diff options
Diffstat (limited to 'src/libsyntax/parse')
| -rw-r--r-- | src/libsyntax/parse/diagnostics.rs | 304 | ||||
| -rw-r--r-- | src/libsyntax/parse/lexer/mod.rs | 57 | ||||
| -rw-r--r-- | src/libsyntax/parse/lexer/tokentrees.rs | 67 | ||||
| -rw-r--r-- | src/libsyntax/parse/mod.rs | 56 | ||||
| -rw-r--r-- | src/libsyntax/parse/parser.rs | 303 | ||||
| -rw-r--r-- | src/libsyntax/parse/token.rs | 6 |
6 files changed, 458 insertions, 335 deletions
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/mod.rs b/src/libsyntax/parse/lexer/mod.rs index e76605cde32..c97d8040761 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) } @@ -251,14 +234,7 @@ 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, } } @@ -1582,10 +1558,10 @@ mod tests { use crate::feature_gate::UnstableFeatures; use crate::parse::token; use crate::diagnostics::plugin::ErrorMap; - use crate::with_globals; + use crate::with_default_globals; use std::io; use std::path::PathBuf; - use syntax_pos::{BytePos, Span, NO_EXPANSION}; + use syntax_pos::{BytePos, Span, NO_EXPANSION, edition::Edition}; use rustc_data_structures::fx::{FxHashSet, FxHashMap}; use rustc_data_structures::sync::Lock; @@ -1605,6 +1581,7 @@ mod tests { raw_identifier_spans: Lock::new(Vec::new()), registered_diagnostics: Lock::new(ErrorMap::new()), buffered_lints: Lock::new(vec![]), + edition: Edition::from_session(), ambiguous_block_expr_parse: Lock::new(FxHashMap::default()), } } @@ -1625,7 +1602,7 @@ mod tests { #[test] fn t1() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); let mut string_reader = setup(&sm, @@ -1673,7 +1650,7 @@ mod tests { #[test] fn doublecolonparsing() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); check_tokenization(setup(&sm, &sh, "a b".to_string()), @@ -1683,7 +1660,7 @@ mod tests { #[test] fn dcparsing_2() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); check_tokenization(setup(&sm, &sh, "a::b".to_string()), @@ -1693,7 +1670,7 @@ mod tests { #[test] fn dcparsing_3() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); check_tokenization(setup(&sm, &sh, "a ::b".to_string()), @@ -1703,7 +1680,7 @@ mod tests { #[test] fn dcparsing_4() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); check_tokenization(setup(&sm, &sh, "a:: b".to_string()), @@ -1713,7 +1690,7 @@ mod tests { #[test] fn character_a() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); assert_eq!(setup(&sm, &sh, "'a'".to_string()).next_token().tok, @@ -1723,7 +1700,7 @@ mod tests { #[test] fn character_space() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); assert_eq!(setup(&sm, &sh, "' '".to_string()).next_token().tok, @@ -1733,7 +1710,7 @@ mod tests { #[test] fn character_escaped() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); assert_eq!(setup(&sm, &sh, "'\\n'".to_string()).next_token().tok, @@ -1743,7 +1720,7 @@ mod tests { #[test] fn lifetime_name() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); assert_eq!(setup(&sm, &sh, "'abc".to_string()).next_token().tok, @@ -1753,7 +1730,7 @@ mod tests { #[test] fn raw_string() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); assert_eq!(setup(&sm, &sh, "r###\"\"#a\\b\x00c\"\"###".to_string()) @@ -1765,7 +1742,7 @@ mod tests { #[test] fn literal_suffixes() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); macro_rules! test { @@ -1811,7 +1788,7 @@ mod tests { #[test] fn nested_block_comments() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); let mut lexer = setup(&sm, &sh, "/* /* */ */'a'".to_string()); @@ -1826,7 +1803,7 @@ mod tests { #[test] fn crlf_comments() { - with_globals(|| { + with_default_globals(|| { let sm = Lrc::new(SourceMap::new(FilePathMapping::empty())); let sh = mk_sess(sm.clone()); let mut lexer = setup(&sm, &sh, "// test\r\n/// test\r\n".to_string()); 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/mod.rs b/src/libsyntax/parse/mod.rs index 655baafcd2d..1073fc6f3ab 100644 --- a/src/libsyntax/parse/mod.rs +++ b/src/libsyntax/parse/mod.rs @@ -13,6 +13,7 @@ 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 syntax_pos::edition::Edition; use rustc_data_structures::fx::{FxHashSet, FxHashMap}; use std::borrow::Cow; @@ -38,6 +39,7 @@ pub struct ParseSess { pub span_diagnostic: Handler, pub unstable_features: UnstableFeatures, pub config: CrateConfig, + pub edition: Edition, pub missing_fragment_specifiers: Lock<FxHashSet<Span>>, /// Places where raw identifiers were used. This is used for feature-gating raw identifiers. pub raw_identifier_spans: Lock<Vec<Span>>, @@ -74,6 +76,7 @@ impl ParseSess { included_mod_stack: Lock::new(vec![]), source_map, buffered_lints: Lock::new(vec![]), + edition: Edition::from_session(), ambiguous_block_expr_parse: Lock::new(FxHashMap::default()), } } @@ -290,22 +293,22 @@ 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)), @@ -329,6 +332,23 @@ pub fn stream_to_parser(sess: &ParseSess, stream: TokenStream) -> Parser<'_> { Parser::new(sess, stream, None, true, false) } +/// Given stream, the `ParseSess` and the base directory, produces a parser. +/// +/// Use this function when you are creating a parser from the token stream +/// and also care about the current working directory of the parser (e.g., +/// you are trying to resolve modules defined inside a macro invocation). +/// +/// # Note +/// +/// The main usage of this function is outside of rustc, for those who uses +/// libsyntax as a library. Please do not remove this function while refactoring +/// just because it is not used in rustc codebase! +pub fn stream_to_parser_with_base_dir<'a>(sess: &'a ParseSess, + stream: TokenStream, + base_dir: Directory<'a>) -> Parser<'a> { + Parser::new(sess, stream, Some(base_dir), true, false) +} + /// A sequence separator. pub struct SeqSep { /// The seperator token. @@ -363,7 +383,7 @@ mod tests { use crate::tokenstream::{DelimSpan, TokenTree}; use crate::util::parser_testing::string_to_stream; use crate::util::parser_testing::{string_to_expr, string_to_item}; - use crate::with_globals; + use crate::with_default_globals; use syntax_pos::{Span, BytePos, Pos, NO_EXPANSION}; /// Parses an item. @@ -382,7 +402,7 @@ mod tests { #[should_panic] #[test] fn bad_path_expr_1() { - with_globals(|| { + with_default_globals(|| { string_to_expr("::abc::def::return".to_string()); }) } @@ -390,7 +410,7 @@ mod tests { // check the token-tree-ization of macros #[test] fn string_to_tts_macro () { - with_globals(|| { + with_default_globals(|| { use crate::symbol::sym; let tts: Vec<_> = @@ -447,7 +467,7 @@ mod tests { #[test] fn string_to_tts_1() { - with_globals(|| { + with_default_globals(|| { let tts = string_to_stream("fn a (b : i32) { b; }".to_string()); let expected = TokenStream::new(vec![ @@ -480,7 +500,7 @@ mod tests { } #[test] fn parse_use() { - with_globals(|| { + with_default_globals(|| { let use_s = "use foo::bar::baz;"; let vitem = string_to_item(use_s.to_string()).unwrap(); let vitem_s = item_to_string(&vitem); @@ -494,7 +514,7 @@ mod tests { } #[test] fn parse_extern_crate() { - with_globals(|| { + with_default_globals(|| { let ex_s = "extern crate foo;"; let vitem = string_to_item(ex_s.to_string()).unwrap(); let vitem_s = item_to_string(&vitem); @@ -531,7 +551,7 @@ mod tests { } #[test] fn span_of_self_arg_pat_idents_are_correct() { - with_globals(|| { + with_default_globals(|| { let srcs = ["impl z { fn a (&self, &myarg: i32) {} }", "impl z { fn a (&mut self, &myarg: i32) {} }", @@ -551,7 +571,7 @@ mod tests { } #[test] fn parse_exprs () { - with_globals(|| { + with_default_globals(|| { // just make sure that they parse.... string_to_expr("3 + 4".to_string()); string_to_expr("a::z.froob(b,&(987+3))".to_string()); @@ -559,7 +579,7 @@ mod tests { } #[test] fn attrs_fix_bug () { - with_globals(|| { + with_default_globals(|| { string_to_item("pub fn mk_file_writer(path: &Path, flags: &[FileFlag]) -> Result<Box<Writer>, String> { #[cfg(windows)] @@ -576,7 +596,7 @@ mod tests { } #[test] fn crlf_doc_comments() { - with_globals(|| { + with_default_globals(|| { use crate::symbol::sym; let sess = ParseSess::new(FilePathMapping::empty()); @@ -613,7 +633,7 @@ mod tests { new_parser_from_source_str(sess, name, source).parse_expr() } - with_globals(|| { + with_default_globals(|| { let sess = ParseSess::new(FilePathMapping::empty()); let expr = parse_expr_from_source_str(PathBuf::from("foo").into(), "foo!( fn main() { body } )".to_string(), &sess).unwrap(); @@ -637,7 +657,7 @@ mod tests { // See `recurse_into_file_modules` in the parser. #[test] fn out_of_line_mod() { - with_globals(|| { + with_default_globals(|| { let sess = ParseSess::new(FilePathMapping::empty()); let item = parse_item_from_source_str( PathBuf::from("foo").into(), diff --git a/src/libsyntax/parse/parser.rs b/src/libsyntax/parse/parser.rs index 2d6c8c54075..24d120376de 100644 --- a/src/libsyntax/parse/parser.rs +++ b/src/libsyntax/parse/parser.rs @@ -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()), @@ -673,56 +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. @@ -2343,7 +2293,7 @@ impl<'a> Parser<'a> { }) } - fn mk_expr(&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 }) } @@ -2629,13 +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)?; @@ -2700,6 +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, @@ -2848,10 +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; @@ -2912,6 +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)?; @@ -3150,23 +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() && @@ -3539,8 +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) => { @@ -3569,58 +3526,6 @@ impl<'a> Parser<'a> { 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>> { @@ -4902,92 +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(); @@ -6891,26 +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)> { @@ -7212,7 +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, @@ -8608,21 +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 049fb6cb78b..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)] @@ -74,7 +74,7 @@ pub enum Lit { } #[cfg(target_arch = "x86_64")] -static_assert!(MEM_SIZE_OF_LIT: mem::size_of::<Lit>() == 8); +static_assert_size!(Lit, 8); impl Lit { crate fn literal_name(&self) -> &'static str { @@ -220,7 +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. |
