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, Nonterminal, 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, MultiSpan,
38 use rustc_session::parse::ParseSess;
39 use rustc_span::source_map::{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 `bump` to ensure that
127 // `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 // This type is used a lot, e.g. it's cloned when matching many declarative macro rules. Make sure
154 // it doesn't unintentionally get bigger.
155 #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))]
156 rustc_data_structures::static_assert_size!(Parser<'_>, 328);
158 /// Stores span information about a closure.
160 pub struct ClosureSpans {
161 pub whole_closure: Span,
162 pub closing_pipe: Span,
166 /// Indicates a range of tokens that should be replaced by
167 /// the tokens in the provided vector. This is used in two
168 /// places during token collection:
170 /// 1. During the parsing of an AST node that may have a `#[derive]`
171 /// attribute, we parse a nested AST node that has `#[cfg]` or `#[cfg_attr]`
172 /// In this case, we use a `ReplaceRange` to replace the entire inner AST node
173 /// with `FlatToken::AttrTarget`, allowing us to perform eager cfg-expansion
174 /// on an `AttrAnnotatedTokenStream`
176 /// 2. When we parse an inner attribute while collecting tokens. We
177 /// remove inner attributes from the token stream entirely, and
178 /// instead track them through the `attrs` field on the AST node.
179 /// This allows us to easily manipulate them (for example, removing
180 /// the first macro inner attribute to invoke a proc-macro).
181 /// When create a `TokenStream`, the inner attributes get inserted
182 /// into the proper place in the token stream.
183 pub type ReplaceRange = (Range<u32>, Vec<(FlatToken, Spacing)>);
185 /// Controls how we capture tokens. Capturing can be expensive,
186 /// so we try to avoid performing capturing in cases where
187 /// we will never need an `AttrAnnotatedTokenStream`
188 #[derive(Copy, Clone)]
190 /// We aren't performing any capturing - this is the default mode.
192 /// We are capturing tokens
197 struct CaptureState {
198 capturing: Capturing,
199 replace_ranges: Vec<ReplaceRange>,
200 inner_attr_ranges: FxHashMap<AttrId, ReplaceRange>,
203 impl<'a> Drop for Parser<'a> {
205 emit_unclosed_delims(&mut self.unclosed_delims, &self.sess);
211 // The current (innermost) frame. `frame` and `stack` could be combined,
212 // but it's faster to have them separately to access `frame` directly
213 // rather than via something like `stack.last().unwrap()` or
214 // `stack[stack.len() - 1]`.
215 frame: TokenCursorFrame,
216 // Additional frames that enclose `frame`.
217 stack: Vec<TokenCursorFrame>,
218 desugar_doc_comments: bool,
219 // Counts the number of calls to `{,inlined_}next`.
220 num_next_calls: usize,
221 // During parsing, we may sometimes need to 'unglue' a
222 // glued token into two component tokens
223 // (e.g. '>>' into '>' and '>), so that the parser
224 // can consume them one at a time. This process
225 // bypasses the normal capturing mechanism
226 // (e.g. `num_next_calls` will not be incremented),
227 // since the 'unglued' tokens due not exist in
228 // the original `TokenStream`.
230 // If we end up consuming both unglued tokens,
231 // then this is not an issue - we'll end up
232 // capturing the single 'glued' token.
234 // However, in certain circumstances, we may
235 // want to capture just the first 'unglued' token.
236 // For example, capturing the `Vec<u8>`
237 // in `Option<Vec<u8>>` requires us to unglue
238 // the trailing `>>` token. The `break_last_token`
239 // field is used to track this token - it gets
240 // appended to the captured stream when
241 // we evaluate a `LazyTokenStream`
242 break_last_token: bool,
246 struct TokenCursorFrame {
247 delim_sp: Option<(DelimToken, DelimSpan)>,
248 tree_cursor: tokenstream::Cursor,
251 impl TokenCursorFrame {
252 fn new(delim_sp: Option<(DelimToken, DelimSpan)>, tts: TokenStream) -> Self {
253 TokenCursorFrame { delim_sp, tree_cursor: tts.into_trees() }
258 fn next(&mut self, desugar_doc_comments: bool) -> (Token, Spacing) {
259 self.inlined_next(desugar_doc_comments)
262 /// This always-inlined version should only be used on hot code paths.
264 fn inlined_next(&mut self, desugar_doc_comments: bool) -> (Token, Spacing) {
266 // FIXME: we currently don't return `NoDelim` open/close delims. To fix #67062 we will
267 // need to, whereupon the `delim != DelimToken::NoDelim` conditions below can be
269 if let Some((tree, spacing)) = self.frame.tree_cursor.next_with_spacing_ref() {
271 &TokenTree::Token(ref token) => match (desugar_doc_comments, token) {
272 (true, &Token { kind: token::DocComment(_, attr_style, data), span }) => {
273 return self.desugar(attr_style, data, span);
275 _ => return (token.clone(), *spacing),
277 &TokenTree::Delimited(sp, delim, ref tts) => {
278 // Set `open_delim` to true here because we deal with it immediately.
279 let frame = TokenCursorFrame::new(Some((delim, sp)), tts.clone());
280 self.stack.push(mem::replace(&mut self.frame, frame));
281 if delim != DelimToken::NoDelim {
282 return (Token::new(token::OpenDelim(delim), sp.open), Spacing::Alone);
284 // No open delimeter to return; continue on to the next iteration.
287 } else if let Some(frame) = self.stack.pop() {
288 if let Some((delim, span)) = self.frame.delim_sp && delim != DelimToken::NoDelim {
290 return (Token::new(token::CloseDelim(delim), span.close), Spacing::Alone);
293 // No close delimiter to return; continue on to the next iteration.
295 return (Token::new(token::Eof, DUMMY_SP), Spacing::Alone);
300 fn desugar(&mut self, attr_style: AttrStyle, data: Symbol, span: Span) -> (Token, Spacing) {
301 // Searches for the occurrences of `"#*` and returns the minimum number of `#`s
302 // required to wrap the text.
303 let mut num_of_hashes = 0;
305 for ch in data.as_str().chars() {
308 '#' if count > 0 => count + 1,
311 num_of_hashes = cmp::max(num_of_hashes, count);
314 let delim_span = DelimSpan::from_single(span);
315 let body = TokenTree::Delimited(
319 TokenTree::token(token::Ident(sym::doc, false), span),
320 TokenTree::token(token::Eq, span),
321 TokenTree::token(TokenKind::lit(token::StrRaw(num_of_hashes), data, None), span),
325 .collect::<TokenStream>(),
328 self.stack.push(mem::replace(
330 TokenCursorFrame::new(
332 if attr_style == AttrStyle::Inner {
333 [TokenTree::token(token::Pound, span), TokenTree::token(token::Not, span), body]
336 .collect::<TokenStream>()
338 [TokenTree::token(token::Pound, span), body]
341 .collect::<TokenStream>()
346 self.next(/* desugar_doc_comments */ false)
350 #[derive(Debug, Clone, PartialEq)]
363 fn to_string(&self) -> String {
365 TokenType::Token(ref t) => format!("`{}`", pprust::token_kind_to_string(t)),
366 TokenType::Keyword(kw) => format!("`{}`", kw),
367 TokenType::Operator => "an operator".to_string(),
368 TokenType::Lifetime => "lifetime".to_string(),
369 TokenType::Ident => "identifier".to_string(),
370 TokenType::Path => "path".to_string(),
371 TokenType::Type => "type".to_string(),
372 TokenType::Const => "a const expression".to_string(),
377 #[derive(Copy, Clone, Debug)]
378 enum TokenExpectType {
383 /// A sequence separator.
385 /// The separator token.
386 sep: Option<TokenKind>,
387 /// `true` if a trailing separator is allowed.
388 trailing_sep_allowed: bool,
392 fn trailing_allowed(t: TokenKind) -> SeqSep {
393 SeqSep { sep: Some(t), trailing_sep_allowed: true }
396 fn none() -> SeqSep {
397 SeqSep { sep: None, trailing_sep_allowed: false }
401 pub enum FollowedByType {
406 fn token_descr_opt(token: &Token) -> Option<&'static str> {
407 Some(match token.kind {
408 _ if token.is_special_ident() => "reserved identifier",
409 _ if token.is_used_keyword() => "keyword",
410 _ if token.is_unused_keyword() => "reserved keyword",
411 token::DocComment(..) => "doc comment",
416 pub(super) fn token_descr(token: &Token) -> String {
417 let token_str = pprust::token_to_string(token);
418 match token_descr_opt(token) {
419 Some(prefix) => format!("{} `{}`", prefix, token_str),
420 _ => format!("`{}`", token_str),
424 impl<'a> Parser<'a> {
428 desugar_doc_comments: bool,
429 subparser_name: Option<&'static str>,
431 let mut parser = Parser {
433 token: Token::dummy(),
434 token_spacing: Spacing::Alone,
435 prev_token: Token::dummy(),
437 restrictions: Restrictions::empty(),
438 expected_tokens: Vec::new(),
439 token_cursor: TokenCursor {
440 frame: TokenCursorFrame::new(None, tokens),
443 desugar_doc_comments,
444 break_last_token: false,
446 desugar_doc_comments,
447 unmatched_angle_bracket_count: 0,
448 max_angle_bracket_count: 0,
449 unclosed_delims: Vec::new(),
450 last_unexpected_token_span: None,
451 last_type_ascription: None,
453 capture_state: CaptureState {
454 capturing: Capturing::No,
455 replace_ranges: Vec::new(),
456 inner_attr_ranges: Default::default(),
458 current_closure: None,
461 // Make parser point to the first token.
467 pub fn unexpected<T>(&mut self) -> PResult<'a, T> {
468 match self.expect_one_of(&[], &[]) {
470 // We can get `Ok(true)` from `recover_closing_delimiter`
471 // which is called in `expected_one_of_not_found`.
472 Ok(_) => FatalError.raise(),
476 /// Expects and consumes the token `t`. Signals an error if the next token is not `t`.
477 pub fn expect(&mut self, t: &TokenKind) -> PResult<'a, bool /* recovered */> {
478 if self.expected_tokens.is_empty() {
479 if self.token == *t {
483 self.unexpected_try_recover(t)
486 self.expect_one_of(slice::from_ref(t), &[])
490 /// Expect next token to be edible or inedible token. If edible,
491 /// then consume it; if inedible, then return without consuming
492 /// anything. Signal a fatal error if next token is unexpected.
493 pub fn expect_one_of(
495 edible: &[TokenKind],
496 inedible: &[TokenKind],
497 ) -> PResult<'a, bool /* recovered */> {
498 if edible.contains(&self.token.kind) {
501 } else if inedible.contains(&self.token.kind) {
502 // leave it in the input
504 } else if self.last_unexpected_token_span == Some(self.token.span) {
507 self.expected_one_of_not_found(edible, inedible)
511 // Public for rustfmt usage.
512 pub fn parse_ident(&mut self) -> PResult<'a, Ident> {
513 self.parse_ident_common(true)
516 fn ident_or_err(&mut self) -> PResult<'a, (Ident, /* is_raw */ bool)> {
517 self.token.ident().ok_or_else(|| match self.prev_token.kind {
518 TokenKind::DocComment(..) => {
519 self.span_err(self.prev_token.span, Error::UselessDocComment)
521 _ => self.expected_ident_found(),
525 fn parse_ident_common(&mut self, recover: bool) -> PResult<'a, Ident> {
526 let (ident, is_raw) = self.ident_or_err()?;
527 if !is_raw && ident.is_reserved() {
528 let mut err = self.expected_ident_found();
539 /// Checks if the next token is `tok`, and returns `true` if so.
541 /// This method will automatically add `tok` to `expected_tokens` if `tok` is not
543 fn check(&mut self, tok: &TokenKind) -> bool {
544 let is_present = self.token == *tok;
546 self.expected_tokens.push(TokenType::Token(tok.clone()));
551 /// Consumes a token 'tok' if it exists. Returns whether the given token was present.
552 pub fn eat(&mut self, tok: &TokenKind) -> bool {
553 let is_present = self.check(tok);
560 /// If the next token is the given keyword, returns `true` without eating it.
561 /// An expectation is also added for diagnostics purposes.
562 fn check_keyword(&mut self, kw: Symbol) -> bool {
563 self.expected_tokens.push(TokenType::Keyword(kw));
564 self.token.is_keyword(kw)
567 /// If the next token is the given keyword, eats it and returns `true`.
568 /// Otherwise, returns `false`. An expectation is also added for diagnostics purposes.
569 // Public for rustfmt usage.
570 pub fn eat_keyword(&mut self, kw: Symbol) -> bool {
571 if self.check_keyword(kw) {
579 fn eat_keyword_noexpect(&mut self, kw: Symbol) -> bool {
580 if self.token.is_keyword(kw) {
588 /// If the given word is not a keyword, signals an error.
589 /// If the next token is not the given word, signals an error.
590 /// Otherwise, eats it.
591 fn expect_keyword(&mut self, kw: Symbol) -> PResult<'a, ()> {
592 if !self.eat_keyword(kw) { self.unexpected() } else { Ok(()) }
595 /// Is the given keyword `kw` followed by a non-reserved identifier?
596 fn is_kw_followed_by_ident(&self, kw: Symbol) -> bool {
597 self.token.is_keyword(kw) && self.look_ahead(1, |t| t.is_ident() && !t.is_reserved_ident())
600 fn check_or_expected(&mut self, ok: bool, typ: TokenType) -> bool {
604 self.expected_tokens.push(typ);
609 fn check_ident(&mut self) -> bool {
610 self.check_or_expected(self.token.is_ident(), TokenType::Ident)
613 fn check_path(&mut self) -> bool {
614 self.check_or_expected(self.token.is_path_start(), TokenType::Path)
617 fn check_type(&mut self) -> bool {
618 self.check_or_expected(self.token.can_begin_type(), TokenType::Type)
621 fn check_const_arg(&mut self) -> bool {
622 self.check_or_expected(self.token.can_begin_const_arg(), TokenType::Const)
625 fn check_inline_const(&self, dist: usize) -> bool {
626 self.is_keyword_ahead(dist, &[kw::Const])
627 && self.look_ahead(dist + 1, |t| match t.kind {
628 token::Interpolated(ref nt) => matches!(**nt, token::NtBlock(..)),
629 token::OpenDelim(DelimToken::Brace) => true,
634 /// Checks to see if the next token is either `+` or `+=`.
635 /// Otherwise returns `false`.
636 fn check_plus(&mut self) -> bool {
637 self.check_or_expected(
638 self.token.is_like_plus(),
639 TokenType::Token(token::BinOp(token::Plus)),
643 /// Eats the expected token if it's present possibly breaking
644 /// compound tokens like multi-character operators in process.
645 /// Returns `true` if the token was eaten.
646 fn break_and_eat(&mut self, expected: TokenKind) -> bool {
647 if self.token.kind == expected {
651 match self.token.kind.break_two_token_op() {
652 Some((first, second)) if first == expected => {
653 let first_span = self.sess.source_map().start_point(self.token.span);
654 let second_span = self.token.span.with_lo(first_span.hi());
655 self.token = Token::new(first, first_span);
656 // Keep track of this token - if we end token capturing now,
657 // we'll want to append this token to the captured stream.
659 // If we consume any additional tokens, then this token
660 // is not needed (we'll capture the entire 'glued' token),
661 // and `bump` will set this field to `None`
662 self.token_cursor.break_last_token = true;
663 // Use the spacing of the glued token as the spacing
664 // of the unglued second token.
665 self.bump_with((Token::new(second, second_span), self.token_spacing));
669 self.expected_tokens.push(TokenType::Token(expected));
675 /// Eats `+` possibly breaking tokens like `+=` in process.
676 fn eat_plus(&mut self) -> bool {
677 self.break_and_eat(token::BinOp(token::Plus))
680 /// Eats `&` possibly breaking tokens like `&&` in process.
681 /// Signals an error if `&` is not eaten.
682 fn expect_and(&mut self) -> PResult<'a, ()> {
683 if self.break_and_eat(token::BinOp(token::And)) { Ok(()) } else { self.unexpected() }
686 /// Eats `|` possibly breaking tokens like `||` in process.
687 /// Signals an error if `|` was not eaten.
688 fn expect_or(&mut self) -> PResult<'a, ()> {
689 if self.break_and_eat(token::BinOp(token::Or)) { Ok(()) } else { self.unexpected() }
692 /// Eats `<` possibly breaking tokens like `<<` in process.
693 fn eat_lt(&mut self) -> bool {
694 let ate = self.break_and_eat(token::Lt);
696 // See doc comment for `unmatched_angle_bracket_count`.
697 self.unmatched_angle_bracket_count += 1;
698 self.max_angle_bracket_count += 1;
699 debug!("eat_lt: (increment) count={:?}", self.unmatched_angle_bracket_count);
704 /// Eats `<` possibly breaking tokens like `<<` in process.
705 /// Signals an error if `<` was not eaten.
706 fn expect_lt(&mut self) -> PResult<'a, ()> {
707 if self.eat_lt() { Ok(()) } else { self.unexpected() }
710 /// Eats `>` possibly breaking tokens like `>>` in process.
711 /// Signals an error if `>` was not eaten.
712 fn expect_gt(&mut self) -> PResult<'a, ()> {
713 if self.break_and_eat(token::Gt) {
714 // See doc comment for `unmatched_angle_bracket_count`.
715 if self.unmatched_angle_bracket_count > 0 {
716 self.unmatched_angle_bracket_count -= 1;
717 debug!("expect_gt: (decrement) count={:?}", self.unmatched_angle_bracket_count);
725 fn expect_any_with_type(&mut self, kets: &[&TokenKind], expect: TokenExpectType) -> bool {
726 kets.iter().any(|k| match expect {
727 TokenExpectType::Expect => self.check(k),
728 TokenExpectType::NoExpect => self.token == **k,
732 fn parse_seq_to_before_tokens<T>(
736 expect: TokenExpectType,
737 mut f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
738 ) -> PResult<'a, (Vec<T>, bool /* trailing */, bool /* recovered */)> {
739 let mut first = true;
740 let mut recovered = false;
741 let mut trailing = false;
743 let unclosed_delims = !self.unclosed_delims.is_empty();
745 while !self.expect_any_with_type(kets, expect) {
746 if let token::CloseDelim(..) | token::Eof = self.token.kind {
749 if let Some(ref t) = sep.sep {
753 match self.expect(t) {
755 self.current_closure.take();
758 self.current_closure.take();
762 Err(mut expect_err) => {
763 let sp = self.prev_token.span.shrink_to_hi();
764 let token_str = pprust::token_kind_to_string(t);
766 match self.current_closure.take() {
767 Some(closure_spans) if self.token.kind == TokenKind::Semi => {
768 // Finding a semicolon instead of a comma
769 // after a closure body indicates that the
770 // closure body may be a block but the user
771 // forgot to put braces around its
774 self.recover_missing_braces_around_closure_body(
783 // Attempt to keep parsing if it was a similar separator.
784 if let Some(ref tokens) = t.similar_tokens() {
785 if tokens.contains(&self.token.kind) && !unclosed_delims {
792 // If this was a missing `@` in a binding pattern
793 // bail with a suggestion
794 // https://github.com/rust-lang/rust/issues/72373
795 if self.prev_token.is_ident() && self.token.kind == token::DotDot {
797 "if you meant to bind the contents of \
798 the rest of the array pattern into `{}`, use `@`",
799 pprust::token_to_string(&self.prev_token)
802 .span_suggestion_verbose(
803 self.prev_token.span.shrink_to_hi().until(self.token.span),
806 Applicability::MaybeIncorrect,
812 // Attempt to keep parsing if it was an omitted separator.
815 // Parsed successfully, therefore most probably the code only
816 // misses a separator.
818 .span_suggestion_short(
820 &format!("missing `{}`", token_str),
822 Applicability::MaybeIncorrect,
830 // Parsing failed, therefore it must be something more serious
831 // than just a missing separator.
842 if sep.trailing_sep_allowed && self.expect_any_with_type(kets, expect) {
851 Ok((v, trailing, recovered))
854 fn recover_missing_braces_around_closure_body(
856 closure_spans: ClosureSpans,
857 mut expect_err: DiagnosticBuilder<'_, ErrorGuaranteed>,
858 ) -> PResult<'a, ()> {
859 let initial_semicolon = self.token.span;
861 while self.eat(&TokenKind::Semi) {
862 let _ = self.parse_stmt(ForceCollect::Yes)?;
865 expect_err.set_primary_message(
866 "closure bodies that contain statements must be surrounded by braces",
869 let preceding_pipe_span = closure_spans.closing_pipe;
870 let following_token_span = self.token.span;
872 let mut first_note = MultiSpan::from(vec![initial_semicolon]);
873 first_note.push_span_label(
875 "this `;` turns the preceding closure into a statement".to_string(),
877 first_note.push_span_label(
879 "this expression is a statement because of the trailing semicolon".to_string(),
881 expect_err.span_note(first_note, "statement found outside of a block");
883 let mut second_note = MultiSpan::from(vec![closure_spans.whole_closure]);
884 second_note.push_span_label(
885 closure_spans.whole_closure,
886 "this is the parsed closure...".to_string(),
888 second_note.push_span_label(
889 following_token_span,
890 "...but likely you meant the closure to end here".to_string(),
892 expect_err.span_note(second_note, "the closure body may be incorrectly delimited");
894 expect_err.set_span(vec![preceding_pipe_span, following_token_span]);
896 let opening_suggestion_str = " {".to_string();
897 let closing_suggestion_str = "}".to_string();
899 expect_err.multipart_suggestion(
902 (preceding_pipe_span.shrink_to_hi(), opening_suggestion_str),
903 (following_token_span.shrink_to_lo(), closing_suggestion_str),
905 Applicability::MaybeIncorrect,
913 /// Parses a sequence, not including the closing delimiter. The function
914 /// `f` must consume tokens until reaching the next separator or
916 fn parse_seq_to_before_end<T>(
920 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
921 ) -> PResult<'a, (Vec<T>, bool, bool)> {
922 self.parse_seq_to_before_tokens(&[ket], sep, TokenExpectType::Expect, f)
925 /// Parses a sequence, including the closing delimiter. The function
926 /// `f` must consume tokens until reaching the next separator or
928 fn parse_seq_to_end<T>(
932 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
933 ) -> PResult<'a, (Vec<T>, bool /* trailing */)> {
934 let (val, trailing, recovered) = self.parse_seq_to_before_end(ket, sep, f)?;
941 /// Parses a sequence, including the closing delimiter. The function
942 /// `f` must consume tokens until reaching the next separator or
944 fn parse_unspanned_seq<T>(
949 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
950 ) -> PResult<'a, (Vec<T>, bool)> {
952 self.parse_seq_to_end(ket, sep, f)
955 fn parse_delim_comma_seq<T>(
958 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
959 ) -> PResult<'a, (Vec<T>, bool)> {
960 self.parse_unspanned_seq(
961 &token::OpenDelim(delim),
962 &token::CloseDelim(delim),
963 SeqSep::trailing_allowed(token::Comma),
968 fn parse_paren_comma_seq<T>(
970 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
971 ) -> PResult<'a, (Vec<T>, bool)> {
972 self.parse_delim_comma_seq(token::Paren, f)
975 /// Advance the parser by one token using provided token as the next one.
976 fn bump_with(&mut self, next: (Token, Spacing)) {
977 self.inlined_bump_with(next)
980 /// This always-inlined version should only be used on hot code paths.
982 fn inlined_bump_with(&mut self, (next_token, next_spacing): (Token, Spacing)) {
983 // Update the current and previous tokens.
984 self.prev_token = mem::replace(&mut self.token, next_token);
985 self.token_spacing = next_spacing;
988 self.expected_tokens.clear();
991 /// Advance the parser by one token.
992 pub fn bump(&mut self) {
993 // Note: destructuring here would give nicer code, but it was found in #96210 to be slower
994 // than `.0`/`.1` access.
995 let mut next = self.token_cursor.inlined_next(self.desugar_doc_comments);
996 self.token_cursor.num_next_calls += 1;
997 // We've retrieved an token from the underlying
998 // cursor, so we no longer need to worry about
999 // an unglued token. See `break_and_eat` for more details
1000 self.token_cursor.break_last_token = false;
1001 if next.0.span.is_dummy() {
1002 // Tweak the location for better diagnostics, but keep syntactic context intact.
1003 let fallback_span = self.token.span;
1004 next.0.span = fallback_span.with_ctxt(next.0.span.ctxt());
1006 debug_assert!(!matches!(
1008 token::OpenDelim(token::NoDelim) | token::CloseDelim(token::NoDelim)
1010 self.inlined_bump_with(next)
1013 /// Look-ahead `dist` tokens of `self.token` and get access to that token there.
1014 /// When `dist == 0` then the current token is looked at.
1015 pub fn look_ahead<R>(&self, dist: usize, looker: impl FnOnce(&Token) -> R) -> R {
1017 return looker(&self.token);
1020 let frame = &self.token_cursor.frame;
1021 if let Some((delim, span)) = frame.delim_sp && delim != DelimToken::NoDelim {
1022 let all_normal = (0..dist).all(|i| {
1023 let token = frame.tree_cursor.look_ahead(i);
1024 !matches!(token, Some(TokenTree::Delimited(_, DelimToken::NoDelim, _)))
1027 return match frame.tree_cursor.look_ahead(dist - 1) {
1028 Some(tree) => match tree {
1029 TokenTree::Token(token) => looker(token),
1030 TokenTree::Delimited(dspan, delim, _) => {
1031 looker(&Token::new(token::OpenDelim(*delim), dspan.open))
1034 None => looker(&Token::new(token::CloseDelim(delim), span.close)),
1039 let mut cursor = self.token_cursor.clone();
1041 let mut token = Token::dummy();
1043 token = cursor.next(/* desugar_doc_comments */ false).0;
1046 token::OpenDelim(token::NoDelim) | token::CloseDelim(token::NoDelim)
1052 return looker(&token);
1055 /// Returns whether any of the given keywords are `dist` tokens ahead of the current one.
1056 fn is_keyword_ahead(&self, dist: usize, kws: &[Symbol]) -> bool {
1057 self.look_ahead(dist, |t| kws.iter().any(|&kw| t.is_keyword(kw)))
1060 /// Parses asyncness: `async` or nothing.
1061 fn parse_asyncness(&mut self) -> Async {
1062 if self.eat_keyword(kw::Async) {
1063 let span = self.prev_token.uninterpolated_span();
1064 Async::Yes { span, closure_id: DUMMY_NODE_ID, return_impl_trait_id: DUMMY_NODE_ID }
1070 /// Parses unsafety: `unsafe` or nothing.
1071 fn parse_unsafety(&mut self) -> Unsafe {
1072 if self.eat_keyword(kw::Unsafe) {
1073 Unsafe::Yes(self.prev_token.uninterpolated_span())
1079 /// Parses constness: `const` or nothing.
1080 fn parse_constness(&mut self) -> Const {
1081 // Avoid const blocks to be parsed as const items
1082 if self.look_ahead(1, |t| t != &token::OpenDelim(DelimToken::Brace))
1083 && self.eat_keyword(kw::Const)
1085 Const::Yes(self.prev_token.uninterpolated_span())
1091 /// Parses inline const expressions.
1092 fn parse_const_block(&mut self, span: Span, pat: bool) -> PResult<'a, P<Expr>> {
1094 self.sess.gated_spans.gate(sym::inline_const_pat, span);
1096 self.sess.gated_spans.gate(sym::inline_const, span);
1098 self.eat_keyword(kw::Const);
1099 let (attrs, blk) = self.parse_inner_attrs_and_block()?;
1100 let anon_const = AnonConst {
1102 value: self.mk_expr(blk.span, ExprKind::Block(blk, None), AttrVec::new()),
1104 let blk_span = anon_const.value.span;
1105 Ok(self.mk_expr(span.to(blk_span), ExprKind::ConstBlock(anon_const), AttrVec::from(attrs)))
1108 /// Parses mutability (`mut` or nothing).
1109 fn parse_mutability(&mut self) -> Mutability {
1110 if self.eat_keyword(kw::Mut) { Mutability::Mut } else { Mutability::Not }
1113 /// Possibly parses mutability (`const` or `mut`).
1114 fn parse_const_or_mut(&mut self) -> Option<Mutability> {
1115 if self.eat_keyword(kw::Mut) {
1116 Some(Mutability::Mut)
1117 } else if self.eat_keyword(kw::Const) {
1118 Some(Mutability::Not)
1124 fn parse_field_name(&mut self) -> PResult<'a, Ident> {
1125 if let token::Literal(token::Lit { kind: token::Integer, symbol, suffix }) = self.token.kind
1127 self.expect_no_suffix(self.token.span, "a tuple index", suffix);
1129 Ok(Ident::new(symbol, self.prev_token.span))
1131 self.parse_ident_common(true)
1135 fn parse_mac_args(&mut self) -> PResult<'a, P<MacArgs>> {
1136 self.parse_mac_args_common(true).map(P)
1139 fn parse_attr_args(&mut self) -> PResult<'a, MacArgs> {
1140 self.parse_mac_args_common(false)
1143 fn parse_mac_args_common(&mut self, delimited_only: bool) -> PResult<'a, MacArgs> {
1145 if self.check(&token::OpenDelim(DelimToken::Paren))
1146 || self.check(&token::OpenDelim(DelimToken::Bracket))
1147 || self.check(&token::OpenDelim(DelimToken::Brace))
1149 match self.parse_token_tree() {
1150 TokenTree::Delimited(dspan, delim, tokens) =>
1151 // We've confirmed above that there is a delimiter so unwrapping is OK.
1153 MacArgs::Delimited(dspan, MacDelimiter::from_token(delim).unwrap(), tokens)
1155 _ => unreachable!(),
1157 } else if !delimited_only {
1158 if self.eat(&token::Eq) {
1159 let eq_span = self.prev_token.span;
1161 // Collect tokens because they are used during lowering to HIR.
1162 let expr = self.parse_expr_force_collect()?;
1163 let span = expr.span;
1165 let token_kind = token::Interpolated(Lrc::new(token::NtExpr(expr)));
1166 MacArgs::Eq(eq_span, Token::new(token_kind, span))
1171 return self.unexpected();
1176 fn parse_or_use_outer_attributes(
1178 already_parsed_attrs: Option<AttrWrapper>,
1179 ) -> PResult<'a, AttrWrapper> {
1180 if let Some(attrs) = already_parsed_attrs {
1183 self.parse_outer_attributes()
1187 /// Parses a single token tree from the input.
1188 pub(crate) fn parse_token_tree(&mut self) -> TokenTree {
1189 match self.token.kind {
1190 token::OpenDelim(..) => {
1191 // Grab the tokens from this frame.
1192 let frame = &self.token_cursor.frame;
1193 let stream = frame.tree_cursor.stream.clone();
1194 let (delim, span) = frame.delim_sp.unwrap();
1196 // Advance the token cursor through the entire delimited
1197 // sequence. After getting the `OpenDelim` we are *within* the
1198 // delimited sequence, i.e. at depth `d`. After getting the
1199 // matching `CloseDelim` we are *after* the delimited sequence,
1200 // i.e. at depth `d - 1`.
1201 let target_depth = self.token_cursor.stack.len() - 1;
1203 // Advance one token at a time, so `TokenCursor::next()`
1204 // can capture these tokens if necessary.
1206 if self.token_cursor.stack.len() == target_depth {
1207 debug_assert!(matches!(self.token.kind, token::CloseDelim(_)));
1212 // Consume close delimiter
1214 TokenTree::Delimited(span, delim, stream)
1216 token::CloseDelim(_) | token::Eof => unreachable!(),
1219 TokenTree::Token(self.prev_token.clone())
1224 /// Parses a stream of tokens into a list of `TokenTree`s, up to EOF.
1225 pub fn parse_all_token_trees(&mut self) -> PResult<'a, Vec<TokenTree>> {
1226 let mut tts = Vec::new();
1227 while self.token != token::Eof {
1228 tts.push(self.parse_token_tree());
1233 pub fn parse_tokens(&mut self) -> TokenStream {
1234 let mut result = Vec::new();
1236 match self.token.kind {
1237 token::Eof | token::CloseDelim(..) => break,
1238 _ => result.push(self.parse_token_tree().into()),
1241 TokenStream::new(result)
1244 /// Evaluates the closure with restrictions in place.
1246 /// Afters the closure is evaluated, restrictions are reset.
1247 fn with_res<T>(&mut self, res: Restrictions, f: impl FnOnce(&mut Self) -> T) -> T {
1248 let old = self.restrictions;
1249 self.restrictions = res;
1251 self.restrictions = old;
1255 fn is_crate_vis(&self) -> bool {
1256 self.token.is_keyword(kw::Crate) && self.look_ahead(1, |t| t != &token::ModSep)
1259 /// Parses `pub`, `pub(crate)` and `pub(in path)` plus shortcuts `crate` for `pub(crate)`,
1260 /// `pub(self)` for `pub(in self)` and `pub(super)` for `pub(in super)`.
1261 /// If the following element can't be a tuple (i.e., it's a function definition), then
1262 /// it's not a tuple struct field), and the contents within the parentheses aren't valid,
1263 /// so emit a proper diagnostic.
1264 // Public for rustfmt usage.
1265 pub fn parse_visibility(&mut self, fbt: FollowedByType) -> PResult<'a, Visibility> {
1266 maybe_whole!(self, NtVis, |x| x.into_inner());
1268 self.expected_tokens.push(TokenType::Keyword(kw::Crate));
1269 if self.is_crate_vis() {
1270 self.bump(); // `crate`
1271 self.sess.gated_spans.gate(sym::crate_visibility_modifier, self.prev_token.span);
1272 return Ok(Visibility {
1273 span: self.prev_token.span,
1274 kind: VisibilityKind::Crate(CrateSugar::JustCrate),
1279 if !self.eat_keyword(kw::Pub) {
1280 // We need a span for our `Spanned<VisibilityKind>`, but there's inherently no
1281 // keyword to grab a span from for inherited visibility; an empty span at the
1282 // beginning of the current token would seem to be the "Schelling span".
1283 return Ok(Visibility {
1284 span: self.token.span.shrink_to_lo(),
1285 kind: VisibilityKind::Inherited,
1289 let lo = self.prev_token.span;
1291 if self.check(&token::OpenDelim(token::Paren)) {
1292 // We don't `self.bump()` the `(` yet because this might be a struct definition where
1293 // `()` or a tuple might be allowed. For example, `struct Struct(pub (), pub (usize));`.
1294 // Because of this, we only `bump` the `(` if we're assured it is appropriate to do so
1295 // by the following tokens.
1296 if self.is_keyword_ahead(1, &[kw::Crate]) && self.look_ahead(2, |t| t != &token::ModSep)
1297 // account for `pub(crate::foo)`
1299 // Parse `pub(crate)`.
1301 self.bump(); // `crate`
1302 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1303 let vis = VisibilityKind::Crate(CrateSugar::PubCrate);
1304 return Ok(Visibility {
1305 span: lo.to(self.prev_token.span),
1309 } else if self.is_keyword_ahead(1, &[kw::In]) {
1310 // Parse `pub(in path)`.
1312 self.bump(); // `in`
1313 let path = self.parse_path(PathStyle::Mod)?; // `path`
1314 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1315 let vis = VisibilityKind::Restricted { path: P(path), id: ast::DUMMY_NODE_ID };
1316 return Ok(Visibility {
1317 span: lo.to(self.prev_token.span),
1321 } else if self.look_ahead(2, |t| t == &token::CloseDelim(token::Paren))
1322 && self.is_keyword_ahead(1, &[kw::Super, kw::SelfLower])
1324 // Parse `pub(self)` or `pub(super)`.
1326 let path = self.parse_path(PathStyle::Mod)?; // `super`/`self`
1327 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1328 let vis = VisibilityKind::Restricted { path: P(path), id: ast::DUMMY_NODE_ID };
1329 return Ok(Visibility {
1330 span: lo.to(self.prev_token.span),
1334 } else if let FollowedByType::No = fbt {
1335 // Provide this diagnostic if a type cannot follow;
1336 // in particular, if this is not a tuple struct.
1337 self.recover_incorrect_vis_restriction()?;
1338 // Emit diagnostic, but continue with public visibility.
1342 Ok(Visibility { span: lo, kind: VisibilityKind::Public, tokens: None })
1345 /// Recovery for e.g. `pub(something) fn ...` or `struct X { pub(something) y: Z }`
1346 fn recover_incorrect_vis_restriction(&mut self) -> PResult<'a, ()> {
1348 let path = self.parse_path(PathStyle::Mod)?;
1349 self.expect(&token::CloseDelim(token::Paren))?; // `)`
1351 let msg = "incorrect visibility restriction";
1352 let suggestion = r##"some possible visibility restrictions are:
1353 `pub(crate)`: visible only on the current crate
1354 `pub(super)`: visible only in the current module's parent
1355 `pub(in path::to::module)`: visible only on the specified path"##;
1357 let path_str = pprust::path_to_string(&path);
1359 struct_span_err!(self.sess.span_diagnostic, path.span, E0704, "{}", msg)
1363 &format!("make this visible only to module `{}` with `in`", path_str),
1364 format!("in {}", path_str),
1365 Applicability::MachineApplicable,
1372 /// Parses `extern string_literal?`.
1373 fn parse_extern(&mut self) -> Extern {
1374 if self.eat_keyword(kw::Extern) { Extern::from_abi(self.parse_abi()) } else { Extern::None }
1377 /// Parses a string literal as an ABI spec.
1378 fn parse_abi(&mut self) -> Option<StrLit> {
1379 match self.parse_str_lit() {
1380 Ok(str_lit) => Some(str_lit),
1381 Err(Some(lit)) => match lit.kind {
1382 ast::LitKind::Err(_) => None,
1384 self.struct_span_err(lit.span, "non-string ABI literal")
1387 "specify the ABI with a string literal",
1388 "\"C\"".to_string(),
1389 Applicability::MaybeIncorrect,
1399 pub fn collect_tokens_no_attrs<R: AstLike>(
1401 f: impl FnOnce(&mut Self) -> PResult<'a, R>,
1402 ) -> PResult<'a, R> {
1403 // The only reason to call `collect_tokens_no_attrs` is if you want tokens, so use
1404 // `ForceCollect::Yes`
1405 self.collect_tokens_trailing_token(
1406 AttrWrapper::empty(),
1408 |this, _attrs| Ok((f(this)?, TrailingToken::None)),
1413 fn is_import_coupler(&mut self) -> bool {
1414 self.check(&token::ModSep)
1415 && self.look_ahead(1, |t| {
1416 *t == token::OpenDelim(token::Brace) || *t == token::BinOp(token::Star)
1420 pub fn clear_expected_tokens(&mut self) {
1421 self.expected_tokens.clear();
1425 crate fn make_unclosed_delims_error(
1426 unmatched: UnmatchedBrace,
1428 ) -> Option<DiagnosticBuilder<'_, ErrorGuaranteed>> {
1429 // `None` here means an `Eof` was found. We already emit those errors elsewhere, we add them to
1430 // `unmatched_braces` only for error recovery in the `Parser`.
1431 let found_delim = unmatched.found_delim?;
1432 let span: MultiSpan = if let Some(sp) = unmatched.unclosed_span {
1433 vec![unmatched.found_span, sp].into()
1435 unmatched.found_span.into()
1437 let mut err = sess.span_diagnostic.struct_span_err(
1440 "mismatched closing delimiter: `{}`",
1441 pprust::token_kind_to_string(&token::CloseDelim(found_delim)),
1444 err.span_label(unmatched.found_span, "mismatched closing delimiter");
1445 if let Some(sp) = unmatched.candidate_span {
1446 err.span_label(sp, "closing delimiter possibly meant for this");
1448 if let Some(sp) = unmatched.unclosed_span {
1449 err.span_label(sp, "unclosed delimiter");
1454 pub fn emit_unclosed_delims(unclosed_delims: &mut Vec<UnmatchedBrace>, sess: &ParseSess) {
1455 *sess.reached_eof.borrow_mut() |=
1456 unclosed_delims.iter().any(|unmatched_delim| unmatched_delim.found_delim.is_none());
1457 for unmatched in unclosed_delims.drain(..) {
1458 if let Some(mut e) = make_unclosed_delims_error(unmatched, sess) {
1464 /// A helper struct used when building an `AttrAnnotatedTokenStream` from
1465 /// a `LazyTokenStream`. Both delimiter and non-delimited tokens
1466 /// are stored as `FlatToken::Token`. A vector of `FlatToken`s
1467 /// is then 'parsed' to build up an `AttrAnnotatedTokenStream` with nested
1468 /// `AttrAnnotatedTokenTree::Delimited` tokens
1469 #[derive(Debug, Clone)]
1470 pub enum FlatToken {
1471 /// A token - this holds both delimiter (e.g. '{' and '}')
1472 /// and non-delimiter tokens
1474 /// Holds the `AttributesData` for an AST node. The
1475 /// `AttributesData` is inserted directly into the
1476 /// constructed `AttrAnnotatedTokenStream` as
1477 /// an `AttrAnnotatedTokenTree::Attributes`
1478 AttrTarget(AttributesData),
1479 /// A special 'empty' token that is ignored during the conversion
1480 /// to an `AttrAnnotatedTokenStream`. This is used to simplify the
1481 /// handling of replace ranges.