13 use crate::lexer::UnmatchedBrace;
14 pub use attr_wrapper::AttrWrapper;
15 pub use diagnostics::AttemptLocalParseRecovery;
16 use diagnostics::Error;
17 pub(crate) use item::FnParseMode;
18 pub use pat::{CommaRecoveryMode, RecoverColon, RecoverComma};
19 pub use path::PathStyle;
21 use rustc_ast::ptr::P;
22 use rustc_ast::token::{self, DelimToken, Token, TokenKind};
23 use rustc_ast::tokenstream::AttributesData;
24 use rustc_ast::tokenstream::{self, DelimSpan, Spacing};
25 use rustc_ast::tokenstream::{TokenStream, TokenTree};
26 use rustc_ast::AttrId;
27 use rustc_ast::DUMMY_NODE_ID;
28 use rustc_ast::{self as ast, AnonConst, AstLike, AttrStyle, AttrVec, Const, CrateSugar, Extern};
29 use rustc_ast::{Async, Expr, ExprKind, MacArgs, MacDelimiter, Mutability, StrLit, Unsafe};
30 use rustc_ast::{Visibility, VisibilityKind};
31 use rustc_ast_pretty::pprust;
32 use rustc_data_structures::fx::FxHashMap;
33 use rustc_data_structures::sync::Lrc;
34 use rustc_errors::PResult;
36 struct_span_err, Applicability, DiagnosticBuilder, ErrorGuaranteed, FatalError,
38 use rustc_session::parse::ParseSess;
39 use rustc_span::source_map::{MultiSpan, Span, DUMMY_SP};
40 use rustc_span::symbol::{kw, sym, Ident, Symbol};
44 use std::{cmp, mem, slice};
47 struct Restrictions: u8 {
48 const STMT_EXPR = 1 << 0;
49 const NO_STRUCT_LITERAL = 1 << 1;
50 const CONST_EXPR = 1 << 2;
54 #[derive(Clone, Copy, PartialEq, Debug)]
61 #[derive(Clone, Copy, PartialEq, Debug)]
67 /// Whether or not we should force collection of tokens for an AST node,
68 /// regardless of whether or not it has attributes
69 #[derive(Clone, Copy, PartialEq)]
70 pub enum ForceCollect {
75 #[derive(Debug, Eq, PartialEq)]
76 pub enum TrailingToken {
79 /// If the trailing token is a comma, then capture it
80 /// Otherwise, ignore the trailing token
84 /// Like `maybe_whole_expr`, but for things other than expressions.
86 macro_rules! maybe_whole {
87 ($p:expr, $constructor:ident, |$x:ident| $e:expr) => {
88 if let token::Interpolated(nt) = &$p.token.kind {
89 if let token::$constructor(x) = &**nt {
98 /// If the next tokens are ill-formed `$ty::` recover them as `<$ty>::`.
100 macro_rules! maybe_recover_from_interpolated_ty_qpath {
101 ($self: expr, $allow_qpath_recovery: expr) => {
102 if $allow_qpath_recovery
103 && $self.look_ahead(1, |t| t == &token::ModSep)
104 && let token::Interpolated(nt) = &$self.token.kind
105 && let token::NtTy(ty) = &**nt
109 return $self.maybe_recover_from_bad_qpath_stage_2($self.prev_token.span, ty);
115 pub struct Parser<'a> {
116 pub sess: &'a ParseSess,
117 /// The current token.
119 /// The spacing for the current token
120 pub token_spacing: Spacing,
121 /// The previous token.
122 pub prev_token: Token,
123 pub capture_cfg: bool,
124 restrictions: Restrictions,
125 expected_tokens: Vec<TokenType>,
126 // Important: This must only be advanced from `next_tok`
127 // to ensure that `token_cursor.num_next_calls` is updated properly
128 token_cursor: TokenCursor,
129 desugar_doc_comments: bool,
130 /// This field is used to keep track of how many left angle brackets we have seen. This is
131 /// required in order to detect extra leading left angle brackets (`<` characters) and error
134 /// See the comments in the `parse_path_segment` function for more details.
135 unmatched_angle_bracket_count: u32,
136 max_angle_bracket_count: u32,
137 /// A list of all unclosed delimiters found by the lexer. If an entry is used for error recovery
138 /// it gets removed from here. Every entry left at the end gets emitted as an independent
140 pub(super) unclosed_delims: Vec<UnmatchedBrace>,
141 last_unexpected_token_span: Option<Span>,
142 /// Span pointing at the `:` for the last type ascription the parser has seen, and whether it
143 /// looked like it could have been a mistyped path or literal `Option:Some(42)`).
144 pub last_type_ascription: Option<(Span, bool /* likely path typo */)>,
145 /// If present, this `Parser` is not parsing Rust code but rather a macro call.
146 subparser_name: Option<&'static str>,
147 capture_state: CaptureState,
148 /// This allows us to recover when the user forget to add braces around
149 /// multiple statements in the closure body.
150 pub current_closure: Option<ClosureSpans>,
153 /// Stores span informations about a closure.
155 pub struct ClosureSpans {
156 pub whole_closure: Span,
157 pub closing_pipe: Span,
161 /// Indicates a range of tokens that should be replaced by
162 /// the tokens in the provided vector. This is used in two
163 /// places during token collection:
165 /// 1. During the parsing of an AST node that may have a `#[derive]`
166 /// attribute, we parse a nested AST node that has `#[cfg]` or `#[cfg_attr]`
167 /// In this case, we use a `ReplaceRange` to replace the entire inner AST node
168 /// with `FlatToken::AttrTarget`, allowing us to perform eager cfg-expansion
169 /// on an `AttrAnnotatedTokenStream`
171 /// 2. When we parse an inner attribute while collecting tokens. We
172 /// remove inner attributes from the token stream entirely, and
173 /// instead track them through the `attrs` field on the AST node.
174 /// This allows us to easily manipulate them (for example, removing
175 /// the first macro inner attribute to invoke a proc-macro).
176 /// When create a `TokenStream`, the inner attributes get inserted
177 /// into the proper place in the token stream.
178 pub type ReplaceRange = (Range<u32>, Vec<(FlatToken, Spacing)>);
180 /// Controls how we capture tokens. Capturing can be expensive,
181 /// so we try to avoid performing capturing in cases where
182 /// we will never need an `AttrAnnotatedTokenStream`
183 #[derive(Copy, Clone)]
185 /// We aren't performing any capturing - this is the default mode.
187 /// We are capturing tokens
192 struct CaptureState {
193 capturing: Capturing,
194 replace_ranges: Vec<ReplaceRange>,
195 inner_attr_ranges: FxHashMap<AttrId, ReplaceRange>,
198 impl<'a> Drop for Parser<'a> {
200 emit_unclosed_delims(&mut self.unclosed_delims, &self.sess);
206 frame: TokenCursorFrame,
207 stack: Vec<TokenCursorFrame>,
208 desugar_doc_comments: bool,
209 // Counts the number of calls to `next` or `next_desugared`,
210 // depending on whether `desugar_doc_comments` is set.
211 num_next_calls: usize,
212 // During parsing, we may sometimes need to 'unglue' a
213 // glued token into two component tokens
214 // (e.g. '>>' into '>' and '>), so that the parser
215 // can consume them one at a time. This process
216 // bypasses the normal capturing mechanism
217 // (e.g. `num_next_calls` will not be incremented),
218 // since the 'unglued' tokens due not exist in
219 // the original `TokenStream`.
221 // If we end up consuming both unglued tokens,
222 // then this is not an issue - we'll end up
223 // capturing the single 'glued' token.
225 // However, in certain circumstances, we may
226 // want to capture just the first 'unglued' token.
227 // For example, capturing the `Vec<u8>`
228 // in `Option<Vec<u8>>` requires us to unglue
229 // the trailing `>>` token. The `break_last_token`
230 // field is used to track this token - it gets
231 // appended to the captured stream when
232 // we evaluate a `LazyTokenStream`
233 break_last_token: bool,
237 struct TokenCursorFrame {
238 delim: token::DelimToken,
241 tree_cursor: tokenstream::Cursor,
245 impl TokenCursorFrame {
246 fn new(span: DelimSpan, delim: DelimToken, tts: TokenStream) -> Self {
251 tree_cursor: tts.into_trees(),
258 fn next(&mut self) -> (Token, Spacing) {
260 let (tree, spacing) = if !self.frame.open_delim {
261 self.frame.open_delim = true;
262 TokenTree::open_tt(self.frame.span, self.frame.delim).into()
263 } else if let Some(tree) = self.frame.tree_cursor.next_with_spacing() {
265 } else if !self.frame.close_delim {
266 self.frame.close_delim = true;
267 TokenTree::close_tt(self.frame.span, self.frame.delim).into()
268 } else if let Some(frame) = self.stack.pop() {
272 (TokenTree::Token(Token::new(token::Eof, DUMMY_SP)), Spacing::Alone)
276 TokenTree::Token(token) => {
277 return (token, spacing);
279 TokenTree::Delimited(sp, delim, tts) => {
280 let frame = TokenCursorFrame::new(sp, delim, tts);
281 self.stack.push(mem::replace(&mut self.frame, frame));
287 fn next_desugared(&mut self) -> (Token, Spacing) {
288 let (data, attr_style, sp) = match self.next() {
289 (Token { kind: token::DocComment(_, attr_style, data), span }, _) => {
290 (data, attr_style, span)
295 // Searches for the occurrences of `"#*` and returns the minimum number of `#`s
296 // required to wrap the text.
297 let mut num_of_hashes = 0;
299 for ch in data.as_str().chars() {
302 '#' if count > 0 => count + 1,
305 num_of_hashes = cmp::max(num_of_hashes, count);
308 let delim_span = DelimSpan::from_single(sp);
309 let body = TokenTree::Delimited(
313 TokenTree::token(token::Ident(sym::doc, false), sp),
314 TokenTree::token(token::Eq, sp),
315 TokenTree::token(TokenKind::lit(token::StrRaw(num_of_hashes), data, None), sp),
319 .collect::<TokenStream>(),
322 self.stack.push(mem::replace(
324 TokenCursorFrame::new(
327 if attr_style == AttrStyle::Inner {
328 [TokenTree::token(token::Pound, sp), TokenTree::token(token::Not, sp), body]
331 .collect::<TokenStream>()
333 [TokenTree::token(token::Pound, sp), body]
336 .collect::<TokenStream>()
345 #[derive(Debug, Clone, PartialEq)]
358 fn to_string(&self) -> String {
360 TokenType::Token(ref t) => format!("`{}`", pprust::token_kind_to_string(t)),
361 TokenType::Keyword(kw) => format!("`{}`", kw),
362 TokenType::Operator => "an operator".to_string(),
363 TokenType::Lifetime => "lifetime".to_string(),
364 TokenType::Ident => "identifier".to_string(),
365 TokenType::Path => "path".to_string(),
366 TokenType::Type => "type".to_string(),
367 TokenType::Const => "a const expression".to_string(),
372 #[derive(Copy, Clone, Debug)]
373 enum TokenExpectType {
378 /// A sequence separator.
380 /// The separator token.
381 sep: Option<TokenKind>,
382 /// `true` if a trailing separator is allowed.
383 trailing_sep_allowed: bool,
387 fn trailing_allowed(t: TokenKind) -> SeqSep {
388 SeqSep { sep: Some(t), trailing_sep_allowed: true }
391 fn none() -> SeqSep {
392 SeqSep { sep: None, trailing_sep_allowed: false }
396 pub enum FollowedByType {
401 fn token_descr_opt(token: &Token) -> Option<&'static str> {
402 Some(match token.kind {
403 _ if token.is_special_ident() => "reserved identifier",
404 _ if token.is_used_keyword() => "keyword",
405 _ if token.is_unused_keyword() => "reserved keyword",
406 token::DocComment(..) => "doc comment",
411 pub(super) fn token_descr(token: &Token) -> String {
412 let token_str = pprust::token_to_string(token);
413 match token_descr_opt(token) {
414 Some(prefix) => format!("{} `{}`", prefix, token_str),
415 _ => format!("`{}`", token_str),
419 impl<'a> Parser<'a> {
423 desugar_doc_comments: bool,
424 subparser_name: Option<&'static str>,
426 let mut start_frame = TokenCursorFrame::new(DelimSpan::dummy(), token::NoDelim, tokens);
427 start_frame.open_delim = true;
428 start_frame.close_delim = true;
430 let mut parser = Parser {
432 token: Token::dummy(),
433 token_spacing: Spacing::Alone,
434 prev_token: Token::dummy(),
436 restrictions: Restrictions::empty(),
437 expected_tokens: Vec::new(),
438 token_cursor: TokenCursor {
442 desugar_doc_comments,
443 break_last_token: false,
445 desugar_doc_comments,
446 unmatched_angle_bracket_count: 0,
447 max_angle_bracket_count: 0,
448 unclosed_delims: Vec::new(),
449 last_unexpected_token_span: None,
450 last_type_ascription: None,
452 capture_state: CaptureState {
453 capturing: Capturing::No,
454 replace_ranges: Vec::new(),
455 inner_attr_ranges: Default::default(),
457 current_closure: None,
460 // Make parser point to the first token.
466 fn next_tok(&mut self, fallback_span: Span) -> (Token, Spacing) {
468 let (mut next, spacing) = if self.desugar_doc_comments {
469 self.token_cursor.next_desugared()
471 self.token_cursor.next()
473 self.token_cursor.num_next_calls += 1;
474 // We've retrieved an token from the underlying
475 // cursor, so we no longer need to worry about
476 // an unglued token. See `break_and_eat` for more details
477 self.token_cursor.break_last_token = false;
478 if next.span.is_dummy() {
479 // Tweak the location for better diagnostics, but keep syntactic context intact.
480 next.span = fallback_span.with_ctxt(next.span.ctxt());
484 token::OpenDelim(token::NoDelim) | token::CloseDelim(token::NoDelim)
488 return (next, spacing);
492 pub fn unexpected<T>(&mut self) -> PResult<'a, T> {
493 match self.expect_one_of(&[], &[]) {
495 // We can get `Ok(true)` from `recover_closing_delimiter`
496 // which is called in `expected_one_of_not_found`.
497 Ok(_) => FatalError.raise(),
501 /// Expects and consumes the token `t`. Signals an error if the next token is not `t`.
502 pub fn expect(&mut self, t: &TokenKind) -> PResult<'a, bool /* recovered */> {
503 if self.expected_tokens.is_empty() {
504 if self.token == *t {
508 self.unexpected_try_recover(t)
511 self.expect_one_of(slice::from_ref(t), &[])
515 /// Expect next token to be edible or inedible token. If edible,
516 /// then consume it; if inedible, then return without consuming
517 /// anything. Signal a fatal error if next token is unexpected.
518 pub fn expect_one_of(
520 edible: &[TokenKind],
521 inedible: &[TokenKind],
522 ) -> PResult<'a, bool /* recovered */> {
523 if edible.contains(&self.token.kind) {
526 } else if inedible.contains(&self.token.kind) {
527 // leave it in the input
529 } else if self.last_unexpected_token_span == Some(self.token.span) {
532 self.expected_one_of_not_found(edible, inedible)
536 // Public for rustfmt usage.
537 pub fn parse_ident(&mut self) -> PResult<'a, Ident> {
538 self.parse_ident_common(true)
541 fn ident_or_err(&mut self) -> PResult<'a, (Ident, /* is_raw */ bool)> {
542 self.token.ident().ok_or_else(|| match self.prev_token.kind {
543 TokenKind::DocComment(..) => {
544 self.span_err(self.prev_token.span, Error::UselessDocComment)
546 _ => self.expected_ident_found(),
550 fn parse_ident_common(&mut self, recover: bool) -> PResult<'a, Ident> {
551 let (ident, is_raw) = self.ident_or_err()?;
552 if !is_raw && ident.is_reserved() {
553 let mut err = self.expected_ident_found();
564 /// Checks if the next token is `tok`, and returns `true` if so.
566 /// This method will automatically add `tok` to `expected_tokens` if `tok` is not
568 fn check(&mut self, tok: &TokenKind) -> bool {
569 let is_present = self.token == *tok;
571 self.expected_tokens.push(TokenType::Token(tok.clone()));
576 /// Consumes a token 'tok' if it exists. Returns whether the given token was present.
577 pub fn eat(&mut self, tok: &TokenKind) -> bool {
578 let is_present = self.check(tok);
585 /// If the next token is the given keyword, returns `true` without eating it.
586 /// An expectation is also added for diagnostics purposes.
587 fn check_keyword(&mut self, kw: Symbol) -> bool {
588 self.expected_tokens.push(TokenType::Keyword(kw));
589 self.token.is_keyword(kw)
592 /// If the next token is the given keyword, eats it and returns `true`.
593 /// Otherwise, returns `false`. An expectation is also added for diagnostics purposes.
594 // Public for rustfmt usage.
595 pub fn eat_keyword(&mut self, kw: Symbol) -> bool {
596 if self.check_keyword(kw) {
604 fn eat_keyword_noexpect(&mut self, kw: Symbol) -> bool {
605 if self.token.is_keyword(kw) {
613 /// If the given word is not a keyword, signals an error.
614 /// If the next token is not the given word, signals an error.
615 /// Otherwise, eats it.
616 fn expect_keyword(&mut self, kw: Symbol) -> PResult<'a, ()> {
617 if !self.eat_keyword(kw) { self.unexpected() } else { Ok(()) }
620 /// Is the given keyword `kw` followed by a non-reserved identifier?
621 fn is_kw_followed_by_ident(&self, kw: Symbol) -> bool {
622 self.token.is_keyword(kw) && self.look_ahead(1, |t| t.is_ident() && !t.is_reserved_ident())
625 fn check_or_expected(&mut self, ok: bool, typ: TokenType) -> bool {
629 self.expected_tokens.push(typ);
634 fn check_ident(&mut self) -> bool {
635 self.check_or_expected(self.token.is_ident(), TokenType::Ident)
638 fn check_path(&mut self) -> bool {
639 self.check_or_expected(self.token.is_path_start(), TokenType::Path)
642 fn check_type(&mut self) -> bool {
643 self.check_or_expected(self.token.can_begin_type(), TokenType::Type)
646 fn check_const_arg(&mut self) -> bool {
647 self.check_or_expected(self.token.can_begin_const_arg(), TokenType::Const)
650 fn check_inline_const(&self, dist: usize) -> bool {
651 self.is_keyword_ahead(dist, &[kw::Const])
652 && self.look_ahead(dist + 1, |t| match t.kind {
653 token::Interpolated(ref nt) => matches!(**nt, token::NtBlock(..)),
654 token::OpenDelim(DelimToken::Brace) => true,
659 /// Checks to see if the next token is either `+` or `+=`.
660 /// Otherwise returns `false`.
661 fn check_plus(&mut self) -> bool {
662 self.check_or_expected(
663 self.token.is_like_plus(),
664 TokenType::Token(token::BinOp(token::Plus)),
668 /// Eats the expected token if it's present possibly breaking
669 /// compound tokens like multi-character operators in process.
670 /// Returns `true` if the token was eaten.
671 fn break_and_eat(&mut self, expected: TokenKind) -> bool {
672 if self.token.kind == expected {
676 match self.token.kind.break_two_token_op() {
677 Some((first, second)) if first == expected => {
678 let first_span = self.sess.source_map().start_point(self.token.span);
679 let second_span = self.token.span.with_lo(first_span.hi());
680 self.token = Token::new(first, first_span);
681 // Keep track of this token - if we end token capturing now,
682 // we'll want to append this token to the captured stream.
684 // If we consume any additional tokens, then this token
685 // is not needed (we'll capture the entire 'glued' token),
686 // and `next_tok` will set this field to `None`
687 self.token_cursor.break_last_token = true;
688 // Use the spacing of the glued token as the spacing
689 // of the unglued second token.
690 self.bump_with((Token::new(second, second_span), self.token_spacing));
694 self.expected_tokens.push(TokenType::Token(expected));
700 /// Eats `+` possibly breaking tokens like `+=` in process.
701 fn eat_plus(&mut self) -> bool {
702 self.break_and_eat(token::BinOp(token::Plus))
705 /// Eats `&` possibly breaking tokens like `&&` in process.
706 /// Signals an error if `&` is not eaten.
707 fn expect_and(&mut self) -> PResult<'a, ()> {
708 if self.break_and_eat(token::BinOp(token::And)) { Ok(()) } else { self.unexpected() }
711 /// Eats `|` possibly breaking tokens like `||` in process.
712 /// Signals an error if `|` was not eaten.
713 fn expect_or(&mut self) -> PResult<'a, ()> {
714 if self.break_and_eat(token::BinOp(token::Or)) { Ok(()) } else { self.unexpected() }
717 /// Eats `<` possibly breaking tokens like `<<` in process.
718 fn eat_lt(&mut self) -> bool {
719 let ate = self.break_and_eat(token::Lt);
721 // See doc comment for `unmatched_angle_bracket_count`.
722 self.unmatched_angle_bracket_count += 1;
723 self.max_angle_bracket_count += 1;
724 debug!("eat_lt: (increment) count={:?}", self.unmatched_angle_bracket_count);
729 /// Eats `<` possibly breaking tokens like `<<` in process.
730 /// Signals an error if `<` was not eaten.
731 fn expect_lt(&mut self) -> PResult<'a, ()> {
732 if self.eat_lt() { Ok(()) } else { self.unexpected() }
735 /// Eats `>` possibly breaking tokens like `>>` in process.
736 /// Signals an error if `>` was not eaten.
737 fn expect_gt(&mut self) -> PResult<'a, ()> {
738 if self.break_and_eat(token::Gt) {
739 // See doc comment for `unmatched_angle_bracket_count`.
740 if self.unmatched_angle_bracket_count > 0 {
741 self.unmatched_angle_bracket_count -= 1;
742 debug!("expect_gt: (decrement) count={:?}", self.unmatched_angle_bracket_count);
750 fn expect_any_with_type(&mut self, kets: &[&TokenKind], expect: TokenExpectType) -> bool {
751 kets.iter().any(|k| match expect {
752 TokenExpectType::Expect => self.check(k),
753 TokenExpectType::NoExpect => self.token == **k,
757 fn parse_seq_to_before_tokens<T>(
761 expect: TokenExpectType,
762 mut f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
763 ) -> PResult<'a, (Vec<T>, bool /* trailing */, bool /* recovered */)> {
764 let mut first = true;
765 let mut recovered = false;
766 let mut trailing = false;
768 let unclosed_delims = !self.unclosed_delims.is_empty();
770 while !self.expect_any_with_type(kets, expect) {
771 if let token::CloseDelim(..) | token::Eof = self.token.kind {
774 if let Some(ref t) = sep.sep {
778 match self.expect(t) {
780 self.current_closure.take();
783 self.current_closure.take();
787 Err(mut expect_err) => {
788 let sp = self.prev_token.span.shrink_to_hi();
789 let token_str = pprust::token_kind_to_string(t);
791 match self.current_closure.take() {
792 Some(closure_spans) if self.token.kind == TokenKind::Semi => {
793 // Finding a semicolon instead of a comma
794 // after a closure body indicates that the
795 // closure body may be a block but the user
796 // forgot to put braces around its
799 self.recover_missing_braces_around_closure_body(
808 // Attempt to keep parsing if it was a similar separator.
809 if let Some(ref tokens) = t.similar_tokens() {
810 if tokens.contains(&self.token.kind) && !unclosed_delims {
817 // If this was a missing `@` in a binding pattern
818 // bail with a suggestion
819 // https://github.com/rust-lang/rust/issues/72373
820 if self.prev_token.is_ident() && self.token.kind == token::DotDot {
822 "if you meant to bind the contents of \
823 the rest of the array pattern into `{}`, use `@`",
824 pprust::token_to_string(&self.prev_token)
827 .span_suggestion_verbose(
828 self.prev_token.span.shrink_to_hi().until(self.token.span),
831 Applicability::MaybeIncorrect,
837 // Attempt to keep parsing if it was an omitted separator.
840 // Parsed successfully, therefore most probably the code only
841 // misses a separator.
843 .span_suggestion_short(
845 &format!("missing `{}`", token_str),
847 Applicability::MaybeIncorrect,
855 // Parsing failed, therefore it must be something more serious
856 // than just a missing separator.
867 if sep.trailing_sep_allowed && self.expect_any_with_type(kets, expect) {
876 Ok((v, trailing, recovered))
879 fn recover_missing_braces_around_closure_body(
881 closure_spans: ClosureSpans,
882 mut expect_err: DiagnosticBuilder<'_, ErrorGuaranteed>,
883 ) -> PResult<'a, ()> {
884 let initial_semicolon = self.token.span;
886 while self.eat(&TokenKind::Semi) {
887 let _ = self.parse_stmt(ForceCollect::Yes)?;
890 expect_err.set_primary_message(
891 "closure bodies that contain statements must be surrounded by braces",
894 let preceding_pipe_span = closure_spans.closing_pipe;
895 let following_token_span = self.token.span;
897 let mut first_note = MultiSpan::from(vec![initial_semicolon]);
898 first_note.push_span_label(
900 "this `;` turns the preceding closure into a statement".to_string(),
902 first_note.push_span_label(
904 "this expression is a statement because of the trailing semicolon".to_string(),
906 expect_err.span_note(first_note, "statement found outside of a block");
908 let mut second_note = MultiSpan::from(vec![closure_spans.whole_closure]);
909 second_note.push_span_label(
910 closure_spans.whole_closure,
911 "this is the parsed closure...".to_string(),
913 second_note.push_span_label(
914 following_token_span,
915 "...but likely you meant the closure to end here".to_string(),
917 expect_err.span_note(second_note, "the closure body may be incorrectly delimited");
919 expect_err.set_span(vec![preceding_pipe_span, following_token_span]);
921 let opening_suggestion_str = " {".to_string();
922 let closing_suggestion_str = "}".to_string();
924 expect_err.multipart_suggestion(
927 (preceding_pipe_span.shrink_to_hi(), opening_suggestion_str),
928 (following_token_span.shrink_to_lo(), closing_suggestion_str),
930 Applicability::MaybeIncorrect,
938 /// Parses a sequence, not including the closing delimiter. The function
939 /// `f` must consume tokens until reaching the next separator or
941 fn parse_seq_to_before_end<T>(
945 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
946 ) -> PResult<'a, (Vec<T>, bool, bool)> {
947 self.parse_seq_to_before_tokens(&[ket], sep, TokenExpectType::Expect, f)
950 /// Parses a sequence, including the closing delimiter. The function
951 /// `f` must consume tokens until reaching the next separator or
953 fn parse_seq_to_end<T>(
957 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
958 ) -> PResult<'a, (Vec<T>, bool /* trailing */)> {
959 let (val, trailing, recovered) = self.parse_seq_to_before_end(ket, sep, f)?;
966 /// Parses a sequence, including the closing delimiter. The function
967 /// `f` must consume tokens until reaching the next separator or
969 fn parse_unspanned_seq<T>(
974 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
975 ) -> PResult<'a, (Vec<T>, bool)> {
977 self.parse_seq_to_end(ket, sep, f)
980 fn parse_delim_comma_seq<T>(
983 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
984 ) -> PResult<'a, (Vec<T>, bool)> {
985 self.parse_unspanned_seq(
986 &token::OpenDelim(delim),
987 &token::CloseDelim(delim),
988 SeqSep::trailing_allowed(token::Comma),
993 fn parse_paren_comma_seq<T>(
995 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
996 ) -> PResult<'a, (Vec<T>, bool)> {
997 self.parse_delim_comma_seq(token::Paren, f)
1000 /// Advance the parser by one token using provided token as the next one.
1001 fn bump_with(&mut self, (next_token, next_spacing): (Token, Spacing)) {
1002 // Bumping after EOF is a bad sign, usually an infinite loop.
1003 if self.prev_token.kind == TokenKind::Eof {
1004 let msg = "attempted to bump the parser past EOF (may be stuck in a loop)";
1005 self.span_bug(self.token.span, msg);
1008 // Update the current and previous tokens.
1009 self.prev_token = mem::replace(&mut self.token, next_token);
1010 self.token_spacing = next_spacing;
1013 self.expected_tokens.clear();
1016 /// Advance the parser by one token.
1017 pub fn bump(&mut self) {
1018 let next_token = self.next_tok(self.token.span);
1019 self.bump_with(next_token);
1022 /// Look-ahead `dist` tokens of `self.token` and get access to that token there.
1023 /// When `dist == 0` then the current token is looked at.
1024 pub fn look_ahead<R>(&self, dist: usize, looker: impl FnOnce(&Token) -> R) -> R {
1026 return looker(&self.token);
1029 let frame = &self.token_cursor.frame;
1030 if frame.delim != DelimToken::NoDelim {
1031 let all_normal = (0..dist).all(|i| {
1032 let token = frame.tree_cursor.look_ahead(i);
1033 !matches!(token, Some(TokenTree::Delimited(_, DelimToken::NoDelim, _)))
1036 return match frame.tree_cursor.look_ahead(dist - 1) {
1037 Some(tree) => match tree {
1038 TokenTree::Token(token) => looker(token),
1039 TokenTree::Delimited(dspan, delim, _) => {
1040 looker(&Token::new(token::OpenDelim(*delim), dspan.open))
1043 None => looker(&Token::new(token::CloseDelim(frame.delim), frame.span.close)),
1048 let mut cursor = self.token_cursor.clone();
1050 let mut token = Token::dummy();
1052 token = cursor.next().0;
1055 token::OpenDelim(token::NoDelim) | token::CloseDelim(token::NoDelim)
1061 return looker(&token);
1064 /// Returns whether any of the given keywords are `dist` tokens ahead of the current one.
1065 fn is_keyword_ahead(&self, dist: usize, kws: &[Symbol]) -> bool {
1066 self.look_ahead(dist, |t| kws.iter().any(|&kw| t.is_keyword(kw)))
1069 /// Parses asyncness: `async` or nothing.
1070 fn parse_asyncness(&mut self) -> Async {
1071 if self.eat_keyword(kw::Async) {
1072 let span = self.prev_token.uninterpolated_span();
1073 Async::Yes { span, closure_id: DUMMY_NODE_ID, return_impl_trait_id: DUMMY_NODE_ID }
1079 /// Parses unsafety: `unsafe` or nothing.
1080 fn parse_unsafety(&mut self) -> Unsafe {
1081 if self.eat_keyword(kw::Unsafe) {
1082 Unsafe::Yes(self.prev_token.uninterpolated_span())
1088 /// Parses constness: `const` or nothing.
1089 fn parse_constness(&mut self) -> Const {
1090 // Avoid const blocks to be parsed as const items
1091 if self.look_ahead(1, |t| t != &token::OpenDelim(DelimToken::Brace))
1092 && self.eat_keyword(kw::Const)
1094 Const::Yes(self.prev_token.uninterpolated_span())
1100 /// Parses inline const expressions.
1101 fn parse_const_block(&mut self, span: Span, pat: bool) -> PResult<'a, P<Expr>> {
1103 self.sess.gated_spans.gate(sym::inline_const_pat, span);
1105 self.sess.gated_spans.gate(sym::inline_const, span);
1107 self.eat_keyword(kw::Const);
1108 let blk = self.parse_block()?;
1109 let anon_const = AnonConst {
1111 value: self.mk_expr(blk.span, ExprKind::Block(blk, None), AttrVec::new()),
1113 let blk_span = anon_const.value.span;
1114 Ok(self.mk_expr(span.to(blk_span), ExprKind::ConstBlock(anon_const), AttrVec::new()))
1117 /// Parses mutability (`mut` or nothing).
1118 fn parse_mutability(&mut self) -> Mutability {
1119 if self.eat_keyword(kw::Mut) { Mutability::Mut } else { Mutability::Not }
1122 /// Possibly parses mutability (`const` or `mut`).
1123 fn parse_const_or_mut(&mut self) -> Option<Mutability> {
1124 if self.eat_keyword(kw::Mut) {
1125 Some(Mutability::Mut)
1126 } else if self.eat_keyword(kw::Const) {
1127 Some(Mutability::Not)
1133 fn parse_field_name(&mut self) -> PResult<'a, Ident> {
1134 if let token::Literal(token::Lit { kind: token::Integer, symbol, suffix }) = self.token.kind
1136 self.expect_no_suffix(self.token.span, "a tuple index", suffix);
1138 Ok(Ident::new(symbol, self.prev_token.span))
1140 self.parse_ident_common(true)
1144 fn parse_mac_args(&mut self) -> PResult<'a, P<MacArgs>> {
1145 self.parse_mac_args_common(true).map(P)
1148 fn parse_attr_args(&mut self) -> PResult<'a, MacArgs> {
1149 self.parse_mac_args_common(false)
1152 fn parse_mac_args_common(&mut self, delimited_only: bool) -> PResult<'a, MacArgs> {
1154 if self.check(&token::OpenDelim(DelimToken::Paren))
1155 || self.check(&token::OpenDelim(DelimToken::Bracket))
1156 || self.check(&token::OpenDelim(DelimToken::Brace))
1158 match self.parse_token_tree() {
1159 TokenTree::Delimited(dspan, delim, tokens) =>
1160 // We've confirmed above that there is a delimiter so unwrapping is OK.
1162 MacArgs::Delimited(dspan, MacDelimiter::from_token(delim).unwrap(), tokens)
1164 _ => unreachable!(),
1166 } else if !delimited_only {
1167 if self.eat(&token::Eq) {
1168 let eq_span = self.prev_token.span;
1170 // Collect tokens because they are used during lowering to HIR.
1171 let expr = self.parse_expr_force_collect()?;
1172 let span = expr.span;
1174 let token_kind = token::Interpolated(Lrc::new(token::NtExpr(expr)));
1175 MacArgs::Eq(eq_span, Token::new(token_kind, span))
1180 return self.unexpected();
1185 fn parse_or_use_outer_attributes(
1187 already_parsed_attrs: Option<AttrWrapper>,
1188 ) -> PResult<'a, AttrWrapper> {
1189 if let Some(attrs) = already_parsed_attrs {
1192 self.parse_outer_attributes()
1196 /// Parses a single token tree from the input.
1197 pub(crate) fn parse_token_tree(&mut self) -> TokenTree {
1198 match self.token.kind {
1199 token::OpenDelim(..) => {
1200 let depth = self.token_cursor.stack.len();
1202 // We keep advancing the token cursor until we hit
1203 // the matching `CloseDelim` token.
1204 while !(depth == self.token_cursor.stack.len()
1205 && matches!(self.token.kind, token::CloseDelim(_)))
1207 // Advance one token at a time, so `TokenCursor::next()`
1208 // can capture these tokens if necessary.
1211 // We are still inside the frame corresponding
1212 // to the delimited stream we captured, so grab
1213 // the tokens from this frame.
1214 let frame = &self.token_cursor.frame;
1215 let stream = frame.tree_cursor.stream.clone();
1216 let span = frame.span;
1217 let delim = frame.delim;
1218 // Consume close delimiter
1220 TokenTree::Delimited(span, delim, stream)
1222 token::CloseDelim(_) | token::Eof => unreachable!(),
1225 TokenTree::Token(self.prev_token.clone())
1230 /// Parses a stream of tokens into a list of `TokenTree`s, up to EOF.
1231 pub fn parse_all_token_trees(&mut self) -> PResult<'a, Vec<TokenTree>> {
1232 let mut tts = Vec::new();
1233 while self.token != token::Eof {
1234 tts.push(self.parse_token_tree());
1239 pub fn parse_tokens(&mut self) -> TokenStream {
1240 let mut result = Vec::new();
1242 match self.token.kind {
1243 token::Eof | token::CloseDelim(..) => break,
1244 _ => result.push(self.parse_token_tree().into()),
1247 TokenStream::new(result)
1250 /// Evaluates the closure with restrictions in place.
1252 /// Afters the closure is evaluated, restrictions are reset.
1253 fn with_res<T>(&mut self, res: Restrictions, f: impl FnOnce(&mut Self) -> T) -> T {
1254 let old = self.restrictions;
1255 self.restrictions = res;
1257 self.restrictions = old;
1261 fn is_crate_vis(&self) -> bool {
1262 self.token.is_keyword(kw::Crate) && self.look_ahead(1, |t| t != &token::ModSep)
1265 /// Parses `pub`, `pub(crate)` and `pub(in path)` plus shortcuts `crate` for `pub(crate)`,
1266 /// `pub(self)` for `pub(in self)` and `pub(super)` for `pub(in super)`.
1267 /// If the following element can't be a tuple (i.e., it's a function definition), then
1268 /// it's not a tuple struct field), and the contents within the parentheses aren't valid,
1269 /// so emit a proper diagnostic.
1270 // Public for rustfmt usage.
1271 pub fn parse_visibility(&mut self, fbt: FollowedByType) -> PResult<'a, Visibility> {
1272 maybe_whole!(self, NtVis, |x| x);
1274 self.expected_tokens.push(TokenType::Keyword(kw::Crate));
1275 if self.is_crate_vis() {
1276 self.bump(); // `crate`
1277 self.sess.gated_spans.gate(sym::crate_visibility_modifier, self.prev_token.span);
1278 return Ok(Visibility {
1279 span: self.prev_token.span,
1280 kind: VisibilityKind::Crate(CrateSugar::JustCrate),
1285 if !self.eat_keyword(kw::Pub) {
1286 // We need a span for our `Spanned<VisibilityKind>`, but there's inherently no
1287 // keyword to grab a span from for inherited visibility; an empty span at the
1288 // beginning of the current token would seem to be the "Schelling span".
1289 return Ok(Visibility {
1290 span: self.token.span.shrink_to_lo(),
1291 kind: VisibilityKind::Inherited,
1295 let lo = self.prev_token.span;
1297 if self.check(&token::OpenDelim(token::Paren)) {
1298 // We don't `self.bump()` the `(` yet because this might be a struct definition where
1299 // `()` or a tuple might be allowed. For example, `struct Struct(pub (), pub (usize));`.
1300 // Because of this, we only `bump` the `(` if we're assured it is appropriate to do so
1301 // by the following tokens.
1302 if self.is_keyword_ahead(1, &[kw::Crate]) && self.look_ahead(2, |t| t != &token::ModSep)
1303 // account for `pub(crate::foo)`
1305 // Parse `pub(crate)`.
1307 self.bump(); // `crate`
1308 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1309 let vis = VisibilityKind::Crate(CrateSugar::PubCrate);
1310 return Ok(Visibility {
1311 span: lo.to(self.prev_token.span),
1315 } else if self.is_keyword_ahead(1, &[kw::In]) {
1316 // Parse `pub(in path)`.
1318 self.bump(); // `in`
1319 let path = self.parse_path(PathStyle::Mod)?; // `path`
1320 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1321 let vis = VisibilityKind::Restricted { path: P(path), id: ast::DUMMY_NODE_ID };
1322 return Ok(Visibility {
1323 span: lo.to(self.prev_token.span),
1327 } else if self.look_ahead(2, |t| t == &token::CloseDelim(token::Paren))
1328 && self.is_keyword_ahead(1, &[kw::Super, kw::SelfLower])
1330 // Parse `pub(self)` or `pub(super)`.
1332 let path = self.parse_path(PathStyle::Mod)?; // `super`/`self`
1333 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1334 let vis = VisibilityKind::Restricted { path: P(path), id: ast::DUMMY_NODE_ID };
1335 return Ok(Visibility {
1336 span: lo.to(self.prev_token.span),
1340 } else if let FollowedByType::No = fbt {
1341 // Provide this diagnostic if a type cannot follow;
1342 // in particular, if this is not a tuple struct.
1343 self.recover_incorrect_vis_restriction()?;
1344 // Emit diagnostic, but continue with public visibility.
1348 Ok(Visibility { span: lo, kind: VisibilityKind::Public, tokens: None })
1351 /// Recovery for e.g. `pub(something) fn ...` or `struct X { pub(something) y: Z }`
1352 fn recover_incorrect_vis_restriction(&mut self) -> PResult<'a, ()> {
1354 let path = self.parse_path(PathStyle::Mod)?;
1355 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1357 let msg = "incorrect visibility restriction";
1358 let suggestion = r##"some possible visibility restrictions are:
1359 `pub(crate)`: visible only on the current crate
1360 `pub(super)`: visible only in the current module's parent
1361 `pub(in path::to::module)`: visible only on the specified path"##;
1363 let path_str = pprust::path_to_string(&path);
1365 struct_span_err!(self.sess.span_diagnostic, path.span, E0704, "{}", msg)
1369 &format!("make this visible only to module `{}` with `in`", path_str),
1370 format!("in {}", path_str),
1371 Applicability::MachineApplicable,
1378 /// Parses `extern string_literal?`.
1379 fn parse_extern(&mut self) -> Extern {
1380 if self.eat_keyword(kw::Extern) { Extern::from_abi(self.parse_abi()) } else { Extern::None }
1383 /// Parses a string literal as an ABI spec.
1384 fn parse_abi(&mut self) -> Option<StrLit> {
1385 match self.parse_str_lit() {
1386 Ok(str_lit) => Some(str_lit),
1387 Err(Some(lit)) => match lit.kind {
1388 ast::LitKind::Err(_) => None,
1390 self.struct_span_err(lit.span, "non-string ABI literal")
1393 "specify the ABI with a string literal",
1394 "\"C\"".to_string(),
1395 Applicability::MaybeIncorrect,
1405 pub fn collect_tokens_no_attrs<R: AstLike>(
1407 f: impl FnOnce(&mut Self) -> PResult<'a, R>,
1408 ) -> PResult<'a, R> {
1409 // The only reason to call `collect_tokens_no_attrs` is if you want tokens, so use
1410 // `ForceCollect::Yes`
1411 self.collect_tokens_trailing_token(
1412 AttrWrapper::empty(),
1414 |this, _attrs| Ok((f(this)?, TrailingToken::None)),
1419 fn is_import_coupler(&mut self) -> bool {
1420 self.check(&token::ModSep)
1421 && self.look_ahead(1, |t| {
1422 *t == token::OpenDelim(token::Brace) || *t == token::BinOp(token::Star)
1426 pub fn clear_expected_tokens(&mut self) {
1427 self.expected_tokens.clear();
1431 crate fn make_unclosed_delims_error(
1432 unmatched: UnmatchedBrace,
1434 ) -> Option<DiagnosticBuilder<'_, ErrorGuaranteed>> {
1435 // `None` here means an `Eof` was found. We already emit those errors elsewhere, we add them to
1436 // `unmatched_braces` only for error recovery in the `Parser`.
1437 let found_delim = unmatched.found_delim?;
1438 let span: MultiSpan = if let Some(sp) = unmatched.unclosed_span {
1439 vec![unmatched.found_span, sp].into()
1441 unmatched.found_span.into()
1443 let mut err = sess.span_diagnostic.struct_span_err(
1446 "mismatched closing delimiter: `{}`",
1447 pprust::token_kind_to_string(&token::CloseDelim(found_delim)),
1450 err.span_label(unmatched.found_span, "mismatched closing delimiter");
1451 if let Some(sp) = unmatched.candidate_span {
1452 err.span_label(sp, "closing delimiter possibly meant for this");
1454 if let Some(sp) = unmatched.unclosed_span {
1455 err.span_label(sp, "unclosed delimiter");
1460 pub fn emit_unclosed_delims(unclosed_delims: &mut Vec<UnmatchedBrace>, sess: &ParseSess) {
1461 *sess.reached_eof.borrow_mut() |=
1462 unclosed_delims.iter().any(|unmatched_delim| unmatched_delim.found_delim.is_none());
1463 for unmatched in unclosed_delims.drain(..) {
1464 if let Some(mut e) = make_unclosed_delims_error(unmatched, sess) {
1470 /// A helper struct used when building an `AttrAnnotatedTokenStream` from
1471 /// a `LazyTokenStream`. Both delimiter and non-delimited tokens
1472 /// are stored as `FlatToken::Token`. A vector of `FlatToken`s
1473 /// is then 'parsed' to build up an `AttrAnnotatedTokenStream` with nested
1474 /// `AttrAnnotatedTokenTree::Delimited` tokens
1475 #[derive(Debug, Clone)]
1476 pub enum FlatToken {
1477 /// A token - this holds both delimiter (e.g. '{' and '}')
1478 /// and non-delimiter tokens
1480 /// Holds the `AttributesData` for an AST node. The
1481 /// `AttributesData` is inserted directly into the
1482 /// constructed `AttrAnnotatedTokenStream` as
1483 /// an `AttrAnnotatedTokenTree::Attributes`
1484 AttrTarget(AttributesData),
1485 /// A special 'empty' token that is ignored during the conversion
1486 /// to an `AttrAnnotatedTokenStream`. This is used to simplify the
1487 /// handling of replace ranges.