diff options
| author | Alex Crichton <alex@alexcrichton.com> | 2019-05-21 15:37:07 -0700 |
|---|---|---|
| committer | Alex Crichton <alex@alexcrichton.com> | 2019-05-21 15:37:07 -0700 |
| commit | fe3dd0b50fef21d14591c960a9610bafb224cdbf (patch) | |
| tree | ee0b62d8e500d4ce4b6f50b4fe5d9056b9826072 /src/libsyntax/parse | |
| parent | e764f475ca7fffd6167ea991afc7d1b2b3f642dc (diff) | |
| parent | 50a0defd5a93523067ef239936cc2e0755220904 (diff) | |
| download | rust-fe3dd0b50fef21d14591c960a9610bafb224cdbf.tar.gz rust-fe3dd0b50fef21d14591c960a9610bafb224cdbf.zip | |
Merge remote-tracking branch 'origin/master' into azure-pipelines
Diffstat (limited to 'src/libsyntax/parse')
| -rw-r--r-- | src/libsyntax/parse/classify.rs | 13 | ||||
| -rw-r--r-- | src/libsyntax/parse/diagnostics.rs | 304 | ||||
| -rw-r--r-- | src/libsyntax/parse/lexer/comments.rs | 29 | ||||
| -rw-r--r-- | src/libsyntax/parse/lexer/mod.rs | 45 | ||||
| -rw-r--r-- | src/libsyntax/parse/lexer/tokentrees.rs | 67 | ||||
| -rw-r--r-- | src/libsyntax/parse/literal.rs | 487 | ||||
| -rw-r--r-- | src/libsyntax/parse/mod.rs | 329 | ||||
| -rw-r--r-- | src/libsyntax/parse/parser.rs | 540 | ||||
| -rw-r--r-- | src/libsyntax/parse/token.rs | 30 |
9 files changed, 986 insertions, 858 deletions
diff --git a/src/libsyntax/parse/classify.rs b/src/libsyntax/parse/classify.rs index b4103440e35..dfd6f451c28 100644 --- a/src/libsyntax/parse/classify.rs +++ b/src/libsyntax/parse/classify.rs @@ -25,16 +25,3 @@ pub fn expr_requires_semi_to_be_stmt(e: &ast::Expr) -> bool { _ => true, } } - -/// this statement requires a semicolon after it. -/// note that in one case (`stmt_semi`), we've already -/// seen the semicolon, and thus don't need another. -pub fn stmt_ends_with_semi(stmt: &ast::StmtKind) -> bool { - match *stmt { - ast::StmtKind::Local(_) => true, - ast::StmtKind::Expr(ref e) => expr_requires_semi_to_be_stmt(e), - ast::StmtKind::Item(_) | - ast::StmtKind::Semi(..) | - ast::StmtKind::Mac(..) => false, - } -} diff --git a/src/libsyntax/parse/diagnostics.rs b/src/libsyntax/parse/diagnostics.rs index 32e1ee94f0d..1a2393be806 100644 --- a/src/libsyntax/parse/diagnostics.rs +++ b/src/libsyntax/parse/diagnostics.rs @@ -1,14 +1,16 @@ use crate::ast; -use crate::ast::{Expr, ExprKind, Item, ItemKind, Pat, PatKind, QSelf, Ty, TyKind}; -use crate::parse::parser::PathStyle; +use crate::ast::{BlockCheckMode, Expr, ExprKind, Item, ItemKind, Pat, PatKind, QSelf, Ty, TyKind}; +use crate::parse::parser::{BlockMode, PathStyle, TokenType, SemiColonMode}; use crate::parse::token; use crate::parse::PResult; use crate::parse::Parser; use crate::print::pprust; use crate::ptr::P; +use crate::symbol::keywords; use crate::ThinVec; -use errors::Applicability; +use errors::{Applicability, DiagnosticBuilder}; use syntax_pos::Span; +use log::debug; pub trait RecoverQPath: Sized + 'static { const PATH_STYLE: PathStyle = PathStyle::Expr; @@ -223,4 +225,300 @@ impl<'a> Parser<'a> { false } } + + /// Consume alternative await syntaxes like `await <expr>`, `await? <expr>`, `await(<expr>)` + /// and `await { <expr> }`. + crate fn parse_incorrect_await_syntax( + &mut self, + lo: Span, + await_sp: Span, + ) -> PResult<'a, (Span, ExprKind)> { + let is_question = self.eat(&token::Question); // Handle `await? <expr>`. + let expr = if self.token == token::OpenDelim(token::Brace) { + // Handle `await { <expr> }`. + // This needs to be handled separatedly from the next arm to avoid + // interpreting `await { <expr> }?` as `<expr>?.await`. + self.parse_block_expr( + None, + self.span, + BlockCheckMode::Default, + ThinVec::new(), + ) + } else { + self.parse_expr() + }.map_err(|mut err| { + err.span_label(await_sp, "while parsing this incorrect await expression"); + err + })?; + let expr_str = self.sess.source_map().span_to_snippet(expr.span) + .unwrap_or_else(|_| pprust::expr_to_string(&expr)); + let suggestion = format!("{}.await{}", expr_str, if is_question { "?" } else { "" }); + let sp = lo.to(expr.span); + let app = match expr.node { + ExprKind::Try(_) => Applicability::MaybeIncorrect, // `await <expr>?` + _ => Applicability::MachineApplicable, + }; + self.struct_span_err(sp, "incorrect use of `await`") + .span_suggestion(sp, "`await` is a postfix operation", suggestion, app) + .emit(); + Ok((sp, ExprKind::Await(ast::AwaitOrigin::FieldLike, expr))) + } + + /// If encountering `future.await()`, consume and emit error. + crate fn recover_from_await_method_call(&mut self) { + if self.token == token::OpenDelim(token::Paren) && + self.look_ahead(1, |t| t == &token::CloseDelim(token::Paren)) + { + // future.await() + let lo = self.span; + self.bump(); // ( + let sp = lo.to(self.span); + self.bump(); // ) + self.struct_span_err(sp, "incorrect use of `await`") + .span_suggestion( + sp, + "`await` is not a method call, remove the parentheses", + String::new(), + Applicability::MachineApplicable, + ).emit() + } + } + + crate fn could_ascription_be_path(&self, node: &ast::ExprKind) -> bool { + self.token.is_ident() && + if let ast::ExprKind::Path(..) = node { true } else { false } && + !self.token.is_reserved_ident() && // v `foo:bar(baz)` + self.look_ahead(1, |t| t == &token::OpenDelim(token::Paren)) || + self.look_ahead(1, |t| t == &token::Lt) && // `foo:bar<baz` + self.look_ahead(2, |t| t.is_ident()) || + self.look_ahead(1, |t| t == &token::Colon) && // `foo:bar:baz` + self.look_ahead(2, |t| t.is_ident()) || + self.look_ahead(1, |t| t == &token::ModSep) && // `foo:bar::baz` + self.look_ahead(2, |t| t.is_ident()) + } + + crate fn bad_type_ascription( + &self, + err: &mut DiagnosticBuilder<'a>, + lhs_span: Span, + cur_op_span: Span, + next_sp: Span, + maybe_path: bool, + ) { + err.span_label(self.span, "expecting a type here because of type ascription"); + let cm = self.sess.source_map(); + let next_pos = cm.lookup_char_pos(next_sp.lo()); + let op_pos = cm.lookup_char_pos(cur_op_span.hi()); + if op_pos.line != next_pos.line { + err.span_suggestion( + cur_op_span, + "try using a semicolon", + ";".to_string(), + Applicability::MaybeIncorrect, + ); + } else { + if maybe_path { + err.span_suggestion( + cur_op_span, + "maybe you meant to write a path separator here", + "::".to_string(), + Applicability::MaybeIncorrect, + ); + } else { + err.note("type ascription is a nightly-only feature that lets \ + you annotate an expression with a type: `<expr>: <type>`") + .span_note( + lhs_span, + "this expression expects an ascribed type after the colon", + ) + .help("this might be indicative of a syntax error elsewhere"); + } + } + } + + crate fn recover_seq_parse_error( + &mut self, + delim: token::DelimToken, + lo: Span, + result: PResult<'a, P<Expr>>, + ) -> P<Expr> { + match result { + Ok(x) => x, + Err(mut err) => { + err.emit(); + // recover from parse error + self.consume_block(delim); + self.mk_expr(lo.to(self.prev_span), ExprKind::Err, ThinVec::new()) + } + } + } + + crate fn recover_closing_delimiter( + &mut self, + tokens: &[token::Token], + mut err: DiagnosticBuilder<'a>, + ) -> PResult<'a, bool> { + let mut pos = None; + // we want to use the last closing delim that would apply + for (i, unmatched) in self.unclosed_delims.iter().enumerate().rev() { + if tokens.contains(&token::CloseDelim(unmatched.expected_delim)) + && Some(self.span) > unmatched.unclosed_span + { + pos = Some(i); + } + } + match pos { + Some(pos) => { + // Recover and assume that the detected unclosed delimiter was meant for + // this location. Emit the diagnostic and act as if the delimiter was + // present for the parser's sake. + + // Don't attempt to recover from this unclosed delimiter more than once. + let unmatched = self.unclosed_delims.remove(pos); + let delim = TokenType::Token(token::CloseDelim(unmatched.expected_delim)); + + // We want to suggest the inclusion of the closing delimiter where it makes + // the most sense, which is immediately after the last token: + // + // {foo(bar {}} + // - ^ + // | | + // | help: `)` may belong here (FIXME: #58270) + // | + // unclosed delimiter + if let Some(sp) = unmatched.unclosed_span { + err.span_label(sp, "unclosed delimiter"); + } + err.span_suggestion_short( + self.sess.source_map().next_point(self.prev_span), + &format!("{} may belong here", delim.to_string()), + delim.to_string(), + Applicability::MaybeIncorrect, + ); + err.emit(); + self.expected_tokens.clear(); // reduce errors + Ok(true) + } + _ => Err(err), + } + } + + /// Recover from `pub` keyword in places where it seems _reasonable_ but isn't valid. + crate fn eat_bad_pub(&mut self) { + if self.token.is_keyword(keywords::Pub) { + match self.parse_visibility(false) { + Ok(vis) => { + self.diagnostic() + .struct_span_err(vis.span, "unnecessary visibility qualifier") + .span_label(vis.span, "`pub` not permitted here") + .emit(); + } + Err(mut err) => err.emit(), + } + } + } + + // Eat tokens until we can be relatively sure we reached the end of the + // statement. This is something of a best-effort heuristic. + // + // We terminate when we find an unmatched `}` (without consuming it). + crate fn recover_stmt(&mut self) { + self.recover_stmt_(SemiColonMode::Ignore, BlockMode::Ignore) + } + + // If `break_on_semi` is `Break`, then we will stop consuming tokens after + // finding (and consuming) a `;` outside of `{}` or `[]` (note that this is + // approximate - it can mean we break too early due to macros, but that + // should only lead to sub-optimal recovery, not inaccurate parsing). + // + // If `break_on_block` is `Break`, then we will stop consuming tokens + // after finding (and consuming) a brace-delimited block. + crate fn recover_stmt_(&mut self, break_on_semi: SemiColonMode, break_on_block: BlockMode) { + let mut brace_depth = 0; + let mut bracket_depth = 0; + let mut in_block = false; + debug!("recover_stmt_ enter loop (semi={:?}, block={:?})", + break_on_semi, break_on_block); + loop { + debug!("recover_stmt_ loop {:?}", self.token); + match self.token { + token::OpenDelim(token::DelimToken::Brace) => { + brace_depth += 1; + self.bump(); + if break_on_block == BlockMode::Break && + brace_depth == 1 && + bracket_depth == 0 { + in_block = true; + } + } + token::OpenDelim(token::DelimToken::Bracket) => { + bracket_depth += 1; + self.bump(); + } + token::CloseDelim(token::DelimToken::Brace) => { + if brace_depth == 0 { + debug!("recover_stmt_ return - close delim {:?}", self.token); + break; + } + brace_depth -= 1; + self.bump(); + if in_block && bracket_depth == 0 && brace_depth == 0 { + debug!("recover_stmt_ return - block end {:?}", self.token); + break; + } + } + token::CloseDelim(token::DelimToken::Bracket) => { + bracket_depth -= 1; + if bracket_depth < 0 { + bracket_depth = 0; + } + self.bump(); + } + token::Eof => { + debug!("recover_stmt_ return - Eof"); + break; + } + token::Semi => { + self.bump(); + if break_on_semi == SemiColonMode::Break && + brace_depth == 0 && + bracket_depth == 0 { + debug!("recover_stmt_ return - Semi"); + break; + } + } + token::Comma if break_on_semi == SemiColonMode::Comma && + brace_depth == 0 && + bracket_depth == 0 => + { + debug!("recover_stmt_ return - Semi"); + break; + } + _ => { + self.bump() + } + } + } + } + + crate fn consume_block(&mut self, delim: token::DelimToken) { + let mut brace_depth = 0; + loop { + if self.eat(&token::OpenDelim(delim)) { + brace_depth += 1; + } else if self.eat(&token::CloseDelim(delim)) { + if brace_depth == 0 { + return; + } else { + brace_depth -= 1; + continue; + } + } else if self.token == token::Eof || self.eat(&token::CloseDelim(token::NoDelim)) { + return; + } else { + self.bump(); + } + } + } + } diff --git a/src/libsyntax/parse/lexer/comments.rs b/src/libsyntax/parse/lexer/comments.rs index 74fff3324ea..97d3fc002e9 100644 --- a/src/libsyntax/parse/lexer/comments.rs +++ b/src/libsyntax/parse/lexer/comments.rs @@ -3,8 +3,7 @@ pub use CommentStyle::*; use crate::ast; use crate::source_map::SourceMap; use crate::parse::lexer::{is_block_doc_comment, is_pattern_whitespace}; -use crate::parse::lexer::{self, ParseSess, StringReader, TokenAndSpan}; -use crate::print::pprust; +use crate::parse::lexer::{self, ParseSess, StringReader}; use syntax_pos::{BytePos, CharPos, Pos, FileName}; use log::debug; @@ -339,16 +338,9 @@ fn consume_comment(rdr: &mut StringReader<'_>, debug!("<<< consume comment"); } -#[derive(Clone)] -pub struct Literal { - pub lit: String, - pub pos: BytePos, -} - // it appears this function is called only from pprust... that's // probably not a good thing. -pub fn gather_comments_and_literals(sess: &ParseSess, path: FileName, srdr: &mut dyn Read) - -> (Vec<Comment>, Vec<Literal>) +pub fn gather_comments(sess: &ParseSess, path: FileName, srdr: &mut dyn Read) -> Vec<Comment> { let mut src = String::new(); srdr.read_to_string(&mut src).unwrap(); @@ -357,7 +349,6 @@ pub fn gather_comments_and_literals(sess: &ParseSess, path: FileName, srdr: &mut let mut rdr = lexer::StringReader::new_raw(sess, source_file, None); let mut comments: Vec<Comment> = Vec::new(); - let mut literals: Vec<Literal> = Vec::new(); let mut code_to_the_left = false; // Only code let mut anything_to_the_left = false; // Code or comments @@ -382,26 +373,12 @@ pub fn gather_comments_and_literals(sess: &ParseSess, path: FileName, srdr: &mut } } - let bstart = rdr.pos; rdr.next_token(); - // discard, and look ahead; we're working with internal state - let TokenAndSpan { tok, sp } = rdr.peek(); - if tok.is_lit() { - rdr.with_str_from(bstart, |s| { - debug!("tok lit: {}", s); - literals.push(Literal { - lit: s.to_string(), - pos: sp.lo(), - }); - }) - } else { - debug!("tok: {}", pprust::token_to_string(&tok)); - } code_to_the_left = true; anything_to_the_left = true; } - (comments, literals) + comments } #[cfg(test)] diff --git a/src/libsyntax/parse/lexer/mod.rs b/src/libsyntax/parse/lexer/mod.rs index 2882acb0e78..47da3ee6a6c 100644 --- a/src/libsyntax/parse/lexer/mod.rs +++ b/src/libsyntax/parse/lexer/mod.rs @@ -62,19 +62,7 @@ pub struct StringReader<'a> { // cache a direct reference to the source text, so that we don't have to // retrieve it via `self.source_file.src.as_ref().unwrap()` all the time. src: Lrc<String>, - token: token::Token, - span: Span, - /// The raw source span which *does not* take `override_span` into account - span_src_raw: Span, - /// Stack of open delimiters and their spans. Used for error message. - open_braces: Vec<(token::DelimToken, Span)>, - crate unmatched_braces: Vec<UnmatchedBrace>, - /// The type and spans for all braces - /// - /// Used only for error recovery when arriving to EOF with mismatched braces. - matching_delim_spans: Vec<(token::DelimToken, Span, Span)>, - crate override_span: Option<Span>, - last_unclosed_found_span: Option<Span>, + override_span: Option<Span>, } impl<'a> StringReader<'a> { @@ -121,8 +109,6 @@ impl<'a> StringReader<'a> { sp: self.peek_span, }; self.advance_token()?; - self.span_src_raw = self.peek_span_src_raw; - Ok(ret_val) } @@ -159,9 +145,6 @@ impl<'a> StringReader<'a> { } } - self.token = t.tok.clone(); - self.span = t.sp; - Ok(t) } @@ -251,29 +234,10 @@ impl<'a> StringReader<'a> { peek_span_src_raw: syntax_pos::DUMMY_SP, src, fatal_errs: Vec::new(), - token: token::Eof, - span: syntax_pos::DUMMY_SP, - span_src_raw: syntax_pos::DUMMY_SP, - open_braces: Vec::new(), - unmatched_braces: Vec::new(), - matching_delim_spans: Vec::new(), override_span, - last_unclosed_found_span: None, } } - pub fn new(sess: &'a ParseSess, - source_file: Lrc<syntax_pos::SourceFile>, - override_span: Option<Span>) -> Self { - let mut sr = StringReader::new_raw(sess, source_file, override_span); - if sr.advance_token().is_err() { - sr.emit_fatal_errors(); - FatalError.raise(); - } - - sr - } - pub fn new_or_buffered_errs(sess: &'a ParseSess, source_file: Lrc<syntax_pos::SourceFile>, override_span: Option<Span>) -> Result<Self, Vec<Diagnostic>> { @@ -1627,7 +1591,12 @@ mod tests { teststr: String) -> StringReader<'a> { let sf = sm.new_source_file(PathBuf::from(teststr.clone()).into(), teststr); - StringReader::new(sess, sf, None) + let mut sr = StringReader::new_raw(sess, sf, None); + if sr.advance_token().is_err() { + sr.emit_fatal_errors(); + FatalError.raise(); + } + sr } #[test] diff --git a/src/libsyntax/parse/lexer/tokentrees.rs b/src/libsyntax/parse/lexer/tokentrees.rs index 0db36c84cdf..4bfc5bb16c0 100644 --- a/src/libsyntax/parse/lexer/tokentrees.rs +++ b/src/libsyntax/parse/lexer/tokentrees.rs @@ -1,13 +1,46 @@ +use syntax_pos::Span; + use crate::print::pprust::token_to_string; use crate::parse::lexer::{StringReader, UnmatchedBrace}; use crate::parse::{token, PResult}; use crate::tokenstream::{DelimSpan, IsJoint::*, TokenStream, TokenTree, TreeAndJoint}; impl<'a> StringReader<'a> { + crate fn into_token_trees(self) -> (PResult<'a, TokenStream>, Vec<UnmatchedBrace>) { + let mut tt_reader = TokenTreesReader { + string_reader: self, + token: token::Eof, + span: syntax_pos::DUMMY_SP, + open_braces: Vec::new(), + unmatched_braces: Vec::new(), + matching_delim_spans: Vec::new(), + last_unclosed_found_span: None, + }; + let res = tt_reader.parse_all_token_trees(); + (res, tt_reader.unmatched_braces) + } +} + +struct TokenTreesReader<'a> { + string_reader: StringReader<'a>, + token: token::Token, + span: Span, + /// Stack of open delimiters and their spans. Used for error message. + open_braces: Vec<(token::DelimToken, Span)>, + unmatched_braces: Vec<UnmatchedBrace>, + /// The type and spans for all braces + /// + /// Used only for error recovery when arriving to EOF with mismatched braces. + matching_delim_spans: Vec<(token::DelimToken, Span, Span)>, + last_unclosed_found_span: Option<Span>, +} + +impl<'a> TokenTreesReader<'a> { // Parse a stream of tokens into a list of `TokenTree`s, up to an `Eof`. - crate fn parse_all_token_trees(&mut self) -> PResult<'a, TokenStream> { + fn parse_all_token_trees(&mut self) -> PResult<'a, TokenStream> { let mut tts = Vec::new(); + self.real_token(); while self.token != token::Eof { tts.push(self.parse_token_tree()?); } @@ -34,11 +67,12 @@ impl<'a> StringReader<'a> { } fn parse_token_tree(&mut self) -> PResult<'a, TreeAndJoint> { - let sm = self.sess.source_map(); + let sm = self.string_reader.sess.source_map(); match self.token { token::Eof => { let msg = "this file contains an un-closed delimiter"; - let mut err = self.sess.span_diagnostic.struct_span_err(self.span, msg); + let mut err = self.string_reader.sess.span_diagnostic + .struct_span_err(self.span, msg); for &(_, sp) in &self.open_braces { err.span_label(sp, "un-closed delimiter"); } @@ -46,13 +80,12 @@ impl<'a> StringReader<'a> { if let Some((delim, _)) = self.open_braces.last() { if let Some((_, open_sp, close_sp)) = self.matching_delim_spans.iter() .filter(|(d, open_sp, close_sp)| { - - if let Some(close_padding) = sm.span_to_margin(*close_sp) { - if let Some(open_padding) = sm.span_to_margin(*open_sp) { - return delim == d && close_padding != open_padding; + if let Some(close_padding) = sm.span_to_margin(*close_sp) { + if let Some(open_padding) = sm.span_to_margin(*open_sp) { + return delim == d && close_padding != open_padding; + } } - } - false + false }).next() // these are in reverse order as they get inserted on close, but { // we want the last open/first close err.span_label( @@ -164,7 +197,8 @@ impl<'a> StringReader<'a> { // matching opening delimiter). let token_str = token_to_string(&self.token); let msg = format!("unexpected close delimiter: `{}`", token_str); - let mut err = self.sess.span_diagnostic.struct_span_err(self.span, &msg); + let mut err = self.string_reader.sess.span_diagnostic + .struct_span_err(self.span, &msg); err.span_label(self.span, "unexpected close delimiter"); Err(err) }, @@ -173,11 +207,20 @@ impl<'a> StringReader<'a> { // Note that testing for joint-ness here is done via the raw // source span as the joint-ness is a property of the raw source // rather than wanting to take `override_span` into account. - let raw = self.span_src_raw; + // Additionally, we actually check if the *next* pair of tokens + // is joint, but this is equivalent to checking the current pair. + let raw = self.string_reader.peek_span_src_raw; self.real_token(); - let is_joint = raw.hi() == self.span_src_raw.lo() && token::is_op(&self.token); + let is_joint = raw.hi() == self.string_reader.peek_span_src_raw.lo() + && token::is_op(&self.token); Ok((tt, if is_joint { Joint } else { NonJoint })) } } } + + fn real_token(&mut self) { + let t = self.string_reader.real_token(); + self.token = t.tok; + self.span = t.sp; + } } diff --git a/src/libsyntax/parse/literal.rs b/src/libsyntax/parse/literal.rs new file mode 100644 index 00000000000..53195421ddc --- /dev/null +++ b/src/libsyntax/parse/literal.rs @@ -0,0 +1,487 @@ +//! Code related to parsing literals. + +use crate::ast::{self, Ident, Lit, LitKind}; +use crate::parse::parser::Parser; +use crate::parse::PResult; +use crate::parse::token::{self, Token}; +use crate::parse::unescape::{unescape_str, unescape_char, unescape_byte_str, unescape_byte}; +use crate::print::pprust; +use crate::symbol::{keywords, Symbol}; +use crate::tokenstream::{TokenStream, TokenTree}; + +use errors::{Applicability, Handler}; +use log::debug; +use rustc_data_structures::sync::Lrc; +use syntax_pos::Span; + +use std::ascii; + +macro_rules! err { + ($opt_diag:expr, |$span:ident, $diag:ident| $($body:tt)*) => { + match $opt_diag { + Some(($span, $diag)) => { $($body)* } + None => return None, + } + } +} + +impl LitKind { + /// Converts literal token with a suffix into a semantic literal. + /// Works speculatively and may return `None` if diagnostic handler is not passed. + /// If diagnostic handler is passed, always returns `Some`, + /// possibly after reporting non-fatal errors and recovery. + fn from_lit_token( + lit: token::Lit, + suf: Option<Symbol>, + diag: Option<(Span, &Handler)> + ) -> Option<LitKind> { + if suf.is_some() && !lit.may_have_suffix() { + err!(diag, |span, diag| { + expect_no_suffix(span, diag, &format!("a {}", lit.literal_name()), suf) + }); + } + + Some(match lit { + token::Bool(i) => { + assert!(i == keywords::True.name() || i == keywords::False.name()); + LitKind::Bool(i == keywords::True.name()) + } + token::Byte(i) => { + match unescape_byte(&i.as_str()) { + Ok(c) => LitKind::Byte(c), + Err(_) => LitKind::Err(i), + } + }, + token::Char(i) => { + match unescape_char(&i.as_str()) { + Ok(c) => LitKind::Char(c), + Err(_) => LitKind::Err(i), + } + }, + token::Err(i) => LitKind::Err(i), + + // There are some valid suffixes for integer and float literals, + // so all the handling is done internally. + token::Integer(s) => return integer_lit(&s.as_str(), suf, diag), + token::Float(s) => return float_lit(&s.as_str(), suf, diag), + + token::Str_(mut sym) => { + // If there are no characters requiring special treatment we can + // reuse the symbol from the Token. Otherwise, we must generate a + // new symbol because the string in the LitKind is different to the + // string in the Token. + let mut has_error = false; + let s = &sym.as_str(); + if s.as_bytes().iter().any(|&c| c == b'\\' || c == b'\r') { + let mut buf = String::with_capacity(s.len()); + unescape_str(s, &mut |_, unescaped_char| { + match unescaped_char { + Ok(c) => buf.push(c), + Err(_) => has_error = true, + } + }); + if has_error { + return Some(LitKind::Err(sym)); + } + sym = Symbol::intern(&buf) + } + + LitKind::Str(sym, ast::StrStyle::Cooked) + } + token::StrRaw(mut sym, n) => { + // Ditto. + let s = &sym.as_str(); + if s.contains('\r') { + sym = Symbol::intern(&raw_str_lit(s)); + } + LitKind::Str(sym, ast::StrStyle::Raw(n)) + } + token::ByteStr(i) => { + let s = &i.as_str(); + let mut buf = Vec::with_capacity(s.len()); + let mut has_error = false; + unescape_byte_str(s, &mut |_, unescaped_byte| { + match unescaped_byte { + Ok(c) => buf.push(c), + Err(_) => has_error = true, + } + }); + if has_error { + return Some(LitKind::Err(i)); + } + buf.shrink_to_fit(); + LitKind::ByteStr(Lrc::new(buf)) + } + token::ByteStrRaw(i, _) => { + LitKind::ByteStr(Lrc::new(i.to_string().into_bytes())) + } + }) + } + + /// Attempts to recover a token from semantic literal. + /// This function is used when the original token doesn't exist (e.g. the literal is created + /// by an AST-based macro) or unavailable (e.g. from HIR pretty-printing). + pub fn to_lit_token(&self) -> (token::Lit, Option<Symbol>) { + match *self { + LitKind::Str(string, ast::StrStyle::Cooked) => { + let escaped = string.as_str().escape_default().to_string(); + (token::Lit::Str_(Symbol::intern(&escaped)), None) + } + LitKind::Str(string, ast::StrStyle::Raw(n)) => { + (token::Lit::StrRaw(string, n), None) + } + LitKind::ByteStr(ref bytes) => { + let string = bytes.iter().cloned().flat_map(ascii::escape_default) + .map(Into::<char>::into).collect::<String>(); + (token::Lit::ByteStr(Symbol::intern(&string)), None) + } + LitKind::Byte(byte) => { + let string: String = ascii::escape_default(byte).map(Into::<char>::into).collect(); + (token::Lit::Byte(Symbol::intern(&string)), None) + } + LitKind::Char(ch) => { + let string: String = ch.escape_default().map(Into::<char>::into).collect(); + (token::Lit::Char(Symbol::intern(&string)), None) + } + LitKind::Int(n, ty) => { + let suffix = match ty { + ast::LitIntType::Unsigned(ty) => Some(Symbol::intern(ty.ty_to_string())), + ast::LitIntType::Signed(ty) => Some(Symbol::intern(ty.ty_to_string())), + ast::LitIntType::Unsuffixed => None, + }; + (token::Lit::Integer(Symbol::intern(&n.to_string())), suffix) + } + LitKind::Float(symbol, ty) => { + (token::Lit::Float(symbol), Some(Symbol::intern(ty.ty_to_string()))) + } + LitKind::FloatUnsuffixed(symbol) => (token::Lit::Float(symbol), None), + LitKind::Bool(value) => { + let kw = if value { keywords::True } else { keywords::False }; + (token::Lit::Bool(kw.name()), None) + } + LitKind::Err(val) => (token::Lit::Err(val), None), + } + } +} + +impl Lit { + /// Converts literal token with a suffix into an AST literal. + /// Works speculatively and may return `None` if diagnostic handler is not passed. + /// If diagnostic handler is passed, may return `Some`, + /// possibly after reporting non-fatal errors and recovery, or `None` for irrecoverable errors. + crate fn from_token( + token: &token::Token, + span: Span, + diag: Option<(Span, &Handler)>, + ) -> Option<Lit> { + let (token, suffix) = match *token { + token::Ident(ident, false) if ident.name == keywords::True.name() || + ident.name == keywords::False.name() => + (token::Bool(ident.name), None), + token::Literal(token, suffix) => + (token, suffix), + token::Interpolated(ref nt) => { + if let token::NtExpr(expr) | token::NtLiteral(expr) = &**nt { + if let ast::ExprKind::Lit(lit) = &expr.node { + return Some(lit.clone()); + } + } + return None; + } + _ => return None, + }; + + let node = LitKind::from_lit_token(token, suffix, diag)?; + Some(Lit { node, token, suffix, span }) + } + + /// Attempts to recover an AST literal from semantic literal. + /// This function is used when the original token doesn't exist (e.g. the literal is created + /// by an AST-based macro) or unavailable (e.g. from HIR pretty-printing). + pub fn from_lit_kind(node: LitKind, span: Span) -> Lit { + let (token, suffix) = node.to_lit_token(); + Lit { node, token, suffix, span } + } + + /// Losslessly convert an AST literal into a token stream. + crate fn tokens(&self) -> TokenStream { + let token = match self.token { + token::Bool(symbol) => Token::Ident(Ident::with_empty_ctxt(symbol), false), + token => Token::Literal(token, self.suffix), + }; + TokenTree::Token(self.span, token).into() + } +} + +impl<'a> Parser<'a> { + /// Matches `lit = true | false | token_lit`. + crate fn parse_lit(&mut self) -> PResult<'a, Lit> { + let diag = Some((self.span, &self.sess.span_diagnostic)); + if let Some(lit) = Lit::from_token(&self.token, self.span, diag) { + self.bump(); + return Ok(lit); + } else if self.token == token::Dot { + // Recover `.4` as `0.4`. + let recovered = self.look_ahead(1, |t| { + if let token::Literal(token::Integer(val), suf) = *t { + let next_span = self.look_ahead_span(1); + if self.span.hi() == next_span.lo() { + let sym = String::from("0.") + &val.as_str(); + let token = token::Literal(token::Float(Symbol::intern(&sym)), suf); + return Some((token, self.span.to(next_span))); + } + } + None + }); + if let Some((token, span)) = recovered { + self.diagnostic() + .struct_span_err(span, "float literals must have an integer part") + .span_suggestion( + span, + "must have an integer part", + pprust::token_to_string(&token), + Applicability::MachineApplicable, + ) + .emit(); + let diag = Some((span, &self.sess.span_diagnostic)); + if let Some(lit) = Lit::from_token(&token, span, diag) { + self.bump(); + self.bump(); + return Ok(lit); + } + } + } + + Err(self.span_fatal(self.span, &format!("unexpected token: {}", self.this_token_descr()))) + } +} + +crate fn expect_no_suffix(sp: Span, diag: &Handler, kind: &str, suffix: Option<ast::Name>) { + match suffix { + None => {/* everything ok */} + Some(suf) => { + let text = suf.as_str(); + if text.is_empty() { + diag.span_bug(sp, "found empty literal suffix in Some") + } + let mut err = if kind == "a tuple index" && + ["i32", "u32", "isize", "usize"].contains(&text.to_string().as_str()) + { + // #59553: warn instead of reject out of hand to allow the fix to percolate + // through the ecosystem when people fix their macros + let mut err = diag.struct_span_warn( + sp, + &format!("suffixes on {} are invalid", kind), + ); + err.note(&format!( + "`{}` is *temporarily* accepted on tuple index fields as it was \ + incorrectly accepted on stable for a few releases", + text, + )); + err.help( + "on proc macros, you'll want to use `syn::Index::from` or \ + `proc_macro::Literal::*_unsuffixed` for code that will desugar \ + to tuple field access", + ); + err.note( + "for more context, see https://github.com/rust-lang/rust/issues/60210", + ); + err + } else { + diag.struct_span_err(sp, &format!("suffixes on {} are invalid", kind)) + }; + err.span_label(sp, format!("invalid suffix `{}`", text)); + err.emit(); + } + } +} + +/// Parses a string representing a raw string literal into its final form. The +/// only operation this does is convert embedded CRLF into a single LF. +fn raw_str_lit(lit: &str) -> String { + debug!("raw_str_lit: given {}", lit.escape_default()); + let mut res = String::with_capacity(lit.len()); + + let mut chars = lit.chars().peekable(); + while let Some(c) = chars.next() { + if c == '\r' { + if *chars.peek().unwrap() != '\n' { + panic!("lexer accepted bare CR"); + } + chars.next(); + res.push('\n'); + } else { + res.push(c); + } + } + + res.shrink_to_fit(); + res +} + +// check if `s` looks like i32 or u1234 etc. +fn looks_like_width_suffix(first_chars: &[char], s: &str) -> bool { + s.starts_with(first_chars) && s[1..].chars().all(|c| c.is_ascii_digit()) +} + +fn filtered_float_lit(data: Symbol, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>) + -> Option<LitKind> { + debug!("filtered_float_lit: {}, {:?}", data, suffix); + let suffix = match suffix { + Some(suffix) => suffix, + None => return Some(LitKind::FloatUnsuffixed(data)), + }; + + Some(match &*suffix.as_str() { + "f32" => LitKind::Float(data, ast::FloatTy::F32), + "f64" => LitKind::Float(data, ast::FloatTy::F64), + suf => { + err!(diag, |span, diag| { + if suf.len() >= 2 && looks_like_width_suffix(&['f'], suf) { + // if it looks like a width, lets try to be helpful. + let msg = format!("invalid width `{}` for float literal", &suf[1..]); + diag.struct_span_err(span, &msg).help("valid widths are 32 and 64").emit() + } else { + let msg = format!("invalid suffix `{}` for float literal", suf); + diag.struct_span_err(span, &msg) + .span_label(span, format!("invalid suffix `{}`", suf)) + .help("valid suffixes are `f32` and `f64`") + .emit(); + } + }); + + LitKind::FloatUnsuffixed(data) + } + }) +} +fn float_lit(s: &str, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>) + -> Option<LitKind> { + debug!("float_lit: {:?}, {:?}", s, suffix); + // FIXME #2252: bounds checking float literals is deferred until trans + + // Strip underscores without allocating a new String unless necessary. + let s2; + let s = if s.chars().any(|c| c == '_') { + s2 = s.chars().filter(|&c| c != '_').collect::<String>(); + &s2 + } else { + s + }; + + filtered_float_lit(Symbol::intern(s), suffix, diag) +} + +fn integer_lit(s: &str, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>) + -> Option<LitKind> { + // s can only be ascii, byte indexing is fine + + // Strip underscores without allocating a new String unless necessary. + let s2; + let mut s = if s.chars().any(|c| c == '_') { + s2 = s.chars().filter(|&c| c != '_').collect::<String>(); + &s2 + } else { + s + }; + + debug!("integer_lit: {}, {:?}", s, suffix); + + let mut base = 10; + let orig = s; + let mut ty = ast::LitIntType::Unsuffixed; + + if s.starts_with('0') && s.len() > 1 { + match s.as_bytes()[1] { + b'x' => base = 16, + b'o' => base = 8, + b'b' => base = 2, + _ => { } + } + } + + // 1f64 and 2f32 etc. are valid float literals. + if let Some(suf) = suffix { + if looks_like_width_suffix(&['f'], &suf.as_str()) { + let err = match base { + 16 => Some("hexadecimal float literal is not supported"), + 8 => Some("octal float literal is not supported"), + 2 => Some("binary float literal is not supported"), + _ => None, + }; + if let Some(err) = err { + err!(diag, |span, diag| { + diag.struct_span_err(span, err) + .span_label(span, "not supported") + .emit(); + }); + } + return filtered_float_lit(Symbol::intern(s), Some(suf), diag) + } + } + + if base != 10 { + s = &s[2..]; + } + + if let Some(suf) = suffix { + if suf.as_str().is_empty() { + err!(diag, |span, diag| diag.span_bug(span, "found empty literal suffix in Some")); + } + ty = match &*suf.as_str() { + "isize" => ast::LitIntType::Signed(ast::IntTy::Isize), + "i8" => ast::LitIntType::Signed(ast::IntTy::I8), + "i16" => ast::LitIntType::Signed(ast::IntTy::I16), + "i32" => ast::LitIntType::Signed(ast::IntTy::I32), + "i64" => ast::LitIntType::Signed(ast::IntTy::I64), + "i128" => ast::LitIntType::Signed(ast::IntTy::I128), + "usize" => ast::LitIntType::Unsigned(ast::UintTy::Usize), + "u8" => ast::LitIntType::Unsigned(ast::UintTy::U8), + "u16" => ast::LitIntType::Unsigned(ast::UintTy::U16), + "u32" => ast::LitIntType::Unsigned(ast::UintTy::U32), + "u64" => ast::LitIntType::Unsigned(ast::UintTy::U64), + "u128" => ast::LitIntType::Unsigned(ast::UintTy::U128), + suf => { + // i<digits> and u<digits> look like widths, so lets + // give an error message along those lines + err!(diag, |span, diag| { + if looks_like_width_suffix(&['i', 'u'], suf) { + let msg = format!("invalid width `{}` for integer literal", &suf[1..]); + diag.struct_span_err(span, &msg) + .help("valid widths are 8, 16, 32, 64 and 128") + .emit(); + } else { + let msg = format!("invalid suffix `{}` for numeric literal", suf); + diag.struct_span_err(span, &msg) + .span_label(span, format!("invalid suffix `{}`", suf)) + .help("the suffix must be one of the integral types \ + (`u32`, `isize`, etc)") + .emit(); + } + }); + + ty + } + } + } + + debug!("integer_lit: the type is {:?}, base {:?}, the new string is {:?}, the original \ + string was {:?}, the original suffix was {:?}", ty, base, s, orig, suffix); + + Some(match u128::from_str_radix(s, base) { + Ok(r) => LitKind::Int(r, ty), + Err(_) => { + // small bases are lexed as if they were base 10, e.g, the string + // might be `0b10201`. This will cause the conversion above to fail, + // but these cases have errors in the lexer: we don't want to emit + // two errors, and we especially don't want to emit this error since + // it isn't necessarily true. + let already_errored = base < 10 && + s.chars().any(|c| c.to_digit(10).map_or(false, |d| d >= base)); + + if !already_errored { + err!(diag, |span, diag| diag.span_err(span, "int literal is too large")); + } + LitKind::Int(0, ty) + } + }) +} diff --git a/src/libsyntax/parse/mod.rs b/src/libsyntax/parse/mod.rs index be44b964ba5..0611c1d9b42 100644 --- a/src/libsyntax/parse/mod.rs +++ b/src/libsyntax/parse/mod.rs @@ -5,7 +5,6 @@ use crate::early_buffered_lints::{BufferedEarlyLint, BufferedEarlyLintId}; use crate::source_map::{SourceMap, FilePathMapping}; use crate::feature_gate::UnstableFeatures; use crate::parse::parser::Parser; -use crate::symbol::Symbol; use crate::syntax::parse::parser::emit_unclosed_delims; use crate::tokenstream::{TokenStream, TokenTree}; use crate::diagnostics::plugin::ErrorMap; @@ -14,7 +13,6 @@ use crate::print::pprust::token_to_string; use errors::{Applicability, FatalError, Level, Handler, ColorConfig, Diagnostic, DiagnosticBuilder}; use rustc_data_structures::sync::{Lrc, Lock}; use syntax_pos::{Span, SourceFile, FileName, MultiSpan}; -use log::debug; use rustc_data_structures::fx::{FxHashSet, FxHashMap}; use std::borrow::Cow; @@ -25,18 +23,15 @@ pub type PResult<'a, T> = Result<T, DiagnosticBuilder<'a>>; #[macro_use] pub mod parser; - +pub mod attr; pub mod lexer; pub mod token; -pub mod attr; -pub mod diagnostics; - -pub mod classify; - -pub(crate) mod unescape; -use unescape::{unescape_str, unescape_char, unescape_byte_str, unescape_byte}; -pub(crate) mod unescape_error_reporting; +crate mod classify; +crate mod diagnostics; +crate mod literal; +crate mod unescape; +crate mod unescape_error_reporting; /// Info about a parsing session. pub struct ParseSess { @@ -295,22 +290,22 @@ pub fn source_file_to_stream( } /// Given a source file, produces a sequence of token trees. Returns any buffered errors from -/// parsing the token tream. +/// parsing the token stream. pub fn maybe_file_to_stream( sess: &ParseSess, source_file: Lrc<SourceFile>, override_span: Option<Span>, ) -> Result<(TokenStream, Vec<lexer::UnmatchedBrace>), Vec<Diagnostic>> { - let mut srdr = lexer::StringReader::new_or_buffered_errs(sess, source_file, override_span)?; - srdr.real_token(); + let srdr = lexer::StringReader::new_or_buffered_errs(sess, source_file, override_span)?; + let (token_trees, unmatched_braces) = srdr.into_token_trees(); - match srdr.parse_all_token_trees() { - Ok(stream) => Ok((stream, srdr.unmatched_braces)), + match token_trees { + Ok(stream) => Ok((stream, unmatched_braces)), Err(err) => { let mut buffer = Vec::with_capacity(1); err.buffer(&mut buffer); // Not using `emit_unclosed_delims` to use `db.buffer` - for unmatched in srdr.unmatched_braces { + for unmatched in unmatched_braces { let mut db = sess.span_diagnostic.struct_span_err(unmatched.found_span, &format!( "incorrect close delimiter: `{}`", token_to_string(&token::Token::CloseDelim(unmatched.found_delim)), @@ -334,284 +329,6 @@ pub fn stream_to_parser(sess: &ParseSess, stream: TokenStream) -> Parser<'_> { Parser::new(sess, stream, None, true, false) } -/// Parses a string representing a raw string literal into its final form. The -/// only operation this does is convert embedded CRLF into a single LF. -fn raw_str_lit(lit: &str) -> String { - debug!("raw_str_lit: given {}", lit.escape_default()); - let mut res = String::with_capacity(lit.len()); - - let mut chars = lit.chars().peekable(); - while let Some(c) = chars.next() { - if c == '\r' { - if *chars.peek().unwrap() != '\n' { - panic!("lexer accepted bare CR"); - } - chars.next(); - res.push('\n'); - } else { - res.push(c); - } - } - - res.shrink_to_fit(); - res -} - -// check if `s` looks like i32 or u1234 etc. -fn looks_like_width_suffix(first_chars: &[char], s: &str) -> bool { - s.starts_with(first_chars) && s[1..].chars().all(|c| c.is_ascii_digit()) -} - -macro_rules! err { - ($opt_diag:expr, |$span:ident, $diag:ident| $($body:tt)*) => { - match $opt_diag { - Some(($span, $diag)) => { $($body)* } - None => return None, - } - } -} - -crate fn lit_token(lit: token::Lit, suf: Option<Symbol>, diag: Option<(Span, &Handler)>) - -> (bool /* suffix illegal? */, Option<ast::LitKind>) { - use ast::LitKind; - - match lit { - token::Byte(i) => { - let lit_kind = match unescape_byte(&i.as_str()) { - Ok(c) => LitKind::Byte(c), - Err(_) => LitKind::Err(i), - }; - (true, Some(lit_kind)) - }, - token::Char(i) => { - let lit_kind = match unescape_char(&i.as_str()) { - Ok(c) => LitKind::Char(c), - Err(_) => LitKind::Err(i), - }; - (true, Some(lit_kind)) - }, - token::Err(i) => (true, Some(LitKind::Err(i))), - - // There are some valid suffixes for integer and float literals, - // so all the handling is done internally. - token::Integer(s) => (false, integer_lit(&s.as_str(), suf, diag)), - token::Float(s) => (false, float_lit(&s.as_str(), suf, diag)), - - token::Str_(mut sym) => { - // If there are no characters requiring special treatment we can - // reuse the symbol from the Token. Otherwise, we must generate a - // new symbol because the string in the LitKind is different to the - // string in the Token. - let mut has_error = false; - let s = &sym.as_str(); - if s.as_bytes().iter().any(|&c| c == b'\\' || c == b'\r') { - let mut buf = String::with_capacity(s.len()); - unescape_str(s, &mut |_, unescaped_char| { - match unescaped_char { - Ok(c) => buf.push(c), - Err(_) => has_error = true, - } - }); - if has_error { - return (true, Some(LitKind::Err(sym))); - } - sym = Symbol::intern(&buf) - } - - (true, Some(LitKind::Str(sym, ast::StrStyle::Cooked))) - } - token::StrRaw(mut sym, n) => { - // Ditto. - let s = &sym.as_str(); - if s.contains('\r') { - sym = Symbol::intern(&raw_str_lit(s)); - } - (true, Some(LitKind::Str(sym, ast::StrStyle::Raw(n)))) - } - token::ByteStr(i) => { - let s = &i.as_str(); - let mut buf = Vec::with_capacity(s.len()); - let mut has_error = false; - unescape_byte_str(s, &mut |_, unescaped_byte| { - match unescaped_byte { - Ok(c) => buf.push(c), - Err(_) => has_error = true, - } - }); - if has_error { - return (true, Some(LitKind::Err(i))); - } - buf.shrink_to_fit(); - (true, Some(LitKind::ByteStr(Lrc::new(buf)))) - } - token::ByteStrRaw(i, _) => { - (true, Some(LitKind::ByteStr(Lrc::new(i.to_string().into_bytes())))) - } - } -} - -fn filtered_float_lit(data: Symbol, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>) - -> Option<ast::LitKind> { - debug!("filtered_float_lit: {}, {:?}", data, suffix); - let suffix = match suffix { - Some(suffix) => suffix, - None => return Some(ast::LitKind::FloatUnsuffixed(data)), - }; - - Some(match &*suffix.as_str() { - "f32" => ast::LitKind::Float(data, ast::FloatTy::F32), - "f64" => ast::LitKind::Float(data, ast::FloatTy::F64), - suf => { - err!(diag, |span, diag| { - if suf.len() >= 2 && looks_like_width_suffix(&['f'], suf) { - // if it looks like a width, lets try to be helpful. - let msg = format!("invalid width `{}` for float literal", &suf[1..]); - diag.struct_span_err(span, &msg).help("valid widths are 32 and 64").emit() - } else { - let msg = format!("invalid suffix `{}` for float literal", suf); - diag.struct_span_err(span, &msg) - .span_label(span, format!("invalid suffix `{}`", suf)) - .help("valid suffixes are `f32` and `f64`") - .emit(); - } - }); - - ast::LitKind::FloatUnsuffixed(data) - } - }) -} -fn float_lit(s: &str, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>) - -> Option<ast::LitKind> { - debug!("float_lit: {:?}, {:?}", s, suffix); - // FIXME #2252: bounds checking float literals is deferred until trans - - // Strip underscores without allocating a new String unless necessary. - let s2; - let s = if s.chars().any(|c| c == '_') { - s2 = s.chars().filter(|&c| c != '_').collect::<String>(); - &s2 - } else { - s - }; - - filtered_float_lit(Symbol::intern(s), suffix, diag) -} - -fn integer_lit(s: &str, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>) - -> Option<ast::LitKind> { - // s can only be ascii, byte indexing is fine - - // Strip underscores without allocating a new String unless necessary. - let s2; - let mut s = if s.chars().any(|c| c == '_') { - s2 = s.chars().filter(|&c| c != '_').collect::<String>(); - &s2 - } else { - s - }; - - debug!("integer_lit: {}, {:?}", s, suffix); - - let mut base = 10; - let orig = s; - let mut ty = ast::LitIntType::Unsuffixed; - - if s.starts_with('0') && s.len() > 1 { - match s.as_bytes()[1] { - b'x' => base = 16, - b'o' => base = 8, - b'b' => base = 2, - _ => { } - } - } - - // 1f64 and 2f32 etc. are valid float literals. - if let Some(suf) = suffix { - if looks_like_width_suffix(&['f'], &suf.as_str()) { - let err = match base { - 16 => Some("hexadecimal float literal is not supported"), - 8 => Some("octal float literal is not supported"), - 2 => Some("binary float literal is not supported"), - _ => None, - }; - if let Some(err) = err { - err!(diag, |span, diag| { - diag.struct_span_err(span, err) - .span_label(span, "not supported") - .emit(); - }); - } - return filtered_float_lit(Symbol::intern(s), Some(suf), diag) - } - } - - if base != 10 { - s = &s[2..]; - } - - if let Some(suf) = suffix { - if suf.as_str().is_empty() { - err!(diag, |span, diag| diag.span_bug(span, "found empty literal suffix in Some")); - } - ty = match &*suf.as_str() { - "isize" => ast::LitIntType::Signed(ast::IntTy::Isize), - "i8" => ast::LitIntType::Signed(ast::IntTy::I8), - "i16" => ast::LitIntType::Signed(ast::IntTy::I16), - "i32" => ast::LitIntType::Signed(ast::IntTy::I32), - "i64" => ast::LitIntType::Signed(ast::IntTy::I64), - "i128" => ast::LitIntType::Signed(ast::IntTy::I128), - "usize" => ast::LitIntType::Unsigned(ast::UintTy::Usize), - "u8" => ast::LitIntType::Unsigned(ast::UintTy::U8), - "u16" => ast::LitIntType::Unsigned(ast::UintTy::U16), - "u32" => ast::LitIntType::Unsigned(ast::UintTy::U32), - "u64" => ast::LitIntType::Unsigned(ast::UintTy::U64), - "u128" => ast::LitIntType::Unsigned(ast::UintTy::U128), - suf => { - // i<digits> and u<digits> look like widths, so lets - // give an error message along those lines - err!(diag, |span, diag| { - if looks_like_width_suffix(&['i', 'u'], suf) { - let msg = format!("invalid width `{}` for integer literal", &suf[1..]); - diag.struct_span_err(span, &msg) - .help("valid widths are 8, 16, 32, 64 and 128") - .emit(); - } else { - let msg = format!("invalid suffix `{}` for numeric literal", suf); - diag.struct_span_err(span, &msg) - .span_label(span, format!("invalid suffix `{}`", suf)) - .help("the suffix must be one of the integral types \ - (`u32`, `isize`, etc)") - .emit(); - } - }); - - ty - } - } - } - - debug!("integer_lit: the type is {:?}, base {:?}, the new string is {:?}, the original \ - string was {:?}, the original suffix was {:?}", ty, base, s, orig, suffix); - - Some(match u128::from_str_radix(s, base) { - Ok(r) => ast::LitKind::Int(r, ty), - Err(_) => { - // small bases are lexed as if they were base 10, e.g, the string - // might be `0b10201`. This will cause the conversion above to fail, - // but these cases have errors in the lexer: we don't want to emit - // two errors, and we especially don't want to emit this error since - // it isn't necessarily true. - let already_errored = base < 10 && - s.chars().any(|c| c.to_digit(10).map_or(false, |d| d >= base)); - - if !already_errored { - err!(diag, |span, diag| diag.span_err(span, "int literal is too large")); - } - ast::LitKind::Int(0, ty) - } - }) -} - /// A sequence separator. pub struct SeqSep { /// The seperator token. @@ -674,6 +391,8 @@ mod tests { #[test] fn string_to_tts_macro () { with_globals(|| { + use crate::symbol::sym; + let tts: Vec<_> = string_to_stream("macro_rules! zip (($a)=>($a))".to_string()).trees().collect(); let tts: &[TokenTree] = &tts[..]; @@ -686,8 +405,8 @@ mod tests { Some(&TokenTree::Token(_, token::Ident(name_zip, false))), Some(&TokenTree::Delimited(_, macro_delim, ref macro_tts)), ) - if name_macro_rules.name == "macro_rules" - && name_zip.name == "zip" => { + if name_macro_rules.name == sym::macro_rules + && name_zip.name.as_str() == "zip" => { let tts = ¯o_tts.trees().collect::<Vec<_>>(); match (tts.len(), tts.get(0), tts.get(1), tts.get(2)) { ( @@ -704,7 +423,7 @@ mod tests { Some(&TokenTree::Token(_, token::Dollar)), Some(&TokenTree::Token(_, token::Ident(ident, false))), ) - if first_delim == token::Paren && ident.name == "a" => {}, + if first_delim == token::Paren && ident.name.as_str() == "a" => {}, _ => panic!("value 3: {:?} {:?}", first_delim, first_tts), } let tts = &second_tts.trees().collect::<Vec<_>>(); @@ -714,7 +433,7 @@ mod tests { Some(&TokenTree::Token(_, token::Dollar)), Some(&TokenTree::Token(_, token::Ident(ident, false))), ) - if second_delim == token::Paren && ident.name == "a" => {}, + if second_delim == token::Paren && ident.name.as_str() == "a" => {}, _ => panic!("value 4: {:?} {:?}", second_delim, second_tts), } }, @@ -858,20 +577,22 @@ mod tests { #[test] fn crlf_doc_comments() { with_globals(|| { + use crate::symbol::sym; + let sess = ParseSess::new(FilePathMapping::empty()); let name_1 = FileName::Custom("crlf_source_1".to_string()); let source = "/// doc comment\r\nfn foo() {}".to_string(); let item = parse_item_from_source_str(name_1, source, &sess) .unwrap().unwrap(); - let doc = first_attr_value_str_by_name(&item.attrs, "doc").unwrap(); - assert_eq!(doc, "/// doc comment"); + let doc = first_attr_value_str_by_name(&item.attrs, sym::doc).unwrap(); + assert_eq!(doc.as_str(), "/// doc comment"); let name_2 = FileName::Custom("crlf_source_2".to_string()); let source = "/// doc comment\r\n/// line 2\r\nfn foo() {}".to_string(); let item = parse_item_from_source_str(name_2, source, &sess) .unwrap().unwrap(); - let docs = item.attrs.iter().filter(|a| a.path == "doc") + let docs = item.attrs.iter().filter(|a| a.path == sym::doc) .map(|a| a.value_str().unwrap().to_string()).collect::<Vec<_>>(); let b: &[_] = &["/// doc comment".to_string(), "/// line 2".to_string()]; assert_eq!(&docs[..], b); @@ -879,8 +600,8 @@ mod tests { let name_3 = FileName::Custom("clrf_source_3".to_string()); let source = "/** doc comment\r\n * with CRLF */\r\nfn foo() {}".to_string(); let item = parse_item_from_source_str(name_3, source, &sess).unwrap().unwrap(); - let doc = first_attr_value_str_by_name(&item.attrs, "doc").unwrap(); - assert_eq!(doc, "/** doc comment\n * with CRLF */"); + let doc = first_attr_value_str_by_name(&item.attrs, sym::doc).unwrap(); + assert_eq!(doc.as_str(), "/** doc comment\n * with CRLF */"); }); } diff --git a/src/libsyntax/parse/parser.rs b/src/libsyntax/parse/parser.rs index d97d1e2f0f4..24d120376de 100644 --- a/src/libsyntax/parse/parser.rs +++ b/src/libsyntax/parse/parser.rs @@ -15,7 +15,7 @@ use crate::ast::{ForeignItem, ForeignItemKind, FunctionRetTy}; use crate::ast::{GenericParam, GenericParamKind}; use crate::ast::GenericArg; use crate::ast::{Ident, ImplItem, IsAsync, IsAuto, Item, ItemKind}; -use crate::ast::{Label, Lifetime, Lit, LitKind}; +use crate::ast::{Label, Lifetime}; use crate::ast::{Local, LocalSource}; use crate::ast::MacStmtStyle; use crate::ast::{Mac, Mac_, MacDelimiter}; @@ -35,7 +35,7 @@ use crate::ast::{RangeEnd, RangeSyntax}; use crate::{ast, attr}; use crate::ext::base::DummyResult; use crate::source_map::{self, SourceMap, Spanned, respan}; -use crate::parse::{self, SeqSep, classify, token}; +use crate::parse::{SeqSep, classify, literal, token}; use crate::parse::lexer::{TokenAndSpan, UnmatchedBrace}; use crate::parse::lexer::comments::{doc_comment_style, strip_doc_comment_decoration}; use crate::parse::token::DelimToken; @@ -46,7 +46,7 @@ use crate::ptr::P; use crate::parse::PResult; use crate::ThinVec; use crate::tokenstream::{self, DelimSpan, TokenTree, TokenStream, TreeAndJoint}; -use crate::symbol::{Symbol, keywords}; +use crate::symbol::{keywords, sym, Symbol}; use errors::{Applicability, DiagnosticBuilder, DiagnosticId, FatalError}; use rustc_target::spec::abi::{self, Abi}; @@ -104,14 +104,14 @@ pub enum PathStyle { } #[derive(Clone, Copy, PartialEq, Debug)] -enum SemiColonMode { +crate enum SemiColonMode { Break, Ignore, Comma, } #[derive(Clone, Copy, PartialEq, Debug)] -enum BlockMode { +crate enum BlockMode { Break, Ignore, } @@ -352,7 +352,7 @@ impl TokenCursor { let body = TokenTree::Delimited( delim_span, token::Bracket, - [TokenTree::Token(sp, token::Ident(ast::Ident::from_str("doc"), false)), + [TokenTree::Token(sp, token::Ident(ast::Ident::with_empty_ctxt(sym::doc), false)), TokenTree::Token(sp, token::Eq), TokenTree::Token(sp, token::Literal( token::StrRaw(Symbol::intern(&stripped), num_of_hashes), None)) @@ -389,7 +389,7 @@ crate enum TokenType { } impl TokenType { - fn to_string(&self) -> String { + crate fn to_string(&self) -> String { match *self { TokenType::Token(ref t) => format!("`{}`", pprust::token_to_string(t)), TokenType::Keyword(kw) => format!("`{}`", kw.name()), @@ -613,7 +613,7 @@ impl<'a> Parser<'a> { }) } - fn this_token_descr(&self) -> String { + crate fn this_token_descr(&self) -> String { if let Some(prefix) = self.token_descr() { format!("{} `{}`", prefix, self.this_token_to_string()) } else { @@ -621,11 +621,6 @@ impl<'a> Parser<'a> { } } - fn unexpected_last<T>(&self, t: &token::Token) -> PResult<'a, T> { - let token_str = pprust::token_to_string(t); - Err(self.span_fatal(self.prev_span, &format!("unexpected token: `{}`", token_str))) - } - crate fn unexpected<T>(&mut self) -> PResult<'a, T> { match self.expect_one_of(&[], &[]) { Err(e) => Err(e), @@ -678,56 +673,6 @@ impl<'a> Parser<'a> { } } - fn recover_closing_delimiter( - &mut self, - tokens: &[token::Token], - mut err: DiagnosticBuilder<'a>, - ) -> PResult<'a, bool> { - let mut pos = None; - // we want to use the last closing delim that would apply - for (i, unmatched) in self.unclosed_delims.iter().enumerate().rev() { - if tokens.contains(&token::CloseDelim(unmatched.expected_delim)) - && Some(self.span) > unmatched.unclosed_span - { - pos = Some(i); - } - } - match pos { - Some(pos) => { - // Recover and assume that the detected unclosed delimiter was meant for - // this location. Emit the diagnostic and act as if the delimiter was - // present for the parser's sake. - - // Don't attempt to recover from this unclosed delimiter more than once. - let unmatched = self.unclosed_delims.remove(pos); - let delim = TokenType::Token(token::CloseDelim(unmatched.expected_delim)); - - // We want to suggest the inclusion of the closing delimiter where it makes - // the most sense, which is immediately after the last token: - // - // {foo(bar {}} - // - ^ - // | | - // | help: `)` may belong here (FIXME: #58270) - // | - // unclosed delimiter - if let Some(sp) = unmatched.unclosed_span { - err.span_label(sp, "unclosed delimiter"); - } - err.span_suggestion_short( - self.sess.source_map().next_point(self.prev_span), - &format!("{} may belong here", delim.to_string()), - delim.to_string(), - Applicability::MaybeIncorrect, - ); - err.emit(); - self.expected_tokens.clear(); // reduce errors - Ok(true) - } - _ => Err(err), - } - } - /// Expect next token to be edible or inedible token. If edible, /// then consume it; if inedible, then return without consuming /// anything. Signal a fatal error if next token is unexpected. @@ -1109,43 +1054,7 @@ impl<'a> Parser<'a> { } fn expect_no_suffix(&self, sp: Span, kind: &str, suffix: Option<ast::Name>) { - match suffix { - None => {/* everything ok */} - Some(suf) => { - let text = suf.as_str(); - if text.is_empty() { - self.span_bug(sp, "found empty literal suffix in Some") - } - let mut err = if kind == "a tuple index" && - ["i32", "u32", "isize", "usize"].contains(&text.to_string().as_str()) - { - // #59553: warn instead of reject out of hand to allow the fix to percolate - // through the ecosystem when people fix their macros - let mut err = self.struct_span_warn( - sp, - &format!("suffixes on {} are invalid", kind), - ); - err.note(&format!( - "`{}` is *temporarily* accepted on tuple index fields as it was \ - incorrectly accepted on stable for a few releases", - text, - )); - err.help( - "on proc macros, you'll want to use `syn::Index::from` or \ - `proc_macro::Literal::*_unsuffixed` for code that will desugar \ - to tuple field access", - ); - err.note( - "for more context, see https://github.com/rust-lang/rust/issues/60210", - ); - err - } else { - self.struct_span_err(sp, &format!("suffixes on {} are invalid", kind)) - }; - err.span_label(sp, format!("invalid suffix `{}`", text)); - err.emit(); - } - } + literal::expect_no_suffix(sp, &self.sess.span_diagnostic, kind, suffix) } /// Attempts to consume a `<`. If `<<` is seen, replaces it with a single @@ -1423,7 +1332,7 @@ impl<'a> Parser<'a> { }) } - fn look_ahead_span(&self, dist: usize) -> Span { + crate fn look_ahead_span(&self, dist: usize) -> Span { if dist == 0 { return self.span } @@ -1452,9 +1361,6 @@ impl<'a> Parser<'a> { crate fn struct_span_err<S: Into<MultiSpan>>(&self, sp: S, m: &str) -> DiagnosticBuilder<'a> { self.sess.span_diagnostic.struct_span_err(sp, m) } - fn struct_span_warn<S: Into<MultiSpan>>(&self, sp: S, m: &str) -> DiagnosticBuilder<'a> { - self.sess.span_diagnostic.struct_span_warn(sp, m) - } crate fn span_bug<S: Into<MultiSpan>>(&self, sp: S, m: &str) -> ! { self.sess.span_diagnostic.span_bug(sp, m) } @@ -1877,7 +1783,7 @@ impl<'a> Parser<'a> { Ok(MutTy { ty: t, mutbl: mutbl }) } - fn is_named_argument(&mut self) -> bool { + fn is_named_argument(&self) -> bool { let offset = match self.token { token::Interpolated(ref nt) => match **nt { token::NtPat(..) => return self.look_ahead(1, |t| t == &token::Colon), @@ -1925,8 +1831,6 @@ impl<'a> Parser<'a> { /// This version of parse arg doesn't necessarily require identifier names. fn parse_arg_general(&mut self, require_name: bool, is_trait_item: bool, allow_c_variadic: bool) -> PResult<'a, Arg> { - maybe_whole!(self, NtArg, |x| x); - if let Ok(Some(_)) = self.parse_self_arg() { let mut err = self.struct_span_err(self.prev_span, "unexpected `self` argument in function"); @@ -2069,88 +1973,6 @@ impl<'a> Parser<'a> { } } - /// Matches `token_lit = LIT_INTEGER | ...`. - fn parse_lit_token(&mut self) -> PResult<'a, LitKind> { - let out = match self.token { - token::Interpolated(ref nt) => match **nt { - token::NtExpr(ref v) | token::NtLiteral(ref v) => match v.node { - ExprKind::Lit(ref lit) => { lit.node.clone() } - _ => { return self.unexpected_last(&self.token); } - }, - _ => { return self.unexpected_last(&self.token); } - }, - token::Literal(lit, suf) => { - let diag = Some((self.span, &self.sess.span_diagnostic)); - let (suffix_illegal, result) = parse::lit_token(lit, suf, diag); - - if suffix_illegal { - let sp = self.span; - self.expect_no_suffix(sp, &format!("a {}", lit.literal_name()), suf) - } - - result.unwrap() - } - token::Dot if self.look_ahead(1, |t| match t { - token::Literal(parse::token::Lit::Integer(_) , _) => true, - _ => false, - }) => { // recover from `let x = .4;` - let lo = self.span; - self.bump(); - if let token::Literal( - parse::token::Lit::Integer(val), - suffix, - ) = self.token { - let suffix = suffix.and_then(|s| { - let s = s.as_str(); - if s == "f32" { - Some("f32") - } else if s == "f64" { - Some("f64") - } else { - None - } - }).unwrap_or(""); - self.bump(); - let sp = lo.to(self.prev_span); - let mut err = self.diagnostic() - .struct_span_err(sp, "float literals must have an integer part"); - err.span_suggestion( - sp, - "must have an integer part", - format!("0.{}{}", val, suffix), - Applicability::MachineApplicable, - ); - err.emit(); - return Ok(match suffix { - "f32" => ast::LitKind::Float(val, ast::FloatTy::F32), - "f64" => ast::LitKind::Float(val, ast::FloatTy::F64), - _ => ast::LitKind::FloatUnsuffixed(val), - }); - } else { - unreachable!(); - }; - } - _ => { return self.unexpected_last(&self.token); } - }; - - self.bump(); - Ok(out) - } - - /// Matches `lit = true | false | token_lit`. - crate fn parse_lit(&mut self) -> PResult<'a, Lit> { - let lo = self.span; - let lit = if self.eat_keyword(keywords::True) { - LitKind::Bool(true) - } else if self.eat_keyword(keywords::False) { - LitKind::Bool(false) - } else { - let lit = self.parse_lit_token()?; - lit - }; - Ok(source_map::Spanned { node: lit, span: lo.to(self.prev_span) }) - } - /// Matches `'-' lit | lit` (cf. `ast_validation::AstValidator::check_expr_within_pat`). crate fn parse_literal_maybe_minus(&mut self) -> PResult<'a, P<Expr>> { maybe_whole_expr!(self); @@ -2471,27 +2293,27 @@ impl<'a> Parser<'a> { }) } - fn mk_expr(&mut self, span: Span, node: ExprKind, attrs: ThinVec<Attribute>) -> P<Expr> { + crate fn mk_expr(&self, span: Span, node: ExprKind, attrs: ThinVec<Attribute>) -> P<Expr> { P(Expr { node, span, attrs, id: ast::DUMMY_NODE_ID }) } - fn mk_unary(&mut self, unop: ast::UnOp, expr: P<Expr>) -> ast::ExprKind { + fn mk_unary(&self, unop: ast::UnOp, expr: P<Expr>) -> ast::ExprKind { ExprKind::Unary(unop, expr) } - fn mk_binary(&mut self, binop: ast::BinOp, lhs: P<Expr>, rhs: P<Expr>) -> ast::ExprKind { + fn mk_binary(&self, binop: ast::BinOp, lhs: P<Expr>, rhs: P<Expr>) -> ast::ExprKind { ExprKind::Binary(binop, lhs, rhs) } - fn mk_call(&mut self, f: P<Expr>, args: Vec<P<Expr>>) -> ast::ExprKind { + fn mk_call(&self, f: P<Expr>, args: Vec<P<Expr>>) -> ast::ExprKind { ExprKind::Call(f, args) } - fn mk_index(&mut self, expr: P<Expr>, idx: P<Expr>) -> ast::ExprKind { + fn mk_index(&self, expr: P<Expr>, idx: P<Expr>) -> ast::ExprKind { ExprKind::Index(expr, idx) } - fn mk_range(&mut self, + fn mk_range(&self, start: Option<P<Expr>>, end: Option<P<Expr>>, limits: RangeLimits) @@ -2503,7 +2325,7 @@ impl<'a> Parser<'a> { } } - fn mk_assign_op(&mut self, binop: ast::BinOp, + fn mk_assign_op(&self, binop: ast::BinOp, lhs: P<Expr>, rhs: P<Expr>) -> ast::ExprKind { ExprKind::AssignOp(binop, lhs, rhs) } @@ -2643,13 +2465,12 @@ impl<'a> Parser<'a> { hi = path.span; return Ok(self.mk_expr(lo.to(hi), ExprKind::Path(Some(qself), path), attrs)); } - if self.span.rust_2018() && self.check_keyword(keywords::Async) - { - if self.is_async_block() { // check for `async {` and `async move {` - return self.parse_async_block(attrs); + if self.span.rust_2018() && self.check_keyword(keywords::Async) { + return if self.is_async_block() { // check for `async {` and `async move {` + self.parse_async_block(attrs) } else { - return self.parse_lambda_expr(attrs); - } + self.parse_lambda_expr(attrs) + }; } if self.check_keyword(keywords::Move) || self.check_keyword(keywords::Static) { return self.parse_lambda_expr(attrs); @@ -2758,13 +2579,9 @@ impl<'a> Parser<'a> { db.note("variable declaration using `let` is a statement"); return Err(db); } else if self.span.rust_2018() && self.eat_keyword(keywords::Await) { - // FIXME: remove this branch when `await!` is no longer supported - // https://github.com/rust-lang/rust/issues/60610 - self.expect(&token::Not)?; - self.expect(&token::OpenDelim(token::Paren))?; - let expr = self.parse_expr()?; - self.expect(&token::CloseDelim(token::Paren))?; - ex = ExprKind::Await(ast::AwaitOrigin::MacroLike, expr); + let (await_hi, e_kind) = self.parse_await_macro_or_alt(lo, self.prev_span)?; + hi = await_hi; + ex = e_kind; } else if self.token.is_path_start() { let path = self.parse_path(PathStyle::Expr)?; @@ -2829,6 +2646,31 @@ impl<'a> Parser<'a> { self.maybe_recover_from_bad_qpath(expr, true) } + /// Parse `await!(<expr>)` calls, or alternatively recover from incorrect but reasonable + /// alternative syntaxes `await <expr>`, `await? <expr>`, `await(<expr>)` and + /// `await { <expr> }`. + fn parse_await_macro_or_alt( + &mut self, + lo: Span, + await_sp: Span, + ) -> PResult<'a, (Span, ExprKind)> { + if self.token == token::Not { + // Handle correct `await!(<expr>)`. + // FIXME: make this an error when `await!` is no longer supported + // https://github.com/rust-lang/rust/issues/60610 + self.expect(&token::Not)?; + self.expect(&token::OpenDelim(token::Paren))?; + let expr = self.parse_expr().map_err(|mut err| { + err.span_label(await_sp, "while parsing this await macro call"); + err + })?; + self.expect(&token::CloseDelim(token::Paren))?; + Ok((self.prev_span, ExprKind::Await(ast::AwaitOrigin::MacroLike, expr))) + } else { // Handle `await <expr>`. + self.parse_incorrect_await_syntax(lo, await_sp) + } + } + fn maybe_parse_struct_expr( &mut self, lo: Span, @@ -2977,10 +2819,13 @@ impl<'a> Parser<'a> { } /// Parses a block or unsafe block. - fn parse_block_expr(&mut self, opt_label: Option<Label>, - lo: Span, blk_mode: BlockCheckMode, - outer_attrs: ThinVec<Attribute>) - -> PResult<'a, P<Expr>> { + crate fn parse_block_expr( + &mut self, + opt_label: Option<Label>, + lo: Span, + blk_mode: BlockCheckMode, + outer_attrs: ThinVec<Attribute>, + ) -> PResult<'a, P<Expr>> { self.expect(&token::OpenDelim(token::Brace))?; let mut attrs = outer_attrs; @@ -3041,6 +2886,7 @@ impl<'a> Parser<'a> { ExprKind::Await(ast::AwaitOrigin::FieldLike, self_arg), ThinVec::new(), ); + self.recover_from_await_method_call(); return Ok(await_expr); } let segment = self.parse_path_segment(PathStyle::Expr)?; @@ -3279,23 +3125,6 @@ impl<'a> Parser<'a> { return Ok(e); } - fn recover_seq_parse_error( - &mut self, - delim: token::DelimToken, - lo: Span, - result: PResult<'a, P<Expr>>, - ) -> P<Expr> { - match result { - Ok(x) => x, - Err(mut err) => { - err.emit(); - // recover from parse error - self.consume_block(delim); - self.mk_expr(lo.to(self.prev_span), ExprKind::Err, ThinVec::new()) - } - } - } - crate fn process_potential_macro_variable(&mut self) { let (token, span) = match self.token { token::Dollar if self.span.ctxt() != syntax_pos::hygiene::SyntaxContext::empty() && @@ -3574,7 +3403,8 @@ impl<'a> Parser<'a> { } else { self.restrictions }; - if op.precedence() < min_prec { + let prec = op.precedence(); + if prec < min_prec { break; } // Check for deprecated `...` syntax @@ -3615,8 +3445,7 @@ impl<'a> Parser<'a> { // We have 2 alternatives here: `x..y`/`x..=y` and `x..`/`x..=` The other // two variants are handled with `parse_prefix_range_expr` call above. let rhs = if self.is_at_start_of_range_notation_rhs() { - Some(self.parse_assoc_expr_with(op.precedence() + 1, - LhsExpr::NotYetParsed)?) + Some(self.parse_assoc_expr_with(prec + 1, LhsExpr::NotYetParsed)?) } else { None }; @@ -3636,28 +3465,18 @@ impl<'a> Parser<'a> { break } - let rhs = match op.fixity() { - Fixity::Right => self.with_res( - restrictions - Restrictions::STMT_EXPR, - |this| { - this.parse_assoc_expr_with(op.precedence(), - LhsExpr::NotYetParsed) - }), - Fixity::Left => self.with_res( - restrictions - Restrictions::STMT_EXPR, - |this| { - this.parse_assoc_expr_with(op.precedence() + 1, - LhsExpr::NotYetParsed) - }), + let fixity = op.fixity(); + let prec_adjustment = match fixity { + Fixity::Right => 0, + Fixity::Left => 1, // We currently have no non-associative operators that are not handled above by // the special cases. The code is here only for future convenience. - Fixity::None => self.with_res( - restrictions - Restrictions::STMT_EXPR, - |this| { - this.parse_assoc_expr_with(op.precedence() + 1, - LhsExpr::NotYetParsed) - }), - }?; + Fixity::None => 1, + }; + let rhs = self.with_res( + restrictions - Restrictions::STMT_EXPR, + |this| this.parse_assoc_expr_with(prec + prec_adjustment, LhsExpr::NotYetParsed) + )?; // Make sure that the span of the parent node is larger than the span of lhs and rhs, // including the attributes. @@ -3678,8 +3497,7 @@ impl<'a> Parser<'a> { let binary = self.mk_binary(source_map::respan(cur_op_span, ast_op), lhs, rhs); self.mk_expr(span, binary, ThinVec::new()) } - AssocOp::Assign => - self.mk_expr(span, ExprKind::Assign(lhs, rhs), ThinVec::new()), + AssocOp::Assign => self.mk_expr(span, ExprKind::Assign(lhs, rhs), ThinVec::new()), AssocOp::ObsoleteInPlace => self.mk_expr(span, ExprKind::ObsoleteInPlace(lhs, rhs), ThinVec::new()), AssocOp::AssignOp(k) => { @@ -3703,63 +3521,11 @@ impl<'a> Parser<'a> { } }; - if op.fixity() == Fixity::None { break } + if let Fixity::None = fixity { break } } Ok(lhs) } - fn could_ascription_be_path(&self, node: &ast::ExprKind) -> bool { - self.token.is_ident() && - if let ast::ExprKind::Path(..) = node { true } else { false } && - !self.token.is_reserved_ident() && // v `foo:bar(baz)` - self.look_ahead(1, |t| t == &token::OpenDelim(token::Paren)) || - self.look_ahead(1, |t| t == &token::Lt) && // `foo:bar<baz` - self.look_ahead(2, |t| t.is_ident()) || - self.look_ahead(1, |t| t == &token::Colon) && // `foo:bar:baz` - self.look_ahead(2, |t| t.is_ident()) || - self.look_ahead(1, |t| t == &token::ModSep) && // `foo:bar::baz` - self.look_ahead(2, |t| t.is_ident()) - } - - fn bad_type_ascription( - &self, - err: &mut DiagnosticBuilder<'a>, - lhs_span: Span, - cur_op_span: Span, - next_sp: Span, - maybe_path: bool, - ) { - err.span_label(self.span, "expecting a type here because of type ascription"); - let cm = self.sess.source_map(); - let next_pos = cm.lookup_char_pos(next_sp.lo()); - let op_pos = cm.lookup_char_pos(cur_op_span.hi()); - if op_pos.line != next_pos.line { - err.span_suggestion( - cur_op_span, - "try using a semicolon", - ";".to_string(), - Applicability::MaybeIncorrect, - ); - } else { - if maybe_path { - err.span_suggestion( - cur_op_span, - "maybe you meant to write a path separator here", - "::".to_string(), - Applicability::MaybeIncorrect, - ); - } else { - err.note("type ascription is a nightly-only feature that lets \ - you annotate an expression with a type: `<expr>: <type>`"); - err.span_note( - lhs_span, - "this expression expects an ascribed type after the colon", - ); - err.help("this might be indicative of a syntax error elsewhere"); - } - } - } - fn parse_assoc_op_cast(&mut self, lhs: P<Expr>, lhs_span: Span, expr_kind: fn(P<Expr>, P<Ty>) -> ExprKind) -> PResult<'a, P<Expr>> { @@ -3840,7 +3606,7 @@ impl<'a> Parser<'a> { /// Produce an error if comparison operators are chained (RFC #558). /// We only need to check lhs, not rhs, because all comparison ops /// have same precedence and are left-associative - fn check_no_chained_comparison(&mut self, lhs: &Expr, outer_op: &AssocOp) { + fn check_no_chained_comparison(&self, lhs: &Expr, outer_op: &AssocOp) { debug_assert!(outer_op.is_comparison(), "check_no_chained_comparison: {:?} is not comparison", outer_op); @@ -4179,8 +3945,6 @@ impl<'a> Parser<'a> { } crate fn parse_arm(&mut self) -> PResult<'a, Arm> { - maybe_whole!(self, NtArm, |x| x); - let attrs = self.parse_outer_attributes()?; let pats = self.parse_pats()?; let guard = if self.eat_keyword(keywords::If) { @@ -5043,92 +4807,6 @@ impl<'a> Parser<'a> { Ok(self.parse_stmt_(true)) } - // Eat tokens until we can be relatively sure we reached the end of the - // statement. This is something of a best-effort heuristic. - // - // We terminate when we find an unmatched `}` (without consuming it). - fn recover_stmt(&mut self) { - self.recover_stmt_(SemiColonMode::Ignore, BlockMode::Ignore) - } - - // If `break_on_semi` is `Break`, then we will stop consuming tokens after - // finding (and consuming) a `;` outside of `{}` or `[]` (note that this is - // approximate - it can mean we break too early due to macros, but that - // should only lead to sub-optimal recovery, not inaccurate parsing). - // - // If `break_on_block` is `Break`, then we will stop consuming tokens - // after finding (and consuming) a brace-delimited block. - fn recover_stmt_(&mut self, break_on_semi: SemiColonMode, break_on_block: BlockMode) { - let mut brace_depth = 0; - let mut bracket_depth = 0; - let mut in_block = false; - debug!("recover_stmt_ enter loop (semi={:?}, block={:?})", - break_on_semi, break_on_block); - loop { - debug!("recover_stmt_ loop {:?}", self.token); - match self.token { - token::OpenDelim(token::DelimToken::Brace) => { - brace_depth += 1; - self.bump(); - if break_on_block == BlockMode::Break && - brace_depth == 1 && - bracket_depth == 0 { - in_block = true; - } - } - token::OpenDelim(token::DelimToken::Bracket) => { - bracket_depth += 1; - self.bump(); - } - token::CloseDelim(token::DelimToken::Brace) => { - if brace_depth == 0 { - debug!("recover_stmt_ return - close delim {:?}", self.token); - break; - } - brace_depth -= 1; - self.bump(); - if in_block && bracket_depth == 0 && brace_depth == 0 { - debug!("recover_stmt_ return - block end {:?}", self.token); - break; - } - } - token::CloseDelim(token::DelimToken::Bracket) => { - bracket_depth -= 1; - if bracket_depth < 0 { - bracket_depth = 0; - } - self.bump(); - } - token::Eof => { - debug!("recover_stmt_ return - Eof"); - break; - } - token::Semi => { - self.bump(); - if break_on_semi == SemiColonMode::Break && - brace_depth == 0 && - bracket_depth == 0 { - debug!("recover_stmt_ return - Semi"); - break; - } - } - token::Comma => { - if break_on_semi == SemiColonMode::Comma && - brace_depth == 0 && - bracket_depth == 0 { - debug!("recover_stmt_ return - Semi"); - break; - } else { - self.bump(); - } - } - _ => { - self.bump() - } - } - } - } - fn parse_stmt_(&mut self, macro_legacy_warnings: bool) -> Option<Stmt> { self.parse_stmt_without_recovery(macro_legacy_warnings).unwrap_or_else(|mut e| { e.emit(); @@ -5137,7 +4815,7 @@ impl<'a> Parser<'a> { }) } - fn is_async_block(&mut self) -> bool { + fn is_async_block(&self) -> bool { self.token.is_keyword(keywords::Async) && ( ( // `async move {` @@ -5149,19 +4827,19 @@ impl<'a> Parser<'a> { ) } - fn is_async_fn(&mut self) -> bool { + fn is_async_fn(&self) -> bool { self.token.is_keyword(keywords::Async) && self.look_ahead(1, |t| t.is_keyword(keywords::Fn)) } - fn is_do_catch_block(&mut self) -> bool { + fn is_do_catch_block(&self) -> bool { self.token.is_keyword(keywords::Do) && self.look_ahead(1, |t| t.is_keyword(keywords::Catch)) && self.look_ahead(2, |t| *t == token::OpenDelim(token::Brace)) && !self.restrictions.contains(Restrictions::NO_STRUCT_LITERAL) } - fn is_try_block(&mut self) -> bool { + fn is_try_block(&self) -> bool { self.token.is_keyword(keywords::Try) && self.look_ahead(1, |t| *t == token::OpenDelim(token::Brace)) && self.span.rust_2018() && @@ -5183,7 +4861,7 @@ impl<'a> Parser<'a> { self.look_ahead(1, |t| t.is_keyword(keywords::Type)) } - fn is_auto_trait_item(&mut self) -> bool { + fn is_auto_trait_item(&self) -> bool { // auto trait (self.token.is_keyword(keywords::Auto) && self.look_ahead(1, |t| t.is_keyword(keywords::Trait))) @@ -5225,7 +4903,7 @@ impl<'a> Parser<'a> { (ident, ast::MacroDef { tokens: tokens.into(), legacy: false }) } - token::Ident(ident, _) if ident.name == "macro_rules" && + token::Ident(ident, _) if ident.name == sym::macro_rules && self.look_ahead(1, |t| *t == token::Not) => { let prev_span = self.prev_span; self.complain_if_pub_macro(&vis.node, prev_span); @@ -5445,7 +5123,7 @@ impl<'a> Parser<'a> { } /// Checks if this expression is a successfully parsed statement. - fn expr_is_complete(&mut self, e: &Expr) -> bool { + fn expr_is_complete(&self, e: &Expr) -> bool { self.restrictions.contains(Restrictions::STMT_EXPR) && !classify::expr_requires_semi_to_be_stmt(e) } @@ -5915,8 +5593,6 @@ impl<'a> Parser<'a> { /// | ( < lifetimes , typaramseq ( , )? > ) /// where typaramseq = ( typaram ) | ( typaram , typaramseq ) fn parse_generics(&mut self) -> PResult<'a, ast::Generics> { - maybe_whole!(self, NtGenerics, |x| x); - let span_lo = self.span; if self.eat_lt() { let params = self.parse_generic_params()?; @@ -6169,8 +5845,6 @@ impl<'a> Parser<'a> { /// where T : Trait<U, V> + 'b, 'a : 'b /// ``` fn parse_where_clause(&mut self) -> PResult<'a, WhereClause> { - maybe_whole!(self, NtWhereClause, |x| x); - let mut where_clause = WhereClause { id: ast::DUMMY_NODE_ID, predicates: Vec::new(), @@ -6517,7 +6191,7 @@ impl<'a> Parser<'a> { Ok((id, generics)) } - fn mk_item(&mut self, span: Span, ident: Ident, node: ItemKind, vis: Visibility, + fn mk_item(&self, span: Span, ident: Ident, node: ItemKind, vis: Visibility, attrs: Vec<Attribute>) -> P<Item> { P(Item { ident, @@ -6549,7 +6223,7 @@ impl<'a> Parser<'a> { /// Returns `true` if we are looking at `const ID` /// (returns `false` for things like `const fn`, etc.). - fn is_const_item(&mut self) -> bool { + fn is_const_item(&self) -> bool { self.token.is_keyword(keywords::Const) && !self.look_ahead(1, |t| t.is_keyword(keywords::Fn)) && !self.look_ahead(1, |t| t.is_keyword(keywords::Unsafe)) @@ -6657,7 +6331,7 @@ impl<'a> Parser<'a> { }) } - fn complain_if_pub_macro(&mut self, vis: &VisibilityKind, sp: Span) { + fn complain_if_pub_macro(&self, vis: &VisibilityKind, sp: Span) { match *vis { VisibilityKind::Inherited => {} _ => { @@ -6686,7 +6360,7 @@ impl<'a> Parser<'a> { } } - fn missing_assoc_item_kind_err(&mut self, item_type: &str, prev_span: Span) + fn missing_assoc_item_kind_err(&self, item_type: &str, prev_span: Span) -> DiagnosticBuilder<'a> { let expected_kinds = if item_type == "extern" { @@ -7036,26 +6710,6 @@ impl<'a> Parser<'a> { Ok((class_name, ItemKind::Union(vdata, generics), None)) } - fn consume_block(&mut self, delim: token::DelimToken) { - let mut brace_depth = 0; - loop { - if self.eat(&token::OpenDelim(delim)) { - brace_depth += 1; - } else if self.eat(&token::CloseDelim(delim)) { - if brace_depth == 0 { - return; - } else { - brace_depth -= 1; - continue; - } - } else if self.token == token::Eof || self.eat(&token::CloseDelim(token::NoDelim)) { - return; - } else { - self.bump(); - } - } - } - fn parse_record_struct_body( &mut self, ) -> PResult<'a, (Vec<StructField>, /* recovered */ bool)> { @@ -7357,7 +7011,8 @@ impl<'a> Parser<'a> { let attr = Attribute { id: attr::mk_attr_id(), style: ast::AttrStyle::Outer, - path: ast::Path::from_ident(Ident::from_str("warn_directory_ownership")), + path: ast::Path::from_ident( + Ident::with_empty_ctxt(sym::warn_directory_ownership)), tokens: TokenStream::empty(), is_sugared_doc: false, span: syntax_pos::DUMMY_SP, @@ -7389,7 +7044,7 @@ impl<'a> Parser<'a> { } fn push_directory(&mut self, id: Ident, attrs: &[Attribute]) { - if let Some(path) = attr::first_attr_value_str_by_name(attrs, "path") { + if let Some(path) = attr::first_attr_value_str_by_name(attrs, sym::path) { self.directory.path.to_mut().push(&path.as_str()); self.directory.ownership = DirectoryOwnership::Owned { relative: None }; } else { @@ -7409,7 +7064,7 @@ impl<'a> Parser<'a> { } pub fn submod_path_from_attr(attrs: &[Attribute], dir_path: &Path) -> Option<PathBuf> { - if let Some(s) = attr::first_attr_value_str_by_name(attrs, "path") { + if let Some(s) = attr::first_attr_value_str_by_name(attrs, sym::path) { let s = s.as_str(); // On windows, the base path might have the form @@ -8753,21 +8408,6 @@ impl<'a> Parser<'a> { ).emit(); } - /// Recover from `pub` keyword in places where it seems _reasonable_ but isn't valid. - fn eat_bad_pub(&mut self) { - if self.token.is_keyword(keywords::Pub) { - match self.parse_visibility(false) { - Ok(vis) => { - let mut err = self.diagnostic() - .struct_span_err(vis.span, "unnecessary visibility qualifier"); - err.span_label(vis.span, "`pub` not permitted here"); - err.emit(); - } - Err(mut err) => err.emit(), - } - } - } - /// When lowering a `async fn` to the HIR, we need to move all of the arguments of the function /// into the generated closure so that they are dropped when the future is polled and not when /// it is created. diff --git a/src/libsyntax/parse/token.rs b/src/libsyntax/parse/token.rs index fd7a39c576d..068fc41c87a 100644 --- a/src/libsyntax/parse/token.rs +++ b/src/libsyntax/parse/token.rs @@ -19,7 +19,7 @@ use log::info; use std::fmt; use std::mem; #[cfg(target_arch = "x86_64")] -use rustc_data_structures::static_assert; +use rustc_data_structures::static_assert_size; use rustc_data_structures::sync::Lrc; #[derive(Clone, PartialEq, RustcEncodable, RustcDecodable, Hash, Debug, Copy)] @@ -61,6 +61,7 @@ impl DelimToken { #[derive(Clone, PartialEq, RustcEncodable, RustcDecodable, Hash, Debug, Copy)] pub enum Lit { + Bool(ast::Name), // AST only, must never appear in a `Token` Byte(ast::Name), Char(ast::Name), Err(ast::Name), @@ -72,9 +73,13 @@ pub enum Lit { ByteStrRaw(ast::Name, u16), /* raw byte str delimited by n hash symbols */ } +#[cfg(target_arch = "x86_64")] +static_assert_size!(Lit, 8); + impl Lit { crate fn literal_name(&self) -> &'static str { match *self { + Bool(_) => panic!("literal token contains `Lit::Bool`"), Byte(_) => "byte literal", Char(_) => "char literal", Err(_) => "invalid literal", @@ -85,6 +90,13 @@ impl Lit { } } + crate fn may_have_suffix(&self) -> bool { + match *self { + Integer(..) | Float(..) => true, + _ => false, + } + } + // See comments in `Nonterminal::to_tokenstream` for why we care about // *probably* equal here rather than actual equality fn probably_equal_for_proc_macro(&self, other: &Lit) -> bool { @@ -208,7 +220,7 @@ pub enum Token { // `Token` is used a lot. Make sure it doesn't unintentionally get bigger. #[cfg(target_arch = "x86_64")] -static_assert!(MEM_SIZE_OF_STATEMENT: mem::size_of::<Token>() == 16); +static_assert_size!(Token, 16); impl Token { /// Recovers a `Token` from an `ast::Ident`. This creates a raw identifier if necessary. @@ -585,14 +597,12 @@ pub enum Nonterminal { NtPath(ast::Path), NtVis(ast::Visibility), NtTT(TokenTree), - // These are not exposed to macros, but are used by quasiquote. - NtArm(ast::Arm), - NtImplItem(ast::ImplItem), + // Used only for passing items to proc macro attributes (they are not + // strictly necessary for that, `Annotatable` can be converted into + // tokens directly, but doing that naively regresses pretty-printing). NtTraitItem(ast::TraitItem), + NtImplItem(ast::ImplItem), NtForeignItem(ast::ForeignItem), - NtGenerics(ast::Generics), - NtWhereClause(ast::WhereClause), - NtArg(ast::Arg), } impl PartialEq for Nonterminal { @@ -625,13 +635,9 @@ impl fmt::Debug for Nonterminal { NtMeta(..) => f.pad("NtMeta(..)"), NtPath(..) => f.pad("NtPath(..)"), NtTT(..) => f.pad("NtTT(..)"), - NtArm(..) => f.pad("NtArm(..)"), NtImplItem(..) => f.pad("NtImplItem(..)"), NtTraitItem(..) => f.pad("NtTraitItem(..)"), NtForeignItem(..) => f.pad("NtForeignItem(..)"), - NtGenerics(..) => f.pad("NtGenerics(..)"), - NtWhereClause(..) => f.pad("NtWhereClause(..)"), - NtArg(..) => f.pad("NtArg(..)"), NtVis(..) => f.pad("NtVis(..)"), NtLifetime(..) => f.pad("NtLifetime(..)"), } |
