diff options
Diffstat (limited to 'compiler/rustc_parse/src/parser')
| -rw-r--r-- | compiler/rustc_parse/src/parser/attr_wrapper.rs | 133 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/diagnostics.rs | 85 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/expr.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/mod.rs | 12 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/nonterminal.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/pat.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/ty.rs | 7 |
7 files changed, 183 insertions, 72 deletions
diff --git a/compiler/rustc_parse/src/parser/attr_wrapper.rs b/compiler/rustc_parse/src/parser/attr_wrapper.rs index 49df2811d52..205cca830b2 100644 --- a/compiler/rustc_parse/src/parser/attr_wrapper.rs +++ b/compiler/rustc_parse/src/parser/attr_wrapper.rs @@ -1,3 +1,4 @@ +use std::borrow::Cow; use std::{iter, mem}; use rustc_ast::token::{Delimiter, Token, TokenKind}; @@ -6,6 +7,7 @@ use rustc_ast::tokenstream::{ Spacing, ToAttrTokenStream, }; use rustc_ast::{self as ast, AttrVec, Attribute, HasAttrs, HasTokens}; +use rustc_data_structures::fx::FxHashSet; use rustc_errors::PResult; use rustc_session::parse::ParseSess; use rustc_span::{sym, Span, DUMMY_SP}; @@ -134,9 +136,8 @@ impl ToAttrTokenStream for LazyAttrTokenStreamImpl { node_replacements.array_windows() { assert!( - node_range.0.end <= next_node_range.0.start - || node_range.0.end >= next_node_range.0.end, - "Node ranges should be disjoint or nested: ({:?}, {:?}) ({:?}, {:?})", + node_range.0.end <= next_node_range.0.start, + "Node ranges should be disjoint: ({:?}, {:?}) ({:?}, {:?})", node_range, tokens, next_node_range, @@ -144,20 +145,8 @@ impl ToAttrTokenStream for LazyAttrTokenStreamImpl { ); } - // Process the replace ranges, starting from the highest start - // position and working our way back. If have tokens like: - // - // `#[cfg(FALSE)] struct Foo { #[cfg(FALSE)] field: bool }` - // - // Then we will generate replace ranges for both - // the `#[cfg(FALSE)] field: bool` and the entire - // `#[cfg(FALSE)] struct Foo { #[cfg(FALSE)] field: bool }` - // - // By starting processing from the replace range with the greatest - // start position, we ensure that any (outer) replace range which - // encloses another (inner) replace range will fully overwrite the - // inner range's replacement. - for (node_range, target) in node_replacements.into_iter().rev() { + // Process the replace ranges. + for (node_range, target) in node_replacements.into_iter() { assert!( !node_range.0.is_empty(), "Cannot replace an empty node range: {:?}", @@ -234,6 +223,8 @@ impl<'a> Parser<'a> { force_collect: ForceCollect, f: impl FnOnce(&mut Self, AttrVec) -> PResult<'a, (R, Trailing, UsePreAttrPos)>, ) -> PResult<'a, R> { + let possible_capture_mode = self.capture_cfg; + // We must collect if anything could observe the collected tokens, i.e. // if any of the following conditions hold. // - We are force collecting tokens (because force collection requires @@ -244,9 +235,9 @@ impl<'a> Parser<'a> { // - Our target supports custom inner attributes (custom // inner attribute invocation might require token capturing). || R::SUPPORTS_CUSTOM_INNER_ATTRS - // - We are in `capture_cfg` mode (which requires tokens if + // - We are in "possible capture mode" (which requires tokens if // the parsed node has `#[cfg]` or `#[cfg_attr]` attributes). - || self.capture_cfg; + || possible_capture_mode; if !needs_collection { return Ok(f(self, attrs.attrs)?.0); } @@ -267,18 +258,48 @@ impl<'a> Parser<'a> { res? }; - // When we're not in `capture_cfg` mode, then skip collecting and - // return early if either of the following conditions hold. // - `None`: Our target doesn't support tokens at all (e.g. `NtIdent`). + // - `Some(None)`: Our target supports tokens and has none. // - `Some(Some(_))`: Our target already has tokens set (e.g. we've - // parsed something like `#[my_attr] $item`). The actual parsing code - // takes care of prepending any attributes to the nonterminal, so we - // don't need to modify the already captured tokens. + // parsed something like `#[my_attr] $item`). + let ret_can_hold_tokens = matches!(ret.tokens_mut(), Some(None)); + + // Ignore any attributes we've previously processed. This happens when + // an inner call to `collect_tokens` returns an AST node and then an + // outer call ends up with the same AST node without any additional + // wrapping layer. + let mut seen_indices = FxHashSet::default(); + for (i, attr) in ret.attrs().iter().enumerate() { + let is_unseen = self.capture_state.seen_attrs.insert(attr.id); + if !is_unseen { + seen_indices.insert(i); + } + } + let ret_attrs: Cow<'_, [Attribute]> = + if seen_indices.is_empty() { + Cow::Borrowed(ret.attrs()) + } else { + let ret_attrs = + ret.attrs() + .iter() + .enumerate() + .filter_map(|(i, attr)| { + if seen_indices.contains(&i) { None } else { Some(attr.clone()) } + }) + .collect(); + Cow::Owned(ret_attrs) + }; + + // When we're not in "definite capture mode", then skip collecting and + // return early if `ret` doesn't support tokens or already has some. // // Note that this check is independent of `force_collect`. There's no // need to collect tokens when we don't support tokens or already have // tokens. - if !self.capture_cfg && matches!(ret.tokens_mut(), None | Some(Some(_))) { + let definite_capture_mode = self.capture_cfg + && matches!(self.capture_state.capturing, Capturing::Yes) + && has_cfg_or_cfg_attr(&ret_attrs); + if !definite_capture_mode && !ret_can_hold_tokens { return Ok(ret); } @@ -297,12 +318,12 @@ impl<'a> Parser<'a> { // outer and inner attributes. So this check is more precise than // the earlier `needs_tokens` check, and we don't need to // check `R::SUPPORTS_CUSTOM_INNER_ATTRS`.) - || needs_tokens(ret.attrs()) - // - We are in `capture_cfg` mode and there are `#[cfg]` or - // `#[cfg_attr]` attributes. (During normal non-`capture_cfg` - // parsing, we don't need any special capturing for those - // attributes, because they're builtin.) - || (self.capture_cfg && has_cfg_or_cfg_attr(ret.attrs())); + || needs_tokens(&ret_attrs) + // - We are in "definite capture mode", which requires that there + // are `#[cfg]` or `#[cfg_attr]` attributes. (During normal + // non-`capture_cfg` parsing, we don't need any special capturing + // for those attributes, because they're builtin.) + || definite_capture_mode; if !needs_collection { return Ok(ret); } @@ -336,7 +357,7 @@ impl<'a> Parser<'a> { // `Parser::parse_inner_attributes`, and pair them in a `ParserReplacement` with `None`, // which means the relevant tokens will be removed. (More details below.) let mut inner_attr_parser_replacements = Vec::new(); - for attr in ret.attrs() { + for attr in ret_attrs.iter() { if attr.style == ast::AttrStyle::Inner { if let Some(inner_attr_parser_range) = self.capture_state.inner_attr_parser_ranges.remove(&attr.id) @@ -359,11 +380,10 @@ impl<'a> Parser<'a> { // from `ParserRange` form to `NodeRange` form. We will perform the actual // replacement only when we convert the `LazyAttrTokenStream` to an // `AttrTokenStream`. - self.capture_state.parser_replacements - [parser_replacements_start..parser_replacements_end] - .iter() - .cloned() - .chain(inner_attr_parser_replacements.iter().cloned()) + self.capture_state + .parser_replacements + .drain(parser_replacements_start..parser_replacements_end) + .chain(inner_attr_parser_replacements.into_iter()) .map(|(parser_range, data)| { (NodeRange::new(parser_range, collect_pos.start_pos), data) }) @@ -399,20 +419,12 @@ impl<'a> Parser<'a> { break_last_token: self.break_last_token, node_replacements, }); + let mut tokens_used = false; - // If we support tokens and don't already have them, store the newly captured tokens. - if let Some(target_tokens @ None) = ret.tokens_mut() { - *target_tokens = Some(tokens.clone()); - } - - // If `capture_cfg` is set and we're inside a recursive call to - // `collect_tokens`, then we need to register a replace range if we - // have `#[cfg]` or `#[cfg_attr]`. This allows us to run eager - // cfg-expansion on the captured token stream. - if self.capture_cfg - && matches!(self.capture_state.capturing, Capturing::Yes) - && has_cfg_or_cfg_attr(ret.attrs()) - { + // If in "definite capture mode" we need to register a replace range + // for the `#[cfg]` and/or `#[cfg_attr]` attrs. This allows us to run + // eager cfg-expansion on the captured token stream. + if definite_capture_mode { assert!(!self.break_last_token, "Should not have unglued last token with cfg attr"); // What is the status here when parsing the example code at the top of this method? @@ -429,7 +441,9 @@ impl<'a> Parser<'a> { // cfg-expand this AST node. let start_pos = if has_outer_attrs { attrs.start_pos.unwrap() } else { collect_pos.start_pos }; - let target = AttrsTarget { attrs: ret.attrs().iter().cloned().collect(), tokens }; + let target = + AttrsTarget { attrs: ret_attrs.iter().cloned().collect(), tokens: tokens.clone() }; + tokens_used = true; self.capture_state .parser_replacements .push((ParserRange(start_pos..end_pos), Some(target))); @@ -438,7 +452,16 @@ impl<'a> Parser<'a> { // the outermost call to this method. self.capture_state.parser_replacements.clear(); self.capture_state.inner_attr_parser_ranges.clear(); + self.capture_state.seen_attrs.clear(); } + + // If we support tokens and don't already have them, store the newly captured tokens. + if let Some(target_tokens @ None) = ret.tokens_mut() { + tokens_used = true; + *target_tokens = Some(tokens); + } + + assert!(tokens_used); // check we didn't create `tokens` unnecessarily Ok(ret) } } @@ -510,9 +533,11 @@ fn make_attr_token_stream( } /// Tokens are needed if: -/// - any non-single-segment attributes (other than doc comments) are present; or -/// - any `cfg_attr` attributes are present; -/// - any single-segment, non-builtin attributes are present. +/// - any non-single-segment attributes (other than doc comments) are present, +/// e.g. `rustfmt::skip`; or +/// - any `cfg_attr` attributes are present; or +/// - any single-segment, non-builtin attributes are present, e.g. `derive`, +/// `test`, `global_allocator`. fn needs_tokens(attrs: &[ast::Attribute]) -> bool { attrs.iter().any(|attr| match attr.ident() { None => !attr.is_doc_comment(), diff --git a/compiler/rustc_parse/src/parser/diagnostics.rs b/compiler/rustc_parse/src/parser/diagnostics.rs index fcdc10c0837..f6f66821df7 100644 --- a/compiler/rustc_parse/src/parser/diagnostics.rs +++ b/compiler/rustc_parse/src/parser/diagnostics.rs @@ -19,8 +19,9 @@ use rustc_errors::{ Subdiagnostic, }; use rustc_session::errors::ExprParenthesesNeeded; +use rustc_span::edit_distance::find_best_match_for_name; use rustc_span::source_map::Spanned; -use rustc_span::symbol::{kw, sym, Ident}; +use rustc_span::symbol::{kw, sym, AllKeywords, Ident}; use rustc_span::{BytePos, Span, SpanSnippetError, Symbol, DUMMY_SP}; use thin_vec::{thin_vec, ThinVec}; use tracing::{debug, trace}; @@ -203,6 +204,37 @@ impl std::fmt::Display for UnaryFixity { } } +#[derive(Debug, rustc_macros::Subdiagnostic)] +#[suggestion( + parse_misspelled_kw, + applicability = "machine-applicable", + code = "{similar_kw}", + style = "verbose" +)] +struct MisspelledKw { + similar_kw: String, + #[primary_span] + span: Span, + is_incorrect_case: bool, +} + +/// Checks if the given `lookup` identifier is similar to any keyword symbol in `candidates`. +fn find_similar_kw(lookup: Ident, candidates: &[Symbol]) -> Option<MisspelledKw> { + let lowercase = lookup.name.as_str().to_lowercase(); + let lowercase_sym = Symbol::intern(&lowercase); + if candidates.contains(&lowercase_sym) { + Some(MisspelledKw { similar_kw: lowercase, span: lookup.span, is_incorrect_case: true }) + } else if let Some(similar_sym) = find_best_match_for_name(candidates, lookup.name, None) { + Some(MisspelledKw { + similar_kw: similar_sym.to_string(), + span: lookup.span, + is_incorrect_case: false, + }) + } else { + None + } +} + struct MultiSugg { msg: String, patches: Vec<(Span, String)>, @@ -638,9 +670,9 @@ impl<'a> Parser<'a> { let concat = Symbol::intern(&format!("{prev}{cur}")); let ident = Ident::new(concat, DUMMY_SP); if ident.is_used_keyword() || ident.is_reserved() || ident.is_raw_guess() { - let span = self.prev_token.span.to(self.token.span); + let concat_span = self.prev_token.span.to(self.token.span); err.span_suggestion_verbose( - span, + concat_span, format!("consider removing the space to spell keyword `{concat}`"), concat, Applicability::MachineApplicable, @@ -741,9 +773,55 @@ impl<'a> Parser<'a> { err.span_label(sp, label_exp); err.span_label(self.token.span, "unexpected token"); } + + // Check for misspelled keywords if there are no suggestions added to the diagnostic. + if err.suggestions.as_ref().is_ok_and(|code_suggestions| code_suggestions.is_empty()) { + self.check_for_misspelled_kw(&mut err, &expected); + } Err(err) } + /// Checks if the current token or the previous token are misspelled keywords + /// and adds a helpful suggestion. + fn check_for_misspelled_kw(&self, err: &mut Diag<'_>, expected: &[TokenType]) { + let Some((curr_ident, _)) = self.token.ident() else { + return; + }; + let expected_tokens: &[TokenType] = + expected.len().checked_sub(10).map_or(&expected, |index| &expected[index..]); + let expected_keywords: Vec<Symbol> = expected_tokens + .iter() + .filter_map(|token| if let TokenType::Keyword(kw) = token { Some(*kw) } else { None }) + .collect(); + + // When there are a few keywords in the last ten elements of `self.expected_tokens` and the current + // token is an identifier, it's probably a misspelled keyword. + // This handles code like `async Move {}`, misspelled `if` in match guard, misspelled `else` in `if`-`else` + // and mispelled `where` in a where clause. + if !expected_keywords.is_empty() + && !curr_ident.is_used_keyword() + && let Some(misspelled_kw) = find_similar_kw(curr_ident, &expected_keywords) + { + err.subdiagnostic(misspelled_kw); + } else if let Some((prev_ident, _)) = self.prev_token.ident() + && !prev_ident.is_used_keyword() + { + // We generate a list of all keywords at runtime rather than at compile time + // so that it gets generated only when the diagnostic needs it. + // Also, it is unlikely that this list is generated multiple times because the + // parser halts after execution hits this path. + let all_keywords = AllKeywords::new().collect_used(|| prev_ident.span.edition()); + + // Otherwise, check the previous token with all the keywords as possible candidates. + // This handles code like `Struct Human;` and `While a < b {}`. + // We check the previous token only when the current token is an identifier to avoid false + // positives like suggesting keyword `for` for `extern crate foo {}`. + if let Some(misspelled_kw) = find_similar_kw(prev_ident, &all_keywords) { + err.subdiagnostic(misspelled_kw); + } + } + } + /// The user has written `#[attr] expr` which is unsupported. (#106020) pub(super) fn attr_on_non_tail_expr(&self, expr: &Expr) -> ErrorGuaranteed { // Missing semicolon typo error. @@ -846,6 +924,7 @@ impl<'a> Parser<'a> { ); } } + err.emit() } diff --git a/compiler/rustc_parse/src/parser/expr.rs b/compiler/rustc_parse/src/parser/expr.rs index 84684e808d9..ecc4cd96faf 100644 --- a/compiler/rustc_parse/src/parser/expr.rs +++ b/compiler/rustc_parse/src/parser/expr.rs @@ -2050,7 +2050,7 @@ impl<'a> Parser<'a> { }; // On an error path, eagerly consider a lifetime to be an unclosed character lit, if that // makes sense. - if let Some(ident) = self.token.lifetime() + if let Some((ident, IdentIsRaw::No)) = self.token.lifetime() && could_be_unclosed_char_literal(ident) { let lt = self.expect_lifetime(); @@ -2925,9 +2925,9 @@ impl<'a> Parser<'a> { } pub(crate) fn eat_label(&mut self) -> Option<Label> { - if let Some(ident) = self.token.lifetime() { + if let Some((ident, is_raw)) = self.token.lifetime() { // Disallow `'fn`, but with a better error message than `expect_lifetime`. - if ident.without_first_quote().is_reserved() { + if matches!(is_raw, IdentIsRaw::No) && ident.without_first_quote().is_reserved() { self.dcx().emit_err(errors::InvalidLabel { span: ident.span, name: ident.name }); } diff --git a/compiler/rustc_parse/src/parser/mod.rs b/compiler/rustc_parse/src/parser/mod.rs index 61e3fa2e6c5..3ee6e742d1b 100644 --- a/compiler/rustc_parse/src/parser/mod.rs +++ b/compiler/rustc_parse/src/parser/mod.rs @@ -35,6 +35,7 @@ use rustc_ast_pretty::pprust; use rustc_data_structures::fx::FxHashMap; use rustc_data_structures::sync::Lrc; use rustc_errors::{Applicability, Diag, FatalError, MultiSpan, PResult}; +use rustc_index::interval::IntervalSet; use rustc_session::parse::ParseSess; use rustc_span::symbol::{kw, sym, Ident, Symbol}; use rustc_span::{Span, DUMMY_SP}; @@ -183,7 +184,7 @@ pub struct Parser<'a> { // This type is used a lot, e.g. it's cloned when matching many declarative macro rules with nonterminals. Make sure // it doesn't unintentionally get bigger. #[cfg(target_pointer_width = "64")] -rustc_data_structures::static_assert_size!(Parser<'_>, 256); +rustc_data_structures::static_assert_size!(Parser<'_>, 288); /// Stores span information about a closure. #[derive(Clone, Debug)] @@ -238,7 +239,8 @@ impl NodeRange { // is the position of the function's start token. This gives // `NodeRange(10..15)`. fn new(ParserRange(parser_range): ParserRange, start_pos: u32) -> NodeRange { - assert!(parser_range.start >= start_pos && parser_range.end >= start_pos); + assert!(!parser_range.is_empty()); + assert!(parser_range.start >= start_pos); NodeRange((parser_range.start - start_pos)..(parser_range.end - start_pos)) } } @@ -260,6 +262,9 @@ struct CaptureState { capturing: Capturing, parser_replacements: Vec<ParserReplacement>, inner_attr_parser_ranges: FxHashMap<AttrId, ParserRange>, + // `IntervalSet` is good for perf because attrs are mostly added to this + // set in contiguous ranges. + seen_attrs: IntervalSet<AttrId>, } /// Iterator over a `TokenStream` that produces `Token`s. It's a bit odd that @@ -457,6 +462,7 @@ impl<'a> Parser<'a> { capturing: Capturing::No, parser_replacements: Vec::new(), inner_attr_parser_ranges: Default::default(), + seen_attrs: IntervalSet::new(u32::MAX as usize), }, current_closure: None, recovery: Recovery::Allowed, @@ -1666,7 +1672,7 @@ enum FlatToken { pub enum ParseNtResult { Tt(TokenTree), Ident(Ident, IdentIsRaw), - Lifetime(Ident), + Lifetime(Ident, IdentIsRaw), /// This case will eventually be removed, along with `Token::Interpolate`. Nt(Lrc<Nonterminal>), diff --git a/compiler/rustc_parse/src/parser/nonterminal.rs b/compiler/rustc_parse/src/parser/nonterminal.rs index e66d0df012b..44b169c881e 100644 --- a/compiler/rustc_parse/src/parser/nonterminal.rs +++ b/compiler/rustc_parse/src/parser/nonterminal.rs @@ -88,7 +88,7 @@ impl<'a> Parser<'a> { }, NonterminalKind::Pat(pat_kind) => token.can_begin_pattern(pat_kind), NonterminalKind::Lifetime => match &token.kind { - token::Lifetime(_) | token::NtLifetime(..) => true, + token::Lifetime(..) | token::NtLifetime(..) => true, _ => false, }, NonterminalKind::TT | NonterminalKind::Item | NonterminalKind::Stmt => { @@ -171,9 +171,9 @@ impl<'a> Parser<'a> { NonterminalKind::Lifetime => { // We want to keep `'keyword` parsing, just like `keyword` is still // an ident for nonterminal purposes. - return if let Some(ident) = self.token.lifetime() { + return if let Some((ident, is_raw)) = self.token.lifetime() { self.bump(); - Ok(ParseNtResult::Lifetime(ident)) + Ok(ParseNtResult::Lifetime(ident, is_raw)) } else { Err(self.dcx().create_err(UnexpectedNonterminal::Lifetime { span: self.token.span, diff --git a/compiler/rustc_parse/src/parser/pat.rs b/compiler/rustc_parse/src/parser/pat.rs index f87b5649654..cbd35ffdfa9 100644 --- a/compiler/rustc_parse/src/parser/pat.rs +++ b/compiler/rustc_parse/src/parser/pat.rs @@ -1,6 +1,6 @@ use rustc_ast::mut_visit::{walk_pat, MutVisitor}; use rustc_ast::ptr::P; -use rustc_ast::token::{self, BinOpToken, Delimiter, Token}; +use rustc_ast::token::{self, BinOpToken, Delimiter, IdentIsRaw, Token}; use rustc_ast::{ self as ast, AttrVec, BindingMode, ByRef, Expr, ExprKind, MacCall, Mutability, Pat, PatField, PatFieldsRest, PatKind, Path, QSelf, RangeEnd, RangeSyntax, @@ -548,7 +548,7 @@ impl<'a> Parser<'a> { None => PatKind::Path(qself, path), } } - } else if let Some(lt) = self.token.lifetime() + } else if let Some((lt, IdentIsRaw::No)) = self.token.lifetime() // In pattern position, we're totally fine with using "next token isn't colon" // as a heuristic. We could probably just always try to recover if it's a lifetime, // because we never have `'a: label {}` in a pattern position anyways, but it does @@ -689,7 +689,7 @@ impl<'a> Parser<'a> { /// Parse `&pat` / `&mut pat`. fn parse_pat_deref(&mut self, expected: Option<Expected>) -> PResult<'a, PatKind> { self.expect_and()?; - if let Some(lifetime) = self.token.lifetime() { + if let Some((lifetime, _)) = self.token.lifetime() { self.bump(); // `'a` self.dcx().emit_err(UnexpectedLifetimeInPattern { diff --git a/compiler/rustc_parse/src/parser/ty.rs b/compiler/rustc_parse/src/parser/ty.rs index 52f0b1c1b04..dd1cc75c7ff 100644 --- a/compiler/rustc_parse/src/parser/ty.rs +++ b/compiler/rustc_parse/src/parser/ty.rs @@ -1,5 +1,5 @@ use rustc_ast::ptr::P; -use rustc_ast::token::{self, BinOpToken, Delimiter, Token, TokenKind}; +use rustc_ast::token::{self, BinOpToken, Delimiter, IdentIsRaw, Token, TokenKind}; use rustc_ast::util::case::Case; use rustc_ast::{ self as ast, BareFnTy, BoundAsyncness, BoundConstness, BoundPolarity, FnRetTy, GenericBound, @@ -1285,8 +1285,9 @@ impl<'a> Parser<'a> { /// Parses a single lifetime `'a` or panics. pub(super) fn expect_lifetime(&mut self) -> Lifetime { - if let Some(ident) = self.token.lifetime() { - if ident.without_first_quote().is_reserved() + if let Some((ident, is_raw)) = self.token.lifetime() { + if matches!(is_raw, IdentIsRaw::No) + && ident.without_first_quote().is_reserved() && ![kw::UnderscoreLifetime, kw::StaticLifetime].contains(&ident.name) { self.dcx().emit_err(errors::KeywordLifetime { span: ident.span }); |