diff options
Diffstat (limited to 'src')
79 files changed, 9345 insertions, 4689 deletions
diff --git a/src/bootstrap/src/bin/rustc.rs b/src/bootstrap/src/bin/rustc.rs index 5865df67b66..f15b76fa85c 100644 --- a/src/bootstrap/src/bin/rustc.rs +++ b/src/bootstrap/src/bin/rustc.rs @@ -179,6 +179,16 @@ fn main() { } } + // Here we pass additional paths that essentially act as a sysroot. + // These are used to load rustc crates (e.g. `extern crate rustc_ast;`) + // for rustc_private tools, so that we do not have to copy them into the + // actual sysroot of the compiler that builds the tool. + if let Ok(dirs) = env::var("RUSTC_ADDITIONAL_SYSROOT_PATHS") { + for dir in dirs.split(",") { + cmd.arg(format!("-L{dir}")); + } + } + // Force all crates compiled by this compiler to (a) be unstable and (b) // allow the `rustc_private` feature to link to other unstable crates // also in the sysroot. We also do this for host crates, since those diff --git a/src/bootstrap/src/core/build_steps/check.rs b/src/bootstrap/src/core/build_steps/check.rs index 4a110b733e1..bebae893ee7 100644 --- a/src/bootstrap/src/core/build_steps/check.rs +++ b/src/bootstrap/src/core/build_steps/check.rs @@ -1,5 +1,8 @@ //! Implementation of compiling the compiler and standard library, in "check"-based modes. +use std::fs; +use std::path::{Path, PathBuf}; + use crate::core::build_steps::compile::{ add_to_sysroot, run_cargo, rustc_cargo, rustc_cargo_env, std_cargo, std_crates_for_run_make, }; @@ -9,11 +12,11 @@ use crate::core::build_steps::tool::{ prepare_tool_cargo, }; use crate::core::builder::{ - self, Alias, Builder, Kind, RunConfig, ShouldRun, Step, StepMetadata, crate_description, + self, Alias, Builder, Cargo, Kind, RunConfig, ShouldRun, Step, StepMetadata, crate_description, }; use crate::core::config::TargetSelection; use crate::utils::build_stamp::{self, BuildStamp}; -use crate::{CodegenBackendKind, Compiler, Mode, Subcommand}; +use crate::{CodegenBackendKind, Compiler, Mode, Subcommand, t}; #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Std { @@ -33,7 +36,7 @@ impl Std { } impl Step for Std { - type Output = (); + type Output = BuildStamp; const DEFAULT: bool = true; fn should_run(run: ShouldRun<'_>) -> ShouldRun<'_> { @@ -60,13 +63,14 @@ impl Step for Std { let crates = std_crates_for_run_make(&run); run.builder.ensure(Std { - build_compiler: prepare_compiler_for_check(run.builder, run.target, Mode::Std), + build_compiler: prepare_compiler_for_check(run.builder, run.target, Mode::Std) + .build_compiler(), target: run.target, crates, }); } - fn run(self, builder: &Builder<'_>) { + fn run(self, builder: &Builder<'_>) -> Self::Output { let build_compiler = self.build_compiler; let target = self.target; @@ -93,18 +97,27 @@ impl Step for Std { Kind::Check, format_args!("library artifacts{}", crate_description(&self.crates)), Mode::Std, - self.build_compiler, + build_compiler, target, ); - let stamp = build_stamp::libstd_stamp(builder, build_compiler, target).with_prefix("check"); - run_cargo(builder, cargo, builder.config.free_args.clone(), &stamp, vec![], true, false); + let check_stamp = + build_stamp::libstd_stamp(builder, build_compiler, target).with_prefix("check"); + run_cargo( + builder, + cargo, + builder.config.free_args.clone(), + &check_stamp, + vec![], + true, + false, + ); drop(_guard); // don't check test dependencies if we haven't built libtest if !self.crates.iter().any(|krate| krate == "test") { - return; + return check_stamp; } // Then run cargo again, once we've put the rmeta files for the library @@ -137,10 +150,11 @@ impl Step for Std { Kind::Check, "library test/bench/example targets", Mode::Std, - self.build_compiler, + build_compiler, target, ); run_cargo(builder, cargo, builder.config.free_args.clone(), &stamp, vec![], true, false); + check_stamp } fn metadata(&self) -> Option<StepMetadata> { @@ -148,12 +162,135 @@ impl Step for Std { } } -/// Checks rustc using `build_compiler` and copies the built -/// .rmeta files into the sysroot of `build_compiler`. +/// Represents a proof that rustc was **checked**. +/// Contains directories with .rmeta files generated by checking rustc for a specific +/// target. +#[derive(Debug, Clone, PartialEq, Eq, Hash)] +struct RmetaSysroot { + host_dir: PathBuf, + target_dir: PathBuf, +} + +impl RmetaSysroot { + /// Copy rmeta artifacts from the given `stamp` into a sysroot located at `directory`. + fn from_stamp( + builder: &Builder<'_>, + stamp: BuildStamp, + target: TargetSelection, + directory: &Path, + ) -> Self { + let host_dir = directory.join("host"); + let target_dir = directory.join(target); + let _ = fs::remove_dir_all(directory); + t!(fs::create_dir_all(directory)); + add_to_sysroot(builder, &target_dir, &host_dir, &stamp); + + Self { host_dir, target_dir } + } + + /// Configure the given cargo invocation so that the compiled crate will be able to use + /// rustc .rmeta artifacts that were previously generated. + fn configure_cargo(&self, cargo: &mut Cargo) { + cargo.append_to_env( + "RUSTC_ADDITIONAL_SYSROOT_PATHS", + format!("{},{}", self.host_dir.to_str().unwrap(), self.target_dir.to_str().unwrap()), + ",", + ); + } +} + +/// Checks rustc using the given `build_compiler` for the given `target`, and produces +/// a sysroot in the build directory that stores the generated .rmeta files. +/// +/// This step exists so that we can store the generated .rmeta artifacts into a separate +/// directory, instead of copying them into the sysroot of `build_compiler`, which would +/// "pollute" it (that is especially problematic for the external stage0 rustc). +#[derive(Debug, Clone, PartialEq, Eq, Hash)] +struct PrepareRustcRmetaSysroot { + build_compiler: CompilerForCheck, + target: TargetSelection, +} + +impl PrepareRustcRmetaSysroot { + fn new(build_compiler: CompilerForCheck, target: TargetSelection) -> Self { + Self { build_compiler, target } + } +} + +impl Step for PrepareRustcRmetaSysroot { + type Output = RmetaSysroot; + + fn should_run(run: ShouldRun<'_>) -> ShouldRun<'_> { + run.never() + } + + fn run(self, builder: &Builder<'_>) -> Self::Output { + // Check rustc + let stamp = builder.ensure(Rustc::from_build_compiler( + self.build_compiler.clone(), + self.target, + vec![], + )); + + let build_compiler = self.build_compiler.build_compiler(); + + // Copy the generated rmeta artifacts to a separate directory + let dir = builder + .out + .join(build_compiler.host) + .join(format!("stage{}-rustc-rmeta-artifacts", build_compiler.stage + 1)); + RmetaSysroot::from_stamp(builder, stamp, self.target, &dir) + } +} + +/// Checks std using the given `build_compiler` for the given `target`, and produces +/// a sysroot in the build directory that stores the generated .rmeta files. +/// +/// This step exists so that we can store the generated .rmeta artifacts into a separate +/// directory, instead of copying them into the sysroot of `build_compiler`, which would +/// "pollute" it (that is especially problematic for the external stage0 rustc). +#[derive(Debug, Clone, PartialEq, Eq, Hash)] +struct PrepareStdRmetaSysroot { + build_compiler: Compiler, + target: TargetSelection, +} + +impl PrepareStdRmetaSysroot { + fn new(build_compiler: Compiler, target: TargetSelection) -> Self { + Self { build_compiler, target } + } +} + +impl Step for PrepareStdRmetaSysroot { + type Output = RmetaSysroot; + + fn should_run(run: ShouldRun<'_>) -> ShouldRun<'_> { + run.never() + } + + fn run(self, builder: &Builder<'_>) -> Self::Output { + // Check std + let stamp = builder.ensure(Std { + build_compiler: self.build_compiler, + target: self.target, + crates: vec![], + }); + + // Copy the generated rmeta artifacts to a separate directory + let dir = builder + .out + .join(self.build_compiler.host) + .join(format!("stage{}-std-rmeta-artifacts", self.build_compiler.stage)); + + RmetaSysroot::from_stamp(builder, stamp, self.target, &dir) + } +} + +/// Checks rustc using `build_compiler`. #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Rustc { /// Compiler that will check this rustc. - pub build_compiler: Compiler, + pub build_compiler: CompilerForCheck, pub target: TargetSelection, /// Whether to build only a subset of crates. /// @@ -166,12 +303,20 @@ pub struct Rustc { impl Rustc { pub fn new(builder: &Builder<'_>, target: TargetSelection, crates: Vec<String>) -> Self { let build_compiler = prepare_compiler_for_check(builder, target, Mode::Rustc); + Self::from_build_compiler(build_compiler, target, crates) + } + + fn from_build_compiler( + build_compiler: CompilerForCheck, + target: TargetSelection, + crates: Vec<String>, + ) -> Self { Self { build_compiler, target, crates } } } impl Step for Rustc { - type Output = (); + type Output = BuildStamp; const IS_HOST: bool = true; const DEFAULT: bool = true; @@ -191,8 +336,8 @@ impl Step for Rustc { /// created will also be linked into the sysroot directory. /// /// If we check a stage 2 compiler, we will have to first build a stage 1 compiler to check it. - fn run(self, builder: &Builder<'_>) { - let build_compiler = self.build_compiler; + fn run(self, builder: &Builder<'_>) -> Self::Output { + let build_compiler = self.build_compiler.build_compiler; let target = self.target; let mut cargo = builder::Cargo::new( @@ -205,6 +350,7 @@ impl Step for Rustc { ); rustc_cargo(builder, &mut cargo, target, &build_compiler, &self.crates); + self.build_compiler.configure_cargo(&mut cargo); // Explicitly pass -p for all compiler crates -- this will force cargo // to also check the tests/benches/examples for these crates, rather @@ -217,7 +363,7 @@ impl Step for Rustc { Kind::Check, format_args!("compiler artifacts{}", crate_description(&self.crates)), Mode::Rustc, - self.build_compiler, + self.build_compiler.build_compiler(), target, ); @@ -226,13 +372,12 @@ impl Step for Rustc { run_cargo(builder, cargo, builder.config.free_args.clone(), &stamp, vec![], true, false); - let libdir = builder.sysroot_target_libdir(build_compiler, target); - let hostdir = builder.sysroot_target_libdir(build_compiler, build_compiler.host); - add_to_sysroot(builder, &libdir, &hostdir, &stamp); + stamp } fn metadata(&self) -> Option<StepMetadata> { - let metadata = StepMetadata::check("rustc", self.target).built_by(self.build_compiler); + let metadata = StepMetadata::check("rustc", self.target) + .built_by(self.build_compiler.build_compiler()); let metadata = if self.crates.is_empty() { metadata } else { @@ -242,45 +387,101 @@ impl Step for Rustc { } } +/// Represents a compiler that can check something. +/// +/// If the compiler was created for `Mode::ToolRustc` or `Mode::Codegen`, it will also contain +/// .rmeta artifacts from rustc that was already checked using `build_compiler`. +/// +/// All steps that use this struct in a "general way" (i.e. they don't know exactly what kind of +/// thing is being built) should call `configure_cargo` to ensure that the rmeta artifacts are +/// properly linked, if present. +#[derive(Debug, Clone, PartialEq, Eq, Hash)] +pub struct CompilerForCheck { + build_compiler: Compiler, + rustc_rmeta_sysroot: Option<RmetaSysroot>, + std_rmeta_sysroot: Option<RmetaSysroot>, +} + +impl CompilerForCheck { + pub fn build_compiler(&self) -> Compiler { + self.build_compiler + } + + /// If there are any rustc rmeta artifacts available, configure the Cargo invocation + /// so that the artifact being built can find them. + pub fn configure_cargo(&self, cargo: &mut Cargo) { + if let Some(sysroot) = &self.rustc_rmeta_sysroot { + sysroot.configure_cargo(cargo); + } + if let Some(sysroot) = &self.std_rmeta_sysroot { + sysroot.configure_cargo(cargo); + } + } +} + +/// Prepare the standard library for checking something (that requires stdlib) using +/// `build_compiler`. +fn prepare_std( + builder: &Builder<'_>, + build_compiler: Compiler, + target: TargetSelection, +) -> Option<RmetaSysroot> { + // We need to build the host stdlib even if we only check, to compile build scripts and proc + // macros + builder.std(build_compiler, builder.host_target); + + // If we're cross-compiling, we generate the rmeta files for the given target + // This check has to be here, because if we generate both .so and .rmeta files, rustc will fail, + // as it will have multiple candidates for linking. + if builder.host_target != target { + Some(builder.ensure(PrepareStdRmetaSysroot::new(build_compiler, target))) + } else { + None + } +} + /// Prepares a compiler that will check something with the given `mode`. pub fn prepare_compiler_for_check( builder: &Builder<'_>, target: TargetSelection, mode: Mode, -) -> Compiler { +) -> CompilerForCheck { let host = builder.host_target; - match mode { + let mut rustc_rmeta_sysroot = None; + let mut std_rmeta_sysroot = None; + let build_compiler = match mode { Mode::ToolBootstrap => builder.compiler(0, host), + // We could also only check std here and use `prepare_std`, but `ToolTarget` is currently + // only used for running in-tree Clippy on bootstrap tools, so it does not seem worth it to + // optimize it. Therefore, here we build std for the target, instead of just checking it. Mode::ToolTarget => get_tool_target_compiler(builder, ToolTargetBuildMode::Build(target)), Mode::ToolStd => { if builder.config.compile_time_deps { // When --compile-time-deps is passed, we can't use any rustc // other than the bootstrap compiler. Luckily build scripts and // proc macros for tools are unlikely to need nightly. - return builder.compiler(0, host); + builder.compiler(0, host) + } else { + // These tools require the local standard library to be checked + let build_compiler = builder.compiler(builder.top_stage, host); + std_rmeta_sysroot = prepare_std(builder, build_compiler, target); + build_compiler } - - // These tools require the local standard library to be checked - let build_compiler = builder.compiler(builder.top_stage, host); - - // We need to build the host stdlib to check the tool itself. - // We need to build the target stdlib so that the tool can link to it. - builder.std(build_compiler, host); - // We could only check this library in theory, but `check::Std` doesn't copy rmetas - // into `build_compiler`'s sysroot to avoid clashes with `.rlibs`, so we build it - // instead. - builder.std(build_compiler, target); - build_compiler } Mode::ToolRustc | Mode::Codegen => { // Check Rustc to produce the required rmeta artifacts for rustc_private, and then // return the build compiler that was used to check rustc. // We do not need to check examples/tests/etc. of Rustc for rustc_private, so we pass // an empty set of crates, which will avoid using `cargo -p`. - let check = Rustc::new(builder, target, vec![]); - let build_compiler = check.build_compiler; - builder.ensure(check); + let compiler_for_rustc = prepare_compiler_for_check(builder, target, Mode::Rustc); + rustc_rmeta_sysroot = Some( + builder.ensure(PrepareRustcRmetaSysroot::new(compiler_for_rustc.clone(), target)), + ); + let build_compiler = compiler_for_rustc.build_compiler(); + + // To check a rustc_private tool, we also need to check std that it will link to + std_rmeta_sysroot = prepare_std(builder, build_compiler, target); build_compiler } Mode::Rustc => { @@ -294,15 +495,8 @@ pub fn prepare_compiler_for_check( let stage = if host == target { builder.top_stage - 1 } else { builder.top_stage }; let build_compiler = builder.compiler(stage, host); - // Build host std for compiling build scripts - builder.std(build_compiler, build_compiler.host); - - // Build target std so that the checked rustc can link to it during the check - // FIXME: maybe we can a way to only do a check of std here? - // But for that we would have to copy the stdlib rmetas to the sysroot of the build - // compiler, which conflicts with std rlibs, if we also build std. - builder.std(build_compiler, target); - + // To check rustc, we need to check std that it will link to + std_rmeta_sysroot = prepare_std(builder, build_compiler, target); build_compiler } Mode::Std => { @@ -311,13 +505,14 @@ pub fn prepare_compiler_for_check( // stage 0 stdlib is used to compile build scripts and proc macros. builder.compiler(builder.top_stage, host) } - } + }; + CompilerForCheck { build_compiler, rustc_rmeta_sysroot, std_rmeta_sysroot } } /// Check the Cranelift codegen backend. #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct CraneliftCodegenBackend { - build_compiler: Compiler, + build_compiler: CompilerForCheck, target: TargetSelection, } @@ -332,12 +527,14 @@ impl Step for CraneliftCodegenBackend { } fn make_run(run: RunConfig<'_>) { - let build_compiler = prepare_compiler_for_check(run.builder, run.target, Mode::Codegen); - run.builder.ensure(CraneliftCodegenBackend { build_compiler, target: run.target }); + run.builder.ensure(CraneliftCodegenBackend { + build_compiler: prepare_compiler_for_check(run.builder, run.target, Mode::Codegen), + target: run.target, + }); } fn run(self, builder: &Builder<'_>) { - let build_compiler = self.build_compiler; + let build_compiler = self.build_compiler.build_compiler(); let target = self.target; let mut cargo = builder::Cargo::new( @@ -353,12 +550,13 @@ impl Step for CraneliftCodegenBackend { .arg("--manifest-path") .arg(builder.src.join("compiler/rustc_codegen_cranelift/Cargo.toml")); rustc_cargo_env(builder, &mut cargo, target); + self.build_compiler.configure_cargo(&mut cargo); let _guard = builder.msg( Kind::Check, "rustc_codegen_cranelift", Mode::Codegen, - self.build_compiler, + build_compiler, target, ); @@ -376,7 +574,7 @@ impl Step for CraneliftCodegenBackend { fn metadata(&self) -> Option<StepMetadata> { Some( StepMetadata::check("rustc_codegen_cranelift", self.target) - .built_by(self.build_compiler), + .built_by(self.build_compiler.build_compiler()), ) } } @@ -384,7 +582,7 @@ impl Step for CraneliftCodegenBackend { /// Check the GCC codegen backend. #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct GccCodegenBackend { - build_compiler: Compiler, + build_compiler: CompilerForCheck, target: TargetSelection, } @@ -399,8 +597,10 @@ impl Step for GccCodegenBackend { } fn make_run(run: RunConfig<'_>) { - let build_compiler = prepare_compiler_for_check(run.builder, run.target, Mode::Codegen); - run.builder.ensure(GccCodegenBackend { build_compiler, target: run.target }); + run.builder.ensure(GccCodegenBackend { + build_compiler: prepare_compiler_for_check(run.builder, run.target, Mode::Codegen), + target: run.target, + }); } fn run(self, builder: &Builder<'_>) { @@ -410,7 +610,7 @@ impl Step for GccCodegenBackend { return; } - let build_compiler = self.build_compiler; + let build_compiler = self.build_compiler.build_compiler(); let target = self.target; let mut cargo = builder::Cargo::new( @@ -424,14 +624,10 @@ impl Step for GccCodegenBackend { cargo.arg("--manifest-path").arg(builder.src.join("compiler/rustc_codegen_gcc/Cargo.toml")); rustc_cargo_env(builder, &mut cargo, target); + self.build_compiler.configure_cargo(&mut cargo); - let _guard = builder.msg( - Kind::Check, - "rustc_codegen_gcc", - Mode::Codegen, - self.build_compiler, - target, - ); + let _guard = + builder.msg(Kind::Check, "rustc_codegen_gcc", Mode::Codegen, build_compiler, target); let stamp = build_stamp::codegen_backend_stamp( builder, @@ -445,7 +641,10 @@ impl Step for GccCodegenBackend { } fn metadata(&self) -> Option<StepMetadata> { - Some(StepMetadata::check("rustc_codegen_gcc", self.target).built_by(self.build_compiler)) + Some( + StepMetadata::check("rustc_codegen_gcc", self.target) + .built_by(self.build_compiler.build_compiler()), + ) } } @@ -467,8 +666,8 @@ macro_rules! tool_check_step { ) => { #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct $name { - pub build_compiler: Compiler, - pub target: TargetSelection, + compiler: CompilerForCheck, + target: TargetSelection, } impl Step for $name { @@ -486,7 +685,7 @@ macro_rules! tool_check_step { let builder = run.builder; let mode = $mode(builder); - let build_compiler = prepare_compiler_for_check(run.builder, target, mode); + let compiler = prepare_compiler_for_check(run.builder, target, mode); // It doesn't make sense to cross-check bootstrap tools if mode == Mode::ToolBootstrap && target != run.builder.host_target { @@ -494,11 +693,11 @@ macro_rules! tool_check_step { return; }; - run.builder.ensure($name { target, build_compiler }); + run.builder.ensure($name { target, compiler }); } fn run(self, builder: &Builder<'_>) { - let Self { target, build_compiler } = self; + let Self { target, compiler } = self; let allow_features = { let mut _value = ""; $( _value = $allow_features; )? @@ -506,11 +705,11 @@ macro_rules! tool_check_step { }; let extra_features: &[&str] = &[$($($enable_features),*)?]; let mode = $mode(builder); - run_tool_check_step(builder, build_compiler, target, $path, mode, allow_features, extra_features); + run_tool_check_step(builder, compiler, target, $path, mode, allow_features, extra_features); } fn metadata(&self) -> Option<StepMetadata> { - Some(StepMetadata::check(stringify!($name), self.target).built_by(self.build_compiler)) + Some(StepMetadata::check(stringify!($name), self.target).built_by(self.compiler.build_compiler)) } } } @@ -519,7 +718,7 @@ macro_rules! tool_check_step { /// Used by the implementation of `Step::run` in `tool_check_step!`. fn run_tool_check_step( builder: &Builder<'_>, - build_compiler: Compiler, + compiler: CompilerForCheck, target: TargetSelection, path: &str, mode: Mode, @@ -528,6 +727,8 @@ fn run_tool_check_step( ) { let display_name = path.rsplit('/').next().unwrap(); + let build_compiler = compiler.build_compiler(); + let extra_features = extra_features.iter().map(|f| f.to_string()).collect::<Vec<String>>(); let mut cargo = prepare_tool_cargo( builder, @@ -544,6 +745,7 @@ fn run_tool_check_step( &extra_features, ); cargo.allow_features(allow_features); + compiler.configure_cargo(&mut cargo); // FIXME: check bootstrap doesn't currently work when multiple targets are checked // FIXME: rust-analyzer does not work with --all-targets diff --git a/src/bootstrap/src/core/build_steps/clippy.rs b/src/bootstrap/src/core/build_steps/clippy.rs index 3c4aa0886c2..05f8b240291 100644 --- a/src/bootstrap/src/core/build_steps/clippy.rs +++ b/src/bootstrap/src/core/build_steps/clippy.rs @@ -18,7 +18,7 @@ use build_helper::exit; use super::compile::{run_cargo, rustc_cargo, std_cargo}; use super::tool::{SourceType, prepare_tool_cargo}; use crate::builder::{Builder, ShouldRun}; -use crate::core::build_steps::check::prepare_compiler_for_check; +use crate::core::build_steps::check::{CompilerForCheck, prepare_compiler_for_check}; use crate::core::build_steps::compile::std_crates_for_run_make; use crate::core::builder; use crate::core::builder::{Alias, Kind, RunConfig, Step, StepMetadata, crate_description}; @@ -231,7 +231,7 @@ impl Step for Std { /// in-tree rustc. #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Rustc { - build_compiler: Compiler, + build_compiler: CompilerForCheck, target: TargetSelection, config: LintConfig, /// Whether to lint only a subset of crates. @@ -271,7 +271,7 @@ impl Step for Rustc { } fn run(self, builder: &Builder<'_>) { - let build_compiler = self.build_compiler; + let build_compiler = self.build_compiler.build_compiler(); let target = self.target; let mut cargo = builder::Cargo::new( @@ -284,6 +284,7 @@ impl Step for Rustc { ); rustc_cargo(builder, &mut cargo, target, &build_compiler, &self.crates); + self.build_compiler.configure_cargo(&mut cargo); // Explicitly pass -p for all compiler crates -- this will force cargo // to also lint the tests/benches/examples for these crates, rather @@ -312,13 +313,16 @@ impl Step for Rustc { } fn metadata(&self) -> Option<StepMetadata> { - Some(StepMetadata::clippy("rustc", self.target).built_by(self.build_compiler)) + Some( + StepMetadata::clippy("rustc", self.target) + .built_by(self.build_compiler.build_compiler()), + ) } } #[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct CodegenGcc { - build_compiler: Compiler, + build_compiler: CompilerForCheck, target: TargetSelection, config: LintConfig, } @@ -347,10 +351,10 @@ impl Step for CodegenGcc { } fn run(self, builder: &Builder<'_>) -> Self::Output { - let build_compiler = self.build_compiler; + let build_compiler = self.build_compiler.build_compiler(); let target = self.target; - let cargo = prepare_tool_cargo( + let mut cargo = prepare_tool_cargo( builder, build_compiler, Mode::Codegen, @@ -360,6 +364,7 @@ impl Step for CodegenGcc { SourceType::InTree, &[], ); + self.build_compiler.configure_cargo(&mut cargo); let _guard = builder.msg(Kind::Clippy, "rustc_codegen_gcc", Mode::ToolRustc, build_compiler, target); @@ -379,7 +384,10 @@ impl Step for CodegenGcc { } fn metadata(&self) -> Option<StepMetadata> { - Some(StepMetadata::clippy("rustc_codegen_gcc", self.target).built_by(self.build_compiler)) + Some( + StepMetadata::clippy("rustc_codegen_gcc", self.target) + .built_by(self.build_compiler.build_compiler()), + ) } } @@ -396,7 +404,7 @@ macro_rules! lint_any { #[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct $name { - build_compiler: Compiler, + build_compiler: CompilerForCheck, target: TargetSelection, config: LintConfig, } @@ -419,9 +427,9 @@ macro_rules! lint_any { } fn run(self, builder: &Builder<'_>) -> Self::Output { - let build_compiler = self.build_compiler; + let build_compiler = self.build_compiler.build_compiler(); let target = self.target; - let cargo = prepare_tool_cargo( + let mut cargo = prepare_tool_cargo( builder, build_compiler, $mode, @@ -431,6 +439,7 @@ macro_rules! lint_any { SourceType::InTree, &[], ); + self.build_compiler.configure_cargo(&mut cargo); let _guard = builder.msg( Kind::Clippy, @@ -456,7 +465,7 @@ macro_rules! lint_any { } fn metadata(&self) -> Option<StepMetadata> { - Some(StepMetadata::clippy($readable_name, self.target).built_by(self.build_compiler)) + Some(StepMetadata::clippy($readable_name, self.target).built_by(self.build_compiler.build_compiler())) } } )+ diff --git a/src/bootstrap/src/core/build_steps/compile.rs b/src/bootstrap/src/core/build_steps/compile.rs index 997a152a31f..004b132f28f 100644 --- a/src/bootstrap/src/core/build_steps/compile.rs +++ b/src/bootstrap/src/core/build_steps/compile.rs @@ -12,6 +12,7 @@ use std::ffi::OsStr; use std::io::BufReader; use std::io::prelude::*; use std::path::{Path, PathBuf}; +use std::time::SystemTime; use std::{env, fs, str}; use serde_derive::Deserialize; @@ -38,7 +39,7 @@ use crate::{ }; /// Build a standard library for the given `target` using the given `build_compiler`. -#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)] +#[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Std { pub target: TargetSelection, /// Compiler that builds the standard library. @@ -933,13 +934,22 @@ fn cp_rustc_component_to_ci_sysroot(builder: &Builder<'_>, sysroot: &Path, conte } } +/// Represents information about a built rustc. +#[derive(Clone, Debug)] +pub struct BuiltRustc { + /// The compiler that actually built this *rustc*. + /// This can be different from the *build_compiler* passed to the `Rustc` step because of + /// uplifting. + pub build_compiler: Compiler, +} + /// Build rustc using the passed `build_compiler`. /// /// - Makes sure that `build_compiler` has a standard library prepared for its host target, /// so that it can compile build scripts and proc macros when building this `rustc`. /// - Makes sure that `build_compiler` has a standard library prepared for `target`, /// so that the built `rustc` can *link to it* and use it at runtime. -#[derive(Debug, PartialOrd, Ord, Clone, PartialEq, Eq, Hash)] +#[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Rustc { /// The target on which rustc will run (its host). pub target: TargetSelection, @@ -960,7 +970,7 @@ impl Rustc { } impl Step for Rustc { - type Output = (); + type Output = BuiltRustc; const IS_HOST: bool = true; const DEFAULT: bool = false; @@ -1000,7 +1010,7 @@ impl Step for Rustc { /// This will build the compiler for a particular stage of the build using /// the `build_compiler` targeting the `target` architecture. The artifacts /// created will also be linked into the sysroot directory. - fn run(self, builder: &Builder<'_>) { + fn run(self, builder: &Builder<'_>) -> Self::Output { let build_compiler = self.build_compiler; let target = self.target; @@ -1016,7 +1026,7 @@ impl Step for Rustc { &sysroot, builder.config.ci_rustc_dev_contents(), ); - return; + return BuiltRustc { build_compiler }; } // Build a standard library for `target` using the `build_compiler`. @@ -1028,9 +1038,9 @@ impl Step for Rustc { builder.info("WARNING: Using a potentially old librustc. This may not behave well."); builder.info("WARNING: Use `--keep-stage-std` if you want to rebuild the compiler when it changes"); - builder.ensure(RustcLink::from_rustc(self, build_compiler)); + builder.ensure(RustcLink::from_rustc(self)); - return; + return BuiltRustc { build_compiler }; } // The stage of the compiler that we're building @@ -1042,21 +1052,35 @@ impl Step for Rustc { && !builder.config.full_bootstrap && (target == builder.host_target || builder.hosts.contains(&target)) { - // If we're cross-compiling, the earliest rustc that we could have is stage 2. - // If we're not cross-compiling, then we should have rustc stage 1. - let stage_to_uplift = if target == builder.host_target { 1 } else { 2 }; - let rustc_to_uplift = builder.compiler(stage_to_uplift, target); - let msg = if rustc_to_uplift.host == target { - format!("Uplifting rustc (stage{} -> stage{stage})", rustc_to_uplift.stage,) + // Here we need to determine the **build compiler** that built the stage that we will + // be uplifting. We cannot uplift stage 1, as it has a different ABI than stage 2+, + // so we always uplift the stage2 compiler (compiled with stage 1). + let uplift_build_compiler = builder.compiler(1, build_compiler.host); + let msg = if uplift_build_compiler.host == target { + format!("Uplifting rustc (stage2 -> stage{stage})") } else { format!( - "Uplifting rustc (stage{}:{} -> stage{stage}:{target})", - rustc_to_uplift.stage, rustc_to_uplift.host, + "Uplifting rustc (stage2:{} -> stage{stage}:{target})", + uplift_build_compiler.host ) }; builder.info(&msg); - builder.ensure(RustcLink::from_rustc(self, rustc_to_uplift)); - return; + + // Here the compiler that built the rlibs (`uplift_build_compiler`) can be different + // from the compiler whose sysroot should be modified in this step. So we need to copy + // the (previously built) rlibs into the correct sysroot. + builder.ensure(RustcLink::from_build_compiler_and_sysroot( + // This is the compiler that actually built the rustc rlibs + uplift_build_compiler, + // We copy the rlibs into the sysroot of `build_compiler` + build_compiler, + target, + self.crates, + )); + + // Here we have performed an uplift, so we return the actual build compiler that "built" + // this rustc. + return BuiltRustc { build_compiler: uplift_build_compiler }; } // Build a standard library for the current host target using the `build_compiler`. @@ -1129,10 +1153,8 @@ impl Step for Rustc { strip_debug(builder, target, &target_root_dir.join("rustc-main")); } - builder.ensure(RustcLink::from_rustc( - self, - builder.compiler(build_compiler.stage, builder.config.host_target), - )); + builder.ensure(RustcLink::from_rustc(self)); + BuiltRustc { build_compiler } } fn metadata(&self) -> Option<StepMetadata> { @@ -1441,31 +1463,51 @@ fn rustc_llvm_env(builder: &Builder<'_>, cargo: &mut Cargo, target: TargetSelect } } -/// `RustcLink` copies all of the rlibs from the rustc build into the previous stage's sysroot. +/// `RustcLink` copies compiler rlibs from a rustc build into a compiler sysroot. +/// It works with (potentially up to) three compilers: +/// - `build_compiler` is a compiler that built rustc rlibs +/// - `sysroot_compiler` is a compiler into whose sysroot we will copy the rlibs +/// - In most situations, `build_compiler` == `sysroot_compiler` +/// - `target_compiler` is the compiler whose rlibs were built. It is not represented explicitly +/// in this step, rather we just read the rlibs from a rustc build stamp of `build_compiler`. +/// /// This is necessary for tools using `rustc_private`, where the previous compiler will build /// a tool against the next compiler. /// To build a tool against a compiler, the rlibs of that compiler that it links against /// must be in the sysroot of the compiler that's doing the compiling. #[derive(Debug, Clone, PartialEq, Eq, Hash)] struct RustcLink { - /// The compiler whose rlibs we are copying around. - pub compiler: Compiler, - /// This is the compiler into whose sysroot we want to copy the rlibs into. - pub previous_stage_compiler: Compiler, - pub target: TargetSelection, + /// This compiler **built** some rustc, whose rlibs we will copy into a sysroot. + build_compiler: Compiler, + /// This is the compiler into whose sysroot we want to copy the built rlibs. + /// In most cases, it will correspond to `build_compiler`. + sysroot_compiler: Compiler, + target: TargetSelection, /// Not actually used; only present to make sure the cache invalidation is correct. crates: Vec<String>, } impl RustcLink { - fn from_rustc(rustc: Rustc, host_compiler: Compiler) -> Self { + /// Copy rlibs from the build compiler that build this `rustc` into the sysroot of that + /// build compiler. + fn from_rustc(rustc: Rustc) -> Self { Self { - compiler: host_compiler, - previous_stage_compiler: rustc.build_compiler, + build_compiler: rustc.build_compiler, + sysroot_compiler: rustc.build_compiler, target: rustc.target, crates: rustc.crates, } } + + /// Copy rlibs **built** by `build_compiler` into the sysroot of `sysroot_compiler`. + fn from_build_compiler_and_sysroot( + build_compiler: Compiler, + sysroot_compiler: Compiler, + target: TargetSelection, + crates: Vec<String>, + ) -> Self { + Self { build_compiler, sysroot_compiler, target, crates } + } } impl Step for RustcLink { @@ -1477,14 +1519,14 @@ impl Step for RustcLink { /// Same as `std_link`, only for librustc fn run(self, builder: &Builder<'_>) { - let compiler = self.compiler; - let previous_stage_compiler = self.previous_stage_compiler; + let build_compiler = self.build_compiler; + let sysroot_compiler = self.sysroot_compiler; let target = self.target; add_to_sysroot( builder, - &builder.sysroot_target_libdir(previous_stage_compiler, target), - &builder.sysroot_target_libdir(previous_stage_compiler, compiler.host), - &build_stamp::librustc_stamp(builder, compiler, target), + &builder.sysroot_target_libdir(sysroot_compiler, target), + &builder.sysroot_target_libdir(sysroot_compiler, sysroot_compiler.host), + &build_stamp::librustc_stamp(builder, build_compiler, target), ); } } @@ -1918,7 +1960,7 @@ impl Step for Sysroot { /// linker wrappers (LLD, LLVM bitcode linker, etc.). /// /// This will assemble a compiler in `build/$target/stage$stage`. -#[derive(Debug, PartialOrd, Ord, Clone, PartialEq, Eq, Hash)] +#[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Assemble { /// The compiler which we will produce in this step. Assemble itself will /// take care of ensuring that the necessary prerequisites to do so exist, @@ -2099,7 +2141,10 @@ impl Step for Assemble { "target_compiler.host" = ?target_compiler.host, "building compiler libraries to link to" ); - builder.ensure(Rustc::new(build_compiler, target_compiler.host)); + + // It is possible that an uplift has happened, so we override build_compiler here. + let BuiltRustc { build_compiler } = + builder.ensure(Rustc::new(build_compiler, target_compiler.host)); let stage = target_compiler.stage; let host = target_compiler.host; @@ -2286,6 +2331,7 @@ impl Step for Assemble { /// /// For a particular stage this will link the file listed in `stamp` into the /// `sysroot_dst` provided. +#[track_caller] pub fn add_to_sysroot( builder: &Builder<'_>, sysroot_dst: &Path, @@ -2568,7 +2614,17 @@ pub fn strip_debug(builder: &Builder<'_>, target: TargetSelection, path: &Path) } let previous_mtime = t!(t!(path.metadata()).modified()); - command("strip").arg("--strip-debug").arg(path).run_capture(builder); + let stamp = BuildStamp::new(path.parent().unwrap()) + .with_prefix(path.file_name().unwrap().to_str().unwrap()) + .with_prefix("strip") + .add_stamp(previous_mtime.duration_since(SystemTime::UNIX_EPOCH).unwrap().as_nanos()); + + // Running strip can be relatively expensive (~1s on librustc_driver.so), so we don't rerun it + // if the file is unchanged. + if !stamp.is_up_to_date() { + command("strip").arg("--strip-debug").arg(path).run_capture(builder); + } + t!(stamp.write()); let file = t!(fs::File::open(path)); diff --git a/src/bootstrap/src/core/build_steps/dist.rs b/src/bootstrap/src/core/build_steps/dist.rs index 414f4464d1e..beb71e70035 100644 --- a/src/bootstrap/src/core/build_steps/dist.rs +++ b/src/bootstrap/src/core/build_steps/dist.rs @@ -54,7 +54,7 @@ fn should_build_extended_tool(builder: &Builder<'_>, tool: &str) -> bool { builder.config.tools.as_ref().is_none_or(|tools| tools.contains(tool)) } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Docs { pub host: TargetSelection, } @@ -91,7 +91,7 @@ impl Step for Docs { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct JsonDocs { build_compiler: Compiler, target: TargetSelection, @@ -354,7 +354,7 @@ fn get_cc_search_dirs( (bin_path, lib_path) } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Mingw { pub host: TargetSelection, } @@ -394,7 +394,7 @@ impl Step for Mingw { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Rustc { pub compiler: Compiler, } @@ -730,7 +730,7 @@ fn copy_target_libs( } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Std { pub compiler: Compiler, pub target: TargetSelection, @@ -785,7 +785,7 @@ impl Step for Std { /// `rust.download-rustc`. /// /// (Don't confuse this with [`RustDev`], without the `c`!) -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct RustcDev { pub compiler: Compiler, pub target: TargetSelection, @@ -916,6 +916,12 @@ fn copy_src_dirs( exclude_dirs: &[&str], dst_dir: &Path, ) { + // The src directories should be relative to `base`, we depend on them not being absolute + // paths below. + for src_dir in src_dirs { + assert!(Path::new(src_dir).is_relative()); + } + // Iterating, filtering and copying a large number of directories can be quite slow. // Avoid doing it in dry run (and thus also tests). if builder.config.dry_run() { @@ -923,6 +929,7 @@ fn copy_src_dirs( } fn filter_fn(exclude_dirs: &[&str], dir: &str, path: &Path) -> bool { + // The paths are relative, e.g. `llvm-project/...`. let spath = match path.to_str() { Some(path) => path, None => return false, @@ -930,65 +937,53 @@ fn copy_src_dirs( if spath.ends_with('~') || spath.ends_with(".pyc") { return false; } + // Normalize slashes + let spath = spath.replace("\\", "/"); - const LLVM_PROJECTS: &[&str] = &[ + static LLVM_PROJECTS: &[&str] = &[ "llvm-project/clang", - "llvm-project\\clang", "llvm-project/libunwind", - "llvm-project\\libunwind", "llvm-project/lld", - "llvm-project\\lld", "llvm-project/lldb", - "llvm-project\\lldb", "llvm-project/llvm", - "llvm-project\\llvm", "llvm-project/compiler-rt", - "llvm-project\\compiler-rt", "llvm-project/cmake", - "llvm-project\\cmake", "llvm-project/runtimes", - "llvm-project\\runtimes", "llvm-project/third-party", - "llvm-project\\third-party", ]; - if spath.contains("llvm-project") - && !spath.ends_with("llvm-project") - && !LLVM_PROJECTS.iter().any(|path| spath.contains(path)) - { - return false; - } + if spath.starts_with("llvm-project") && spath != "llvm-project" { + if !LLVM_PROJECTS.iter().any(|path| spath.starts_with(path)) { + return false; + } - // Keep only these third party libraries - const LLVM_THIRD_PARTY: &[&str] = - &["llvm-project/third-party/siphash", "llvm-project\\third-party\\siphash"]; - if (spath.starts_with("llvm-project/third-party") - || spath.starts_with("llvm-project\\third-party")) - && !(spath.ends_with("llvm-project/third-party") - || spath.ends_with("llvm-project\\third-party")) - && !LLVM_THIRD_PARTY.iter().any(|path| spath.contains(path)) - { - return false; - } + // Keep siphash third-party dependency + if spath.starts_with("llvm-project/third-party") + && spath != "llvm-project/third-party" + && !spath.starts_with("llvm-project/third-party/siphash") + { + return false; + } - const LLVM_TEST: &[&str] = &["llvm-project/llvm/test", "llvm-project\\llvm\\test"]; - if LLVM_TEST.iter().any(|path| spath.contains(path)) - && (spath.ends_with(".ll") || spath.ends_with(".td") || spath.ends_with(".s")) - { - return false; + if spath.starts_with("llvm-project/llvm/test") + && (spath.ends_with(".ll") || spath.ends_with(".td") || spath.ends_with(".s")) + { + return false; + } } // Cargo tests use some files like `.gitignore` that we would otherwise exclude. - const CARGO_TESTS: &[&str] = &["tools/cargo/tests", "tools\\cargo\\tests"]; - if CARGO_TESTS.iter().any(|path| spath.contains(path)) { + if spath.starts_with("tools/cargo/tests") { return true; } - let full_path = Path::new(dir).join(path); - if exclude_dirs.iter().any(|excl| full_path == Path::new(excl)) { - return false; + if !exclude_dirs.is_empty() { + let full_path = Path::new(dir).join(path); + if exclude_dirs.iter().any(|excl| full_path == Path::new(excl)) { + return false; + } } - let excludes = [ + static EXCLUDES: &[&str] = &[ "CVS", "RCS", "SCCS", @@ -1011,7 +1006,15 @@ fn copy_src_dirs( ".hgrags", "_darcs", ]; - !path.iter().map(|s| s.to_str().unwrap()).any(|s| excludes.contains(&s)) + + // We want to check if any component of `path` doesn't contain the strings in `EXCLUDES`. + // However, since we traverse directories top-down in `Builder::cp_link_filtered`, + // it is enough to always check only the last component: + // - If the path is a file, we will iterate to it and then check it's filename + // - If the path is a dir, if it's dir name contains an excluded string, we will not even + // recurse into it. + let last_component = path.iter().next_back().map(|s| s.to_str().unwrap()).unwrap(); + !EXCLUDES.contains(&last_component) } // Copy the directories using our filter @@ -1023,7 +1026,7 @@ fn copy_src_dirs( } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Src; impl Step for Src { @@ -1084,7 +1087,7 @@ impl Step for Src { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct PlainSourceTarball; impl Step for PlainSourceTarball { @@ -1230,7 +1233,7 @@ impl Step for PlainSourceTarball { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Cargo { pub build_compiler: Compiler, pub target: TargetSelection, @@ -1284,7 +1287,7 @@ impl Step for Cargo { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct RustAnalyzer { pub build_compiler: Compiler, pub target: TargetSelection, @@ -1560,7 +1563,7 @@ impl Step for Rustfmt { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Extended { stage: u32, host: TargetSelection, @@ -2401,7 +2404,7 @@ impl Step for LlvmTools { /// Distributes the `llvm-bitcode-linker` tool so that it can be used by a compiler whose host /// is `target`. -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct LlvmBitcodeLinker { /// The linker will be compiled by this compiler. pub build_compiler: Compiler, diff --git a/src/bootstrap/src/core/build_steps/doc.rs b/src/bootstrap/src/core/build_steps/doc.rs index f6b27d83120..7fe19c00ef5 100644 --- a/src/bootstrap/src/core/build_steps/doc.rs +++ b/src/bootstrap/src/core/build_steps/doc.rs @@ -580,7 +580,7 @@ impl Step for SharedAssets { } /// Document the standard library using `build_compiler`. -#[derive(Debug, Clone, Hash, PartialEq, Eq, PartialOrd, Ord)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Std { build_compiler: Compiler, target: TargetSelection, @@ -715,7 +715,7 @@ impl Step for Std { /// or remote link. const STD_PUBLIC_CRATES: [&str; 5] = ["core", "alloc", "std", "proc_macro", "test"]; -#[derive(Debug, Copy, Clone, Hash, PartialEq, Eq, PartialOrd, Ord)] +#[derive(Debug, Copy, Clone, Hash, PartialEq, Eq)] pub enum DocumentationFormat { Html, Json, @@ -1230,7 +1230,7 @@ fn symlink_dir_force(config: &Config, original: &Path, link: &Path) { } /// Builds the Rust compiler book. -#[derive(Ord, PartialOrd, Debug, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct RustcBook { build_compiler: Compiler, target: TargetSelection, @@ -1334,7 +1334,7 @@ impl Step for RustcBook { /// Documents the reference. /// It has to always be done using a stage 1+ compiler, because it references in-tree /// compiler/stdlib concepts. -#[derive(Ord, PartialOrd, Debug, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct Reference { build_compiler: Compiler, target: TargetSelection, diff --git a/src/bootstrap/src/core/build_steps/run.rs b/src/bootstrap/src/core/build_steps/run.rs index 7f1a7d00257..c6288f63847 100644 --- a/src/bootstrap/src/core/build_steps/run.rs +++ b/src/bootstrap/src/core/build_steps/run.rs @@ -17,7 +17,7 @@ use crate::core::config::flags::get_completion; use crate::utils::exec::command; use crate::{Mode, t}; -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct BuildManifest; impl Step for BuildManifest { @@ -56,7 +56,7 @@ impl Step for BuildManifest { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct BumpStage0; impl Step for BumpStage0 { @@ -78,7 +78,7 @@ impl Step for BumpStage0 { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct ReplaceVersionPlaceholder; impl Step for ReplaceVersionPlaceholder { @@ -169,7 +169,7 @@ impl Step for Miri { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct CollectLicenseMetadata; impl Step for CollectLicenseMetadata { @@ -200,7 +200,7 @@ impl Step for CollectLicenseMetadata { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct GenerateCopyright; impl Step for GenerateCopyright { @@ -264,7 +264,7 @@ impl Step for GenerateCopyright { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct GenerateWindowsSys; impl Step for GenerateWindowsSys { @@ -326,7 +326,7 @@ impl Step for GenerateCompletions { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct UnicodeTableGenerator; impl Step for UnicodeTableGenerator { @@ -348,7 +348,7 @@ impl Step for UnicodeTableGenerator { } } -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct FeaturesStatusDump; impl Step for FeaturesStatusDump { @@ -408,7 +408,7 @@ impl Step for CyclicStep { /// /// The coverage-dump tool is an internal detail of coverage tests, so this run /// step is only needed when testing coverage-dump manually. -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct CoverageDump; impl Step for CoverageDump { diff --git a/src/bootstrap/src/core/build_steps/test.rs b/src/bootstrap/src/core/build_steps/test.rs index 4006bed4ac5..e7a57a7f375 100644 --- a/src/bootstrap/src/core/build_steps/test.rs +++ b/src/bootstrap/src/core/build_steps/test.rs @@ -2732,7 +2732,7 @@ fn prepare_cargo_test( /// FIXME(Zalathar): Try to split this into two separate steps: a user-visible /// step for testing standard library crates, and an internal step used for both /// library crates and compiler crates. -#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)] +#[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct Crate { pub compiler: Compiler, pub target: TargetSelection, @@ -3747,7 +3747,7 @@ impl Step for TestFloatParse { /// Runs the tool `src/tools/collect-license-metadata` in `ONLY_CHECK=1` mode, /// which verifies that `license-metadata.json` is up-to-date and therefore /// running the tool normally would not update anything. -#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)] +#[derive(Debug, Clone, Hash, PartialEq, Eq)] pub struct CollectLicenseMetadata; impl Step for CollectLicenseMetadata { diff --git a/src/bootstrap/src/core/builder/cargo.rs b/src/bootstrap/src/core/builder/cargo.rs index 3ce21eb151c..69a744a86cb 100644 --- a/src/bootstrap/src/core/builder/cargo.rs +++ b/src/bootstrap/src/core/builder/cargo.rs @@ -101,6 +101,7 @@ pub struct Cargo { impl Cargo { /// Calls [`Builder::cargo`] and [`Cargo::configure_linker`] to prepare an invocation of `cargo` /// to be run. + #[track_caller] pub fn new( builder: &Builder<'_>, compiler: Compiler, @@ -139,6 +140,7 @@ impl Cargo { /// Same as [`Cargo::new`] except this one doesn't configure the linker with /// [`Cargo::configure_linker`]. + #[track_caller] pub fn new_for_mir_opt_tests( builder: &Builder<'_>, compiler: Compiler, @@ -186,6 +188,32 @@ impl Cargo { self } + /// Append a value to an env var of the cargo command instance. + /// If the variable was unset previously, this is equivalent to [`Cargo::env`]. + /// If the variable was already set, this will append `delimiter` and then `value` to it. + /// + /// Note that this only considers the existence of the env. var. configured on this `Cargo` + /// instance. It does not look at the environment of this process. + pub fn append_to_env( + &mut self, + key: impl AsRef<OsStr>, + value: impl AsRef<OsStr>, + delimiter: impl AsRef<OsStr>, + ) -> &mut Cargo { + assert_ne!(key.as_ref(), "RUSTFLAGS"); + assert_ne!(key.as_ref(), "RUSTDOCFLAGS"); + + let key = key.as_ref(); + if let Some((_, Some(previous_value))) = self.command.get_envs().find(|(k, _)| *k == key) { + let mut combined: OsString = previous_value.to_os_string(); + combined.push(delimiter.as_ref()); + combined.push(value.as_ref()); + self.env(key, combined) + } else { + self.env(key, value) + } + } + pub fn add_rustc_lib_path(&mut self, builder: &Builder<'_>) { builder.add_rustc_lib_path(self.compiler, &mut self.command); } @@ -396,6 +424,7 @@ impl From<Cargo> for BootstrapCommand { impl Builder<'_> { /// Like [`Builder::cargo`], but only passes flags that are valid for all commands. + #[track_caller] pub fn bare_cargo( &self, compiler: Compiler, @@ -480,6 +509,7 @@ impl Builder<'_> { /// scoped by `mode`'s output directory, it will pass the `--target` flag for the specified /// `target`, and will be executing the Cargo command `cmd`. `cmd` can be `miri-cmd` for /// commands to be run with Miri. + #[track_caller] fn cargo( &self, compiler: Compiler, diff --git a/src/bootstrap/src/core/builder/tests.rs b/src/bootstrap/src/core/builder/tests.rs index a9398a654e9..f4266a6085b 100644 --- a/src/bootstrap/src/core/builder/tests.rs +++ b/src/bootstrap/src/core/builder/tests.rs @@ -1569,7 +1569,7 @@ mod snapshot { [build] llvm <host> [build] rustc 0 <host> -> rustc 1 <host> [build] rustc 1 <host> -> std 1 <host> - [build] rustc 1 <host> -> std 1 <target1> + [check] rustc 1 <host> -> std 1 <target1> [check] rustc 1 <host> -> rustc 2 <target1> (73 crates) [check] rustc 1 <host> -> rustc 2 <target1> [check] rustc 1 <host> -> Rustdoc 2 <target1> diff --git a/src/bootstrap/src/core/config/mod.rs b/src/bootstrap/src/core/config/mod.rs index 285d20917e7..8c5f9037251 100644 --- a/src/bootstrap/src/core/config/mod.rs +++ b/src/bootstrap/src/core/config/mod.rs @@ -324,7 +324,7 @@ impl FromStr for LlvmLibunwind { } } -#[derive(Debug, Default, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)] +#[derive(Debug, Default, Copy, Clone, PartialEq, Eq, Hash)] pub enum SplitDebuginfo { Packed, Unpacked, diff --git a/src/bootstrap/src/core/config/target_selection.rs b/src/bootstrap/src/core/config/target_selection.rs index ebd3fe7a8c6..40b63a7f9c7 100644 --- a/src/bootstrap/src/core/config/target_selection.rs +++ b/src/bootstrap/src/core/config/target_selection.rs @@ -14,7 +14,7 @@ pub struct TargetSelection { } /// Newtype over `Vec<TargetSelection>` so we can implement custom parsing logic -#[derive(Clone, Default, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)] +#[derive(Clone, Default, PartialEq, Eq, Hash, Debug)] pub struct TargetSelectionList(pub Vec<TargetSelection>); pub fn target_selection_list(s: &str) -> Result<TargetSelectionList, String> { diff --git a/src/bootstrap/src/lib.rs b/src/bootstrap/src/lib.rs index 706a3cbb210..8f766ed00a5 100644 --- a/src/bootstrap/src/lib.rs +++ b/src/bootstrap/src/lib.rs @@ -279,7 +279,7 @@ pub enum DependencyType { /// /// These entries currently correspond to the various output directories of the /// build system, with each mod generating output in a different directory. -#[derive(Debug, Hash, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)] +#[derive(Debug, Hash, Clone, Copy, PartialEq, Eq)] pub enum Mode { /// Build the standard library, placing output in the "stageN-std" directory. Std, @@ -357,7 +357,7 @@ pub enum RemapScheme { NonCompiler, } -#[derive(Debug, Hash, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)] +#[derive(Debug, Hash, Clone, Copy, PartialEq, Eq)] pub enum CLang { C, Cxx, @@ -1743,6 +1743,7 @@ impl Build { /// /// If `src` is a symlink, `src` will be resolved to the actual path /// and copied to `dst` instead of the symlink itself. + #[track_caller] pub fn resolve_symlink_and_copy(&self, src: &Path, dst: &Path) { self.copy_link_internal(src, dst, true); } @@ -1751,6 +1752,7 @@ impl Build { /// Attempts to use hard links if possible, falling back to copying. /// You can neither rely on this being a copy nor it being a link, /// so do not write to dst. + #[track_caller] pub fn copy_link(&self, src: &Path, dst: &Path, file_type: FileType) { self.copy_link_internal(src, dst, false); @@ -1765,6 +1767,7 @@ impl Build { } } + #[track_caller] fn copy_link_internal(&self, src: &Path, dst: &Path, dereference_symlinks: bool) { if self.config.dry_run() { return; @@ -1773,6 +1776,10 @@ impl Build { if src == dst { return; } + + #[cfg(feature = "tracing")] + let _span = trace_io!("file-copy-link", ?src, ?dst); + if let Err(e) = fs::remove_file(dst) && cfg!(windows) && e.kind() != io::ErrorKind::NotFound @@ -1815,6 +1822,7 @@ impl Build { /// Links the `src` directory recursively to `dst`. Both are assumed to exist /// when this function is called. /// Will attempt to use hard links if possible and fall back to copying. + #[track_caller] pub fn cp_link_r(&self, src: &Path, dst: &Path) { if self.config.dry_run() { return; @@ -1837,12 +1845,14 @@ impl Build { /// Will attempt to use hard links if possible and fall back to copying. /// Unwanted files or directories can be skipped /// by returning `false` from the filter function. + #[track_caller] pub fn cp_link_filtered(&self, src: &Path, dst: &Path, filter: &dyn Fn(&Path) -> bool) { // Immediately recurse with an empty relative path self.cp_link_filtered_recurse(src, dst, Path::new(""), filter) } // Inner function does the actual work + #[track_caller] fn cp_link_filtered_recurse( &self, src: &Path, @@ -1862,7 +1872,6 @@ impl Build { self.create_dir(&dst); self.cp_link_filtered_recurse(&path, &dst, &relative, filter); } else { - let _ = fs::remove_file(&dst); self.copy_link(&path, &dst, FileType::Regular); } } @@ -1904,10 +1913,15 @@ impl Build { t!(fs::read_to_string(path)) } + #[track_caller] fn create_dir(&self, dir: &Path) { if self.config.dry_run() { return; } + + #[cfg(feature = "tracing")] + let _span = trace_io!("dir-create", ?dir); + t!(fs::create_dir_all(dir)) } @@ -1915,6 +1929,10 @@ impl Build { if self.config.dry_run() { return; } + + #[cfg(feature = "tracing")] + let _span = trace_io!("dir-remove", ?dir); + t!(fs::remove_dir_all(dir)) } diff --git a/src/bootstrap/src/utils/tracing.rs b/src/bootstrap/src/utils/tracing.rs index 428ba013c98..472781ffa73 100644 --- a/src/bootstrap/src/utils/tracing.rs +++ b/src/bootstrap/src/utils/tracing.rs @@ -49,13 +49,36 @@ macro_rules! error { } #[cfg(feature = "tracing")] +pub const IO_SPAN_TARGET: &str = "IO"; + +/// Create a tracing span around an I/O operation, if tracing is enabled. +/// Note that at least one tracing value field has to be passed to this macro, otherwise it will not +/// compile. +#[macro_export] +macro_rules! trace_io { + ($name:expr, $($args:tt)*) => { + ::tracing::trace_span!( + target: $crate::utils::tracing::IO_SPAN_TARGET, + $name, + $($args)*, + location = $crate::utils::tracing::format_location(*::std::panic::Location::caller()) + ).entered() + } +} + +#[cfg(feature = "tracing")] +pub fn format_location(location: std::panic::Location<'static>) -> String { + format!("{}:{}", location.file(), location.line()) +} + +#[cfg(feature = "tracing")] const COMMAND_SPAN_TARGET: &str = "COMMAND"; #[cfg(feature = "tracing")] pub fn trace_cmd(command: &crate::BootstrapCommand) -> tracing::span::EnteredSpan { let fingerprint = command.fingerprint(); let location = command.get_created_location(); - let location = format!("{}:{}", location.file(), location.line()); + let location = format_location(location); tracing::span!( target: COMMAND_SPAN_TARGET, @@ -84,6 +107,7 @@ mod inner { use std::fmt::Debug; use std::fs::File; use std::io::Write; + use std::path::{Path, PathBuf}; use std::sync::atomic::Ordering; use chrono::{DateTime, Utc}; @@ -93,8 +117,8 @@ mod inner { use tracing_subscriber::registry::{LookupSpan, SpanRef}; use tracing_subscriber::{EnvFilter, Layer}; + use super::{COMMAND_SPAN_TARGET, IO_SPAN_TARGET}; use crate::STEP_SPAN_TARGET; - use crate::utils::tracing::COMMAND_SPAN_TARGET; pub fn setup_tracing(env_name: &str) -> TracingGuard { let filter = EnvFilter::from_env(env_name); @@ -291,6 +315,23 @@ mod inner { Ok(()) } + // Write fields while treating the "location" field specially, and assuming that it + // contains the source file location relevant to the span. + let write_with_location = |writer: &mut W| -> std::io::Result<()> { + if let Some(values) = field_values { + write_fields( + writer, + values.fields.iter().filter(|(name, _)| *name != "location"), + )?; + let location = + &values.fields.iter().find(|(name, _)| *name == "location").unwrap().1; + let (filename, line) = location.rsplit_once(':').unwrap(); + let filename = shorten_filename(filename); + write!(writer, " ({filename}:{line})",)?; + } + Ok(()) + }; + // We handle steps specially. We instrument them dynamically in `Builder::ensure`, // and we want to have custom name for each step span. But tracing doesn't allow setting // dynamic span names. So we detect step spans here and override their name. @@ -311,17 +352,11 @@ mod inner { // Executed command COMMAND_SPAN_TARGET => { write!(writer, "{}", span.name())?; - if let Some(values) = field_values { - write_fields( - writer, - values.fields.iter().filter(|(name, _)| *name != "location"), - )?; - write!( - writer, - " ({})", - values.fields.iter().find(|(name, _)| *name == "location").unwrap().1 - )?; - } + write_with_location(writer)?; + } + IO_SPAN_TARGET => { + write!(writer, "{}", span.name())?; + write_with_location(writer)?; } // Other span _ => { @@ -342,21 +377,10 @@ mod inner { writer: &mut W, metadata: &'static tracing::Metadata<'static>, ) -> std::io::Result<()> { - use std::path::{Path, PathBuf}; - if let Some(filename) = metadata.file() { - // Keep only the module name and file name to make it shorter - let filename: PathBuf = Path::new(filename) - .components() - // Take last two path components - .rev() - .take(2) - .collect::<Vec<_>>() - .into_iter() - .rev() - .collect(); - - write!(writer, " ({}", filename.display())?; + let filename = shorten_filename(filename); + + write!(writer, " ({filename}")?; if let Some(line) = metadata.line() { write!(writer, ":{line}")?; } @@ -365,6 +389,21 @@ mod inner { Ok(()) } + /// Keep only the module name and file name to make it shorter + fn shorten_filename(filename: &str) -> String { + Path::new(filename) + .components() + // Take last two path components + .rev() + .take(2) + .collect::<Vec<_>>() + .into_iter() + .rev() + .collect::<PathBuf>() + .display() + .to_string() + } + impl<S> Layer<S> for TracingPrinter where S: Subscriber, diff --git a/src/build_helper/src/util.rs b/src/build_helper/src/util.rs index a8355f774e9..1bdbb7515e2 100644 --- a/src/build_helper/src/util.rs +++ b/src/build_helper/src/util.rs @@ -3,6 +3,8 @@ use std::io::{BufRead, BufReader}; use std::path::Path; use std::process::Command; +use crate::ci::CiEnv; + /// Invokes `build_helper::util::detail_exit` with `cfg!(test)` /// /// This is a macro instead of a function so that it uses `cfg(test)` in the *calling* crate, not in build helper. @@ -20,6 +22,15 @@ pub fn detail_exit(code: i32, is_test: bool) -> ! { if is_test { panic!("status code: {code}"); } else { + // If we're in CI, print the current bootstrap invocation command, to make it easier to + // figure out what exactly has failed. + if CiEnv::is_ci() { + // Skip the first argument, as it will be some absolute path to the bootstrap binary. + let bootstrap_args = + std::env::args().skip(1).map(|a| a.to_string()).collect::<Vec<_>>().join(" "); + eprintln!("Bootstrap failed while executing `{bootstrap_args}`"); + } + // otherwise, exit with provided status code std::process::exit(code); } diff --git a/src/ci/docker/host-x86_64/tidy/Dockerfile b/src/ci/docker/host-x86_64/tidy/Dockerfile index ee1ae5410ee..c8558689d3b 100644 --- a/src/ci/docker/host-x86_64/tidy/Dockerfile +++ b/src/ci/docker/host-x86_64/tidy/Dockerfile @@ -45,4 +45,4 @@ RUN bash -c 'npm install -g eslint@$(cat /tmp/eslint.version)' # NOTE: intentionally uses python2 for x.py so we can test it still works. # validate-toolstate only runs in our CI, so it's ok for it to only support python3. ENV SCRIPT TIDY_PRINT_DIFF=1 python2.7 ../x.py test --stage 0 \ - src/tools/tidy tidyselftest --extra-checks=py,cpp,js + src/tools/tidy tidyselftest --extra-checks=py,cpp,js,spellcheck diff --git a/src/ci/docker/host-x86_64/tidy/eslint.version b/src/ci/docker/host-x86_64/tidy/eslint.version index 1acea15afd6..42890ac0095 100644 --- a/src/ci/docker/host-x86_64/tidy/eslint.version +++ b/src/ci/docker/host-x86_64/tidy/eslint.version @@ -1 +1 @@ -8.6.0 \ No newline at end of file +8.57.1 diff --git a/src/ci/scripts/free-disk-space-windows-wait.py b/src/ci/scripts/free-disk-space-windows-wait.py index b8612bb71c2..d510781d534 100644 --- a/src/ci/scripts/free-disk-space-windows-wait.py +++ b/src/ci/scripts/free-disk-space-windows-wait.py @@ -61,12 +61,27 @@ def read_pid_from_file() -> int: ) from e -def main() -> int: - pid = read_pid_from_file() +def wait_for_process(pid: int): + timeout_duration_seconds = 5 * 60 + interval_seconds = 3 + max_attempts = timeout_duration_seconds / interval_seconds + attempts = 0 # Poll until process exits while is_process_running(pid): - time.sleep(3) + if attempts >= max_attempts: + print( + "::warning::Timeout expired while waiting for the disk cleanup process to finish." + ) + break + time.sleep(interval_seconds) + attempts += 1 + + +def main() -> int: + pid = read_pid_from_file() + + wait_for_process(pid) print_logs() diff --git a/src/doc/rustc-dev-guide/src/building/bootstrapping/debugging-bootstrap.md b/src/doc/rustc-dev-guide/src/building/bootstrapping/debugging-bootstrap.md index fb90c0fdb43..93b11c0690a 100644 --- a/src/doc/rustc-dev-guide/src/building/bootstrapping/debugging-bootstrap.md +++ b/src/doc/rustc-dev-guide/src/building/bootstrapping/debugging-bootstrap.md @@ -81,9 +81,11 @@ There are two orthogonal ways to control which kind of tracing logs you want: - If you select a level, all events/spans with an equal or higher priority level will be shown. 2. You can also control the log **target**, e.g. `bootstrap` or `bootstrap::core::config` or a custom target like `CONFIG_HANDLING` or `STEP`. - Custom targets are used to limit what kinds of spans you are interested in, as the `BOOTSTRAP_TRACING=trace` output can be quite verbose. Currently, you can use the following custom targets: - - `CONFIG_HANDLING`: show spans related to config handling - - `STEP`: show all executed steps. Note that executed commands have `info` event level. - - `COMMAND`: show all executed commands. Note that executed commands have `trace` event level. + - `CONFIG_HANDLING`: show spans related to config handling. + - `STEP`: show all executed steps. Executed commands have `info` event level. + - `COMMAND`: show all executed commands. Executed commands have `trace` event level. + - `IO`: show performed I/O operations. Executed commands have `trace` event level. + - Note that many I/O are currently not being traced. You can of course combine them (custom target logs are typically gated behind `TRACE` log level additionally): diff --git a/src/doc/rustc-dev-guide/src/sanitizers.md b/src/doc/rustc-dev-guide/src/sanitizers.md index 29d9056c15d..34c78d4d952 100644 --- a/src/doc/rustc-dev-guide/src/sanitizers.md +++ b/src/doc/rustc-dev-guide/src/sanitizers.md @@ -45,7 +45,7 @@ implementation: [marked][sanitizer-attribute] with appropriate LLVM attribute: `SanitizeAddress`, `SanitizeHWAddress`, `SanitizeMemory`, or `SanitizeThread`. By default all functions are instrumented, but this - behaviour can be changed with `#[no_sanitize(...)]`. + behaviour can be changed with `#[sanitize(xyz = "on|off")]`. * The decision whether to perform instrumentation or not is possible only at a function granularity. In the cases were those decision differ between diff --git a/src/doc/rustc/src/codegen-options/index.md b/src/doc/rustc/src/codegen-options/index.md index 07eafdf4c4c..445b10188e3 100644 --- a/src/doc/rustc/src/codegen-options/index.md +++ b/src/doc/rustc/src/codegen-options/index.md @@ -375,12 +375,12 @@ linking time. It takes one of the following values: * `y`, `yes`, `on`, `true`, `fat`, or no value: perform "fat" LTO which attempts to perform optimizations across all crates within the dependency graph. -* `n`, `no`, `off`, `false`: disables LTO. * `thin`: perform ["thin" LTO](http://blog.llvm.org/2016/06/thinlto-scalable-and-incremental-lto.html). This is similar to "fat", but takes substantially less time to run while still achieving performance gains similar to "fat". For larger projects like the Rust compiler, ThinLTO can even result in better performance than fat LTO. +* `n`, `no`, `off`, `false`: disables LTO. If `-C lto` is not specified, then the compiler will attempt to perform "thin local LTO" which performs "thin" LTO on the local crate only across its diff --git a/src/doc/rustc/src/platform-support/openharmony.md b/src/doc/rustc/src/platform-support/openharmony.md index 3acdc3707a8..de3b83d6c2c 100644 --- a/src/doc/rustc/src/platform-support/openharmony.md +++ b/src/doc/rustc/src/platform-support/openharmony.md @@ -16,7 +16,7 @@ system. ## Target maintainers [@Amanieu](https://github.com/Amanieu) -[@lubinglun](https://github.com/lubinglun) +[@cceerczw](https://github.com/cceerczw) ## Requirements diff --git a/src/doc/rustc/src/target-tier-policy.md b/src/doc/rustc/src/target-tier-policy.md index a0acfdf0e4a..28d3dc32a63 100644 --- a/src/doc/rustc/src/target-tier-policy.md +++ b/src/doc/rustc/src/target-tier-policy.md @@ -534,10 +534,10 @@ tests, and will reject patches that fail to build or pass the testsuite on a target. We hold tier 1 targets to our highest standard of requirements. A proposed new tier 1 target must be reviewed and approved by the compiler team -based on these requirements. In addition, the release team must approve the -viability and value of supporting the target. For a tier 1 target, this will +based on these requirements. In addition, the infra team must approve the +viability of supporting the target. For a tier 1 target, this will typically take place via a full RFC proposing the target, to be jointly -reviewed and approved by the compiler team and release team. +reviewed and approved by the compiler team and infra team. In addition, the infrastructure team must approve the integration of the target into Continuous Integration (CI), and the tier 1 CI-related requirements. This @@ -617,7 +617,7 @@ including the infrastructure team in the RFC proposing the target. A tier 1 target may be demoted if it no longer meets these requirements but still meets the requirements for a lower tier. Any proposal for demotion of a tier 1 target requires a full RFC process, with approval by the compiler and -release teams. Any such proposal will be communicated widely to the Rust +infra teams. Any such proposal will be communicated widely to the Rust community, both when initially proposed and before being dropped from a stable release. A tier 1 target is highly unlikely to be directly removed without first being demoted to tier 2 or tier 3. (The amount of time between such @@ -628,7 +628,7 @@ planned and scheduled action.) Raising the baseline expectations of a tier 1 target (such as the minimum CPU features or OS version required) requires the approval of the compiler and -release teams, and should be widely communicated as well, but does not +infra teams, and should be widely communicated as well, but does not necessarily require a full RFC. ### Tier 1 with host tools @@ -638,11 +638,11 @@ host (such as `rustc` and `cargo`). This allows the target to be used as a development platform, not just a compilation target. A proposed new tier 1 target with host tools must be reviewed and approved by -the compiler team based on these requirements. In addition, the release team -must approve the viability and value of supporting host tools for the target. +the compiler team based on these requirements. In addition, the infra team +must approve the viability of supporting host tools for the target. For a tier 1 target, this will typically take place via a full RFC proposing the target, to be jointly reviewed and approved by the compiler team and -release team. +infra team. In addition, the infrastructure team must approve the integration of the target's host tools into Continuous Integration (CI), and the CI-related @@ -697,7 +697,7 @@ target with host tools may be demoted (including having its host tools dropped, or being demoted to tier 2 with host tools) if it no longer meets these requirements but still meets the requirements for a lower tier. Any proposal for demotion of a tier 1 target (with or without host tools) requires a full -RFC process, with approval by the compiler and release teams. Any such proposal +RFC process, with approval by the compiler and infra teams. Any such proposal will be communicated widely to the Rust community, both when initially proposed and before being dropped from a stable release. diff --git a/src/doc/unstable-book/src/compiler-flags/indirect-branch-cs-prefix.md b/src/doc/unstable-book/src/compiler-flags/indirect-branch-cs-prefix.md new file mode 100644 index 00000000000..040e2d41701 --- /dev/null +++ b/src/doc/unstable-book/src/compiler-flags/indirect-branch-cs-prefix.md @@ -0,0 +1,19 @@ +# `indirect-branch-cs-prefix` + +The tracking issue for this feature is: https://github.com/rust-lang/rust/issues/116852. + +------------------------ + +Option `-Zindirect-branch-cs-prefix` controls whether a `cs` prefix is added to +`call` and `jmp` to indirect thunks. + +It is equivalent to [Clang]'s and [GCC]'s `-mindirect-branch-cs-prefix`. The +Linux kernel uses it for RETPOLINE builds. For details, see +[LLVM commit 6f867f910283] ("[X86] Support ``-mindirect-branch-cs-prefix`` for +call and jmp to indirect thunk") which introduces the feature. + +Only x86 and x86_64 are supported. + +[Clang]: https://clang.llvm.org/docs/ClangCommandLineReference.html#cmdoption-clang-mindirect-branch-cs-prefix +[GCC]: https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html#index-mindirect-branch-cs-prefix +[LLVM commit 6f867f910283]: https://github.com/llvm/llvm-project/commit/6f867f9102838ebe314c1f3661fdf95700386e5a diff --git a/src/doc/unstable-book/src/language-features/no-sanitize.md b/src/doc/unstable-book/src/language-features/no-sanitize.md deleted file mode 100644 index 28c683934d4..00000000000 --- a/src/doc/unstable-book/src/language-features/no-sanitize.md +++ /dev/null @@ -1,29 +0,0 @@ -# `no_sanitize` - -The tracking issue for this feature is: [#39699] - -[#39699]: https://github.com/rust-lang/rust/issues/39699 - ------------------------- - -The `no_sanitize` attribute can be used to selectively disable sanitizer -instrumentation in an annotated function. This might be useful to: avoid -instrumentation overhead in a performance critical function, or avoid -instrumenting code that contains constructs unsupported by given sanitizer. - -The precise effect of this annotation depends on particular sanitizer in use. -For example, with `no_sanitize(thread)`, the thread sanitizer will no longer -instrument non-atomic store / load operations, but it will instrument atomic -operations to avoid reporting false positives and provide meaning full stack -traces. - -## Examples - -``` rust -#![feature(no_sanitize)] - -#[no_sanitize(address)] -fn foo() { - // ... -} -``` diff --git a/src/doc/unstable-book/src/language-features/sanitize.md b/src/doc/unstable-book/src/language-features/sanitize.md new file mode 100644 index 00000000000..fcd099cb36e --- /dev/null +++ b/src/doc/unstable-book/src/language-features/sanitize.md @@ -0,0 +1,73 @@ +# `sanitize` + +The tracking issue for this feature is: [#39699] + +[#39699]: https://github.com/rust-lang/rust/issues/39699 + +------------------------ + +The `sanitize` attribute can be used to selectively disable or enable sanitizer +instrumentation in an annotated function. This might be useful to: avoid +instrumentation overhead in a performance critical function, or avoid +instrumenting code that contains constructs unsupported by given sanitizer. + +The precise effect of this annotation depends on particular sanitizer in use. +For example, with `sanitize(thread = "off")`, the thread sanitizer will no +longer instrument non-atomic store / load operations, but it will instrument +atomic operations to avoid reporting false positives and provide meaning full +stack traces. + +This attribute was previously named `no_sanitize`. + +## Examples + +``` rust +#![feature(sanitize)] + +#[sanitize(address = "off")] +fn foo() { + // ... +} +``` + +It is also possible to disable sanitizers for entire modules and enable them +for single items or functions. + +```rust +#![feature(sanitize)] + +#[sanitize(address = "off")] +mod foo { + fn unsanitized() { + // ... + } + + #[sanitize(address = "on")] + fn sanitized() { + // ... + } +} +``` + +It's also applicable to impl blocks. + +```rust +#![feature(sanitize)] + +trait MyTrait { + fn foo(&self); + fn bar(&self); +} + +#[sanitize(address = "off")] +impl MyTrait for () { + fn foo(&self) { + // ... + } + + #[sanitize(address = "on")] + fn bar(&self) { + // ... + } +} +``` diff --git a/src/etc/htmldocck.py b/src/etc/htmldocck.py index 72975cc6206..8d7f7341c2e 100755 --- a/src/etc/htmldocck.py +++ b/src/etc/htmldocck.py @@ -15,6 +15,7 @@ import os.path import re import shlex from collections import namedtuple +from pathlib import Path try: from html.parser import HTMLParser @@ -242,6 +243,11 @@ class CachedFiles(object): return self.last_path def get_absolute_path(self, path): + if "*" in path: + paths = list(Path(self.root).glob(path)) + if len(paths) != 1: + raise FailedCheck("glob path does not resolve to one file") + path = str(paths[0]) return os.path.join(self.root, path) def get_file(self, path): diff --git a/src/librustdoc/Cargo.toml b/src/librustdoc/Cargo.toml index fdde8309cf9..5d36ffc2d3a 100644 --- a/src/librustdoc/Cargo.toml +++ b/src/librustdoc/Cargo.toml @@ -21,6 +21,7 @@ rustdoc-json-types = { path = "../rustdoc-json-types" } serde = { version = "1.0", features = ["derive"] } serde_json = "1.0" smallvec = "1.8.1" +stringdex = { version = "0.0.1-alpha4" } tempfile = "3" threadpool = "1.8.1" tracing = "0.1" diff --git a/src/librustdoc/build.rs b/src/librustdoc/build.rs index 07269d5bdc2..5b497183ae6 100644 --- a/src/librustdoc/build.rs +++ b/src/librustdoc/build.rs @@ -10,6 +10,7 @@ fn main() { "static/css/normalize.css", "static/js/main.js", "static/js/search.js", + "static/js/stringdex.js", "static/js/settings.js", "static/js/src-script.js", "static/js/storage.js", diff --git a/src/librustdoc/clean/types.rs b/src/librustdoc/clean/types.rs index 26b087feb16..46aaa0068de 100644 --- a/src/librustdoc/clean/types.rs +++ b/src/librustdoc/clean/types.rs @@ -1552,10 +1552,10 @@ impl Type { matches!(self, Type::Path { path: Path { res: Res::Def(DefKind::TyAlias, _), .. } }) } - /// Check if two types are "the same" for documentation purposes. + /// Check if this type is a subtype of another type for documentation purposes. /// /// This is different from `Eq`, because it knows that things like - /// `Placeholder` are possible matches for everything. + /// `Infer` and generics have special subtyping rules. /// /// This relation is not commutative when generics are involved: /// @@ -1566,8 +1566,8 @@ impl Type { /// let cache = Cache::new(false); /// let generic = Type::Generic(rustc_span::symbol::sym::Any); /// let unit = Type::Primitive(PrimitiveType::Unit); - /// assert!(!generic.is_same(&unit, &cache)); - /// assert!(unit.is_same(&generic, &cache)); + /// assert!(!generic.is_doc_subtype_of(&unit, &cache)); + /// assert!(unit.is_doc_subtype_of(&generic, &cache)); /// ``` /// /// An owned type is also the same as its borrowed variants (this is commutative), diff --git a/src/librustdoc/formats/cache.rs b/src/librustdoc/formats/cache.rs index 80399cf3842..cb6837dd614 100644 --- a/src/librustdoc/formats/cache.rs +++ b/src/librustdoc/formats/cache.rs @@ -1,6 +1,5 @@ use std::mem; -use rustc_ast::join_path_syms; use rustc_data_structures::fx::{FxHashMap, FxHashSet, FxIndexMap, FxIndexSet}; use rustc_hir::StabilityLevel; use rustc_hir::def_id::{CrateNum, DefId, DefIdMap, DefIdSet}; @@ -48,7 +47,7 @@ pub(crate) struct Cache { /// Similar to `paths`, but only holds external paths. This is only used for /// generating explicit hyperlinks to other crates. - pub(crate) external_paths: FxHashMap<DefId, (Vec<Symbol>, ItemType)>, + pub(crate) external_paths: FxIndexMap<DefId, (Vec<Symbol>, ItemType)>, /// Maps local `DefId`s of exported types to fully qualified paths. /// Unlike 'paths', this mapping ignores any renames that occur @@ -574,7 +573,6 @@ fn add_item_to_search_index(tcx: TyCtxt<'_>, cache: &mut Cache, item: &clean::It clean::ItemKind::ImportItem(import) => import.source.did.unwrap_or(item_def_id), _ => item_def_id, }; - let path = join_path_syms(parent_path); let impl_id = if let Some(ParentStackItem::Impl { item_id, .. }) = cache.parent_stack.last() { item_id.as_def_id() } else { @@ -593,11 +591,11 @@ fn add_item_to_search_index(tcx: TyCtxt<'_>, cache: &mut Cache, item: &clean::It ty: item.type_(), defid: Some(defid), name, - path, + module_path: parent_path.to_vec(), desc, parent: parent_did, parent_idx: None, - exact_path: None, + exact_module_path: None, impl_id, search_type, aliases, diff --git a/src/librustdoc/formats/item_type.rs b/src/librustdoc/formats/item_type.rs index 1dba84aa44c..142a9d7d8af 100644 --- a/src/librustdoc/formats/item_type.rs +++ b/src/librustdoc/formats/item_type.rs @@ -4,7 +4,7 @@ use std::fmt; use rustc_hir::def::{CtorOf, DefKind, MacroKinds}; use rustc_span::hygiene::MacroKind; -use serde::{Serialize, Serializer}; +use serde::{Deserialize, Deserializer, Serialize, Serializer, de}; use crate::clean; @@ -68,6 +68,52 @@ impl Serialize for ItemType { } } +impl<'de> Deserialize<'de> for ItemType { + fn deserialize<D>(deserializer: D) -> Result<ItemType, D::Error> + where + D: Deserializer<'de>, + { + struct ItemTypeVisitor; + impl<'de> de::Visitor<'de> for ItemTypeVisitor { + type Value = ItemType; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "an integer between 0 and 25") + } + fn visit_u64<E: de::Error>(self, v: u64) -> Result<ItemType, E> { + Ok(match v { + 0 => ItemType::Keyword, + 1 => ItemType::Primitive, + 2 => ItemType::Module, + 3 => ItemType::ExternCrate, + 4 => ItemType::Import, + 5 => ItemType::Struct, + 6 => ItemType::Enum, + 7 => ItemType::Function, + 8 => ItemType::TypeAlias, + 9 => ItemType::Static, + 10 => ItemType::Trait, + 11 => ItemType::Impl, + 12 => ItemType::TyMethod, + 13 => ItemType::Method, + 14 => ItemType::StructField, + 15 => ItemType::Variant, + 16 => ItemType::Macro, + 17 => ItemType::AssocType, + 18 => ItemType::Constant, + 19 => ItemType::AssocConst, + 20 => ItemType::Union, + 21 => ItemType::ForeignType, + 23 => ItemType::ProcAttribute, + 24 => ItemType::ProcDerive, + 25 => ItemType::TraitAlias, + _ => return Err(E::missing_field("unknown number")), + }) + } + } + deserializer.deserialize_any(ItemTypeVisitor) + } +} + impl<'a> From<&'a clean::Item> for ItemType { fn from(item: &'a clean::Item) -> ItemType { let kind = match &item.kind { @@ -198,6 +244,10 @@ impl ItemType { pub(crate) fn is_adt(&self) -> bool { matches!(self, ItemType::Struct | ItemType::Union | ItemType::Enum) } + /// Keep this the same as isFnLikeTy in search.js + pub(crate) fn is_fn_like(&self) -> bool { + matches!(self, ItemType::Function | ItemType::Method | ItemType::TyMethod) + } } impl fmt::Display for ItemType { diff --git a/src/librustdoc/html/layout.rs b/src/librustdoc/html/layout.rs index 2782e8e0058..5db742bdebf 100644 --- a/src/librustdoc/html/layout.rs +++ b/src/librustdoc/html/layout.rs @@ -27,6 +27,7 @@ pub(crate) struct Layout { pub(crate) struct Page<'a> { pub(crate) title: &'a str, + pub(crate) short_title: &'a str, pub(crate) css_class: &'a str, pub(crate) root_path: &'a str, pub(crate) static_root_path: Option<&'a str>, diff --git a/src/librustdoc/html/render/context.rs b/src/librustdoc/html/render/context.rs index 5ceb1fc988d..e4fca09d64f 100644 --- a/src/librustdoc/html/render/context.rs +++ b/src/librustdoc/html/render/context.rs @@ -204,6 +204,18 @@ impl<'tcx> Context<'tcx> { if !is_module { title.push_str(it.name.unwrap().as_str()); } + let short_title; + let short_title = if is_module { + let module_name = self.current.last().unwrap(); + short_title = if it.is_crate() { + format!("Crate {module_name}") + } else { + format!("Module {module_name}") + }; + &short_title[..] + } else { + it.name.as_ref().unwrap().as_str() + }; if !it.is_primitive() && !it.is_keyword() { if !is_module { title.push_str(" in "); @@ -240,6 +252,7 @@ impl<'tcx> Context<'tcx> { root_path: &self.root_path(), static_root_path: self.shared.static_root_path.as_deref(), title: &title, + short_title, description: &desc, resource_suffix: &self.shared.resource_suffix, rust_logo: has_doc_flag(self.tcx(), LOCAL_CRATE.as_def_id(), sym::rust_logo), @@ -617,6 +630,7 @@ impl<'tcx> FormatRenderer<'tcx> for Context<'tcx> { let shared = &self.shared; let mut page = layout::Page { title: "List of all items in this crate", + short_title: "All", css_class: "mod sys", root_path: "../", static_root_path: shared.static_root_path.as_deref(), diff --git a/src/librustdoc/html/render/mod.rs b/src/librustdoc/html/render/mod.rs index a46253237db..8d7f0577506 100644 --- a/src/librustdoc/html/render/mod.rs +++ b/src/librustdoc/html/render/mod.rs @@ -130,11 +130,11 @@ pub(crate) struct IndexItem { pub(crate) ty: ItemType, pub(crate) defid: Option<DefId>, pub(crate) name: Symbol, - pub(crate) path: String, + pub(crate) module_path: Vec<Symbol>, pub(crate) desc: String, pub(crate) parent: Option<DefId>, - pub(crate) parent_idx: Option<isize>, - pub(crate) exact_path: Option<String>, + pub(crate) parent_idx: Option<usize>, + pub(crate) exact_module_path: Option<Vec<Symbol>>, pub(crate) impl_id: Option<DefId>, pub(crate) search_type: Option<IndexItemFunctionType>, pub(crate) aliases: Box<[Symbol]>, @@ -150,6 +150,19 @@ struct RenderType { } impl RenderType { + fn size(&self) -> usize { + let mut size = 1; + if let Some(generics) = &self.generics { + size += generics.iter().map(RenderType::size).sum::<usize>(); + } + if let Some(bindings) = &self.bindings { + for (_, constraints) in bindings.iter() { + size += 1; + size += constraints.iter().map(RenderType::size).sum::<usize>(); + } + } + size + } // Types are rendered as lists of lists, because that's pretty compact. // The contents of the lists are always integers in self-terminating hex // form, handled by `RenderTypeId::write_to_string`, so no commas are @@ -191,6 +204,62 @@ impl RenderType { write_optional_id(self.id, string); } } + fn read_from_bytes(string: &[u8]) -> (RenderType, usize) { + let mut i = 0; + if string[i] == b'{' { + i += 1; + let (id, offset) = RenderTypeId::read_from_bytes(&string[i..]); + i += offset; + let generics = if string[i] == b'{' { + i += 1; + let mut generics = Vec::new(); + while string[i] != b'}' { + let (ty, offset) = RenderType::read_from_bytes(&string[i..]); + i += offset; + generics.push(ty); + } + assert!(string[i] == b'}'); + i += 1; + Some(generics) + } else { + None + }; + let bindings = if string[i] == b'{' { + i += 1; + let mut bindings = Vec::new(); + while string[i] == b'{' { + i += 1; + let (binding, boffset) = RenderTypeId::read_from_bytes(&string[i..]); + i += boffset; + let mut bconstraints = Vec::new(); + assert!(string[i] == b'{'); + i += 1; + while string[i] != b'}' { + let (constraint, coffset) = RenderType::read_from_bytes(&string[i..]); + i += coffset; + bconstraints.push(constraint); + } + assert!(string[i] == b'}'); + i += 1; + bindings.push((binding.unwrap(), bconstraints)); + assert!(string[i] == b'}'); + i += 1; + } + assert!(string[i] == b'}'); + i += 1; + Some(bindings) + } else { + None + }; + assert!(string[i] == b'}'); + i += 1; + (RenderType { id, generics, bindings }, i) + } else { + let (id, offset) = RenderTypeId::read_from_bytes(string); + i += offset; + (RenderType { id, generics: None, bindings: None }, i) + } + } } #[derive(Clone, Copy, Debug, Eq, PartialEq)] @@ -212,7 +281,20 @@ impl RenderTypeId { RenderTypeId::Index(idx) => (*idx).try_into().unwrap(), _ => panic!("must convert render types to indexes before serializing"), }; - search_index::encode::write_vlqhex_to_string(id, string); + search_index::encode::write_signed_vlqhex_to_string(id, string); + } + fn read_from_bytes(string: &[u8]) -> (Option<RenderTypeId>, usize) { + let Some((value, offset)) = search_index::encode::read_signed_vlqhex_from_string(string) + else { + return (None, 0); + }; + let value = isize::try_from(value).unwrap(); + let ty = match value { + ..0 => Some(RenderTypeId::Index(value)), + 0 => None, + 1.. => Some(RenderTypeId::Index(value - 1)), + }; + (ty, offset) } } @@ -226,12 +308,64 @@ pub(crate) struct IndexItemFunctionType { } impl IndexItemFunctionType { - fn write_to_string<'a>( - &'a self, - string: &mut String, - backref_queue: &mut VecDeque<&'a IndexItemFunctionType>, - ) { - assert!(backref_queue.len() <= 16); + fn size(&self) -> usize { + self.inputs.iter().map(RenderType::size).sum::<usize>() + + self.output.iter().map(RenderType::size).sum::<usize>() + + self + .where_clause + .iter() + .map(|constraints| constraints.iter().map(RenderType::size).sum::<usize>()) + .sum::<usize>() + } + fn read_from_string_without_param_names(string: &[u8]) -> (IndexItemFunctionType, usize) { + let mut i = 0; + if string[i] == b'`' { + return ( + IndexItemFunctionType { + inputs: Vec::new(), + output: Vec::new(), + where_clause: Vec::new(), + param_names: Vec::new(), + }, + 1, + ); + } + assert_eq!(b'{', string[i]); + i += 1; + fn read_args_from_string(string: &[u8]) -> (Vec<RenderType>, usize) { + let mut i = 0; + let mut params = Vec::new(); + if string[i] == b'{' { + // multiple params + i += 1; + while string[i] != b'}' { + let (ty, offset) = RenderType::read_from_bytes(&string[i..]); + i += offset; + params.push(ty); + } + i += 1; + } else if string[i] != b'}' { + let (tyid, offset) = RenderTypeId::read_from_bytes(&string[i..]); + params.push(RenderType { id: tyid, generics: None, bindings: None }); + i += offset; + } + (params, i) + } + let (inputs, offset) = read_args_from_string(&string[i..]); + i += offset; + let (output, offset) = read_args_from_string(&string[i..]); + i += offset; + let mut where_clause = Vec::new(); + while string[i] != b'}' { + let (constraint, offset) = read_args_from_string(&string[i..]); + i += offset; + where_clause.push(constraint); + } + assert_eq!(b'}', string[i], "{} {}", String::from_utf8_lossy(&string), i); + i += 1; + (IndexItemFunctionType { inputs, output, where_clause, param_names: Vec::new() }, i) + } + fn write_to_string_without_param_names<'a>(&'a self, string: &mut String) { // If we couldn't figure out a type, just write 0, // which is encoded as `` ` `` (see RenderTypeId::write_to_string). let has_missing = self @@ -241,18 +375,7 @@ impl IndexItemFunctionType { .any(|i| i.id.is_none() && i.generics.is_none()); if has_missing { string.push('`'); - } else if let Some(idx) = backref_queue.iter().position(|other| *other == self) { - // The backref queue has 16 items, so backrefs use - // a single hexit, disjoint from the ones used for numbers. - string.push( - char::try_from('0' as u32 + u32::try_from(idx).unwrap()) - .expect("last possible value is '?'"), - ); } else { - backref_queue.push_front(self); - if backref_queue.len() > 16 { - backref_queue.pop_back(); - } string.push('{'); match &self.inputs[..] { [one] if one.generics.is_none() && one.bindings.is_none() => { diff --git a/src/librustdoc/html/render/print_item.rs b/src/librustdoc/html/render/print_item.rs index 759f53974f5..407238d66b8 100644 --- a/src/librustdoc/html/render/print_item.rs +++ b/src/librustdoc/html/render/print_item.rs @@ -35,6 +35,7 @@ use crate::html::format::{ visibility_print_with_space, }; use crate::html::markdown::{HeadingOffset, MarkdownSummaryLine}; +use crate::html::render::sidebar::filters; use crate::html::render::{document_full, document_item_info}; use crate::html::url_parts_builder::UrlPartsBuilder; diff --git a/src/librustdoc/html/render/search_index.rs b/src/librustdoc/html/render/search_index.rs index e2f86b8a854..41657e290ea 100644 --- a/src/librustdoc/html/render/search_index.rs +++ b/src/librustdoc/html/render/search_index.rs @@ -1,72 +1,1169 @@ pub(crate) mod encode; +use std::collections::BTreeSet; use std::collections::hash_map::Entry; -use std::collections::{BTreeMap, VecDeque}; +use std::path::Path; -use encode::{bitmap_to_string, write_vlqhex_to_string}; use rustc_ast::join_path_syms; -use rustc_data_structures::fx::{FxHashMap, FxIndexMap}; +use rustc_data_structures::fx::{FxHashMap, FxHashSet, FxIndexMap}; use rustc_middle::ty::TyCtxt; use rustc_span::def_id::DefId; use rustc_span::sym; use rustc_span::symbol::{Symbol, kw}; -use serde::ser::{Serialize, SerializeSeq, SerializeStruct, Serializer}; +use serde::de::{self, Deserializer, Error as _}; +use serde::ser::{SerializeSeq, Serializer}; +use serde::{Deserialize, Serialize}; +use stringdex::internals as stringdex_internals; use thin_vec::ThinVec; use tracing::instrument; use crate::clean::types::{Function, Generics, ItemId, Type, WherePredicate}; use crate::clean::{self, utils}; +use crate::error::Error; use crate::formats::cache::{Cache, OrphanImplItem}; use crate::formats::item_type::ItemType; use crate::html::markdown::short_markdown_summary; -use crate::html::render::ordered_json::OrderedJson; use crate::html::render::{self, IndexItem, IndexItemFunctionType, RenderType, RenderTypeId}; -/// The serialized search description sharded version -/// -/// The `index` is a JSON-encoded list of names and other information. -/// -/// The desc has newlined descriptions, split up by size into 128KiB shards. -/// For example, `(4, "foo\nbar\nbaz\nquux")`. -/// -/// There is no single, optimal size for these shards, because it depends on -/// configuration values that we can't predict or control, such as the version -/// of HTTP used (HTTP/1.1 would work better with larger files, while HTTP/2 -/// and 3 are more agnostic), transport compression (gzip, zstd, etc), whether -/// the search query is going to produce a large number of results or a small -/// number, the bandwidth delay product of the network... -/// -/// Gzipping some standard library descriptions to guess what transport -/// compression will do, the compressed file sizes can be as small as 4.9KiB -/// or as large as 18KiB (ignoring the final 1.9KiB shard of leftovers). -/// A "reasonable" range for files is for them to be bigger than 1KiB, -/// since that's about the amount of data that can be transferred in a -/// single TCP packet, and 64KiB, the maximum amount of data that -/// TCP can transfer in a single round trip without extensions. -/// -/// [1]: https://en.wikipedia.org/wiki/Maximum_transmission_unit#MTUs_for_common_media -/// [2]: https://en.wikipedia.org/wiki/Sliding_window_protocol#Basic_concept -/// [3]: https://learn.microsoft.com/en-us/troubleshoot/windows-server/networking/description-tcp-features +#[derive(Clone, Debug, Default, Deserialize, Serialize)] pub(crate) struct SerializedSearchIndex { - pub(crate) index: OrderedJson, - pub(crate) desc: Vec<(usize, String)>, + // data from disk + names: Vec<String>, + path_data: Vec<Option<PathData>>, + entry_data: Vec<Option<EntryData>>, + descs: Vec<String>, + function_data: Vec<Option<FunctionData>>, + alias_pointers: Vec<Option<usize>>, + // inverted index for concrete types and generics + type_data: Vec<Option<TypeData>>, + /// inverted index of generics + /// + /// - The outermost list has one entry per alpha-normalized generic. + /// + /// - The second layer is sorted by number of types that appear in the + /// type signature. The search engine iterates over these in order from + /// smallest to largest. Functions with less stuff in their type + /// signature are more likely to be what the user wants, because we never + /// show functions that are *missing* parts of the query, so removing.. + /// + /// - The final layer is the list of functions. + generic_inverted_index: Vec<Vec<Vec<u32>>>, + // generated in-memory backref cache + #[serde(skip)] + crate_paths_index: FxHashMap<(ItemType, Vec<Symbol>), usize>, +} + +impl SerializedSearchIndex { + fn load(doc_root: &Path, resource_suffix: &str) -> Result<SerializedSearchIndex, Error> { + let mut names: Vec<String> = Vec::new(); + let mut path_data: Vec<Option<PathData>> = Vec::new(); + let mut entry_data: Vec<Option<EntryData>> = Vec::new(); + let mut descs: Vec<String> = Vec::new(); + let mut function_data: Vec<Option<FunctionData>> = Vec::new(); + let mut type_data: Vec<Option<TypeData>> = Vec::new(); + let mut alias_pointers: Vec<Option<usize>> = Vec::new(); + + let mut generic_inverted_index: Vec<Vec<Vec<u32>>> = Vec::new(); + + match perform_read_strings(resource_suffix, doc_root, "name", &mut names) { + Ok(()) => { + perform_read_serde(resource_suffix, doc_root, "path", &mut path_data)?; + perform_read_serde(resource_suffix, doc_root, "entry", &mut entry_data)?; + perform_read_strings(resource_suffix, doc_root, "desc", &mut descs)?; + perform_read_serde(resource_suffix, doc_root, "function", &mut function_data)?; + perform_read_serde(resource_suffix, doc_root, "type", &mut type_data)?; + perform_read_serde(resource_suffix, doc_root, "alias", &mut alias_pointers)?; + perform_read_postings( + resource_suffix, + doc_root, + "generic_inverted_index", + &mut generic_inverted_index, + )?; + } + Err(_) => { + names.clear(); + } + } + fn perform_read_strings( + resource_suffix: &str, + doc_root: &Path, + column_name: &str, + column: &mut Vec<String>, + ) -> Result<(), Error> { + let root_path = doc_root.join(format!("search.index/root{resource_suffix}.js")); + let column_path = doc_root.join(format!("search.index/{column_name}/")); + stringdex_internals::read_data_from_disk_column( + root_path, + column_name.as_bytes(), + column_path.clone(), + &mut |_id, item| { + column.push(String::from_utf8(item.to_vec())?); + Ok(()) + }, + ) + .map_err( + |error: stringdex_internals::ReadDataError<Box<dyn std::error::Error>>| Error { + file: column_path, + error: format!("failed to read column from disk: {error}"), + }, + ) + } + fn perform_read_serde( + resource_suffix: &str, + doc_root: &Path, + column_name: &str, + column: &mut Vec<Option<impl for<'de> Deserialize<'de> + 'static>>, + ) -> Result<(), Error> { + let root_path = doc_root.join(format!("search.index/root{resource_suffix}.js")); + let column_path = doc_root.join(format!("search.index/{column_name}/")); + stringdex_internals::read_data_from_disk_column( + root_path, + column_name.as_bytes(), + column_path.clone(), + &mut |_id, item| { + if item.is_empty() { + column.push(None); + } else { + column.push(Some(serde_json::from_slice(item)?)); + } + Ok(()) + }, + ) + .map_err( + |error: stringdex_internals::ReadDataError<Box<dyn std::error::Error>>| Error { + file: column_path, + error: format!("failed to read column from disk: {error}"), + }, + ) + } + fn perform_read_postings( + resource_suffix: &str, + doc_root: &Path, + column_name: &str, + column: &mut Vec<Vec<Vec<u32>>>, + ) -> Result<(), Error> { + let root_path = doc_root.join(format!("search.index/root{resource_suffix}.js")); + let column_path = doc_root.join(format!("search.index/{column_name}/")); + stringdex_internals::read_data_from_disk_column( + root_path, + column_name.as_bytes(), + column_path.clone(), + &mut |_id, buf| { + let mut postings = Vec::new(); + encode::read_postings_from_string(&mut postings, buf); + column.push(postings); + Ok(()) + }, + ) + .map_err( + |error: stringdex_internals::ReadDataError<Box<dyn std::error::Error>>| Error { + file: column_path, + error: format!("failed to read column from disk: {error}"), + }, + ) + } + + assert_eq!(names.len(), path_data.len()); + assert_eq!(path_data.len(), entry_data.len()); + assert_eq!(entry_data.len(), descs.len()); + assert_eq!(descs.len(), function_data.len()); + assert_eq!(function_data.len(), type_data.len()); + assert_eq!(type_data.len(), alias_pointers.len()); + + // generic_inverted_index is not the same length as other columns, + // because it's actually a completely different set of objects + + let mut crate_paths_index: FxHashMap<(ItemType, Vec<Symbol>), usize> = FxHashMap::default(); + for (i, (name, path_data)) in names.iter().zip(path_data.iter()).enumerate() { + if let Some(path_data) = path_data { + let full_path = if path_data.module_path.is_empty() { + vec![Symbol::intern(name)] + } else { + let mut full_path = path_data.module_path.to_vec(); + full_path.push(Symbol::intern(name)); + full_path + }; + crate_paths_index.insert((path_data.ty, full_path), i); + } + } + + Ok(SerializedSearchIndex { + names, + path_data, + entry_data, + descs, + function_data, + type_data, + alias_pointers, + generic_inverted_index, + crate_paths_index, + }) + } + fn push( + &mut self, + name: String, + path_data: Option<PathData>, + entry_data: Option<EntryData>, + desc: String, + function_data: Option<FunctionData>, + type_data: Option<TypeData>, + alias_pointer: Option<usize>, + ) -> usize { + let index = self.names.len(); + assert_eq!(self.names.len(), self.path_data.len()); + if let Some(path_data) = &path_data + && let name = Symbol::intern(&name) + && let fqp = if path_data.module_path.is_empty() { + vec![name] + } else { + let mut v = path_data.module_path.clone(); + v.push(name); + v + } + && let Some(&other_path) = self.crate_paths_index.get(&(path_data.ty, fqp)) + && self.path_data.get(other_path).map_or(false, Option::is_some) + { + self.path_data.push(None); + } else { + self.path_data.push(path_data); + } + self.names.push(name); + assert_eq!(self.entry_data.len(), self.descs.len()); + self.entry_data.push(entry_data); + assert_eq!(self.descs.len(), self.function_data.len()); + self.descs.push(desc); + assert_eq!(self.function_data.len(), self.type_data.len()); + self.function_data.push(function_data); + assert_eq!(self.type_data.len(), self.alias_pointers.len()); + self.type_data.push(type_data); + self.alias_pointers.push(alias_pointer); + index + } + fn push_path(&mut self, name: String, path_data: PathData) -> usize { + self.push(name, Some(path_data), None, String::new(), None, None, None) + } + fn push_type(&mut self, name: String, path_data: PathData, type_data: TypeData) -> usize { + self.push(name, Some(path_data), None, String::new(), None, Some(type_data), None) + } + fn push_alias(&mut self, name: String, alias_pointer: usize) -> usize { + self.push(name, None, None, String::new(), None, None, Some(alias_pointer)) + } + + fn get_id_by_module_path(&mut self, path: &[Symbol]) -> usize { + let ty = if path.len() == 1 { ItemType::ExternCrate } else { ItemType::Module }; + match self.crate_paths_index.entry((ty, path.to_vec())) { + Entry::Occupied(index) => *index.get(), + Entry::Vacant(slot) => { + slot.insert(self.path_data.len()); + let (name, module_path) = path.split_last().unwrap(); + self.push_path( + name.as_str().to_string(), + PathData { ty, module_path: module_path.to_vec(), exact_module_path: None }, + ) + } + } + } + + pub(crate) fn union(mut self, other: &SerializedSearchIndex) -> SerializedSearchIndex { + let other_entryid_offset = self.names.len(); + let mut map_other_pathid_to_self_pathid: Vec<usize> = Vec::new(); + let mut skips = FxHashSet::default(); + for (other_pathid, other_path_data) in other.path_data.iter().enumerate() { + if let Some(other_path_data) = other_path_data { + let mut fqp = other_path_data.module_path.clone(); + let name = Symbol::intern(&other.names[other_pathid]); + fqp.push(name); + let self_pathid = other_entryid_offset + other_pathid; + let self_pathid = match self.crate_paths_index.entry((other_path_data.ty, fqp)) { + Entry::Vacant(slot) => { + slot.insert(self_pathid); + self_pathid + } + Entry::Occupied(existing_entryid) => { + skips.insert(other_pathid); + let self_pathid = *existing_entryid.get(); + let new_type_data = match ( + self.type_data[self_pathid].take(), + other.type_data[other_pathid].as_ref(), + ) { + (Some(self_type_data), None) => Some(self_type_data), + (None, Some(other_type_data)) => Some(TypeData { + search_unbox: other_type_data.search_unbox, + inverted_function_signature_index: other_type_data + .inverted_function_signature_index + .iter() + .cloned() + .map(|mut list: Vec<u32>| { + for fnid in &mut list { + assert!( + other.function_data + [usize::try_from(*fnid).unwrap()] + .is_some(), + ); + // this is valid because we call `self.push()` once, exactly, for every entry, + // even if we're just pushing a tombstone + *fnid += u32::try_from(other_entryid_offset).unwrap(); + } + list + }) + .collect(), + }), + (Some(mut self_type_data), Some(other_type_data)) => { + for (size, other_list) in other_type_data + .inverted_function_signature_index + .iter() + .enumerate() + { + while self_type_data.inverted_function_signature_index.len() + <= size + { + self_type_data + .inverted_function_signature_index + .push(Vec::new()); + } + self_type_data.inverted_function_signature_index[size].extend( + other_list.iter().copied().map(|fnid| { + assert!( + other.function_data[usize::try_from(fnid).unwrap()] + .is_some(), + ); + // this is valid because we call `self.push()` once, exactly, for every entry, + // even if we're just pushing a tombstone + fnid + u32::try_from(other_entryid_offset).unwrap() + }), + ) + } + Some(self_type_data) + } + (None, None) => None, + }; + self.type_data[self_pathid] = new_type_data; + self_pathid + } + }; + map_other_pathid_to_self_pathid.push(self_pathid); + } else { + // if this gets used, we want it to crash + // this should be impossible as a valid index, since some of the + // memory must be used for stuff other than the list + map_other_pathid_to_self_pathid.push(!0); + } + } + for other_entryid in 0..other.names.len() { + if skips.contains(&other_entryid) { + // we push tombstone entries to keep the IDs lined up + self.push(String::new(), None, None, String::new(), None, None, None); + } else { + self.push( + other.names[other_entryid].clone(), + other.path_data[other_entryid].clone(), + other.entry_data[other_entryid].as_ref().map(|other_entry_data| EntryData { + parent: other_entry_data + .parent + .map(|parent| map_other_pathid_to_self_pathid[parent]) + .clone(), + module_path: other_entry_data + .module_path + .map(|path| map_other_pathid_to_self_pathid[path]) + .clone(), + exact_module_path: other_entry_data + .exact_module_path + .map(|exact_path| map_other_pathid_to_self_pathid[exact_path]) + .clone(), + krate: map_other_pathid_to_self_pathid[other_entry_data.krate], + ..other_entry_data.clone() + }), + other.descs[other_entryid].clone(), + other.function_data[other_entryid].as_ref().map(|function_data| FunctionData { + function_signature: { + let (mut func, _offset) = + IndexItemFunctionType::read_from_string_without_param_names( + function_data.function_signature.as_bytes(), + ); + fn map_fn_sig_item( + map_other_pathid_to_self_pathid: &mut Vec<usize>, + ty: &mut RenderType, + ) { + match ty.id { + None => {} + Some(RenderTypeId::Index(generic)) if generic < 0 => {} + Some(RenderTypeId::Index(id)) => { + let id = usize::try_from(id).unwrap(); + let id = map_other_pathid_to_self_pathid[id]; + assert!(id != !0); + ty.id = + Some(RenderTypeId::Index(isize::try_from(id).unwrap())); + } + _ => unreachable!(), + } + if let Some(generics) = &mut ty.generics { + for generic in generics { + map_fn_sig_item(map_other_pathid_to_self_pathid, generic); + } + } + if let Some(bindings) = &mut ty.bindings { + for (param, constraints) in bindings { + *param = match *param { + param @ RenderTypeId::Index(generic) if generic < 0 => { + param + } + RenderTypeId::Index(id) => { + let id = usize::try_from(id).unwrap(); + let id = map_other_pathid_to_self_pathid[id]; + assert!(id != !0); + RenderTypeId::Index(isize::try_from(id).unwrap()) + } + _ => unreachable!(), + }; + for constraint in constraints { + map_fn_sig_item( + map_other_pathid_to_self_pathid, + constraint, + ); + } + } + } + } + for input in &mut func.inputs { + map_fn_sig_item(&mut map_other_pathid_to_self_pathid, input); + } + for output in &mut func.output { + map_fn_sig_item(&mut map_other_pathid_to_self_pathid, output); + } + for clause in &mut func.where_clause { + for entry in clause { + map_fn_sig_item(&mut map_other_pathid_to_self_pathid, entry); + } + } + let mut result = + String::with_capacity(function_data.function_signature.len()); + func.write_to_string_without_param_names(&mut result); + result + }, + param_names: function_data.param_names.clone(), + }), + other.type_data[other_entryid].as_ref().map(|type_data| TypeData { + inverted_function_signature_index: type_data + .inverted_function_signature_index + .iter() + .cloned() + .map(|mut list| { + for fnid in &mut list { + assert!( + other.function_data[usize::try_from(*fnid).unwrap()] + .is_some(), + ); + // this is valid because we call `self.push()` once, exactly, for every entry, + // even if we're just pushing a tombstone + *fnid += u32::try_from(other_entryid_offset).unwrap(); + } + list + }) + .collect(), + search_unbox: type_data.search_unbox, + }), + other.alias_pointers[other_entryid] + .map(|alias_pointer| alias_pointer + other_entryid_offset), + ); + } + } + for (i, other_generic_inverted_index) in other.generic_inverted_index.iter().enumerate() { + for (size, other_list) in other_generic_inverted_index.iter().enumerate() { + let self_generic_inverted_index = match self.generic_inverted_index.get_mut(i) { + Some(self_generic_inverted_index) => self_generic_inverted_index, + None => { + self.generic_inverted_index.push(Vec::new()); + self.generic_inverted_index.last_mut().unwrap() + } + }; + while self_generic_inverted_index.len() <= size { + self_generic_inverted_index.push(Vec::new()); + } + self_generic_inverted_index[size].extend( + other_list + .iter() + .copied() + .map(|fnid| fnid + u32::try_from(other_entryid_offset).unwrap()), + ); + } + } + self + } + + pub(crate) fn sort(self) -> SerializedSearchIndex { + let mut idlist: Vec<usize> = (0..self.names.len()).collect(); + // nameless entries are tombstones, and will be removed after sorting + // sort shorter names first, so that we can present them in order out of search.js + idlist.sort_by_key(|&id| { + ( + self.names[id].is_empty(), + self.names[id].len(), + &self.names[id], + self.entry_data[id].as_ref().map_or("", |entry| self.names[entry.krate].as_str()), + self.path_data[id].as_ref().map_or(&[][..], |entry| &entry.module_path[..]), + ) + }); + let map = FxHashMap::from_iter( + idlist.iter().enumerate().map(|(new_id, &old_id)| (old_id, new_id)), + ); + let mut new = SerializedSearchIndex::default(); + for &id in &idlist { + if self.names[id].is_empty() { + break; + } + new.push( + self.names[id].clone(), + self.path_data[id].clone(), + self.entry_data[id].as_ref().map( + |EntryData { + krate, + ty, + module_path, + exact_module_path, + parent, + deprecated, + associated_item_disambiguator, + }| EntryData { + krate: *map.get(krate).unwrap(), + ty: *ty, + module_path: module_path.and_then(|path_id| map.get(&path_id).copied()), + exact_module_path: exact_module_path + .and_then(|path_id| map.get(&path_id).copied()), + parent: parent.and_then(|path_id| map.get(&path_id).copied()), + deprecated: *deprecated, + associated_item_disambiguator: associated_item_disambiguator.clone(), + }, + ), + self.descs[id].clone(), + self.function_data[id].as_ref().map( + |FunctionData { function_signature, param_names }| FunctionData { + function_signature: { + let (mut func, _offset) = + IndexItemFunctionType::read_from_string_without_param_names( + function_signature.as_bytes(), + ); + fn map_fn_sig_item(map: &FxHashMap<usize, usize>, ty: &mut RenderType) { + match ty.id { + None => {} + Some(RenderTypeId::Index(generic)) if generic < 0 => {} + Some(RenderTypeId::Index(id)) => { + let id = usize::try_from(id).unwrap(); + let id = *map.get(&id).unwrap(); + assert!(id != !0); + ty.id = + Some(RenderTypeId::Index(isize::try_from(id).unwrap())); + } + _ => unreachable!(), + } + if let Some(generics) = &mut ty.generics { + for generic in generics { + map_fn_sig_item(map, generic); + } + } + if let Some(bindings) = &mut ty.bindings { + for (param, constraints) in bindings { + *param = match *param { + param @ RenderTypeId::Index(generic) if generic < 0 => { + param + } + RenderTypeId::Index(id) => { + let id = usize::try_from(id).unwrap(); + let id = *map.get(&id).unwrap(); + assert!(id != !0); + RenderTypeId::Index(isize::try_from(id).unwrap()) + } + _ => unreachable!(), + }; + for constraint in constraints { + map_fn_sig_item(map, constraint); + } + } + } + } + for input in &mut func.inputs { + map_fn_sig_item(&map, input); + } + for output in &mut func.output { + map_fn_sig_item(&map, output); + } + for clause in &mut func.where_clause { + for entry in clause { + map_fn_sig_item(&map, entry); + } + } + let mut result = String::with_capacity(function_signature.len()); + func.write_to_string_without_param_names(&mut result); + result + }, + param_names: param_names.clone(), + }, + ), + self.type_data[id].as_ref().map( + |TypeData { search_unbox, inverted_function_signature_index }| { + let inverted_function_signature_index: Vec<Vec<u32>> = + inverted_function_signature_index + .iter() + .cloned() + .map(|mut list| { + for id in &mut list { + *id = u32::try_from( + *map.get(&usize::try_from(*id).unwrap()).unwrap(), + ) + .unwrap(); + } + list.sort(); + list + }) + .collect(); + TypeData { search_unbox: *search_unbox, inverted_function_signature_index } + }, + ), + self.alias_pointers[id].and_then(|alias| map.get(&alias).copied()), + ); + } + new.generic_inverted_index = self + .generic_inverted_index + .into_iter() + .map(|mut postings| { + for list in postings.iter_mut() { + let mut new_list: Vec<u32> = list + .iter() + .copied() + .filter_map(|id| u32::try_from(*map.get(&usize::try_from(id).ok()?)?).ok()) + .collect(); + new_list.sort(); + *list = new_list; + } + postings + }) + .collect(); + new + } + + pub(crate) fn write_to(self, doc_root: &Path, resource_suffix: &str) -> Result<(), Error> { + let SerializedSearchIndex { + names, + path_data, + entry_data, + descs, + function_data, + type_data, + alias_pointers, + generic_inverted_index, + crate_paths_index: _, + } = self; + let mut serialized_root = Vec::new(); + serialized_root.extend_from_slice(br#"rr_('{"normalizedName":{"I":""#); + let normalized_names = names + .iter() + .map(|name| { + if name.contains("_") { + name.replace("_", "").to_ascii_lowercase() + } else { + name.to_ascii_lowercase() + } + }) + .collect::<Vec<String>>(); + let names_search_tree = stringdex_internals::tree::encode_search_tree_ukkonen( + normalized_names.iter().map(|name| name.as_bytes()), + ); + let dir_path = doc_root.join(format!("search.index/")); + let _ = std::fs::remove_dir_all(&dir_path); // if already missing, no problem + stringdex_internals::write_tree_to_disk( + &names_search_tree, + &dir_path, + &mut serialized_root, + ) + .map_err(|error| Error { + file: dir_path, + error: format!("failed to write name tree to disk: {error}"), + })?; + std::mem::drop(names_search_tree); + serialized_root.extend_from_slice(br#"","#); + serialized_root.extend_from_slice(&perform_write_strings( + doc_root, + "normalizedName", + normalized_names.into_iter(), + )?); + serialized_root.extend_from_slice(br#"},"crateNames":{"#); + let mut crates: Vec<&[u8]> = entry_data + .iter() + .filter_map(|entry_data| Some(names[entry_data.as_ref()?.krate].as_bytes())) + .collect(); + crates.sort(); + crates.dedup(); + serialized_root.extend_from_slice(&perform_write_strings( + doc_root, + "crateNames", + crates.into_iter(), + )?); + serialized_root.extend_from_slice(br#"},"name":{"#); + serialized_root.extend_from_slice(&perform_write_strings(doc_root, "name", names.iter())?); + serialized_root.extend_from_slice(br#"},"path":{"#); + serialized_root.extend_from_slice(&perform_write_serde(doc_root, "path", path_data)?); + serialized_root.extend_from_slice(br#"},"entry":{"#); + serialized_root.extend_from_slice(&perform_write_serde(doc_root, "entry", entry_data)?); + serialized_root.extend_from_slice(br#"},"desc":{"#); + serialized_root.extend_from_slice(&perform_write_strings( + doc_root, + "desc", + descs.into_iter(), + )?); + serialized_root.extend_from_slice(br#"},"function":{"#); + serialized_root.extend_from_slice(&perform_write_serde( + doc_root, + "function", + function_data, + )?); + serialized_root.extend_from_slice(br#"},"type":{"#); + serialized_root.extend_from_slice(&perform_write_serde(doc_root, "type", type_data)?); + serialized_root.extend_from_slice(br#"},"alias":{"#); + serialized_root.extend_from_slice(&perform_write_serde(doc_root, "alias", alias_pointers)?); + serialized_root.extend_from_slice(br#"},"generic_inverted_index":{"#); + serialized_root.extend_from_slice(&perform_write_postings( + doc_root, + "generic_inverted_index", + generic_inverted_index, + )?); + serialized_root.extend_from_slice(br#"}}')"#); + fn perform_write_strings( + doc_root: &Path, + dirname: &str, + mut column: impl Iterator<Item = impl AsRef<[u8]> + Clone> + ExactSizeIterator, + ) -> Result<Vec<u8>, Error> { + let dir_path = doc_root.join(format!("search.index/{dirname}")); + stringdex_internals::write_data_to_disk(&mut column, &dir_path).map_err(|error| Error { + file: dir_path, + error: format!("failed to write column to disk: {error}"), + }) + } + fn perform_write_serde( + doc_root: &Path, + dirname: &str, + column: Vec<Option<impl Serialize>>, + ) -> Result<Vec<u8>, Error> { + perform_write_strings( + doc_root, + dirname, + column.into_iter().map(|value| { + if let Some(value) = value { + serde_json::to_vec(&value).unwrap() + } else { + Vec::new() + } + }), + ) + } + fn perform_write_postings( + doc_root: &Path, + dirname: &str, + column: Vec<Vec<Vec<u32>>>, + ) -> Result<Vec<u8>, Error> { + perform_write_strings( + doc_root, + dirname, + column.into_iter().map(|postings| { + let mut buf = Vec::new(); + encode::write_postings_to_string(&postings, &mut buf); + buf + }), + ) + } + std::fs::write( + doc_root.join(format!("search.index/root{resource_suffix}.js")), + serialized_root, + ) + .map_err(|error| Error { + file: doc_root.join(format!("search.index/root{resource_suffix}.js")), + error: format!("failed to write root to disk: {error}"), + })?; + Ok(()) + } +} + +#[derive(Clone, Debug)] +struct EntryData { + krate: usize, + ty: ItemType, + module_path: Option<usize>, + exact_module_path: Option<usize>, + parent: Option<usize>, + deprecated: bool, + associated_item_disambiguator: Option<String>, +} + +impl Serialize for EntryData { + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + let mut seq = serializer.serialize_seq(None)?; + seq.serialize_element(&self.krate)?; + seq.serialize_element(&self.ty)?; + seq.serialize_element(&self.module_path.map(|id| id + 1).unwrap_or(0))?; + seq.serialize_element(&self.exact_module_path.map(|id| id + 1).unwrap_or(0))?; + seq.serialize_element(&self.parent.map(|id| id + 1).unwrap_or(0))?; + seq.serialize_element(&if self.deprecated { 1 } else { 0 })?; + if let Some(disambig) = &self.associated_item_disambiguator { + seq.serialize_element(&disambig)?; + } + seq.end() + } +} + +impl<'de> Deserialize<'de> for EntryData { + fn deserialize<D>(deserializer: D) -> Result<EntryData, D::Error> + where + D: Deserializer<'de>, + { + struct EntryDataVisitor; + impl<'de> de::Visitor<'de> for EntryDataVisitor { + type Value = EntryData; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "path data") + } + fn visit_seq<A: de::SeqAccess<'de>>(self, mut v: A) -> Result<EntryData, A::Error> { + let krate: usize = + v.next_element()?.ok_or_else(|| A::Error::missing_field("krate"))?; + let ty: ItemType = + v.next_element()?.ok_or_else(|| A::Error::missing_field("ty"))?; + let module_path: SerializedOptional32 = + v.next_element()?.ok_or_else(|| A::Error::missing_field("module_path"))?; + let exact_module_path: SerializedOptional32 = v + .next_element()? + .ok_or_else(|| A::Error::missing_field("exact_module_path"))?; + let parent: SerializedOptional32 = + v.next_element()?.ok_or_else(|| A::Error::missing_field("parent"))?; + let deprecated: u32 = v.next_element()?.unwrap_or(0); + let associated_item_disambiguator: Option<String> = v.next_element()?; + Ok(EntryData { + krate, + ty, + module_path: Option::<i32>::from(module_path).map(|path| path as usize), + exact_module_path: Option::<i32>::from(exact_module_path) + .map(|path| path as usize), + parent: Option::<i32>::from(parent).map(|path| path as usize), + deprecated: deprecated != 0, + associated_item_disambiguator, + }) + } + } + deserializer.deserialize_any(EntryDataVisitor) + } +} + +#[derive(Clone, Debug)] +struct PathData { + ty: ItemType, + module_path: Vec<Symbol>, + exact_module_path: Option<Vec<Symbol>>, } -const DESC_INDEX_SHARD_LEN: usize = 128 * 1024; +impl Serialize for PathData { + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + let mut seq = serializer.serialize_seq(None)?; + seq.serialize_element(&self.ty)?; + seq.serialize_element(&if self.module_path.is_empty() { + String::new() + } else { + join_path_syms(&self.module_path) + })?; + if let Some(ref path) = self.exact_module_path { + seq.serialize_element(&if path.is_empty() { + String::new() + } else { + join_path_syms(path) + })?; + } + seq.end() + } +} + +impl<'de> Deserialize<'de> for PathData { + fn deserialize<D>(deserializer: D) -> Result<PathData, D::Error> + where + D: Deserializer<'de>, + { + struct PathDataVisitor; + impl<'de> de::Visitor<'de> for PathDataVisitor { + type Value = PathData; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "path data") + } + fn visit_seq<A: de::SeqAccess<'de>>(self, mut v: A) -> Result<PathData, A::Error> { + let ty: ItemType = + v.next_element()?.ok_or_else(|| A::Error::missing_field("ty"))?; + let module_path: String = + v.next_element()?.ok_or_else(|| A::Error::missing_field("module_path"))?; + let exact_module_path: Option<String> = + v.next_element()?.and_then(SerializedOptionalString::into); + Ok(PathData { + ty, + module_path: if module_path.is_empty() { + vec![] + } else { + module_path.split("::").map(Symbol::intern).collect() + }, + exact_module_path: exact_module_path.map(|path| { + if path.is_empty() { + vec![] + } else { + path.split("::").map(Symbol::intern).collect() + } + }), + }) + } + } + deserializer.deserialize_any(PathDataVisitor) + } +} + +#[derive(Clone, Debug)] +struct TypeData { + /// If set to "true", the generics can be matched without having to + /// mention the type itself. The truth table, assuming `Unboxable` + /// has `search_unbox = true` and `Inner` has `search_unbox = false` + /// + /// | **query** | `Unboxable<Inner>` | `Inner` | `Inner<Unboxable>` | + /// |--------------------|--------------------|---------|--------------------| + /// | `Inner` | yes | yes | yes | + /// | `Unboxable` | yes | no | no | + /// | `Unboxable<Inner>` | yes | no | no | + /// | `Inner<Unboxable>` | no | no | yes | + search_unbox: bool, + /// List of functions that mention this type in their type signature. + /// + /// - The outermost list has one entry per alpha-normalized generic. + /// + /// - The second layer is sorted by number of types that appear in the + /// type signature. The search engine iterates over these in order from + /// smallest to largest. Functions with less stuff in their type + /// signature are more likely to be what the user wants, because we never + /// show functions that are *missing* parts of the query, so removing.. + /// + /// - The final layer is the list of functions. + inverted_function_signature_index: Vec<Vec<u32>>, +} + +impl Serialize for TypeData { + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + if self.search_unbox || !self.inverted_function_signature_index.is_empty() { + let mut seq = serializer.serialize_seq(None)?; + if !self.inverted_function_signature_index.is_empty() { + let mut buf = Vec::new(); + encode::write_postings_to_string(&self.inverted_function_signature_index, &mut buf); + let mut serialized_result = Vec::new(); + stringdex_internals::encode::write_base64_to_bytes(&buf, &mut serialized_result); + seq.serialize_element(&String::from_utf8(serialized_result).unwrap())?; + } + if self.search_unbox { + seq.serialize_element(&1)?; + } + seq.end() + } else { + None::<()>.serialize(serializer) + } + } +} + +impl<'de> Deserialize<'de> for TypeData { + fn deserialize<D>(deserializer: D) -> Result<TypeData, D::Error> + where + D: Deserializer<'de>, + { + struct TypeDataVisitor; + impl<'de> de::Visitor<'de> for TypeDataVisitor { + type Value = TypeData; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "type data") + } + fn visit_none<E>(self) -> Result<TypeData, E> { + Ok(TypeData { inverted_function_signature_index: vec![], search_unbox: false }) + } + fn visit_seq<A: de::SeqAccess<'de>>(self, mut v: A) -> Result<TypeData, A::Error> { + let inverted_function_signature_index: String = + v.next_element()?.unwrap_or(String::new()); + let search_unbox: u32 = v.next_element()?.unwrap_or(0); + let mut idx: Vec<u8> = Vec::new(); + stringdex_internals::decode::read_base64_from_bytes( + inverted_function_signature_index.as_bytes(), + &mut idx, + ) + .unwrap(); + let mut inverted_function_signature_index = Vec::new(); + encode::read_postings_from_string(&mut inverted_function_signature_index, &idx); + Ok(TypeData { inverted_function_signature_index, search_unbox: search_unbox == 1 }) + } + } + deserializer.deserialize_any(TypeDataVisitor) + } +} + +enum SerializedOptionalString { + None, + Some(String), +} + +impl From<SerializedOptionalString> for Option<String> { + fn from(me: SerializedOptionalString) -> Option<String> { + match me { + SerializedOptionalString::Some(string) => Some(string), + SerializedOptionalString::None => None, + } + } +} + +impl Serialize for SerializedOptionalString { + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + match self { + SerializedOptionalString::Some(string) => string.serialize(serializer), + SerializedOptionalString::None => 0.serialize(serializer), + } + } +} +impl<'de> Deserialize<'de> for SerializedOptionalString { + fn deserialize<D>(deserializer: D) -> Result<SerializedOptionalString, D::Error> + where + D: Deserializer<'de>, + { + struct SerializedOptionalStringVisitor; + impl<'de> de::Visitor<'de> for SerializedOptionalStringVisitor { + type Value = SerializedOptionalString; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "0 or string") + } + fn visit_u64<E: de::Error>(self, v: u64) -> Result<SerializedOptionalString, E> { + if v != 0 { + return Err(E::missing_field("not 0")); + } + Ok(SerializedOptionalString::None) + } + fn visit_string<E: de::Error>(self, v: String) -> Result<SerializedOptionalString, E> { + Ok(SerializedOptionalString::Some(v)) + } + fn visit_str<E: de::Error>(self, v: &str) -> Result<SerializedOptionalString, E> { + Ok(SerializedOptionalString::Some(v.to_string())) + } + } + deserializer.deserialize_any(SerializedOptionalStringVisitor) + } +} + +enum SerializedOptional32 { + None, + Some(i32), +} + +impl From<SerializedOptional32> for Option<i32> { + fn from(me: SerializedOptional32) -> Option<i32> { + match me { + SerializedOptional32::Some(number) => Some(number), + SerializedOptional32::None => None, + } + } +} + +impl Serialize for SerializedOptional32 { + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + match self { + &SerializedOptional32::Some(number) if number < 0 => number.serialize(serializer), + &SerializedOptional32::Some(number) => (number + 1).serialize(serializer), + &SerializedOptional32::None => 0.serialize(serializer), + } + } +} +impl<'de> Deserialize<'de> for SerializedOptional32 { + fn deserialize<D>(deserializer: D) -> Result<SerializedOptional32, D::Error> + where + D: Deserializer<'de>, + { + struct SerializedOptional32Visitor; + impl<'de> de::Visitor<'de> for SerializedOptional32Visitor { + type Value = SerializedOptional32; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "integer") + } + fn visit_i64<E: de::Error>(self, v: i64) -> Result<SerializedOptional32, E> { + Ok(match v { + 0 => SerializedOptional32::None, + v if v < 0 => SerializedOptional32::Some(v as i32), + v => SerializedOptional32::Some(v as i32 - 1), + }) + } + fn visit_u64<E: de::Error>(self, v: u64) -> Result<SerializedOptional32, E> { + Ok(match v { + 0 => SerializedOptional32::None, + v => SerializedOptional32::Some(v as i32 - 1), + }) + } + } + deserializer.deserialize_any(SerializedOptional32Visitor) + } +} + +#[derive(Clone, Debug)] +pub struct FunctionData { + function_signature: String, + param_names: Vec<String>, +} + +impl Serialize for FunctionData { + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + let mut seq = serializer.serialize_seq(None)?; + seq.serialize_element(&self.function_signature)?; + seq.serialize_element(&self.param_names)?; + seq.end() + } +} + +impl<'de> Deserialize<'de> for FunctionData { + fn deserialize<D>(deserializer: D) -> Result<FunctionData, D::Error> + where + D: Deserializer<'de>, + { + struct FunctionDataVisitor; + impl<'de> de::Visitor<'de> for FunctionDataVisitor { + type Value = FunctionData; + fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(formatter, "fn data") + } + fn visit_seq<A: de::SeqAccess<'de>>(self, mut v: A) -> Result<FunctionData, A::Error> { + let function_signature: String = v + .next_element()? + .ok_or_else(|| A::Error::missing_field("function_signature"))?; + let param_names: Vec<String> = + v.next_element()?.ok_or_else(|| A::Error::missing_field("param_names"))?; + Ok(FunctionData { function_signature, param_names }) + } + } + deserializer.deserialize_any(FunctionDataVisitor) + } +} /// Builds the search index from the collected metadata pub(crate) fn build_index( krate: &clean::Crate, cache: &mut Cache, tcx: TyCtxt<'_>, -) -> SerializedSearchIndex { - // Maps from ID to position in the `crate_paths` array. - let mut itemid_to_pathid = FxHashMap::default(); - let mut primitives = FxHashMap::default(); - let mut associated_types = FxHashMap::default(); - - // item type, display path, re-exported internal path - let mut crate_paths: Vec<(ItemType, Vec<Symbol>, Option<Vec<Symbol>>, bool)> = vec![]; + doc_root: &Path, + resource_suffix: &str, +) -> Result<SerializedSearchIndex, Error> { + let mut search_index = std::mem::take(&mut cache.search_index); // Attach all orphan items to the type's definition if the type // has since been learned. @@ -74,15 +1171,15 @@ pub(crate) fn build_index( { if let Some((fqp, _)) = cache.paths.get(&parent) { let desc = short_markdown_summary(&item.doc_value(), &item.link_names(cache)); - cache.search_index.push(IndexItem { + search_index.push(IndexItem { ty: item.type_(), defid: item.item_id.as_def_id(), name: item.name.unwrap(), - path: join_path_syms(&fqp[..fqp.len() - 1]), + module_path: fqp[..fqp.len() - 1].to_vec(), desc, parent: Some(parent), parent_idx: None, - exact_path: None, + exact_module_path: None, impl_id, search_type: get_function_type_for_search( item, @@ -97,85 +1194,299 @@ pub(crate) fn build_index( } } + // Sort search index items. This improves the compressibility of the search index. + search_index.sort_unstable_by(|k1, k2| { + // `sort_unstable_by_key` produces lifetime errors + // HACK(rustdoc): should not be sorting `CrateNum` or `DefIndex`, this will soon go away, too + let k1 = + (&k1.module_path, k1.name.as_str(), &k1.ty, k1.parent.map(|id| (id.index, id.krate))); + let k2 = + (&k2.module_path, k2.name.as_str(), &k2.ty, k2.parent.map(|id| (id.index, id.krate))); + Ord::cmp(&k1, &k2) + }); + + // Now, convert to an on-disk search index format + // + // if there's already a search index, load it into memory and add the new entries to it + // otherwise, do nothing + let mut serialized_index = SerializedSearchIndex::load(doc_root, resource_suffix)?; + + // The crate always goes first in this list + let crate_name = krate.name(tcx); let crate_doc = short_markdown_summary(&krate.module.doc_value(), &krate.module.link_names(cache)); + let crate_idx = { + let crate_path = (ItemType::ExternCrate, vec![crate_name]); + match serialized_index.crate_paths_index.entry(crate_path) { + Entry::Occupied(index) => { + let index = *index.get(); + serialized_index.descs[index] = crate_doc; + for type_data in serialized_index.type_data.iter_mut() { + if let Some(TypeData { inverted_function_signature_index, .. }) = type_data { + for list in &mut inverted_function_signature_index[..] { + list.retain(|fnid| { + serialized_index.entry_data[usize::try_from(*fnid).unwrap()] + .as_ref() + .unwrap() + .krate + != index + }); + } + } + } + for i in (index + 1)..serialized_index.entry_data.len() { + // if this crate has been built before, replace its stuff with new + if let Some(EntryData { krate, .. }) = serialized_index.entry_data[i] + && krate == index + { + serialized_index.entry_data[i] = None; + serialized_index.descs[i] = String::new(); + serialized_index.function_data[i] = None; + if serialized_index.path_data[i].is_none() { + serialized_index.names[i] = String::new(); + } + } + if let Some(alias_pointer) = serialized_index.alias_pointers[i] + && serialized_index.entry_data[alias_pointer].is_none() + { + serialized_index.alias_pointers[i] = None; + if serialized_index.path_data[i].is_none() + && serialized_index.entry_data[i].is_none() + { + serialized_index.names[i] = String::new(); + } + } + } + index + } + Entry::Vacant(slot) => { + let krate = serialized_index.names.len(); + slot.insert(krate); + serialized_index.push( + crate_name.as_str().to_string(), + Some(PathData { + ty: ItemType::ExternCrate, + module_path: vec![], + exact_module_path: None, + }), + Some(EntryData { + krate, + ty: ItemType::ExternCrate, + module_path: None, + exact_module_path: None, + parent: None, + deprecated: false, + associated_item_disambiguator: None, + }), + crate_doc, + None, + None, + None, + ); + krate + } + } + }; + + // First, populate associated item parents + let crate_items: Vec<&mut IndexItem> = search_index + .iter_mut() + .map(|item| { + item.parent_idx = item.parent.and_then(|defid| { + cache.paths.get(&defid).map(|&(ref fqp, ty)| { + let pathid = serialized_index.names.len(); + match serialized_index.crate_paths_index.entry((ty, fqp.clone())) { + Entry::Occupied(entry) => *entry.get(), + Entry::Vacant(entry) => { + entry.insert(pathid); + let (name, path) = fqp.split_last().unwrap(); + serialized_index.push_path( + name.as_str().to_string(), + PathData { + ty, + module_path: path.to_vec(), + exact_module_path: if let Some(exact_path) = + cache.exact_paths.get(&defid) + && let Some((name2, exact_path)) = exact_path.split_last() + && name == name2 + { + Some(exact_path.to_vec()) + } else { + None + }, + }, + ); + usize::try_from(pathid).unwrap() + } + } + }) + }); + + if let Some(defid) = item.defid + && item.parent_idx.is_none() + { + // If this is a re-export, retain the original path. + // Associated items don't use this. + // Their parent carries the exact fqp instead. + let exact_fqp = cache + .exact_paths + .get(&defid) + .or_else(|| cache.external_paths.get(&defid).map(|(fqp, _)| fqp)); + item.exact_module_path = exact_fqp.and_then(|fqp| { + // Re-exports only count if the name is exactly the same. + // This is a size optimization, since it means we only need + // to store the name once (and the path is re-used for everything + // exported from this same module). It's also likely to Do + // What I Mean, since if a re-export changes the name, it might + // also be a change in semantic meaning. + if fqp.last() != Some(&item.name) { + return None; + } + let path = + if item.ty == ItemType::Macro && tcx.has_attr(defid, sym::macro_export) { + // `#[macro_export]` always exports to the crate root. + vec![tcx.crate_name(defid.krate)] + } else { + if fqp.len() < 2 { + return None; + } + fqp[..fqp.len() - 1].to_vec() + }; + if path == item.module_path { + return None; + } + Some(path) + }); + } else if let Some(parent_idx) = item.parent_idx { + let i = usize::try_from(parent_idx).unwrap(); + item.module_path = + serialized_index.path_data[i].as_ref().unwrap().module_path.clone(); + item.exact_module_path = + serialized_index.path_data[i].as_ref().unwrap().exact_module_path.clone(); + } - #[derive(Eq, Ord, PartialEq, PartialOrd)] - struct SerSymbolAsStr(Symbol); + &mut *item + }) + .collect(); - impl Serialize for SerSymbolAsStr { - fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> - where - S: Serializer, + // Now, find anywhere that the same name is used for two different items + // these need a disambiguator hash for lints + let mut associated_item_duplicates = FxHashMap::<(usize, ItemType, Symbol), usize>::default(); + for item in crate_items.iter().map(|x| &*x) { + if item.impl_id.is_some() + && let Some(parent_idx) = item.parent_idx { - self.0.as_str().serialize(serializer) + let count = + associated_item_duplicates.entry((parent_idx, item.ty, item.name)).or_insert(0); + *count += 1; } } - type AliasMap = BTreeMap<SerSymbolAsStr, Vec<usize>>; - // Aliases added through `#[doc(alias = "...")]`. Since a few items can have the same alias, - // we need the alias element to have an array of items. - let mut aliases: AliasMap = BTreeMap::new(); + // now populate the actual entries, type data, and function data + for item in crate_items { + assert_eq!( + item.parent.is_some(), + item.parent_idx.is_some(), + "`{}` is missing idx", + item.name + ); - // Sort search index items. This improves the compressibility of the search index. - cache.search_index.sort_unstable_by(|k1, k2| { - // `sort_unstable_by_key` produces lifetime errors - // HACK(rustdoc): should not be sorting `CrateNum` or `DefIndex`, this will soon go away, too - let k1 = (&k1.path, k1.name.as_str(), &k1.ty, k1.parent.map(|id| (id.index, id.krate))); - let k2 = (&k2.path, k2.name.as_str(), &k2.ty, k2.parent.map(|id| (id.index, id.krate))); - Ord::cmp(&k1, &k2) - }); + let module_path = Some(serialized_index.get_id_by_module_path(&item.module_path)); + let exact_module_path = item + .exact_module_path + .as_ref() + .map(|path| serialized_index.get_id_by_module_path(path)); + + let new_entry_id = serialized_index.push( + item.name.as_str().to_string(), + None, + Some(EntryData { + ty: item.ty, + parent: item.parent_idx, + module_path, + exact_module_path, + deprecated: item.deprecation.is_some(), + associated_item_disambiguator: if let Some(impl_id) = item.impl_id + && let Some(parent_idx) = item.parent_idx + && associated_item_duplicates + .get(&(parent_idx, item.ty, item.name)) + .copied() + .unwrap_or(0) + > 1 + { + Some(render::get_id_for_impl(tcx, ItemId::DefId(impl_id))) + } else { + None + }, + krate: crate_idx, + }), + item.desc.to_string(), + None, // filled in after all the types have been indexed + None, + None, + ); - // Set up alias indexes. - for (i, item) in cache.search_index.iter().enumerate() { + // Aliases + // ------- for alias in &item.aliases[..] { - aliases.entry(SerSymbolAsStr(*alias)).or_default().push(i); + serialized_index.push_alias(alias.as_str().to_string(), new_entry_id); } - } - - // Reduce `DefId` in paths into smaller sequential numbers, - // and prune the paths that do not appear in the index. - let mut lastpath = ""; - let mut lastpathid = 0isize; - // First, on function signatures - let mut search_index = std::mem::take(&mut cache.search_index); - for item in search_index.iter_mut() { - fn insert_into_map<F: std::hash::Hash + Eq>( - map: &mut FxHashMap<F, isize>, - itemid: F, - lastpathid: &mut isize, - crate_paths: &mut Vec<(ItemType, Vec<Symbol>, Option<Vec<Symbol>>, bool)>, - item_type: ItemType, + // Function signature reverse index + // -------------------------------- + fn insert_into_map( + ty: ItemType, path: &[Symbol], exact_path: Option<&[Symbol]>, search_unbox: bool, + serialized_index: &mut SerializedSearchIndex, + used_in_function_signature: &mut BTreeSet<isize>, ) -> RenderTypeId { - match map.entry(itemid) { - Entry::Occupied(entry) => RenderTypeId::Index(*entry.get()), + let pathid = serialized_index.names.len(); + let pathid = match serialized_index.crate_paths_index.entry((ty, path.to_vec())) { + Entry::Occupied(entry) => { + let id = *entry.get(); + if serialized_index.type_data[id].as_mut().is_none() { + serialized_index.type_data[id] = Some(TypeData { + search_unbox, + inverted_function_signature_index: Vec::new(), + }); + } else if search_unbox { + serialized_index.type_data[id].as_mut().unwrap().search_unbox = true; + } + id + } Entry::Vacant(entry) => { - let pathid = *lastpathid; entry.insert(pathid); - *lastpathid += 1; - crate_paths.push(( - item_type, - path.to_vec(), - exact_path.map(|path| path.to_vec()), - search_unbox, - )); - RenderTypeId::Index(pathid) + let (name, path) = path.split_last().unwrap(); + serialized_index.push_type( + name.to_string(), + PathData { + ty, + module_path: path.to_vec(), + exact_module_path: if let Some(exact_path) = exact_path + && let Some((name2, exact_path)) = exact_path.split_last() + && name == name2 + { + Some(exact_path.to_vec()) + } else { + None + }, + }, + TypeData { search_unbox, inverted_function_signature_index: Vec::new() }, + ); + pathid } - } + }; + used_in_function_signature.insert(isize::try_from(pathid).unwrap()); + RenderTypeId::Index(isize::try_from(pathid).unwrap()) } fn convert_render_type_id( id: RenderTypeId, cache: &mut Cache, - itemid_to_pathid: &mut FxHashMap<ItemId, isize>, - primitives: &mut FxHashMap<Symbol, isize>, - associated_types: &mut FxHashMap<Symbol, isize>, - lastpathid: &mut isize, - crate_paths: &mut Vec<(ItemType, Vec<Symbol>, Option<Vec<Symbol>>, bool)>, + serialized_index: &mut SerializedSearchIndex, + used_in_function_signature: &mut BTreeSet<isize>, tcx: TyCtxt<'_>, ) -> Option<RenderTypeId> { use crate::clean::PrimitiveType; @@ -192,39 +1503,55 @@ pub(crate) fn build_index( }; match id { RenderTypeId::Mut => Some(insert_into_map( - primitives, - kw::Mut, - lastpathid, - crate_paths, ItemType::Keyword, &[kw::Mut], None, search_unbox, + serialized_index, + used_in_function_signature, )), RenderTypeId::DefId(defid) => { if let Some(&(ref fqp, item_type)) = paths.get(&defid).or_else(|| external_paths.get(&defid)) { - let exact_fqp = exact_paths - .get(&defid) - .or_else(|| external_paths.get(&defid).map(|(fqp, _)| fqp)) - // Re-exports only count if the name is exactly the same. - // This is a size optimization, since it means we only need - // to store the name once (and the path is re-used for everything - // exported from this same module). It's also likely to Do - // What I Mean, since if a re-export changes the name, it might - // also be a change in semantic meaning. - .filter(|this_fqp| this_fqp.last() == fqp.last()); - Some(insert_into_map( - itemid_to_pathid, - ItemId::DefId(defid), - lastpathid, - crate_paths, - item_type, - fqp, - exact_fqp.map(|x| &x[..]).filter(|exact_fqp| exact_fqp != fqp), - search_unbox, - )) + if tcx.lang_items().fn_mut_trait() == Some(defid) + || tcx.lang_items().fn_once_trait() == Some(defid) + || tcx.lang_items().fn_trait() == Some(defid) + { + let name = *fqp.last().unwrap(); + // Make absolutely sure we use this single, correct path, + // because search.js needs to match. If we don't do this, + // there are three different paths that these traits may + // appear to come from. + Some(insert_into_map( + item_type, + &[sym::core, sym::ops, name], + Some(&[sym::core, sym::ops, name]), + search_unbox, + serialized_index, + used_in_function_signature, + )) + } else { + let exact_fqp = exact_paths + .get(&defid) + .or_else(|| external_paths.get(&defid).map(|(fqp, _)| fqp)) + .map(|v| &v[..]) + // Re-exports only count if the name is exactly the same. + // This is a size optimization, since it means we only need + // to store the name once (and the path is re-used for everything + // exported from this same module). It's also likely to Do + // What I Mean, since if a re-export changes the name, it might + // also be a change in semantic meaning. + .filter(|this_fqp| this_fqp.last() == fqp.last()); + Some(insert_into_map( + item_type, + fqp, + exact_fqp, + search_unbox, + serialized_index, + used_in_function_signature, + )) + } } else { None } @@ -232,26 +1559,25 @@ pub(crate) fn build_index( RenderTypeId::Primitive(primitive) => { let sym = primitive.as_sym(); Some(insert_into_map( - primitives, - sym, - lastpathid, - crate_paths, ItemType::Primitive, &[sym], None, search_unbox, + serialized_index, + used_in_function_signature, )) } - RenderTypeId::Index(_) => Some(id), + RenderTypeId::Index(index) => { + used_in_function_signature.insert(index); + Some(id) + } RenderTypeId::AssociatedType(sym) => Some(insert_into_map( - associated_types, - sym, - lastpathid, - crate_paths, ItemType::AssocType, &[sym], None, search_unbox, + serialized_index, + used_in_function_signature, )), } } @@ -259,11 +1585,8 @@ pub(crate) fn build_index( fn convert_render_type( ty: &mut RenderType, cache: &mut Cache, - itemid_to_pathid: &mut FxHashMap<ItemId, isize>, - primitives: &mut FxHashMap<Symbol, isize>, - associated_types: &mut FxHashMap<Symbol, isize>, - lastpathid: &mut isize, - crate_paths: &mut Vec<(ItemType, Vec<Symbol>, Option<Vec<Symbol>>, bool)>, + serialized_index: &mut SerializedSearchIndex, + used_in_function_signature: &mut BTreeSet<isize>, tcx: TyCtxt<'_>, ) { if let Some(generics) = &mut ty.generics { @@ -271,11 +1594,8 @@ pub(crate) fn build_index( convert_render_type( item, cache, - itemid_to_pathid, - primitives, - associated_types, - lastpathid, - crate_paths, + serialized_index, + used_in_function_signature, tcx, ); } @@ -285,11 +1605,8 @@ pub(crate) fn build_index( let converted_associated_type = convert_render_type_id( *associated_type, cache, - itemid_to_pathid, - primitives, - associated_types, - lastpathid, - crate_paths, + serialized_index, + used_in_function_signature, tcx, ); let Some(converted_associated_type) = converted_associated_type else { @@ -300,11 +1617,8 @@ pub(crate) fn build_index( convert_render_type( constraint, cache, - itemid_to_pathid, - primitives, - associated_types, - lastpathid, - crate_paths, + serialized_index, + used_in_function_signature, tcx, ); } @@ -318,24 +1632,74 @@ pub(crate) fn build_index( ty.id = convert_render_type_id( id, cache, - itemid_to_pathid, - primitives, - associated_types, - lastpathid, - crate_paths, + serialized_index, + used_in_function_signature, tcx, ); + use crate::clean::PrimitiveType; + // These cases are added to the inverted index, but not actually included + // in the signature. There's a matching set of cases in the + // `unifyFunctionTypeIsMatchCandidate` function, for the slow path. + match id { + // typeNameIdOfArrayOrSlice + RenderTypeId::Primitive(PrimitiveType::Array | PrimitiveType::Slice) => { + insert_into_map( + ItemType::Primitive, + &[Symbol::intern("[]")], + None, + false, + serialized_index, + used_in_function_signature, + ); + } + RenderTypeId::Primitive(PrimitiveType::Tuple | PrimitiveType::Unit) => { + // typeNameIdOfArrayOrSlice + insert_into_map( + ItemType::Primitive, + &[Symbol::intern("()")], + None, + false, + serialized_index, + used_in_function_signature, + ); + } + // typeNameIdOfHof + RenderTypeId::Primitive(PrimitiveType::Fn) => { + insert_into_map( + ItemType::Primitive, + &[Symbol::intern("->")], + None, + false, + serialized_index, + used_in_function_signature, + ); + } + RenderTypeId::DefId(did) + if tcx.lang_items().fn_mut_trait() == Some(did) + || tcx.lang_items().fn_once_trait() == Some(did) + || tcx.lang_items().fn_trait() == Some(did) => + { + insert_into_map( + ItemType::Primitive, + &[Symbol::intern("->")], + None, + false, + serialized_index, + used_in_function_signature, + ); + } + // not special + _ => {} + } } if let Some(search_type) = &mut item.search_type { + let mut used_in_function_signature = BTreeSet::new(); for item in &mut search_type.inputs { convert_render_type( item, cache, - &mut itemid_to_pathid, - &mut primitives, - &mut associated_types, - &mut lastpathid, - &mut crate_paths, + &mut serialized_index, + &mut used_in_function_signature, tcx, ); } @@ -343,11 +1707,8 @@ pub(crate) fn build_index( convert_render_type( item, cache, - &mut itemid_to_pathid, - &mut primitives, - &mut associated_types, - &mut lastpathid, - &mut crate_paths, + &mut serialized_index, + &mut used_in_function_signature, tcx, ); } @@ -356,464 +1717,56 @@ pub(crate) fn build_index( convert_render_type( trait_, cache, - &mut itemid_to_pathid, - &mut primitives, - &mut associated_types, - &mut lastpathid, - &mut crate_paths, + &mut serialized_index, + &mut used_in_function_signature, tcx, ); } } - } - } - - let Cache { ref paths, ref exact_paths, ref external_paths, .. } = *cache; - - // Then, on parent modules - let crate_items: Vec<&IndexItem> = search_index - .iter_mut() - .map(|item| { - item.parent_idx = - item.parent.and_then(|defid| match itemid_to_pathid.entry(ItemId::DefId(defid)) { - Entry::Occupied(entry) => Some(*entry.get()), - Entry::Vacant(entry) => { - let pathid = lastpathid; - entry.insert(pathid); - lastpathid += 1; - - if let Some(&(ref fqp, short)) = paths.get(&defid) { - let exact_fqp = exact_paths - .get(&defid) - .or_else(|| external_paths.get(&defid).map(|(fqp, _)| fqp)) - .filter(|exact_fqp| { - exact_fqp.last() == Some(&item.name) && *exact_fqp != fqp - }); - crate_paths.push(( - short, - fqp.clone(), - exact_fqp.cloned(), - utils::has_doc_flag(tcx, defid, sym::search_unbox), - )); - Some(pathid) - } else { - None - } - } - }); - - if let Some(defid) = item.defid - && item.parent_idx.is_none() - { - // If this is a re-export, retain the original path. - // Associated items don't use this. - // Their parent carries the exact fqp instead. - let exact_fqp = exact_paths - .get(&defid) - .or_else(|| external_paths.get(&defid).map(|(fqp, _)| fqp)); - item.exact_path = exact_fqp.and_then(|fqp| { - // Re-exports only count if the name is exactly the same. - // This is a size optimization, since it means we only need - // to store the name once (and the path is re-used for everything - // exported from this same module). It's also likely to Do - // What I Mean, since if a re-export changes the name, it might - // also be a change in semantic meaning. - if fqp.last() != Some(&item.name) { - return None; - } - let path = - if item.ty == ItemType::Macro && tcx.has_attr(defid, sym::macro_export) { - // `#[macro_export]` always exports to the crate root. - tcx.crate_name(defid.krate).to_string() - } else { - if fqp.len() < 2 { - return None; - } - join_path_syms(&fqp[..fqp.len() - 1]) - }; - if path == item.path { - return None; - } - Some(path) - }); - } else if let Some(parent_idx) = item.parent_idx { - let i = <isize as TryInto<usize>>::try_into(parent_idx).unwrap(); - item.path = { - let p = &crate_paths[i].1; - join_path_syms(&p[..p.len() - 1]) - }; - item.exact_path = - crate_paths[i].2.as_ref().map(|xp| join_path_syms(&xp[..xp.len() - 1])); - } - - // Omit the parent path if it is same to that of the prior item. - if lastpath == item.path { - item.path.clear(); - } else { - lastpath = &item.path; - } - - &*item - }) - .collect(); - - // Find associated items that need disambiguators - let mut associated_item_duplicates = FxHashMap::<(isize, ItemType, Symbol), usize>::default(); - - for &item in &crate_items { - if item.impl_id.is_some() - && let Some(parent_idx) = item.parent_idx - { - let count = - associated_item_duplicates.entry((parent_idx, item.ty, item.name)).or_insert(0); - *count += 1; - } - } - - let associated_item_disambiguators = crate_items - .iter() - .enumerate() - .filter_map(|(index, item)| { - let impl_id = ItemId::DefId(item.impl_id?); - let parent_idx = item.parent_idx?; - let count = *associated_item_duplicates.get(&(parent_idx, item.ty, item.name))?; - if count > 1 { Some((index, render::get_id_for_impl(tcx, impl_id))) } else { None } - }) - .collect::<Vec<_>>(); - - struct CrateData<'a> { - items: Vec<&'a IndexItem>, - paths: Vec<(ItemType, Vec<Symbol>, Option<Vec<Symbol>>, bool)>, - // The String is alias name and the vec is the list of the elements with this alias. - // - // To be noted: the `usize` elements are indexes to `items`. - aliases: &'a AliasMap, - // Used when a type has more than one impl with an associated item with the same name. - associated_item_disambiguators: &'a Vec<(usize, String)>, - // A list of shard lengths encoded as vlqhex. See the comment in write_vlqhex_to_string - // for information on the format. - desc_index: String, - // A list of items with no description. This is eventually turned into a bitmap. - empty_desc: Vec<u32>, - } - - struct Paths { - ty: ItemType, - name: Symbol, - path: Option<usize>, - exact_path: Option<usize>, - search_unbox: bool, - } - - impl Serialize for Paths { - fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> - where - S: Serializer, - { - let mut seq = serializer.serialize_seq(None)?; - seq.serialize_element(&self.ty)?; - seq.serialize_element(self.name.as_str())?; - if let Some(ref path) = self.path { - seq.serialize_element(path)?; - } - if let Some(ref path) = self.exact_path { - assert!(self.path.is_some()); - seq.serialize_element(path)?; - } - if self.search_unbox { - if self.path.is_none() { - seq.serialize_element(&None::<u8>)?; - } - if self.exact_path.is_none() { - seq.serialize_element(&None::<u8>)?; - } - seq.serialize_element(&1)?; - } - seq.end() - } - } - - impl Serialize for CrateData<'_> { - fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> - where - S: Serializer, - { - let mut extra_paths = FxHashMap::default(); - // We need to keep the order of insertion, hence why we use an `IndexMap`. Then we will - // insert these "extra paths" (which are paths of items from external crates) into the - // `full_paths` list at the end. - let mut revert_extra_paths = FxIndexMap::default(); - let mut mod_paths = FxHashMap::default(); - for (index, item) in self.items.iter().enumerate() { - if item.path.is_empty() { - continue; - } - mod_paths.insert(&item.path, index); - } - let mut paths = Vec::with_capacity(self.paths.len()); - for &(ty, ref path, ref exact, search_unbox) in &self.paths { - if path.len() < 2 { - paths.push(Paths { - ty, - name: path[0], - path: None, - exact_path: None, - search_unbox, - }); - continue; - } - let full_path = join_path_syms(&path[..path.len() - 1]); - let full_exact_path = exact - .as_ref() - .filter(|exact| exact.last() == path.last() && exact.len() >= 2) - .map(|exact| join_path_syms(&exact[..exact.len() - 1])); - let exact_path = extra_paths.len() + self.items.len(); - let exact_path = full_exact_path.as_ref().map(|full_exact_path| match extra_paths - .entry(full_exact_path.clone()) - { - Entry::Occupied(entry) => *entry.get(), - Entry::Vacant(entry) => { - if let Some(index) = mod_paths.get(&full_exact_path) { - return *index; - } - entry.insert(exact_path); - if !revert_extra_paths.contains_key(&exact_path) { - revert_extra_paths.insert(exact_path, full_exact_path.clone()); - } - exact_path - } - }); - if let Some(index) = mod_paths.get(&full_path) { - paths.push(Paths { - ty, - name: *path.last().unwrap(), - path: Some(*index), - exact_path, - search_unbox, - }); - continue; - } - // It means it comes from an external crate so the item and its path will be - // stored into another array. + let search_type_size = search_type.size() + + // Artificially give struct fields a size of 8 instead of their real + // size of 2. This is because search.js sorts them to the end, so + // by pushing them down, we prevent them from blocking real 2-arity functions. // - // `index` is put after the last `mod_paths` - let index = extra_paths.len() + self.items.len(); - match extra_paths.entry(full_path.clone()) { - Entry::Occupied(entry) => { - paths.push(Paths { - ty, - name: *path.last().unwrap(), - path: Some(*entry.get()), - exact_path, - search_unbox, - }); - } - Entry::Vacant(entry) => { - entry.insert(index); - if !revert_extra_paths.contains_key(&index) { - revert_extra_paths.insert(index, full_path); - } - paths.push(Paths { - ty, - name: *path.last().unwrap(), - path: Some(index), - exact_path, - search_unbox, - }); - } - } - } - - // Direct exports use adjacent arrays for the current crate's items, - // but re-exported exact paths don't. - let mut re_exports = Vec::new(); - for (item_index, item) in self.items.iter().enumerate() { - if let Some(exact_path) = item.exact_path.as_ref() { - if let Some(path_index) = mod_paths.get(&exact_path) { - re_exports.push((item_index, *path_index)); - } else { - let path_index = extra_paths.len() + self.items.len(); - let path_index = match extra_paths.entry(exact_path.clone()) { - Entry::Occupied(entry) => *entry.get(), - Entry::Vacant(entry) => { - entry.insert(path_index); - if !revert_extra_paths.contains_key(&path_index) { - revert_extra_paths.insert(path_index, exact_path.clone()); - } - path_index - } - }; - re_exports.push((item_index, path_index)); - } - } - } - - let mut names = Vec::with_capacity(self.items.len()); - let mut types = String::with_capacity(self.items.len()); - let mut full_paths = Vec::with_capacity(self.items.len()); - let mut parents = String::with_capacity(self.items.len()); - let mut parents_backref_queue = VecDeque::new(); - let mut functions = String::with_capacity(self.items.len()); - let mut deprecated = Vec::with_capacity(self.items.len()); - - let mut type_backref_queue = VecDeque::new(); - - let mut last_name = None; - for (index, item) in self.items.iter().enumerate() { - let n = item.ty as u8; - let c = char::from(n + b'A'); - assert!(c <= 'z', "item types must fit within ASCII printables"); - types.push(c); - - assert_eq!( - item.parent.is_some(), - item.parent_idx.is_some(), - "`{}` is missing idx", - item.name - ); - assert!( - parents_backref_queue.len() <= 16, - "the string encoding only supports 16 slots of lookback" - ); - let parent: i32 = item.parent_idx.map(|x| x + 1).unwrap_or(0).try_into().unwrap(); - if let Some(idx) = parents_backref_queue.iter().position(|p: &i32| *p == parent) { - parents.push( - char::try_from('0' as u32 + u32::try_from(idx).unwrap()) - .expect("last possible value is '?'"), - ); - } else if parent == 0 { - write_vlqhex_to_string(parent, &mut parents); - } else { - parents_backref_queue.push_front(parent); - write_vlqhex_to_string(parent, &mut parents); - if parents_backref_queue.len() > 16 { - parents_backref_queue.pop_back(); - } - } - - if Some(item.name.as_str()) == last_name { - names.push(""); + // The number 8 is arbitrary. We want it big, but not enormous, + // because the postings list has to fill in an empty array for each + // unoccupied size. + if item.ty.is_fn_like() { 0 } else { 16 }; + serialized_index.function_data[new_entry_id] = Some(FunctionData { + function_signature: { + let mut function_signature = String::new(); + search_type.write_to_string_without_param_names(&mut function_signature); + function_signature + }, + param_names: search_type + .param_names + .iter() + .map(|sym| sym.map(|sym| sym.to_string()).unwrap_or(String::new())) + .collect::<Vec<String>>(), + }); + for index in used_in_function_signature { + let postings = if index >= 0 { + assert!(serialized_index.path_data[index as usize].is_some()); + &mut serialized_index.type_data[index as usize] + .as_mut() + .unwrap() + .inverted_function_signature_index } else { - names.push(item.name.as_str()); - last_name = Some(item.name.as_str()); - } - - if !item.path.is_empty() { - full_paths.push((index, &item.path)); - } - - match &item.search_type { - Some(ty) => ty.write_to_string(&mut functions, &mut type_backref_queue), - None => functions.push('`'), - } - - if item.deprecation.is_some() { - // bitmasks always use 1-indexing for items, with 0 as the crate itself - deprecated.push(u32::try_from(index + 1).unwrap()); - } - } - - for (index, path) in &revert_extra_paths { - full_paths.push((*index, path)); - } - - let param_names: Vec<(usize, String)> = { - let mut prev = Vec::new(); - let mut result = Vec::new(); - for (index, item) in self.items.iter().enumerate() { - if let Some(ty) = &item.search_type - && let my = ty - .param_names - .iter() - .filter_map(|sym| sym.map(|sym| sym.to_string())) - .collect::<Vec<_>>() - && my != prev - { - result.push((index, my.join(","))); - prev = my; + let generic_id = usize::try_from(-index).unwrap() - 1; + for _ in serialized_index.generic_inverted_index.len()..=generic_id { + serialized_index.generic_inverted_index.push(Vec::new()); } + &mut serialized_index.generic_inverted_index[generic_id] + }; + while postings.len() <= search_type_size { + postings.push(Vec::new()); } - result - }; - - let has_aliases = !self.aliases.is_empty(); - let mut crate_data = - serializer.serialize_struct("CrateData", if has_aliases { 13 } else { 12 })?; - crate_data.serialize_field("t", &types)?; - crate_data.serialize_field("n", &names)?; - crate_data.serialize_field("q", &full_paths)?; - crate_data.serialize_field("i", &parents)?; - crate_data.serialize_field("f", &functions)?; - crate_data.serialize_field("D", &self.desc_index)?; - crate_data.serialize_field("p", &paths)?; - crate_data.serialize_field("r", &re_exports)?; - crate_data.serialize_field("b", &self.associated_item_disambiguators)?; - crate_data.serialize_field("c", &bitmap_to_string(&deprecated))?; - crate_data.serialize_field("e", &bitmap_to_string(&self.empty_desc))?; - crate_data.serialize_field("P", ¶m_names)?; - if has_aliases { - crate_data.serialize_field("a", &self.aliases)?; + postings[search_type_size].push(new_entry_id as u32); } - crate_data.end() } } - let (empty_desc, desc) = { - let mut empty_desc = Vec::new(); - let mut result = Vec::new(); - let mut set = String::new(); - let mut len: usize = 0; - let mut item_index: u32 = 0; - for desc in std::iter::once(&crate_doc).chain(crate_items.iter().map(|item| &item.desc)) { - if desc.is_empty() { - empty_desc.push(item_index); - item_index += 1; - continue; - } - if set.len() >= DESC_INDEX_SHARD_LEN { - result.push((len, std::mem::take(&mut set))); - len = 0; - } else if len != 0 { - set.push('\n'); - } - set.push_str(desc); - len += 1; - item_index += 1; - } - result.push((len, std::mem::take(&mut set))); - (empty_desc, result) - }; - - let desc_index = { - let mut desc_index = String::with_capacity(desc.len() * 4); - for &(len, _) in desc.iter() { - write_vlqhex_to_string(len.try_into().unwrap(), &mut desc_index); - } - desc_index - }; - - assert_eq!( - crate_items.len() + 1, - desc.iter().map(|(len, _)| *len).sum::<usize>() + empty_desc.len() - ); - - // The index, which is actually used to search, is JSON - // It uses `JSON.parse(..)` to actually load, since JSON - // parses faster than the full JavaScript syntax. - let crate_name = krate.name(tcx); - let data = CrateData { - items: crate_items, - paths: crate_paths, - aliases: &aliases, - associated_item_disambiguators: &associated_item_disambiguators, - desc_index, - empty_desc, - }; - let index = OrderedJson::array_unsorted([ - OrderedJson::serialize(crate_name.as_str()).unwrap(), - OrderedJson::serialize(data).unwrap(), - ]); - SerializedSearchIndex { index, desc } + Ok(serialized_index.sort()) } pub(crate) fn get_function_type_for_search( diff --git a/src/librustdoc/html/render/search_index/encode.rs b/src/librustdoc/html/render/search_index/encode.rs index de2f54558ff..d15e13a2d37 100644 --- a/src/librustdoc/html/render/search_index/encode.rs +++ b/src/librustdoc/html/render/search_index/encode.rs @@ -1,6 +1,4 @@ -use base64::prelude::*; - -pub(crate) fn write_vlqhex_to_string(n: i32, string: &mut String) { +pub(crate) fn write_signed_vlqhex_to_string(n: i32, string: &mut String) { let (sign, magnitude): (bool, u32) = if n >= 0 { (false, n.try_into().unwrap()) } else { (true, (-n).try_into().unwrap()) }; // zig-zag encoding @@ -37,206 +35,66 @@ pub(crate) fn write_vlqhex_to_string(n: i32, string: &mut String) { } } -// Used during bitmap encoding -enum Container { - /// number of ones, bits - Bits(Box<[u64; 1024]>), - /// list of entries - Array(Vec<u16>), - /// list of (start, len-1) - Run(Vec<(u16, u16)>), -} -impl Container { - fn popcount(&self) -> u32 { - match self { - Container::Bits(bits) => bits.iter().copied().map(|x| x.count_ones()).sum(), - Container::Array(array) => { - array.len().try_into().expect("array can't be bigger than 2**32") - } - Container::Run(runs) => { - runs.iter().copied().map(|(_, lenm1)| u32::from(lenm1) + 1).sum() - } +pub fn read_signed_vlqhex_from_string(string: &[u8]) -> Option<(i32, usize)> { + let mut n = 0i32; + let mut i = 0; + while let Some(&c) = string.get(i) { + i += 1; + n = (n << 4) | i32::from(c & 0xF); + if c >= 96 { + // zig-zag encoding + let (sign, magnitude) = (n & 1, n >> 1); + let value = if sign == 0 { 1 } else { -1 } * magnitude; + return Some((value, i)); } } - fn push(&mut self, value: u16) { - match self { - Container::Bits(bits) => bits[value as usize >> 6] |= 1 << (value & 0x3F), - Container::Array(array) => { - array.push(value); - if array.len() >= 4096 { - let array = std::mem::take(array); - *self = Container::Bits(Box::new([0; 1024])); - for value in array { - self.push(value); - } - } - } - Container::Run(runs) => { - if let Some(r) = runs.last_mut() - && r.0 + r.1 + 1 == value - { - r.1 += 1; - } else { - runs.push((value, 0)); - } - } + None +} + +pub fn write_postings_to_string(postings: &[Vec<u32>], buf: &mut Vec<u8>) { + for list in postings { + if list.is_empty() { + buf.push(0); + continue; } - } - fn try_make_run(&mut self) -> bool { - match self { - Container::Bits(bits) => { - let mut r: u64 = 0; - for (i, chunk) in bits.iter().copied().enumerate() { - let next_chunk = - i.checked_add(1).and_then(|i| bits.get(i)).copied().unwrap_or(0); - r += !chunk & u64::from((chunk << 1).count_ones()); - r += !next_chunk & u64::from((chunk >> 63).count_ones()); - } - if (2 + 4 * r) >= 8192 { - return false; - } - let bits = std::mem::replace(bits, Box::new([0; 1024])); - *self = Container::Run(Vec::new()); - for (i, bits) in bits.iter().copied().enumerate() { - if bits == 0 { - continue; - } - for j in 0..64 { - let value = (u16::try_from(i).unwrap() << 6) | j; - if bits & (1 << j) != 0 { - self.push(value); - } - } - } - true + let len_before = buf.len(); + stringdex::internals::encode::write_bitmap_to_bytes(&list, &mut *buf).unwrap(); + let len_after = buf.len(); + if len_after - len_before > 1 + (4 * list.len()) && list.len() < 0x3a { + buf.truncate(len_before); + buf.push(list.len() as u8); + for &item in list { + buf.push(item as u8); + buf.push((item >> 8) as u8); + buf.push((item >> 16) as u8); + buf.push((item >> 24) as u8); } - Container::Array(array) if array.len() <= 5 => false, - Container::Array(array) => { - let mut r = 0; - let mut prev = None; - for value in array.iter().copied() { - if value.checked_sub(1) != prev { - r += 1; - } - prev = Some(value); - } - if 2 + 4 * r >= 2 * array.len() + 2 { - return false; - } - let array = std::mem::take(array); - *self = Container::Run(Vec::new()); - for value in array { - self.push(value); - } - true - } - Container::Run(_) => true, } } } -// checked against roaring-rs in -// https://gitlab.com/notriddle/roaring-test -pub(crate) fn write_bitmap_to_bytes( - domain: &[u32], - mut out: impl std::io::Write, -) -> std::io::Result<()> { - // https://arxiv.org/pdf/1603.06549.pdf - let mut keys = Vec::<u16>::new(); - let mut containers = Vec::<Container>::new(); - let mut key: u16; - let mut domain_iter = domain.iter().copied().peekable(); - let mut has_run = false; - while let Some(entry) = domain_iter.next() { - key = (entry >> 16).try_into().expect("shifted off the top 16 bits, so it should fit"); - let value: u16 = (entry & 0x00_00_FF_FF).try_into().expect("AND 16 bits, so it should fit"); - let mut container = Container::Array(vec![value]); - while let Some(entry) = domain_iter.peek().copied() { - let entry_key: u16 = - (entry >> 16).try_into().expect("shifted off the top 16 bits, so it should fit"); - if entry_key != key { - break; +pub fn read_postings_from_string(postings: &mut Vec<Vec<u32>>, mut buf: &[u8]) { + use stringdex::internals::decode::RoaringBitmap; + while let Some(&c) = buf.get(0) { + if c < 0x3a { + buf = &buf[1..]; + let mut slot = Vec::new(); + for _ in 0..c { + slot.push( + (buf[0] as u32) + | ((buf[1] as u32) << 8) + | ((buf[2] as u32) << 16) + | ((buf[3] as u32) << 24), + ); + buf = &buf[4..]; } - domain_iter.next().expect("peeking just succeeded"); - container - .push((entry & 0x00_00_FF_FF).try_into().expect("AND 16 bits, so it should fit")); - } - keys.push(key); - has_run = container.try_make_run() || has_run; - containers.push(container); - } - // https://github.com/RoaringBitmap/RoaringFormatSpec - const SERIAL_COOKIE_NO_RUNCONTAINER: u32 = 12346; - const SERIAL_COOKIE: u32 = 12347; - const NO_OFFSET_THRESHOLD: u32 = 4; - let size: u32 = containers.len().try_into().unwrap(); - let start_offset = if has_run { - out.write_all(&u32::to_le_bytes(SERIAL_COOKIE | ((size - 1) << 16)))?; - for set in containers.chunks(8) { - let mut b = 0; - for (i, container) in set.iter().enumerate() { - if matches!(container, &Container::Run(..)) { - b |= 1 << i; - } - } - out.write_all(&[b])?; - } - if size < NO_OFFSET_THRESHOLD { - 4 + 4 * size + size.div_ceil(8) + postings.push(slot); } else { - 4 + 8 * size + size.div_ceil(8) - } - } else { - out.write_all(&u32::to_le_bytes(SERIAL_COOKIE_NO_RUNCONTAINER))?; - out.write_all(&u32::to_le_bytes(containers.len().try_into().unwrap()))?; - 4 + 4 + 4 * size + 4 * size - }; - for (&key, container) in keys.iter().zip(&containers) { - // descriptive header - let key: u32 = key.into(); - let count: u32 = container.popcount() - 1; - out.write_all(&u32::to_le_bytes((count << 16) | key))?; - } - if !has_run || size >= NO_OFFSET_THRESHOLD { - // offset header - let mut starting_offset = start_offset; - for container in &containers { - out.write_all(&u32::to_le_bytes(starting_offset))?; - starting_offset += match container { - Container::Bits(_) => 8192u32, - Container::Array(array) => u32::try_from(array.len()).unwrap() * 2, - Container::Run(runs) => 2 + u32::try_from(runs.len()).unwrap() * 4, - }; + let (bitmap, consumed_bytes_len) = + RoaringBitmap::from_bytes(buf).unwrap_or_else(|| (RoaringBitmap::default(), 0)); + assert_ne!(consumed_bytes_len, 0); + postings.push(bitmap.to_vec()); + buf = &buf[consumed_bytes_len..]; } } - for container in &containers { - match container { - Container::Bits(bits) => { - for chunk in bits.iter() { - out.write_all(&u64::to_le_bytes(*chunk))?; - } - } - Container::Array(array) => { - for value in array.iter() { - out.write_all(&u16::to_le_bytes(*value))?; - } - } - Container::Run(runs) => { - out.write_all(&u16::to_le_bytes(runs.len().try_into().unwrap()))?; - for (start, lenm1) in runs.iter().copied() { - out.write_all(&u16::to_le_bytes(start))?; - out.write_all(&u16::to_le_bytes(lenm1))?; - } - } - } - } - Ok(()) -} - -pub(crate) fn bitmap_to_string(domain: &[u32]) -> String { - let mut buf = Vec::new(); - let mut strbuf = String::new(); - write_bitmap_to_bytes(domain, &mut buf).unwrap(); - BASE64_STANDARD.encode_string(&buf, &mut strbuf); - strbuf } diff --git a/src/librustdoc/html/render/write_shared.rs b/src/librustdoc/html/render/write_shared.rs index 1f691392b17..e37a5246a76 100644 --- a/src/librustdoc/html/render/write_shared.rs +++ b/src/librustdoc/html/render/write_shared.rs @@ -65,17 +65,17 @@ pub(crate) fn write_shared( // Write shared runs within a flock; disable thread dispatching of IO temporarily. let _lock = try_err!(flock::Lock::new(&lock_file, true, true, true), &lock_file); - let SerializedSearchIndex { index, desc } = build_index(krate, &mut cx.shared.cache, tcx); - write_search_desc(cx, krate, &desc)?; // does not need to be merged + let search_index = + build_index(krate, &mut cx.shared.cache, tcx, &cx.dst, &cx.shared.resource_suffix)?; let crate_name = krate.name(cx.tcx()); let crate_name = crate_name.as_str(); // rand let crate_name_json = OrderedJson::serialize(crate_name).unwrap(); // "rand" let external_crates = hack_get_external_crate_names(&cx.dst, &cx.shared.resource_suffix)?; let info = CrateInfo { - version: CrateInfoVersion::V1, + version: CrateInfoVersion::V2, src_files_js: SourcesPart::get(cx, &crate_name_json)?, - search_index_js: SearchIndexPart::get(index, &cx.shared.resource_suffix)?, + search_index, all_crates: AllCratesPart::get(crate_name_json.clone(), &cx.shared.resource_suffix)?, crates_index: CratesIndexPart::get(crate_name, &external_crates)?, trait_impl: TraitAliasPart::get(cx, &crate_name_json)?, @@ -141,7 +141,7 @@ pub(crate) fn write_not_crate_specific( resource_suffix: &str, include_sources: bool, ) -> Result<(), Error> { - write_rendered_cross_crate_info(crates, dst, opt, include_sources)?; + write_rendered_cross_crate_info(crates, dst, opt, include_sources, resource_suffix)?; write_static_files(dst, opt, style_files, css_file_extension, resource_suffix)?; Ok(()) } @@ -151,13 +151,18 @@ fn write_rendered_cross_crate_info( dst: &Path, opt: &RenderOptions, include_sources: bool, + resource_suffix: &str, ) -> Result<(), Error> { let m = &opt.should_merge; if opt.should_emit_crate() { if include_sources { write_rendered_cci::<SourcesPart, _>(SourcesPart::blank, dst, crates, m)?; } - write_rendered_cci::<SearchIndexPart, _>(SearchIndexPart::blank, dst, crates, m)?; + crates + .iter() + .fold(SerializedSearchIndex::default(), |a, b| a.union(&b.search_index)) + .sort() + .write_to(dst, resource_suffix)?; write_rendered_cci::<AllCratesPart, _>(AllCratesPart::blank, dst, crates, m)?; } write_rendered_cci::<TraitAliasPart, _>(TraitAliasPart::blank, dst, crates, m)?; @@ -215,38 +220,12 @@ fn write_static_files( Ok(()) } -/// Write the search description shards to disk -fn write_search_desc( - cx: &mut Context<'_>, - krate: &Crate, - search_desc: &[(usize, String)], -) -> Result<(), Error> { - let crate_name = krate.name(cx.tcx()).to_string(); - let encoded_crate_name = OrderedJson::serialize(&crate_name).unwrap(); - let path = PathBuf::from_iter([&cx.dst, Path::new("search.desc"), Path::new(&crate_name)]); - if path.exists() { - try_err!(fs::remove_dir_all(&path), &path); - } - for (i, (_, part)) in search_desc.iter().enumerate() { - let filename = static_files::suffix_path( - &format!("{crate_name}-desc-{i}-.js"), - &cx.shared.resource_suffix, - ); - let path = path.join(filename); - let part = OrderedJson::serialize(part).unwrap(); - let part = format!("searchState.loadedDescShard({encoded_crate_name}, {i}, {part})"); - create_parents(&path)?; - try_err!(fs::write(&path, part), &path); - } - Ok(()) -} - /// Contains pre-rendered contents to insert into the CCI template #[derive(Serialize, Deserialize, Clone, Debug)] pub(crate) struct CrateInfo { version: CrateInfoVersion, src_files_js: PartsAndLocations<SourcesPart>, - search_index_js: PartsAndLocations<SearchIndexPart>, + search_index: SerializedSearchIndex, all_crates: PartsAndLocations<AllCratesPart>, crates_index: PartsAndLocations<CratesIndexPart>, trait_impl: PartsAndLocations<TraitAliasPart>, @@ -277,7 +256,7 @@ impl CrateInfo { /// to provide better diagnostics about including an invalid file. #[derive(Serialize, Deserialize, Clone, Debug)] enum CrateInfoVersion { - V1, + V2, } /// Paths (relative to the doc root) and their pre-merge contents @@ -332,36 +311,6 @@ trait CciPart: Sized + fmt::Display + DeserializeOwned + 'static { } #[derive(Serialize, Deserialize, Clone, Default, Debug)] -struct SearchIndex; -type SearchIndexPart = Part<SearchIndex, EscapedJson>; -impl CciPart for SearchIndexPart { - type FileFormat = sorted_template::Js; - fn from_crate_info(crate_info: &CrateInfo) -> &PartsAndLocations<Self> { - &crate_info.search_index_js - } -} - -impl SearchIndexPart { - fn blank() -> SortedTemplate<<Self as CciPart>::FileFormat> { - SortedTemplate::from_before_after( - r"var searchIndex = new Map(JSON.parse('[", - r"]')); -if (typeof exports !== 'undefined') exports.searchIndex = searchIndex; -else if (window.initSearch) window.initSearch(searchIndex);", - ) - } - - fn get( - search_index: OrderedJson, - resource_suffix: &str, - ) -> Result<PartsAndLocations<Self>, Error> { - let path = suffix_path("search-index.js", resource_suffix); - let search_index = EscapedJson::from(search_index); - Ok(PartsAndLocations::with(path, search_index)) - } -} - -#[derive(Serialize, Deserialize, Clone, Default, Debug)] struct AllCrates; type AllCratesPart = Part<AllCrates, OrderedJson>; impl CciPart for AllCratesPart { @@ -426,6 +375,7 @@ impl CratesIndexPart { fn blank(cx: &Context<'_>) -> SortedTemplate<<Self as CciPart>::FileFormat> { let page = layout::Page { title: "Index of crates", + short_title: "Crates", css_class: "mod sys", root_path: "./", static_root_path: cx.shared.static_root_path.as_deref(), diff --git a/src/librustdoc/html/render/write_shared/tests.rs b/src/librustdoc/html/render/write_shared/tests.rs index 6f185e85345..1989a1f87aa 100644 --- a/src/librustdoc/html/render/write_shared/tests.rs +++ b/src/librustdoc/html/render/write_shared/tests.rs @@ -30,14 +30,6 @@ fn sources_template() { } #[test] -fn sources_parts() { - let parts = - SearchIndexPart::get(OrderedJson::serialize(["foo", "bar"]).unwrap(), "suffix").unwrap(); - assert_eq!(&parts.parts[0].0, Path::new("search-indexsuffix.js")); - assert_eq!(&parts.parts[0].1.to_string(), r#"["foo","bar"]"#); -} - -#[test] fn all_crates_template() { let mut template = AllCratesPart::blank(); assert_eq!(but_last_line(&template.to_string()), r"window.ALL_CRATES = [];"); @@ -55,31 +47,6 @@ fn all_crates_parts() { } #[test] -fn search_index_template() { - let mut template = SearchIndexPart::blank(); - assert_eq!( - but_last_line(&template.to_string()), - r"var searchIndex = new Map(JSON.parse('[]')); -if (typeof exports !== 'undefined') exports.searchIndex = searchIndex; -else if (window.initSearch) window.initSearch(searchIndex);" - ); - template.append(EscapedJson::from(OrderedJson::serialize([1, 2]).unwrap()).to_string()); - assert_eq!( - but_last_line(&template.to_string()), - r"var searchIndex = new Map(JSON.parse('[[1,2]]')); -if (typeof exports !== 'undefined') exports.searchIndex = searchIndex; -else if (window.initSearch) window.initSearch(searchIndex);" - ); - template.append(EscapedJson::from(OrderedJson::serialize([4, 3]).unwrap()).to_string()); - assert_eq!( - but_last_line(&template.to_string()), - r"var searchIndex = new Map(JSON.parse('[[1,2],[4,3]]')); -if (typeof exports !== 'undefined') exports.searchIndex = searchIndex; -else if (window.initSearch) window.initSearch(searchIndex);" - ); -} - -#[test] fn crates_index_part() { let external_crates = ["bar".to_string(), "baz".to_string()]; let mut parts = CratesIndexPart::get("foo", &external_crates).unwrap(); diff --git a/src/librustdoc/html/sources.rs b/src/librustdoc/html/sources.rs index c34b3154269..9c5518a780e 100644 --- a/src/librustdoc/html/sources.rs +++ b/src/librustdoc/html/sources.rs @@ -230,6 +230,7 @@ impl SourceCollector<'_, '_> { ); let page = layout::Page { title: &title, + short_title: &src_fname.to_string_lossy(), css_class: "src", root_path: &root_path, static_root_path: shared.static_root_path.as_deref(), diff --git a/src/librustdoc/html/static/css/rustdoc.css b/src/librustdoc/html/static/css/rustdoc.css index c48863b4681..dc27d7943d9 100644 --- a/src/librustdoc/html/static/css/rustdoc.css +++ b/src/librustdoc/html/static/css/rustdoc.css @@ -258,6 +258,17 @@ h1, h2, h3, h4 { padding-bottom: 6px; margin-bottom: 15px; } +.search-results-main-heading { + grid-template-areas: + "main-heading-breadcrumbs main-heading-placeholder" + "main-heading-breadcrumbs main-heading-toolbar " + "main-heading-h1 main-heading-toolbar "; +} +.search-results-main-heading nav.sub { + grid-area: main-heading-h1; + align-items: end; + margin: 4px 0 8px 0; +} .rustdoc-breadcrumbs { grid-area: main-heading-breadcrumbs; line-height: 1.25; @@ -265,6 +276,16 @@ h1, h2, h3, h4 { position: relative; z-index: 1; } +.search-switcher { + grid-area: main-heading-breadcrumbs; + line-height: 1.5; + display: flex; + color: var(--main-color); + align-items: baseline; + white-space: nowrap; + padding-top: 8px; + min-height: 34px; +} .rustdoc-breadcrumbs a { padding: 5px 0 7px; } @@ -305,7 +326,7 @@ h4.code-header { #crate-search, h1, h2, h3, h4, h5, h6, .sidebar, -.mobile-topbar, +rustdoc-topbar, .search-input, .search-results .result-name, .item-table dt > a, @@ -317,6 +338,7 @@ rustdoc-toolbar, summary.hideme, .scraped-example-list, .rustdoc-breadcrumbs, +.search-switcher, /* This selector is for the items listed in the "all items" page. */ ul.all-items { font-family: "Fira Sans", Arial, NanumBarunGothic, sans-serif; @@ -329,7 +351,7 @@ a.anchor, .rust a, .sidebar h2 a, .sidebar h3 a, -.mobile-topbar h2 a, +rustdoc-topbar h2 a, h1 a, .search-results a, .search-results li, @@ -616,7 +638,7 @@ img { color: var(--sidebar-resizer-active); } -.sidebar, .mobile-topbar, .sidebar-menu-toggle, +.sidebar, rustdoc-topbar, .sidebar-menu-toggle, #src-sidebar { background-color: var(--sidebar-background-color); } @@ -857,7 +879,7 @@ ul.block, .block li, .block ul { margin-bottom: 1rem; } -.mobile-topbar { +rustdoc-topbar { display: none; } @@ -1098,16 +1120,15 @@ div.where { nav.sub { flex-grow: 1; flex-flow: row nowrap; - margin: 4px 0 0 0; display: flex; - align-items: center; + align-items: start; + margin-top: 4px; } .search-form { position: relative; display: flex; height: 34px; flex-grow: 1; - margin-bottom: 4px; } .src nav.sub { margin: 0 0 -10px 0; @@ -1208,27 +1229,14 @@ table, margin-left: 0; } -.search-results-title { - margin-top: 0; - white-space: nowrap; - /* flex layout allows shrinking the <select> appropriately if it becomes too large */ - display: flex; - /* make things look like in a line, despite the fact that we're using a layout - with boxes (i.e. from the flex layout) */ - align-items: baseline; -} -.search-results-title + .sub-heading { - color: var(--main-color); - display: flex; - align-items: baseline; - white-space: nowrap; -} #crate-search-div { /* ensures that 100% in properties of #crate-search-div:after are relative to the size of this div */ position: relative; /* allows this div (and with it the <select>-element "#crate-search") to be shrunk */ min-width: 0; + /* keep label text for switcher from moving down when this appears */ + margin-top: -1px; } #crate-search { padding: 0 23px 0 4px; @@ -1294,6 +1302,7 @@ so that we can apply CSS-filters to change the arrow color in themes */ flex-grow: 1; background-color: var(--button-background-color); color: var(--search-color); + max-width: 100%; } .search-input:focus { border-color: var(--search-input-focused-border-color); @@ -1459,14 +1468,14 @@ so that we can apply CSS-filters to change the arrow color in themes */ } #settings.popover { - --popover-arrow-offset: 202px; + --popover-arrow-offset: 196px; top: calc(100% - 16px); } /* use larger max-width for help popover, but not for help.html */ #help.popover { max-width: 600px; - --popover-arrow-offset: 118px; + --popover-arrow-offset: 115px; top: calc(100% - 16px); } @@ -1929,10 +1938,12 @@ a.tooltip:hover::after { color: inherit; } #search-tabs button:not(.selected) { + --search-tab-button-background: var(--search-tab-button-not-selected-background); background-color: var(--search-tab-button-not-selected-background); border-top-color: var(--search-tab-button-not-selected-border-top-color); } #search-tabs button:hover, #search-tabs button.selected { + --search-tab-button-background: var(--search-tab-button-selected-background); background-color: var(--search-tab-button-selected-background); border-top-color: var(--search-tab-button-selected-border-top-color); } @@ -1941,6 +1952,73 @@ a.tooltip:hover::after { font-size: 1rem; font-variant-numeric: tabular-nums; color: var(--search-tab-title-count-color); + position: relative; +} + +#search-tabs .count.loading { + color: transparent; +} + +.search-form.loading { + --search-tab-button-background: var(--button-background-color); +} + +#search-tabs .count.loading::before, +.search-form.loading::before +{ + width: 16px; + height: 16px; + border-radius: 16px; + background: radial-gradient( + var(--search-tab-button-background) 0 50%, + transparent 50% 100% + ), conic-gradient( + var(--code-highlight-kw-color) 0deg 30deg, + var(--code-highlight-prelude-color) 30deg 60deg, + var(--code-highlight-number-color) 90deg 120deg, + var(--code-highlight-lifetime-color ) 120deg 150deg, + var(--code-highlight-comment-color) 150deg 180deg, + var(--code-highlight-self-color) 180deg 210deg, + var(--code-highlight-attribute-color) 210deg 240deg, + var(--code-highlight-literal-color) 210deg 240deg, + var(--code-highlight-macro-color) 240deg 270deg, + var(--code-highlight-question-mark-color) 270deg 300deg, + var(--code-highlight-prelude-val-color) 300deg 330deg, + var(--code-highlight-doc-comment-color) 330deg 360deg + ); + content: ""; + position: absolute; + left: 2px; + top: 2px; + animation: rotating 1.25s linear infinite; +} +#search-tabs .count.loading::after, +.search-form.loading::after +{ + width: 18px; + height: 18px; + border-radius: 18px; + background: conic-gradient( + var(--search-tab-button-background) 0deg 180deg, + transparent 270deg 360deg + ); + content: ""; + position: absolute; + left: 1px; + top: 1px; + animation: rotating 0.66s linear infinite; +} + +.search-form.loading::before { + left: auto; + right: 9px; + top: 8px; +} + +.search-form.loading::after { + left: auto; + right: 8px; + top: 8px; } #search .error code { @@ -1974,7 +2052,7 @@ a.tooltip:hover::after { border-bottom: 1px solid var(--border-color); } -#settings-menu, #help-button, button#toggle-all-docs { +#search-button, .settings-menu, .help-menu, button#toggle-all-docs { margin-left: var(--button-left-margin); display: flex; line-height: 1.25; @@ -1989,69 +2067,100 @@ a.tooltip:hover::after { display: flex; margin-right: 4px; position: fixed; + margin-top: 25px; + left: 6px; height: 34px; width: 34px; + z-index: calc(var(--desktop-sidebar-z-index) + 1); } .hide-sidebar #sidebar-button { left: 6px; background-color: var(--main-background-color); - z-index: 1; } .src #sidebar-button { + margin-top: 0; + top: 8px; left: 8px; - z-index: calc(var(--desktop-sidebar-z-index) + 1); + border-color: var(--border-color); } .hide-sidebar .src #sidebar-button { position: static; } -#settings-menu > a, #help-button > a, #sidebar-button > a, button#toggle-all-docs { +#search-button > a, +.settings-menu > a, +.help-menu > a, +#sidebar-button > a, +button#toggle-all-docs { display: flex; align-items: center; justify-content: center; flex-direction: column; } -#settings-menu > a, #help-button > a, button#toggle-all-docs { +#search-button > a, +.settings-menu > a, +.help-menu > a, +button#toggle-all-docs { border: 1px solid transparent; border-radius: var(--button-border-radius); color: var(--main-color); } -#settings-menu > a, #help-button > a, button#toggle-all-docs { +#search-button > a, .settings-menu > a, .help-menu > a, button#toggle-all-docs { width: 80px; border-radius: var(--toolbar-button-border-radius); } -#settings-menu > a, #help-button > a { +#search-button > a, .settings-menu > a, .help-menu > a { min-width: 0; } #sidebar-button > a { - background-color: var(--sidebar-background-color); + border: solid 1px transparent; + border-radius: var(--button-border-radius); + background-color: var(--button-background-color); width: 33px; } -#sidebar-button > a:hover, #sidebar-button > a:focus-visible { - background-color: var(--main-background-color); +.src #sidebar-button > a { + background-color: var(--sidebar-background-color); + border-color: var(--border-color); } -#settings-menu > a:hover, #settings-menu > a:focus-visible, -#help-button > a:hover, #help-button > a:focus-visible, +#search-button > a:hover, #search-button > a:focus-visible, +.settings-menu > a:hover, .settings-menu > a:focus-visible, +.help-menu > a:hover, #help-menu > a:focus-visible, +#sidebar-button > a:hover, #sidebar-button > a:focus-visible, +#copy-path:hover, #copy-path:focus-visible, button#toggle-all-docs:hover, button#toggle-all-docs:focus-visible { border-color: var(--settings-button-border-focus); text-decoration: none; } -#settings-menu > a::before { +#search-button > a::before { + /* Magnifying glass */ + content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" \ + width="18" height="18" viewBox="0 0 16 16">\ + <circle r="5" cy="7" cx="7" style="fill:none;stroke:black;stroke-width:3"/><path \ + d="M14.5,14.5 12,12" style="fill:none;stroke:black;stroke-width:3;stroke-linecap:round">\ + </path><desc>Search</desc>\ + </svg>'); + width: 18px; + height: 18px; + filter: var(--settings-menu-filter); +} + +.settings-menu > a::before { /* Wheel <https://www.svgrepo.com/svg/384069/settings-cog-gear> */ content: url('data:image/svg+xml,<svg width="18" height="18" viewBox="0 0 12 12" \ enable-background="new 0 0 12 12" xmlns="http://www.w3.org/2000/svg">\ - <path d="M10.25,6c0-0.1243286-0.0261841-0.241333-0.0366211-0.362915l1.6077881-1.5545654l\ - -1.25-2.1650391 c0,0-1.2674561,0.3625488-2.1323853,0.6099854c-0.2034912-0.1431885-0.421875\ - -0.2639771-0.6494751-0.3701782L7.25,0h-2.5 c0,0-0.3214111,1.2857666-0.5393066,2.1572876\ - C3.9830933,2.2634888,3.7647095,2.3842773,3.5612183,2.5274658L1.428833,1.9174805 \ - l-1.25,2.1650391c0,0,0.9641113,0.9321899,1.6077881,1.5545654C1.7761841,5.758667,\ - 1.75,5.8756714,1.75,6 s0.0261841,0.241333,0.0366211,0.362915L0.178833,7.9174805l1.25,\ - 2.1650391l2.1323853-0.6099854 c0.2034912,0.1432495,0.421875,0.2639771,0.6494751,0.3701782\ - L4.75,12h2.5l0.5393066-2.1572876 c0.2276001-0.1062012,0.4459839-0.2269287,0.6494751\ - -0.3701782l2.1323853,0.6099854l1.25-2.1650391L10.2133789,6.362915 C10.2238159,6.241333,\ - 10.25,6.1243286,10.25,6z M6,7.5C5.1715698,7.5,4.5,6.8284302,4.5,6S5.1715698,4.5,6,4.5S7.5\ - ,5.1715698,7.5,6 S6.8284302,7.5,6,7.5z" fill="black"/></svg>'); + <path d="m4.75 0s-0.32117 1.286-0.53906 2.1576c-0.2276 0.1062-0.44625 \ + 0.2266-0.64974 0.36979l-2.1328-0.60938-1.25 2.1641s0.9644 0.93231 1.6081 1.5547c-0.010437 \ + 0.12158-0.036458 0.23895-0.036458 0.36328s0.026021 0.2417 0.036458 0.36328l-1.6081 \ + 1.5547 1.25 2.1641 2.1328-0.60937c0.20349 0.14325 0.42214 0.26359 0.64974 0.36979l0.53906 \ + 2.1576h2.5l0.53906-2.1576c0.2276-0.1062 0.44625-0.22654 0.64974-0.36979l2.1328 0.60937 \ + 1.25-2.1641-1.6081-1.5547c0.010437-0.12158 0.036458-0.23895 \ + 0.036458-0.36328s-0.02602-0.2417-0.03646-0.36328l1.6081-1.5547-1.25-2.1641s-1.2679 \ + 0.36194-2.1328 0.60938c-0.20349-0.14319-0.42214-0.26359-0.64974-0.36979l-0.53906-2.1576\ + zm1.25 2.5495c1.9058-2.877e-4 3.4508 1.5447 3.4505 3.4505 2.877e-4 1.9058-1.5447 3.4508-3.4505 \ + 3.4505-1.9058 2.877e-4 -3.4508-1.5447-3.4505-3.4505-2.877e-4 -1.9058 1.5447-3.4508 \ + 3.4505-3.4505z" fill="black"/>\ + <circle cx="6" cy="6" r="1.75" fill="none" stroke="black" stroke-width="1"/></svg>'); width: 18px; height: 18px; filter: var(--settings-menu-filter); @@ -2067,36 +2176,51 @@ button#toggle-all-docs::before { filter: var(--settings-menu-filter); } -button#toggle-all-docs.will-expand::before { - /* Custom arrow icon */ - content: url('data:image/svg+xml,<svg width="18" height="18" viewBox="0 0 12 12" \ - enable-background="new 0 0 12 12" xmlns="http://www.w3.org/2000/svg">\ - <path d="M2,5l4,-4l4,4M2,7l4,4l4,-4" stroke="black" fill="none" stroke-width="2px"/></svg>'); -} - -#help-button > a::before { - /* Question mark with circle */ - content: url('data:image/svg+xml,<svg width="18" height="18" viewBox="0 0 12 12" \ - enable-background="new 0 0 12 12" xmlns="http://www.w3.org/2000/svg" fill="none">\ - <circle r="5.25" cx="6" cy="6" stroke-width="1.25" stroke="black"/>\ - <text x="6" y="7" style="font:8px sans-serif;font-weight:1000" text-anchor="middle" \ - dominant-baseline="middle" fill="black">?</text></svg>'); +.help-menu > a::before { + /* Question mark with "circle" */ + content: url('data:image/svg+xml,\ + <svg width="18" height="18" enable-background="new 0 0 12 12" fill="none" \ + version="1.1" viewBox="0 0 12 12" xmlns="http://www.w3.org/2000/svg"> \ + <path d="m6.007 0.6931c2.515 0 5.074 1.908 5.074 5.335 0 3.55-2.567 5.278-5.088 \ + 5.278-2.477 0-5.001-1.742-5.001-5.3 0-3.38 2.527-5.314 5.014-5.314z" stroke="black" \ + stroke-width="1.5"/>\ + <path d="m5.999 7.932c0.3111 0 0.7062 0.2915 0.7062 0.7257 0 0.5458-0.3951 \ + 0.8099-0.7081 0.8099-0.2973 0-0.7023-0.266-0.7023-0.7668 0-0.4695 0.3834-0.7688 \ + 0.7042-0.7688z" fill="black"/>\ + <path d="m4.281 3.946c0.0312-0.03057 0.06298-0.06029 0.09528-0.08916 0.4833-0.432 1.084-0.6722 \ + 1.634-0.6722 1.141 0 1.508 1.043 1.221 1.621-0.2753 0.5542-1.061 0.5065-1.273 \ + 1.595-0.05728 0.2939 0.0134 0.9812 0.0134 1.205" fill="none" stroke="black" \ + stroke-width="1.25"/>\ + </svg>'); width: 18px; height: 18px; filter: var(--settings-menu-filter); } +/* design hack to cope with "Help" being far shorter than "Settings" etc */ +.help-menu > a { + width: 74px; +} +.help-menu > a > .label { + padding-right: 1px; +} +#toggle-all-docs:not(.will-expand) > .label { + padding-left: 1px; +} + +#search-button > a::before, button#toggle-all-docs::before, -#help-button > a::before, -#settings-menu > a::before { +.help-menu > a::before, +.settings-menu > a::before { filter: var(--settings-menu-filter); margin: 8px; } @media not (pointer: coarse) { + #search-button > a:hover::before, button#toggle-all-docs:hover::before, - #help-button > a:hover::before, - #settings-menu > a:hover::before { + .help-menu > a:hover::before, + .settings-menu > a:hover::before { filter: var(--settings-menu-hover-filter); } } @@ -2122,9 +2246,9 @@ rustdoc-toolbar span.label { /* sidebar resizer image */ content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 22 22" \ fill="none" stroke="black">\ - <rect x="1" y="1" width="20" height="20" ry="1.5" stroke-width="1.5" stroke="%23777"/>\ - <circle cx="4.375" cy="4.375" r="1" stroke-width=".75"/>\ - <path d="m7.6121 3v16 M5.375 7.625h-2 m2 3h-2 m2 3h-2" stroke-width="1.25"/></svg>'); + <rect x="1" y="2" width="20" height="18" ry="1.5" stroke-width="1.5" stroke="%23777"/>\ + <circle cx="4.375" cy="5.375" r="1" stroke-width=".75"/>\ + <path d="m7.6121 4v14 M5.375 8.625h-2 m2 3h-2 m2 3h-2" stroke-width="1.25"/></svg>'); width: 22px; height: 22px; } @@ -2137,7 +2261,8 @@ rustdoc-toolbar span.label { margin-left: 10px; padding: 0; padding-left: 2px; - border: 0; + border: solid 1px transparent; + border-radius: var(--button-border-radius); font-size: 0; } #copy-path::before { @@ -2159,7 +2284,7 @@ rustdoc-toolbar span.label { transform: rotate(360deg); } } -#settings-menu.rotate > a img { +.settings-menu.rotate > a img { animation: rotating 2s linear infinite; } @@ -2402,6 +2527,9 @@ However, it's not needed with smaller screen width because the doc/code block is opacity: 0.75; filter: var(--mobile-sidebar-menu-filter); } +.src #sidebar-button > a:hover { + background: var(--main-background-color); +} .sidebar-menu-toggle:hover::before, .sidebar-menu-toggle:active::before, .sidebar-menu-toggle:focus::before { @@ -2410,8 +2538,8 @@ However, it's not needed with smaller screen width because the doc/code block is /* Media Queries */ -/* Make sure all the buttons line wrap at the same time */ @media (max-width: 850px) { + /* Make sure all the buttons line wrap at the same time */ #search-tabs .count { display: block; } @@ -2421,6 +2549,81 @@ However, it's not needed with smaller screen width because the doc/code block is .side-by-side > div { width: auto; } + + /* Text label takes up too much space at this size. */ + .main-heading { + grid-template-areas: + "main-heading-breadcrumbs main-heading-toolbar" + "main-heading-h1 main-heading-toolbar" + "main-heading-sub-heading main-heading-toolbar"; + } + .search-results-main-heading { + display: grid; + grid-template-areas: + "main-heading-breadcrumbs main-heading-toolbar" + "main-heading-breadcrumbs main-heading-toolbar" + "main-heading-h1 main-heading-toolbar"; + } + rustdoc-toolbar { + margin-top: -10px; + display: grid; + grid-template-areas: + "x settings help" + "search summary summary"; + grid-template-rows: 35px 1fr; + } + .search-results-main-heading rustdoc-toolbar { + display: grid; + grid-template-areas: + "settings help" + "search search"; + } + .search-results-main-heading #toggle-all-docs { + display: none; + } + rustdoc-toolbar .settings-menu span.label, + rustdoc-toolbar .help-menu span.label + { + display: none; + } + rustdoc-toolbar .settings-menu { + grid-area: settings; + } + rustdoc-toolbar .help-menu { + grid-area: help; + } + rustdoc-toolbar .settings-menu { + grid-area: settings; + } + rustdoc-toolbar #search-button { + grid-area: search; + } + rustdoc-toolbar #toggle-all-docs { + grid-area: summary; + } + rustdoc-toolbar .settings-menu, + rustdoc-toolbar .help-menu { + height: 35px; + } + rustdoc-toolbar .settings-menu > a, + rustdoc-toolbar .help-menu > a { + border-radius: 2px; + text-align: center; + width: 34px; + padding: 5px 0; + } + rustdoc-toolbar .settings-menu > a:before, + rustdoc-toolbar .help-menu > a:before { + margin: 0 4px; + } + #settings.popover { + top: 16px; + --popover-arrow-offset: 58px; + } + #help.popover { + top: 16px; + --popover-arrow-offset: 16px; + } } /* @@ -2435,7 +2638,7 @@ in src-script.js and main.js /* When linking to an item with an `id` (for instance, by clicking a link in the sidebar, or visiting a URL with a fragment like `#method.new`, we don't want the item to be obscured - by the topbar. Anything with an `id` gets scroll-margin-top equal to .mobile-topbar's size. + by the topbar. Anything with an `id` gets scroll-margin-top equal to rustdoc-topbar's size. */ *[id] { scroll-margin-top: 45px; @@ -2451,18 +2654,32 @@ in src-script.js and main.js visibility: hidden; } - /* Text label takes up too much space at this size. */ - rustdoc-toolbar span.label { + + /* Pull settings and help up into the top bar. */ + rustdoc-topbar span.label, + html:not(.hide-sidebar) .rustdoc:not(.src) rustdoc-toolbar .settings-menu > a, + html:not(.hide-sidebar) .rustdoc:not(.src) rustdoc-toolbar .help-menu > a + { display: none; } - #settings-menu > a, #help-button > a, button#toggle-all-docs { + rustdoc-topbar .settings-menu > a, + rustdoc-topbar .help-menu > a { width: 33px; + line-height: 0; + } + rustdoc-topbar .settings-menu > a:hover, + rustdoc-topbar .help-menu > a:hover { + border: none; + background: var(--main-background-color); + border-radius: 0; } #settings.popover { - --popover-arrow-offset: 86px; + top: 32px; + --popover-arrow-offset: 48px; } #help.popover { - --popover-arrow-offset: 48px; + top: 32px; + --popover-arrow-offset: 12px; } .rustdoc { @@ -2471,13 +2688,13 @@ in src-script.js and main.js display: block; } - main { + html:not(.hide-sidebar) main { padding-left: 15px; padding-top: 0px; } /* Hide the logo and item name from the sidebar. Those are displayed - in the mobile-topbar instead. */ + in the rustdoc-topbar instead. */ .sidebar .logo-container, .sidebar .location, .sidebar-resizer { @@ -2510,6 +2727,9 @@ in src-script.js and main.js height: 100vh; border: 0; } + html .src main { + padding: 18px 0; + } .src .search-form { margin-left: 40px; } @@ -2529,9 +2749,9 @@ in src-script.js and main.js left: 0; } - .mobile-topbar h2 { + rustdoc-topbar > h2 { padding-bottom: 0; - margin: auto 0.5em auto auto; + margin: auto; overflow: hidden; /* Rare exception to specifying font sizes in rem. Since the topbar height is specified in pixels, this also has to be specified in @@ -2540,32 +2760,34 @@ in src-script.js and main.js font-size: 24px; white-space: nowrap; text-overflow: ellipsis; + text-align: center; } - .mobile-topbar .logo-container > img { + rustdoc-topbar .logo-container > img { max-width: 35px; max-height: 35px; margin: 5px 0 5px 20px; } - .mobile-topbar { + rustdoc-topbar { display: flex; flex-direction: row; position: sticky; z-index: 10; - font-size: 2rem; height: 45px; width: 100%; left: 0; top: 0; } - .hide-sidebar .mobile-topbar { + .hide-sidebar rustdoc-topbar { display: none; } .sidebar-menu-toggle { - width: 45px; + /* prevent flexbox shrinking */ + width: 41px; + min-width: 41px; border: none; line-height: 0; } @@ -2591,9 +2813,13 @@ in src-script.js and main.js #sidebar-button > a::before { content: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" \ viewBox="0 0 22 22" fill="none" stroke="black">\ - <rect x="1" y="1" width="20" height="20" ry="1.5" stroke-width="1.5" stroke="%23777"/>\ - <circle cx="4.375" cy="4.375" r="1" stroke-width=".75"/>\ - <path d="m3 7.375h16m0-3h-4" stroke-width="1.25"/></svg>'); + <rect x="1" y="2" width="20" height="18" ry="1.5" stroke-width="1.5" stroke="%23777"/>\ + <g fill="black" stroke="none">\ + <circle cx="4.375" cy="5.375" r="1" stroke-width=".75"/>\ + <circle cx="17.375" cy="5.375" r="1" stroke-width=".75"/>\ + <circle cx="14.375" cy="5.375" r="1" stroke-width=".75"/>\ + </g>\ + <path d="m3 8.375h16" stroke-width="1.25"/></svg>'); width: 22px; height: 22px; } @@ -3283,7 +3509,7 @@ Original by Dempfi (https://github.com/dempfi/ayu) border-bottom: 1px solid rgba(242, 151, 24, 0.3); } -:root[data-theme="ayu"] #settings-menu > a img, +:root[data-theme="ayu"] .settings-menu > a img, :root[data-theme="ayu"] #sidebar-button > a::before { filter: invert(100); } diff --git a/src/librustdoc/html/static/js/main.js b/src/librustdoc/html/static/js/main.js index 8e3d07b3a1c..20fc6b75d37 100644 --- a/src/librustdoc/html/static/js/main.js +++ b/src/librustdoc/html/static/js/main.js @@ -54,23 +54,6 @@ function showMain() { window.rootPath = getVar("root-path"); window.currentCrate = getVar("current-crate"); -function setMobileTopbar() { - // FIXME: It would be nicer to generate this text content directly in HTML, - // but with the current code it's hard to get the right information in the right place. - const mobileTopbar = document.querySelector(".mobile-topbar"); - const locationTitle = document.querySelector(".sidebar h2.location"); - if (mobileTopbar) { - const mobileTitle = document.createElement("h2"); - mobileTitle.className = "location"; - if (hasClass(document.querySelector(".rustdoc"), "crate")) { - mobileTitle.innerHTML = `Crate <a href="#">${window.currentCrate}</a>`; - } else if (locationTitle) { - mobileTitle.innerHTML = locationTitle.innerHTML; - } - mobileTopbar.appendChild(mobileTitle); - } -} - /** * Gets the human-readable string for the virtual-key code of the * given KeyboardEvent, ev. @@ -84,6 +67,7 @@ function setMobileTopbar() { * So I guess you could say things are getting pretty interoperable. * * @param {KeyboardEvent} ev + * @returns {string} */ function getVirtualKey(ev) { if ("key" in ev && typeof ev.key !== "undefined") { @@ -98,18 +82,8 @@ function getVirtualKey(ev) { } const MAIN_ID = "main-content"; -const SETTINGS_BUTTON_ID = "settings-menu"; const ALTERNATIVE_DISPLAY_ID = "alternative-display"; const NOT_DISPLAYED_ID = "not-displayed"; -const HELP_BUTTON_ID = "help-button"; - -function getSettingsButton() { - return document.getElementById(SETTINGS_BUTTON_ID); -} - -function getHelpButton() { - return document.getElementById(HELP_BUTTON_ID); -} // Returns the current URL without any query parameter or hash. function getNakedUrl() { @@ -174,7 +148,7 @@ function getNotDisplayedElem() { * contains the displayed element (there can be only one at the same time!). So basically, we switch * elements between the two `<section>` elements. * - * @param {HTMLElement|null} elemToDisplay + * @param {Element|null} elemToDisplay */ function switchDisplayedElement(elemToDisplay) { const el = getAlternativeDisplayElem(); @@ -239,14 +213,14 @@ function preLoadCss(cssUrl) { document.head.append(script); } - const settingsButton = getSettingsButton(); - if (settingsButton) { - settingsButton.onclick = event => { + onEachLazy(document.querySelectorAll(".settings-menu"), settingsMenu => { + /** @param {MouseEvent} event */ + settingsMenu.querySelector("a").onclick = event => { if (event.ctrlKey || event.altKey || event.metaKey) { return; } window.hideAllModals(false); - addClass(getSettingsButton(), "rotate"); + addClass(settingsMenu, "rotate"); event.preventDefault(); // Sending request for the CSS and the JS files at the same time so it will // hopefully be loaded when the JS will generate the settings content. @@ -268,15 +242,42 @@ function preLoadCss(cssUrl) { } }, 0); }; - } + }); window.searchState = { rustdocToolbar: document.querySelector("rustdoc-toolbar"), loadingText: "Loading search results...", - // This will always be an HTMLInputElement, but tsc can't see that - // @ts-expect-error - input: document.getElementsByClassName("search-input")[0], - outputElement: () => { + inputElement: () => { + let el = document.getElementsByClassName("search-input")[0]; + if (!el) { + const out = nonnull(nonnull(window.searchState.outputElement()).parentElement); + const hdr = document.createElement("div"); + hdr.className = "main-heading search-results-main-heading"; + const params = window.searchState.getQueryStringParams(); + const autofocusParam = params.search === "" ? "autofocus" : ""; + hdr.innerHTML = `<nav class="sub"> + <form class="search-form loading"> + <span></span> <!-- This empty span is a hacky fix for Safari: see #93184 --> + <input + ${autofocusParam} + class="search-input" + name="search" + aria-label="Run search in the documentation" + autocomplete="off" + spellcheck="false" + placeholder="Type ‘S’ or ‘/’ to search, ‘?’ for more options…" + type="search"> + </form> + </nav><div class="search-switcher"></div>`; + out.insertBefore(hdr, window.searchState.outputElement()); + el = document.getElementsByClassName("search-input")[0]; + } + if (el instanceof HTMLInputElement) { + return el; + } + return null; + }, + containerElement: () => { let el = document.getElementById("search"); if (!el) { el = document.createElement("section"); @@ -285,6 +286,19 @@ function preLoadCss(cssUrl) { } return el; }, + outputElement: () => { + const container = window.searchState.containerElement(); + if (!container) { + return null; + } + let el = container.querySelector(".search-out"); + if (!el) { + el = document.createElement("div"); + el.className = "search-out"; + container.appendChild(el); + } + return el; + }, title: document.title, titleBeforeSearch: document.title, timeout: null, @@ -303,25 +317,52 @@ function preLoadCss(cssUrl) { } }, isDisplayed: () => { - const outputElement = window.searchState.outputElement(); - return !!outputElement && - !!outputElement.parentElement && - outputElement.parentElement.id === ALTERNATIVE_DISPLAY_ID; + const container = window.searchState.containerElement(); + if (!container) { + return false; + } + return !!container.parentElement && container.parentElement.id === + ALTERNATIVE_DISPLAY_ID; }, // Sets the focus on the search bar at the top of the page focus: () => { - window.searchState.input && window.searchState.input.focus(); + const inputElement = window.searchState.inputElement(); + window.searchState.showResults(); + if (inputElement) { + inputElement.focus(); + // Avoid glitch if something focuses the search button after clicking. + requestAnimationFrame(() => inputElement.focus()); + } }, // Removes the focus from the search bar. defocus: () => { - window.searchState.input && window.searchState.input.blur(); + nonnull(window.searchState.inputElement()).blur(); }, - showResults: search => { - if (search === null || typeof search === "undefined") { - search = window.searchState.outputElement(); + toggle: () => { + if (window.searchState.isDisplayed()) { + window.searchState.defocus(); + window.searchState.hideResults(); + } else { + window.searchState.focus(); } - switchDisplayedElement(search); + }, + showResults: () => { document.title = window.searchState.title; + if (window.searchState.isDisplayed()) { + return; + } + const search = window.searchState.containerElement(); + switchDisplayedElement(search); + const btn = document.querySelector("#search-button a"); + if (browserSupportsHistoryApi() && btn instanceof HTMLAnchorElement && + window.searchState.getQueryStringParams().search === undefined + ) { + history.pushState(null, "", btn.href); + } + const btnLabel = document.querySelector("#search-button a span.label"); + if (btnLabel) { + btnLabel.innerHTML = "Exit"; + } }, removeQueryParameters: () => { // We change the document title. @@ -334,6 +375,10 @@ function preLoadCss(cssUrl) { switchDisplayedElement(null); // We also remove the query parameter from the URL. window.searchState.removeQueryParameters(); + const btnLabel = document.querySelector("#search-button a span.label"); + if (btnLabel) { + btnLabel.innerHTML = "Search"; + } }, getQueryStringParams: () => { /** @type {Object.<any, string>} */ @@ -348,11 +393,11 @@ function preLoadCss(cssUrl) { return params; }, setup: () => { - const search_input = window.searchState.input; + let searchLoaded = false; + const search_input = window.searchState.inputElement(); if (!search_input) { return; } - let searchLoaded = false; // If you're browsing the nightly docs, the page might need to be refreshed for the // search to work because the hash of the JS scripts might have changed. function sendSearchForm() { @@ -363,21 +408,102 @@ function preLoadCss(cssUrl) { if (!searchLoaded) { searchLoaded = true; // @ts-expect-error - loadScript(getVar("static-root-path") + getVar("search-js"), sendSearchForm); - loadScript(resourcePath("search-index", ".js"), sendSearchForm); + window.rr_ = data => { + // @ts-expect-error + window.searchIndex = data; + }; + if (!window.StringdexOnload) { + window.StringdexOnload = []; + } + window.StringdexOnload.push(() => { + loadScript( + // @ts-expect-error + getVar("static-root-path") + getVar("search-js"), + sendSearchForm, + ); + }); + // @ts-expect-error + loadScript(getVar("static-root-path") + getVar("stringdex-js"), sendSearchForm); + loadScript(resourcePath("search.index/root", ".js"), sendSearchForm); } } search_input.addEventListener("focus", () => { - window.searchState.origPlaceholder = search_input.placeholder; - search_input.placeholder = "Type your search here."; loadSearch(); }); - if (search_input.value !== "") { - loadSearch(); + const btn = document.getElementById("search-button"); + if (btn) { + btn.onclick = event => { + if (event.ctrlKey || event.altKey || event.metaKey) { + return; + } + event.preventDefault(); + window.searchState.toggle(); + loadSearch(); + }; + } + + // Push and pop states are used to add search results to the browser + // history. + if (browserSupportsHistoryApi()) { + // Store the previous <title> so we can revert back to it later. + const previousTitle = document.title; + + window.addEventListener("popstate", e => { + const params = window.searchState.getQueryStringParams(); + // Revert to the previous title manually since the History + // API ignores the title parameter. + document.title = previousTitle; + // Synchronize search bar with query string state and + // perform the search. This will empty the bar if there's + // nothing there, which lets you really go back to a + // previous state with nothing in the bar. + const inputElement = window.searchState.inputElement(); + if (params.search !== undefined && inputElement !== null) { + loadSearch(); + inputElement.value = params.search; + // Some browsers fire "onpopstate" for every page load + // (Chrome), while others fire the event only when actually + // popping a state (Firefox), which is why search() is + // called both here and at the end of the startSearch() + // function. + e.preventDefault(); + window.searchState.showResults(); + if (params.search === "") { + window.searchState.focus(); + } + } else { + // When browsing back from search results the main page + // visibility must be reset. + window.searchState.hideResults(); + } + }); } + // This is required in firefox to avoid this problem: Navigating to a search result + // with the keyboard, hitting enter, and then hitting back would take you back to + // the doc page, rather than the search that should overlay it. + // This was an interaction between the back-forward cache and our handlers + // that try to sync state between the URL and the search input. To work around it, + // do a small amount of re-init on page show. + window.onpageshow = () => { + const inputElement = window.searchState.inputElement(); + const qSearch = window.searchState.getQueryStringParams().search; + if (qSearch !== undefined && inputElement !== null) { + if (inputElement.value === "") { + inputElement.value = qSearch; + } + window.searchState.showResults(); + if (qSearch === "") { + loadSearch(); + window.searchState.focus(); + } + } else { + window.searchState.hideResults(); + } + }; + const params = window.searchState.getQueryStringParams(); if (params.search !== undefined) { window.searchState.setLoadingSearch(); @@ -386,13 +512,9 @@ function preLoadCss(cssUrl) { }, setLoadingSearch: () => { const search = window.searchState.outputElement(); - if (!search) { - return; - } - search.innerHTML = "<h3 class=\"search-loading\">" + - window.searchState.loadingText + - "</h3>"; - window.searchState.showResults(search); + nonnull(search).innerHTML = "<h3 class=\"search-loading\">" + + window.searchState.loadingText + "</h3>"; + window.searchState.showResults(); }, descShards: new Map(), loadDesc: async function({descShard, descIndex}) { @@ -1500,15 +1622,13 @@ function preLoadCss(cssUrl) { // @ts-expect-error function helpBlurHandler(event) { - // @ts-expect-error - if (!getHelpButton().contains(document.activeElement) && - // @ts-expect-error - !getHelpButton().contains(event.relatedTarget) && - // @ts-expect-error - !getSettingsButton().contains(document.activeElement) && - // @ts-expect-error - !getSettingsButton().contains(event.relatedTarget) - ) { + const isInPopover = onEachLazy( + document.querySelectorAll(".settings-menu, .help-menu"), + menu => { + return menu.contains(document.activeElement) || menu.contains(event.relatedTarget); + }, + ); + if (!isInPopover) { window.hidePopoverMenus(); } } @@ -1529,7 +1649,7 @@ function preLoadCss(cssUrl) { ["⏎", "Go to active search result"], ["+", "Expand all sections"], ["-", "Collapse all sections"], - // for the sake of brevity, we don't say "inherint impl blocks", + // for the sake of brevity, we don't say "inherit impl blocks", // although that would be more correct, // since trait impl blocks are collapsed by - ["_", "Collapse all sections, including impl blocks"], @@ -1571,10 +1691,9 @@ function preLoadCss(cssUrl) { const container = document.createElement("div"); if (!isHelpPage) { - container.className = "popover"; + container.className = "popover content"; } container.id = "help"; - container.style.display = "none"; const side_by_side = document.createElement("div"); side_by_side.className = "side-by-side"; @@ -1590,17 +1709,16 @@ function preLoadCss(cssUrl) { help_section.appendChild(container); // @ts-expect-error document.getElementById("main-content").appendChild(help_section); - container.style.display = "block"; } else { - const help_button = getHelpButton(); - // @ts-expect-error - help_button.appendChild(container); - - container.onblur = helpBlurHandler; - // @ts-expect-error - help_button.onblur = helpBlurHandler; - // @ts-expect-error - help_button.children[0].onblur = helpBlurHandler; + onEachLazy(document.getElementsByClassName("help-menu"), menu => { + if (menu.offsetWidth !== 0) { + menu.appendChild(container); + container.onblur = helpBlurHandler; + menu.onblur = helpBlurHandler; + menu.children[0].onblur = helpBlurHandler; + return true; + } + }); } return container; @@ -1621,80 +1739,57 @@ function preLoadCss(cssUrl) { * Hide all the popover menus. */ window.hidePopoverMenus = () => { - onEachLazy(document.querySelectorAll("rustdoc-toolbar .popover"), elem => { + onEachLazy(document.querySelectorAll(".settings-menu .popover"), elem => { elem.style.display = "none"; }); - const button = getHelpButton(); - if (button) { - removeClass(button, "help-open"); - } + onEachLazy(document.querySelectorAll(".help-menu .popover"), elem => { + elem.parentElement.removeChild(elem); + }); }; /** - * Returns the help menu element (not the button). - * - * @param {boolean} buildNeeded - If this argument is `false`, the help menu element won't be - * built if it doesn't exist. - * - * @return {HTMLElement} - */ - function getHelpMenu(buildNeeded) { - // @ts-expect-error - let menu = getHelpButton().querySelector(".popover"); - if (!menu && buildNeeded) { - menu = buildHelpMenu(); - } - // @ts-expect-error - return menu; - } - - /** * Show the help popup menu. */ function showHelp() { + window.hideAllModals(false); // Prevent `blur` events from being dispatched as a result of closing // other modals. - const button = getHelpButton(); - addClass(button, "help-open"); - // @ts-expect-error - button.querySelector("a").focus(); - const menu = getHelpMenu(true); - if (menu.style.display === "none") { - // @ts-expect-error - window.hideAllModals(); - menu.style.display = ""; - } + onEachLazy(document.querySelectorAll(".help-menu a"), menu => { + if (menu.offsetWidth !== 0) { + menu.focus(); + return true; + } + }); + buildHelpMenu(); } - const helpLink = document.querySelector(`#${HELP_BUTTON_ID} > a`); if (isHelpPage) { buildHelpMenu(); - } else if (helpLink) { - helpLink.addEventListener("click", event => { - // By default, have help button open docs in a popover. - // If user clicks with a moderator, though, use default browser behavior, - // probably opening in a new window or tab. - if (!helpLink.contains(helpLink) || - // @ts-expect-error - event.ctrlKey || - // @ts-expect-error - event.altKey || - // @ts-expect-error - event.metaKey) { - return; - } - event.preventDefault(); - const menu = getHelpMenu(true); - const shouldShowHelp = menu.style.display === "none"; - if (shouldShowHelp) { - showHelp(); - } else { - window.hidePopoverMenus(); - } + } else { + onEachLazy(document.querySelectorAll(".help-menu > a"), helpLink => { + helpLink.addEventListener( + "click", + /** @param {MouseEvent} event */ + event => { + // By default, have help button open docs in a popover. + // If user clicks with a moderator, though, use default browser behavior, + // probably opening in a new window or tab. + if (event.ctrlKey || + event.altKey || + event.metaKey) { + return; + } + event.preventDefault(); + if (document.getElementById("help")) { + window.hidePopoverMenus(); + } else { + showHelp(); + } + }, + ); }); } - setMobileTopbar(); addSidebarItems(); addSidebarCrates(); onHashChange(null); @@ -1746,7 +1841,15 @@ function preLoadCss(cssUrl) { // On larger, "desktop-sized" viewports (though that includes many // tablets), it's fixed-position, appears in the left side margin, // and it can be activated by resizing the sidebar into nothing. - const sidebarButton = document.getElementById("sidebar-button"); + let sidebarButton = document.getElementById("sidebar-button"); + const body = document.querySelector(".main-heading"); + if (!sidebarButton && body) { + sidebarButton = document.createElement("div"); + sidebarButton.id = "sidebar-button"; + const path = `${window.rootPath}${window.currentCrate}/all.html`; + sidebarButton.innerHTML = `<a href="${path}" title="show sidebar"></a>`; + body.insertBefore(sidebarButton, body.firstChild); + } if (sidebarButton) { sidebarButton.addEventListener("click", e => { removeClass(document.documentElement, "hide-sidebar"); diff --git a/src/librustdoc/html/static/js/rustdoc.d.ts b/src/librustdoc/html/static/js/rustdoc.d.ts index 3d30a7adb98..56581aebf06 100644 --- a/src/librustdoc/html/static/js/rustdoc.d.ts +++ b/src/librustdoc/html/static/js/rustdoc.d.ts @@ -2,6 +2,8 @@ // not put into the JavaScript we include as part of the documentation. It is used for // type checking. See README.md in this directory for more info. +import { RoaringBitmap } from "./stringdex"; + /* eslint-disable */ declare global { /** Search engine data used by main.js and search.js */ @@ -10,6 +12,17 @@ declare global { declare function nonnull(x: T|null, msg: string|undefined); /** Defined and documented in `storage.js` */ declare function nonundef(x: T|undefined, msg: string|undefined); + interface PromiseConstructor { + /** + * Polyfill + * @template T + */ + withResolvers: function(): { + "promise": Promise<T>, + "resolve": (function(T): void), + "reject": (function(any): void) + }; + } interface Window { /** Make the current theme easy to find */ currentTheme: HTMLLinkElement|null; @@ -95,29 +108,28 @@ declare namespace rustdoc { interface SearchState { rustdocToolbar: HTMLElement|null; loadingText: string; - input: HTMLInputElement|null; + inputElement: function(): HTMLInputElement|null; + containerElement: function(): Element|null; title: string; titleBeforeSearch: string; - timeout: number|null; + timeout: ReturnType<typeof setTimeout>|null; currentTab: number; - focusedByTab: [number|null, number|null, number|null]; + focusedByTab: [Element|null, Element|null, Element|null]; clearInputTimeout: function; - outputElement(): HTMLElement|null; - focus(); - defocus(); - // note: an optional param is not the same as - // a nullable/undef-able param. - showResults(elem?: HTMLElement|null); - removeQueryParameters(); - hideResults(); - getQueryStringParams(): Object.<any, string>; - origPlaceholder: string; + outputElement: function(): Element|null; + focus: function(); + defocus: function(); + toggle: function(); + showResults: function(); + removeQueryParameters: function(); + hideResults: function(); + getQueryStringParams: function(): Object.<any, string>; setup: function(); setLoadingSearch(); descShards: Map<string, SearchDescShard[]>; loadDesc: function({descShard: SearchDescShard, descIndex: number}): Promise<string|null>; - loadedDescShard(string, number, string); - isDisplayed(): boolean, + loadedDescShard: function(string, number, string); + isDisplayed: function(): boolean; } interface SearchDescShard { @@ -131,12 +143,13 @@ declare namespace rustdoc { * A single parsed "atom" in a search query. For example, * * std::fmt::Formatter, Write -> Result<()> - * ┏━━━━━━━━━━━━━━━━━━ ┌──── ┏━━━━━┅┅┅┅┄┄┄┄┄┄┄┄┄┄┄┄┄┄┐ - * ┃ │ ┗ QueryElement { ┊ - * ┃ │ name: Result ┊ - * ┃ │ generics: [ ┊ - * ┃ │ QueryElement ┘ - * ┃ │ name: () + * ┏━━━━━━━━━━━━━━━━━━ ┌──── ┏━━━━━┅┅┅┅┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┐ + * ┃ │ ┗ QueryElement { ┊ + * ┃ │ name: Result ┊ + * ┃ │ generics: [ ┊ + * ┃ │ QueryElement { ┘ + * ┃ │ name: () + * ┃ │ } * ┃ │ ] * ┃ │ } * ┃ └ QueryElement { @@ -156,14 +169,14 @@ declare namespace rustdoc { normalizedPathLast: string, generics: Array<QueryElement>, bindings: Map<number, Array<QueryElement>>, - typeFilter: number|null, + typeFilter: number, } /** * Same as QueryElement, but bindings and typeFilter support strings */ interface ParserQueryElement { - name: string|null, + name: string, id: number|null, fullPath: Array<string>, pathWithoutLast: Array<string>, @@ -172,7 +185,7 @@ declare namespace rustdoc { generics: Array<ParserQueryElement>, bindings: Map<string, Array<ParserQueryElement>>, bindingName: {name: string|null, generics: ParserQueryElement[]}|null, - typeFilter: number|string|null, + typeFilter: string|null, } /** @@ -215,35 +228,74 @@ declare namespace rustdoc { /** * An entry in the search index database. */ + interface EntryData { + krate: number, + ty: ItemType, + modulePath: number?, + exactModulePath: number?, + parent: number?, + deprecated: boolean, + associatedItemDisambiguator: string?, + } + + /** + * A path in the search index database + */ + interface PathData { + ty: ItemType, + modulePath: string, + exactModulePath: string?, + } + + /** + * A function signature in the search index database + * + * Note that some non-function items (eg. constants, struct fields) have a function signature so they can appear in type-based search. + */ + interface FunctionData { + functionSignature: FunctionSearchType|null, + paramNames: string[], + elemCount: number, + } + + /** + * A function signature in the search index database + */ + interface TypeData { + searchUnbox: boolean, + invertedFunctionSignatureIndex: RoaringBitmap[], + } + + /** + * A search entry of some sort. + */ interface Row { - crate: string, - descShard: SearchDescShard, id: number, - // This is the name of the item. For doc aliases, if you want the name of the aliased - // item, take a look at `Row.original.name`. + crate: string, + ty: ItemType, name: string, normalizedName: string, - word: string, - paramNames: string[], - parent: ({ty: number, name: string, path: string, exactPath: string}|null|undefined), - path: string, - ty: number, - type: FunctionSearchType | null, - descIndex: number, - bitIndex: number, - implDisambiguator: String | null, - is_alias?: boolean, - original?: Row, + modulePath: string, + exactModulePath: string, + entry: EntryData?, + path: PathData?, + type: FunctionData?, + deprecated: boolean, + parent: { path: PathData, name: string}?, } + type ItemType = 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | + 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | + 21 | 22 | 23 | 24 | 25 | 26; + /** * The viewmodel for the search engine results page. */ interface ResultsTable { - in_args: Array<ResultObject>, - returned: Array<ResultObject>, - others: Array<ResultObject>, - query: ParsedQuery, + in_args: AsyncGenerator<ResultObject>, + returned: AsyncGenerator<ResultObject>, + others: AsyncGenerator<ResultObject>, + query: ParsedQuery<rustdoc.ParserQueryElement>, } type Results = { max_dist?: number } & Map<number, ResultObject> @@ -252,25 +304,41 @@ declare namespace rustdoc { * An annotated `Row`, used in the viewmodel. */ interface ResultObject { - desc: string, + desc: Promise<string|null>, displayPath: string, fullPath: string, href: string, id: number, dist: number, path_dist: number, - name: string, - normalizedName: string, - word: string, index: number, - parent: (Object|undefined), - path: string, - ty: number, + parent: ({ + path: string, + exactPath: string, + name: string, + ty: number, + }|undefined), type?: FunctionSearchType, paramNames?: string[], displayTypeSignature: Promise<rustdoc.DisplayTypeSignature> | null, item: Row, - dontValidate?: boolean, + is_alias: boolean, + alias?: string, + } + + /** + * An annotated `Row`, used in the viewmodel. + */ + interface PlainResultObject { + id: number, + dist: number, + path_dist: number, + index: number, + elems: rustdoc.QueryElement[], + returned: rustdoc.QueryElement[], + is_alias: boolean, + alias?: string, + original?: rustdoc.Rlow, } /** @@ -364,7 +432,19 @@ declare namespace rustdoc { * Numeric IDs are *ONE-indexed* into the paths array (`p`). Zero is used as a sentinel for `null` * because `null` is four bytes while `0` is one byte. */ - type RawFunctionType = number | [number, Array<RawFunctionType>]; + type RawFunctionType = number | [number, Array<RawFunctionType>] | [number, Array<RawFunctionType>, Array<[RawFunctionType, RawFunctionType[]]>]; + + /** + * Utility typedef for deserializing compact JSON. + * + * R is the required part, O is the optional part, which goes afterward. + * For example, `ArrayWithOptionals<[A, B], [C, D]>` matches + * `[A, B] | [A, B, C] | [A, B, C, D]`. + */ + type ArrayWithOptionals<R extends any[], O extends any[]> = + O extends [infer First, ...infer Rest] ? + R | ArrayWithOptionals<[...R, First], Rest> : + R; /** * The type signature entry in the decoded search index. @@ -382,8 +462,8 @@ declare namespace rustdoc { */ interface FunctionType { id: null|number, - ty: number|null, - name?: string, + ty: ItemType, + name: string|null, path: string|null, exactPath: string|null, unboxFlag: boolean, @@ -403,70 +483,6 @@ declare namespace rustdoc { bindings: Map<number, FingerprintableType[]>; }; - /** - * The raw search data for a given crate. `n`, `t`, `d`, `i`, and `f` - * are arrays with the same length. `q`, `a`, and `c` use a sparse - * representation for compactness. - * - * `n[i]` contains the name of an item. - * - * `t[i]` contains the type of that item - * (as a string of characters that represent an offset in `itemTypes`). - * - * `d[i]` contains the description of that item. - * - * `q` contains the full paths of the items. For compactness, it is a set of - * (index, path) pairs used to create a map. If a given index `i` is - * not present, this indicates "same as the last index present". - * - * `i[i]` contains an item's parent, usually a module. For compactness, - * it is a set of indexes into the `p` array. - * - * `f` contains function signatures, or `0` if the item isn't a function. - * More information on how they're encoded can be found in rustc-dev-guide - * - * Functions are themselves encoded as arrays. The first item is a list of - * types representing the function's inputs, and the second list item is a list - * of types representing the function's output. Tuples are flattened. - * Types are also represented as arrays; the first item is an index into the `p` - * array, while the second is a list of types representing any generic parameters. - * - * b[i] contains an item's impl disambiguator. This is only present if an item - * is defined in an impl block and, the impl block's type has more than one associated - * item with the same name. - * - * `a` defines aliases with an Array of pairs: [name, offset], where `offset` - * points into the n/t/d/q/i/f arrays. - * - * `doc` contains the description of the crate. - * - * `p` is a list of path/type pairs. It is used for parents and function parameters. - * The first item is the type, the second is the name, the third is the visible path (if any) and - * the fourth is the canonical path used for deduplication (if any). - * - * `r` is the canonical path used for deduplication of re-exported items. - * It is not used for associated items like methods (that's the fourth element - * of `p`) but is used for modules items like free functions. - * - * `c` is an array of item indices that are deprecated. - */ - type RawSearchIndexCrate = { - doc: string, - a: { [key: string]: number[] }, - n: Array<string>, - t: string, - D: string, - e: string, - q: Array<[number, string]>, - i: string, - f: string, - p: Array<[number, string] | [number, string, number] | [number, string, number, number] | [number, string, number, number, string]>, - b: Array<[number, String]>, - c: string, - r: Array<[number, number]>, - P: Array<[number, string]>, - }; - type VlqData = VlqData[] | number; /** diff --git a/src/librustdoc/html/static/js/search.js b/src/librustdoc/html/static/js/search.js index 505652c0f4a..d55208150b8 100644 --- a/src/librustdoc/html/static/js/search.js +++ b/src/librustdoc/html/static/js/search.js @@ -1,9 +1,16 @@ // ignore-tidy-filelength -/* global addClass, getNakedUrl, getSettingValue, getVar */ -/* global onEachLazy, removeClass, searchState, browserSupportsHistoryApi, exports */ +/* global addClass, getNakedUrl, getVar, nonnull, getSettingValue */ +/* global onEachLazy, removeClass, searchState, browserSupportsHistoryApi */ "use strict"; +/** + * @param {stringdex.Stringdex} Stringdex + * @param {typeof stringdex.RoaringBitmap} RoaringBitmap + * @param {stringdex.Hooks} hooks + */ +const initSearch = async function(Stringdex, RoaringBitmap, hooks) { + // polyfill // https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/toSpliced if (!Array.prototype.toSpliced) { @@ -20,31 +27,65 @@ if (!Array.prototype.toSpliced) { * * @template T * @param {Iterable<T>} arr - * @param {function(T): any} func + * @param {function(T): Promise<any>} func * @param {function(T): boolean} funcBtwn */ -function onEachBtwn(arr, func, funcBtwn) { +async function onEachBtwnAsync(arr, func, funcBtwn) { let skipped = true; for (const value of arr) { if (!skipped) { funcBtwn(value); } - skipped = func(value); + skipped = await func(value); } } /** - * Convert any `undefined` to `null`. - * - * @template T - * @param {T|undefined} x - * @returns {T|null} + * Allow the browser to redraw. + * @returns {Promise<void>} */ -function undef2null(x) { - if (x !== undefined) { - return x; - } - return null; +const yieldToBrowser = typeof window !== "undefined" && window.requestIdleCallback ? + function() { + return new Promise((resolve, _reject) => { + window.requestIdleCallback(resolve); + }); + } : + function() { + return new Promise((resolve, _reject) => { + setTimeout(resolve, 0); + }); + }; + +/** + * Promise-based timer wrapper. + * @param {number} ms + * @returns {Promise<void>} + */ +const timeout = function(ms) { + return new Promise((resolve, _reject) => { + setTimeout(resolve, ms); + }); +}; + +if (!Promise.withResolvers) { + /** + * Polyfill + * @template T + * @returns {{ + "promise": Promise<T>, + "resolve": (function(T): void), + "reject": (function(any): void) + }} + */ + Promise.withResolvers = () => { + let resolve, reject; + const promise = new Promise((res, rej) => { + resolve = res; + reject = rej; + }); + // @ts-expect-error + return {promise, resolve, reject}; + }; } // ==================== Core search logic begin ==================== @@ -81,13 +122,22 @@ const itemTypes = [ ]; // used for special search precedence -const TY_PRIMITIVE = itemTypes.indexOf("primitive"); -const TY_GENERIC = itemTypes.indexOf("generic"); -const TY_IMPORT = itemTypes.indexOf("import"); -const TY_TRAIT = itemTypes.indexOf("trait"); -const TY_FN = itemTypes.indexOf("fn"); -const TY_METHOD = itemTypes.indexOf("method"); -const TY_TYMETHOD = itemTypes.indexOf("tymethod"); +/** @type {rustdoc.ItemType} */ +const TY_PRIMITIVE = 1; +/** @type {rustdoc.ItemType} */ +const TY_GENERIC = 26; +/** @type {rustdoc.ItemType} */ +const TY_IMPORT = 4; +/** @type {rustdoc.ItemType} */ +const TY_TRAIT = 10; +/** @type {rustdoc.ItemType} */ +const TY_FN = 7; +/** @type {rustdoc.ItemType} */ +const TY_METHOD = 13; +/** @type {rustdoc.ItemType} */ +const TY_TYMETHOD = 12; +/** @type {rustdoc.ItemType} */ +const TY_ASSOCTYPE = 17; const ROOT_PATH = typeof window !== "undefined" ? window.rootPath : "../"; // Hard limit on how deep to recurse into generics when doing type-driven search. @@ -242,7 +292,9 @@ function isEndCharacter(c) { } /** - * @param {number} ty + * Same thing as ItemType::is_fn_like in item_type.rs + * + * @param {rustdoc.ItemType} ty * @returns */ function isFnLikeTy(ty) { @@ -1023,6 +1075,7 @@ class VlqHexDecoder { this.string = string; this.cons = cons; this.offset = 0; + this.elemCount = 0; /** @type {T[]} */ this.backrefQueue = []; } @@ -1060,6 +1113,7 @@ class VlqHexDecoder { n = (n << 4) | (c & 0xF); const [sign, value] = [n & 1, n >> 1]; this.offset += 1; + this.elemCount += 1; return sign ? -value : value; } /** @@ -1086,1247 +1140,138 @@ class VlqHexDecoder { return result; } } -class RoaringBitmap { - /** @param {string} str */ - constructor(str) { - // https://github.com/RoaringBitmap/RoaringFormatSpec - // - // Roaring bitmaps are used for flags that can be kept in their - // compressed form, even when loaded into memory. This decoder - // turns the containers into objects, but uses byte array - // slices of the original format for the data payload. - const strdecoded = atob(str); - const u8array = new Uint8Array(strdecoded.length); - for (let j = 0; j < strdecoded.length; ++j) { - u8array[j] = strdecoded.charCodeAt(j); - } - const has_runs = u8array[0] === 0x3b; - const size = has_runs ? - ((u8array[2] | (u8array[3] << 8)) + 1) : - ((u8array[4] | (u8array[5] << 8) | (u8array[6] << 16) | (u8array[7] << 24))); - let i = has_runs ? 4 : 8; - let is_run; - if (has_runs) { - const is_run_len = Math.floor((size + 7) / 8); - is_run = u8array.slice(i, i + is_run_len); - i += is_run_len; - } else { - is_run = new Uint8Array(); - } - this.keys = []; - this.cardinalities = []; - for (let j = 0; j < size; ++j) { - this.keys.push(u8array[i] | (u8array[i + 1] << 8)); - i += 2; - this.cardinalities.push((u8array[i] | (u8array[i + 1] << 8)) + 1); - i += 2; - } - this.containers = []; - let offsets = null; - if (!has_runs || this.keys.length >= 4) { - offsets = []; - for (let j = 0; j < size; ++j) { - offsets.push(u8array[i] | (u8array[i + 1] << 8) | (u8array[i + 2] << 16) | - (u8array[i + 3] << 24)); - i += 4; - } - } - for (let j = 0; j < size; ++j) { - if (offsets && offsets[j] !== i) { - // eslint-disable-next-line no-console - console.log(this.containers); - throw new Error(`corrupt bitmap ${j}: ${i} / ${offsets[j]}`); - } - if (is_run[j >> 3] & (1 << (j & 0x7))) { - const runcount = (u8array[i] | (u8array[i + 1] << 8)); - i += 2; - this.containers.push(new RoaringBitmapRun( - runcount, - u8array.slice(i, i + (runcount * 4)), - )); - i += runcount * 4; - } else if (this.cardinalities[j] >= 4096) { - this.containers.push(new RoaringBitmapBits(u8array.slice(i, i + 8192))); - i += 8192; - } else { - const end = this.cardinalities[j] * 2; - this.containers.push(new RoaringBitmapArray( - this.cardinalities[j], - u8array.slice(i, i + end), - )); - i += end; - } - } - } - /** @param {number} keyvalue */ - contains(keyvalue) { - const key = keyvalue >> 16; - const value = keyvalue & 0xFFFF; - // Binary search algorithm copied from - // https://en.wikipedia.org/wiki/Binary_search#Procedure - // - // Format is required by specification to be sorted. - // Because keys are 16 bits and unique, length can't be - // bigger than 2**16, and because we have 32 bits of safe int, - // left + right can't overflow. - let left = 0; - let right = this.keys.length - 1; - while (left <= right) { - const mid = Math.floor((left + right) / 2); - const x = this.keys[mid]; - if (x < key) { - left = mid + 1; - } else if (x > key) { - right = mid - 1; - } else { - return this.containers[mid].contains(value); - } - } - return false; - } -} -class RoaringBitmapRun { - /** - * @param {number} runcount - * @param {Uint8Array} array - */ - constructor(runcount, array) { - this.runcount = runcount; - this.array = array; - } - /** @param {number} value */ - contains(value) { - // Binary search algorithm copied from - // https://en.wikipedia.org/wiki/Binary_search#Procedure - // - // Since runcount is stored as 16 bits, left + right - // can't overflow. - let left = 0; - let right = this.runcount - 1; - while (left <= right) { - const mid = Math.floor((left + right) / 2); - const i = mid * 4; - const start = this.array[i] | (this.array[i + 1] << 8); - const lenm1 = this.array[i + 2] | (this.array[i + 3] << 8); - if ((start + lenm1) < value) { - left = mid + 1; - } else if (start > value) { - right = mid - 1; - } else { - return true; - } - } - return false; - } -} -class RoaringBitmapArray { - /** - * @param {number} cardinality - * @param {Uint8Array} array - */ - constructor(cardinality, array) { - this.cardinality = cardinality; - this.array = array; - } - /** @param {number} value */ - contains(value) { - // Binary search algorithm copied from - // https://en.wikipedia.org/wiki/Binary_search#Procedure - // - // Since cardinality can't be higher than 4096, left + right - // cannot overflow. - let left = 0; - let right = this.cardinality - 1; - while (left <= right) { - const mid = Math.floor((left + right) / 2); - const i = mid * 2; - const x = this.array[i] | (this.array[i + 1] << 8); - if (x < value) { - left = mid + 1; - } else if (x > value) { - right = mid - 1; - } else { - return true; - } - } - return false; - } -} -class RoaringBitmapBits { - /** - * @param {Uint8Array} array - */ - constructor(array) { - this.array = array; - } - /** @param {number} value */ - contains(value) { - return !!(this.array[value >> 3] & (1 << (value & 7))); - } -} +/** @type {Array<string>} */ +const EMPTY_STRING_ARRAY = []; + +/** @type {Array<rustdoc.FunctionType>} */ +const EMPTY_GENERICS_ARRAY = []; + +/** @type {Array<[number, rustdoc.FunctionType[]]>} */ +const EMPTY_BINDINGS_ARRAY = []; + +/** @type {Map<number, Array<any>>} */ +const EMPTY_BINDINGS_MAP = new Map(); /** - * A prefix tree, used for name-based search. - * - * This data structure is used to drive prefix matches, - * such as matching the query "link" to `LinkedList`, - * and Lev-distance matches, such as matching the - * query "hahsmap" to `HashMap`. Substring matches, - * such as "list" to `LinkedList`, are done with a - * tailTable that deep-links into this trie. - * - * children - * : A [sparse array] of subtrees. The array index - * is a charCode. - * - * [sparse array]: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/ - * Indexed_collections#sparse_arrays - * - * matches - * : A list of search index IDs for this node. - * - * @type {{ - * children: NameTrie[], - * matches: number[], - * }} + * @param {string|null} typename + * @returns {number} */ -class NameTrie { - constructor() { - this.children = []; - this.matches = []; +function itemTypeFromName(typename) { + if (typename === null) { + return NO_TYPE_FILTER; } - /** - * @param {string} name - * @param {number} id - * @param {Map<string, NameTrie[]>} tailTable - */ - insert(name, id, tailTable) { - this.insertSubstring(name, 0, id, tailTable); - } - /** - * @param {string} name - * @param {number} substart - * @param {number} id - * @param {Map<string, NameTrie[]>} tailTable - */ - insertSubstring(name, substart, id, tailTable) { - const l = name.length; - if (substart === l) { - this.matches.push(id); - } else { - const sb = name.charCodeAt(substart); - let child; - if (this.children[sb] !== undefined) { - child = this.children[sb]; - } else { - child = new NameTrie(); - this.children[sb] = child; - /** @type {NameTrie[]} */ - let sste; - if (substart >= 2) { - const tail = name.substring(substart - 2, substart + 1); - const entry = tailTable.get(tail); - if (entry !== undefined) { - sste = entry; - } else { - sste = []; - tailTable.set(tail, sste); - } - sste.push(child); - } - } - child.insertSubstring(name, substart + 1, id, tailTable); - } - } - /** - * @param {string} name - * @param {Map<string, NameTrie[]>} tailTable - */ - search(name, tailTable) { - const results = new Set(); - this.searchSubstringPrefix(name, 0, results); - if (results.size < MAX_RESULTS && name.length >= 3) { - const levParams = name.length >= 6 ? - new Lev2TParametricDescription(name.length) : - new Lev1TParametricDescription(name.length); - this.searchLev(name, 0, levParams, results); - const tail = name.substring(0, 3); - const list = tailTable.get(tail); - if (list !== undefined) { - for (const entry of list) { - entry.searchSubstringPrefix(name, 3, results); - } - } - } - return [...results]; - } - /** - * @param {string} name - * @param {number} substart - * @param {Set<number>} results - */ - searchSubstringPrefix(name, substart, results) { - const l = name.length; - if (substart === l) { - for (const match of this.matches) { - results.add(match); - } - // breadth-first traversal orders prefix matches by length - /** @type {NameTrie[]} */ - let unprocessedChildren = []; - for (const child of this.children) { - if (child) { - unprocessedChildren.push(child); - } - } - /** @type {NameTrie[]} */ - let nextSet = []; - while (unprocessedChildren.length !== 0) { - /** @type {NameTrie} */ - // @ts-expect-error - const next = unprocessedChildren.pop(); - for (const child of next.children) { - if (child) { - nextSet.push(child); - } - } - for (const match of next.matches) { - results.add(match); - } - if (unprocessedChildren.length === 0) { - const tmp = unprocessedChildren; - unprocessedChildren = nextSet; - nextSet = tmp; - } - } - } else { - const sb = name.charCodeAt(substart); - if (this.children[sb] !== undefined) { - this.children[sb].searchSubstringPrefix(name, substart + 1, results); - } - } - } - /** - * @param {string} name - * @param {number} substart - * @param {Lev2TParametricDescription|Lev1TParametricDescription} levParams - * @param {Set<number>} results - */ - searchLev(name, substart, levParams, results) { - const stack = [[this, 0]]; - const n = levParams.n; - while (stack.length !== 0) { - // It's not empty - //@ts-expect-error - const [trie, levState] = stack.pop(); - for (const [charCode, child] of trie.children.entries()) { - if (!child) { - continue; - } - const levPos = levParams.getPosition(levState); - const vector = levParams.getVector( - name, - charCode, - levPos, - Math.min(name.length, levPos + (2 * n) + 1), - ); - const newLevState = levParams.transition( - levState, - levPos, - vector, - ); - if (newLevState >= 0) { - stack.push([child, newLevState]); - if (levParams.isAccept(newLevState)) { - for (const match of child.matches) { - results.add(match); - } - } - } - } - } + const index = itemTypes.findIndex(i => i === typename); + if (index < 0) { + throw ["Unknown type filter ", typename]; } + return index; } class DocSearch { /** - * @param {Map<string, rustdoc.RawSearchIndexCrate>} rawSearchIndex * @param {string} rootPath - * @param {rustdoc.SearchState} searchState + * @param {stringdex.Database} database */ - constructor(rawSearchIndex, rootPath, searchState) { - /** - * @type {Map<String, RoaringBitmap>} - */ - this.searchIndexDeprecated = new Map(); - /** - * @type {Map<String, RoaringBitmap>} - */ - this.searchIndexEmptyDesc = new Map(); - /** - * @type {Uint32Array} - */ - this.functionTypeFingerprint = new Uint32Array(0); - /** - * Map from normalized type names to integers. Used to make type search - * more efficient. - * - * @type {Map<string, {id: number, assocOnly: boolean}>} - */ - this.typeNameIdMap = new Map(); - /** - * Map from type ID to associated type name. Used for display, - * not for search. - * - * @type {Map<number, string>} - */ - this.assocTypeIdNameMap = new Map(); - this.ALIASES = new Map(); - this.FOUND_ALIASES = new Set(); + constructor(rootPath, database) { this.rootPath = rootPath; - this.searchState = searchState; + this.database = database; - /** - * Special type name IDs for searching by array. - * @type {number} - */ - this.typeNameIdOfArray = this.buildTypeMapIndex("array"); - /** - * Special type name IDs for searching by slice. - * @type {number} - */ - this.typeNameIdOfSlice = this.buildTypeMapIndex("slice"); - /** - * Special type name IDs for searching by both array and slice (`[]` syntax). - * @type {number} - */ - this.typeNameIdOfArrayOrSlice = this.buildTypeMapIndex("[]"); - /** - * Special type name IDs for searching by tuple. - * @type {number} - */ - this.typeNameIdOfTuple = this.buildTypeMapIndex("tuple"); - /** - * Special type name IDs for searching by unit. - * @type {number} - */ - this.typeNameIdOfUnit = this.buildTypeMapIndex("unit"); - /** - * Special type name IDs for searching by both tuple and unit (`()` syntax). - * @type {number} - */ - this.typeNameIdOfTupleOrUnit = this.buildTypeMapIndex("()"); - /** - * Special type name IDs for searching `fn`. - * @type {number} - */ - this.typeNameIdOfFn = this.buildTypeMapIndex("fn"); - /** - * Special type name IDs for searching `fnmut`. - * @type {number} - */ - this.typeNameIdOfFnMut = this.buildTypeMapIndex("fnmut"); - /** - * Special type name IDs for searching `fnonce`. - * @type {number} - */ - this.typeNameIdOfFnOnce = this.buildTypeMapIndex("fnonce"); - /** - * Special type name IDs for searching higher order functions (`->` syntax). - * @type {number} - */ - this.typeNameIdOfHof = this.buildTypeMapIndex("->"); - /** - * Special type name IDs the output assoc type. - * @type {number} - */ - this.typeNameIdOfOutput = this.buildTypeMapIndex("output", true); - /** - * Special type name IDs for searching by reference. - * @type {number} - */ - this.typeNameIdOfReference = this.buildTypeMapIndex("reference"); + this.typeNameIdOfOutput = -1; + this.typeNameIdOfArray = -1; + this.typeNameIdOfSlice = -1; + this.typeNameIdOfArrayOrSlice = -1; + this.typeNameIdOfTuple = -1; + this.typeNameIdOfUnit = -1; + this.typeNameIdOfTupleOrUnit = -1; + this.typeNameIdOfReference = -1; + this.typeNameIdOfHof = -1; - /** - * Empty, immutable map used in item search types with no bindings. - * - * @type {Map<number, Array<any>>} - */ - this.EMPTY_BINDINGS_MAP = new Map(); + this.utf8decoder = new TextDecoder(); - /** - * Empty, immutable map used in item search types with no bindings. - * - * @type {Array<any>} - */ - this.EMPTY_GENERICS_ARRAY = []; - - /** - * Object pool for function types with no bindings or generics. - * This is reset after loading the index. - * - * @type {Map<number|null, rustdoc.FunctionType>} - */ + /** @type {Map<number|null, rustdoc.FunctionType>} */ this.TYPES_POOL = new Map(); - - /** - * A trie for finding items by name. - * This is used for edit distance and prefix finding. - * - * @type {NameTrie} - */ - this.nameTrie = new NameTrie(); - - /** - * Find items by 3-substring. This is a map from three-char - * prefixes into lists of subtries. - */ - this.tailTable = new Map(); - - /** - * @type {Array<rustdoc.Row>} - */ - this.searchIndex = this.buildIndex(rawSearchIndex); } /** - * Add an item to the type Name->ID map, or, if one already exists, use it. - * Returns the number. If name is "" or null, return null (pure generic). - * - * This is effectively string interning, so that function matching can be - * done more quickly. Two types with the same name but different item kinds - * get the same ID. - * - * @template T extends string - * @overload - * @param {T} name - * @param {boolean=} isAssocType - True if this is an assoc type - * @returns {T extends "" ? null : number} - * - * @param {string} name - * @param {boolean=} isAssocType - * @returns {number | null} - * + * Load search index. If you do not call this function, `execQuery` + * will never fulfill. */ - buildTypeMapIndex(name, isAssocType) { - if (name === "" || name === null) { - return null; - } - - const obj = this.typeNameIdMap.get(name); - if (obj !== undefined) { - obj.assocOnly = !!(isAssocType && obj.assocOnly); - return obj.id; - } else { - const id = this.typeNameIdMap.size; - this.typeNameIdMap.set(name, { id, assocOnly: !!isAssocType }); - return id; - } - } - - /** - * Convert a list of RawFunctionType / ID to object-based FunctionType. - * - * Crates often have lots of functions in them, and it's common to have a large number of - * functions that operate on a small set of data types, so the search index compresses them - * by encoding function parameter and return types as indexes into an array of names. - * - * Even when a general-purpose compression algorithm is used, this is still a win. - * I checked. https://github.com/rust-lang/rust/pull/98475#issue-1284395985 - * - * The format for individual function types is encoded in - * librustdoc/html/render/mod.rs: impl Serialize for RenderType - * - * @param {null|Array<rustdoc.RawFunctionType>} types - * @param {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean - * }>} paths - * @param {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean, - * }>} lowercasePaths - * - * @return {Array<rustdoc.FunctionType>} - */ - buildItemSearchTypeAll(types, paths, lowercasePaths) { - return types && types.length > 0 ? - types.map(type => this.buildItemSearchType(type, paths, lowercasePaths)) : - this.EMPTY_GENERICS_ARRAY; - } - - /** - * Converts a single type. - * - * @param {rustdoc.RawFunctionType} type - * @param {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean - * }>} paths - * @param {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean, - * }>} lowercasePaths - * @param {boolean=} isAssocType - */ - buildItemSearchType(type, paths, lowercasePaths, isAssocType) { - const PATH_INDEX_DATA = 0; - const GENERICS_DATA = 1; - const BINDINGS_DATA = 2; - let pathIndex, generics, bindings; - if (typeof type === "number") { - pathIndex = type; - generics = this.EMPTY_GENERICS_ARRAY; - bindings = this.EMPTY_BINDINGS_MAP; - } else { - pathIndex = type[PATH_INDEX_DATA]; - generics = this.buildItemSearchTypeAll( - type[GENERICS_DATA], - paths, - lowercasePaths, - ); - // @ts-expect-error - if (type.length > BINDINGS_DATA && type[BINDINGS_DATA].length > 0) { - // @ts-expect-error - bindings = new Map(type[BINDINGS_DATA].map(binding => { - const [assocType, constraints] = binding; - // Associated type constructors are represented sloppily in rustdoc's - // type search, to make the engine simpler. - // - // MyType<Output<T>=Result<T>> is equivalent to MyType<Output<Result<T>>=T> - // and both are, essentially - // MyType<Output=(T, Result<T>)>, except the tuple isn't actually there. - // It's more like the value of a type binding is naturally an array, - // which rustdoc calls "constraints". - // - // As a result, the key should never have generics on it. - return [ - this.buildItemSearchType(assocType, paths, lowercasePaths, true).id, - this.buildItemSearchTypeAll(constraints, paths, lowercasePaths), - ]; - })); - } else { - bindings = this.EMPTY_BINDINGS_MAP; - } + async buildIndex() { + const nn = this.database.getIndex("normalizedName"); + if (!nn) { + return; } + // Each of these identifiers are used specially by + // type-driven search. + const [ + // output is the special associated type that goes + // after the arrow: the type checker desugars + // the path `Fn(a) -> b` into `Fn<Output=b, (a)>` + output, + // fn, fnmut, and fnonce all match `->` + fn, + fnMut, + fnOnce, + hof, + // array and slice both match `[]` + array, + slice, + arrayOrSlice, + // tuple and unit both match `()` + tuple, + unit, + tupleOrUnit, + // reference matches `&` + reference, + // never matches `!` + never, + ] = await Promise.all([ + nn.search("output"), + nn.search("fn"), + nn.search("fnmut"), + nn.search("fnonce"), + nn.search("->"), + nn.search("array"), + nn.search("slice"), + nn.search("[]"), + nn.search("tuple"), + nn.search("unit"), + nn.search("()"), + nn.search("reference"), + nn.search("never"), + ]); /** - * @type {rustdoc.FunctionType} - */ - let result; - if (pathIndex < 0) { - // types less than 0 are generic parameters - // the actual names of generic parameters aren't stored, since they aren't API - result = { - id: pathIndex, - name: "", - ty: TY_GENERIC, - path: null, - exactPath: null, - generics, - bindings, - unboxFlag: true, - }; - } else if (pathIndex === 0) { - // `0` is used as a sentinel because it's fewer bytes than `null` - result = { - id: null, - name: "", - ty: null, - path: null, - exactPath: null, - generics, - bindings, - unboxFlag: true, - }; - } else { - const item = lowercasePaths[pathIndex - 1]; - const id = this.buildTypeMapIndex(item.name, isAssocType); - if (isAssocType && id !== null) { - this.assocTypeIdNameMap.set(id, paths[pathIndex - 1].name); - } - result = { - id, - name: paths[pathIndex - 1].name, - ty: item.ty, - path: item.path, - exactPath: item.exactPath, - generics, - bindings, - unboxFlag: item.unboxFlag, - }; - } - const cr = this.TYPES_POOL.get(result.id); - if (cr) { - // Shallow equality check. Since this function is used - // to construct every type object, this should be mostly - // equivalent to a deep equality check, except if there's - // a conflict, we don't keep the old one around, so it's - // not a fully precise implementation of hashcons. - if (cr.generics.length === result.generics.length && - cr.generics !== result.generics && - cr.generics.every((x, i) => result.generics[i] === x) - ) { - result.generics = cr.generics; - } - if (cr.bindings.size === result.bindings.size && cr.bindings !== result.bindings) { - let ok = true; - for (const [k, v] of cr.bindings.entries()) { - // @ts-expect-error - const v2 = result.bindings.get(v); - if (!v2) { - ok = false; - break; + * @param {stringdex.Trie|null|undefined} trie + * @param {rustdoc.ItemType} ty + * @param {string} modulePath + * @returns {Promise<number>} + * */ + const first = async(trie, ty, modulePath) => { + if (trie) { + for (const id of trie.matches().entries()) { + const pathData = await this.getPathData(id); + if (pathData && pathData.ty === ty && pathData.modulePath === modulePath) { + return id; } - if (v !== v2 && v.length === v2.length && v.every((x, i) => v2[i] === x)) { - result.bindings.set(k, v); - } else if (v !== v2) { - ok = false; - break; - } - } - if (ok) { - result.bindings = cr.bindings; } } - if (cr.ty === result.ty && cr.path === result.path - && cr.bindings === result.bindings && cr.generics === result.generics - && cr.ty === result.ty && cr.name === result.name - && cr.unboxFlag === result.unboxFlag - ) { - return cr; - } - } - this.TYPES_POOL.set(result.id, result); - return result; - } - - /** - * Type fingerprints allow fast, approximate matching of types. - * - * This algo creates a compact representation of the type set using a Bloom filter. - * This fingerprint is used three ways: - * - * - It accelerates the matching algorithm by checking the function fingerprint against the - * query fingerprint. If any bits are set in the query but not in the function, it can't - * match. - * - * - The fourth section has the number of items in the set. - * This is the distance function, used for filtering and for sorting. - * - * [^1]: Distance is the relatively naive metric of counting the number of distinct items in - * the function that are not present in the query. - * - * @param {rustdoc.FingerprintableType} type - a single type - * @param {Uint32Array} output - write the fingerprint to this data structure: uses 128 bits - */ - buildFunctionTypeFingerprint(type, output) { - let input = type.id; - // All forms of `[]`/`()`/`->` get collapsed down to one thing in the bloom filter. - // Differentiating between arrays and slices, if the user asks for it, is - // still done in the matching algorithm. - if (input === this.typeNameIdOfArray || input === this.typeNameIdOfSlice) { - input = this.typeNameIdOfArrayOrSlice; - } - if (input === this.typeNameIdOfTuple || input === this.typeNameIdOfUnit) { - input = this.typeNameIdOfTupleOrUnit; - } - if (input === this.typeNameIdOfFn || input === this.typeNameIdOfFnMut || - input === this.typeNameIdOfFnOnce) { - input = this.typeNameIdOfHof; - } - /** - * http://burtleburtle.net/bob/hash/integer.html - * ~~ is toInt32. It's used before adding, so - * the number stays in safe integer range. - * @param {number} k - */ - const hashint1 = k => { - k = (~~k + 0x7ed55d16) + (k << 12); - k = (k ^ 0xc761c23c) ^ (k >>> 19); - k = (~~k + 0x165667b1) + (k << 5); - k = (~~k + 0xd3a2646c) ^ (k << 9); - k = (~~k + 0xfd7046c5) + (k << 3); - return (k ^ 0xb55a4f09) ^ (k >>> 16); - }; - /** @param {number} k */ - const hashint2 = k => { - k = ~k + (k << 15); - k ^= k >>> 12; - k += k << 2; - k ^= k >>> 4; - k = Math.imul(k, 2057); - return k ^ (k >> 16); - }; - if (input !== null) { - const h0a = hashint1(input); - const h0b = hashint2(input); - // Less Hashing, Same Performance: Building a Better Bloom Filter - // doi=10.1.1.72.2442 - const h1a = ~~(h0a + Math.imul(h0b, 2)); - const h1b = ~~(h0a + Math.imul(h0b, 3)); - const h2a = ~~(h0a + Math.imul(h0b, 4)); - const h2b = ~~(h0a + Math.imul(h0b, 5)); - output[0] |= (1 << (h0a % 32)) | (1 << (h1b % 32)); - output[1] |= (1 << (h1a % 32)) | (1 << (h2b % 32)); - output[2] |= (1 << (h2a % 32)) | (1 << (h0b % 32)); - // output[3] is the total number of items in the type signature - output[3] += 1; - } - for (const g of type.generics) { - this.buildFunctionTypeFingerprint(g, output); - } - /** - * @type {{ - * id: number|null, - * ty: number, - * generics: rustdoc.FingerprintableType[], - * bindings: Map<number, rustdoc.FingerprintableType[]> - * }} - */ - const fb = { - id: null, - ty: 0, - generics: this.EMPTY_GENERICS_ARRAY, - bindings: this.EMPTY_BINDINGS_MAP, - }; - for (const [k, v] of type.bindings.entries()) { - fb.id = k; - fb.generics = v; - this.buildFunctionTypeFingerprint(fb, output); - } - } - - /** - * Convert raw search index into in-memory search index. - * - * @param {Map<string, rustdoc.RawSearchIndexCrate>} rawSearchIndex - * @returns {rustdoc.Row[]} - */ - buildIndex(rawSearchIndex) { - /** - * Convert from RawFunctionSearchType to FunctionSearchType. - * - * Crates often have lots of functions in them, and function signatures are sometimes - * complex, so rustdoc uses a pretty tight encoding for them. This function converts it - * to a simpler, object-based encoding so that the actual search code is more readable - * and easier to debug. - * - * The raw function search type format is generated using serde in - * librustdoc/html/render/mod.rs: IndexItemFunctionType::write_to_string - * - * @param {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean - * }>} paths - * @param {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean - * }>} lowercasePaths - * - * @return {function(rustdoc.RawFunctionSearchType): null|rustdoc.FunctionSearchType} - */ - const buildFunctionSearchTypeCallback = (paths, lowercasePaths) => { - /** - * @param {rustdoc.RawFunctionSearchType} functionSearchType - */ - const cb = functionSearchType => { - if (functionSearchType === 0) { - return null; - } - const INPUTS_DATA = 0; - const OUTPUT_DATA = 1; - /** @type {rustdoc.FunctionType[]} */ - let inputs; - /** @type {rustdoc.FunctionType[]} */ - let output; - if (typeof functionSearchType[INPUTS_DATA] === "number") { - inputs = [ - this.buildItemSearchType( - functionSearchType[INPUTS_DATA], - paths, - lowercasePaths, - ), - ]; - } else { - inputs = this.buildItemSearchTypeAll( - functionSearchType[INPUTS_DATA], - paths, - lowercasePaths, - ); - } - if (functionSearchType.length > 1) { - if (typeof functionSearchType[OUTPUT_DATA] === "number") { - output = [ - this.buildItemSearchType( - functionSearchType[OUTPUT_DATA], - paths, - lowercasePaths, - ), - ]; - } else { - output = this.buildItemSearchTypeAll( - // @ts-expect-error - functionSearchType[OUTPUT_DATA], - paths, - lowercasePaths, - ); - } - } else { - output = []; - } - const where_clause = []; - const l = functionSearchType.length; - for (let i = 2; i < l; ++i) { - where_clause.push(typeof functionSearchType[i] === "number" - // @ts-expect-error - ? [this.buildItemSearchType(functionSearchType[i], paths, lowercasePaths)] - : this.buildItemSearchTypeAll( - // @ts-expect-error - functionSearchType[i], - paths, - lowercasePaths, - )); - } - return { - inputs, output, where_clause, - }; - }; - return cb; + return -1; }; - - /** @type {rustdoc.Row[]} */ - const searchIndex = []; - let currentIndex = 0; - let id = 0; - - // Function type fingerprints are 128-bit bloom filters that are used to - // estimate the distance between function and query. - // This loop counts the number of items to allocate a fingerprint for. - for (const crate of rawSearchIndex.values()) { - // Each item gets an entry in the fingerprint array, and the crate - // does, too - id += crate.t.length + 1; - } - this.functionTypeFingerprint = new Uint32Array((id + 1) * 4); - // This loop actually generates the search item indexes, including - // normalized names, type signature objects and fingerprints, and aliases. - id = 0; - - /** @type {Array<[string, { [key: string]: Array<number> }, number]>} */ - const allAliases = []; - for (const [crate, crateCorpus] of rawSearchIndex) { - // a string representing the lengths of each description shard - // a string representing the list of function types - const itemDescShardDecoder = new VlqHexDecoder(crateCorpus.D, noop => { - /** @type {number} */ - // @ts-expect-error - const n = noop; - return n; - }); - let descShard = { - crate, - shard: 0, - start: 0, - len: itemDescShardDecoder.next(), - promise: null, - resolve: null, - }; - const descShardList = [descShard]; - - // Deprecated items and items with no description - this.searchIndexDeprecated.set(crate, new RoaringBitmap(crateCorpus.c)); - this.searchIndexEmptyDesc.set(crate, new RoaringBitmap(crateCorpus.e)); - let descIndex = 0; - - /** - * List of generic function type parameter names. - * Used for display, not for searching. - * @type {string[]} - */ - let lastParamNames = []; - - // This object should have exactly the same set of fields as the "row" - // object defined below. Your JavaScript runtime will thank you. - // https://mathiasbynens.be/notes/shapes-ics - let normalizedName = crate.indexOf("_") === -1 ? crate : crate.replace(/_/g, ""); - const crateRow = { - crate, - ty: 3, // == ExternCrate - name: crate, - path: "", - descShard, - descIndex, - exactPath: "", - desc: crateCorpus.doc, - parent: undefined, - type: null, - paramNames: lastParamNames, - id, - word: crate, - normalizedName, - bitIndex: 0, - implDisambiguator: null, - }; - this.nameTrie.insert(normalizedName, id, this.tailTable); - id += 1; - searchIndex.push(crateRow); - currentIndex += 1; - // it's not undefined - // @ts-expect-error - if (!this.searchIndexEmptyDesc.get(crate).contains(0)) { - descIndex += 1; - } - - // see `RawSearchIndexCrate` in `rustdoc.d.ts` for a more - // up to date description of these fields - const itemTypes = crateCorpus.t; - // an array of (String) item names - const itemNames = crateCorpus.n; - // an array of [(Number) item index, - // (String) full path] - // an item whose index is not present will fall back to the previous present path - // i.e. if indices 4 and 11 are present, but 5-10 and 12-13 are not present, - // 5-10 will fall back to the path for 4 and 12-13 will fall back to the path for 11 - const itemPaths = new Map(crateCorpus.q); - // An array of [(Number) item index, (Number) path index] - // Used to de-duplicate inlined and re-exported stuff - const itemReexports = new Map(crateCorpus.r); - // an array of (Number) the parent path index + 1 to `paths`, or 0 if none - const itemParentIdxDecoder = new VlqHexDecoder(crateCorpus.i, noop => noop); - // a map Number, string for impl disambiguators - const implDisambiguator = new Map(crateCorpus.b); - const rawPaths = crateCorpus.p; - const aliases = crateCorpus.a; - // an array of [(Number) item index, - // (String) comma-separated list of function generic param names] - // an item whose index is not present will fall back to the previous present path - const itemParamNames = new Map(crateCorpus.P); - - /** - * @type {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean - * }>} - */ - const lowercasePaths = []; - /** - * @type {Array<{ - * name: string, - * ty: number, - * path: string|null, - * exactPath: string|null, - * unboxFlag: boolean - * }>} - */ - const paths = []; - - // a string representing the list of function types - const itemFunctionDecoder = new VlqHexDecoder( - crateCorpus.f, - // @ts-expect-error - buildFunctionSearchTypeCallback(paths, lowercasePaths), - ); - - // convert `rawPaths` entries into object form - // generate normalizedPaths for function search mode - let len = rawPaths.length; - let lastPath = undef2null(itemPaths.get(0)); - for (let i = 0; i < len; ++i) { - const elem = rawPaths[i]; - const ty = elem[0]; - const name = elem[1]; - /** - * @param {2|3} idx - * @param {string|null} if_null - * @param {string|null} if_not_found - * @returns {string|null} - */ - const elemPath = (idx, if_null, if_not_found) => { - if (elem.length > idx && elem[idx] !== undefined) { - const p = itemPaths.get(elem[idx]); - if (p !== undefined) { - return p; - } - return if_not_found; - } - return if_null; - }; - const path = elemPath(2, lastPath, null); - const exactPath = elemPath(3, path, path); - const unboxFlag = elem.length > 4 && !!elem[4]; - - lowercasePaths.push({ ty, name: name.toLowerCase(), path, exactPath, unboxFlag }); - paths[i] = { ty, name, path, exactPath, unboxFlag }; - } - - // Convert `item*` into an object form, and construct word indices. - // - // Before any analysis is performed, let's gather the search terms to - // search against apart from the rest of the data. This is a quick - // operation that is cached for the life of the page state so that - // all other search operations have access to this cached data for - // faster analysis operations - lastPath = ""; - len = itemTypes.length; - let lastName = ""; - let lastWord = ""; - for (let i = 0; i < len; ++i) { - const bitIndex = i + 1; - if (descIndex >= descShard.len && - // @ts-expect-error - !this.searchIndexEmptyDesc.get(crate).contains(bitIndex)) { - descShard = { - crate, - shard: descShard.shard + 1, - start: descShard.start + descShard.len, - len: itemDescShardDecoder.next(), - promise: null, - resolve: null, - }; - descIndex = 0; - descShardList.push(descShard); - } - const name = itemNames[i] === "" ? lastName : itemNames[i]; - const word = itemNames[i] === "" ? lastWord : itemNames[i].toLowerCase(); - const pathU = itemPaths.get(i); - const path = pathU !== undefined ? pathU : lastPath; - const paramNameString = itemParamNames.get(i); - const paramNames = paramNameString !== undefined ? - paramNameString.split(",") : - lastParamNames; - const type = itemFunctionDecoder.next(); - if (type !== null) { - if (type) { - const fp = this.functionTypeFingerprint.subarray(id * 4, (id + 1) * 4); - for (const t of type.inputs) { - this.buildFunctionTypeFingerprint(t, fp); - } - for (const t of type.output) { - this.buildFunctionTypeFingerprint(t, fp); - } - for (const w of type.where_clause) { - for (const t of w) { - this.buildFunctionTypeFingerprint(t, fp); - } - } - } - } - // This object should have exactly the same set of fields as the "crateRow" - // object defined above. - const itemParentIdx = itemParentIdxDecoder.next(); - normalizedName = word.indexOf("_") === -1 ? word : word.replace(/_/g, ""); - /** @type {rustdoc.Row} */ - const row = { - crate, - ty: itemTypes.charCodeAt(i) - 65, // 65 = "A" - name, - path, - descShard, - descIndex, - exactPath: itemReexports.has(i) ? - // @ts-expect-error - itemPaths.get(itemReexports.get(i)) : path, - // @ts-expect-error - parent: itemParentIdx > 0 ? paths[itemParentIdx - 1] : undefined, - type, - paramNames, - id, - word, - normalizedName, - bitIndex, - implDisambiguator: undef2null(implDisambiguator.get(i)), - }; - this.nameTrie.insert(normalizedName, id, this.tailTable); - id += 1; - searchIndex.push(row); - lastPath = row.path; - lastParamNames = row.paramNames; - // @ts-expect-error - if (!this.searchIndexEmptyDesc.get(crate).contains(bitIndex)) { - descIndex += 1; - } - lastName = name; - lastWord = word; - } - - if (aliases) { - // We need to add the aliases in `searchIndex` after we finished filling it - // to not mess up indexes. - allAliases.push([crate, aliases, currentIndex]); - } - currentIndex += itemTypes.length; - this.searchState.descShards.set(crate, descShardList); - } - - for (const [crate, aliases, index] of allAliases) { - for (const [alias_name, alias_refs] of Object.entries(aliases)) { - if (!this.ALIASES.has(crate)) { - this.ALIASES.set(crate, new Map()); - } - const word = alias_name.toLowerCase(); - const crate_alias_map = this.ALIASES.get(crate); - if (!crate_alias_map.has(word)) { - crate_alias_map.set(word, []); - } - const aliases_map = crate_alias_map.get(word); - - const normalizedName = word.indexOf("_") === -1 ? word : word.replace(/_/g, ""); - for (const alias of alias_refs) { - const originalIndex = alias + index; - const original = searchIndex[originalIndex]; - /** @type {rustdoc.Row} */ - const row = { - crate, - name: alias_name, - normalizedName, - is_alias: true, - ty: original.ty, - type: original.type, - paramNames: [], - word, - id, - parent: undefined, - original, - path: "", - implDisambiguator: original.implDisambiguator, - // Needed to load the description of the original item. - // @ts-ignore - descShard: original.descShard, - descIndex: original.descIndex, - bitIndex: original.bitIndex, - }; - aliases_map.push(row); - this.nameTrie.insert(normalizedName, id, this.tailTable); - id += 1; - searchIndex.push(row); - } - } - } - // Drop the (rather large) hash table used for reusing function items - this.TYPES_POOL = new Map(); - return searchIndex; + this.typeNameIdOfOutput = await first(output, TY_ASSOCTYPE, ""); + this.typeNameIdOfFnPtr = await first(fn, TY_PRIMITIVE, ""); + this.typeNameIdOfFn = await first(fn, TY_TRAIT, "core::ops"); + this.typeNameIdOfFnMut = await first(fnMut, TY_TRAIT, "core::ops"); + this.typeNameIdOfFnOnce = await first(fnOnce, TY_TRAIT, "core::ops"); + this.typeNameIdOfArray = await first(array, TY_PRIMITIVE, ""); + this.typeNameIdOfSlice = await first(slice, TY_PRIMITIVE, ""); + this.typeNameIdOfArrayOrSlice = await first(arrayOrSlice, TY_PRIMITIVE, ""); + this.typeNameIdOfTuple = await first(tuple, TY_PRIMITIVE, ""); + this.typeNameIdOfUnit = await first(unit, TY_PRIMITIVE, ""); + this.typeNameIdOfTupleOrUnit = await first(tupleOrUnit, TY_PRIMITIVE, ""); + this.typeNameIdOfReference = await first(reference, TY_PRIMITIVE, ""); + this.typeNameIdOfHof = await first(hof, TY_PRIMITIVE, ""); + this.typeNameIdOfNever = await first(never, TY_PRIMITIVE, ""); } /** @@ -2343,41 +1288,6 @@ class DocSearch { */ static parseQuery(userQuery) { /** - * @param {string} typename - * @returns {number} - */ - function itemTypeFromName(typename) { - const index = itemTypes.findIndex(i => i === typename); - if (index < 0) { - throw ["Unknown type filter ", typename]; - } - return index; - } - - /** - * @param {rustdoc.ParserQueryElement} elem - */ - function convertTypeFilterOnElem(elem) { - if (typeof elem.typeFilter === "string") { - let typeFilter = elem.typeFilter; - if (typeFilter === "const") { - typeFilter = "constant"; - } - elem.typeFilter = itemTypeFromName(typeFilter); - } else { - elem.typeFilter = NO_TYPE_FILTER; - } - for (const elem2 of elem.generics) { - convertTypeFilterOnElem(elem2); - } - for (const constraints of elem.bindings.values()) { - for (const constraint of constraints) { - convertTypeFilterOnElem(constraint); - } - } - } - - /** * Takes the user search input and returns an empty `ParsedQuery`. * * @param {string} userQuery @@ -2437,8 +1347,7 @@ class DocSearch { continue; } if (!foundStopChar) { - /** @type String[] */ - let extra = []; + let extra = EMPTY_STRING_ARRAY; if (isLastElemGeneric(query.elems, parserState)) { extra = [" after ", ">"]; } else if (prevIs(parserState, "\"")) { @@ -2515,11 +1424,33 @@ class DocSearch { try { parseInput(query, parserState); + + // Scan for invalid type filters, so that we can report the error + // outside the search loop. + /** @param {rustdoc.ParserQueryElement} elem */ + const checkTypeFilter = elem => { + const ty = itemTypeFromName(elem.typeFilter); + if (ty === TY_GENERIC && elem.generics.length !== 0) { + throw [ + "Generic type parameter ", + elem.name, + " does not accept generic parameters", + ]; + } + for (const generic of elem.generics) { + checkTypeFilter(generic); + } + for (const constraints of elem.bindings.values()) { + for (const constraint of constraints) { + checkTypeFilter(constraint); + } + } + }; for (const elem of query.elems) { - convertTypeFilterOnElem(elem); + checkTypeFilter(elem); } for (const elem of query.returned) { - convertTypeFilterOnElem(elem); + checkTypeFilter(elem); } } catch (err) { query = newParsedQuery(userQuery); @@ -2543,208 +1474,573 @@ class DocSearch { } /** - * Executes the parsed query and builds a {ResultsTable}. - * - * @param {rustdoc.ParsedQuery<rustdoc.ParserQueryElement>} origParsedQuery - * - The parsed user query - * @param {Object} filterCrates - Crate to search in if defined - * @param {string} currentCrate - Current crate, to rank results from this crate higher - * - * @return {Promise<rustdoc.ResultsTable>} + * @param {number} id + * @returns {Promise<string|null>} */ - async execQuery(origParsedQuery, filterCrates, currentCrate) { - /** @type {rustdoc.Results} */ - const results_others = new Map(), - /** @type {rustdoc.Results} */ - results_in_args = new Map(), - /** @type {rustdoc.Results} */ - results_returned = new Map(); - - /** @type {rustdoc.ParsedQuery<rustdoc.QueryElement>} */ - // @ts-expect-error - const parsedQuery = origParsedQuery; + async getName(id) { + const ni = this.database.getData("name"); + if (!ni) { + return null; + } + const name = await ni.at(id); + return name === undefined || name === null ? null : this.utf8decoder.decode(name); + } - const queryLen = - parsedQuery.elems.reduce((acc, next) => acc + next.pathLast.length, 0) + - parsedQuery.returned.reduce((acc, next) => acc + next.pathLast.length, 0); - const maxEditDistance = Math.floor(queryLen / 3); - // We reinitialize the `FOUND_ALIASES` map. - this.FOUND_ALIASES.clear(); + /** + * @param {number} id + * @returns {Promise<string|null>} + */ + async getDesc(id) { + const di = this.database.getData("desc"); + if (!di) { + return null; + } + const desc = await di.at(id); + return desc === undefined || desc === null ? null : this.utf8decoder.decode(desc); + } + /** + * @param {number} id + * @returns {Promise<number|null>} + */ + async getAliasTarget(id) { + const ai = this.database.getData("alias"); + if (!ai) { + return null; + } + const bytes = await ai.at(id); + if (bytes === undefined || bytes === null || bytes.length === 0) { + return null; + } else { + /** @type {string} */ + const encoded = this.utf8decoder.decode(bytes); + /** @type {number|null} */ + const decoded = JSON.parse(encoded); + return decoded; + } + } + + /** + * @param {number} id + * @returns {Promise<rustdoc.EntryData|null>} + */ + async getEntryData(id) { + const ei = this.database.getData("entry"); + if (!ei) { + return null; + } + const encoded = this.utf8decoder.decode(await ei.at(id)); + if (encoded === "" || encoded === undefined || encoded === null) { + return null; + } /** - * @type {Map<string, number>} + * krate, + * ty, + * module_path, + * exact_module_path, + * parent, + * deprecated, + * associated_item_disambiguator + * @type {rustdoc.ArrayWithOptionals<[ + * number, + * rustdoc.ItemType, + * number, + * number, + * number, + * number, + * ], [string]>} */ - const genericSymbols = new Map(); + const raw = JSON.parse(encoded); + return { + krate: raw[0], + ty: raw[1], + modulePath: raw[2] === 0 ? null : raw[2] - 1, + exactModulePath: raw[3] === 0 ? null : raw[3] - 1, + parent: raw[4] === 0 ? null : raw[4] - 1, + deprecated: raw[5] === 1 ? true : false, + associatedItemDisambiguator: raw.length === 6 ? null : raw[6], + }; + } + /** + * @param {number} id + * @returns {Promise<rustdoc.PathData|null>} + */ + async getPathData(id) { + const pi = this.database.getData("path"); + if (!pi) { + return null; + } + const encoded = this.utf8decoder.decode(await pi.at(id)); + if (encoded === "" || encoded === undefined || encoded === null) { + return null; + } /** - * Convert names to ids in parsed query elements. - * This is not used for the "In Names" tab, but is used for the - * "In Params", "In Returns", and "In Function Signature" tabs. - * - * If there is no matching item, but a close-enough match, this - * function also that correction. - * - * See `buildTypeMapIndex` for more information. - * - * @param {rustdoc.QueryElement} elem - * @param {boolean=} isAssocType + * ty, module_path, exact_module_path, search_unbox, inverted_function_signature_index + * @type {rustdoc.ArrayWithOptionals<[rustdoc.ItemType, string], [string|0, 0|1, string]>} */ - const convertNameToId = (elem, isAssocType) => { - const loweredName = elem.pathLast.toLowerCase(); - if (this.typeNameIdMap.has(loweredName) && - // @ts-expect-error - (isAssocType || !this.typeNameIdMap.get(loweredName).assocOnly)) { - // @ts-expect-error - elem.id = this.typeNameIdMap.get(loweredName).id; - } else if (!parsedQuery.literalSearch) { - let match = null; - let matchDist = maxEditDistance + 1; - let matchName = ""; - for (const [name, { id, assocOnly }] of this.typeNameIdMap) { - const dist = Math.min( - editDistance(name, loweredName, maxEditDistance), - editDistance(name, elem.normalizedPathLast, maxEditDistance), - ); - if (dist <= matchDist && dist <= maxEditDistance && - (isAssocType || !assocOnly)) { - if (dist === matchDist && matchName > name) { - continue; - } - match = id; - matchDist = dist; - matchName = name; - } - } - if (match !== null) { - parsedQuery.correction = matchName; - } - elem.id = match; + const raw = JSON.parse(encoded); + return { + ty: raw[0], + modulePath: raw[1], + exactModulePath: raw[2] === 0 || raw[2] === undefined ? raw[1] : raw[2], + }; + } + + /** + * @param {number} id + * @returns {Promise<rustdoc.FunctionData|null>} + */ + async getFunctionData(id) { + const fi = this.database.getData("function"); + if (!fi) { + return null; + } + const encoded = this.utf8decoder.decode(await fi.at(id)); + if (encoded === "" || encoded === undefined || encoded === null) { + return null; + } + /** + * function_signature, param_names + * @type {[string, string[]]} + */ + const raw = JSON.parse(encoded); + + const parser = new VlqHexDecoder(raw[0], async functionSearchType => { + if (typeof functionSearchType === "number") { + return null; } - if ((elem.id === null && parsedQuery.totalElems > 1 && elem.typeFilter === -1 - && elem.generics.length === 0 && elem.bindings.size === 0) - || elem.typeFilter === TY_GENERIC) { - const id = genericSymbols.get(elem.normalizedPathLast); - if (id !== undefined) { - elem.id = id; + const INPUTS_DATA = 0; + const OUTPUT_DATA = 1; + /** @type {Promise<rustdoc.FunctionType[]>} */ + let inputs_; + /** @type {Promise<rustdoc.FunctionType[]>} */ + let output_; + if (typeof functionSearchType[INPUTS_DATA] === "number") { + inputs_ = Promise.all([ + this.buildItemSearchType(functionSearchType[INPUTS_DATA]), + ]); + } else { + // @ts-ignore + inputs_ = this.buildItemSearchTypeAll(functionSearchType[INPUTS_DATA]); + } + if (functionSearchType.length > 1) { + if (typeof functionSearchType[OUTPUT_DATA] === "number") { + output_ = Promise.all([ + this.buildItemSearchType(functionSearchType[OUTPUT_DATA]), + ]); } else { - elem.id = -(genericSymbols.size + 1); - genericSymbols.set(elem.normalizedPathLast, elem.id); - } - if (elem.typeFilter === -1 && elem.normalizedPathLast.length >= 3) { - // Silly heuristic to catch if the user probably meant - // to not write a generic parameter. We don't use it, - // just bring it up. - const maxPartDistance = Math.floor(elem.normalizedPathLast.length / 3); - let matchDist = maxPartDistance + 1; - let matchName = ""; - for (const name of this.typeNameIdMap.keys()) { - const dist = editDistance( - name, - elem.normalizedPathLast, - maxPartDistance, - ); - if (dist <= matchDist && dist <= maxPartDistance) { - if (dist === matchDist && matchName > name) { - continue; - } - matchDist = dist; - matchName = name; - } - } - if (matchName !== "") { - parsedQuery.proposeCorrectionFrom = elem.name; - parsedQuery.proposeCorrectionTo = matchName; - } + // @ts-expect-error + output_ = this.buildItemSearchTypeAll(functionSearchType[OUTPUT_DATA]); } - elem.typeFilter = TY_GENERIC; - } - if (elem.generics.length > 0 && elem.typeFilter === TY_GENERIC) { - // Rust does not have HKT - parsedQuery.error = [ - "Generic type parameter ", - elem.name, - " does not accept generic parameters", - ]; - } - for (const elem2 of elem.generics) { - convertNameToId(elem2); + } else { + output_ = Promise.resolve(EMPTY_GENERICS_ARRAY); } - elem.bindings = new Map(Array.from(elem.bindings.entries()) - .map(entry => { - const [name, constraints] = entry; + /** @type {Promise<rustdoc.FunctionType[]>[]} */ + const where_clause_ = []; + const l = functionSearchType.length; + for (let i = 2; i < l; ++i) { + where_clause_.push(typeof functionSearchType[i] === "number" // @ts-expect-error - if (!this.typeNameIdMap.has(name)) { - parsedQuery.error = [ - "Type parameter ", - // @ts-expect-error - name, - " does not exist", - ]; - return [0, []]; - } - for (const elem2 of constraints) { - convertNameToId(elem2, false); - } - + ? Promise.all([this.buildItemSearchType(functionSearchType[i])]) // @ts-expect-error - return [this.typeNameIdMap.get(name).id, constraints]; - }), - ); + : this.buildItemSearchTypeAll(functionSearchType[i]), + ); + } + const [inputs, output, where_clause] = await Promise.all([ + inputs_, + output_, + Promise.all(where_clause_), + ]); + return { + inputs, output, where_clause, + }; + }); + + return { + functionSignature: await parser.next(), + paramNames: raw[1], + elemCount: parser.elemCount, }; + } - for (const elem of parsedQuery.elems) { - convertNameToId(elem, false); - this.buildFunctionTypeFingerprint(elem, parsedQuery.typeFingerprint); + /** + * @param {number} id + * @returns {Promise<rustdoc.TypeData|null>} + */ + async getTypeData(id) { + const ti = this.database.getData("type"); + if (!ti) { + return null; } - for (const elem of parsedQuery.returned) { - convertNameToId(elem, false); - this.buildFunctionTypeFingerprint(elem, parsedQuery.typeFingerprint); + const encoded = this.utf8decoder.decode(await ti.at(id)); + if (encoded === "" || encoded === undefined || encoded === null) { + return null; } + /** + * function_signature, param_names + * @type {[string, number] | [number] | [string] | [] | null} + */ + const raw = JSON.parse(encoded); + if (!raw || raw.length === 0) { + return null; + } + + let searchUnbox = false; + const invertedFunctionSignatureIndex = []; + + if (typeof raw[0] === "string") { + if (raw[1]) { + searchUnbox = true; + } + // the inverted function signature index is a list of bitmaps, + // by number of types that appear in the function + let i = 0; + const pb = makeUint8ArrayFromBase64(raw[0]); + const l = pb.length; + while (i < l) { + if (pb[i] === 0) { + invertedFunctionSignatureIndex.push(RoaringBitmap.empty()); + i += 1; + } else { + const bitmap = new RoaringBitmap(pb, i); + i += bitmap.consumed_len_bytes; + invertedFunctionSignatureIndex.push(bitmap); + } + } + } else if (raw[0]) { + searchUnbox = true; + } + + return { searchUnbox, invertedFunctionSignatureIndex }; + } + + /** + * @returns {Promise<string[]>} + */ + async getCrateNameList() { + const crateNames = this.database.getData("crateNames"); + if (!crateNames) { + return []; + } + const l = crateNames.length; + const names = []; + for (let i = 0; i < l; ++i) { + names.push(crateNames.at(i).then(name => { + if (name === undefined) { + return ""; + } + return this.utf8decoder.decode(name); + })); + } + return Promise.all(names); + } + + /** + * @param {number} id non-negative generic index + * @returns {Promise<stringdex.RoaringBitmap[]>} + */ + async getGenericInvertedIndex(id) { + const gii = this.database.getData("generic_inverted_index"); + if (!gii) { + return []; + } + const pb = await gii.at(id); + if (pb === undefined || pb === null || pb.length === 0) { + return []; + } + + const invertedFunctionSignatureIndex = []; + // the inverted function signature index is a list of bitmaps, + // by number of types that appear in the function + let i = 0; + const l = pb.length; + while (i < l) { + if (pb[i] === 0) { + invertedFunctionSignatureIndex.push(RoaringBitmap.empty()); + i += 1; + } else { + const bitmap = new RoaringBitmap(pb, i); + i += bitmap.consumed_len_bytes; + invertedFunctionSignatureIndex.push(bitmap); + } + } + return invertedFunctionSignatureIndex; + } + + /** + * @param {number} id + * @returns {Promise<rustdoc.Row?>} + */ + async getRow(id) { + const [name_, entry, path, type] = await Promise.all([ + this.getName(id), + this.getEntryData(id), + this.getPathData(id), + this.getFunctionData(id), + ]); + if (!entry && !path) { + return null; + } + const [ + moduleName, + modulePathData, + exactModuleName, + exactModulePathData, + ] = await Promise.all([ + entry && entry.modulePath !== null ? this.getName(entry.modulePath) : null, + entry && entry.modulePath !== null ? this.getPathData(entry.modulePath) : null, + entry && entry.exactModulePath !== null ? + this.getName(entry.exactModulePath) : + null, + entry && entry.exactModulePath !== null ? + this.getPathData(entry.exactModulePath) : + null, + ]); + const name = name_ === null ? "" : name_; + const normalizedName = (name.indexOf("_") === -1 ? + name : + name.replace(/_/g, "")).toLowerCase(); + const modulePath = modulePathData === null || moduleName === null ? "" : + (modulePathData.modulePath === "" ? + moduleName : + `${modulePathData.modulePath}::${moduleName}`); + const [parentName, parentPath] = entry !== null && entry.parent !== null ? + await Promise.all([this.getName(entry.parent), this.getPathData(entry.parent)]) : + [null, null]; + return { + id, + crate: entry ? nonnull(await this.getName(entry.krate)) : "", + ty: entry ? entry.ty : nonnull(path).ty, + name, + normalizedName, + modulePath, + exactModulePath: exactModulePathData === null || exactModuleName === null ? modulePath : + (exactModulePathData.exactModulePath === "" ? + exactModuleName : + `${exactModulePathData.exactModulePath}::${exactModuleName}`), + entry, + path, + type, + deprecated: entry ? entry.deprecated : false, + parent: parentName !== null && parentPath !== null ? + { name: parentName, path: parentPath } : + null, + }; + } + + /** + * Convert a list of RawFunctionType / ID to object-based FunctionType. + * + * Crates often have lots of functions in them, and it's common to have a large number of + * functions that operate on a small set of data types, so the search index compresses them + * by encoding function parameter and return types as indexes into an array of names. + * + * Even when a general-purpose compression algorithm is used, this is still a win. + * I checked. https://github.com/rust-lang/rust/pull/98475#issue-1284395985 + * + * The format for individual function types is encoded in + * librustdoc/html/render/mod.rs: impl Serialize for RenderType + * + * @param {null|Array<rustdoc.RawFunctionType>} types + * + * @return {Promise<Array<rustdoc.FunctionType>>} + */ + async buildItemSearchTypeAll(types) { + return types && types.length > 0 ? + await Promise.all(types.map(type => this.buildItemSearchType(type))) : + EMPTY_GENERICS_ARRAY; + } + /** + * Converts a single type. + * + * @param {rustdoc.RawFunctionType} type + * @return {Promise<rustdoc.FunctionType>} + */ + async buildItemSearchType(type) { + const PATH_INDEX_DATA = 0; + const GENERICS_DATA = 1; + const BINDINGS_DATA = 2; + let id, generics; /** - * Creates the query results. - * - * @param {Array<rustdoc.ResultObject>} results_in_args - * @param {Array<rustdoc.ResultObject>} results_returned - * @param {Array<rustdoc.ResultObject>} results_others - * @param {rustdoc.ParsedQuery<rustdoc.QueryElement>} parsedQuery - * - * @return {rustdoc.ResultsTable} + * @type {Map<number, rustdoc.FunctionType[]>} */ - function createQueryResults( - results_in_args, - results_returned, - results_others, - parsedQuery) { - return { - "in_args": results_in_args, - "returned": results_returned, - "others": results_others, - "query": parsedQuery, + let bindings; + if (typeof type === "number") { + id = type; + generics = EMPTY_GENERICS_ARRAY; + bindings = EMPTY_BINDINGS_MAP; + } else { + id = type[PATH_INDEX_DATA]; + generics = await this.buildItemSearchTypeAll(type[GENERICS_DATA]); + if (type[BINDINGS_DATA] && type[BINDINGS_DATA].length > 0) { + bindings = new Map((await Promise.all(type[BINDINGS_DATA].map( + /** + * @param {[rustdoc.RawFunctionType, rustdoc.RawFunctionType[]]} binding + * @returns {Promise<[number, rustdoc.FunctionType[]][]>} + */ + async binding => { + const [assocType, constraints] = binding; + // Associated type constructors are represented sloppily in rustdoc's + // type search, to make the engine simpler. + // + // MyType<Output<T>=Result<T>> is equivalent to MyType<Output<Result<T>>=T> + // and both are, essentially + // MyType<Output=(T, Result<T>)>, except the tuple isn't actually there. + // It's more like the value of a type binding is naturally an array, + // which rustdoc calls "constraints". + // + // As a result, the key should never have generics on it. + const [k, v] = await Promise.all([ + this.buildItemSearchType(assocType).then(t => t.id), + this.buildItemSearchTypeAll(constraints), + ]); + return k === null ? EMPTY_BINDINGS_ARRAY : [[k, v]]; + }, + ))).flat()); + } else { + bindings = EMPTY_BINDINGS_MAP; + } + } + /** + * @type {rustdoc.FunctionType} + */ + let result; + if (id < 0) { + // types less than 0 are generic parameters + // the actual names of generic parameters aren't stored, since they aren't API + result = { + id, + name: "", + ty: TY_GENERIC, + path: null, + exactPath: null, + generics, + bindings, + unboxFlag: true, + }; + } else if (id === 0) { + // `0` is used as a sentinel because it's fewer bytes than `null` + result = { + id: null, + name: "", + ty: TY_GENERIC, + path: null, + exactPath: null, + generics, + bindings, + unboxFlag: true, + }; + } else { + const [name, path, type] = await Promise.all([ + this.getName(id - 1), + this.getPathData(id - 1), + this.getTypeData(id - 1), + ]); + if (path === undefined || path === null || type === undefined || type === null) { + return { + id: null, + name: "", + ty: TY_GENERIC, + path: null, + exactPath: null, + generics, + bindings, + unboxFlag: true, + }; + } + result = { + id: id - 1, + name, + ty: path.ty, + path: path.modulePath, + exactPath: path.exactModulePath === null ? path.modulePath : path.exactModulePath, + generics, + bindings, + unboxFlag: type.searchUnbox, }; } + const cr = this.TYPES_POOL.get(result.id); + if (cr) { + // Shallow equality check. Since this function is used + // to construct every type object, this should be mostly + // equivalent to a deep equality check, except if there's + // a conflict, we don't keep the old one around, so it's + // not a fully precise implementation of hashcons. + if (cr.generics.length === result.generics.length && + cr.generics !== result.generics && + cr.generics.every((x, i) => result.generics[i] === x) + ) { + result.generics = cr.generics; + } + if (cr.bindings.size === result.bindings.size && cr.bindings !== result.bindings) { + let ok = true; + for (const [k, v] of cr.bindings.entries()) { + const v2 = result.bindings.get(k); + if (!v2) { + ok = false; + break; + } + if (v !== v2 && v.length === v2.length && v.every((x, i) => v2[i] === x)) { + result.bindings.set(k, v); + } else if (v !== v2) { + ok = false; + break; + } + } + if (ok) { + result.bindings = cr.bindings; + } + } + if (cr.ty === result.ty && cr.path === result.path + && cr.bindings === result.bindings && cr.generics === result.generics + && cr.ty === result.ty && cr.name === result.name + && cr.unboxFlag === result.unboxFlag + ) { + return cr; + } + } + this.TYPES_POOL.set(result.id, result); + return result; + } - // @ts-expect-error + /** + * Executes the parsed query and builds a {ResultsTable}. + * + * @param {rustdoc.ParsedQuery<rustdoc.ParserQueryElement>} parsedQuery + * - The parsed user query + * @param {Object} filterCrates - Crate to search in if defined + * @param {string} currentCrate - Current crate, to rank results from this crate higher + * + * @return {Promise<rustdoc.ResultsTable>} + */ + async execQuery(parsedQuery, filterCrates, currentCrate) { + const queryLen = + parsedQuery.elems.reduce((acc, next) => acc + next.pathLast.length, 0) + + parsedQuery.returned.reduce((acc, next) => acc + next.pathLast.length, 0); + const maxEditDistance = Math.floor(queryLen / 3); + + /** + * @param {rustdoc.Row} item + * @returns {[string, string, string]} + */ const buildHrefAndPath = item => { let displayPath; let href; - if (item.is_alias) { - this.FOUND_ALIASES.add(item.word); - item = item.original; - } const type = itemTypes[item.ty]; const name = item.name; - let path = item.path; - let exactPath = item.exactPath; + let path = item.modulePath; + let exactPath = item.exactModulePath; if (type === "mod") { displayPath = path + "::"; href = this.rootPath + path.replace(/::/g, "/") + "/" + name + "/index.html"; } else if (type === "import") { - displayPath = item.path + "::"; - href = this.rootPath + item.path.replace(/::/g, "/") + + displayPath = item.modulePath + "::"; + href = this.rootPath + item.modulePath.replace(/::/g, "/") + "/index.html#reexport." + name; } else if (type === "primitive" || type === "keyword") { displayPath = ""; @@ -2754,13 +2050,13 @@ class DocSearch { } else if (type === "externcrate") { displayPath = ""; href = this.rootPath + name + "/index.html"; - } else if (item.parent !== undefined) { + } else if (item.parent) { const myparent = item.parent; let anchor = type + "." + name; - const parentType = itemTypes[myparent.ty]; + const parentType = itemTypes[myparent.path.ty]; let pageType = parentType; let pageName = myparent.name; - exactPath = `${myparent.exactPath}::${myparent.name}`; + exactPath = `${myparent.path.exactModulePath}::${myparent.name}`; if (parentType === "primitive") { displayPath = myparent.name + "::"; @@ -2768,9 +2064,9 @@ class DocSearch { } else if (type === "structfield" && parentType === "variant") { // Structfields belonging to variants are special: the // final path element is the enum name. - const enumNameIdx = item.path.lastIndexOf("::"); - const enumName = item.path.substr(enumNameIdx + 2); - path = item.path.substr(0, enumNameIdx); + const enumNameIdx = item.modulePath.lastIndexOf("::"); + const enumName = item.modulePath.substr(enumNameIdx + 2); + path = item.modulePath.substr(0, enumNameIdx); displayPath = path + "::" + enumName + "::" + myparent.name + "::"; anchor = "variant." + myparent.name + ".field." + name; pageType = "enum"; @@ -2778,16 +2074,16 @@ class DocSearch { } else { displayPath = path + "::" + myparent.name + "::"; } - if (item.implDisambiguator !== null) { - anchor = item.implDisambiguator + "/" + anchor; + if (item.entry && item.entry.associatedItemDisambiguator !== null) { + anchor = item.entry.associatedItemDisambiguator + "/" + anchor; } href = this.rootPath + path.replace(/::/g, "/") + "/" + pageType + "." + pageName + ".html#" + anchor; } else { - displayPath = item.path + "::"; - href = this.rootPath + item.path.replace(/::/g, "/") + + displayPath = item.modulePath + "::"; + href = this.rootPath + item.modulePath.replace(/::/g, "/") + "/" + type + "." + name + ".html"; } return [displayPath, href, `${exactPath}::${name}`]; @@ -2810,82 +2106,16 @@ class DocSearch { * Add extra data to result objects, and filter items that have been * marked for removal. * - * @param {rustdoc.ResultObject[]} results - * @param {"sig"|"elems"|"returned"|null} typeInfo - * @returns {rustdoc.ResultObject[]} - */ - const transformResults = (results, typeInfo) => { - const duplicates = new Set(); - const out = []; - - for (const result of results) { - if (result.id !== -1) { - const res = buildHrefAndPath(this.searchIndex[result.id]); - // many of these properties don't strictly need to be - // copied over, but copying them over satisfies tsc, - // and hopefully plays nice with the shape optimization - // of the browser engine. - /** @type {rustdoc.ResultObject} */ - const obj = Object.assign({ - parent: result.parent, - type: result.type, - dist: result.dist, - path_dist: result.path_dist, - index: result.index, - desc: result.desc, - item: result.item, - displayPath: pathSplitter(res[0]), - fullPath: "", - href: "", - displayTypeSignature: null, - }, this.searchIndex[result.id]); - - // To be sure than it some items aren't considered as duplicate. - obj.fullPath = res[2] + "|" + obj.ty; - - if (duplicates.has(obj.fullPath)) { - continue; - } - - // Exports are specifically not shown if the items they point at - // are already in the results. - if (obj.ty === TY_IMPORT && duplicates.has(res[2])) { - continue; - } - if (duplicates.has(res[2] + "|" + TY_IMPORT)) { - continue; - } - duplicates.add(obj.fullPath); - duplicates.add(res[2]); - - if (typeInfo !== null) { - obj.displayTypeSignature = - // @ts-expect-error - this.formatDisplayTypeSignature(obj, typeInfo); - } - - obj.href = res[1]; - out.push(obj); - if (out.length >= MAX_RESULTS) { - break; - } - } - } - return out; - }; - - /** - * Add extra data to result objects, and filter items that have been - * marked for removal. - * * The output is formatted as an array of hunks, where odd numbered * hunks are highlighted and even numbered ones are not. * * @param {rustdoc.ResultObject} obj * @param {"sig"|"elems"|"returned"|null} typeInfo + * @param {rustdoc.QueryElement[]} elems + * @param {rustdoc.QueryElement[]} returned * @returns {Promise<rustdoc.DisplayTypeSignature>} */ - this.formatDisplayTypeSignature = async(obj, typeInfo) => { + const formatDisplayTypeSignature = async(obj, typeInfo, elems, returned) => { const objType = obj.type; if (!objType) { return {type: [], mappedNames: new Map(), whereClause: new Map()}; @@ -2897,13 +2127,13 @@ class DocSearch { if (typeInfo !== "elems" && typeInfo !== "returned") { fnInputs = unifyFunctionTypes( objType.inputs, - parsedQuery.elems, + elems, objType.where_clause, null, mgensScratch => { fnOutput = unifyFunctionTypes( objType.output, - parsedQuery.returned, + returned, objType.where_clause, mgensScratch, mgensOut => { @@ -2917,10 +2147,9 @@ class DocSearch { 0, ); } else { - const arr = typeInfo === "elems" ? objType.inputs : objType.output; const highlighted = unifyFunctionTypes( - arr, - parsedQuery.elems, + typeInfo === "elems" ? objType.inputs : objType.output, + typeInfo === "elems" ? elems : returned, objType.where_clause, null, mgensOut => { @@ -2969,15 +2198,15 @@ class DocSearch { } }; - parsedQuery.elems.forEach(remapQuery); - parsedQuery.returned.forEach(remapQuery); + elems.forEach(remapQuery); + returned.forEach(remapQuery); /** * Write text to a highlighting array. * Index 0 is not highlighted, index 1 is highlighted, * index 2 is not highlighted, etc. * - * @param {{name?: string, highlighted?: boolean}} fnType - input + * @param {{name: string|null, highlighted?: boolean}} fnType - input * @param {string[]} result */ const pushText = (fnType, result) => { @@ -3004,8 +2233,9 @@ class DocSearch { * * @param {rustdoc.HighlightedFunctionType} fnType - input * @param {string[]} result + * @returns {Promise<void>} */ - const writeHof = (fnType, result) => { + const writeHof = async(fnType, result) => { const hofOutput = fnType.bindings.get(this.typeNameIdOfOutput) || []; const hofInputs = fnType.generics; pushText(fnType, result); @@ -3016,7 +2246,7 @@ class DocSearch { pushText({ name: ", ", highlighted: false }, result); } needsComma = true; - writeFn(fnType, result); + await writeFn(fnType, result); } pushText({ name: hofOutput.length === 0 ? ")" : ") -> ", @@ -3031,7 +2261,7 @@ class DocSearch { pushText({ name: ", ", highlighted: false }, result); } needsComma = true; - writeFn(fnType, result); + await writeFn(fnType, result); } if (hofOutput.length > 1) { pushText({name: ")", highlighted: false}, result); @@ -3044,8 +2274,9 @@ class DocSearch { * * @param {rustdoc.HighlightedFunctionType} fnType * @param {string[]} result + * @returns {Promise<boolean>} */ - const writeSpecialPrimitive = (fnType, result) => { + const writeSpecialPrimitive = async(fnType, result) => { if (fnType.id === this.typeNameIdOfArray || fnType.id === this.typeNameIdOfSlice || fnType.id === this.typeNameIdOfTuple || fnType.id === this.typeNameIdOfUnit) { const [ob, sb] = @@ -3054,7 +2285,7 @@ class DocSearch { ["[", "]"] : ["(", ")"]; pushText({ name: ob, highlighted: fnType.highlighted }, result); - onEachBtwn( + await onEachBtwnAsync( fnType.generics, nested => writeFn(nested, result), // @ts-expect-error @@ -3065,11 +2296,11 @@ class DocSearch { } else if (fnType.id === this.typeNameIdOfReference) { pushText({ name: "&", highlighted: fnType.highlighted }, result); let prevHighlighted = false; - onEachBtwn( + await onEachBtwnAsync( fnType.generics, - value => { + async value => { prevHighlighted = !!value.highlighted; - writeFn(value, result); + await writeFn(value, result); }, // @ts-expect-error value => pushText({ @@ -3078,8 +2309,16 @@ class DocSearch { }, result), ); return true; - } else if (fnType.id === this.typeNameIdOfFn) { - writeHof(fnType, result); + } else if ( + fnType.id === this.typeNameIdOfFn || + fnType.id === this.typeNameIdOfFnMut || + fnType.id === this.typeNameIdOfFnOnce || + fnType.id === this.typeNameIdOfFnPtr + ) { + await writeHof(fnType, result); + return true; + } else if (fnType.id === this.typeNameIdOfNever) { + pushText({ name: "!", highlighted: fnType.highlighted }, result); return true; } return false; @@ -3091,8 +2330,9 @@ class DocSearch { * * @param {rustdoc.HighlightedFunctionType} fnType * @param {string[]} result + * @returns {Promise<void>} */ - const writeFn = (fnType, result) => { + const writeFn = async(fnType, result) => { if (fnType.id !== null && fnType.id < 0) { if (fnParamNames[-1 - fnType.id] === "") { // Normally, there's no need to shown an unhighlighted @@ -3101,7 +2341,7 @@ class DocSearch { fnType.generics : objType.where_clause[-1 - fnType.id]; for (const nested of generics) { - writeFn(nested, result); + await writeFn(nested, result); } return; } else if (mgens) { @@ -3120,7 +2360,7 @@ class DocSearch { }, result); /** @type{string[]} */ const where = []; - onEachBtwn( + await onEachBtwnAsync( fnType.generics, nested => writeFn(nested, where), // @ts-expect-error @@ -3131,32 +2371,61 @@ class DocSearch { } } else { if (fnType.ty === TY_PRIMITIVE) { - if (writeSpecialPrimitive(fnType, result)) { + if (await writeSpecialPrimitive(fnType, result)) { return; } } else if (fnType.ty === TY_TRAIT && ( fnType.id === this.typeNameIdOfFn || - fnType.id === this.typeNameIdOfFnMut || - fnType.id === this.typeNameIdOfFnOnce)) { - writeHof(fnType, result); + fnType.id === this.typeNameIdOfFnMut || + fnType.id === this.typeNameIdOfFnOnce || + fnType.id === this.typeNameIdOfFnPtr + )) { + await writeHof(fnType, result); + return; + } else if (fnType.name === "" && + fnType.bindings.size === 0 && + fnType.generics.length !== 0 + ) { + pushText({ name: "impl ", highlighted: false }, result); + if (fnType.generics.length > 1) { + pushText({ name: "(", highlighted: false }, result); + } + await onEachBtwnAsync( + fnType.generics, + value => writeFn(value, result), + // @ts-expect-error + () => pushText({ name: ", ", highlighted: false }, result), + ); + if (fnType.generics.length > 1) { + pushText({ name: ")", highlighted: false }, result); + } return; } pushText(fnType, result); let hasBindings = false; if (fnType.bindings.size > 0) { - onEachBtwn( - fnType.bindings, - ([key, values]) => { - const name = this.assocTypeIdNameMap.get(key); + await onEachBtwnAsync( + await Promise.all([...fnType.bindings.entries()].map( + /** + * @param {[number, rustdoc.HighlightedFunctionType[]]} param0 + * @returns {Promise<[ + * string|null, + * rustdoc.HighlightedFunctionType[], + * ]>} + */ + async([key, values]) => [await this.getName(key), values], + )), + async([name, values]) => { // @ts-expect-error if (values.length === 1 && values[0].id < 0 && // @ts-expect-error - `${fnType.name}::${name}` === fnParamNames[-1 - values[0].id]) { + `${fnType.name}::${name}` === fnParamNames[-1 - values[0].id] + ) { // the internal `Item=Iterator::Item` type variable should be // shown in the where clause and name mapping output, but is // redundant in this spot for (const value of values) { - writeFn(value, []); + await writeFn(value, []); } return true; } @@ -3169,7 +2438,7 @@ class DocSearch { name: values.length !== 1 ? "=(" : "=", highlighted: false, }, result); - onEachBtwn( + await onEachBtwnAsync( values || [], value => writeFn(value, result), // @ts-expect-error @@ -3186,7 +2455,7 @@ class DocSearch { if (fnType.generics.length > 0) { pushText({ name: hasBindings ? ", " : "<", highlighted: false }, result); } - onEachBtwn( + await onEachBtwnAsync( fnType.generics, value => writeFn(value, result), // @ts-expect-error @@ -3199,14 +2468,14 @@ class DocSearch { }; /** @type {string[]} */ const type = []; - onEachBtwn( + await onEachBtwnAsync( fnInputs, fnType => writeFn(fnType, type), // @ts-expect-error () => pushText({ name: ", ", highlighted: false }, type), ); pushText({ name: " -> ", highlighted: false }, type); - onEachBtwn( + await onEachBtwnAsync( fnOutput, fnType => writeFn(fnType, type), // @ts-expect-error @@ -3217,176 +2486,252 @@ class DocSearch { }; /** - * This function takes a result map, and sorts it by various criteria, including edit - * distance, substring match, and the crate it comes from. + * Add extra data to result objects, and filter items that have been + * marked for removal. * - * @param {rustdoc.Results} results + * @param {[rustdoc.PlainResultObject, rustdoc.Row][]} results * @param {"sig"|"elems"|"returned"|null} typeInfo - * @param {string} preferredCrate - * @returns {Promise<rustdoc.ResultObject[]>} + * @param {Set<string>} duplicates + * @returns {rustdoc.ResultObject[]} */ - const sortResults = async(results, typeInfo, preferredCrate) => { - const userQuery = parsedQuery.userQuery; - const normalizedUserQuery = parsedQuery.userQuery.toLowerCase(); - const isMixedCase = normalizedUserQuery !== userQuery; - const result_list = []; - const isReturnTypeQuery = parsedQuery.elems.length === 0 || - typeInfo === "returned"; - for (const result of results.values()) { - result.item = this.searchIndex[result.id]; - result.word = this.searchIndex[result.id].word; - if (isReturnTypeQuery) { - // We are doing a return-type based search, deprioritize "clone-like" results, - // ie. functions that also take the queried type as an argument. - const resultItemType = result.item && result.item.type; - if (!resultItemType) { + const transformResults = (results, typeInfo, duplicates) => { + const out = []; + + for (const [result, item] of results) { + if (item.id !== -1) { + const res = buildHrefAndPath(item); + // many of these properties don't strictly need to be + // copied over, but copying them over satisfies tsc, + // and hopefully plays nice with the shape optimization + // of the browser engine. + /** @type {rustdoc.ResultObject} */ + const obj = Object.assign({ + parent: item.parent ? { + path: item.parent.path.modulePath, + exactPath: item.parent.path.exactModulePath || + item.parent.path.modulePath, + name: item.parent.name, + ty: item.parent.path.ty, + } : undefined, + type: item.type && item.type.functionSignature ? + item.type.functionSignature : + undefined, + paramNames: item.type && item.type.paramNames ? + item.type.paramNames : + undefined, + dist: result.dist, + path_dist: result.path_dist, + index: result.index, + desc: this.getDesc(result.id), + item, + displayPath: pathSplitter(res[0]), + fullPath: "", + href: "", + displayTypeSignature: null, + }, result); + + // To be sure than it some items aren't considered as duplicate. + obj.fullPath = res[2] + "|" + obj.item.ty; + + if (duplicates.has(obj.fullPath)) { continue; } - const inputs = resultItemType.inputs; - const where_clause = resultItemType.where_clause; - if (containsTypeFromQuery(inputs, where_clause)) { - result.path_dist *= 100; - result.dist *= 100; + + // Exports are specifically not shown if the items they point at + // are already in the results. + if (obj.item.ty === TY_IMPORT && duplicates.has(res[2])) { + continue; + } + if (duplicates.has(res[2] + "|" + TY_IMPORT)) { + continue; + } + duplicates.add(obj.fullPath); + duplicates.add(res[2]); + + if (typeInfo !== null) { + obj.displayTypeSignature = formatDisplayTypeSignature( + obj, + typeInfo, + result.elems, + result.returned, + ); + } + + obj.href = res[1]; + out.push(obj); + if (out.length >= MAX_RESULTS) { + break; } } - result_list.push(result); } - result_list.sort((aaa, bbb) => { - /** @type {number} */ - let a; - /** @type {number} */ - let b; + return out; + }; - // sort by exact case-sensitive match - if (isMixedCase) { - a = Number(aaa.item.name !== userQuery); - b = Number(bbb.item.name !== userQuery); - if (a !== b) { - return a - b; + const sortAndTransformResults = + /** + * @this {DocSearch} + * @param {Array<rustdoc.PlainResultObject|null>} results + * @param {"sig"|"elems"|"returned"|null} typeInfo + * @param {string} preferredCrate + * @param {Set<string>} duplicates + * @returns {AsyncGenerator<rustdoc.ResultObject, number>} + */ + async function*(results, typeInfo, preferredCrate, duplicates) { + const userQuery = parsedQuery.userQuery; + const normalizedUserQuery = parsedQuery.userQuery.toLowerCase(); + const isMixedCase = normalizedUserQuery !== userQuery; + /** + * @type {[rustdoc.PlainResultObject, rustdoc.Row][]} + */ + const result_list = []; + for (const result of results.values()) { + if (!result) { + continue; + } + /** + * @type {rustdoc.Row?} + */ + const item = await this.getRow(result.id); + if (!item) { + continue; + } + if (filterCrates !== null && item.crate !== filterCrates) { + continue; + } + if (item) { + result_list.push([result, item]); + } else { + continue; } } - // sort by exact match with regard to the last word (mismatch goes later) - a = Number(aaa.word !== normalizedUserQuery); - b = Number(bbb.word !== normalizedUserQuery); - if (a !== b) { - return a - b; - } + result_list.sort(([aaa, aai], [bbb, bbi]) => { + /** @type {number} */ + let a; + /** @type {number} */ + let b; + + if (typeInfo === null) { + // in name based search... + + // sort by exact case-sensitive match + if (isMixedCase) { + a = Number(aai.name !== userQuery); + b = Number(bbi.name !== userQuery); + if (a !== b) { + return a - b; + } + } - // sort by index of keyword in item name (no literal occurrence goes later) - a = Number(aaa.index < 0); - b = Number(bbb.index < 0); - if (a !== b) { - return a - b; - } + // sort by exact match with regard to the last word (mismatch goes later) + a = Number(aai.normalizedName !== normalizedUserQuery); + b = Number(bbi.normalizedName !== normalizedUserQuery); + if (a !== b) { + return a - b; + } - // in type based search, put functions first - if (parsedQuery.hasReturnArrow) { - a = Number(!isFnLikeTy(aaa.item.ty)); - b = Number(!isFnLikeTy(bbb.item.ty)); + // sort by index of keyword in item name (no literal occurrence goes later) + a = Number(aaa.index < 0); + b = Number(bbb.index < 0); + if (a !== b) { + return a - b; + } + } + + // Sort by distance in the path part, if specified + // (less changes required to match means higher rankings) + a = Number(aaa.path_dist); + b = Number(bbb.path_dist); if (a !== b) { return a - b; } - } - // Sort by distance in the path part, if specified - // (less changes required to match means higher rankings) - a = Number(aaa.path_dist); - b = Number(bbb.path_dist); - if (a !== b) { - return a - b; - } - - // (later literal occurrence, if any, goes later) - a = Number(aaa.index); - b = Number(bbb.index); - if (a !== b) { - return a - b; - } + // (later literal occurrence, if any, goes later) + a = Number(aaa.index); + b = Number(bbb.index); + if (a !== b) { + return a - b; + } - // Sort by distance in the name part, the last part of the path - // (less changes required to match means higher rankings) - a = Number(aaa.dist); - b = Number(bbb.dist); - if (a !== b) { - return a - b; - } + // Sort by distance in the name part, the last part of the path + // (less changes required to match means higher rankings) + a = Number(aaa.dist); + b = Number(bbb.dist); + if (a !== b) { + return a - b; + } - // sort deprecated items later - a = Number( - // @ts-expect-error - this.searchIndexDeprecated.get(aaa.item.crate).contains(aaa.item.bitIndex), - ); - b = Number( - // @ts-expect-error - this.searchIndexDeprecated.get(bbb.item.crate).contains(bbb.item.bitIndex), - ); - if (a !== b) { - return a - b; - } + // sort aliases lower + a = Number(aaa.is_alias); + b = Number(bbb.is_alias); + if (a !== b) { + return a - b; + } - // sort by crate (current crate comes first) - a = Number(aaa.item.crate !== preferredCrate); - b = Number(bbb.item.crate !== preferredCrate); - if (a !== b) { - return a - b; - } + // sort deprecated items later + a = Number(aai.deprecated); + b = Number(bbi.deprecated); + if (a !== b) { + return a - b; + } - // sort by item name length (longer goes later) - a = Number(aaa.word.length); - b = Number(bbb.word.length); - if (a !== b) { - return a - b; - } + // sort by crate (current crate comes first) + a = Number(aai.crate !== preferredCrate); + b = Number(bbi.crate !== preferredCrate); + if (a !== b) { + return a - b; + } - // sort doc alias items later - a = Number(aaa.item.is_alias === true); - b = Number(bbb.item.is_alias === true); - if (a !== b) { - return a - b; - } + // sort by item name length (longer goes later) + a = Number(aai.normalizedName.length); + b = Number(bbi.normalizedName.length); + if (a !== b) { + return a - b; + } - // sort by item name (lexicographically larger goes later) - let aw = aaa.word; - let bw = bbb.word; - if (aw !== bw) { - return (aw > bw ? +1 : -1); - } + // sort by item name (lexicographically larger goes later) + let aw = aai.normalizedName; + let bw = bbi.normalizedName; + if (aw !== bw) { + return (aw > bw ? +1 : -1); + } - // sort by description (no description goes later) - a = Number( - // @ts-expect-error - this.searchIndexEmptyDesc.get(aaa.item.crate).contains(aaa.item.bitIndex), - ); - b = Number( - // @ts-expect-error - this.searchIndexEmptyDesc.get(bbb.item.crate).contains(bbb.item.bitIndex), - ); - if (a !== b) { - return a - b; - } + // sort by description (no description goes later) + const di = this.database.getData("desc"); + if (di) { + a = Number(di.isEmpty(aaa.id)); + b = Number(di.isEmpty(bbb.id)); + if (a !== b) { + return a - b; + } + } - // sort by type (later occurrence in `itemTypes` goes later) - a = Number(aaa.item.ty); - b = Number(bbb.item.ty); - if (a !== b) { - return a - b; - } + // sort by type (later occurrence in `itemTypes` goes later) + a = Number(aai.ty); + b = Number(bbi.ty); + if (a !== b) { + return a - b; + } - // sort by path (lexicographically larger goes later) - aw = aaa.item.path; - bw = bbb.item.path; - if (aw !== bw) { - return (aw > bw ? +1 : -1); - } + // sort by path (lexicographically larger goes later) + const ap = aai.modulePath; + const bp = bbi.modulePath; + aw = ap === undefined ? "" : ap; + bw = bp === undefined ? "" : bp; + if (aw !== bw) { + return (aw > bw ? +1 : -1); + } - // que sera, sera - return 0; - }); + // que sera, sera + return 0; + }); - return transformResults(result_list, typeInfo); - }; + const transformed_result_list = transformResults(result_list, typeInfo, duplicates); + yield* transformed_result_list; + return transformed_result_list.length; + } + .bind(this); /** * This function checks if a list of search query `queryElems` can all be found in the @@ -3938,6 +3283,8 @@ class DocSearch { } return true; } else { + // For these special cases, matching code need added to the inverted index. + // search_index.rs -> convert_render_type does this if (queryElem.id === this.typeNameIdOfArrayOrSlice && (fnType.id === this.typeNameIdOfSlice || fnType.id === this.typeNameIdOfArray) ) { @@ -3948,10 +3295,12 @@ class DocSearch { ) { // () matches primitive:tuple or primitive:unit // if it matches, then we're fine, and this is an appropriate match candidate - } else if (queryElem.id === this.typeNameIdOfHof && - (fnType.id === this.typeNameIdOfFn || fnType.id === this.typeNameIdOfFnMut || - fnType.id === this.typeNameIdOfFnOnce) - ) { + } else if (queryElem.id === this.typeNameIdOfHof && ( + fnType.id === this.typeNameIdOfFn || + fnType.id === this.typeNameIdOfFnMut || + fnType.id === this.typeNameIdOfFnOnce || + fnType.id === this.typeNameIdOfFnPtr + )) { // -> matches fn, fnonce, and fnmut // if it matches, then we're fine, and this is an appropriate match candidate } else if (fnType.id !== queryElem.id || queryElem.id === null) { @@ -4134,21 +3483,13 @@ class DocSearch { * This function checks if the given list contains any * (non-generic) types mentioned in the query. * + * @param {rustdoc.QueryElement[]} elems * @param {rustdoc.FunctionType[]} list - A list of function types. * @param {rustdoc.FunctionType[][]} where_clause - Trait bounds for generic items. */ - function containsTypeFromQuery(list, where_clause) { + function containsTypeFromQuery(elems, list, where_clause) { if (!list) return false; - for (const ty of parsedQuery.returned) { - // negative type ids are generics - if (ty.id !== null && ty.id < 0) { - continue; - } - if (checkIfInList(list, ty, where_clause, null, 0)) { - return true; - } - } - for (const ty of parsedQuery.elems) { + for (const ty of elems) { if (ty.id !== null && ty.id < 0) { continue; } @@ -4240,10 +3581,10 @@ class DocSearch { /** * Compute an "edit distance" that ignores missing path elements. * @param {string[]} contains search query path - * @param {rustdoc.Row} ty indexed item + * @param {string[]} path indexed page path * @returns {null|number} edit distance */ - function checkPath(contains, ty) { + function checkPath(contains, path) { if (contains.length === 0) { return 0; } @@ -4251,11 +3592,6 @@ class DocSearch { contains.reduce((acc, next) => acc + next.length, 0) / 3, ); let ret_dist = maxPathEditDistance + 1; - const path = ty.path.split("::"); - - if (ty.parent && ty.parent.name) { - path.push(ty.parent.name.toLowerCase()); - } const length = path.length; const clength = contains.length; @@ -4281,7 +3617,32 @@ class DocSearch { return ret_dist > maxPathEditDistance ? null : ret_dist; } - // @ts-expect-error + /** + * Compute an "edit distance" that ignores missing path elements. + * @param {string[]} contains search query path + * @param {rustdoc.Row} row indexed item + * @returns {null|number} edit distance + */ + function checkRowPath(contains, row) { + if (contains.length === 0) { + return 0; + } + + const path = row.modulePath.split("::"); + + if (row.parent && row.parent.name) { + path.push(row.parent.name.toLowerCase()); + } + + return checkPath(contains, path); + } + + /** + * + * @param {number} filter + * @param {rustdoc.ItemType} type + * @returns + */ function typePassesFilter(filter, type) { // No filter or Exact mach if (filter <= NO_TYPE_FILTER || filter === type) return true; @@ -4303,366 +3664,839 @@ class DocSearch { return false; } - // @ts-expect-error - const handleAliases = async(ret, query, filterCrates, currentCrate) => { - const lowerQuery = query.toLowerCase(); - if (this.FOUND_ALIASES.has(lowerQuery)) { - return; - } - this.FOUND_ALIASES.add(lowerQuery); - // We separate aliases and crate aliases because we want to have current crate - // aliases to be before the others in the displayed results. - // @ts-expect-error - const aliases = []; - // @ts-expect-error - const crateAliases = []; - if (filterCrates !== null) { - if (this.ALIASES.has(filterCrates) - && this.ALIASES.get(filterCrates).has(lowerQuery)) { - const query_aliases = this.ALIASES.get(filterCrates).get(lowerQuery); - for (const alias of query_aliases) { - aliases.push(alias); - } + const innerRunNameQuery = + /** + * @this {DocSearch} + * @param {string} currentCrate + * @returns {AsyncGenerator<rustdoc.ResultObject>} + */ + async function*(currentCrate) { + const index = this.database.getIndex("normalizedName"); + if (!index) { + return; } - } else { - for (const [crate, crateAliasesIndex] of this.ALIASES) { - if (crateAliasesIndex.has(lowerQuery)) { - // @ts-expect-error - const pushTo = crate === currentCrate ? crateAliases : aliases; - const query_aliases = crateAliasesIndex.get(lowerQuery); - for (const alias of query_aliases) { - pushTo.push(alias); + const idDuplicates = new Set(); + const pathDuplicates = new Set(); + let count = 0; + const prefixResults = []; + const normalizedUserQuery = parsedQuery.userQuery + .replace(/[_"]/g, "") + .toLowerCase(); + /** + * @param {string} name + * @param {number} alias + * @param {number} dist + * @param {number} index + * @returns {Promise<rustdoc.PlainResultObject?>} + */ + const handleAlias = async(name, alias, dist, index) => { + return { + id: alias, + dist, + path_dist: 0, + index, + alias: name, + is_alias: true, + elems: [], // only used in type-based queries + returned: [], // only used in type-based queries + original: await this.getRow(alias), + }; + }; + /** + * @param {Promise<rustdoc.PlainResultObject|null>[]} data + * @returns {AsyncGenerator<rustdoc.ResultObject, boolean>} + */ + const flush = async function* (data) { + const satr = sortAndTransformResults( + await Promise.all(data), + null, + currentCrate, + pathDuplicates, + ); + data.length = 0; + for await (const processed of satr) { + yield processed; + count += 1; + if ((count & 0x7F) === 0) { + await yieldToBrowser(); + } + if (count >= MAX_RESULTS) { + return true; } } - } - } - - // @ts-expect-error - const sortFunc = (aaa, bbb) => { - if (aaa.original.path < bbb.original.path) { - return 1; - } else if (aaa.original.path === bbb.original.path) { - return 0; - } - return -1; - }; - // @ts-expect-error - crateAliases.sort(sortFunc); - aliases.sort(sortFunc); - - // @ts-expect-error - const pushFunc = alias => { - // Cloning `alias` to prevent its fields to be updated. - alias = {...alias}; - const res = buildHrefAndPath(alias); - alias.displayPath = pathSplitter(res[0]); - alias.fullPath = alias.displayPath + alias.name; - alias.href = res[1]; - - ret.others.unshift(alias); - if (ret.others.length > MAX_RESULTS) { - ret.others.pop(); - } - }; - - aliases.forEach(pushFunc); - // @ts-expect-error - crateAliases.forEach(pushFunc); - }; - - /** - * This function adds the given result into the provided `results` map if it matches the - * following condition: - * - * * If it is a "literal search" (`parsedQuery.literalSearch`), then `dist` must be 0. - * * If it is not a "literal search", `dist` must be <= `maxEditDistance`. - * - * The `results` map contains information which will be used to sort the search results: - * - * * `fullId` is an `integer`` used as the key of the object we use for the `results` map. - * * `id` is the index in the `searchIndex` array for this element. - * * `index` is an `integer`` used to sort by the position of the word in the item's name. - * * `dist` is the main metric used to sort the search results. - * * `path_dist` is zero if a single-component search query is used, otherwise it's the - * distance computed for everything other than the last path component. - * - * @param {rustdoc.Results} results - * @param {number} fullId - * @param {number} id - * @param {number} index - * @param {number} dist - * @param {number} path_dist - * @param {number} maxEditDistance - */ - function addIntoResults(results, fullId, id, index, dist, path_dist, maxEditDistance) { - if (dist <= maxEditDistance || index !== -1) { - if (results.has(fullId)) { - const result = results.get(fullId); - if (result === undefined || result.dontValidate || result.dist <= dist) { - return; + return false; + }; + const aliasResults = await index.search(normalizedUserQuery); + if (aliasResults) { + for (const id of aliasResults.matches().entries()) { + const [name, alias] = await Promise.all([ + this.getName(id), + this.getAliasTarget(id), + ]); + if (name !== null && + alias !== null && + !idDuplicates.has(id) && + name.replace(/[_"]/g, "").toLowerCase() === normalizedUserQuery + ) { + prefixResults.push(handleAlias(name, alias, 0, 0)); + idDuplicates.add(id); + } } } - // @ts-expect-error - results.set(fullId, { - id: id, - index: index, - dontValidate: parsedQuery.literalSearch, - dist: dist, - path_dist: path_dist, - }); - } - } - - /** - * This function is called in case the query has more than one element. In this case, it'll - * try to match the items which validates all the elements. For `aa -> bb` will look for - * functions which have a parameter `aa` and has `bb` in its returned values. - * - * @param {rustdoc.Row} row - * @param {number} pos - Position in the `searchIndex`. - * @param {rustdoc.Results} results - */ - function handleArgs(row, pos, results) { - if (!row || (filterCrates !== null && row.crate !== filterCrates)) { - return; - } - const rowType = row.type; - if (!rowType) { - return; - } - - const tfpDist = compareTypeFingerprints( - row.id, - parsedQuery.typeFingerprint, - ); - if (tfpDist === null) { - return; - } - // @ts-expect-error - if (results.size >= MAX_RESULTS && tfpDist > results.max_dist) { - return; - } - - // If the result is too "bad", we return false and it ends this search. - if (!unifyFunctionTypes( - rowType.inputs, - parsedQuery.elems, - rowType.where_clause, - null, - // @ts-expect-error - mgens => { - return unifyFunctionTypes( - rowType.output, - parsedQuery.returned, - rowType.where_clause, - mgens, - checkTypeMgensForConflict, - 0, // unboxing depth - ); - }, - 0, // unboxing depth - )) { - return; - } - - results.max_dist = Math.max(results.max_dist || 0, tfpDist); - addIntoResults(results, row.id, pos, 0, tfpDist, 0, Number.MAX_VALUE); - } - - /** - * Compare the query fingerprint with the function fingerprint. - * - * @param {number} fullId - The function - * @param {Uint32Array} queryFingerprint - The query - * @returns {number|null} - Null if non-match, number if distance - * This function might return 0! - */ - const compareTypeFingerprints = (fullId, queryFingerprint) => { - const fh0 = this.functionTypeFingerprint[fullId * 4]; - const fh1 = this.functionTypeFingerprint[(fullId * 4) + 1]; - const fh2 = this.functionTypeFingerprint[(fullId * 4) + 2]; - const [qh0, qh1, qh2] = queryFingerprint; - // Approximate set intersection with bloom filters. - // This can be larger than reality, not smaller, because hashes have - // the property that if they've got the same value, they hash to the - // same thing. False positives exist, but not false negatives. - const [in0, in1, in2] = [fh0 & qh0, fh1 & qh1, fh2 & qh2]; - // Approximate the set of items in the query but not the function. - // This might be smaller than reality, but cannot be bigger. - // - // | in_ | qh_ | XOR | Meaning | - // | --- | --- | --- | ------------------------------------------------ | - // | 0 | 0 | 0 | Not present | - // | 1 | 0 | 1 | IMPOSSIBLE because `in_` is `fh_ & qh_` | - // | 1 | 1 | 0 | If one or both is false positive, false negative | - // | 0 | 1 | 1 | Since in_ has no false negatives, must be real | - if ((in0 ^ qh0) || (in1 ^ qh1) || (in2 ^ qh2)) { - return null; - } - return this.functionTypeFingerprint[(fullId * 4) + 3]; - }; - - - const innerRunQuery = () => { - if (parsedQuery.foundElems === 1 && !parsedQuery.hasReturnArrow) { + if (parsedQuery.error !== null || parsedQuery.elems.length === 0) { + yield* flush(prefixResults); + return; + } const elem = parsedQuery.elems[0]; - // use arrow functions to preserve `this`. - /** @type {function(number): void} */ - const handleNameSearch = id => { - const row = this.searchIndex[id]; - if (!typePassesFilter(elem.typeFilter, row.ty) || + const typeFilter = itemTypeFromName(elem.typeFilter); + /** + * @param {number} id + * @returns {Promise<rustdoc.PlainResultObject?>} + */ + const handleNameSearch = async id => { + const row = await this.getRow(id); + if (!row || !row.entry) { + return null; + } + if (!typePassesFilter(typeFilter, row.ty) || (filterCrates !== null && row.crate !== filterCrates)) { - return; + return null; } + /** @type {number|null} */ let pathDist = 0; if (elem.fullPath.length > 1) { - - const maybePathDist = checkPath(elem.pathWithoutLast, row); - if (maybePathDist === null) { - return; + pathDist = checkRowPath(elem.pathWithoutLast, row); + if (pathDist === null) { + return null; } - pathDist = maybePathDist; } if (parsedQuery.literalSearch) { - if (row.word === elem.pathLast) { - addIntoResults(results_others, row.id, id, 0, 0, pathDist, 0); - } + return row.name.toLowerCase() === elem.pathLast ? { + id, + dist: 0, + path_dist: 0, + index: 0, + elems: [], // only used in type-based queries + returned: [], // only used in type-based queries + is_alias: false, + } : null; } else { - addIntoResults( - results_others, - row.id, + return { id, - row.normalizedName.indexOf(elem.normalizedPathLast), - editDistance( + dist: editDistance( row.normalizedName, elem.normalizedPathLast, maxEditDistance, ), - pathDist, - maxEditDistance, - ); + path_dist: pathDist, + index: row.normalizedName.indexOf(elem.normalizedPathLast), + elems: [], // only used in type-based queries + returned: [], // only used in type-based queries + is_alias: false, + }; } }; - if (elem.normalizedPathLast !== "") { - const last = elem.normalizedPathLast; - for (const id of this.nameTrie.search(last, this.tailTable)) { - handleNameSearch(id); + if (elem.normalizedPathLast === "") { + // faster full-table scan for this specific case. + const nameData = this.database.getData("name"); + const l = nameData ? nameData.length : 0; + for (let id = 0; id < l; ++id) { + if (!idDuplicates.has(id)) { + idDuplicates.add(id); + prefixResults.push(handleNameSearch(id)); + } + if (yield* flush(prefixResults)) { + return; + } } + return; } - const length = this.searchIndex.length; - - for (let i = 0, nSearchIndex = length; i < nSearchIndex; ++i) { - // queries that end in :: bypass the trie - if (elem.normalizedPathLast === "") { - handleNameSearch(i); + const results = await index.search(elem.normalizedPathLast); + if (results) { + for await (const result of results.prefixMatches()) { + for (const id of result.entries()) { + if (!idDuplicates.has(id)) { + idDuplicates.add(id); + prefixResults.push(handleNameSearch(id)); + const [name, alias] = await Promise.all([ + this.getName(id), + this.getAliasTarget(id), + ]); + if (name !== null && alias !== null) { + prefixResults.push(handleAlias(name, alias, 0, 0)); + } + } + } + if (yield* flush(prefixResults)) { + return; + } } - const row = this.searchIndex[i]; - if (filterCrates !== null && row.crate !== filterCrates) { - continue; + if (yield* flush(prefixResults)) { + return; } - const tfpDist = compareTypeFingerprints( - row.id, - parsedQuery.typeFingerprint, - ); - if (tfpDist !== null) { - const in_args = row.type && row.type.inputs - && checkIfInList(row.type.inputs, elem, row.type.where_clause, null, 0); - const returned = row.type && row.type.output - && checkIfInList(row.type.output, elem, row.type.where_clause, null, 0); - if (in_args) { - results_in_args.max_dist = Math.max( - results_in_args.max_dist || 0, - tfpDist, - ); - const maxDist = results_in_args.size < MAX_RESULTS ? - (tfpDist + 1) : - results_in_args.max_dist; - addIntoResults(results_in_args, row.id, i, -1, tfpDist, 0, maxDist); - } - if (returned) { - results_returned.max_dist = Math.max( - results_returned.max_dist || 0, - tfpDist, - ); - const maxDist = results_returned.size < MAX_RESULTS ? - (tfpDist + 1) : - results_returned.max_dist; - addIntoResults(results_returned, row.id, i, -1, tfpDist, 0, maxDist); + } + const levSearchResults = index.searchLev(elem.normalizedPathLast); + const levResults = []; + for await (const levResult of levSearchResults) { + for (const id of levResult.matches().entries()) { + if (!idDuplicates.has(id)) { + idDuplicates.add(id); + levResults.push(handleNameSearch(id)); + const [name, alias] = await Promise.all([ + this.getName(id), + this.getAliasTarget(id), + ]); + if (name !== null && alias !== null) { + levResults.push(handleAlias( + name, + alias, + editDistance(elem.normalizedPathLast, name, maxEditDistance), + name.indexOf(elem.normalizedPathLast), + )); + } } } } - } else if (parsedQuery.foundElems > 0) { - // Sort input and output so that generic type variables go first and - // types with generic parameters go last. - // That's because of the way unification is structured: it eats off - // the end, and hits a fast path if the last item is a simple atom. - /** @type {function(rustdoc.QueryElement, rustdoc.QueryElement): number} */ - const sortQ = (a, b) => { - const ag = a.generics.length === 0 && a.bindings.size === 0; - const bg = b.generics.length === 0 && b.bindings.size === 0; - if (ag !== bg) { - // unary `+` converts booleans into integers. - return +ag - +bg; + yield* flush(levResults); + if (results) { + const substringResults = []; + for await (const result of results.substringMatches()) { + for (const id of result.entries()) { + if (!idDuplicates.has(id)) { + idDuplicates.add(id); + substringResults.push(handleNameSearch(id)); + const [name, alias] = await Promise.all([ + this.getName(id), + this.getAliasTarget(id), + ]); + if (name !== null && alias !== null) { + levResults.push(handleAlias( + name, + alias, + editDistance( + elem.normalizedPathLast, + name, + maxEditDistance, + ), + name.indexOf(elem.normalizedPathLast), + )); + } + } + } + if (yield* flush(substringResults)) { + return; + } } - const ai = a.id !== null && a.id > 0; - const bi = b.id !== null && b.id > 0; - return +ai - +bi; - }; - parsedQuery.elems.sort(sortQ); - parsedQuery.returned.sort(sortQ); - for (let i = 0, nSearchIndex = this.searchIndex.length; i < nSearchIndex; ++i) { - handleArgs(this.searchIndex[i], i, results_others); } } - }; + .bind(this); - if (parsedQuery.error === null) { - innerRunQuery(); - } + const innerRunTypeQuery = + /** + * @this {DocSearch} + * @param {rustdoc.ParserQueryElement[]} inputs + * @param {rustdoc.ParserQueryElement[]} output + * @param {"sig"|"elems"|"returned"|null} typeInfo + * @param {string} currentCrate + * @returns {AsyncGenerator<rustdoc.ResultObject>} + */ + async function*(inputs, output, typeInfo, currentCrate) { + const index = this.database.getIndex("normalizedName"); + if (!index) { + return; + } + /** @type {Map<string, number>} */ + const genericMap = new Map(); + /** + * @template Q + * @typedef {{ + * invertedIndex: stringdex.RoaringBitmap[], + * queryElem: Q, + * }} PostingsList + */ + /** @type {stringdex.RoaringBitmap[]} */ + const empty_inverted_index = []; + /** @type {PostingsList<any>[]} */ + const empty_postings_list = []; + /** @type {stringdex.RoaringBitmap[]} */ + const everything_inverted_index = []; + for (let i = 0; i < 64; ++i) { + everything_inverted_index.push(RoaringBitmap.everything()); + } + /** + * @type {PostingsList<rustdoc.QueryElement[]>} + */ + const everything_postings_list = { + invertedIndex: everything_inverted_index, + queryElem: [], + }; + /** + * @type {PostingsList<rustdoc.QueryElement[]>[]} + */ + const nested_everything_postings_list = [everything_postings_list]; + /** + * @param {...stringdex.RoaringBitmap[]} idx + * @returns {stringdex.RoaringBitmap[]} + */ + const intersectInvertedIndexes = (...idx) => { + let i = 0; + const l = idx.length; + while (i < l - 1 && idx[i] === everything_inverted_index) { + i += 1; + } + const result = [...idx[i]]; + for (; i < l; ++i) { + if (idx[i] === everything_inverted_index) { + continue; + } + if (idx[i].length < result.length) { + result.length = idx[i].length; + } + for (let j = 0; j < result.length; ++j) { + result[j] = result[j].intersection(idx[i][j]); + } + } + return result; + }; + /** + * Fetch a bitmap of potentially-matching functions, + * plus a list of query elements annotated with the correct IDs. + * + * More than one ID can exist because, for example, q=`Iter` can match + * `std::vec::Iter`, or `std::btree_set::Iter`, or anything else, and those + * items different IDs. What's worse, q=`Iter<Iter>` has N**2 possible + * matches, because it could be `vec::Iter<btree_set::Iter>`, + * `btree_set::Iter<vec::Iter>`, `vec::Iter<vec::Iter>`, + * `btree_set::Iter<btree_set::Iter>`, + * or anything else. This function returns all possible permutations. + * + * @param {rustdoc.ParserQueryElement|null} elem + * @returns {Promise<PostingsList<rustdoc.QueryElement>[]>} + */ + const unpackPostingsList = async elem => { + if (!elem) { + return empty_postings_list; + } + const typeFilter = itemTypeFromName(elem.typeFilter); + const searchResults = await index.search(elem.normalizedPathLast); + /** + * @type {Promise<[ + * number, + * string|null, + * rustdoc.TypeData|null, + * rustdoc.PathData|null, + * ]>[]} + * */ + const typePromises = []; + if (typeFilter !== TY_GENERIC && searchResults) { + for (const id of searchResults.matches().entries()) { + typePromises.push(Promise.all([ + this.getName(id), + this.getTypeData(id), + this.getPathData(id), + ]).then(([name, typeData, pathData]) => + [id, name, typeData, pathData])); + } + } + const types = (await Promise.all(typePromises)) + .filter(([_id, name, ty, path]) => + name !== null && name.toLowerCase() === elem.pathLast && + ty && !ty.invertedFunctionSignatureIndex.every(bitmap => { + return bitmap.isEmpty(); + }) && + path && path.ty !== TY_ASSOCTYPE && + (elem.pathWithoutLast.length === 0 || + checkPath( + elem.pathWithoutLast, + path.modulePath.split("::"), + ) === 0), + ); + if (types.length === 0) { + const areGenericsAllowed = typeFilter === TY_GENERIC || ( + typeFilter === -1 && + (parsedQuery.totalElems > 1 || parsedQuery.hasReturnArrow) && + elem.pathWithoutLast.length === 0 && + elem.generics.length === 0 && + elem.bindings.size === 0 + ); + if (typeFilter !== TY_GENERIC && + (elem.name.length >= 3 || !areGenericsAllowed) + ) { + /** @type {string|null} */ + let chosenName = null; + /** @type {rustdoc.TypeData[]} */ + let chosenType = []; + /** @type {rustdoc.PathData[]} */ + let chosenPath = []; + /** @type {number[]} */ + let chosenId = []; + let chosenDist = Number.MAX_SAFE_INTEGER; + const levResults = index.searchLev(elem.normalizedPathLast); + for await (const searchResults of levResults) { + for (const id of searchResults.matches().entries()) { + const [name, ty, path] = await Promise.all([ + this.getName(id), + this.getTypeData(id), + this.getPathData(id), + ]); + if (name !== null && ty !== null && path !== null && + !ty.invertedFunctionSignatureIndex.every(bitmap => { + return bitmap.isEmpty(); + }) && + path.ty !== TY_ASSOCTYPE + ) { + let dist = editDistance( + name, + elem.pathLast, + maxEditDistance, + ); + if (elem.pathWithoutLast.length !== 0) { + const pathDist = checkPath( + elem.pathWithoutLast, + path.modulePath.split("::"), + ); + // guaranteed to be higher than the path limit + dist += pathDist === null ? + Number.MAX_SAFE_INTEGER : + pathDist; + } + if (name === chosenName) { + chosenId.push(id); + chosenType.push(ty); + chosenPath.push(path); + } else if (dist < chosenDist) { + chosenName = name; + chosenId = [id]; + chosenType = [ty]; + chosenPath = [path]; + chosenDist = dist; + } + } + } + if (chosenId.length !== 0) { + // searchLev returns results in order + // if we have working matches, we're done + break; + } + } + if (areGenericsAllowed) { + parsedQuery.proposeCorrectionFrom = elem.name; + parsedQuery.proposeCorrectionTo = chosenName; + } else { + parsedQuery.correction = chosenName; + for (let i = 0; i < chosenType.length; ++i) { + types.push([ + chosenId[i], + chosenName, + chosenType[i], + chosenPath[i], + ]); + } + } + } + if (areGenericsAllowed) { + let genericId = genericMap.get(elem.normalizedPathLast); + if (genericId === undefined) { + genericId = genericMap.size; + genericMap.set(elem.normalizedPathLast, genericId); + } + return [{ + invertedIndex: await this.getGenericInvertedIndex(genericId), + queryElem: { + name: elem.name, + id: (-genericId) - 1, + typeFilter: TY_GENERIC, + generics: [], + bindings: EMPTY_BINDINGS_MAP, + fullPath: elem.fullPath, + pathLast: elem.pathLast, + normalizedPathLast: elem.normalizedPathLast, + pathWithoutLast: elem.pathWithoutLast, + }, + }]; + } + } + types.sort(([_i, name1, _t, pathData1], [_i2, name2, _t2, pathData2]) => { + const p1 = !pathData1 ? "" : pathData1.modulePath; + const p2 = !pathData2 ? "" : pathData2.modulePath; + const n1 = name1 === null ? "" : name1; + const n2 = name2 === null ? "" : name2; + if (p1.length !== p2.length) { + return p1.length > p2.length ? +1 : -1; + } + if (n1.length !== n2.length) { + return n1.length > n2.length ? +1 : -1; + } + if (n1 !== n2) { + return n1 > n2 ? +1 : -1; + } + if (p1 !== p2) { + return p1 > p2 ? +1 : -1; + } + return 0; + }); + /** @type {PostingsList<rustdoc.QueryElement>[]} */ + const results = []; + for (const [id, _name, typeData] of types) { + if (!typeData || typeData.invertedFunctionSignatureIndex.every(bitmap => { + return bitmap.isEmpty(); + })) { + continue; + } + const upla = await unpackPostingsListAll(elem.generics); + const uplb = await unpackPostingsListBindings(elem.bindings); + for (const {invertedIndex: genericsIdx, queryElem: generics} of upla) { + for (const {invertedIndex: bindingsIdx, queryElem: bindings} of uplb) { + results.push({ + invertedIndex: intersectInvertedIndexes( + typeData.invertedFunctionSignatureIndex, + genericsIdx, + bindingsIdx, + ), + queryElem: { + name: elem.name, + id, + typeFilter, + generics, + bindings, + fullPath: elem.fullPath, + pathLast: elem.pathLast, + normalizedPathLast: elem.normalizedPathLast, + pathWithoutLast: elem.pathWithoutLast, + }, + }); + if ((results.length & 0x7F) === 0) { + await yieldToBrowser(); + } + } + } + } + return results; + }; + /** + * Fetch all possible matching permutations of a list of query elements. + * + * The empty list returns an "identity postings list", with a bitmap that + * matches everything and an empty list of elems. This allows you to safely + * take the intersection of this bitmap. + * + * @param {(rustdoc.ParserQueryElement|null)[]|null} elems + * @returns {Promise<PostingsList<rustdoc.QueryElement[]>[]>} + */ + const unpackPostingsListAll = async elems => { + if (!elems || elems.length === 0) { + return nested_everything_postings_list; + } + const [firstPostingsList, remainingAll] = await Promise.all([ + unpackPostingsList(elems[0]), + unpackPostingsListAll(elems.slice(1)), + ]); + /** @type {PostingsList<rustdoc.QueryElement[]>[]} */ + const results = []; + for (const { + invertedIndex: firstIdx, + queryElem: firstElem, + } of firstPostingsList) { + for (const { + invertedIndex: remainingIdx, + queryElem: remainingElems, + } of remainingAll) { + results.push({ + invertedIndex: intersectInvertedIndexes(firstIdx, remainingIdx), + queryElem: [firstElem, ...remainingElems], + }); + if ((results.length & 0x7F) === 0) { + await yieldToBrowser(); + } + } + } + return results; + }; + /** + * Fetch all possible matching permutations of a map query element bindings. + * + * The empty list returns an "identity postings list", with a bitmap that + * matches everything and an empty list of elems. This allows you to safely + * take the intersection of this bitmap. + * + * Heads up! This function mutates the Map that you provide. + * Before passing an actual parser item to it, make sure to clone the map. + * + * @param {Map<string, rustdoc.ParserQueryElement[]>} elems + * @returns {Promise<PostingsList< + * Map<number, rustdoc.QueryElement[]>, + * >[]>} + */ + const unpackPostingsListBindings = async elems => { + if (!elems) { + return [{ + invertedIndex: everything_inverted_index, + queryElem: new Map(), + }]; + } + const firstKey = elems.keys().next().value; + if (firstKey === undefined) { + return [{ + invertedIndex: everything_inverted_index, + queryElem: new Map(), + }]; + } + const firstList = elems.get(firstKey); + if (firstList === undefined) { + return [{ + invertedIndex: everything_inverted_index, + queryElem: new Map(), + }]; + } + const firstKeyIds = await index.search(firstKey); + if (!firstKeyIds) { + // User specified a non-existent key. + return [{ + invertedIndex: empty_inverted_index, + queryElem: new Map(), + }]; + } + elems.delete(firstKey); + const [firstPostingsList, remainingAll] = await Promise.all([ + unpackPostingsListAll(firstList), + unpackPostingsListBindings(elems), + ]); + /** @type {PostingsList<Map<number, rustdoc.QueryElement[]>>[]} */ + const results = []; + for (const keyId of firstKeyIds.matches().entries()) { + for (const { + invertedIndex: firstIdx, + queryElem: firstElem, + } of firstPostingsList) { + for (const { + invertedIndex: remainingIdx, + queryElem: remainingElems, + } of remainingAll) { + const elems = new Map(remainingElems); + elems.set(keyId, firstElem); + results.push({ + invertedIndex: intersectInvertedIndexes(firstIdx, remainingIdx), + queryElem: elems, + }); + if ((results.length & 0x7F) === 0) { + await yieldToBrowser(); + } + } + } + } + elems.set(firstKey, firstList); + if (results.length === 0) { + // User specified a non-existent key. + return [{ + invertedIndex: empty_inverted_index, + queryElem: new Map(), + }]; + } + return results; + }; - const isType = parsedQuery.foundElems !== 1 || parsedQuery.hasReturnArrow; - const [sorted_in_args, sorted_returned, sorted_others] = await Promise.all([ - sortResults(results_in_args, "elems", currentCrate), - sortResults(results_returned, "returned", currentCrate), - // @ts-expect-error - sortResults(results_others, (isType ? "query" : null), currentCrate), - ]); - const ret = createQueryResults( - sorted_in_args, - sorted_returned, - sorted_others, - parsedQuery); - await handleAliases(ret, parsedQuery.userQuery.replace(/"/g, ""), - filterCrates, currentCrate); - await Promise.all([ret.others, ret.returned, ret.in_args].map(async list => { - const descs = await Promise.all(list.map(result => { - // @ts-expect-error - return this.searchIndexEmptyDesc.get(result.crate).contains(result.bitIndex) ? - "" : - // @ts-expect-error - this.searchState.loadDesc(result); - })); - for (const [i, result] of list.entries()) { - // @ts-expect-error - result.desc = descs[i]; + // finally, we can do the actual unification loop + const [allInputs, allOutput] = await Promise.all([ + unpackPostingsListAll(inputs), + unpackPostingsListAll(output), + ]); + let checkCounter = 0; + /** + * Finally, we can perform an incremental search, sorted by the number of + * entries that match a given query. + * + * The outer list gives the number of elements. The inner one is separate + * for each distinct name resolution. + * + * @type {{ + * bitmap: stringdex.RoaringBitmap, + * inputs: rustdoc.QueryElement[], + * output: rustdoc.QueryElement[], + * }[][]} + */ + const queryPlan = []; + for (const {invertedIndex: inputsIdx, queryElem: inputs} of allInputs) { + for (const {invertedIndex: outputIdx, queryElem: output} of allOutput) { + const invertedIndex = intersectInvertedIndexes(inputsIdx, outputIdx); + for (const [size, bitmap] of invertedIndex.entries()) { + checkCounter += 1; + if ((checkCounter & 0x7F) === 0) { + await yieldToBrowser(); + } + if (!queryPlan[size]) { + queryPlan[size] = []; + } + queryPlan[size].push({ + bitmap, inputs, output, + }); + } + } + } + const resultPromises = []; + const dedup = new Set(); + let resultCounter = 0; + const isReturnTypeQuery = inputs.length === 0; + /** @type {rustdoc.PlainResultObject[]} */ + const pushToBottom = []; + plan: for (const queryStep of queryPlan) { + for (const {bitmap, inputs, output} of queryStep) { + for (const id of bitmap.entries()) { + checkCounter += 1; + if ((checkCounter & 0x7F) === 0) { + await yieldToBrowser(); + } + resultPromises.push(this.getFunctionData(id).then(async fnData => { + if (!fnData || !fnData.functionSignature) { + return null; + } + checkCounter += 1; + if ((checkCounter & 0x7F) === 0) { + await yieldToBrowser(); + } + const functionSignature = fnData.functionSignature; + if (!unifyFunctionTypes( + functionSignature.inputs, + inputs, + functionSignature.where_clause, + null, + mgens => { + return !!unifyFunctionTypes( + functionSignature.output, + output, + functionSignature.where_clause, + mgens, + checkTypeMgensForConflict, + 0, // unboxing depth + ); + }, + 0, // unboxing depth + )) { + return null; + } + const result = { + id, + dist: fnData.elemCount, + path_dist: 0, + index: -1, + elems: inputs, + returned: output, + is_alias: false, + }; + const entry = await this.getEntryData(id); + if ((entry && !isFnLikeTy(entry.ty)) || + (isReturnTypeQuery && + functionSignature && + containsTypeFromQuery( + output, + functionSignature.inputs, + functionSignature.where_clause, + ) + ) + ) { + pushToBottom.push(result); + return null; + } + return result; + })); + } + } + for await (const result of sortAndTransformResults( + await Promise.all(resultPromises), + typeInfo, + currentCrate, + dedup, + )) { + if (resultCounter >= MAX_RESULTS) { + break plan; + } + yield result; + resultCounter += 1; + } + resultPromises.length = 0; + } + if (resultCounter >= MAX_RESULTS) { + return; + } + for await (const result of sortAndTransformResults( + await Promise.all(pushToBottom), + typeInfo, + currentCrate, + dedup, + )) { + if (resultCounter >= MAX_RESULTS) { + break; + } + yield result; + resultCounter += 1; + } } - })); - if (parsedQuery.error !== null && ret.others.length !== 0) { - // It means some doc aliases were found so let's "remove" the error! - ret.query.error = null; + .bind(this); + + if (parsedQuery.foundElems === 1 && !parsedQuery.hasReturnArrow) { + // We never want the main tab to delay behind the other two tabs. + // This is a bit of a hack (because JS's scheduler doesn't have much of an API), + // along with making innerRunTypeQuery yield to the UI thread. + const { + promise: donePromise, + resolve: doneResolve, + reject: doneReject, + } = Promise.withResolvers(); + const doneTimeout = timeout(250); + return { + "in_args": (async function*() { + await Promise.race([donePromise, doneTimeout]); + yield* innerRunTypeQuery(parsedQuery.elems, [], "elems", currentCrate); + })(), + "returned": (async function*() { + await Promise.race([donePromise, doneTimeout]); + yield* innerRunTypeQuery([], parsedQuery.elems, "returned", currentCrate); + })(), + "others": (async function*() { + try { + yield* innerRunNameQuery(currentCrate); + doneResolve(null); + } catch (e) { + doneReject(e); + throw e; + } + })(), + "query": parsedQuery, + }; + } else if (parsedQuery.error !== null) { + return { + "in_args": (async function*() {})(), + "returned": (async function*() {})(), + "others": innerRunNameQuery(currentCrate), + "query": parsedQuery, + }; + } else { + const typeInfo = parsedQuery.elems.length === 0 ? + "returned" : ( + parsedQuery.returned.length === 0 ? "elems" : "sig" + ); + return { + "in_args": (async function*() {})(), + "returned": (async function*() {})(), + "others": parsedQuery.foundElems === 0 ? + (async function*() {})() : + innerRunTypeQuery( + parsedQuery.elems, + parsedQuery.returned, + typeInfo, + currentCrate, + ), + "query": parsedQuery, + }; } - return ret; } } // ==================== Core search logic end ==================== -/** @type {Map<string, rustdoc.RawSearchIndexCrate>} */ -let rawSearchIndex; -// @ts-expect-error +/** @type {DocSearch} */ let docSearch; const longItemTypes = [ "keyword", @@ -4762,12 +4596,8 @@ function buildUrl(search, filterCrates) { function getFilterCrates() { const elem = document.getElementById("crate-search"); - if (elem && - // @ts-expect-error - elem.value !== "all crates" && - // @ts-expect-error - window.searchIndex.has(elem.value) - ) { + // @ts-expect-error + if (elem && elem.value !== "all crates") { // @ts-expect-error return elem.value; } @@ -4777,8 +4607,7 @@ function getFilterCrates() { // @ts-expect-error function nextTab(direction) { const next = (searchState.currentTab + direction + 3) % searchState.focusedByTab.length; - // @ts-expect-error - searchState.focusedByTab[searchState.currentTab] = document.activeElement; + window.searchState.focusedByTab[searchState.currentTab] = document.activeElement; printTab(next); focusSearchResult(); } @@ -4790,133 +4619,182 @@ function focusSearchResult() { document.querySelectorAll(".search-results.active a").item(0) || document.querySelectorAll("#search-tabs button").item(searchState.currentTab); searchState.focusedByTab[searchState.currentTab] = null; - if (target) { - // @ts-expect-error + if (target && target instanceof HTMLElement) { target.focus(); } } /** * Render a set of search results for a single tab. - * @param {Array<?>} array - The search results for this tab - * @param {rustdoc.ParsedQuery<rustdoc.QueryElement>} query + * @param {AsyncGenerator<rustdoc.ResultObject>} results - The search results for this tab + * @param {rustdoc.ParsedQuery<rustdoc.ParserQueryElement>} query * @param {boolean} display - True if this is the active tab + * @param {function(number, HTMLElement): any} finishedCallback + * @param {boolean} isTypeSearch + * @returns {Promise<HTMLElement>} */ -async function addTab(array, query, display) { +async function addTab(results, query, display, finishedCallback, isTypeSearch) { const extraClass = display ? " active" : ""; - const output = document.createElement( - array.length === 0 && query.error === null ? "div" : "ul", - ); - if (array.length > 0) { - output.className = "search-results " + extraClass; + /** @type {HTMLElement} */ + let output = document.createElement("ul"); + output.className = "search-results " + extraClass; - const lis = Promise.all(array.map(async item => { - const name = item.is_alias ? item.original.name : item.name; - const type = itemTypes[item.ty]; - const longType = longItemTypes[item.ty]; - const typeName = longType.length !== 0 ? `${longType}` : "?"; + let count = 0; + + /** @type {Promise<string|null>[]} */ + const descList = []; + + /** @param {rustdoc.ResultObject} obj */ + const addNextResultToOutput = async obj => { + count += 1; + + const name = obj.item.name; + const type = itemTypes[obj.item.ty]; + const longType = longItemTypes[obj.item.ty]; + const typeName = longType.length !== 0 ? `${longType}` : "?"; - const link = document.createElement("a"); - link.className = "result-" + type; - link.href = item.href; + const link = document.createElement("a"); + link.className = "result-" + type; + link.href = obj.href; - const resultName = document.createElement("span"); - resultName.className = "result-name"; + const resultName = document.createElement("span"); + resultName.className = "result-name"; - resultName.insertAdjacentHTML( - "beforeend", - `<span class="typename">${typeName}</span>`); - link.appendChild(resultName); + resultName.insertAdjacentHTML( + "beforeend", + `<span class="typename">${typeName}</span>`); + link.appendChild(resultName); - let alias = " "; - if (item.is_alias) { - alias = ` <div class="alias">\ -<b>${item.name}</b><i class="grey"> - see </i>\ + let alias = " "; + if (obj.alias !== undefined) { + alias = ` <div class="alias">\ +<b>${obj.alias}</b><i class="grey"> - see </i>\ </div>`; - } - resultName.insertAdjacentHTML( - "beforeend", - `<div class="path">${alias}\ -${item.displayPath}<span class="${type}">${name}</span>\ + } + resultName.insertAdjacentHTML( + "beforeend", + `<div class="path">${alias}\ +${obj.displayPath}<span class="${type}">${name}</span>\ </div>`); - const description = document.createElement("div"); - description.className = "desc"; - description.insertAdjacentHTML("beforeend", item.desc); - if (item.displayTypeSignature) { - const {type, mappedNames, whereClause} = await item.displayTypeSignature; - const displayType = document.createElement("div"); - // @ts-expect-error - type.forEach((value, index) => { - if (index % 2 !== 0) { - const highlight = document.createElement("strong"); - highlight.appendChild(document.createTextNode(value)); - displayType.appendChild(highlight); - } else { - displayType.appendChild(document.createTextNode(value)); + const description = document.createElement("div"); + description.className = "desc"; + obj.desc.then(desc => { + if (desc !== null) { + description.insertAdjacentHTML("beforeend", desc); + } + }); + descList.push(obj.desc); + if (obj.displayTypeSignature) { + const {type, mappedNames, whereClause} = await obj.displayTypeSignature; + const displayType = document.createElement("div"); + type.forEach((value, index) => { + if (index % 2 !== 0) { + const highlight = document.createElement("strong"); + highlight.appendChild(document.createTextNode(value)); + displayType.appendChild(highlight); + } else { + displayType.appendChild(document.createTextNode(value)); + } + }); + if (mappedNames.size > 0 || whereClause.size > 0) { + let addWhereLineFn = () => { + const line = document.createElement("div"); + line.className = "where"; + line.appendChild(document.createTextNode("where")); + displayType.appendChild(line); + addWhereLineFn = () => {}; + }; + for (const [qname, name] of mappedNames) { + // don't care unless the generic name is different + if (name === qname) { + continue; } - }); - if (mappedNames.size > 0 || whereClause.size > 0) { - let addWhereLineFn = () => { - const line = document.createElement("div"); - line.className = "where"; - line.appendChild(document.createTextNode("where")); - displayType.appendChild(line); - addWhereLineFn = () => {}; - }; - for (const [qname, name] of mappedNames) { - // don't care unless the generic name is different - if (name === qname) { - continue; - } - addWhereLineFn(); - const line = document.createElement("div"); - line.className = "where"; - line.appendChild(document.createTextNode(` ${qname} matches `)); - const lineStrong = document.createElement("strong"); - lineStrong.appendChild(document.createTextNode(name)); - line.appendChild(lineStrong); - displayType.appendChild(line); + addWhereLineFn(); + const line = document.createElement("div"); + line.className = "where"; + line.appendChild(document.createTextNode(` ${qname} matches `)); + const lineStrong = document.createElement("strong"); + lineStrong.appendChild(document.createTextNode(name)); + line.appendChild(lineStrong); + displayType.appendChild(line); + } + for (const [name, innerType] of whereClause) { + // don't care unless there's at least one highlighted entry + if (innerType.length <= 1) { + continue; } - for (const [name, innerType] of whereClause) { - // don't care unless there's at least one highlighted entry - if (innerType.length <= 1) { - continue; + addWhereLineFn(); + const line = document.createElement("div"); + line.className = "where"; + line.appendChild(document.createTextNode(` ${name}: `)); + innerType.forEach((value, index) => { + if (index % 2 !== 0) { + const highlight = document.createElement("strong"); + highlight.appendChild(document.createTextNode(value)); + line.appendChild(highlight); + } else { + line.appendChild(document.createTextNode(value)); } - addWhereLineFn(); - const line = document.createElement("div"); - line.className = "where"; - line.appendChild(document.createTextNode(` ${name}: `)); - // @ts-expect-error - innerType.forEach((value, index) => { - if (index % 2 !== 0) { - const highlight = document.createElement("strong"); - highlight.appendChild(document.createTextNode(value)); - line.appendChild(highlight); - } else { - line.appendChild(document.createTextNode(value)); - } - }); - displayType.appendChild(line); - } + }); + displayType.appendChild(line); } - displayType.className = "type-signature"; - link.appendChild(displayType); } + displayType.className = "type-signature"; + link.appendChild(displayType); + } - link.appendChild(description); - return link; - })); - lis.then(lis => { - for (const li of lis) { - output.appendChild(li); + link.appendChild(description); + output.appendChild(link); + + results.next().then(async nextResult => { + if (nextResult.value) { + addNextResultToOutput(nextResult.value); + } else { + await Promise.all(descList); + // need to make sure the element is shown before + // running this callback + yieldToBrowser().then(() => finishedCallback(count, output)); } }); - } else if (query.error === null) { - const dlroChannel = `https://doc.rust-lang.org/${getVar("channel")}`; + }; + const firstResult = await results.next(); + let correctionOutput = ""; + if (query.correction !== null && isTypeSearch) { + const orig = query.returned.length > 0 + ? query.returned[0].name + : query.elems[0].name; + correctionOutput = "<h3 class=\"search-corrections\">" + + `Type "${orig}" not found. ` + + "Showing results for closest type name " + + `"${query.correction}" instead.</h3>`; + } + if (query.proposeCorrectionFrom !== null && isTypeSearch) { + const orig = query.proposeCorrectionFrom; + const targ = query.proposeCorrectionTo; + correctionOutput = "<h3 class=\"search-corrections\">" + + `Type "${orig}" not found and used as generic parameter. ` + + `Consider searching for "${targ}" instead.</h3>`; + } + if (firstResult.value) { + if (correctionOutput !== "") { + const h3 = document.createElement("h3"); + h3.innerHTML = correctionOutput; + output.appendChild(h3); + } + await addNextResultToOutput(firstResult.value); + } else { + output = document.createElement("div"); + if (correctionOutput !== "") { + const h3 = document.createElement("h3"); + h3.innerHTML = correctionOutput; + output.appendChild(h3); + } output.className = "search-failed" + extraClass; - output.innerHTML = "No results :(<br/>" + + const dlroChannel = `https://doc.rust-lang.org/${getVar("channel")}`; + if (query.userQuery !== "") { + output.innerHTML += "No results :(<br/>" + "Try on <a href=\"https://duckduckgo.com/?q=" + encodeURIComponent("rust " + query.userQuery) + "\">DuckDuckGo</a>?<br/><br/>" + @@ -4929,192 +4807,198 @@ ${item.displayPath}<span class="${type}">${name}</span>\ "introductions to language features and the language itself.</li><li><a " + "href=\"https://docs.rs\">Docs.rs</a> for documentation of crates released on" + " <a href=\"https://crates.io/\">crates.io</a>.</li></ul>"; + } + output.innerHTML += "Example searches:<ul>" + + "<li><a href=\"" + getNakedUrl() + "?search=std::vec\">std::vec</a></li>" + + "<li><a href=\"" + getNakedUrl() + "?search=u32+->+bool\">u32 -> bool</a></li>" + + "<li><a href=\"" + getNakedUrl() + "?search=Option<T>,+(T+->+U)+->+Option<U>\">" + + "Option<T>, (T -> U) -> Option<U></a></li>" + + "</ul>"; + // need to make sure the element is shown before + // running this callback + yieldToBrowser().then(() => finishedCallback(0, output)); } return output; } -// @ts-expect-error -function makeTabHeader(tabNb, text, nbElems) { - // https://blog.horizon-eda.org/misc/2020/02/19/ui.html - // - // CSS runs with `font-variant-numeric: tabular-nums` to ensure all - // digits are the same width. \u{2007} is a Unicode space character - // that is defined to be the same width as a digit. - const fmtNbElems = - nbElems < 10 ? `\u{2007}(${nbElems})\u{2007}\u{2007}` : - nbElems < 100 ? `\u{2007}(${nbElems})\u{2007}` : `\u{2007}(${nbElems})`; - if (searchState.currentTab === tabNb) { - return "<button class=\"selected\">" + text + - "<span class=\"count\">" + fmtNbElems + "</span></button>"; - } - return "<button>" + text + "<span class=\"count\">" + fmtNbElems + "</span></button>"; +/** + * returns [tab, output] + * @param {number} tabNb + * @param {string} text + * @param {AsyncGenerator<rustdoc.ResultObject>} results + * @param {rustdoc.ParsedQuery<rustdoc.ParserQueryElement>} query + * @param {boolean} isTypeSearch + * @param {boolean} goToFirst + * @returns {[HTMLElement, Promise<HTMLElement>]} + */ +function makeTab(tabNb, text, results, query, isTypeSearch, goToFirst) { + const isCurrentTab = window.searchState.currentTab === tabNb; + const tabButton = document.createElement("button"); + tabButton.appendChild(document.createTextNode(text)); + tabButton.className = isCurrentTab ? "selected" : ""; + const tabCount = document.createElement("span"); + tabCount.className = "count loading"; + tabCount.innerHTML = "\u{2007}(\u{2007})\u{2007}\u{2007}"; + tabButton.appendChild(tabCount); + return [ + tabButton, + addTab(results, query, isCurrentTab, (count, output) => { + const search = window.searchState.outputElement(); + const error = query.error; + if (count === 0 && error !== null && search) { + error.forEach((value, index) => { + value = value.split("<").join("<").split(">").join(">"); + if (index % 2 !== 0) { + error[index] = `<code>${value.replaceAll(" ", " ")}</code>`; + } else { + error[index] = value; + } + }); + const errorReport = document.createElement("h3"); + errorReport.className = "error"; + errorReport.innerHTML = `Query parser error: "${error.join("")}".`; + search.insertBefore(errorReport, search.firstElementChild); + } else if (goToFirst || + (count === 1 && getSettingValue("go-to-only-result") === "true") + ) { + // Needed to force re-execution of JS when coming back to a page. Let's take this + // scenario as example: + // + // 1. You have the "Directly go to item in search if there is only one result" + // option enabled. + // 2. You make a search which results only one result, leading you automatically to + // this result. + // 3. You go back to previous page. + // + // Now, without the call below, the JS will not be re-executed and the previous + // state will be used, starting search again since the search input is not empty, + // leading you back to the previous page again. + window.onunload = () => { }; + window.searchState.removeQueryParameters(); + const a = output.querySelector("a"); + if (a) { + a.click(); + return; + } + } + + // https://blog.horizon-eda.org/misc/2020/02/19/ui.html + // + // CSS runs with `font-variant-numeric: tabular-nums` to ensure all + // digits are the same width. \u{2007} is a Unicode space character + // that is defined to be the same width as a digit. + const fmtNbElems = + count < 10 ? `\u{2007}(${count})\u{2007}\u{2007}` : + count < 100 ? `\u{2007}(${count})\u{2007}` : `\u{2007}(${count})`; + tabCount.innerHTML = fmtNbElems; + tabCount.className = "count"; + }, isTypeSearch), + ]; } /** + * @param {DocSearch} docSearch * @param {rustdoc.ResultsTable} results - * @param {boolean} go_to_first + * @param {boolean} goToFirst * @param {string} filterCrates */ -async function showResults(results, go_to_first, filterCrates) { - const search = searchState.outputElement(); - if (go_to_first || (results.others.length === 1 - && getSettingValue("go-to-only-result") === "true") - ) { - // Needed to force re-execution of JS when coming back to a page. Let's take this - // scenario as example: - // - // 1. You have the "Directly go to item in search if there is only one result" option - // enabled. - // 2. You make a search which results only one result, leading you automatically to - // this result. - // 3. You go back to previous page. - // - // Now, without the call below, the JS will not be re-executed and the previous state - // will be used, starting search again since the search input is not empty, leading you - // back to the previous page again. - window.onunload = () => { }; - searchState.removeQueryParameters(); - const elem = document.createElement("a"); - elem.href = results.others[0].href; - removeClass(elem, "active"); - // For firefox, we need the element to be in the DOM so it can be clicked. - document.body.appendChild(elem); - elem.click(); - return; - } - if (results.query === undefined) { - // @ts-expect-error - results.query = DocSearch.parseQuery(searchState.input.value); - } +async function showResults(docSearch, results, goToFirst, filterCrates) { + const search = window.searchState.outputElement(); - currentResults = results.query.userQuery; - - // Navigate to the relevant tab if the current tab is empty, like in case users search - // for "-> String". If they had selected another tab previously, they have to click on - // it again. - let currentTab = searchState.currentTab; - if ((currentTab === 0 && results.others.length === 0) || - (currentTab === 1 && results.in_args.length === 0) || - (currentTab === 2 && results.returned.length === 0)) { - if (results.others.length !== 0) { - currentTab = 0; - } else if (results.in_args.length) { - currentTab = 1; - } else if (results.returned.length) { - currentTab = 2; - } + if (!search) { + return; } let crates = ""; - if (rawSearchIndex.size > 1) { - crates = "<div class=\"sub-heading\"> in <div id=\"crate-search-div\">" + + const crateNames = await docSearch.getCrateNameList(); + if (crateNames.length > 1) { + crates = " in <div id=\"crate-search-div\">" + "<select id=\"crate-search\"><option value=\"all crates\">all crates</option>"; - for (const c of rawSearchIndex.keys()) { + const l = crateNames.length; + for (let i = 0; i < l; i += 1) { + const c = crateNames[i]; crates += `<option value="${c}" ${c === filterCrates && "selected"}>${c}</option>`; } - crates += "</select></div></div>"; + crates += "</select></div>"; } + nonnull(document.querySelector(".search-switcher")).innerHTML = `Search results${crates}`; - let output = `<div class="main-heading">\ - <h1 class="search-results-title">Results</h1>${crates}</div>`; + /** @type {[HTMLElement, Promise<HTMLElement>][]} */ + const tabs = []; + searchState.currentTab = 0; if (results.query.error !== null) { - const error = results.query.error; - // @ts-expect-error - error.forEach((value, index) => { - value = value.split("<").join("<").split(">").join(">"); - if (index % 2 !== 0) { - error[index] = `<code>${value.replaceAll(" ", " ")}</code>`; - } else { - error[index] = value; - } - }); - output += `<h3 class="error">Query parser error: "${error.join("")}".</h3>`; - output += "<div id=\"search-tabs\">" + - makeTabHeader(0, "In Names", results.others.length) + - "</div>"; - currentTab = 0; - } else if (results.query.foundElems <= 1 && results.query.returned.length === 0) { - output += "<div id=\"search-tabs\">" + - makeTabHeader(0, "In Names", results.others.length) + - makeTabHeader(1, "In Parameters", results.in_args.length) + - makeTabHeader(2, "In Return Types", results.returned.length) + - "</div>"; + tabs.push(makeTab(0, "In Names", results.others, results.query, false, goToFirst)); + } else if ( + results.query.foundElems <= 1 && + results.query.returned.length === 0 && + !results.query.hasReturnArrow + ) { + tabs.push(makeTab(0, "In Names", results.others, results.query, false, goToFirst)); + tabs.push(makeTab(1, "In Parameters", results.in_args, results.query, true, false)); + tabs.push(makeTab(2, "In Return Types", results.returned, results.query, true, false)); } else { const signatureTabTitle = results.query.elems.length === 0 ? "In Function Return Types" : results.query.returned.length === 0 ? "In Function Parameters" : "In Function Signatures"; - output += "<div id=\"search-tabs\">" + - makeTabHeader(0, signatureTabTitle, results.others.length) + - "</div>"; - currentTab = 0; + tabs.push(makeTab(0, signatureTabTitle, results.others, results.query, true, goToFirst)); } - if (results.query.correction !== null) { - const orig = results.query.returned.length > 0 - ? results.query.returned[0].name - : results.query.elems[0].name; - output += "<h3 class=\"search-corrections\">" + - `Type "${orig}" not found. ` + - "Showing results for closest type name " + - `"${results.query.correction}" instead.</h3>`; - } - if (results.query.proposeCorrectionFrom !== null) { - const orig = results.query.proposeCorrectionFrom; - const targ = results.query.proposeCorrectionTo; - output += "<h3 class=\"search-corrections\">" + - `Type "${orig}" not found and used as generic parameter. ` + - `Consider searching for "${targ}" instead.</h3>`; - } - - const [ret_others, ret_in_args, ret_returned] = await Promise.all([ - addTab(results.others, results.query, currentTab === 0), - addTab(results.in_args, results.query, currentTab === 1), - addTab(results.returned, results.query, currentTab === 2), - ]); + const tabsElem = document.createElement("div"); + tabsElem.id = "search-tabs"; const resultsElem = document.createElement("div"); resultsElem.id = "results"; - resultsElem.appendChild(ret_others); - resultsElem.appendChild(ret_in_args); - resultsElem.appendChild(ret_returned); - // @ts-expect-error - search.innerHTML = output; - if (searchState.rustdocToolbar) { - // @ts-expect-error - search.querySelector(".main-heading").appendChild(searchState.rustdocToolbar); + search.innerHTML = ""; + for (const [tab, output] of tabs) { + tabsElem.appendChild(tab); + const placeholder = document.createElement("div"); + output.then(output => { + if (placeholder.parentElement) { + placeholder.parentElement.replaceChild(output, placeholder); + } + }); + resultsElem.appendChild(placeholder); + } + + if (window.searchState.rustdocToolbar) { + nonnull( + nonnull(window.searchState.containerElement()) + .querySelector(".main-heading"), + ).appendChild(window.searchState.rustdocToolbar); } const crateSearch = document.getElementById("crate-search"); if (crateSearch) { crateSearch.addEventListener("input", updateCrate); } - // @ts-expect-error + search.appendChild(tabsElem); search.appendChild(resultsElem); // Reset focused elements. - searchState.showResults(search); - // @ts-expect-error - const elems = document.getElementById("search-tabs").childNodes; - // @ts-expect-error - searchState.focusedByTab = []; + window.searchState.showResults(); + window.searchState.focusedByTab = [null, null, null]; let i = 0; - for (const elem of elems) { + for (const elem of tabsElem.childNodes) { const j = i; // @ts-expect-error elem.onclick = () => printTab(j); - searchState.focusedByTab.push(null); + window.searchState.focusedByTab[i] = null; i += 1; } - printTab(currentTab); + printTab(0); } // @ts-expect-error function updateSearchHistory(url) { + const btn = document.querySelector("#search-button a"); + if (btn instanceof HTMLAnchorElement) { + btn.href = url; + } if (!browserSupportsHistoryApi()) { return; } const params = searchState.getQueryStringParams(); - if (!history.state && !params.search) { + if (!history.state && params.search === undefined) { history.pushState(null, "", url); } else { history.replaceState(null, "", url); @@ -5127,8 +5011,8 @@ function updateSearchHistory(url) { * @param {boolean} [forced] */ async function search(forced) { - // @ts-expect-error - const query = DocSearch.parseQuery(searchState.input.value.trim()); + const query = DocSearch.parseQuery(nonnull(window.searchState.inputElement()).value.trim()); + let filterCrates = getFilterCrates(); // @ts-expect-error @@ -5138,6 +5022,7 @@ async function search(forced) { } return; } + currentResults = query.userQuery; searchState.setLoadingSearch(); @@ -5149,6 +5034,12 @@ async function search(forced) { filterCrates = params["filter-crate"]; } + if (filterCrates !== null && + (await docSearch.getCrateNameList()).indexOf(filterCrates) === -1 + ) { + filterCrates = null; + } + // Update document title to maintain a meaningful browser history searchState.title = "\"" + query.userQuery + "\" Search - Rust"; @@ -5157,6 +5048,7 @@ async function search(forced) { updateSearchHistory(buildUrl(query.userQuery, filterCrates)); await showResults( + docSearch, // @ts-expect-error await docSearch.execQuery(query, filterCrates, window.currentCrate), params.go_to_first, @@ -5176,16 +5068,14 @@ function onSearchSubmit(e) { } function putBackSearch() { - const search_input = searchState.input; - if (!searchState.input) { + const search_input = window.searchState.inputElement(); + if (!search_input) { return; } - // @ts-expect-error if (search_input.value !== "" && !searchState.isDisplayed()) { searchState.showResults(); if (browserSupportsHistoryApi()) { history.replaceState(null, "", - // @ts-expect-error buildUrl(search_input.value, getFilterCrates())); } document.title = searchState.title; @@ -5199,30 +5089,21 @@ function registerSearchEvents() { // but only if the input bar is empty. This avoid the obnoxious issue // where you start trying to do a search, and the index loads, and // suddenly your search is gone! - // @ts-expect-error - if (searchState.input.value === "") { - // @ts-expect-error - searchState.input.value = params.search || ""; + const inputElement = nonnull(window.searchState.inputElement()); + if (inputElement.value === "") { + inputElement.value = params.search || ""; } const searchAfter500ms = () => { searchState.clearInputTimeout(); - // @ts-expect-error - if (searchState.input.value.length === 0) { - searchState.hideResults(); - } else { - // @ts-ignore - searchState.timeout = setTimeout(search, 500); - } + window.searchState.timeout = setTimeout(search, 500); }; - // @ts-expect-error - searchState.input.onkeyup = searchAfter500ms; - // @ts-expect-error - searchState.input.oninput = searchAfter500ms; - // @ts-expect-error - document.getElementsByClassName("search-form")[0].onsubmit = onSearchSubmit; - // @ts-expect-error - searchState.input.onchange = e => { + inputElement.onkeyup = searchAfter500ms; + inputElement.oninput = searchAfter500ms; + if (inputElement.form) { + inputElement.form.onsubmit = onSearchSubmit; + } + inputElement.onchange = e => { if (e.target !== document.activeElement) { // To prevent doing anything when it's from a blur event. return; @@ -5234,11 +5115,13 @@ function registerSearchEvents() { // change, though. setTimeout(search, 0); }; - // @ts-expect-error - searchState.input.onpaste = searchState.input.onchange; + inputElement.onpaste = inputElement.onchange; // @ts-expect-error searchState.outputElement().addEventListener("keydown", e => { + if (!(e instanceof KeyboardEvent)) { + return; + } // We only handle unmodified keystrokes here. We don't want to interfere with, // for instance, alt-left and alt-right for history navigation. if (e.altKey || e.ctrlKey || e.shiftKey || e.metaKey) { @@ -5278,88 +5161,23 @@ function registerSearchEvents() { } }); - // @ts-expect-error - searchState.input.addEventListener("keydown", e => { + inputElement.addEventListener("keydown", e => { if (e.which === 40) { // down focusSearchResult(); e.preventDefault(); } }); - // @ts-expect-error - searchState.input.addEventListener("focus", () => { + inputElement.addEventListener("focus", () => { putBackSearch(); }); - - // @ts-expect-error - searchState.input.addEventListener("blur", () => { - if (window.searchState.input) { - window.searchState.input.placeholder = window.searchState.origPlaceholder; - } - }); - - // Push and pop states are used to add search results to the browser - // history. - if (browserSupportsHistoryApi()) { - // Store the previous <title> so we can revert back to it later. - const previousTitle = document.title; - - window.addEventListener("popstate", e => { - const params = searchState.getQueryStringParams(); - // Revert to the previous title manually since the History - // API ignores the title parameter. - document.title = previousTitle; - // When browsing forward to search results the previous - // search will be repeated, so the currentResults are - // cleared to ensure the search is successful. - currentResults = null; - // Synchronize search bar with query string state and - // perform the search. This will empty the bar if there's - // nothing there, which lets you really go back to a - // previous state with nothing in the bar. - if (params.search && params.search.length > 0) { - // @ts-expect-error - searchState.input.value = params.search; - // Some browsers fire "onpopstate" for every page load - // (Chrome), while others fire the event only when actually - // popping a state (Firefox), which is why search() is - // called both here and at the end of the startSearch() - // function. - e.preventDefault(); - search(); - } else { - // @ts-expect-error - searchState.input.value = ""; - // When browsing back from search results the main page - // visibility must be reset. - searchState.hideResults(); - } - }); - } - - // This is required in firefox to avoid this problem: Navigating to a search result - // with the keyboard, hitting enter, and then hitting back would take you back to - // the doc page, rather than the search that should overlay it. - // This was an interaction between the back-forward cache and our handlers - // that try to sync state between the URL and the search input. To work around it, - // do a small amount of re-init on page show. - window.onpageshow = () => { - const qSearch = searchState.getQueryStringParams().search; - // @ts-expect-error - if (searchState.input.value === "" && qSearch) { - // @ts-expect-error - searchState.input.value = qSearch; - } - search(); - }; } // @ts-expect-error function updateCrate(ev) { if (ev.target.value === "all crates") { // If we don't remove it from the URL, it'll be picked up again by the search. - // @ts-expect-error - const query = searchState.input.value.trim(); + const query = nonnull(window.searchState.inputElement()).value.trim(); updateSearchHistory(buildUrl(query, null)); } // In case you "cut" the entry from the search input, then change the crate filter @@ -5369,522 +5187,91 @@ function updateCrate(ev) { search(true); } -// Parts of this code are based on Lucene, which is licensed under the -// Apache/2.0 license. -// More information found here: -// https://fossies.org/linux/lucene/lucene/core/src/java/org/apache/lucene/util/automaton/ -// LevenshteinAutomata.java -class ParametricDescription { - // @ts-expect-error - constructor(w, n, minErrors) { - this.w = w; - this.n = n; - this.minErrors = minErrors; - } - // @ts-expect-error - isAccept(absState) { - const state = Math.floor(absState / (this.w + 1)); - const offset = absState % (this.w + 1); - return this.w - offset + this.minErrors[state] <= this.n; - } - // @ts-expect-error - getPosition(absState) { - return absState % (this.w + 1); - } - // @ts-expect-error - getVector(name, charCode, pos, end) { - let vector = 0; - for (let i = pos; i < end; i += 1) { - vector = vector << 1; - if (name.charCodeAt(i) === charCode) { - vector |= 1; - } - } - return vector; - } - // @ts-expect-error - unpack(data, index, bitsPerValue) { - const bitLoc = (bitsPerValue * index); - const dataLoc = bitLoc >> 5; - const bitStart = bitLoc & 31; - if (bitStart + bitsPerValue <= 32) { - // not split - return ((data[dataLoc] >> bitStart) & this.MASKS[bitsPerValue - 1]); - } else { - // split - const part = 32 - bitStart; - return ~~(((data[dataLoc] >> bitStart) & this.MASKS[part - 1]) + - ((data[1 + dataLoc] & this.MASKS[bitsPerValue - part - 1]) << part)); - } +// eslint-disable-next-line max-len +// polyfill https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Uint8Array/fromBase64 +/** + * @type {function(string): Uint8Array} base64 + */ +//@ts-expect-error +const makeUint8ArrayFromBase64 = Uint8Array.fromBase64 ? Uint8Array.fromBase64 : (string => { + const bytes_as_string = atob(string); + const l = bytes_as_string.length; + const bytes = new Uint8Array(l); + for (let i = 0; i < l; ++i) { + bytes[i] = bytes_as_string.charCodeAt(i); } -} -ParametricDescription.prototype.MASKS = new Int32Array([ - 0x1, 0x3, 0x7, 0xF, - 0x1F, 0x3F, 0x7F, 0xFF, - 0x1FF, 0x3F, 0x7FF, 0xFFF, - 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF, - 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, - 0x1FFFFF, 0x3FFFFF, 0x7FFFFF, 0xFFFFFF, - 0x1FFFFFF, 0x3FFFFFF, 0x7FFFFFF, 0xFFFFFFF, - 0x1FFFFFFF, 0x3FFFFFFF, 0x7FFFFFFF, 0xFFFFFFFF, -]); - -// The following code was generated with the moman/finenight pkg -// This package is available under the MIT License, see NOTICE.txt -// for more details. -// This class is auto-generated, Please do not modify it directly. -// You should modify the https://gitlab.com/notriddle/createAutomata.py instead. -// The following code was generated with the moman/finenight pkg -// This package is available under the MIT License, see NOTICE.txt -// for more details. -// This class is auto-generated, Please do not modify it directly. -// You should modify https://gitlab.com/notriddle/moman-rustdoc instead. - -class Lev2TParametricDescription extends ParametricDescription { - /** - * @param {number} absState - * @param {number} position - * @param {number} vector - * @returns {number} - */ - transition(absState, position, vector) { - let state = Math.floor(absState / (this.w + 1)); - let offset = absState % (this.w + 1); - - if (position === this.w) { - if (state < 3) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 3) + state; - offset += this.unpack(this.offsetIncrs0, loc, 1); - state = this.unpack(this.toStates0, loc, 2) - 1; - } - } else if (position === this.w - 1) { - if (state < 5) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 5) + state; - offset += this.unpack(this.offsetIncrs1, loc, 1); - state = this.unpack(this.toStates1, loc, 3) - 1; - } - } else if (position === this.w - 2) { - if (state < 13) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 13) + state; - offset += this.unpack(this.offsetIncrs2, loc, 2); - state = this.unpack(this.toStates2, loc, 4) - 1; - } - } else if (position === this.w - 3) { - if (state < 28) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 28) + state; - offset += this.unpack(this.offsetIncrs3, loc, 2); - state = this.unpack(this.toStates3, loc, 5) - 1; - } - } else if (position === this.w - 4) { - if (state < 45) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 45) + state; - offset += this.unpack(this.offsetIncrs4, loc, 3); - state = this.unpack(this.toStates4, loc, 6) - 1; - } - } else { - if (state < 45) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 45) + state; - offset += this.unpack(this.offsetIncrs5, loc, 3); - state = this.unpack(this.toStates5, loc, 6) - 1; - } - } + return bytes; +}); - if (state === -1) { - // null state - return -1; - } else { - // translate back to abs - return Math.imul(state, this.w + 1) + offset; - } - } - // state map - // 0 -> [(0, 0)] - // 1 -> [(0, 1)] - // 2 -> [(0, 2)] - // 3 -> [(0, 1), (1, 1)] - // 4 -> [(0, 2), (1, 2)] - // 5 -> [(0, 1), (1, 1), (2, 1)] - // 6 -> [(0, 2), (1, 2), (2, 2)] - // 7 -> [(0, 1), (2, 1)] - // 8 -> [(0, 1), (2, 2)] - // 9 -> [(0, 2), (2, 1)] - // 10 -> [(0, 2), (2, 2)] - // 11 -> [t(0, 1), (0, 1), (1, 1), (2, 1)] - // 12 -> [t(0, 2), (0, 2), (1, 2), (2, 2)] - // 13 -> [(0, 2), (1, 2), (2, 2), (3, 2)] - // 14 -> [(0, 1), (1, 1), (3, 2)] - // 15 -> [(0, 1), (2, 2), (3, 2)] - // 16 -> [(0, 1), (3, 2)] - // 17 -> [(0, 1), t(1, 2), (2, 2), (3, 2)] - // 18 -> [(0, 2), (1, 2), (3, 1)] - // 19 -> [(0, 2), (1, 2), (3, 2)] - // 20 -> [(0, 2), (1, 2), t(1, 2), (2, 2), (3, 2)] - // 21 -> [(0, 2), (2, 1), (3, 1)] - // 22 -> [(0, 2), (2, 2), (3, 2)] - // 23 -> [(0, 2), (3, 1)] - // 24 -> [(0, 2), (3, 2)] - // 25 -> [(0, 2), t(1, 2), (1, 2), (2, 2), (3, 2)] - // 26 -> [t(0, 2), (0, 2), (1, 2), (2, 2), (3, 2)] - // 27 -> [t(0, 2), (0, 2), (1, 2), (3, 1)] - // 28 -> [(0, 2), (1, 2), (2, 2), (3, 2), (4, 2)] - // 29 -> [(0, 2), (1, 2), (2, 2), (4, 2)] - // 30 -> [(0, 2), (1, 2), (2, 2), t(2, 2), (3, 2), (4, 2)] - // 31 -> [(0, 2), (1, 2), (3, 2), (4, 2)] - // 32 -> [(0, 2), (1, 2), (4, 2)] - // 33 -> [(0, 2), (1, 2), t(1, 2), (2, 2), (3, 2), (4, 2)] - // 34 -> [(0, 2), (1, 2), t(2, 2), (2, 2), (3, 2), (4, 2)] - // 35 -> [(0, 2), (2, 1), (4, 2)] - // 36 -> [(0, 2), (2, 2), (3, 2), (4, 2)] - // 37 -> [(0, 2), (2, 2), (4, 2)] - // 38 -> [(0, 2), (3, 2), (4, 2)] - // 39 -> [(0, 2), (4, 2)] - // 40 -> [(0, 2), t(1, 2), (1, 2), (2, 2), (3, 2), (4, 2)] - // 41 -> [(0, 2), t(2, 2), (2, 2), (3, 2), (4, 2)] - // 42 -> [t(0, 2), (0, 2), (1, 2), (2, 2), (3, 2), (4, 2)] - // 43 -> [t(0, 2), (0, 2), (1, 2), (2, 2), (4, 2)] - // 44 -> [t(0, 2), (0, 2), (1, 2), (2, 2), t(2, 2), (3, 2), (4, 2)] - - - /** @param {number} w - length of word being checked */ - constructor(w) { - super(w, 2, new Int32Array([ - 0,1,2,0,1,-1,0,-1,0,-1,0,-1,0,-1,-1,-1,-1,-1,-2,-1,-1,-2,-1,-2, - -1,-1,-1,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2, - ])); +if (ROOT_PATH === null) { + return; +} +const database = await Stringdex.loadDatabase(hooks); +if (typeof window !== "undefined") { + docSearch = new DocSearch(ROOT_PATH, database); + await docSearch.buildIndex(); + onEachLazy(document.querySelectorAll( + ".search-form.loading", + ), form => { + removeClass(form, "loading"); + }); + registerSearchEvents(); + // If there's a search term in the URL, execute the search now. + if (window.searchState.getQueryStringParams().search !== undefined) { + search(); } +} else if (typeof exports !== "undefined") { + docSearch = new DocSearch(ROOT_PATH, database); + await docSearch.buildIndex(); + return { docSearch, DocSearch }; } +}; -Lev2TParametricDescription.prototype.toStates0 = /*2 bits per value */ new Int32Array([ - 0xe, -]); -Lev2TParametricDescription.prototype.offsetIncrs0 = /*1 bits per value */ new Int32Array([ - 0x0, -]); - -Lev2TParametricDescription.prototype.toStates1 = /*3 bits per value */ new Int32Array([ - 0x1a688a2c, -]); -Lev2TParametricDescription.prototype.offsetIncrs1 = /*1 bits per value */ new Int32Array([ - 0x3e0, -]); - -Lev2TParametricDescription.prototype.toStates2 = /*4 bits per value */ new Int32Array([ - 0x70707054,0xdc07035,0x3dd3a3a,0x2323213a, - 0x15435223,0x22545432,0x5435, -]); -Lev2TParametricDescription.prototype.offsetIncrs2 = /*2 bits per value */ new Int32Array([ - 0x80000,0x55582088,0x55555555,0x55, -]); - -Lev2TParametricDescription.prototype.toStates3 = /*5 bits per value */ new Int32Array([ - 0x1c0380a4,0x700a570,0xca529c0,0x180a00, - 0xa80af180,0xc5498e60,0x5a546398,0x8c4300e8, - 0xac18c601,0xd8d43501,0x863500ad,0x51976d6a, - 0x8ca0180a,0xc3501ac2,0xb0c5be16,0x76dda8a5, - 0x18c4519,0xc41294a,0xe248d231,0x1086520c, - 0xce31ac42,0x13946358,0x2d0348c4,0x6732d494, - 0x1ad224a5,0xd635ad4b,0x520c4139,0xce24948, - 0x22110a52,0x58ce729d,0xc41394e3,0x941cc520, - 0x90e732d4,0x4729d224,0x39ce35ad, -]); -Lev2TParametricDescription.prototype.offsetIncrs3 = /*2 bits per value */ new Int32Array([ - 0x80000,0xc0c830,0x300f3c30,0x2200fcff, - 0xcaa00a08,0x3c2200a8,0xa8fea00a,0x55555555, - 0x55555555,0x55555555,0x55555555,0x55555555, - 0x55555555,0x55555555, -]); - -Lev2TParametricDescription.prototype.toStates4 = /*6 bits per value */ new Int32Array([ - 0x801c0144,0x1453803,0x14700038,0xc0005145, - 0x1401,0x14,0x140000,0x0, - 0x510000,0x6301f007,0x301f00d1,0xa186178, - 0xc20ca0c3,0xc20c30,0xc30030c,0xc00c00cd, - 0xf0c00c30,0x4c054014,0xc30944c3,0x55150c34, - 0x8300550,0x430c0143,0x50c31,0xc30850c, - 0xc3143000,0x50053c50,0x5130d301,0x850d30c2, - 0x30a08608,0xc214414,0x43142145,0x21450031, - 0x1400c314,0x4c143145,0x32832803,0x28014d6c, - 0xcd34a0c3,0x1c50c76,0x1c314014,0x430c30c3, - 0x1431,0xc300500,0xca00d303,0xd36d0e40, - 0x90b0e400,0xcb2abb2c,0x70c20ca1,0x2c32ca2c, - 0xcd2c70cb,0x31c00c00,0x34c2c32c,0x5583280, - 0x558309b7,0x6cd6ca14,0x430850c7,0x51c51401, - 0x1430c714,0xc3087,0x71451450,0xca00d30, - 0xc26dc156,0xb9071560,0x1cb2abb2,0xc70c2144, - 0xb1c51ca1,0x1421c70c,0xc51c00c3,0x30811c51, - 0x24324308,0xc51031c2,0x70820820,0x5c33830d, - 0xc33850c3,0x30c30c30,0xc30c31c,0x451450c3, - 0x20c20c20,0xda0920d,0x5145914f,0x36596114, - 0x51965865,0xd9643653,0x365a6590,0x51964364, - 0x43081505,0x920b2032,0x2c718b28,0xd7242249, - 0x35cb28b0,0x2cb3872c,0x972c30d7,0xb0c32cb2, - 0x4e1c75c,0xc80c90c2,0x62ca2482,0x4504171c, - 0xd65d9610,0x33976585,0xd95cb5d,0x4b5ca5d7, - 0x73975c36,0x10308138,0xc2245105,0x41451031, - 0x14e24208,0xc35c3387,0x51453851,0x1c51c514, - 0xc70c30c3,0x20451450,0x14f1440c,0x4f0da092, - 0x4513d41,0x6533944d,0x1350e658,0xe1545055, - 0x64365a50,0x5519383,0x51030815,0x28920718, - 0x441c718b,0x714e2422,0x1c35cb28,0x4e1c7387, - 0xb28e1c51,0x5c70c32c,0xc204e1c7,0x81c61440, - 0x1c62ca24,0xd04503ce,0x85d63944,0x39338e65, - 0x8e154387,0x364b5ca3,0x38739738, -]); -Lev2TParametricDescription.prototype.offsetIncrs4 = /*3 bits per value */ new Int32Array([ - 0x10000000,0xc00000,0x60061,0x400, - 0x0,0x80010008,0x249248a4,0x8229048, - 0x2092,0x6c3603,0xb61b6c30,0x6db6036d, - 0xdb6c0,0x361b0180,0x91b72000,0xdb11b71b, - 0x6db6236,0x1008200,0x12480012,0x24924906, - 0x48200049,0x80410002,0x24000900,0x4924a489, - 0x10822492,0x20800125,0x48360,0x9241b692, - 0x6da4924,0x40009268,0x241b010,0x291b4900, - 0x6d249249,0x49493423,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x2492, -]); - -Lev2TParametricDescription.prototype.toStates5 = /*6 bits per value */ new Int32Array([ - 0x801c0144,0x1453803,0x14700038,0xc0005145, - 0x1401,0x14,0x140000,0x0, - 0x510000,0x4e00e007,0xe0051,0x3451451c, - 0xd015000,0x30cd0000,0xc30c30c,0xc30c30d4, - 0x40c30c30,0x7c01c014,0xc03458c0,0x185e0c07, - 0x2830c286,0x830c3083,0xc30030,0x33430c, - 0x30c3003,0x70051030,0x16301f00,0x8301f00d, - 0x30a18617,0xc20ca0c,0x431420c3,0xb1450c51, - 0x14314315,0x4f143145,0x34c05401,0x4c30944c, - 0x55150c3,0x30830055,0x1430c014,0xc00050c3, - 0xc30850,0xc314300,0x150053c5,0x25130d30, - 0x5430d30c,0xc0354154,0x300d0c90,0x1cb2cd0c, - 0xc91cb0c3,0x72c30cb2,0x14f1cb2c,0xc34c0540, - 0x34c30944,0x82182214,0x851050c2,0x50851430, - 0x1400c50c,0x30c5085,0x50c51450,0x150053c, - 0xc25130d3,0x8850d30,0x1430a086,0x450c2144, - 0x51cb1c21,0x1c91c70c,0xc71c314b,0x34c1cb1, - 0x6c328328,0xc328014d,0x76cd34a0,0x1401c50c, - 0xc31c3140,0x31430c30,0x14,0x30c3005, - 0xa0ca00d3,0x535b0c,0x4d2830ca,0x514369b3, - 0xc500d01,0x5965965a,0x30d46546,0x6435030c, - 0x8034c659,0xdb439032,0x2c390034,0xcaaecb24, - 0x30832872,0xcb28b1c,0x4b1c32cb,0x70030033, - 0x30b0cb0c,0xe40ca00d,0x400d36d0,0xb2c90b0e, - 0xca1cb2ab,0xa2c70c20,0x6575d95c,0x4315b5ce, - 0x95c53831,0x28034c5d,0x9b705583,0xa1455830, - 0xc76cd6c,0x40143085,0x71451c51,0x871430c, - 0x450000c3,0xd3071451,0x1560ca00,0x560c26dc, - 0xb35b2851,0xc914369,0x1a14500d,0x46593945, - 0xcb2c939,0x94507503,0x328034c3,0x9b70558, - 0xe41c5583,0x72caaeca,0x1c308510,0xc7147287, - 0x50871c32,0x1470030c,0xd307147,0xc1560ca0, - 0x1560c26d,0xabb2b907,0x21441cb2,0x38a1c70c, - 0x8e657394,0x314b1c93,0x39438738,0x43083081, - 0x31c22432,0x820c510,0x830d7082,0x50c35c33, - 0xc30c338,0xc31c30c3,0x50c30c30,0xc204514, - 0x890c90c2,0x31440c70,0xa8208208,0xea0df0c3, - 0x8a231430,0xa28a28a2,0x28a28a1e,0x1861868a, - 0x48308308,0xc3682483,0x14516453,0x4d965845, - 0xd4659619,0x36590d94,0xd969964,0x546590d9, - 0x20c20541,0x920d20c,0x5914f0da,0x96114514, - 0x65865365,0xe89d3519,0x99e7a279,0x9e89e89e, - 0x81821827,0xb2032430,0x18b28920,0x422492c7, - 0xb28b0d72,0x3872c35c,0xc30d72cb,0x32cb2972, - 0x1c75cb0c,0xc90c204e,0xa2482c80,0x24b1c62c, - 0xc3a89089,0xb0ea2e42,0x9669a31c,0xa4966a28, - 0x59a8a269,0x8175e7a,0xb203243,0x718b2892, - 0x4114105c,0x17597658,0x74ce5d96,0x5c36572d, - 0xd92d7297,0xe1ce5d70,0xc90c204,0xca2482c8, - 0x4171c62,0x5d961045,0x976585d6,0x79669533, - 0x964965a2,0x659689e6,0x308175e7,0x24510510, - 0x451031c2,0xe2420841,0x5c338714,0x453851c3, - 0x51c51451,0xc30c31c,0x451450c7,0x41440c20, - 0xc708914,0x82105144,0xf1c58c90,0x1470ea0d, - 0x61861863,0x8a1e85e8,0x8687a8a2,0x3081861, - 0x24853c51,0x5053c368,0x1341144f,0x96194ce5, - 0x1544d439,0x94385514,0xe0d90d96,0x5415464, - 0x4f1440c2,0xf0da0921,0x4513d414,0x533944d0, - 0x350e6586,0x86082181,0xe89e981d,0x18277689, - 0x10308182,0x89207185,0x41c718b2,0x14e24224, - 0xc35cb287,0xe1c73871,0x28e1c514,0xc70c32cb, - 0x204e1c75,0x1c61440c,0xc62ca248,0x90891071, - 0x2e41c58c,0xa31c70ea,0xe86175e7,0xa269a475, - 0x5e7a57a8,0x51030817,0x28920718,0xf38718b, - 0xe5134114,0x39961758,0xe1ce4ce,0x728e3855, - 0x5ce0d92d,0xc204e1ce,0x81c61440,0x1c62ca24, - 0xd04503ce,0x85d63944,0x75338e65,0x5d86075e, - 0x89e69647,0x75e76576, -]); -Lev2TParametricDescription.prototype.offsetIncrs5 = /*3 bits per value */ new Int32Array([ - 0x10000000,0xc00000,0x60061,0x400, - 0x0,0x60000008,0x6b003080,0xdb6ab6db, - 0x2db6,0x800400,0x49245240,0x11482412, - 0x104904,0x40020000,0x92292000,0xa4b25924, - 0x9649658,0xd80c000,0xdb0c001b,0x80db6d86, - 0x6db01b6d,0xc0600003,0x86000d86,0x6db6c36d, - 0xddadb6ed,0x300001b6,0x6c360,0xe37236e4, - 0x46db6236,0xdb6c,0x361b018,0xb91b7200, - 0x6dbb1b71,0x6db763,0x20100820,0x61248001, - 0x92492490,0x24820004,0x8041000,0x92400090, - 0x24924830,0x555b6a49,0x2080012,0x20004804, - 0x49252449,0x84112492,0x4000928,0x240201, - 0x92922490,0x58924924,0x49456,0x120d8082, - 0x6da4800,0x69249249,0x249a01b,0x6c04100, - 0x6d240009,0x92492483,0x24d5adb4,0x60208001, - 0x92000483,0x24925236,0x6846da49,0x10400092, - 0x241b0,0x49291b49,0x636d2492,0x92494935, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924,0x49249249, - 0x92492492,0x24924924,0x49249249,0x92492492, - 0x24924924,0x49249249,0x92492492,0x24924924, - 0x49249249,0x92492492,0x24924924, -]); - -class Lev1TParametricDescription extends ParametricDescription { - /** - * @param {number} absState - * @param {number} position - * @param {number} vector - * @returns {number} - */ - transition(absState, position, vector) { - let state = Math.floor(absState / (this.w + 1)); - let offset = absState % (this.w + 1); - - if (position === this.w) { - if (state < 2) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 2) + state; - offset += this.unpack(this.offsetIncrs0, loc, 1); - state = this.unpack(this.toStates0, loc, 2) - 1; - } - } else if (position === this.w - 1) { - if (state < 3) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 3) + state; - offset += this.unpack(this.offsetIncrs1, loc, 1); - state = this.unpack(this.toStates1, loc, 2) - 1; - } - } else if (position === this.w - 2) { - if (state < 6) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 6) + state; - offset += this.unpack(this.offsetIncrs2, loc, 2); - state = this.unpack(this.toStates2, loc, 3) - 1; +if (typeof window !== "undefined") { + const ROOT_PATH = window.rootPath; + /** @type {stringdex.Callbacks|null} */ + let databaseCallbacks = null; + initSearch(window.Stringdex, window.RoaringBitmap, { + loadRoot: callbacks => { + for (const key in callbacks) { + if (Object.hasOwn(callbacks, key)) { + // @ts-ignore + window[key] = callbacks[key]; + } } - } else { - if (state < 6) { // eslint-disable-line no-lonely-if - const loc = Math.imul(vector, 6) + state; - offset += this.unpack(this.offsetIncrs3, loc, 2); - state = this.unpack(this.toStates3, loc, 3) - 1; + databaseCallbacks = callbacks; + // search.index/root is loaded by main.js, so + // this script doesn't need to launch it, but + // must pick it up + // @ts-ignore + if (window.searchIndex) { + // @ts-ignore + window.rr_(window.searchIndex); } - } - - if (state === -1) { - // null state - return -1; - } else { - // translate back to abs - return Math.imul(state, this.w + 1) + offset; - } - } - - // state map - // 0 -> [(0, 0)] - // 1 -> [(0, 1)] - // 2 -> [(0, 1), (1, 1)] - // 3 -> [(0, 1), (1, 1), (2, 1)] - // 4 -> [(0, 1), (2, 1)] - // 5 -> [t(0, 1), (0, 1), (1, 1), (2, 1)] - - - /** @param {number} w - length of word being checked */ - constructor(w) { - super(w, 1, new Int32Array([0,1,0,-1,-1,-1])); - } -} - -Lev1TParametricDescription.prototype.toStates0 = /*2 bits per value */ new Int32Array([ - 0x2, -]); -Lev1TParametricDescription.prototype.offsetIncrs0 = /*1 bits per value */ new Int32Array([ - 0x0, -]); - -Lev1TParametricDescription.prototype.toStates1 = /*2 bits per value */ new Int32Array([ - 0xa43, -]); -Lev1TParametricDescription.prototype.offsetIncrs1 = /*1 bits per value */ new Int32Array([ - 0x38, -]); - -Lev1TParametricDescription.prototype.toStates2 = /*3 bits per value */ new Int32Array([ - 0x12180003,0xb45a4914,0x69, -]); -Lev1TParametricDescription.prototype.offsetIncrs2 = /*2 bits per value */ new Int32Array([ - 0x558a0000,0x5555, -]); - -Lev1TParametricDescription.prototype.toStates3 = /*3 bits per value */ new Int32Array([ - 0x900c0003,0xa1904864,0x45a49169,0x5a6d196a, - 0x9634, -]); -Lev1TParametricDescription.prototype.offsetIncrs3 = /*2 bits per value */ new Int32Array([ - 0xa0fc0000,0x5555ba08,0x55555555, -]); - -// ==================== -// WARNING: Nothing should be added below this comment: we need the `initSearch` function to -// be called ONLY when the whole file has been parsed and loaded. - -// @ts-expect-error -function initSearch(searchIndex) { - rawSearchIndex = searchIndex; - if (typeof window !== "undefined") { - // @ts-expect-error - docSearch = new DocSearch(rawSearchIndex, ROOT_PATH, searchState); - registerSearchEvents(); - // If there's a search term in the URL, execute the search now. - if (window.searchState.getQueryStringParams().search) { - search(); - } - } else if (typeof exports !== "undefined") { - // @ts-expect-error - docSearch = new DocSearch(rawSearchIndex, ROOT_PATH, searchState); - exports.docSearch = docSearch; - exports.parseQuery = DocSearch.parseQuery; - } -} - -if (typeof exports !== "undefined") { + }, + loadTreeByHash: hashHex => { + const script = document.createElement("script"); + script.src = `${ROOT_PATH}/search.index/${hashHex}.js`; + script.onerror = e => { + if (databaseCallbacks) { + databaseCallbacks.err_rn_(hashHex, e); + } + }; + document.documentElement.appendChild(script); + }, + loadDataByNameAndHash: (name, hashHex) => { + const script = document.createElement("script"); + script.src = `${ROOT_PATH}/search.index/${name}/${hashHex}.js`; + script.onerror = e => { + if (databaseCallbacks) { + databaseCallbacks.err_rd_(hashHex, e); + } + }; + document.documentElement.appendChild(script); + }, + }); +} else if (typeof exports !== "undefined") { + // eslint-disable-next-line no-undef exports.initSearch = initSearch; } - -if (typeof window !== "undefined") { - // @ts-expect-error - window.initSearch = initSearch; - // @ts-expect-error - if (window.searchIndex !== undefined) { - // @ts-expect-error - initSearch(window.searchIndex); - } -} else { - // Running in Node, not a browser. Run initSearch just to produce the - // exports. - initSearch(new Map()); -} diff --git a/src/librustdoc/html/static/js/settings.js b/src/librustdoc/html/static/js/settings.js index 2430b5829b2..347d3d0750e 100644 --- a/src/librustdoc/html/static/js/settings.js +++ b/src/librustdoc/html/static/js/settings.js @@ -1,7 +1,7 @@ // Local js definitions: /* global getSettingValue, updateLocalStorage, updateTheme */ /* global addClass, removeClass, onEach, onEachLazy */ -/* global MAIN_ID, getVar, getSettingsButton, getHelpButton, nonnull */ +/* global MAIN_ID, getVar, nonnull */ "use strict"; @@ -9,18 +9,6 @@ const isSettingsPage = window.location.pathname.endsWith("/settings.html"); /** - * @param {Element} elem - * @param {EventTarget|null} target - */ - function elemContainsTarget(elem, target) { - if (target instanceof Node) { - return elem.contains(target); - } else { - return false; - } - } - - /** * @overload {"theme"|"preferred-dark-theme"|"preferred-light-theme"} * @param {string} settingName * @param {string} value @@ -305,10 +293,12 @@ } } else { el.setAttribute("tabindex", "-1"); - const settingsBtn = getSettingsButton(); - if (settingsBtn !== null) { - settingsBtn.appendChild(el); - } + onEachLazy(document.querySelectorAll(".settings-menu"), menu => { + if (menu.offsetWidth !== 0) { + menu.appendChild(el); + return true; + } + }); } return el; } @@ -317,6 +307,15 @@ function displaySettings() { settingsMenu.style.display = ""; + onEachLazy(document.querySelectorAll(".settings-menu"), menu => { + if (menu.offsetWidth !== 0) { + if (!menu.contains(settingsMenu) && settingsMenu.parentElement) { + settingsMenu.parentElement.removeChild(settingsMenu); + menu.appendChild(settingsMenu); + } + return true; + } + }); onEachLazy(settingsMenu.querySelectorAll("input[type='checkbox']"), el => { const val = getSettingValue(el.id); const checked = val === "true"; @@ -330,40 +329,37 @@ * @param {FocusEvent} event */ function settingsBlurHandler(event) { - const helpBtn = getHelpButton(); - const settingsBtn = getSettingsButton(); - const helpUnfocused = helpBtn === null || - (!helpBtn.contains(document.activeElement) && - !elemContainsTarget(helpBtn, event.relatedTarget)); - const settingsUnfocused = settingsBtn === null || - (!settingsBtn.contains(document.activeElement) && - !elemContainsTarget(settingsBtn, event.relatedTarget)); - if (helpUnfocused && settingsUnfocused) { + const isInPopover = onEachLazy( + document.querySelectorAll(".settings-menu, .help-menu"), + menu => { + return menu.contains(document.activeElement) || menu.contains(event.relatedTarget); + }, + ); + if (!isInPopover) { window.hidePopoverMenus(); } } if (!isSettingsPage) { // We replace the existing "onclick" callback. - // These elements must exist, as (outside of the settings page) - // `settings.js` is only loaded after the settings button is clicked. - const settingsButton = nonnull(getSettingsButton()); const settingsMenu = nonnull(document.getElementById("settings")); - settingsButton.onclick = event => { - if (elemContainsTarget(settingsMenu, event.target)) { - return; - } - event.preventDefault(); - const shouldDisplaySettings = settingsMenu.style.display === "none"; + onEachLazy(document.querySelectorAll(".settings-menu"), settingsButton => { + /** @param {MouseEvent} event */ + settingsButton.querySelector("a").onclick = event => { + if (!(event.target instanceof Element) || settingsMenu.contains(event.target)) { + return; + } + event.preventDefault(); + const shouldDisplaySettings = settingsMenu.style.display === "none"; - window.hideAllModals(false); - if (shouldDisplaySettings) { - displaySettings(); - } - }; - settingsButton.onblur = settingsBlurHandler; - // the settings button should always have a link in it - nonnull(settingsButton.querySelector("a")).onblur = settingsBlurHandler; + window.hideAllModals(false); + if (shouldDisplaySettings) { + displaySettings(); + } + }; + settingsButton.onblur = settingsBlurHandler; + settingsButton.querySelector("a").onblur = settingsBlurHandler; + }); onEachLazy(settingsMenu.querySelectorAll("input"), el => { el.onblur = settingsBlurHandler; }); @@ -377,6 +373,8 @@ if (!isSettingsPage) { displaySettings(); } - removeClass(getSettingsButton(), "rotate"); + onEachLazy(document.querySelectorAll(".settings-menu"), settingsButton => { + removeClass(settingsButton, "rotate"); + }); }, 0); })(); diff --git a/src/librustdoc/html/static/js/storage.js b/src/librustdoc/html/static/js/storage.js index ca13b891638..c055eb0f808 100644 --- a/src/librustdoc/html/static/js/storage.js +++ b/src/librustdoc/html/static/js/storage.js @@ -7,6 +7,7 @@ /** * @import * as rustdoc from "./rustdoc.d.ts"; + * @import * as stringdex from "./stringdex.d.ts"; */ const builtinThemes = ["light", "dark", "ayu"]; @@ -172,7 +173,7 @@ function updateLocalStorage(name, value) { } else { window.localStorage.setItem("rustdoc-" + name, value); } - } catch (e) { + } catch { // localStorage is not accessible, do nothing } } @@ -189,7 +190,7 @@ function updateLocalStorage(name, value) { function getCurrentValue(name) { try { return window.localStorage.getItem("rustdoc-" + name); - } catch (e) { + } catch { return null; } } @@ -375,32 +376,6 @@ window.addEventListener("pageshow", ev => { // That's also why this is in storage.js and not main.js. // // [parser]: https://html.spec.whatwg.org/multipage/parsing.html -class RustdocSearchElement extends HTMLElement { - constructor() { - super(); - } - connectedCallback() { - const rootPath = getVar("root-path"); - const currentCrate = getVar("current-crate"); - this.innerHTML = `<nav class="sub"> - <form class="search-form"> - <span></span> <!-- This empty span is a hacky fix for Safari - See #93184 --> - <div id="sidebar-button" tabindex="-1"> - <a href="${rootPath}${currentCrate}/all.html" title="show sidebar"></a> - </div> - <input - class="search-input" - name="search" - aria-label="Run search in the documentation" - autocomplete="off" - spellcheck="false" - placeholder="Type ‘S’ or ‘/’ to search, ‘?’ for more options…" - type="search"> - </form> - </nav>`; - } -} -window.customElements.define("rustdoc-search", RustdocSearchElement); class RustdocToolbarElement extends HTMLElement { constructor() { super(); @@ -411,11 +386,15 @@ class RustdocToolbarElement extends HTMLElement { return; } const rootPath = getVar("root-path"); + const currentUrl = window.location.href.split("?")[0].split("#")[0]; this.innerHTML = ` - <div id="settings-menu" tabindex="-1"> + <div id="search-button" tabindex="-1"> + <a href="${currentUrl}?search="><span class="label">Search</span></a> + </div> + <div class="settings-menu" tabindex="-1"> <a href="${rootPath}settings.html"><span class="label">Settings</span></a> </div> - <div id="help-button" tabindex="-1"> + <div class="help-menu" tabindex="-1"> <a href="${rootPath}help.html"><span class="label">Help</span></a> </div> <button id="toggle-all-docs" @@ -424,3 +403,31 @@ class="label">Summary</span></button>`; } } window.customElements.define("rustdoc-toolbar", RustdocToolbarElement); +class RustdocTopBarElement extends HTMLElement { + constructor() { + super(); + } + connectedCallback() { + const rootPath = getVar("root-path"); + const tmplt = document.createElement("template"); + tmplt.innerHTML = ` + <slot name="sidebar-menu-toggle"></slot> + <slot></slot> + <slot name="settings-menu"></slot> + <slot name="help-menu"></slot> + `; + const shadow = this.attachShadow({ mode: "open" }); + shadow.appendChild(tmplt.content.cloneNode(true)); + this.innerHTML += ` + <button class="sidebar-menu-toggle" slot="sidebar-menu-toggle" title="show sidebar"> + </button> + <div class="settings-menu" slot="settings-menu" tabindex="-1"> + <a href="${rootPath}settings.html"><span class="label">Settings</span></a> + </div> + <div class="help-menu" slot="help-menu" tabindex="-1"> + <a href="${rootPath}help.html"><span class="label">Help</span></a> + </div> + `; + } +} +window.customElements.define("rustdoc-topbar", RustdocTopBarElement); diff --git a/src/librustdoc/html/static/js/stringdex.d.ts b/src/librustdoc/html/static/js/stringdex.d.ts new file mode 100644 index 00000000000..cf9a8b6b564 --- /dev/null +++ b/src/librustdoc/html/static/js/stringdex.d.ts @@ -0,0 +1,165 @@ +export = stringdex; + +declare namespace stringdex { + /** + * The client interface to Stringdex. + */ + interface Database { + getIndex(colname: string): SearchTree|undefined; + getData(colname: string): DataColumn|undefined; + } + /** + * A search index file. + */ + interface SearchTree { + trie(): Trie; + search(name: Uint8Array|string): Promise<Trie?>; + searchLev(name: Uint8Array|string): AsyncGenerator<Trie>; + } + /** + * A compressed node in the search tree. + * + * This object logically addresses two interleaved trees: + * a "prefix tree", and a "suffix tree". If you ask for + * generic matches, you get both, but if you ask for one + * that excludes suffix-only entries, you'll get prefixes + * alone. + */ + interface Trie { + matches(): RoaringBitmap; + substringMatches(): AsyncGenerator<RoaringBitmap>; + prefixMatches(): AsyncGenerator<RoaringBitmap>; + keys(): Uint8Array; + keysExcludeSuffixOnly(): Uint8Array; + children(): [number, Promise<Trie>][]; + childrenExcludeSuffixOnly(): [number, Promise<Trie>][]; + child(id: number): Promise<Trie>?; + } + /** + * The client interface to Stringdex. + */ + interface DataColumn { + isEmpty(id: number): boolean; + at(id: number): Promise<Uint8Array|undefined>; + length: number, + } + /** + * Callbacks for a host application and VFS backend. + * + * These functions are calleb with mostly-raw data, + * except the JSONP wrapper is removed. For example, + * a file with the contents `rr_('{"A":"B"}')` should, + * after being pulled in, result in the `rr_` callback + * being invoked. + * + * The success callbacks don't need to supply the name of + * the file that succeeded, but, if you want successful error + * reporting, you'll need to remember which files are + * in flight and report the filename as the first parameter. + */ + interface Callbacks { + /** + * Load the root of the search database + * @param {string} dataString + */ + rr_: function(string); + err_rr_: function(any); + /** + * Load a nodefile in the search tree. + * A node file may contain multiple nodes; + * each node has five fields, separated by newlines. + * @param {string} inputBase64 + */ + rn_: function(string); + err_rn_: function(string, any); + /** + * Load a database column partition from a string + * @param {string} dataString + */ + rd_: function(string); + err_rd_: function(string, any); + /** + * Load a database column partition from base64 + * @param {string} dataString + */ + rb_: function(string); + err_rb_: function(string, any); + }; + /** + * Hooks that a VFS layer must provide for stringdex to load data. + * + * When the root is loaded, the Callbacks object is provided. These + * functions should result in callback functions being called with + * the contents of the file, or in error callbacks being invoked with + * the failed-to-load filename. + */ + interface Hooks { + /** + * The first function invoked as part of loading a search database. + * This function must, eventually, invoke `rr_` with the string + * representation of the root file (the function call wrapper, + * `rr_('` and `')`, must be removed). + * + * The supplied callbacks object is used to feed search data back + * to the search engine core. You have to store it, so that + * loadTreeByHash and loadDataByNameAndHash can use it. + * + * If this fails, either throw an exception, or call `err_rr_` + * with the error object. + */ + loadRoot: function(Callbacks); + /** + * Load a subtree file from the search index. + * + * If this function succeeds, call `rn_` on the callbacks + * object. If it fails, call `err_rn_(hashHex, error)`. + * + * @param {string} hashHex + */ + loadTreeByHash: function(string); + /** + * Load a column partition from the search database. + * + * If this function succeeds, call `rd_` or `rb_` on the callbacks + * object. If it fails, call `err_rd_(hashHex, error)`. or `err_rb_`. + * To determine which one, the wrapping function call in the js file + * specifies it. + * + * @param {string} columnName + * @param {string} hashHex + */ + loadDataByNameAndHash: function(string, string); + }; + class RoaringBitmap { + constructor(array: Uint8Array|null, start?: number); + static makeSingleton(number: number); + static everything(): RoaringBitmap; + static empty(): RoaringBitmap; + isEmpty(): boolean; + union(that: RoaringBitmap): RoaringBitmap; + intersection(that: RoaringBitmap): RoaringBitmap; + contains(number: number): boolean; + entries(): Generator<number>; + first(): number|null; + consumed_len_bytes: number; + }; + + type Stringdex = { + /** + * Initialize Stringdex with VFS hooks. + * Returns a database that you can use. + */ + loadDatabase: function(Hooks): Promise<Database>, + }; + + const Stringdex: Stringdex; + const RoaringBitmap: Class<stringdex.RoaringBitmap>; +} + +declare global { + interface Window { + Stringdex: stringdex.Stringdex; + RoaringBitmap: Class<stringdex.RoaringBitmap>; + StringdexOnload: Array<function(stringdex.Stringdex): any>?; + }; +} \ No newline at end of file diff --git a/src/librustdoc/html/static/js/stringdex.js b/src/librustdoc/html/static/js/stringdex.js new file mode 100644 index 00000000000..cb956d926db --- /dev/null +++ b/src/librustdoc/html/static/js/stringdex.js @@ -0,0 +1,3217 @@ +/** + * @import * as stringdex from "./stringdex.d.ts" + */ + +const EMPTY_UINT8 = new Uint8Array(); + +/** + * @property {Uint8Array} keysAndCardinalities + * @property {Uint8Array[]} containers + */ +class RoaringBitmap { + /** + * @param {Uint8Array|null} u8array + * @param {number} [startingOffset] + */ + constructor(u8array, startingOffset) { + const start = startingOffset ? startingOffset : 0; + let i = start; + /** @type {Uint8Array} */ + this.keysAndCardinalities = EMPTY_UINT8; + /** @type {(RoaringBitmapArray|RoaringBitmapBits|RoaringBitmapRun)[]} */ + this.containers = []; + /** @type {number} */ + this.consumed_len_bytes = 0; + if (u8array === null || u8array.length === i || u8array[i] === 0) { + return this; + } else if (u8array[i] > 0xf0) { + // Special representation of tiny sets that are close together + const lspecial = u8array[i] & 0x0f; + this.keysAndCardinalities = new Uint8Array(lspecial * 4); + let pspecial = i + 1; + let key = u8array[pspecial + 2] | (u8array[pspecial + 3] << 8); + let value = u8array[pspecial] | (u8array[pspecial + 1] << 8); + let entry = (key << 16) | value; + let container; + container = new RoaringBitmapArray(1, new Uint8Array(4)); + container.array[0] = value & 0xFF; + container.array[1] = (value >> 8) & 0xFF; + this.containers.push(container); + this.keysAndCardinalities[0] = key; + this.keysAndCardinalities[1] = key >> 8; + pspecial += 4; + for (let ispecial = 1; ispecial < lspecial; ispecial += 1) { + entry += u8array[pspecial] | (u8array[pspecial + 1] << 8); + value = entry & 0xffff; + key = entry >> 16; + container = this.addToArrayAt(key); + const cardinalityOld = container.cardinality; + container.array[cardinalityOld * 2] = value & 0xFF; + container.array[(cardinalityOld * 2) + 1] = (value >> 8) & 0xFF; + container.cardinality = cardinalityOld + 1; + pspecial += 2; + } + this.consumed_len_bytes = pspecial - i; + return this; + } else if (u8array[i] < 0x3a) { + // Special representation of tiny sets with arbitrary 32-bit integers + const lspecial = u8array[i]; + this.keysAndCardinalities = new Uint8Array(lspecial * 4); + let pspecial = i + 1; + for (let ispecial = 0; ispecial < lspecial; ispecial += 1) { + const key = u8array[pspecial + 2] | (u8array[pspecial + 3] << 8); + const value = u8array[pspecial] | (u8array[pspecial + 1] << 8); + const container = this.addToArrayAt(key); + const cardinalityOld = container.cardinality; + container.array[cardinalityOld * 2] = value & 0xFF; + container.array[(cardinalityOld * 2) + 1] = (value >> 8) & 0xFF; + container.cardinality = cardinalityOld + 1; + pspecial += 4; + } + this.consumed_len_bytes = pspecial - i; + return this; + } + // https://github.com/RoaringBitmap/RoaringFormatSpec + // + // Roaring bitmaps are used for flags that can be kept in their + // compressed form, even when loaded into memory. This decoder + // turns the containers into objects, but uses byte array + // slices of the original format for the data payload. + const has_runs = u8array[i] === 0x3b; + if (u8array[i] !== 0x3a && u8array[i] !== 0x3b) { + throw new Error("not a roaring bitmap: " + u8array[i]); + } + const size = has_runs ? + ((u8array[i + 2] | (u8array[i + 3] << 8)) + 1) : + ((u8array[i + 4] | (u8array[i + 5] << 8) | + (u8array[i + 6] << 16) | (u8array[i + 7] << 24))); + i += has_runs ? 4 : 8; + let is_run; + if (has_runs) { + const is_run_len = (size + 7) >> 3; + is_run = new Uint8Array(u8array.buffer, i + u8array.byteOffset, is_run_len); + i += is_run_len; + } else { + is_run = EMPTY_UINT8; + } + this.keysAndCardinalities = u8array.subarray(i, i + (size * 4)); + i += size * 4; + let offsets = null; + if (!has_runs || size >= 4) { + offsets = []; + for (let j = 0; j < size; ++j) { + offsets.push(u8array[i] | (u8array[i + 1] << 8) | (u8array[i + 2] << 16) | + (u8array[i + 3] << 24)); + i += 4; + } + } + for (let j = 0; j < size; ++j) { + if (offsets && offsets[j] !== i - start) { + throw new Error(`corrupt bitmap ${j}: ${i - start} / ${offsets[j]}`); + } + const cardinality = (this.keysAndCardinalities[(j * 4) + 2] | + (this.keysAndCardinalities[(j * 4) + 3] << 8)) + 1; + if (is_run[j >> 3] & (1 << (j & 0x7))) { + const runcount = (u8array[i] | (u8array[i + 1] << 8)); + i += 2; + this.containers.push(new RoaringBitmapRun( + runcount, + new Uint8Array(u8array.buffer, i + u8array.byteOffset, runcount * 4), + )); + i += runcount * 4; + } else if (cardinality >= 4096) { + this.containers.push(new RoaringBitmapBits(new Uint8Array( + u8array.buffer, + i + u8array.byteOffset, 8192, + ))); + i += 8192; + } else { + const end = cardinality * 2; + this.containers.push(new RoaringBitmapArray( + cardinality, + new Uint8Array(u8array.buffer, i + u8array.byteOffset, end), + )); + i += end; + } + } + this.consumed_len_bytes = i - start; + } + /** + * @param {number} number + * @returns {RoaringBitmap} + */ + static makeSingleton(number) { + const result = new RoaringBitmap(null, 0); + result.keysAndCardinalities = Uint8Array.of( + (number >> 16), (number >> 24), + 0, 0, // keysAndCardinalities stores the true cardinality minus 1 + ); + result.containers.push(new RoaringBitmapArray( + 1, + Uint8Array.of(number, number >> 8), + )); + return result; + } + /** @returns {RoaringBitmap} */ + static everything() { + if (EVERYTHING_BITMAP.isEmpty()) { + let i = 0; + const l = 1 << 16; + const everything_range = new RoaringBitmapRun(1, Uint8Array.of(0, 0, 0xff, 0xff)); + EVERYTHING_BITMAP.keysAndCardinalities = new Uint8Array(l * 4); + while (i < l) { + EVERYTHING_BITMAP.containers.push(everything_range); + // key + EVERYTHING_BITMAP.keysAndCardinalities[(i * 4) + 0] = i; + EVERYTHING_BITMAP.keysAndCardinalities[(i * 4) + 1] = i >> 8; + // cardinality (minus one) + EVERYTHING_BITMAP.keysAndCardinalities[(i * 4) + 2] = 0xff; + EVERYTHING_BITMAP.keysAndCardinalities[(i * 4) + 3] = 0xff; + i += 1; + } + } + return EVERYTHING_BITMAP; + } + /** @returns {RoaringBitmap} */ + static empty() { + return EMPTY_BITMAP; + } + /** @returns {boolean} */ + isEmpty() { + return this.containers.length === 0; + } + /** + * Helper function used when constructing bitmaps from lists. + * Returns an array container with at least two free byte slots + * and bumps `this.cardinalities`. + * @param {number} key + * @returns {RoaringBitmapArray} + */ + addToArrayAt(key) { + let mid = this.getContainerId(key); + /** @type {RoaringBitmapArray|RoaringBitmapBits|RoaringBitmapRun} */ + let container; + if (mid === -1) { + container = new RoaringBitmapArray(0, new Uint8Array(2)); + mid = this.containers.length; + this.containers.push(container); + if (mid * 4 > this.keysAndCardinalities.length) { + const keysAndContainers = new Uint8Array(mid * 8); + keysAndContainers.set(this.keysAndCardinalities); + this.keysAndCardinalities = keysAndContainers; + } + this.keysAndCardinalities[(mid * 4) + 0] = key; + this.keysAndCardinalities[(mid * 4) + 1] = key >> 8; + } else { + container = this.containers[mid]; + const cardinalityOld = + this.keysAndCardinalities[(mid * 4) + 2] | + (this.keysAndCardinalities[(mid * 4) + 3] << 8); + const cardinality = cardinalityOld + 1; + this.keysAndCardinalities[(mid * 4) + 2] = cardinality; + this.keysAndCardinalities[(mid * 4) + 3] = cardinality >> 8; + } + // the logic for handing this number is annoying, because keysAndCardinalities stores + // the cardinality *minus one*, so that it can count up to 65536 with only two bytes + // (because empty containers are never stored). + // + // So, if this is a new container, the stored cardinality contains `0 0`, which is + // the proper value of the old cardinality (an imaginary empty container existed). + // If this is adding to an existing container, then the above `else` branch bumps it + // by one, leaving us with a proper value of `cardinality - 1`. + const cardinalityOld = + this.keysAndCardinalities[(mid * 4) + 2] | + (this.keysAndCardinalities[(mid * 4) + 3] << 8); + if (!(container instanceof RoaringBitmapArray) || + container.array.byteLength < ((cardinalityOld + 1) * 2) + ) { + const newBuf = new Uint8Array((cardinalityOld + 1) * 4); + let idx = 0; + for (const cvalue of container.values()) { + newBuf[idx] = cvalue & 0xFF; + newBuf[idx + 1] = (cvalue >> 8) & 0xFF; + idx += 2; + } + if (container instanceof RoaringBitmapArray) { + container.cardinality = cardinalityOld; + container.array = newBuf; + return container; + } + const newcontainer = new RoaringBitmapArray(cardinalityOld, newBuf); + this.containers[mid] = newcontainer; + return newcontainer; + } else { + return container; + } + } + /** + * @param {RoaringBitmap} that + * @returns {RoaringBitmap} + */ + union(that) { + if (this.isEmpty()) { + return that; + } + if (that.isEmpty()) { + return this; + } + if (this === RoaringBitmap.everything() || that === RoaringBitmap.everything()) { + return RoaringBitmap.everything(); + } + let i = 0; + const il = this.containers.length; + let j = 0; + const jl = that.containers.length; + const result = new RoaringBitmap(null, 0); + result.keysAndCardinalities = new Uint8Array((il + jl) * 4); + while (i < il || j < jl) { + const ik = i * 4; + const jk = j * 4; + const k = result.containers.length * 4; + if (j >= jl || (i < il && ( + (this.keysAndCardinalities[ik + 1] < that.keysAndCardinalities[jk + 1]) || + (this.keysAndCardinalities[ik + 1] === that.keysAndCardinalities[jk + 1] && + this.keysAndCardinalities[ik] < that.keysAndCardinalities[jk]) + ))) { + result.keysAndCardinalities[k + 0] = this.keysAndCardinalities[ik + 0]; + result.keysAndCardinalities[k + 1] = this.keysAndCardinalities[ik + 1]; + result.keysAndCardinalities[k + 2] = this.keysAndCardinalities[ik + 2]; + result.keysAndCardinalities[k + 3] = this.keysAndCardinalities[ik + 3]; + result.containers.push(this.containers[i]); + i += 1; + } else if (i >= il || (j < jl && ( + (that.keysAndCardinalities[jk + 1] < this.keysAndCardinalities[ik + 1]) || + (that.keysAndCardinalities[jk + 1] === this.keysAndCardinalities[ik + 1] && + that.keysAndCardinalities[jk] < this.keysAndCardinalities[ik]) + ))) { + result.keysAndCardinalities[k + 0] = that.keysAndCardinalities[jk + 0]; + result.keysAndCardinalities[k + 1] = that.keysAndCardinalities[jk + 1]; + result.keysAndCardinalities[k + 2] = that.keysAndCardinalities[jk + 2]; + result.keysAndCardinalities[k + 3] = that.keysAndCardinalities[jk + 3]; + result.containers.push(that.containers[j]); + j += 1; + } else { + // this key is not smaller than that key + // that key is not smaller than this key + // they must be equal + const thisContainer = this.containers[i]; + const thatContainer = that.containers[j]; + let card = 0; + if (thisContainer instanceof RoaringBitmapBits && + thatContainer instanceof RoaringBitmapBits + ) { + const resultArray = new Uint8Array( + thisContainer.array.length > thatContainer.array.length ? + thisContainer.array.length : + thatContainer.array.length, + ); + let k = 0; + const kl = resultArray.length; + while (k < kl) { + const c = thisContainer.array[k] | thatContainer.array[k]; + resultArray[k] = c; + card += bitCount(c); + k += 1; + } + result.containers.push(new RoaringBitmapBits(resultArray)); + } else { + const thisValues = thisContainer.values(); + const thatValues = thatContainer.values(); + let thisResult = thisValues.next(); + let thatResult = thatValues.next(); + /** @type {Array<number>} */ + const resultValues = []; + while (!thatResult.done || !thisResult.done) { + // generator will definitely implement the iterator protocol correctly + /** @type {number} */ + const thisValue = thisResult.value; + /** @type {number} */ + const thatValue = thatResult.value; + if (thatResult.done || thisValue < thatValue) { + resultValues.push(thisValue); + thisResult = thisValues.next(); + } else if (thisResult.done || thatValue < thisValue) { + resultValues.push(thatValue); + thatResult = thatValues.next(); + } else { + // this value is not smaller than that value + // that value is not smaller than this value + // they must be equal + resultValues.push(thisValue); + thisResult = thisValues.next(); + thatResult = thatValues.next(); + } + } + const resultArray = new Uint8Array(resultValues.length * 2); + let k = 0; + for (const value of resultValues) { + // roaring bitmap is little endian + resultArray[k] = value & 0xFF; + resultArray[k + 1] = (value >> 8) & 0xFF; + k += 2; + } + result.containers.push(new RoaringBitmapArray( + resultValues.length, + resultArray, + )); + card = resultValues.length; + } + result.keysAndCardinalities[k + 0] = this.keysAndCardinalities[ik + 0]; + result.keysAndCardinalities[k + 1] = this.keysAndCardinalities[ik + 1]; + card -= 1; + result.keysAndCardinalities[k + 2] = card; + result.keysAndCardinalities[k + 3] = card >> 8; + i += 1; + j += 1; + } + } + return result; + } + /** + * @param {RoaringBitmap} that + * @returns {RoaringBitmap} + */ + intersection(that) { + if (this.isEmpty() || that.isEmpty()) { + return EMPTY_BITMAP; + } + if (this === RoaringBitmap.everything()) { + return that; + } + if (that === RoaringBitmap.everything()) { + return this; + } + let i = 0; + const il = this.containers.length; + let j = 0; + const jl = that.containers.length; + const result = new RoaringBitmap(null, 0); + result.keysAndCardinalities = new Uint8Array((il > jl ? il : jl) * 4); + while (i < il && j < jl) { + const ik = i * 4; + const jk = j * 4; + const k = result.containers.length * 4; + if (j >= jl || (i < il && ( + (this.keysAndCardinalities[ik + 1] < that.keysAndCardinalities[jk + 1]) || + (this.keysAndCardinalities[ik + 1] === that.keysAndCardinalities[jk + 1] && + this.keysAndCardinalities[ik] < that.keysAndCardinalities[jk]) + ))) { + i += 1; + } else if (i >= il || (j < jl && ( + (that.keysAndCardinalities[jk + 1] < this.keysAndCardinalities[ik + 1]) || + (that.keysAndCardinalities[jk + 1] === this.keysAndCardinalities[ik + 1] && + that.keysAndCardinalities[jk] < this.keysAndCardinalities[ik]) + ))) { + j += 1; + } else { + // this key is not smaller than that key + // that key is not smaller than this key + // they must be equal + const thisContainer = this.containers[i]; + const thatContainer = that.containers[j]; + let card = 0; + if (thisContainer instanceof RoaringBitmapBits && + thatContainer instanceof RoaringBitmapBits + ) { + const resultArray = new Uint8Array( + thisContainer.array.length > thatContainer.array.length ? + thisContainer.array.length : + thatContainer.array.length, + ); + let k = 0; + const kl = resultArray.length; + while (k < kl) { + const c = thisContainer.array[k] & thatContainer.array[k]; + resultArray[k] = c; + card += bitCount(c); + k += 1; + } + if (card !== 0) { + result.containers.push(new RoaringBitmapBits(resultArray)); + } + } else { + const thisValues = thisContainer.values(); + const thatValues = thatContainer.values(); + let thisValue = thisValues.next(); + let thatValue = thatValues.next(); + const resultValues = []; + while (!thatValue.done && !thisValue.done) { + if (thisValue.value < thatValue.value) { + thisValue = thisValues.next(); + } else if (thatValue.value < thisValue.value) { + thatValue = thatValues.next(); + } else { + // this value is not smaller than that value + // that value is not smaller than this value + // they must be equal + resultValues.push(thisValue.value); + thisValue = thisValues.next(); + thatValue = thatValues.next(); + } + } + card = resultValues.length; + if (card !== 0) { + const resultArray = new Uint8Array(resultValues.length * 2); + let k = 0; + for (const value of resultValues) { + // roaring bitmap is little endian + resultArray[k] = value & 0xFF; + resultArray[k + 1] = (value >> 8) & 0xFF; + k += 2; + } + result.containers.push(new RoaringBitmapArray( + resultValues.length, + resultArray, + )); + } + } + if (card !== 0) { + result.keysAndCardinalities[k + 0] = this.keysAndCardinalities[ik + 0]; + result.keysAndCardinalities[k + 1] = this.keysAndCardinalities[ik + 1]; + card -= 1; + result.keysAndCardinalities[k + 2] = card; + result.keysAndCardinalities[k + 3] = card >> 8; + } + i += 1; + j += 1; + } + } + return result; + } + /** @param {number} keyvalue */ + contains(keyvalue) { + const key = keyvalue >> 16; + const value = keyvalue & 0xFFFF; + const mid = this.getContainerId(key); + return mid === -1 ? false : this.containers[mid].contains(value); + } + /** + * @param {number} key + * @returns {number} + */ + getContainerId(key) { + // Binary search algorithm copied from + // https://en.wikipedia.org/wiki/Binary_search#Procedure + // + // Format is required by specification to be sorted. + // Because keys are 16 bits and unique, length can't be + // bigger than 2**16, and because we have 32 bits of safe int, + // left + right can't overflow. + let left = 0; + let right = this.containers.length - 1; + while (left <= right) { + const mid = Math.floor((left + right) / 2); + const x = this.keysAndCardinalities[(mid * 4)] | + (this.keysAndCardinalities[(mid * 4) + 1] << 8); + if (x < key) { + left = mid + 1; + } else if (x > key) { + right = mid - 1; + } else { + return mid; + } + } + return -1; + } + * entries() { + const l = this.containers.length; + for (let i = 0; i < l; ++i) { + const key = this.keysAndCardinalities[i * 4] | + (this.keysAndCardinalities[(i * 4) + 1] << 8); + for (const value of this.containers[i].values()) { + yield (key << 16) | value; + } + } + } + /** + * @returns {number|null} + */ + first() { + for (const entry of this.entries()) { + return entry; + } + return null; + } + /** + * @returns {number} + */ + cardinality() { + let result = 0; + const l = this.containers.length; + for (let i = 0; i < l; ++i) { + const card = this.keysAndCardinalities[(i * 4) + 2] | + (this.keysAndCardinalities[(i * 4) + 3] << 8); + result += card + 1; + } + return result; + } +} + +class RoaringBitmapRun { + /** + * @param {number} runcount + * @param {Uint8Array} array + */ + constructor(runcount, array) { + this.runcount = runcount; + this.array = array; + } + /** @param {number} value */ + contains(value) { + // Binary search algorithm copied from + // https://en.wikipedia.org/wiki/Binary_search#Procedure + // + // Since runcount is stored as 16 bits, left + right + // can't overflow. + let left = 0; + let right = this.runcount - 1; + while (left <= right) { + const mid = (left + right) >> 1; + const i = mid * 4; + const start = this.array[i] | (this.array[i + 1] << 8); + const lenm1 = this.array[i + 2] | (this.array[i + 3] << 8); + if ((start + lenm1) < value) { + left = mid + 1; + } else if (start > value) { + right = mid - 1; + } else { + return true; + } + } + return false; + } + * values() { + let i = 0; + while (i < this.runcount) { + const start = this.array[i * 4] | (this.array[(i * 4) + 1] << 8); + const lenm1 = this.array[(i * 4) + 2] | (this.array[(i * 4) + 3] << 8); + let value = start; + let j = 0; + while (j <= lenm1) { + yield value; + value += 1; + j += 1; + } + i += 1; + } + } +} +class RoaringBitmapArray { + /** + * @param {number} cardinality + * @param {Uint8Array} array + */ + constructor(cardinality, array) { + this.cardinality = cardinality; + this.array = array; + } + /** @param {number} value */ + contains(value) { + // Binary search algorithm copied from + // https://en.wikipedia.org/wiki/Binary_search#Procedure + // + // Since cardinality can't be higher than 4096, left + right + // cannot overflow. + let left = 0; + let right = this.cardinality - 1; + while (left <= right) { + const mid = (left + right) >> 1; + const i = mid * 2; + const x = this.array[i] | (this.array[i + 1] << 8); + if (x < value) { + left = mid + 1; + } else if (x > value) { + right = mid - 1; + } else { + return true; + } + } + return false; + } + /** @returns {Generator<number>} */ + * values() { + let i = 0; + const l = this.cardinality * 2; + while (i < l) { + yield this.array[i] | (this.array[i + 1] << 8); + i += 2; + } + } +} +class RoaringBitmapBits { + /** + * @param {Uint8Array} array + */ + constructor(array) { + this.array = array; + } + /** @param {number} value */ + contains(value) { + return !!(this.array[value >> 3] & (1 << (value & 7))); + } + * values() { + let i = 0; + const l = this.array.length << 3; + while (i < l) { + if (this.contains(i)) { + yield i; + } + i += 1; + } + } +} + +const EMPTY_BITMAP = new RoaringBitmap(null, 0); +EMPTY_BITMAP.consumed_len_bytes = 0; +const EMPTY_BITMAP1 = new RoaringBitmap(null, 0); +EMPTY_BITMAP1.consumed_len_bytes = 1; +const EVERYTHING_BITMAP = new RoaringBitmap(null, 0); + +/** + * A mapping from six byte nodeids to an arbitrary value. + * We don't just use `Map` because that requires double hashing. + * @template T + * @property {Uint8Array} keys + * @property {T[]} values + * @property {number} size + * @property {number} capacityClass + */ +class HashTable { + /** + * Construct an empty hash table. + */ + constructor() { + this.keys = EMPTY_UINT8; + /** @type {(T|undefined)[]} */ + this.values = []; + this.size = 0; + this.capacityClass = 0; + } + /** + * @returns {Generator<[Uint8Array, T]>} + */ + * entries() { + const keys = this.keys; + const values = this.values; + const l = this.values.length; + for (let i = 0; i < l; i += 1) { + const value = values[i]; + if (value !== undefined) { + yield [keys.subarray(i * 6, (i + 1) * 6), value]; + } + } + } + /** + * Add a value to the hash table. + * @param {Uint8Array} key + * @param {T} value + */ + set(key, value) { + // 90 % load factor + if (this.size * 10 >= this.values.length * 9) { + const keys = this.keys; + const values = this.values; + const l = values.length; + this.capacityClass += 1; + const capacity = 1 << this.capacityClass; + this.keys = new Uint8Array(capacity * 6); + this.values = []; + for (let i = 0; i < capacity; i += 1) { + this.values.push(undefined); + } + this.size = 0; + for (let i = 0; i < l; i += 1) { + const oldValue = values[i]; + if (oldValue !== undefined) { + this.setNoGrow(keys, i * 6, oldValue); + } + } + } + this.setNoGrow(key, 0, value); + } + /** + * @param {Uint8Array} key + * @param {number} start + * @param {T} value + */ + setNoGrow(key, start, value) { + const mask = ~(0xffffffff << this.capacityClass); + const keys = this.keys; + const values = this.values; + const l = 1 << this.capacityClass; + // because we know that our values are already hashed, + // just chop off the lower four bytes + let slot = ( + (key[start + 2] << 24) | + (key[start + 3] << 16) | + (key[start + 4] << 8) | + key[start + 5] + ) & mask; + for (let distance = 0; distance < l; ) { + const j = slot * 6; + const otherValue = values[slot]; + if (otherValue === undefined) { + values[slot] = value; + const keysStart = slot * 6; + keys[keysStart + 0] = key[start + 0]; + keys[keysStart + 1] = key[start + 1]; + keys[keysStart + 2] = key[start + 2]; + keys[keysStart + 3] = key[start + 3]; + keys[keysStart + 4] = key[start + 4]; + keys[keysStart + 5] = key[start + 5]; + this.size += 1; + break; + } else if ( + key[start + 0] === keys[j + 0] && + key[start + 1] === keys[j + 1] && + key[start + 2] === keys[j + 2] && + key[start + 3] === keys[j + 3] && + key[start + 4] === keys[j + 4] && + key[start + 5] === keys[j + 5] + ) { + values[slot] = value; + break; + } else { + const otherPreferredSlot = ( + (keys[j + 2] << 24) | (keys[j + 3] << 16) | + (keys[j + 4] << 8) | keys[j + 5] + ) & mask; + const otherDistance = otherPreferredSlot <= slot ? + slot - otherPreferredSlot : + (l - otherPreferredSlot) + slot; + if (distance > otherDistance) { + // if the other key is closer to its preferred slot than this one, + // then insert our node in its place and swap + // + // https://cglab.ca/~abeinges/blah/robinhood-part-1/ + const otherKey = keys.slice(j, j + 6); + values[slot] = value; + value = otherValue; + keys[j + 0] = key[start + 0]; + keys[j + 1] = key[start + 1]; + keys[j + 2] = key[start + 2]; + keys[j + 3] = key[start + 3]; + keys[j + 4] = key[start + 4]; + keys[j + 5] = key[start + 5]; + key = otherKey; + start = 0; + distance = otherDistance; + } + distance += 1; + slot = (slot + 1) & mask; + } + } + } + /** + * Retrieve a value + * @param {Uint8Array} key + * @returns {T|undefined} + */ + get(key) { + if (key.length !== 6) { + throw "invalid key"; + } + return this.getWithOffsetKey(key, 0); + } + /** + * Retrieve a value + * @param {Uint8Array} key + * @param {number} start + * @returns {T|undefined} + */ + getWithOffsetKey(key, start) { + const mask = ~(0xffffffff << this.capacityClass); + const keys = this.keys; + const values = this.values; + const l = 1 << this.capacityClass; + // because we know that our values are already hashed, + // just chop off the lower four bytes + let slot = ( + (key[start + 2] << 24) | + (key[start + 3] << 16) | + (key[start + 4] << 8) | + key[start + 5] + ) & mask; + for (let distance = 0; distance < l; distance += 1) { + const j = slot * 6; + const value = values[slot]; + if (value === undefined) { + break; + } else if ( + key[start + 0] === keys[j + 0] && + key[start + 1] === keys[j + 1] && + key[start + 2] === keys[j + 2] && + key[start + 3] === keys[j + 3] && + key[start + 4] === keys[j + 4] && + key[start + 5] === keys[j + 5] + ) { + return value; + } else { + const otherPreferredSlot = ( + (keys[j + 2] << 24) | (keys[j + 3] << 16) | + (keys[j + 4] << 8) | keys[j + 5] + ) & mask; + const otherDistance = otherPreferredSlot <= slot ? + slot - otherPreferredSlot : + (l - otherPreferredSlot) + slot; + if (distance > otherDistance) { + break; + } + } + slot = (slot + 1) & mask; + } + return undefined; + } +} + +/*eslint-disable */ +// ignore-tidy-linelength +/** <https://stackoverflow.com/questions/43122082/efficiently-count-the-number-of-bits-in-an-integer-in-javascript> + * @param {number} n + * @returns {number} + */ +function bitCount(n) { + n = (~~n) - ((n >> 1) & 0x55555555); + n = (n & 0x33333333) + ((n >> 2) & 0x33333333); + return ((n + (n >> 4) & 0xF0F0F0F) * 0x1010101) >> 24; +} +/*eslint-enable */ + +/** + * @param {stringdex.Hooks} hooks + * @returns {Promise<stringdex.Database>} + */ +function loadDatabase(hooks) { + /** @type {stringdex.Callbacks} */ + const callbacks = { + rr_: function(data) { + const dataObj = JSON.parse(data); + for (const colName of Object.keys(dataObj)) { + if (Object.hasOwn(dataObj[colName], "I")) { + registry.searchTreeRoots.set( + colName, + makeSearchTreeFromBase64(dataObj[colName].I)[1], + ); + } + if (Object.hasOwn(dataObj[colName], "N")) { + const counts = []; + const countsstring = dataObj[colName]["N"]; + let i = 0; + const l = countsstring.length; + while (i < l) { + let n = 0; + let c = countsstring.charCodeAt(i); + while (c < 96) { // 96 = "`" + n = (n << 4) | (c & 0xF); + i += 1; + c = countsstring.charCodeAt(i); + } + n = (n << 4) | (c & 0xF); + counts.push(n); + i += 1; + } + registry.dataColumns.set(colName, new DataColumn( + counts, + makeUint8ArrayFromBase64(dataObj[colName]["H"]), + new RoaringBitmap(makeUint8ArrayFromBase64(dataObj[colName]["E"]), 0), + colName, + )); + } + } + const cb = registry.searchTreeRootCallback; + if (cb) { + cb(null, new Database(registry.searchTreeRoots, registry.dataColumns)); + } + }, + err_rr_: function(err) { + const cb = registry.searchTreeRootCallback; + if (cb) { + cb(err, null); + } + }, + rd_: function(dataString) { + const l = dataString.length; + const data = new Uint8Array(l); + for (let i = 0; i < l; ++i) { + data[i] = dataString.charCodeAt(i); + } + loadColumnFromBytes(data); + }, + err_rd_: function(filename, err) { + const nodeid = makeUint8ArrayFromHex(filename); + const cb = registry.dataColumnLoadPromiseCallbacks.get(nodeid); + if (cb) { + cb(err, null); + } + }, + rb_: function(dataString64) { + loadColumnFromBytes(makeUint8ArrayFromBase64(dataString64)); + }, + err_rb_: function(filename, err) { + const nodeid = makeUint8ArrayFromHex(filename); + const cb = registry.dataColumnLoadPromiseCallbacks.get(nodeid); + if (cb) { + cb(err, null); + } + }, + rn_: function(inputBase64) { + const [nodeid, tree] = makeSearchTreeFromBase64(inputBase64); + const cb = registry.searchTreeLoadPromiseCallbacks.get(nodeid); + if (cb) { + cb(null, tree); + registry.searchTreeLoadPromiseCallbacks.set(nodeid, null); + } + }, + err_rn_: function(filename, err) { + const nodeid = makeUint8ArrayFromHex(filename); + const cb = registry.searchTreeLoadPromiseCallbacks.get(nodeid); + if (cb) { + cb(err, null); + } + }, + }; + + /** + * @type {{ + * searchTreeRoots: Map<string, SearchTree>; + * searchTreeLoadPromiseCallbacks: HashTable<(function(any, SearchTree?): any)|null>; + * searchTreePromises: HashTable<Promise<SearchTree>>; + * dataColumnLoadPromiseCallbacks: HashTable<function(any, Uint8Array[]?): any>; + * dataColumns: Map<string, DataColumn>; + * dataColumnsBuckets: Map<string, HashTable<Promise<Uint8Array[]>>>; + * searchTreeLoadByNodeID: function(Uint8Array): Promise<SearchTree>; + * searchTreeRootCallback?: function(any, Database?): any; + * dataLoadByNameAndHash: function(string, Uint8Array): Promise<Uint8Array[]>; + * }} + */ + const registry = { + searchTreeRoots: new Map(), + searchTreeLoadPromiseCallbacks: new HashTable(), + searchTreePromises: new HashTable(), + dataColumnLoadPromiseCallbacks: new HashTable(), + dataColumns: new Map(), + dataColumnsBuckets: new Map(), + searchTreeLoadByNodeID: function(nodeid) { + const existingPromise = registry.searchTreePromises.get(nodeid); + if (existingPromise) { + return existingPromise; + } + /** @type {Promise<SearchTree>} */ + let newPromise; + if ((nodeid[0] & 0x80) !== 0) { + const isWhole = (nodeid[0] & 0x40) !== 0; + let leaves; + if ((nodeid[0] & 0x10) !== 0) { + let id1 = (nodeid[2] << 8) | nodeid[3]; + if ((nodeid[0] & 0x20) !== 0) { + // when data is present, id1 can be up to 20 bits + id1 |= ((nodeid[1] & 0x0f) << 16); + } else { + // otherwise, we fit in 28 + id1 |= ((nodeid[0] & 0x0f) << 24) | (nodeid[1] << 16); + } + const id2 = id1 + ((nodeid[4] << 8) | nodeid[5]); + leaves = RoaringBitmap.makeSingleton(id1) + .union(RoaringBitmap.makeSingleton(id2)); + } else { + leaves = RoaringBitmap.makeSingleton( + (nodeid[2] << 24) | (nodeid[3] << 16) | + (nodeid[4] << 8) | nodeid[5], + ); + } + const data = (nodeid[0] & 0x20) !== 0 ? + Uint8Array.of(((nodeid[0] & 0x0f) << 4) | (nodeid[1] >> 4)) : + EMPTY_UINT8; + newPromise = Promise.resolve(new SearchTree( + EMPTY_SEARCH_TREE_BRANCHES, + EMPTY_SEARCH_TREE_BRANCHES, + data, + isWhole ? leaves : EMPTY_BITMAP, + isWhole ? EMPTY_BITMAP : leaves, + )); + } else { + const hashHex = makeHexFromUint8Array(nodeid); + newPromise = new Promise((resolve, reject) => { + const cb = registry.searchTreeLoadPromiseCallbacks.get(nodeid); + if (cb) { + registry.searchTreeLoadPromiseCallbacks.set(nodeid, (err, data) => { + cb(err, data); + if (data) { + resolve(data); + } else { + reject(err); + } + }); + } else { + registry.searchTreeLoadPromiseCallbacks.set(nodeid, (err, data) => { + if (data) { + resolve(data); + } else { + reject(err); + } + }); + hooks.loadTreeByHash(hashHex); + } + }); + } + registry.searchTreePromises.set(nodeid, newPromise); + return newPromise; + }, + dataLoadByNameAndHash: function(name, hash) { + let dataColumnBuckets = registry.dataColumnsBuckets.get(name); + if (dataColumnBuckets === undefined) { + dataColumnBuckets = new HashTable(); + registry.dataColumnsBuckets.set(name, dataColumnBuckets); + } + const existingBucket = dataColumnBuckets.get(hash); + if (existingBucket) { + return existingBucket; + } + const hashHex = makeHexFromUint8Array(hash); + /** @type {Promise<Uint8Array[]>} */ + const newBucket = new Promise((resolve, reject) => { + const cb = registry.dataColumnLoadPromiseCallbacks.get(hash); + if (cb) { + registry.dataColumnLoadPromiseCallbacks.set(hash, (err, data) => { + cb(err, data); + if (data) { + resolve(data); + } else { + reject(err); + } + }); + } else { + registry.dataColumnLoadPromiseCallbacks.set(hash, (err, data) => { + if (data) { + resolve(data); + } else { + reject(err); + } + }); + hooks.loadDataByNameAndHash(name, hashHex); + } + }); + dataColumnBuckets.set(hash, newBucket); + return newBucket; + }, + }; + + /** + * The set of child subtrees. + * @type {{ + * nodeids: Uint8Array, + * subtrees: Array<Promise<SearchTree>|null>, + * }} + */ + class SearchTreeBranches { + /** + * Construct the subtree list with `length` nulls + * @param {number} length + * @param {Uint8Array} nodeids + */ + constructor(length, nodeids) { + this.nodeids = nodeids; + this.subtrees = []; + for (let i = 0; i < length; ++i) { + this.subtrees.push(null); + } + } + /** + * @param {number} i + * @returns {Uint8Array} + */ + getNodeID(i) { + return new Uint8Array( + this.nodeids.buffer, + this.nodeids.byteOffset + (i * 6), + 6, + ); + } + // https://github.com/microsoft/TypeScript/issues/17227 + /** @returns {Generator<[number, Promise<SearchTree>|null]>} */ + entries() { + throw new Error(); + } + /** + * @param {number} _k + * @returns {number} + */ + getIndex(_k) { + throw new Error(); + } + /** + * @param {number} _i + * @returns {number} + */ + getKey(_i) { + throw new Error(); + } + /** + * @returns {Uint8Array} + */ + getKeys() { + throw new Error(); + } + } + + /** + * A sorted array of search tree branches. + * + * @type {{ + * keys: Uint8Array, + * nodeids: Uint8Array, + * subtrees: Array<Promise<SearchTree>|null>, + * }} + */ + class SearchTreeBranchesArray extends SearchTreeBranches { + /** + * @param {Uint8Array} keys + * @param {Uint8Array} nodeids + */ + constructor(keys, nodeids) { + super(keys.length, nodeids); + this.keys = keys; + let i = 1; + while (i < this.keys.length) { + if (this.keys[i - 1] >= this.keys[i]) { + throw new Error("HERE"); + } + i += 1; + } + } + /** @returns {Generator<[number, Promise<SearchTree>|null]>} */ + * entries() { + let i = 0; + const l = this.keys.length; + while (i < l) { + yield [this.keys[i], this.subtrees[i]]; + i += 1; + } + } + /** + * @param {number} k + * @returns {number} + */ + getIndex(k) { + // Since length can't be bigger than 256, + // left + right can't overflow. + let left = 0; + let right = this.keys.length - 1; + while (left <= right) { + const mid = (left + right) >> 1; + if (this.keys[mid] < k) { + left = mid + 1; + } else if (this.keys[mid] > k) { + right = mid - 1; + } else { + return mid; + } + } + return -1; + } + /** + * @param {number} i + * @returns {number} + */ + getKey(i) { + return this.keys[i]; + } + /** + * @returns {Uint8Array} + */ + getKeys() { + return this.keys; + } + } + + const EMPTY_SEARCH_TREE_BRANCHES = new SearchTreeBranchesArray( + EMPTY_UINT8, + EMPTY_UINT8, + ); + + /** @type {number[]} */ + const SHORT_ALPHABITMAP_CHARS = []; + for (let i = 0x61; i <= 0x7A; ++i) { + if (i === 0x76 || i === 0x71) { + // 24 entries, 26 letters, so we skip q and v + continue; + } + SHORT_ALPHABITMAP_CHARS.push(i); + } + + /** @type {number[]} */ + const LONG_ALPHABITMAP_CHARS = [0x31, 0x32, 0x33, 0x34, 0x35, 0x36]; + for (let i = 0x61; i <= 0x7A; ++i) { + LONG_ALPHABITMAP_CHARS.push(i); + } + + /** + * @param {number[]} alphabitmap_chars + * @param {number} width + * @return {(typeof SearchTreeBranches)&{"ALPHABITMAP_CHARS": number[], "width": number}} + */ + function makeSearchTreeBranchesAlphaBitmapClass(alphabitmap_chars, width) { + const bitwidth = width * 8; + const cls = class SearchTreeBranchesAlphaBitmap extends SearchTreeBranches { + /** + * @param {number} bitmap + * @param {Uint8Array} nodeids + */ + constructor(bitmap, nodeids) { + super(nodeids.length / 6, nodeids); + if (nodeids.length / 6 !== bitCount(bitmap)) { + throw new Error(`mismatch ${bitmap} ${nodeids}`); + } + this.bitmap = bitmap; + this.nodeids = nodeids; + } + /** @returns {Generator<[number, Promise<SearchTree>|null]>} */ + * entries() { + let i = 0; + let j = 0; + while (i < bitwidth) { + if (this.bitmap & (1 << i)) { + yield [alphabitmap_chars[i], this.subtrees[j]]; + j += 1; + } + i += 1; + } + } + /** + * @param {number} k + * @returns {number} + */ + getIndex(k) { + //return this.getKeys().indexOf(k); + const ix = alphabitmap_chars.indexOf(k); + if (ix < 0) { + return ix; + } + const result = bitCount(~(0xffffffff << ix) & this.bitmap); + return result >= this.subtrees.length ? -1 : result; + } + /** + * @param {number} branch_index + * @returns {number} + */ + getKey(branch_index) { + //return this.getKeys()[branch_index]; + let alpha_index = 0; + while (branch_index >= 0) { + if (this.bitmap & (1 << alpha_index)) { + branch_index -= 1; + } + alpha_index += 1; + } + return alphabitmap_chars[alpha_index]; + } + /** + * @returns {Uint8Array} + */ + getKeys() { + const length = bitCount(this.bitmap); + const result = new Uint8Array(length); + let result_index = 0; + for (let alpha_index = 0; alpha_index < bitwidth; ++alpha_index) { + if (this.bitmap & (1 << alpha_index)) { + result[result_index] = alphabitmap_chars[alpha_index]; + result_index += 1; + } + } + return result; + } + }; + cls.ALPHABITMAP_CHARS = alphabitmap_chars; + cls.width = width; + return cls; + } + + const SearchTreeBranchesShortAlphaBitmap = + makeSearchTreeBranchesAlphaBitmapClass(SHORT_ALPHABITMAP_CHARS, 3); + + const SearchTreeBranchesLongAlphaBitmap = + makeSearchTreeBranchesAlphaBitmapClass(LONG_ALPHABITMAP_CHARS, 4); + + /** + * A [suffix tree], used for name-based search. + * + * This data structure is used to drive substring matches, + * such as matching the query "link" to `LinkedList`, + * and Lev-distance matches, such as matching the + * query "hahsmap" to `HashMap`. + * + * [suffix tree]: https://en.wikipedia.org/wiki/Suffix_tree + * + * branches + * : A sorted-array map of subtrees. + * + * data + * : The substring represented by this node. The root node + * is always empty. + * + * leaves_suffix + * : The IDs of every entry that matches. Levenshtein matches + * won't include these. + * + * leaves_whole + * : The IDs of every entry that matches exactly. Levenshtein matches + * will include these. + * + * @type {{ + * might_have_prefix_branches: SearchTreeBranches, + * branches: SearchTreeBranches, + * data: Uint8Array, + * leaves_suffix: RoaringBitmap, + * leaves_whole: RoaringBitmap, + * }} + */ + class SearchTree { + /** + * @param {SearchTreeBranches} branches + * @param {SearchTreeBranches} might_have_prefix_branches + * @param {Uint8Array} data + * @param {RoaringBitmap} leaves_whole + * @param {RoaringBitmap} leaves_suffix + */ + constructor( + branches, + might_have_prefix_branches, + data, + leaves_whole, + leaves_suffix, + ) { + this.might_have_prefix_branches = might_have_prefix_branches; + this.branches = branches; + this.data = data; + this.leaves_suffix = leaves_suffix; + this.leaves_whole = leaves_whole; + } + /** + * Returns the Trie for the root node. + * + * A Trie pointer refers to a single node in a logical decompressed search tree + * (the real search tree is compressed). + * + * @return {Trie} + */ + trie() { + return new Trie(this, 0); + } + + /** + * Return the trie representing `name` + * @param {Uint8Array|string} name + * @returns {Promise<Trie?>} + */ + async search(name) { + if (typeof name === "string") { + const utf8encoder = new TextEncoder(); + name = utf8encoder.encode(name); + } + let trie = this.trie(); + for (const datum of name) { + // code point definitely exists + const newTrie = trie.child(datum); + if (newTrie) { + trie = await newTrie; + } else { + return null; + } + } + return trie; + } + + /** + * @param {Uint8Array|string} name + * @returns {AsyncGenerator<Trie>} + */ + async* searchLev(name) { + if (typeof name === "string") { + const utf8encoder = new TextEncoder(); + name = utf8encoder.encode(name); + } + const w = name.length; + if (w < 3) { + const trie = await this.search(name); + if (trie !== null) { + yield trie; + } + return; + } + const levParams = w >= 6 ? + new Lev2TParametricDescription(w) : + new Lev1TParametricDescription(w); + /** @type {Array<[Promise<Trie>, number]>} */ + const stack = [[Promise.resolve(this.trie()), 0]]; + const n = levParams.n; + while (stack.length !== 0) { + // It's not empty + /** @type {[Promise<Trie>, number]} */ + //@ts-expect-error + const [triePromise, levState] = stack.pop(); + const trie = await triePromise; + for (const byte of trie.keysExcludeSuffixOnly()) { + const levPos = levParams.getPosition(levState); + const vector = levParams.getVector( + name, + byte, + levPos, + Math.min(w, levPos + (2 * n) + 1), + ); + const newLevState = levParams.transition( + levState, + levPos, + vector, + ); + if (newLevState >= 0) { + const child = trie.child(byte); + if (child) { + stack.push([child, newLevState]); + if (levParams.isAccept(newLevState)) { + yield child; + } + } + } + } + } + } + } + + /** + * A representation of a set of strings in the search index, + * as a subset of the entire tree. + */ + class Trie { + /** + * @param {SearchTree} tree + * @param {number} offset + */ + constructor(tree, offset) { + this.tree = tree; + this.offset = offset; + } + + /** + * All exact matches for the string represented by this node. + * @returns {RoaringBitmap} + */ + matches() { + if (this.offset === this.tree.data.length) { + return this.tree.leaves_whole; + } else { + return EMPTY_BITMAP; + } + } + + /** + * All matches for strings that contain the string represented by this node. + * @returns {AsyncGenerator<RoaringBitmap>} + */ + async* substringMatches() { + /** @type {Promise<SearchTree>[]} */ + let layer = [Promise.resolve(this.tree)]; + while (layer.length) { + const current_layer = layer; + layer = []; + for await (const tree of current_layer) { + yield tree.leaves_whole.union(tree.leaves_suffix); + } + /** @type {HashTable<[number, SearchTree][]>} */ + const subnodes = new HashTable(); + for await (const node of current_layer) { + const branches = node.branches; + const l = branches.subtrees.length; + for (let i = 0; i < l; ++i) { + const subtree = branches.subtrees[i]; + if (subtree) { + layer.push(subtree); + } else if (subtree === null) { + const byte = branches.getKey(i); + const newnode = branches.getNodeID(i); + if (!newnode) { + throw new Error(`malformed tree; no node for key ${byte}`); + } else { + let subnode_list = subnodes.get(newnode); + if (!subnode_list) { + subnode_list = [[byte, node]]; + subnodes.set(newnode, subnode_list); + } else { + subnode_list.push([byte, node]); + } + } + } else { + throw new Error(`malformed tree; index ${i} does not exist`); + } + } + } + for (const [newnode, subnode_list] of subnodes.entries()) { + const res = registry.searchTreeLoadByNodeID(newnode); + for (const [byte, node] of subnode_list) { + const branches = node.branches; + const might_have_prefix_branches = node.might_have_prefix_branches; + const i = branches.getIndex(byte); + branches.subtrees[i] = res; + const mhpI = might_have_prefix_branches.getIndex(byte); + if (mhpI !== -1) { + might_have_prefix_branches.subtrees[mhpI] = res; + } + } + layer.push(res); + } + } + } + + /** + * All matches for strings that start with the string represented by this node. + * @returns {AsyncGenerator<RoaringBitmap>} + */ + async* prefixMatches() { + /** @type {{node: Promise<SearchTree>, len: number}[]} */ + let layer = [{node: Promise.resolve(this.tree), len: 0}]; + // https://en.wikipedia.org/wiki/Heap_(data_structure)#Implementation_using_arrays + /** @type {{bitmap: RoaringBitmap, length: number}[]} */ + const backlog = []; + while (layer.length !== 0 || backlog.length !== 0) { + const current_layer = layer; + layer = []; + let minLength = null; + // push every entry in the current layer into the backlog, + // a min-heap of result entries + // we then yield the smallest ones (can't yield bigger ones + // if we want to do them in order) + for (const {node, len} of current_layer) { + const tree = await node; + const length = len + tree.data.length; + if (minLength === null || length < minLength) { + minLength = length; + } + let backlogSlot = backlog.length; + backlog.push({bitmap: tree.leaves_whole, length}); + while (backlogSlot > 0 && + backlog[backlogSlot].length < backlog[(backlogSlot - 1) >> 1].length + ) { + const parentSlot = (backlogSlot - 1) >> 1; + const parent = backlog[parentSlot]; + backlog[parentSlot] = backlog[backlogSlot]; + backlog[backlogSlot] = parent; + backlogSlot = parentSlot; + } + } + // yield nodes in length order, smallest one first + // we know that, whatever the smallest item is + // every child will be bigger than that + while (backlog.length !== 0) { + const backlogEntry = backlog[0]; + if (minLength !== null && backlogEntry.length > minLength) { + break; + } + if (!backlogEntry.bitmap.isEmpty()) { + yield backlogEntry.bitmap; + } + backlog[0] = backlog[backlog.length - 1]; + backlog.length -= 1; + let backlogSlot = 0; + const backlogLength = backlog.length; + while (backlogSlot < backlogLength) { + const leftSlot = (backlogSlot << 1) + 1; + const rightSlot = (backlogSlot << 1) + 2; + let smallest = backlogSlot; + if (leftSlot < backlogLength && + backlog[leftSlot].length < backlog[smallest].length + ) { + smallest = leftSlot; + } + if (rightSlot < backlogLength && + backlog[rightSlot].length < backlog[smallest].length + ) { + smallest = rightSlot; + } + if (smallest === backlogSlot) { + break; + } else { + const tmp = backlog[backlogSlot]; + backlog[backlogSlot] = backlog[smallest]; + backlog[smallest] = tmp; + backlogSlot = smallest; + } + } + } + // if we still have more subtrees to walk, then keep going + /** @type {HashTable<{byte: number, tree: SearchTree, len: number}[]>} */ + const subnodes = new HashTable(); + for await (const {node, len} of current_layer) { + const tree = await node; + const length = len + tree.data.length; + const mhp_branches = tree.might_have_prefix_branches; + const l = mhp_branches.subtrees.length; + for (let i = 0; i < l; ++i) { + const len = length + 1; + const subtree = mhp_branches.subtrees[i]; + if (subtree) { + layer.push({node: subtree, len}); + } else if (subtree === null) { + const byte = mhp_branches.getKey(i); + const newnode = mhp_branches.getNodeID(i); + if (!newnode) { + throw new Error(`malformed tree; no node for key ${byte}`); + } else { + let subnode_list = subnodes.get(newnode); + if (!subnode_list) { + subnode_list = [{byte, tree, len}]; + subnodes.set(newnode, subnode_list); + } else { + subnode_list.push({byte, tree, len}); + } + } + } else { + throw new Error(`malformed tree; index ${i} does not exist`); + } + } + } + for (const [newnode, subnode_list] of subnodes.entries()) { + const res = registry.searchTreeLoadByNodeID(newnode); + let len = Number.MAX_SAFE_INTEGER; + for (const {byte, tree, len: subtreelen} of subnode_list) { + if (subtreelen < len) { + len = subtreelen; + } + const mhp_branches = tree.might_have_prefix_branches; + const i = mhp_branches.getIndex(byte); + mhp_branches.subtrees[i] = res; + const branches = tree.branches; + const bi = branches.getIndex(byte); + branches.subtrees[bi] = res; + } + layer.push({node: res, len}); + } + } + } + + /** + * Returns all keys that are children of this node. + * @returns {Uint8Array} + */ + keys() { + const data = this.tree.data; + if (this.offset === data.length) { + return this.tree.branches.getKeys(); + } else { + return Uint8Array.of(data[this.offset]); + } + } + + /** + * Returns all nodes that are direct children of this node. + * @returns {[number, Promise<Trie>][]} + */ + children() { + const data = this.tree.data; + if (this.offset === data.length) { + /** @type {[number, Promise<Trie>][]} */ + const nodes = []; + let i = 0; + for (const [k, v] of this.tree.branches.entries()) { + /** @type {Promise<SearchTree>} */ + let node; + if (v) { + node = v; + } else { + const newnode = this.tree.branches.getNodeID(i); + if (!newnode) { + throw new Error(`malformed tree; no hash for key ${k}: ${newnode} \ + ${this.tree.branches.nodeids} ${this.tree.branches.getKeys()}`); + } + node = registry.searchTreeLoadByNodeID(newnode); + this.tree.branches.subtrees[i] = node; + const mhpI = this.tree.might_have_prefix_branches.getIndex(k); + if (mhpI !== -1) { + this.tree.might_have_prefix_branches.subtrees[mhpI] = node; + } + } + nodes.push([k, node.then(node => node.trie())]); + i += 1; + } + return nodes; + } else { + /** @type {number} */ + const codePoint = data[this.offset]; + const trie = new Trie(this.tree, this.offset + 1); + return [[codePoint, Promise.resolve(trie)]]; + } + } + + /** + * Returns all keys that are children of this node. + * @returns {Uint8Array} + */ + keysExcludeSuffixOnly() { + const data = this.tree.data; + if (this.offset === data.length) { + return this.tree.might_have_prefix_branches.getKeys(); + } else { + return Uint8Array.of(data[this.offset]); + } + } + + /** + * Returns all nodes that are direct children of this node. + * @returns {[number, Promise<Trie>][]} + */ + childrenExcludeSuffixOnly() { + const data = this.tree.data; + if (this.offset === data.length) { + /** @type {[number, Promise<Trie>][]} */ + const nodes = []; + let i = 0; + for (const [k, v] of this.tree.might_have_prefix_branches.entries()) { + /** @type {Promise<SearchTree>} */ + let node; + if (v) { + node = v; + } else { + const newnode = this.tree.might_have_prefix_branches.getNodeID(i); + if (!newnode) { + throw new Error(`malformed tree; no node for key ${k}`); + } + node = registry.searchTreeLoadByNodeID(newnode); + this.tree.might_have_prefix_branches.subtrees[i] = node; + this.tree.branches.subtrees[this.tree.branches.getIndex(k)] = node; + } + nodes.push([k, node.then(node => node.trie())]); + i += 1; + } + return nodes; + } else { + /** @type {number} */ + const codePoint = data[this.offset]; + const trie = new Trie(this.tree, this.offset + 1); + return [[codePoint, Promise.resolve(trie)]]; + } + } + + /** + * Returns a single node that is a direct child of this node. + * @param {number} byte + * @returns {Promise<Trie>?} + */ + child(byte) { + if (this.offset === this.tree.data.length) { + const i = this.tree.branches.getIndex(byte); + if (i !== -1) { + let branch = this.tree.branches.subtrees[i]; + if (branch === null) { + const newnode = this.tree.branches.getNodeID(i); + if (!newnode) { + throw new Error(`malformed tree; no node for key ${byte}`); + } + branch = registry.searchTreeLoadByNodeID(newnode); + this.tree.branches.subtrees[i] = branch; + const mhpI = this.tree.might_have_prefix_branches.getIndex(byte); + if (mhpI !== -1) { + this.tree.might_have_prefix_branches.subtrees[mhpI] = branch; + } + } + return branch.then(branch => branch.trie()); + } + } else if (this.tree.data[this.offset] === byte) { + return Promise.resolve(new Trie(this.tree, this.offset + 1)); + } + return null; + } + } + + class DataColumn { + /** + * Construct the wrapper object for a data column. + * @param {number[]} counts + * @param {Uint8Array} hashes + * @param {RoaringBitmap} emptyset + * @param {string} name + */ + constructor(counts, hashes, emptyset, name) { + this.hashes = hashes; + this.emptyset = emptyset; + this.name = name; + /** @type {{"hash": Uint8Array, "data": Promise<Uint8Array[]>?, "end": number}[]} */ + this.buckets = []; + this.bucket_keys = []; + const l = counts.length; + let k = 0; + let totalLength = 0; + for (let i = 0; i < l; ++i) { + const count = counts[i]; + totalLength += count; + const start = k; + for (let j = 0; j < count; ++j) { + if (emptyset.contains(k)) { + j -= 1; + } + k += 1; + } + const end = k; + const bucket = {hash: hashes.subarray(i * 6, (i + 1) * 6), data: null, end, count}; + this.buckets.push(bucket); + this.bucket_keys.push(start); + } + this.length = totalLength; + } + /** + * Check if a cell contains the empty string. + * @param {number} id + * @returns {boolean} + */ + isEmpty(id) { + return this.emptyset.contains(id); + } + /** + * Look up a cell by row ID. + * @param {number} id + * @returns {Promise<Uint8Array|undefined>} + */ + async at(id) { + if (this.emptyset.contains(id)) { + return Promise.resolve(EMPTY_UINT8); + } else { + let idx = -1; + while (this.bucket_keys[idx + 1] <= id) { + idx += 1; + } + if (idx === -1 || idx >= this.bucket_keys.length) { + return Promise.resolve(undefined); + } else { + const start = this.bucket_keys[idx]; + const {hash, end} = this.buckets[idx]; + let data = this.buckets[idx].data; + if (data === null) { + const dataSansEmptyset = await registry.dataLoadByNameAndHash( + this.name, + hash, + ); + // After the `await` resolves, another task might fill + // in the data. If so, we should use that. + data = this.buckets[idx].data; + if (data !== null) { + return (await data)[id - start]; + } + /** @type {(Uint8Array[])|null} */ + let dataWithEmptyset = null; + let pos = start; + let insertCount = 0; + while (pos < end) { + if (this.emptyset.contains(pos)) { + if (dataWithEmptyset === null) { + dataWithEmptyset = dataSansEmptyset.splice(0, insertCount); + } else if (insertCount !== 0) { + dataWithEmptyset.push( + ...dataSansEmptyset.splice(0, insertCount), + ); + } + insertCount = 0; + dataWithEmptyset.push(EMPTY_UINT8); + } else { + insertCount += 1; + } + pos += 1; + } + data = Promise.resolve( + dataWithEmptyset === null ? + dataSansEmptyset : + dataWithEmptyset.concat(dataSansEmptyset), + ); + this.buckets[idx].data = data; + } + return (await data)[id - start]; + } + } + } + } + + class Database { + /** + * The primary frontend for accessing data in this index. + * + * @param {Map<string, SearchTree>} searchTreeRoots + * @param {Map<string, DataColumn>} dataColumns + */ + constructor(searchTreeRoots, dataColumns) { + this.searchTreeRoots = searchTreeRoots; + this.dataColumns = dataColumns; + } + /** + * Search a column by name, returning verbatim matched IDs. + * @param {string} colname + * @returns {SearchTree|undefined} + */ + getIndex(colname) { + return this.searchTreeRoots.get(colname); + } + /** + * Look up a cell by column ID and row ID. + * @param {string} colname + * @returns {DataColumn|undefined} + */ + getData(colname) { + return this.dataColumns.get(colname); + } + } + + /** + * Load a data column. + * @param {Uint8Array} data + */ + function loadColumnFromBytes(data) { + const hashBuf = Uint8Array.of(0, 0, 0, 0, 0, 0, 0, 0); + const truncatedHash = hashBuf.subarray(2, 8); + siphashOfBytes(data, 0, 0, 0, 0, hashBuf); + const cb = registry.dataColumnLoadPromiseCallbacks.get(truncatedHash); + if (cb) { + const backrefs = []; + const dataSansEmptyset = []; + let i = 0; + const l = data.length; + while (i < l) { + let c = data[i]; + if (c >= 48 && c <= 63) { // 48 = "0", 63 = "?" + dataSansEmptyset.push(backrefs[c - 48]); + i += 1; + } else { + let n = 0; + while (c < 96) { // 96 = "`" + n = (n << 4) | (c & 0xF); + i += 1; + c = data[i]; + } + n = (n << 4) | (c & 0xF); + i += 1; + const item = data.subarray(i, i + n); + dataSansEmptyset.push(item); + i += n; + backrefs.unshift(item); + if (backrefs.length > 16) { + backrefs.pop(); + } + } + } + cb(null, dataSansEmptyset); + } + } + + /** + * @param {string} inputBase64 + * @returns {[Uint8Array, SearchTree]} + */ + function makeSearchTreeFromBase64(inputBase64) { + const input = makeUint8ArrayFromBase64(inputBase64); + let i = 0; + const l = input.length; + /** @type {HashTable<SearchTree>} */ + const stash = new HashTable(); + const hash = Uint8Array.of(0, 0, 0, 0, 0, 0, 0, 0); + const truncatedHash = new Uint8Array(hash.buffer, 2, 6); + // used for handling compressed (that is, relative-offset) nodes + /** @type {{hash: Uint8Array, used: boolean}[]} */ + const hash_history = []; + /** @type {Uint8Array[]} */ + const data_history = []; + let canonical = EMPTY_UINT8; + /** @type {SearchTree} */ + let tree = new SearchTree( + EMPTY_SEARCH_TREE_BRANCHES, + EMPTY_SEARCH_TREE_BRANCHES, + EMPTY_UINT8, + EMPTY_BITMAP, + EMPTY_BITMAP, + ); + /** + * @param {Uint8Array} input + * @param {number} i + * @param {number} compression_tag + * @returns {{ + * "cpbranches": Uint8Array, + * "csbranches": Uint8Array, + * "might_have_prefix_branches": SearchTreeBranches, + * "branches": SearchTreeBranches, + * "cpnodes": Uint8Array, + * "csnodes": Uint8Array, + * "consumed_len_bytes": number, + * }} + */ + function makeBranchesFromBinaryData( + input, + i, + compression_tag, + ) { + const is_pure_suffixes_only_node = (compression_tag & 0x01) !== 0x00; + const is_stack_compressed = (compression_tag & 0x02) !== 0; + const is_long_compressed = (compression_tag & 0x04) !== 0; + const all_children_are_compressed = + (compression_tag & 0xF0) === 0xF0 && !is_long_compressed; + const any_children_are_compressed = + (compression_tag & 0xF0) !== 0x00 || is_long_compressed; + const start_point = i; + let cplen; + let cslen; + let alphabitmap = null; + if (is_pure_suffixes_only_node) { + cplen = 0; + cslen = input[i]; + i += 1; + if (cslen >= 0xc0) { + alphabitmap = SearchTreeBranchesLongAlphaBitmap; + cslen = cslen & 0x3F; + } else if (cslen >= 0x80) { + alphabitmap = SearchTreeBranchesShortAlphaBitmap; + cslen = cslen & 0x7F; + } + } else { + cplen = input[i]; + i += 1; + cslen = input[i]; + i += 1; + if (cplen === 0xff && cslen === 0xff) { + cplen = 0x100; + cslen = 0; + } else if (cplen >= 0xc0 && cslen >= 0xc0) { + alphabitmap = SearchTreeBranchesLongAlphaBitmap; + cplen = cplen & 0x3F; + cslen = cslen & 0x3F; + } else if (cplen >= 0x80 && cslen >= 0x80) { + alphabitmap = SearchTreeBranchesShortAlphaBitmap; + cplen = cplen & 0x7F; + cslen = cslen & 0x7F; + } + } + let j = 0; + /** @type {Uint8Array} */ + let cpnodes; + if (any_children_are_compressed) { + cpnodes = cplen === 0 ? EMPTY_UINT8 : new Uint8Array(cplen * 6); + while (j < cplen) { + const is_compressed = all_children_are_compressed || + ((0x10 << j) & compression_tag) !== 0; + if (is_compressed) { + let slot = hash_history.length - 1; + if (is_stack_compressed) { + while (hash_history[slot].used) { + slot -= 1; + } + } else { + slot -= input[i]; + i += 1; + } + hash_history[slot].used = true; + cpnodes.set( + hash_history[slot].hash, + j * 6, + ); + } else { + const joff = j * 6; + cpnodes[joff + 0] = input[i + 0]; + cpnodes[joff + 1] = input[i + 1]; + cpnodes[joff + 2] = input[i + 2]; + cpnodes[joff + 3] = input[i + 3]; + cpnodes[joff + 4] = input[i + 4]; + cpnodes[joff + 5] = input[i + 5]; + i += 6; + } + j += 1; + } + } else { + cpnodes = cplen === 0 ? EMPTY_UINT8 : input.subarray(i, i + (cplen * 6)); + i += cplen * 6; + } + j = 0; + /** @type {Uint8Array} */ + let csnodes; + if (any_children_are_compressed) { + csnodes = cslen === 0 ? EMPTY_UINT8 : new Uint8Array(cslen * 6); + while (j < cslen) { + const is_compressed = all_children_are_compressed || + ((0x10 << (cplen + j)) & compression_tag) !== 0; + if (is_compressed) { + let slot = hash_history.length - 1; + if (is_stack_compressed) { + while (hash_history[slot].used) { + slot -= 1; + } + } else { + slot -= input[i]; + i += 1; + } + hash_history[slot].used = true; + csnodes.set( + hash_history[slot].hash, + j * 6, + ); + } else { + const joff = j * 6; + csnodes[joff + 0] = input[i + 0]; + csnodes[joff + 1] = input[i + 1]; + csnodes[joff + 2] = input[i + 2]; + csnodes[joff + 3] = input[i + 3]; + csnodes[joff + 4] = input[i + 4]; + csnodes[joff + 5] = input[i + 5]; + i += 6; + } + j += 1; + } + } else { + csnodes = cslen === 0 ? EMPTY_UINT8 : input.subarray(i, i + (cslen * 6)); + i += cslen * 6; + } + let cpbranches; + let might_have_prefix_branches; + if (cplen === 0) { + cpbranches = EMPTY_UINT8; + might_have_prefix_branches = EMPTY_SEARCH_TREE_BRANCHES; + } else if (alphabitmap) { + cpbranches = new Uint8Array(input.buffer, i + input.byteOffset, alphabitmap.width); + const branchset = (alphabitmap.width === 4 ? (input[i + 3] << 24) : 0) | + (input[i + 2] << 16) | + (input[i + 1] << 8) | + input[i]; + might_have_prefix_branches = new alphabitmap(branchset, cpnodes); + i += alphabitmap.width; + } else { + cpbranches = new Uint8Array(input.buffer, i + input.byteOffset, cplen); + might_have_prefix_branches = new SearchTreeBranchesArray(cpbranches, cpnodes); + i += cplen; + } + let csbranches; + let branches; + if (cslen === 0) { + csbranches = EMPTY_UINT8; + branches = might_have_prefix_branches; + } else if (alphabitmap) { + csbranches = new Uint8Array(input.buffer, i + input.byteOffset, alphabitmap.width); + const branchset = (alphabitmap.width === 4 ? (input[i + 3] << 24) : 0) | + (input[i + 2] << 16) | + (input[i + 1] << 8) | + input[i]; + if (cplen === 0) { + branches = new alphabitmap(branchset, csnodes); + } else { + const cpoffset = i - alphabitmap.width; + const cpbranchset = + (alphabitmap.width === 4 ? (input[cpoffset + 3] << 24) : 0) | + (input[cpoffset + 2] << 16) | + (input[cpoffset + 1] << 8) | + input[cpoffset]; + const hashes = new Uint8Array((cplen + cslen) * 6); + let cpi = 0; + let csi = 0; + let j = 0; + for (let k = 0; k < alphabitmap.ALPHABITMAP_CHARS.length; k += 1) { + if (branchset & (1 << k)) { + hashes[j + 0] = csnodes[csi + 0]; + hashes[j + 1] = csnodes[csi + 1]; + hashes[j + 2] = csnodes[csi + 2]; + hashes[j + 3] = csnodes[csi + 3]; + hashes[j + 4] = csnodes[csi + 4]; + hashes[j + 5] = csnodes[csi + 5]; + j += 6; + csi += 6; + } else if (cpbranchset & (1 << k)) { + hashes[j + 0] = cpnodes[cpi + 0]; + hashes[j + 1] = cpnodes[cpi + 1]; + hashes[j + 2] = cpnodes[cpi + 2]; + hashes[j + 3] = cpnodes[cpi + 3]; + hashes[j + 4] = cpnodes[cpi + 4]; + hashes[j + 5] = cpnodes[cpi + 5]; + j += 6; + cpi += 6; + } + } + branches = new alphabitmap(branchset | cpbranchset, hashes); + } + i += alphabitmap.width; + } else { + csbranches = new Uint8Array(input.buffer, i + input.byteOffset, cslen); + if (cplen === 0) { + branches = new SearchTreeBranchesArray(csbranches, csnodes); + } else { + const branchset = new Uint8Array(cplen + cslen); + const hashes = new Uint8Array((cplen + cslen) * 6); + let cpi = 0; + let csi = 0; + let j = 0; + while (cpi < cplen || csi < cslen) { + if (cpi >= cplen || (csi < cslen && cpbranches[cpi] > csbranches[csi])) { + branchset[j] = csbranches[csi]; + const joff = j * 6; + const csioff = csi * 6; + hashes[joff + 0] = csnodes[csioff + 0]; + hashes[joff + 1] = csnodes[csioff + 1]; + hashes[joff + 2] = csnodes[csioff + 2]; + hashes[joff + 3] = csnodes[csioff + 3]; + hashes[joff + 4] = csnodes[csioff + 4]; + hashes[joff + 5] = csnodes[csioff + 5]; + csi += 1; + } else { + branchset[j] = cpbranches[cpi]; + const joff = j * 6; + const cpioff = cpi * 6; + hashes[joff + 0] = cpnodes[cpioff + 0]; + hashes[joff + 1] = cpnodes[cpioff + 1]; + hashes[joff + 2] = cpnodes[cpioff + 2]; + hashes[joff + 3] = cpnodes[cpioff + 3]; + hashes[joff + 4] = cpnodes[cpioff + 4]; + hashes[joff + 5] = cpnodes[cpioff + 5]; + cpi += 1; + } + j += 1; + } + branches = new SearchTreeBranchesArray(branchset, hashes); + } + i += cslen; + } + return { + consumed_len_bytes: i - start_point, + cpbranches, + csbranches, + cpnodes, + csnodes, + branches, + might_have_prefix_branches, + }; + } + while (i < l) { + const start = i; + let data; + // compression_tag = 1 means pure-suffixes-only, + // which is not considered "compressed" for the purposes of this loop + // because that's the canonical, hashed version of the data + let compression_tag = input[i]; + const is_pure_suffixes_only_node = (compression_tag & 0x01) !== 0; + if (compression_tag > 1) { + // compressed node + const is_long_compressed = (compression_tag & 0x04) !== 0; + const is_data_compressed = (compression_tag & 0x08) !== 0; + i += 1; + if (is_long_compressed) { + compression_tag |= input[i] << 8; + i += 1; + compression_tag |= input[i] << 16; + i += 1; + } + let dlen = input[i]; + i += 1; + if (is_data_compressed) { + data = data_history[data_history.length - dlen - 1]; + dlen = data.length; + } else { + data = dlen === 0 ? + EMPTY_UINT8 : + new Uint8Array(input.buffer, i + input.byteOffset, dlen); + i += dlen; + } + const coffset = i; + const { + cpbranches, + csbranches, + cpnodes, + csnodes, + consumed_len_bytes: branches_consumed_len_bytes, + branches, + might_have_prefix_branches, + } = makeBranchesFromBinaryData(input, i, compression_tag); + i += branches_consumed_len_bytes; + let whole; + let suffix; + if (is_pure_suffixes_only_node) { + whole = EMPTY_BITMAP; + suffix = input[i] === 0 ? + EMPTY_BITMAP1 : + new RoaringBitmap(input, i); + i += suffix.consumed_len_bytes; + } else if (input[i] === 0xff) { + whole = EMPTY_BITMAP; + suffix = EMPTY_BITMAP1; + i += 1; + } else { + whole = input[i] === 0 ? + EMPTY_BITMAP1 : + new RoaringBitmap(input, i); + i += whole.consumed_len_bytes; + suffix = input[i] === 0 ? + EMPTY_BITMAP1 : + new RoaringBitmap(input, i); + i += suffix.consumed_len_bytes; + } + tree = new SearchTree( + branches, + might_have_prefix_branches, + data, + whole, + suffix, + ); + const clen = ( + (is_pure_suffixes_only_node ? 3 : 4) + // lengths of children and data + dlen + + cpnodes.length + csnodes.length + + cpbranches.length + csbranches.length + + whole.consumed_len_bytes + + suffix.consumed_len_bytes + ); + if (canonical.length < clen) { + canonical = new Uint8Array(clen); + } + let ci = 0; + canonical[ci] = is_pure_suffixes_only_node ? 1 : 0; + ci += 1; + canonical[ci] = dlen; + ci += 1; + canonical.set(data, ci); + ci += dlen; + canonical[ci] = input[coffset]; + ci += 1; + if (!is_pure_suffixes_only_node) { + canonical[ci] = input[coffset + 1]; + ci += 1; + } + canonical.set(cpnodes, ci); + ci += cpnodes.length; + canonical.set(csnodes, ci); + ci += csnodes.length; + canonical.set(cpbranches, ci); + ci += cpbranches.length; + canonical.set(csbranches, ci); + ci += csbranches.length; + const leavesOffset = i - whole.consumed_len_bytes - suffix.consumed_len_bytes; + for (let j = leavesOffset; j < i; j += 1) { + canonical[ci + j - leavesOffset] = input[j]; + } + siphashOfBytes(canonical.subarray(0, clen), 0, 0, 0, 0, hash); + hash[2] &= 0x7f; + } else { + // uncompressed node + const dlen = input [i + 1]; + i += 2; + if (dlen === 0) { + data = EMPTY_UINT8; + } else { + data = new Uint8Array(input.buffer, i + input.byteOffset, dlen); + } + i += dlen; + const { + consumed_len_bytes: branches_consumed_len_bytes, + branches, + might_have_prefix_branches, + } = makeBranchesFromBinaryData(input, i, compression_tag); + i += branches_consumed_len_bytes; + let whole; + let suffix; + if (is_pure_suffixes_only_node) { + whole = EMPTY_BITMAP; + suffix = input[i] === 0 ? + EMPTY_BITMAP1 : + new RoaringBitmap(input, i); + i += suffix.consumed_len_bytes; + } else if (input[i] === 0xff) { + whole = EMPTY_BITMAP; + suffix = EMPTY_BITMAP; + i += 1; + } else { + whole = input[i] === 0 ? + EMPTY_BITMAP1 : + new RoaringBitmap(input, i); + i += whole.consumed_len_bytes; + suffix = input[i] === 0 ? + EMPTY_BITMAP1 : + new RoaringBitmap(input, i); + i += suffix.consumed_len_bytes; + } + siphashOfBytes(new Uint8Array( + input.buffer, + start + input.byteOffset, + i - start, + ), 0, 0, 0, 0, hash); + hash[2] &= 0x7f; + tree = new SearchTree( + branches, + might_have_prefix_branches, + data, + whole, + suffix, + ); + } + hash_history.push({hash: truncatedHash.slice(), used: false}); + if (data.length !== 0) { + data_history.push(data); + } + const tree_branch_nodeids = tree.branches.nodeids; + const tree_branch_subtrees = tree.branches.subtrees; + let j = 0; + let lb = tree.branches.subtrees.length; + while (j < lb) { + // node id with a 1 in its most significant bit is inlined, and, so + // it won't be in the stash + if ((tree_branch_nodeids[j * 6] & 0x80) === 0) { + const subtree = stash.getWithOffsetKey(tree_branch_nodeids, j * 6); + if (subtree !== undefined) { + tree_branch_subtrees[j] = Promise.resolve(subtree); + } + } + j += 1; + } + const tree_mhp_branch_nodeids = tree.might_have_prefix_branches.nodeids; + const tree_mhp_branch_subtrees = tree.might_have_prefix_branches.subtrees; + j = 0; + lb = tree.might_have_prefix_branches.subtrees.length; + while (j < lb) { + // node id with a 1 in its most significant bit is inlined, and, so + // it won't be in the stash + if ((tree_mhp_branch_nodeids[j * 6] & 0x80) === 0) { + const subtree = stash.getWithOffsetKey(tree_mhp_branch_nodeids, j * 6); + if (subtree !== undefined) { + tree_mhp_branch_subtrees[j] = Promise.resolve(subtree); + } + } + j += 1; + } + if (i !== l) { + stash.set(truncatedHash, tree); + } + } + return [truncatedHash, tree]; + } + + return new Promise((resolve, reject) => { + registry.searchTreeRootCallback = (error, data) => { + if (data) { + resolve(data); + } else { + reject(error); + } + }; + hooks.loadRoot(callbacks); + }); +} + +if (typeof window !== "undefined") { + window.Stringdex = { + loadDatabase, + }; + window.RoaringBitmap = RoaringBitmap; + if (window.StringdexOnload) { + window.StringdexOnload.forEach(cb => cb(window.Stringdex)); + } +} else { + /** @type {stringdex.Stringdex} */ + // eslint-disable-next-line no-undef + module.exports.Stringdex = { + loadDatabase, + }; + /** @type {stringdex.RoaringBitmap} */ + // eslint-disable-next-line no-undef + module.exports.RoaringBitmap = RoaringBitmap; +} + +// eslint-disable-next-line max-len +// polyfill https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Uint8Array/fromBase64 +/** + * @type {function(string): Uint8Array} base64 + */ +//@ts-expect-error +const makeUint8ArrayFromBase64 = Uint8Array.fromBase64 ? Uint8Array.fromBase64 : (string => { + const bytes_as_string = atob(string); + const l = bytes_as_string.length; + const bytes = new Uint8Array(l); + for (let i = 0; i < l; ++i) { + bytes[i] = bytes_as_string.charCodeAt(i); + } + return bytes; +}); +/** + * @type {function(string): Uint8Array} base64 + */ +//@ts-expect-error +const makeUint8ArrayFromHex = Uint8Array.fromHex ? Uint8Array.fromHex : (string => { + /** @type {Object<string, number>} */ + const alpha = { + "0": 0, "1": 1, + "2": 2, "3": 3, + "4": 4, "5": 5, + "6": 6, "7": 7, + "8": 8, "9": 9, + "a": 10, "b": 11, + "A": 10, "B": 11, + "c": 12, "d": 13, + "C": 12, "D": 13, + "e": 14, "f": 15, + "E": 14, "F": 15, + }; + const l = string.length >> 1; + const bytes = new Uint8Array(l); + for (let i = 0; i < l; i += 1) { + const top = string[i << 1]; + const bottom = string[(i << 1) + 1]; + bytes[i] = (alpha[top] << 4) | alpha[bottom]; + } + return bytes; +}); + +/** + * @type {function(Uint8Array): string} base64 + */ +//@ts-expect-error +const makeHexFromUint8Array = Uint8Array.prototype.toHex ? (array => array.toHex()) : (array => { + /** @type {string[]} */ + const alpha = [ + "0", "1", + "2", "3", + "4", "5", + "6", "7", + "8", "9", + "a", "b", + "c", "d", + "e", "f", + ]; + const l = array.length; + const v = []; + for (let i = 0; i < l; ++i) { + v.push(alpha[array[i] >> 4]); + v.push(alpha[array[i] & 0xf]); + } + return v.join(""); +}); + +////////////// + +/** + * SipHash 1-3 + * @param {Uint8Array} input data to be hashed; all codepoints in string should be less than 256 + * @param {number} k0lo first word of key + * @param {number} k0hi second word of key + * @param {number} k1lo third word of key + * @param {number} k1hi fourth word of key + * @param {Uint8Array} output the data to write (clobber the first eight bytes) + */ +function siphashOfBytes(input, k0lo, k0hi, k1lo, k1hi, output) { + // hash state + // While siphash uses 64 bit state, js only has native support + // for 32 bit numbers. BigInt, unfortunately, doesn't count. + // It's too slow. + let v0lo = k0lo ^ 0x70736575; + let v0hi = k0hi ^ 0x736f6d65; + let v1lo = k1lo ^ 0x6e646f6d; + let v1hi = k1hi ^ 0x646f7261; + let v2lo = k0lo ^ 0x6e657261; + let v2hi = k0hi ^ 0x6c796765; + let v3lo = k1lo ^ 0x79746573; + let v3hi = k1hi ^ 0x74656462; + const inputLength = input.length; + let inputI = 0; + // main hash loop + const left = inputLength & 0x7; + let milo = 0; + let mihi = 0; + while (inputI < inputLength - left) { + u8ToU64le(inputI, inputI + 8); + v3lo ^= milo; + v3hi ^= mihi; + siphashCompress(); + v0lo ^= milo; + v0hi ^= mihi; + inputI += 8; + } + u8ToU64le(inputI, inputI + left); + // finish + const blo = milo; + const bhi = ((inputLength & 0xff) << 24) | mihi; + v3lo ^= blo; + v3hi ^= bhi; + siphashCompress(); + v0lo ^= blo; + v0hi ^= bhi; + v2lo ^= 0xff; + siphashCompress(); + siphashCompress(); + siphashCompress(); + output[7] = (v0lo ^ v1lo ^ v2lo ^ v3lo) & 0xff; + output[6] = (v0lo ^ v1lo ^ v2lo ^ v3lo) >>> 8; + output[5] = (v0lo ^ v1lo ^ v2lo ^ v3lo) >>> 16; + output[4] = (v0lo ^ v1lo ^ v2lo ^ v3lo) >>> 24; + output[3] = (v0hi ^ v1hi ^ v2hi ^ v3hi) & 0xff; + output[2] = (v0hi ^ v1hi ^ v2hi ^ v3hi) >>> 8; + output[1] = (v0hi ^ v1hi ^ v2hi ^ v3hi) >>> 16; + output[0] = (v0hi ^ v1hi ^ v2hi ^ v3hi) >>> 24; + /** + * Convert eight bytes to a single 64-bit number + * @param {number} offset + * @param {number} length + */ + function u8ToU64le(offset, length) { + const n0 = offset < length ? input[offset] & 0xff : 0; + const n1 = offset + 1 < length ? input[offset + 1] & 0xff : 0; + const n2 = offset + 2 < length ? input[offset + 2] & 0xff : 0; + const n3 = offset + 3 < length ? input[offset + 3] & 0xff : 0; + const n4 = offset + 4 < length ? input[offset + 4] & 0xff : 0; + const n5 = offset + 5 < length ? input[offset + 5] & 0xff : 0; + const n6 = offset + 6 < length ? input[offset + 6] & 0xff : 0; + const n7 = offset + 7 < length ? input[offset + 7] & 0xff : 0; + milo = n0 | (n1 << 8) | (n2 << 16) | (n3 << 24); + mihi = n4 | (n5 << 8) | (n6 << 16) | (n7 << 24); + } + function siphashCompress() { + // v0 += v1; + v0hi = (v0hi + v1hi + (((v0lo >>> 0) + (v1lo >>> 0) > 0xffffffff) ? 1 : 0)) | 0; + v0lo = (v0lo + v1lo) | 0; + // rotl(v1, 13) + let v1lo_ = v1lo; + let v1hi_ = v1hi; + v1lo = (v1lo_ << 13) | (v1hi_ >>> 19); + v1hi = (v1hi_ << 13) | (v1lo_ >>> 19); + // v1 ^= v0 + v1lo ^= v0lo; + v1hi ^= v0hi; + // rotl(v0, 32) + const v0lo_ = v0lo; + const v0hi_ = v0hi; + v0lo = v0hi_; + v0hi = v0lo_; + // v2 += v3 + v2hi = (v2hi + v3hi + (((v2lo >>> 0) + (v3lo >>> 0) > 0xffffffff) ? 1 : 0)) | 0; + v2lo = (v2lo + v3lo) | 0; + // rotl(v3, 16) + let v3lo_ = v3lo; + let v3hi_ = v3hi; + v3lo = (v3lo_ << 16) | (v3hi_ >>> 16); + v3hi = (v3hi_ << 16) | (v3lo_ >>> 16); + // v3 ^= v2 + v3lo ^= v2lo; + v3hi ^= v2hi; + // v0 += v3 + v0hi = (v0hi + v3hi + (((v0lo >>> 0) + (v3lo >>> 0) > 0xffffffff) ? 1 : 0)) | 0; + v0lo = (v0lo + v3lo) | 0; + // rotl(v3, 21) + v3lo_ = v3lo; + v3hi_ = v3hi; + v3lo = (v3lo_ << 21) | (v3hi_ >>> 11); + v3hi = (v3hi_ << 21) | (v3lo_ >>> 11); + // v3 ^= v0 + v3lo ^= v0lo; + v3hi ^= v0hi; + // v2 += v1 + v2hi = (v2hi + v1hi + (((v2lo >>> 0) + (v1lo >>> 0) > 0xffffffff) ? 1 : 0)) | 0; + v2lo = (v2lo + v1lo) | 0; + // rotl(v1, 17) + v1lo_ = v1lo; + v1hi_ = v1hi; + v1lo = (v1lo_ << 17) | (v1hi_ >>> 15); + v1hi = (v1hi_ << 17) | (v1lo_ >>> 15); + // v1 ^= v2 + v1lo ^= v2lo; + v1hi ^= v2hi; + // rotl(v2, 32) + const v2lo_ = v2lo; + const v2hi_ = v2hi; + v2lo = v2hi_; + v2hi = v2lo_; + } +} + +////////////// + + +// Parts of this code are based on Lucene, which is licensed under the +// Apache/2.0 license. +// More information found here: +// https://fossies.org/linux/lucene/lucene/core/src/java/org/apache/lucene/util/automaton/ +// LevenshteinAutomata.java +class ParametricDescription { + /** + * @param {number} w + * @param {number} n + * @param {Int32Array} minErrors + */ + constructor(w, n, minErrors) { + this.w = w; + this.n = n; + this.minErrors = minErrors; + } + /** + * @param {number} absState + * @returns {boolean} + */ + isAccept(absState) { + const state = Math.floor(absState / (this.w + 1)); + const offset = absState % (this.w + 1); + return this.w - offset + this.minErrors[state] <= this.n; + } + /** + * @param {number} absState + * @returns {number} + */ + getPosition(absState) { + return absState % (this.w + 1); + } + /** + * @param {Uint8Array} name + * @param {number} charCode + * @param {number} pos + * @param {number} end + * @returns {number} + */ + getVector(name, charCode, pos, end) { + let vector = 0; + for (let i = pos; i < end; i += 1) { + vector = vector << 1; + if (name[i] === charCode) { + vector |= 1; + } + } + return vector; + } + /** + * @param {Int32Array} data + * @param {number} index + * @param {number} bitsPerValue + * @returns {number} + */ + unpack(data, index, bitsPerValue) { + const bitLoc = (bitsPerValue * index); + const dataLoc = bitLoc >> 5; + const bitStart = bitLoc & 31; + if (bitStart + bitsPerValue <= 32) { + // not split + return ((data[dataLoc] >> bitStart) & this.MASKS[bitsPerValue - 1]); + } else { + // split + const part = 32 - bitStart; + return ~~(((data[dataLoc] >> bitStart) & this.MASKS[part - 1]) + + ((data[1 + dataLoc] & this.MASKS[bitsPerValue - part - 1]) << part)); + } + } +} +ParametricDescription.prototype.MASKS = new Int32Array([ + 0x1, 0x3, 0x7, 0xF, + 0x1F, 0x3F, 0x7F, 0xFF, + 0x1FF, 0x3F, 0x7FF, 0xFFF, + 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF, + 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, + 0x1FFFFF, 0x3FFFFF, 0x7FFFFF, 0xFFFFFF, + 0x1FFFFFF, 0x3FFFFFF, 0x7FFFFFF, 0xFFFFFFF, + 0x1FFFFFFF, 0x3FFFFFFF, 0x7FFFFFFF, 0xFFFFFFFF, +]); + +// The following code was generated with the moman/finenight pkg +// This package is available under the MIT License, see NOTICE.txt +// for more details. +// This class is auto-generated, Please do not modify it directly. +// You should modify the https://gitlab.com/notriddle/createAutomata.py instead. +// The following code was generated with the moman/finenight pkg +// This package is available under the MIT License, see NOTICE.txt +// for more details. +// This class is auto-generated, Please do not modify it directly. +// You should modify https://gitlab.com/notriddle/moman-rustdoc instead. + +class Lev2TParametricDescription extends ParametricDescription { + /** + * @param {number} absState + * @param {number} position + * @param {number} vector + * @returns {number} + */ + transition(absState, position, vector) { + let state = Math.floor(absState / (this.w + 1)); + let offset = absState % (this.w + 1); + + if (position === this.w) { + if (state < 3) { + const loc = Math.imul(vector, 3) + state; + offset += this.unpack(this.offsetIncrs0, loc, 1); + state = this.unpack(this.toStates0, loc, 2) - 1; + } + } else if (position === this.w - 1) { + if (state < 5) { + const loc = Math.imul(vector, 5) + state; + offset += this.unpack(this.offsetIncrs1, loc, 1); + state = this.unpack(this.toStates1, loc, 3) - 1; + } + } else if (position === this.w - 2) { + if (state < 13) { + const loc = Math.imul(vector, 13) + state; + offset += this.unpack(this.offsetIncrs2, loc, 2); + state = this.unpack(this.toStates2, loc, 4) - 1; + } + } else if (position === this.w - 3) { + if (state < 28) { + const loc = Math.imul(vector, 28) + state; + offset += this.unpack(this.offsetIncrs3, loc, 2); + state = this.unpack(this.toStates3, loc, 5) - 1; + } + } else if (position === this.w - 4) { + if (state < 45) { + const loc = Math.imul(vector, 45) + state; + offset += this.unpack(this.offsetIncrs4, loc, 3); + state = this.unpack(this.toStates4, loc, 6) - 1; + } + } else { + // eslint-disable-next-line no-lonely-if + if (state < 45) { + const loc = Math.imul(vector, 45) + state; + offset += this.unpack(this.offsetIncrs5, loc, 3); + state = this.unpack(this.toStates5, loc, 6) - 1; + } + } + + if (state === -1) { + // null state + return -1; + } else { + // translate back to abs + return Math.imul(state, this.w + 1) + offset; + } + } + + // state map + // 0 -> [(0, 0)] + // 1 -> [(0, 1)] + // 2 -> [(0, 2)] + // 3 -> [(0, 1), (1, 1)] + // 4 -> [(0, 2), (1, 2)] + // 5 -> [(0, 1), (1, 1), (2, 1)] + // 6 -> [(0, 2), (1, 2), (2, 2)] + // 7 -> [(0, 1), (2, 1)] + // 8 -> [(0, 1), (2, 2)] + // 9 -> [(0, 2), (2, 1)] + // 10 -> [(0, 2), (2, 2)] + // 11 -> [t(0, 1), (0, 1), (1, 1), (2, 1)] + // 12 -> [t(0, 2), (0, 2), (1, 2), (2, 2)] + // 13 -> [(0, 2), (1, 2), (2, 2), (3, 2)] + // 14 -> [(0, 1), (1, 1), (3, 2)] + // 15 -> [(0, 1), (2, 2), (3, 2)] + // 16 -> [(0, 1), (3, 2)] + // 17 -> [(0, 1), t(1, 2), (2, 2), (3, 2)] + // 18 -> [(0, 2), (1, 2), (3, 1)] + // 19 -> [(0, 2), (1, 2), (3, 2)] + // 20 -> [(0, 2), (1, 2), t(1, 2), (2, 2), (3, 2)] + // 21 -> [(0, 2), (2, 1), (3, 1)] + // 22 -> [(0, 2), (2, 2), (3, 2)] + // 23 -> [(0, 2), (3, 1)] + // 24 -> [(0, 2), (3, 2)] + // 25 -> [(0, 2), t(1, 2), (1, 2), (2, 2), (3, 2)] + // 26 -> [t(0, 2), (0, 2), (1, 2), (2, 2), (3, 2)] + // 27 -> [t(0, 2), (0, 2), (1, 2), (3, 1)] + // 28 -> [(0, 2), (1, 2), (2, 2), (3, 2), (4, 2)] + // 29 -> [(0, 2), (1, 2), (2, 2), (4, 2)] + // 30 -> [(0, 2), (1, 2), (2, 2), t(2, 2), (3, 2), (4, 2)] + // 31 -> [(0, 2), (1, 2), (3, 2), (4, 2)] + // 32 -> [(0, 2), (1, 2), (4, 2)] + // 33 -> [(0, 2), (1, 2), t(1, 2), (2, 2), (3, 2), (4, 2)] + // 34 -> [(0, 2), (1, 2), t(2, 2), (2, 2), (3, 2), (4, 2)] + // 35 -> [(0, 2), (2, 1), (4, 2)] + // 36 -> [(0, 2), (2, 2), (3, 2), (4, 2)] + // 37 -> [(0, 2), (2, 2), (4, 2)] + // 38 -> [(0, 2), (3, 2), (4, 2)] + // 39 -> [(0, 2), (4, 2)] + // 40 -> [(0, 2), t(1, 2), (1, 2), (2, 2), (3, 2), (4, 2)] + // 41 -> [(0, 2), t(2, 2), (2, 2), (3, 2), (4, 2)] + // 42 -> [t(0, 2), (0, 2), (1, 2), (2, 2), (3, 2), (4, 2)] + // 43 -> [t(0, 2), (0, 2), (1, 2), (2, 2), (4, 2)] + // 44 -> [t(0, 2), (0, 2), (1, 2), (2, 2), t(2, 2), (3, 2), (4, 2)] + + + /** @param {number} w - length of word being checked */ + constructor(w) { + super(w, 2, new Int32Array([ + 0,1,2,0,1,-1,0,-1,0,-1,0,-1,0,-1,-1,-1,-1,-1,-2,-1,-1,-2,-1,-2, + -1,-1,-1,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2, + ])); + } +} + +Lev2TParametricDescription.prototype.toStates0 = /*2 bits per value */ new Int32Array([ + 0xe, +]); +Lev2TParametricDescription.prototype.offsetIncrs0 = /*1 bits per value */ new Int32Array([ + 0x0, +]); + +Lev2TParametricDescription.prototype.toStates1 = /*3 bits per value */ new Int32Array([ + 0x1a688a2c, +]); +Lev2TParametricDescription.prototype.offsetIncrs1 = /*1 bits per value */ new Int32Array([ + 0x3e0, +]); + +Lev2TParametricDescription.prototype.toStates2 = /*4 bits per value */ new Int32Array([ + 0x70707054,0xdc07035,0x3dd3a3a,0x2323213a, + 0x15435223,0x22545432,0x5435, +]); +Lev2TParametricDescription.prototype.offsetIncrs2 = /*2 bits per value */ new Int32Array([ + 0x80000,0x55582088,0x55555555,0x55, +]); + +Lev2TParametricDescription.prototype.toStates3 = /*5 bits per value */ new Int32Array([ + 0x1c0380a4,0x700a570,0xca529c0,0x180a00, + 0xa80af180,0xc5498e60,0x5a546398,0x8c4300e8, + 0xac18c601,0xd8d43501,0x863500ad,0x51976d6a, + 0x8ca0180a,0xc3501ac2,0xb0c5be16,0x76dda8a5, + 0x18c4519,0xc41294a,0xe248d231,0x1086520c, + 0xce31ac42,0x13946358,0x2d0348c4,0x6732d494, + 0x1ad224a5,0xd635ad4b,0x520c4139,0xce24948, + 0x22110a52,0x58ce729d,0xc41394e3,0x941cc520, + 0x90e732d4,0x4729d224,0x39ce35ad, +]); +Lev2TParametricDescription.prototype.offsetIncrs3 = /*2 bits per value */ new Int32Array([ + 0x80000,0xc0c830,0x300f3c30,0x2200fcff, + 0xcaa00a08,0x3c2200a8,0xa8fea00a,0x55555555, + 0x55555555,0x55555555,0x55555555,0x55555555, + 0x55555555,0x55555555, +]); + +Lev2TParametricDescription.prototype.toStates4 = /*6 bits per value */ new Int32Array([ + 0x801c0144,0x1453803,0x14700038,0xc0005145, + 0x1401,0x14,0x140000,0x0, + 0x510000,0x6301f007,0x301f00d1,0xa186178, + 0xc20ca0c3,0xc20c30,0xc30030c,0xc00c00cd, + 0xf0c00c30,0x4c054014,0xc30944c3,0x55150c34, + 0x8300550,0x430c0143,0x50c31,0xc30850c, + 0xc3143000,0x50053c50,0x5130d301,0x850d30c2, + 0x30a08608,0xc214414,0x43142145,0x21450031, + 0x1400c314,0x4c143145,0x32832803,0x28014d6c, + 0xcd34a0c3,0x1c50c76,0x1c314014,0x430c30c3, + 0x1431,0xc300500,0xca00d303,0xd36d0e40, + 0x90b0e400,0xcb2abb2c,0x70c20ca1,0x2c32ca2c, + 0xcd2c70cb,0x31c00c00,0x34c2c32c,0x5583280, + 0x558309b7,0x6cd6ca14,0x430850c7,0x51c51401, + 0x1430c714,0xc3087,0x71451450,0xca00d30, + 0xc26dc156,0xb9071560,0x1cb2abb2,0xc70c2144, + 0xb1c51ca1,0x1421c70c,0xc51c00c3,0x30811c51, + 0x24324308,0xc51031c2,0x70820820,0x5c33830d, + 0xc33850c3,0x30c30c30,0xc30c31c,0x451450c3, + 0x20c20c20,0xda0920d,0x5145914f,0x36596114, + 0x51965865,0xd9643653,0x365a6590,0x51964364, + 0x43081505,0x920b2032,0x2c718b28,0xd7242249, + 0x35cb28b0,0x2cb3872c,0x972c30d7,0xb0c32cb2, + 0x4e1c75c,0xc80c90c2,0x62ca2482,0x4504171c, + 0xd65d9610,0x33976585,0xd95cb5d,0x4b5ca5d7, + 0x73975c36,0x10308138,0xc2245105,0x41451031, + 0x14e24208,0xc35c3387,0x51453851,0x1c51c514, + 0xc70c30c3,0x20451450,0x14f1440c,0x4f0da092, + 0x4513d41,0x6533944d,0x1350e658,0xe1545055, + 0x64365a50,0x5519383,0x51030815,0x28920718, + 0x441c718b,0x714e2422,0x1c35cb28,0x4e1c7387, + 0xb28e1c51,0x5c70c32c,0xc204e1c7,0x81c61440, + 0x1c62ca24,0xd04503ce,0x85d63944,0x39338e65, + 0x8e154387,0x364b5ca3,0x38739738, +]); +Lev2TParametricDescription.prototype.offsetIncrs4 = /*3 bits per value */ new Int32Array([ + 0x10000000,0xc00000,0x60061,0x400, + 0x0,0x80010008,0x249248a4,0x8229048, + 0x2092,0x6c3603,0xb61b6c30,0x6db6036d, + 0xdb6c0,0x361b0180,0x91b72000,0xdb11b71b, + 0x6db6236,0x1008200,0x12480012,0x24924906, + 0x48200049,0x80410002,0x24000900,0x4924a489, + 0x10822492,0x20800125,0x48360,0x9241b692, + 0x6da4924,0x40009268,0x241b010,0x291b4900, + 0x6d249249,0x49493423,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x2492, +]); + +Lev2TParametricDescription.prototype.toStates5 = /*6 bits per value */ new Int32Array([ + 0x801c0144,0x1453803,0x14700038,0xc0005145, + 0x1401,0x14,0x140000,0x0, + 0x510000,0x4e00e007,0xe0051,0x3451451c, + 0xd015000,0x30cd0000,0xc30c30c,0xc30c30d4, + 0x40c30c30,0x7c01c014,0xc03458c0,0x185e0c07, + 0x2830c286,0x830c3083,0xc30030,0x33430c, + 0x30c3003,0x70051030,0x16301f00,0x8301f00d, + 0x30a18617,0xc20ca0c,0x431420c3,0xb1450c51, + 0x14314315,0x4f143145,0x34c05401,0x4c30944c, + 0x55150c3,0x30830055,0x1430c014,0xc00050c3, + 0xc30850,0xc314300,0x150053c5,0x25130d30, + 0x5430d30c,0xc0354154,0x300d0c90,0x1cb2cd0c, + 0xc91cb0c3,0x72c30cb2,0x14f1cb2c,0xc34c0540, + 0x34c30944,0x82182214,0x851050c2,0x50851430, + 0x1400c50c,0x30c5085,0x50c51450,0x150053c, + 0xc25130d3,0x8850d30,0x1430a086,0x450c2144, + 0x51cb1c21,0x1c91c70c,0xc71c314b,0x34c1cb1, + 0x6c328328,0xc328014d,0x76cd34a0,0x1401c50c, + 0xc31c3140,0x31430c30,0x14,0x30c3005, + 0xa0ca00d3,0x535b0c,0x4d2830ca,0x514369b3, + 0xc500d01,0x5965965a,0x30d46546,0x6435030c, + 0x8034c659,0xdb439032,0x2c390034,0xcaaecb24, + 0x30832872,0xcb28b1c,0x4b1c32cb,0x70030033, + 0x30b0cb0c,0xe40ca00d,0x400d36d0,0xb2c90b0e, + 0xca1cb2ab,0xa2c70c20,0x6575d95c,0x4315b5ce, + 0x95c53831,0x28034c5d,0x9b705583,0xa1455830, + 0xc76cd6c,0x40143085,0x71451c51,0x871430c, + 0x450000c3,0xd3071451,0x1560ca00,0x560c26dc, + 0xb35b2851,0xc914369,0x1a14500d,0x46593945, + 0xcb2c939,0x94507503,0x328034c3,0x9b70558, + 0xe41c5583,0x72caaeca,0x1c308510,0xc7147287, + 0x50871c32,0x1470030c,0xd307147,0xc1560ca0, + 0x1560c26d,0xabb2b907,0x21441cb2,0x38a1c70c, + 0x8e657394,0x314b1c93,0x39438738,0x43083081, + 0x31c22432,0x820c510,0x830d7082,0x50c35c33, + 0xc30c338,0xc31c30c3,0x50c30c30,0xc204514, + 0x890c90c2,0x31440c70,0xa8208208,0xea0df0c3, + 0x8a231430,0xa28a28a2,0x28a28a1e,0x1861868a, + 0x48308308,0xc3682483,0x14516453,0x4d965845, + 0xd4659619,0x36590d94,0xd969964,0x546590d9, + 0x20c20541,0x920d20c,0x5914f0da,0x96114514, + 0x65865365,0xe89d3519,0x99e7a279,0x9e89e89e, + 0x81821827,0xb2032430,0x18b28920,0x422492c7, + 0xb28b0d72,0x3872c35c,0xc30d72cb,0x32cb2972, + 0x1c75cb0c,0xc90c204e,0xa2482c80,0x24b1c62c, + 0xc3a89089,0xb0ea2e42,0x9669a31c,0xa4966a28, + 0x59a8a269,0x8175e7a,0xb203243,0x718b2892, + 0x4114105c,0x17597658,0x74ce5d96,0x5c36572d, + 0xd92d7297,0xe1ce5d70,0xc90c204,0xca2482c8, + 0x4171c62,0x5d961045,0x976585d6,0x79669533, + 0x964965a2,0x659689e6,0x308175e7,0x24510510, + 0x451031c2,0xe2420841,0x5c338714,0x453851c3, + 0x51c51451,0xc30c31c,0x451450c7,0x41440c20, + 0xc708914,0x82105144,0xf1c58c90,0x1470ea0d, + 0x61861863,0x8a1e85e8,0x8687a8a2,0x3081861, + 0x24853c51,0x5053c368,0x1341144f,0x96194ce5, + 0x1544d439,0x94385514,0xe0d90d96,0x5415464, + 0x4f1440c2,0xf0da0921,0x4513d414,0x533944d0, + 0x350e6586,0x86082181,0xe89e981d,0x18277689, + 0x10308182,0x89207185,0x41c718b2,0x14e24224, + 0xc35cb287,0xe1c73871,0x28e1c514,0xc70c32cb, + 0x204e1c75,0x1c61440c,0xc62ca248,0x90891071, + 0x2e41c58c,0xa31c70ea,0xe86175e7,0xa269a475, + 0x5e7a57a8,0x51030817,0x28920718,0xf38718b, + 0xe5134114,0x39961758,0xe1ce4ce,0x728e3855, + 0x5ce0d92d,0xc204e1ce,0x81c61440,0x1c62ca24, + 0xd04503ce,0x85d63944,0x75338e65,0x5d86075e, + 0x89e69647,0x75e76576, +]); +Lev2TParametricDescription.prototype.offsetIncrs5 = /*3 bits per value */ new Int32Array([ + 0x10000000,0xc00000,0x60061,0x400, + 0x0,0x60000008,0x6b003080,0xdb6ab6db, + 0x2db6,0x800400,0x49245240,0x11482412, + 0x104904,0x40020000,0x92292000,0xa4b25924, + 0x9649658,0xd80c000,0xdb0c001b,0x80db6d86, + 0x6db01b6d,0xc0600003,0x86000d86,0x6db6c36d, + 0xddadb6ed,0x300001b6,0x6c360,0xe37236e4, + 0x46db6236,0xdb6c,0x361b018,0xb91b7200, + 0x6dbb1b71,0x6db763,0x20100820,0x61248001, + 0x92492490,0x24820004,0x8041000,0x92400090, + 0x24924830,0x555b6a49,0x2080012,0x20004804, + 0x49252449,0x84112492,0x4000928,0x240201, + 0x92922490,0x58924924,0x49456,0x120d8082, + 0x6da4800,0x69249249,0x249a01b,0x6c04100, + 0x6d240009,0x92492483,0x24d5adb4,0x60208001, + 0x92000483,0x24925236,0x6846da49,0x10400092, + 0x241b0,0x49291b49,0x636d2492,0x92494935, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924,0x49249249, + 0x92492492,0x24924924,0x49249249,0x92492492, + 0x24924924,0x49249249,0x92492492,0x24924924, + 0x49249249,0x92492492,0x24924924, +]); + +class Lev1TParametricDescription extends ParametricDescription { + /** + * @param {number} absState + * @param {number} position + * @param {number} vector + * @returns {number} + */ + transition(absState, position, vector) { + let state = Math.floor(absState / (this.w + 1)); + let offset = absState % (this.w + 1); + + if (position === this.w) { + if (state < 2) { + const loc = Math.imul(vector, 2) + state; + offset += this.unpack(this.offsetIncrs0, loc, 1); + state = this.unpack(this.toStates0, loc, 2) - 1; + } + } else if (position === this.w - 1) { + if (state < 3) { + const loc = Math.imul(vector, 3) + state; + offset += this.unpack(this.offsetIncrs1, loc, 1); + state = this.unpack(this.toStates1, loc, 2) - 1; + } + } else if (position === this.w - 2) { + if (state < 6) { + const loc = Math.imul(vector, 6) + state; + offset += this.unpack(this.offsetIncrs2, loc, 2); + state = this.unpack(this.toStates2, loc, 3) - 1; + } + } else { + // eslint-disable-next-line no-lonely-if + if (state < 6) { + const loc = Math.imul(vector, 6) + state; + offset += this.unpack(this.offsetIncrs3, loc, 2); + state = this.unpack(this.toStates3, loc, 3) - 1; + } + } + + if (state === -1) { + // null state + return -1; + } else { + // translate back to abs + return Math.imul(state, this.w + 1) + offset; + } + } + + // state map + // 0 -> [(0, 0)] + // 1 -> [(0, 1)] + // 2 -> [(0, 1), (1, 1)] + // 3 -> [(0, 1), (1, 1), (2, 1)] + // 4 -> [(0, 1), (2, 1)] + // 5 -> [t(0, 1), (0, 1), (1, 1), (2, 1)] + + + /** @param {number} w - length of word being checked */ + constructor(w) { + super(w, 1, new Int32Array([0,1,0,-1,-1,-1])); + } +} + +Lev1TParametricDescription.prototype.toStates0 = /*2 bits per value */ new Int32Array([ + 0x2, +]); +Lev1TParametricDescription.prototype.offsetIncrs0 = /*1 bits per value */ new Int32Array([ + 0x0, +]); + +Lev1TParametricDescription.prototype.toStates1 = /*2 bits per value */ new Int32Array([ + 0xa43, +]); +Lev1TParametricDescription.prototype.offsetIncrs1 = /*1 bits per value */ new Int32Array([ + 0x38, +]); + +Lev1TParametricDescription.prototype.toStates2 = /*3 bits per value */ new Int32Array([ + 0x12180003,0xb45a4914,0x69, +]); +Lev1TParametricDescription.prototype.offsetIncrs2 = /*2 bits per value */ new Int32Array([ + 0x558a0000,0x5555, +]); + +Lev1TParametricDescription.prototype.toStates3 = /*3 bits per value */ new Int32Array([ + 0x900c0003,0xa1904864,0x45a49169,0x5a6d196a, + 0x9634, +]); +Lev1TParametricDescription.prototype.offsetIncrs3 = /*2 bits per value */ new Int32Array([ + 0xa0fc0000,0x5555ba08,0x55555555, +]); diff --git a/src/librustdoc/html/static/js/tsconfig.json b/src/librustdoc/html/static/js/tsconfig.json index b81099bb9df..42993cb0f2a 100644 --- a/src/librustdoc/html/static/js/tsconfig.json +++ b/src/librustdoc/html/static/js/tsconfig.json @@ -10,6 +10,6 @@ "skipLibCheck": true }, "typeAcquisition": { - "include": ["./rustdoc.d.ts"] + "include": ["./rustdoc.d.ts", "./stringdex.d.ts"] } } diff --git a/src/librustdoc/html/static_files.rs b/src/librustdoc/html/static_files.rs index 45589a37069..e670c2f39e7 100644 --- a/src/librustdoc/html/static_files.rs +++ b/src/librustdoc/html/static_files.rs @@ -80,6 +80,7 @@ static_files! { normalize_css => "static/css/normalize.css", main_js => "static/js/main.js", search_js => "static/js/search.js", + stringdex_js => "static/js/stringdex.js", settings_js => "static/js/settings.js", src_script_js => "static/js/src-script.js", storage_js => "static/js/storage.js", diff --git a/src/librustdoc/html/templates/page.html b/src/librustdoc/html/templates/page.html index 398436e3fe1..1f8ec9f30c5 100644 --- a/src/librustdoc/html/templates/page.html +++ b/src/librustdoc/html/templates/page.html @@ -29,6 +29,7 @@ data-rustdoc-version="{{rustdoc_version}}" {#+ #} data-channel="{{rust_channel}}" {#+ #} data-search-js="{{files.search_js}}" {#+ #} + data-stringdex-js="{{files.stringdex_js}}" {#+ #} data-settings-js="{{files.settings_js}}" {#+ #} > {# #} <script src="{{static_root_path|safe}}{{files.storage_js}}"></script> @@ -72,18 +73,9 @@ <![endif]--> {{ layout.external_html.before_content|safe }} {% if page.css_class != "src" %} - <nav class="mobile-topbar"> {# #} - <button class="sidebar-menu-toggle" title="show sidebar"></button> - {% if !layout.logo.is_empty() || page.rust_logo %} - <a class="logo-container" href="{{page.root_path|safe}}{{display_krate_with_trailing_slash|safe}}index.html"> - {% if page.rust_logo %} - <img class="rust-logo" src="{{static_root_path|safe}}{{files.rust_logo_svg}}" alt=""> - {% else if !layout.logo.is_empty() %} - <img src="{{layout.logo}}" alt=""> - {% endif %} - </a> - {% endif %} - </nav> + <rustdoc-topbar> {# #} + <h2><a href="#">{{page.short_title}}</a></h2> {# #} + </rustdoc-topbar> {% endif %} <nav class="sidebar"> {% if page.css_class != "src" %} @@ -117,9 +109,6 @@ <div class="sidebar-resizer" title="Drag to resize sidebar"></div> {# #} <main> {% if page.css_class != "src" %}<div class="width-limiter">{% endif %} - {# defined in storage.js to avoid duplicating complex UI across every page #} - {# and because the search form only works if JS is enabled anyway #} - <rustdoc-search></rustdoc-search> {# #} <section id="main-content" class="content">{{ content|safe }}</section> {% if page.css_class != "src" %}</div>{% endif %} </main> diff --git a/src/librustdoc/html/templates/print_item.html b/src/librustdoc/html/templates/print_item.html index 62954dbb023..640fd3dfee4 100644 --- a/src/librustdoc/html/templates/print_item.html +++ b/src/librustdoc/html/templates/print_item.html @@ -12,8 +12,8 @@ <h1> {{typ}} <span{% if item_type != "mod" +%} class="{{item_type}}"{% endif %}> - {{name}} - </span> {# #} + {{name|wrapped|safe}} + </span> {# #} <button id="copy-path" title="Copy item path to clipboard"> {# #} Copy item path {# #} </button> {# #} diff --git a/src/tools/cargo b/src/tools/cargo -Subproject 840b83a10fb0e039a83f4d70ad032892c287570 +Subproject 71eb84f21aef43c07580c6aed6f806a6299f504 diff --git a/src/tools/compiletest/src/runtest.rs b/src/tools/compiletest/src/runtest.rs index be4663fffbe..821cb128647 100644 --- a/src/tools/compiletest/src/runtest.rs +++ b/src/tools/compiletest/src/runtest.rs @@ -2151,7 +2151,7 @@ impl<'test> TestCx<'test> { #[rustfmt::skip] let tidy_args = [ - "--new-blocklevel-tags", "rustdoc-search,rustdoc-toolbar", + "--new-blocklevel-tags", "rustdoc-search,rustdoc-toolbar,rustdoc-topbar", "--indent", "yes", "--indent-spaces", "2", "--wrap", "0", diff --git a/src/tools/html-checker/main.rs b/src/tools/html-checker/main.rs index 5cdc4d53ab5..d5335d9e72e 100644 --- a/src/tools/html-checker/main.rs +++ b/src/tools/html-checker/main.rs @@ -31,7 +31,7 @@ fn check_html_file(file: &Path) -> usize { .arg("--mute-id") // this option is useful in case we want to mute more warnings .arg("yes") .arg("--new-blocklevel-tags") - .arg("rustdoc-search,rustdoc-toolbar") // custom elements + .arg("rustdoc-search,rustdoc-toolbar,rustdoc-topbar") // custom elements .arg("--mute") .arg(&to_mute_s) .arg(file); diff --git a/src/tools/miri/Cargo.toml b/src/tools/miri/Cargo.toml index 91dadf78a2f..99111092d39 100644 --- a/src/tools/miri/Cargo.toml +++ b/src/tools/miri/Cargo.toml @@ -78,6 +78,7 @@ native-lib = ["dep:libffi", "dep:libloading", "dep:capstone", "dep:ipc-channel", [lints.rust.unexpected_cfgs] level = "warn" +check-cfg = ['cfg(bootstrap)'] # Be aware that this file is inside a workspace when used via the # submodule in the rustc repo. That means there are many cargo features diff --git a/src/tools/miri/rust-version b/src/tools/miri/rust-version index a399b5cd77e..59adc572eaa 100644 --- a/src/tools/miri/rust-version +++ b/src/tools/miri/rust-version @@ -1 +1 @@ -425a9c0a0e365c0b8c6cfd00c2ded83a73bed9a0 +f605b57042ffeb320d7ae44490113a827139b766 diff --git a/src/tools/miri/src/lib.rs b/src/tools/miri/src/lib.rs index 7d8538d24bb..5ed6d6b346c 100644 --- a/src/tools/miri/src/lib.rs +++ b/src/tools/miri/src/lib.rs @@ -1,3 +1,4 @@ +#![cfg_attr(bootstrap, feature(strict_overflow_ops))] #![feature(abort_unwind)] #![feature(cfg_select)] #![feature(rustc_private)] diff --git a/src/tools/miri/src/machine.rs b/src/tools/miri/src/machine.rs index c76f993b5c9..52c4bd142c6 100644 --- a/src/tools/miri/src/machine.rs +++ b/src/tools/miri/src/machine.rs @@ -1111,6 +1111,7 @@ impl<'tcx> Machine<'tcx> for MiriMachine<'tcx> { ) -> InterpResult<'tcx, Option<(&'tcx mir::Body<'tcx>, ty::Instance<'tcx>)>> { // For foreign items, try to see if we can emulate them. if ecx.tcx.is_foreign_item(instance.def_id()) { + let _trace = enter_trace_span!("emulate_foreign_item"); // An external function call that does not have a MIR body. We either find MIR elsewhere // or emulate its effect. // This will be Ok(None) if we're emulating the intrinsic entirely within Miri (no need @@ -1123,6 +1124,7 @@ impl<'tcx> Machine<'tcx> for MiriMachine<'tcx> { } // Otherwise, load the MIR. + let _trace = enter_trace_span!("load_mir"); interp_ok(Some((ecx.load_mir(instance.def, None)?, instance))) } diff --git a/src/tools/miri/src/shims/foreign_items.rs b/src/tools/miri/src/shims/foreign_items.rs index 1da97cae98c..a700644b95d 100644 --- a/src/tools/miri/src/shims/foreign_items.rs +++ b/src/tools/miri/src/shims/foreign_items.rs @@ -153,7 +153,7 @@ pub trait EvalContextExt<'tcx>: crate::MiriInterpCxExt<'tcx> { return interp_ok(()); } // Skip over items without an explicitly defined symbol name. - if !(attrs.export_name.is_some() + if !(attrs.symbol_name.is_some() || attrs.flags.contains(CodegenFnAttrFlags::NO_MANGLE) || attrs.flags.contains(CodegenFnAttrFlags::RUSTC_STD_INTERNAL_SYMBOL)) { diff --git a/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.rs b/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.rs new file mode 100644 index 00000000000..b91a41d7650 --- /dev/null +++ b/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.rs @@ -0,0 +1,34 @@ +//@revisions: stack tree +//@[tree]compile-flags: -Zmiri-tree-borrows +// Validation forces more things into memory, which we can't have here. +//@compile-flags: -Zmiri-disable-validation +#![feature(custom_mir, core_intrinsics)] +use std::intrinsics::mir::*; + +pub struct S(i32); + +#[custom_mir(dialect = "runtime", phase = "optimized")] +fn main() { + mir! { + let _unit: (); + { + let staging = S(42); // This forces `staging` into memory... + let non_copy = staging; // ... so we move it to a non-inmemory local here. + // This specifically uses a type with scalar representation to tempt Miri to use the + // efficient way of storing local variables (outside adressable memory). + Call(_unit = callee(Move(non_copy), Move(non_copy)), ReturnTo(after_call), UnwindContinue()) + //~[stack]^ ERROR: not granting access + //~[tree]| ERROR: /read access .* forbidden/ + } + after_call = { + Return() + } + } +} + +pub fn callee(x: S, mut y: S) { + // With the setup above, if `x` and `y` are both moved, + // then writing to `y` will change the value stored in `x`! + y.0 = 0; + assert_eq!(x.0, 42); +} diff --git a/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.stack.stderr b/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.stack.stderr new file mode 100644 index 00000000000..0c1100cae63 --- /dev/null +++ b/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.stack.stderr @@ -0,0 +1,25 @@ +error: Undefined Behavior: not granting access to tag <TAG> because that would remove [Unique for <TAG>] which is strongly protected + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | Call(_unit = callee(Move(non_copy), Move(non_copy)), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Undefined Behavior occurred here + | + = help: this indicates a potential bug in the program: it performed an invalid operation, but the Stacked Borrows rules it violated are still experimental + = help: see https://github.com/rust-lang/unsafe-code-guidelines/blob/master/wip/stacked-borrows.md for further information +help: <TAG> was created here, as the root tag for ALLOC + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | Call(_unit = callee(Move(non_copy), Move(non_copy)), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +help: <TAG> is this argument + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | y.0 = 0; + | ^^^^^^^ + = note: BACKTRACE (of the first span): + = note: inside `main` at tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + +note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace + +error: aborting due to 1 previous error + diff --git a/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.tree.stderr b/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.tree.stderr new file mode 100644 index 00000000000..2266a9c39f9 --- /dev/null +++ b/src/tools/miri/tests/fail/function_calls/arg_inplace_locals_alias.tree.stderr @@ -0,0 +1,34 @@ +error: Undefined Behavior: read access through <TAG> (root of the allocation) at ALLOC[0x0] is forbidden + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | Call(_unit = callee(Move(non_copy), Move(non_copy)), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Undefined Behavior occurred here + | + = help: this indicates a potential bug in the program: it performed an invalid operation, but the Tree Borrows rules it violated are still experimental + = help: see https://github.com/rust-lang/unsafe-code-guidelines/blob/master/wip/tree-borrows.md for further information + = help: the accessed tag <TAG> (root of the allocation) is foreign to the protected tag <TAG> (i.e., it is not a child) + = help: this foreign read access would cause the protected tag <TAG> (currently Active) to become Disabled + = help: protected tags must never be Disabled +help: the accessed tag <TAG> was created here + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | Call(_unit = callee(Move(non_copy), Move(non_copy)), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +help: the protected tag <TAG> was created here, in the initial state Reserved + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | y.0 = 0; + | ^^^^^^^ +help: the protected tag <TAG> later transitioned to Active due to a child write access at offsets [0x0..0x4] + --> tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + | +LL | y.0 = 0; + | ^^^^^^^ + = help: this transition corresponds to the first write to a 2-phase borrowed mutable reference + = note: BACKTRACE (of the first span): + = note: inside `main` at tests/fail/function_calls/arg_inplace_locals_alias.rs:LL:CC + +note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace + +error: aborting due to 1 previous error + diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.none.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.none.stderr index 24091547258..d478568ceae 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.none.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.none.stderr @@ -11,8 +11,8 @@ LL | unsafe { ptr.read() }; note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_read.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Uninitialized memory occurred at ALLOC[0x0..0x4], in this allocation: ALLOC (stack variable, size: 4, align: 4) { diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.rs b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.rs index a6e0134bd17..dc22e129e18 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.rs +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.rs @@ -10,11 +10,11 @@ use std::intrinsics::mir::*; pub fn main() { mir! { { - let x = 0; - let ptr = &raw mut x; + let _x = 0; + let ptr = &raw mut _x; // We arrange for `myfun` to have a pointer that aliases // its return place. Even just reading from that pointer is UB. - Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) } after_call = { @@ -25,7 +25,7 @@ pub fn main() { fn myfun(ptr: *mut i32) -> i32 { unsafe { ptr.read() }; - //~[stack]^ ERROR: not granting access + //~[stack]^ ERROR: does not exist in the borrow stack //~[tree]| ERROR: /read access .* forbidden/ //~[none]| ERROR: uninitialized // Without an aliasing model, reads are "fine" but at least they return uninit data. diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.stack.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.stack.stderr index 77cc0332a1c..86adbab353b 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.stack.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.stack.stderr @@ -1,8 +1,8 @@ -error: Undefined Behavior: not granting access to tag <TAG> because that would remove [Unique for <TAG>] which is strongly protected +error: Undefined Behavior: attempting a read access using <TAG> at ALLOC[0x0], but that tag does not exist in the borrow stack for this location --> tests/fail/function_calls/return_pointer_aliasing_read.rs:LL:CC | LL | unsafe { ptr.read() }; - | ^^^^^^^^^^ Undefined Behavior occurred here + | ^^^^^^^^^^ this error occurs as part of an access at ALLOC[0x0..0x4] | = help: this indicates a potential bug in the program: it performed an invalid operation, but the Stacked Borrows rules it violated are still experimental = help: see https://github.com/rust-lang/unsafe-code-guidelines/blob/master/wip/stacked-borrows.md for further information @@ -11,12 +11,12 @@ help: <TAG> was created by a SharedReadWrite retag at offsets [0x0..0x4] | LL | / mir! { LL | | { -LL | | let x = 0; -LL | | let ptr = &raw mut x; +LL | | let _x = 0; +LL | | let ptr = &raw mut _x; ... | LL | | } | |_____^ -help: <TAG> is this argument +help: <TAG> was later invalidated at offsets [0x0..0x4] by a Unique in-place function argument/return passing protection --> tests/fail/function_calls/return_pointer_aliasing_read.rs:LL:CC | LL | unsafe { ptr.read() }; @@ -26,8 +26,8 @@ LL | unsafe { ptr.read() }; note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_read.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ = note: this error originates in the macro `::core::intrinsics::mir::__internal_remove_let` which comes from the expansion of the macro `mir` (in Nightly builds, run with -Z macro-backtrace for more info) note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.tree.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.tree.stderr index 33aad6ee17e..b1aa2ba2886 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.tree.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_read.tree.stderr @@ -14,8 +14,8 @@ help: the accessed tag <TAG> was created here | LL | / mir! { LL | | { -LL | | let x = 0; -LL | | let ptr = &raw mut x; +LL | | let _x = 0; +LL | | let ptr = &raw mut _x; ... | LL | | } | |_____^ @@ -35,8 +35,8 @@ LL | unsafe { ptr.read() }; note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_read.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ = note: this error originates in the macro `::core::intrinsics::mir::__internal_remove_let` which comes from the expansion of the macro `mir` (in Nightly builds, run with -Z macro-backtrace for more info) note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.rs b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.rs index 6155e925c4b..2fddaf37235 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.rs +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.rs @@ -14,7 +14,7 @@ pub fn main() { let ptr = &raw mut _x; // We arrange for `myfun` to have a pointer that aliases // its return place. Writing to that pointer is UB. - Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) } after_call = { @@ -26,7 +26,7 @@ pub fn main() { fn myfun(ptr: *mut i32) -> i32 { // This overwrites the return place, which shouldn't be possible through another pointer. unsafe { ptr.write(0) }; - //~[stack]^ ERROR: strongly protected + //~[stack]^ ERROR: does not exist in the borrow stack //~[tree]| ERROR: /write access .* forbidden/ 13 } diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.stack.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.stack.stderr index 828b2339f8c..faae6172d75 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.stack.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.stack.stderr @@ -1,8 +1,8 @@ -error: Undefined Behavior: not granting access to tag <TAG> because that would remove [Unique for <TAG>] which is strongly protected +error: Undefined Behavior: attempting a write access using <TAG> at ALLOC[0x0], but that tag does not exist in the borrow stack for this location --> tests/fail/function_calls/return_pointer_aliasing_write.rs:LL:CC | LL | unsafe { ptr.write(0) }; - | ^^^^^^^^^^^^ Undefined Behavior occurred here + | ^^^^^^^^^^^^ this error occurs as part of an access at ALLOC[0x0..0x4] | = help: this indicates a potential bug in the program: it performed an invalid operation, but the Stacked Borrows rules it violated are still experimental = help: see https://github.com/rust-lang/unsafe-code-guidelines/blob/master/wip/stacked-borrows.md for further information @@ -16,7 +16,7 @@ LL | | let ptr = &raw mut _x; ... | LL | | } | |_____^ -help: <TAG> is this argument +help: <TAG> was later invalidated at offsets [0x0..0x4] by a Unique in-place function argument/return passing protection --> tests/fail/function_calls/return_pointer_aliasing_write.rs:LL:CC | LL | unsafe { ptr.write(0) }; @@ -26,8 +26,8 @@ LL | unsafe { ptr.write(0) }; note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_write.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ = note: this error originates in the macro `::core::intrinsics::mir::__internal_remove_let` which comes from the expansion of the macro `mir` (in Nightly builds, run with -Z macro-backtrace for more info) note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.tree.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.tree.stderr index 5cb36d99846..0cf449ea3ec 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.tree.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write.tree.stderr @@ -35,8 +35,8 @@ LL | unsafe { ptr.write(0) }; note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_write.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ = note: this error originates in the macro `::core::intrinsics::mir::__internal_remove_let` which comes from the expansion of the macro `mir` (in Nightly builds, run with -Z macro-backtrace for more info) note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs index 37ee7ae1b0d..5f3ecb65022 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs @@ -16,7 +16,7 @@ pub fn main() { let ptr = &raw mut _x; // We arrange for `myfun` to have a pointer that aliases // its return place. Writing to that pointer is UB. - Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) } after_call = { @@ -32,7 +32,7 @@ fn myfun(ptr: *mut i32) -> i32 { fn myfun2(ptr: *mut i32) -> i32 { // This overwrites the return place, which shouldn't be possible through another pointer. unsafe { ptr.write(0) }; - //~[stack]^ ERROR: strongly protected + //~[stack]^ ERROR: does not exist in the borrow stack //~[tree]| ERROR: /write access .* forbidden/ 13 } diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.stack.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.stack.stderr index f5183cfaf5b..1a18857bb17 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.stack.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.stack.stderr @@ -1,8 +1,8 @@ -error: Undefined Behavior: not granting access to tag <TAG> because that would remove [Unique for <TAG>] which is strongly protected +error: Undefined Behavior: attempting a write access using <TAG> at ALLOC[0x0], but that tag does not exist in the borrow stack for this location --> tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs:LL:CC | LL | unsafe { ptr.write(0) }; - | ^^^^^^^^^^^^ Undefined Behavior occurred here + | ^^^^^^^^^^^^ this error occurs as part of an access at ALLOC[0x0..0x4] | = help: this indicates a potential bug in the program: it performed an invalid operation, but the Stacked Borrows rules it violated are still experimental = help: see https://github.com/rust-lang/unsafe-code-guidelines/blob/master/wip/stacked-borrows.md for further information @@ -16,18 +16,18 @@ LL | | let ptr = &raw mut _x; ... | LL | | } | |_____^ -help: <TAG> is this argument +help: <TAG> was later invalidated at offsets [0x0..0x4] by a Unique in-place function argument/return passing protection --> tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs:LL:CC | -LL | unsafe { ptr.write(0) }; - | ^^^^^^^^^^^^^^^^^^^^^^^ +LL | become myfun2(ptr) + | ^^^^^^^^^^^^^^^^^^ = note: BACKTRACE (of the first span): = note: inside `myfun2` at tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs:LL:CC note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ = note: this error originates in the macro `::core::intrinsics::mir::__internal_remove_let` which comes from the expansion of the macro `mir` (in Nightly builds, run with -Z macro-backtrace for more info) note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace diff --git a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.tree.stderr b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.tree.stderr index 8c54017cba3..a006c6feae4 100644 --- a/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.tree.stderr +++ b/src/tools/miri/tests/fail/function_calls/return_pointer_aliasing_write_tail_call.tree.stderr @@ -35,8 +35,8 @@ LL | unsafe { ptr.write(0) }; note: inside `main` --> tests/fail/function_calls/return_pointer_aliasing_write_tail_call.rs:LL:CC | -LL | Call(*ptr = myfun(ptr), ReturnTo(after_call), UnwindContinue()) - | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +LL | Call(_x = myfun(ptr), ReturnTo(after_call), UnwindContinue()) + | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ = note: this error originates in the macro `::core::intrinsics::mir::__internal_remove_let` which comes from the expansion of the macro `mir` (in Nightly builds, run with -Z macro-backtrace for more info) note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace diff --git a/src/tools/nix-dev-shell/shell.nix b/src/tools/nix-dev-shell/shell.nix index 0adbacf7e8d..ad33b121f97 100644 --- a/src/tools/nix-dev-shell/shell.nix +++ b/src/tools/nix-dev-shell/shell.nix @@ -14,6 +14,7 @@ pkgs.mkShell { packages = [ pkgs.git pkgs.nix + pkgs.glibc.static x # Get the runtime deps of the x wrapper ] ++ lists.flatten (attrsets.attrValues env); diff --git a/src/tools/rustdoc-js/tester.js b/src/tools/rustdoc-js/tester.js index 0baa179e16b..a1e632ce743 100644 --- a/src/tools/rustdoc-js/tester.js +++ b/src/tools/rustdoc-js/tester.js @@ -1,7 +1,8 @@ /* global globalThis */ + const fs = require("fs"); const path = require("path"); - +const { isGeneratorObject } = require("util/types"); function arrayToCode(array) { return array.map((value, index) => { @@ -45,23 +46,16 @@ function shouldIgnoreField(fieldName) { } function valueMapper(key, testOutput) { - const isAlias = testOutput["is_alias"]; let value = testOutput[key]; // To make our life easier, if there is a "parent" type, we add it to the path. if (key === "path") { - if (testOutput["parent"] !== undefined) { + if (testOutput["parent"]) { if (value.length > 0) { value += "::" + testOutput["parent"]["name"]; } else { value = testOutput["parent"]["name"]; } - } else if (testOutput["is_alias"]) { - value = valueMapper(key, testOutput["original"]); } - } else if (isAlias && key === "alias") { - value = testOutput["name"]; - } else if (isAlias && ["name"].includes(key)) { - value = testOutput["original"][key]; } return value; } @@ -237,7 +231,7 @@ async function runSearch(query, expected, doSearch, loadedFile, queryName) { const ignore_order = loadedFile.ignore_order; const exact_check = loadedFile.exact_check; - const results = await doSearch(query, loadedFile.FILTER_CRATE); + const { resultsTable } = await doSearch(query, loadedFile.FILTER_CRATE); const error_text = []; for (const key in expected) { @@ -247,37 +241,38 @@ async function runSearch(query, expected, doSearch, loadedFile, queryName) { if (!Object.prototype.hasOwnProperty.call(expected, key)) { continue; } - if (!Object.prototype.hasOwnProperty.call(results, key)) { + if (!Object.prototype.hasOwnProperty.call(resultsTable, key)) { error_text.push("==> Unknown key \"" + key + "\""); break; } const entry = expected[key]; - if (exact_check && entry.length !== results[key].length) { + if (exact_check && entry.length !== resultsTable[key].length) { error_text.push(queryName + "==> Expected exactly " + entry.length + - " results but found " + results[key].length + " in '" + key + "'"); + " results but found " + resultsTable[key].length + " in '" + key + "'"); } let prev_pos = -1; for (const [index, elem] of entry.entries()) { - const entry_pos = lookForEntry(elem, results[key]); + const entry_pos = lookForEntry(elem, resultsTable[key]); if (entry_pos === -1) { error_text.push(queryName + "==> Result not found in '" + key + "': '" + JSON.stringify(elem) + "'"); // By default, we just compare the two first items. let item_to_diff = 0; - if ((!ignore_order || exact_check) && index < results[key].length) { + if ((!ignore_order || exact_check) && index < resultsTable[key].length) { item_to_diff = index; } error_text.push("Diff of first error:\n" + - betterLookingDiff(elem, results[key][item_to_diff])); + betterLookingDiff(elem, resultsTable[key][item_to_diff])); } else if (exact_check === true && prev_pos + 1 !== entry_pos) { error_text.push(queryName + "==> Exact check failed at position " + (prev_pos + 1) + ": expected '" + JSON.stringify(elem) + "' but found '" + - JSON.stringify(results[key][index]) + "'"); + JSON.stringify(resultsTable[key][index]) + "'"); } else if (ignore_order === false && entry_pos < prev_pos) { - error_text.push(queryName + "==> '" + JSON.stringify(elem) + "' was supposed " + - "to be before '" + JSON.stringify(results[key][prev_pos]) + "'"); + error_text.push(queryName + "==> '" + + JSON.stringify(elem) + "' was supposed to be before '" + + JSON.stringify(resultsTable[key][prev_pos]) + "'"); } else { prev_pos = entry_pos; } @@ -286,19 +281,20 @@ async function runSearch(query, expected, doSearch, loadedFile, queryName) { return error_text; } -async function runCorrections(query, corrections, getCorrections, loadedFile) { - const qc = await getCorrections(query, loadedFile.FILTER_CRATE); +async function runCorrections(query, corrections, doSearch, loadedFile) { + const { parsedQuery } = await doSearch(query, loadedFile.FILTER_CRATE); + const qc = parsedQuery.correction; const error_text = []; if (corrections === null) { if (qc !== null) { - error_text.push(`==> expected = null, found = ${qc}`); + error_text.push(`==> [correction] expected = null, found = ${qc}`); } return error_text; } - if (qc !== corrections.toLowerCase()) { - error_text.push(`==> expected = ${corrections}, found = ${qc}`); + if (qc.toLowerCase() !== corrections.toLowerCase()) { + error_text.push(`==> [correction] expected = ${corrections}, found = ${qc}`); } return error_text; @@ -320,7 +316,7 @@ function checkResult(error_text, loadedFile, displaySuccess) { return 1; } -async function runCheckInner(callback, loadedFile, entry, getCorrections, extra) { +async function runCheckInner(callback, loadedFile, entry, extra, doSearch) { if (typeof entry.query !== "string") { console.log("FAILED"); console.log("==> Missing `query` field"); @@ -338,7 +334,7 @@ async function runCheckInner(callback, loadedFile, entry, getCorrections, extra) error_text = await runCorrections( entry.query, entry.correction, - getCorrections, + doSearch, loadedFile, ); if (checkResult(error_text, loadedFile, false) !== 0) { @@ -348,16 +344,16 @@ async function runCheckInner(callback, loadedFile, entry, getCorrections, extra) return true; } -async function runCheck(loadedFile, key, getCorrections, callback) { +async function runCheck(loadedFile, key, doSearch, callback) { const expected = loadedFile[key]; if (Array.isArray(expected)) { for (const entry of expected) { - if (!await runCheckInner(callback, loadedFile, entry, getCorrections, true)) { + if (!await runCheckInner(callback, loadedFile, entry, true, doSearch)) { return 1; } } - } else if (!await runCheckInner(callback, loadedFile, expected, getCorrections, false)) { + } else if (!await runCheckInner(callback, loadedFile, expected, false, doSearch)) { return 1; } console.log("OK"); @@ -368,7 +364,7 @@ function hasCheck(content, checkName) { return content.startsWith(`const ${checkName}`) || content.includes(`\nconst ${checkName}`); } -async function runChecks(testFile, doSearch, parseQuery, getCorrections) { +async function runChecks(testFile, doSearch, parseQuery) { let checkExpected = false; let checkParsed = false; let testFileContent = readFile(testFile); @@ -397,12 +393,12 @@ async function runChecks(testFile, doSearch, parseQuery, getCorrections) { let res = 0; if (checkExpected) { - res += await runCheck(loadedFile, "EXPECTED", getCorrections, (query, expected, text) => { + res += await runCheck(loadedFile, "EXPECTED", doSearch, (query, expected, text) => { return runSearch(query, expected, doSearch, loadedFile, text); }); } if (checkParsed) { - res += await runCheck(loadedFile, "PARSED", getCorrections, (query, expected, text) => { + res += await runCheck(loadedFile, "PARSED", doSearch, (query, expected, text) => { return runParser(query, expected, parseQuery, text); }); } @@ -416,71 +412,89 @@ async function runChecks(testFile, doSearch, parseQuery, getCorrections) { * @param {string} resource_suffix - Version number between filename and .js, e.g. "1.59.0" * @returns {Object} - Object containing keys: `doSearch`, which runs a search * with the loaded index and returns a table of results; `parseQuery`, which is the - * `parseQuery` function exported from the search module; and `getCorrections`, which runs + * `parseQuery` function exported from the search module, which runs * a search but returns type name corrections instead of results. */ -function loadSearchJS(doc_folder, resource_suffix) { - const searchIndexJs = path.join(doc_folder, "search-index" + resource_suffix + ".js"); - const searchIndex = require(searchIndexJs); - - globalThis.searchState = { - descShards: new Map(), - loadDesc: async function({descShard, descIndex}) { - if (descShard.promise === null) { - descShard.promise = new Promise((resolve, reject) => { - descShard.resolve = resolve; - const ds = descShard; - const fname = `${ds.crate}-desc-${ds.shard}-${resource_suffix}.js`; - fs.readFile( - `${doc_folder}/search.desc/${descShard.crate}/${fname}`, - (err, data) => { - if (err) { - reject(err); - } else { - eval(data.toString("utf8")); - } - }, - ); - }); - } - const list = await descShard.promise; - return list[descIndex]; - }, - loadedDescShard: function(crate, shard, data) { - this.descShards.get(crate)[shard].resolve(data.split("\n")); - }, - }; - +async function loadSearchJS(doc_folder, resource_suffix) { const staticFiles = path.join(doc_folder, "static.files"); + const stringdexJs = fs.readdirSync(staticFiles).find(f => f.match(/stringdex.*\.js$/)); + const stringdexModule = require(path.join(staticFiles, stringdexJs)); const searchJs = fs.readdirSync(staticFiles).find(f => f.match(/search.*\.js$/)); const searchModule = require(path.join(staticFiles, searchJs)); - searchModule.initSearch(searchIndex.searchIndex); - const docSearch = searchModule.docSearch; + globalThis.nonnull = (x, msg) => { + if (x === null) { + throw (msg || "unexpected null value!"); + } else { + return x; + } + }; + const { docSearch, DocSearch } = await searchModule.initSearch( + stringdexModule.Stringdex, + stringdexModule.RoaringBitmap, + { + loadRoot: callbacks => { + for (const key in callbacks) { + if (Object.hasOwn(callbacks, key)) { + globalThis[key] = callbacks[key]; + } + } + const rootJs = readFile(path.join(doc_folder, "search.index/root" + + resource_suffix + ".js")); + eval(rootJs); + }, + loadTreeByHash: hashHex => { + const shardJs = readFile(path.join(doc_folder, "search.index/" + hashHex + ".js")); + eval(shardJs); + }, + loadDataByNameAndHash: (name, hashHex) => { + const shardJs = readFile(path.join(doc_folder, "search.index/" + name + "/" + + hashHex + ".js")); + eval(shardJs); + }, + }, + ); return { doSearch: async function(queryStr, filterCrate, currentCrate) { - const result = await docSearch.execQuery(searchModule.parseQuery(queryStr), - filterCrate, currentCrate); + const parsedQuery = DocSearch.parseQuery(queryStr); + const result = await docSearch.execQuery(parsedQuery, filterCrate, currentCrate); + const resultsTable = {}; for (const tab in result) { if (!Object.prototype.hasOwnProperty.call(result, tab)) { continue; } - if (!(result[tab] instanceof Array)) { + if (!isGeneratorObject(result[tab])) { continue; } - for (const entry of result[tab]) { + resultsTable[tab] = []; + for await (const entry of result[tab]) { + const flatEntry = Object.assign({ + crate: entry.item.crate, + name: entry.item.name, + path: entry.item.modulePath, + exactPath: entry.item.exactModulePath, + ty: entry.item.ty, + }, entry); for (const key in entry) { if (!Object.prototype.hasOwnProperty.call(entry, key)) { continue; } - if (key === "displayTypeSignature" && entry.displayTypeSignature !== null) { - const {type, mappedNames, whereClause} = - await entry.displayTypeSignature; - entry.displayType = arrayToCode(type); - entry.displayMappedNames = [...mappedNames.entries()] + if (key === "desc" && entry.desc !== null) { + flatEntry.desc = await entry.desc; + } else if (key === "displayTypeSignature" && + entry.displayTypeSignature !== null + ) { + flatEntry.displayTypeSignature = await entry.displayTypeSignature; + const { + type, + mappedNames, + whereClause, + } = flatEntry.displayTypeSignature; + flatEntry.displayType = arrayToCode(type); + flatEntry.displayMappedNames = [...mappedNames.entries()] .map(([name, qname]) => { return `${name} = ${qname}`; }).join(", "); - entry.displayWhereClause = [...whereClause.entries()] + flatEntry.displayWhereClause = [...whereClause.entries()] .flatMap(([name, value]) => { if (value.length === 0) { return []; @@ -489,16 +503,12 @@ function loadSearchJS(doc_folder, resource_suffix) { }).join(", "); } } + resultsTable[tab].push(flatEntry); } } - return result; + return { resultsTable, parsedQuery }; }, - getCorrections: function(queryStr, filterCrate, currentCrate) { - const parsedQuery = searchModule.parseQuery(queryStr); - docSearch.execQuery(parsedQuery, filterCrate, currentCrate); - return parsedQuery.correction; - }, - parseQuery: searchModule.parseQuery, + parseQuery: DocSearch.parseQuery, }; } @@ -570,7 +580,7 @@ async function main(argv) { return 1; } - const parseAndSearch = loadSearchJS( + const parseAndSearch = await loadSearchJS( opts["doc_folder"], opts["resource_suffix"], ); @@ -579,14 +589,11 @@ async function main(argv) { const doSearch = function(queryStr, filterCrate) { return parseAndSearch.doSearch(queryStr, filterCrate, opts["crate_name"]); }; - const getCorrections = function(queryStr, filterCrate) { - return parseAndSearch.getCorrections(queryStr, filterCrate, opts["crate_name"]); - }; if (opts["test_file"].length !== 0) { for (const file of opts["test_file"]) { process.stdout.write(`Testing ${file} ... `); - errors += await runChecks(file, doSearch, parseAndSearch.parseQuery, getCorrections); + errors += await runChecks(file, doSearch, parseAndSearch.parseQuery); } } else if (opts["test_folder"].length !== 0) { for (const file of fs.readdirSync(opts["test_folder"])) { @@ -595,7 +602,7 @@ async function main(argv) { } process.stdout.write(`Testing ${file} ... `); errors += await runChecks(path.join(opts["test_folder"], file), doSearch, - parseAndSearch.parseQuery, getCorrections); + parseAndSearch.parseQuery); } } return errors > 0 ? 1 : 0; diff --git a/src/tools/tidy/src/extra_checks/mod.rs b/src/tools/tidy/src/extra_checks/mod.rs index f90f716cd95..31169ec5967 100644 --- a/src/tools/tidy/src/extra_checks/mod.rs +++ b/src/tools/tidy/src/extra_checks/mod.rs @@ -41,7 +41,6 @@ const RUFF_CONFIG_PATH: &[&str] = &["src", "tools", "tidy", "config", "ruff.toml const RUFF_CACHE_PATH: &[&str] = &["cache", "ruff_cache"]; const PIP_REQ_PATH: &[&str] = &["src", "tools", "tidy", "config", "requirements.txt"]; -// this must be kept in sync with with .github/workflows/spellcheck.yml const SPELLCHECK_DIRS: &[&str] = &["compiler", "library", "src/bootstrap", "src/librustdoc"]; pub fn check( @@ -51,6 +50,7 @@ pub fn check( librustdoc_path: &Path, tools_path: &Path, npm: &Path, + cargo: &Path, bless: bool, extra_checks: Option<&str>, pos_args: &[String], @@ -63,6 +63,7 @@ pub fn check( librustdoc_path, tools_path, npm, + cargo, bless, extra_checks, pos_args, @@ -78,6 +79,7 @@ fn check_impl( librustdoc_path: &Path, tools_path: &Path, npm: &Path, + cargo: &Path, bless: bool, extra_checks: Option<&str>, pos_args: &[String], @@ -293,7 +295,7 @@ fn check_impl( } else { eprintln!("spellcheck files"); } - spellcheck_runner(&args)?; + spellcheck_runner(root_path, &outdir, &cargo, &args)?; } if js_lint || js_typecheck { @@ -576,34 +578,25 @@ fn shellcheck_runner(args: &[&OsStr]) -> Result<(), Error> { if status.success() { Ok(()) } else { Err(Error::FailedCheck("shellcheck")) } } -/// Check that spellchecker is installed then run it at the given path -fn spellcheck_runner(args: &[&str]) -> Result<(), Error> { - // sync version with .github/workflows/spellcheck.yml - let expected_version = "typos-cli 1.34.0"; - match Command::new("typos").arg("--version").output() { - Ok(o) => { - let stdout = String::from_utf8_lossy(&o.stdout); - if stdout.trim() != expected_version { - return Err(Error::Version { - program: "typos", - required: expected_version, - installed: stdout.trim().to_string(), - }); +/// Ensure that spellchecker is installed then run it at the given path +fn spellcheck_runner( + src_root: &Path, + outdir: &Path, + cargo: &Path, + args: &[&str], +) -> Result<(), Error> { + let bin_path = + crate::ensure_version_or_cargo_install(outdir, cargo, "typos-cli", "typos", "1.34.0")?; + match Command::new(bin_path).current_dir(src_root).args(args).status() { + Ok(status) => { + if status.success() { + Ok(()) + } else { + Err(Error::FailedCheck("typos")) } } - Err(e) if e.kind() == io::ErrorKind::NotFound => { - return Err(Error::MissingReq( - "typos", - "spellcheck file checks", - // sync version with .github/workflows/spellcheck.yml - Some("install tool via `cargo install typos-cli@1.34.0`".to_owned()), - )); - } - Err(e) => return Err(e.into()), + Err(err) => Err(Error::Generic(format!("failed to run typos tool: {err:?}"))), } - - let status = Command::new("typos").args(args).status()?; - if status.success() { Ok(()) } else { Err(Error::FailedCheck("typos")) } } /// Check git for tracked files matching an extension diff --git a/src/tools/tidy/src/lib.rs b/src/tools/tidy/src/lib.rs index 4ea9d051ddb..37713cbf75e 100644 --- a/src/tools/tidy/src/lib.rs +++ b/src/tools/tidy/src/lib.rs @@ -4,7 +4,9 @@ //! to be used by tools. use std::ffi::OsStr; +use std::path::{Path, PathBuf}; use std::process::Command; +use std::{env, io}; use build_helper::ci::CiEnv; use build_helper::git::{GitConfig, get_closest_upstream_commit}; @@ -180,6 +182,70 @@ pub fn files_modified(ci_info: &CiInfo, pred: impl Fn(&str) -> bool) -> bool { !v.is_empty() } +/// If the given executable is installed with the given version, use that, +/// otherwise install via cargo. +pub fn ensure_version_or_cargo_install( + build_dir: &Path, + cargo: &Path, + pkg_name: &str, + bin_name: &str, + version: &str, +) -> io::Result<PathBuf> { + // ignore the process exit code here and instead just let the version number check fail. + // we also importantly don't return if the program wasn't installed, + // instead we want to continue to the fallback. + 'ck: { + // FIXME: rewrite as if-let chain once this crate is 2024 edition. + let Ok(output) = Command::new(bin_name).arg("--version").output() else { + break 'ck; + }; + let Ok(s) = str::from_utf8(&output.stdout) else { + break 'ck; + }; + let Some(v) = s.trim().split_whitespace().last() else { + break 'ck; + }; + if v == version { + return Ok(PathBuf::from(bin_name)); + } + } + + let tool_root_dir = build_dir.join("misc-tools"); + let tool_bin_dir = tool_root_dir.join("bin"); + eprintln!("building external tool {bin_name} from package {pkg_name}@{version}"); + // use --force to ensure that if the required version is bumped, we update it. + // use --target-dir to ensure we have a build cache so repeated invocations aren't slow. + // modify PATH so that cargo doesn't print a warning telling the user to modify the path. + let cargo_exit_code = Command::new(cargo) + .args(["install", "--locked", "--force", "--quiet"]) + .arg("--root") + .arg(&tool_root_dir) + .arg("--target-dir") + .arg(tool_root_dir.join("target")) + .arg(format!("{pkg_name}@{version}")) + .env( + "PATH", + env::join_paths( + env::split_paths(&env::var("PATH").unwrap()) + .chain(std::iter::once(tool_bin_dir.clone())), + ) + .expect("build dir contains invalid char"), + ) + .env("RUSTFLAGS", "-Copt-level=0") + .spawn()? + .wait()?; + if !cargo_exit_code.success() { + return Err(io::Error::other("cargo install failed")); + } + let bin_path = tool_bin_dir.join(bin_name); + assert!( + matches!(bin_path.try_exists(), Ok(true)), + "cargo install did not produce the expected binary" + ); + eprintln!("finished building tool {bin_name}"); + Ok(bin_path) +} + pub mod alphabetical; pub mod bins; pub mod debug_artifacts; diff --git a/src/tools/tidy/src/main.rs b/src/tools/tidy/src/main.rs index cd2567ddb64..bfe30258915 100644 --- a/src/tools/tidy/src/main.rs +++ b/src/tools/tidy/src/main.rs @@ -184,6 +184,7 @@ fn main() { &librustdoc_path, &tools_path, &npm, + &cargo, bless, extra_checks, pos_args |