13 use crate::lexer::UnmatchedBrace;
14 pub use attr_wrapper::AttrWrapper;
15 pub use diagnostics::AttemptLocalParseRecovery;
16 pub(crate) use item::FnParseMode;
17 pub use pat::{CommaRecoveryMode, RecoverColon, RecoverComma};
18 pub use path::PathStyle;
20 use rustc_ast::ptr::P;
21 use rustc_ast::token::{self, Delimiter, Nonterminal, Token, TokenKind};
22 use rustc_ast::tokenstream::AttributesData;
23 use rustc_ast::tokenstream::{self, DelimSpan, Spacing};
24 use rustc_ast::tokenstream::{TokenStream, TokenTree};
25 use rustc_ast::util::case::Case;
26 use rustc_ast::AttrId;
27 use rustc_ast::DUMMY_NODE_ID;
28 use rustc_ast::{self as ast, AnonConst, AttrStyle, Const, DelimArgs, Extern};
29 use rustc_ast::{Async, AttrArgs, AttrArgsEq, Expr, ExprKind, MacDelimiter, Mutability, StrLit};
30 use rustc_ast::{HasAttrs, HasTokens, Unsafe, Visibility, VisibilityKind};
31 use rustc_ast_pretty::pprust;
32 use rustc_data_structures::fx::FxHashMap;
33 use rustc_errors::PResult;
35 Applicability, DiagnosticBuilder, ErrorGuaranteed, FatalError, IntoDiagnostic, MultiSpan,
37 use rustc_session::parse::ParseSess;
38 use rustc_span::source_map::{Span, DUMMY_SP};
39 use rustc_span::symbol::{kw, sym, Ident, Symbol};
42 use std::{cmp, mem, slice};
45 DocCommentDoesNotDocumentAnything, IncorrectVisibilityRestriction, MismatchedClosingDelimiter,
50 struct Restrictions: u8 {
51 const STMT_EXPR = 1 << 0;
52 const NO_STRUCT_LITERAL = 1 << 1;
53 const CONST_EXPR = 1 << 2;
54 const ALLOW_LET = 1 << 3;
58 #[derive(Clone, Copy, PartialEq, Debug)]
65 #[derive(Clone, Copy, PartialEq, Debug)]
71 /// Whether or not we should force collection of tokens for an AST node,
72 /// regardless of whether or not it has attributes
73 #[derive(Clone, Copy, PartialEq)]
74 pub enum ForceCollect {
79 #[derive(Debug, Eq, PartialEq)]
80 pub enum TrailingToken {
84 /// If the trailing token is a comma, then capture it
85 /// Otherwise, ignore the trailing token
89 /// Like `maybe_whole_expr`, but for things other than expressions.
91 macro_rules! maybe_whole {
92 ($p:expr, $constructor:ident, |$x:ident| $e:expr) => {
93 if let token::Interpolated(nt) = &$p.token.kind {
94 if let token::$constructor(x) = &**nt {
103 /// If the next tokens are ill-formed `$ty::` recover them as `<$ty>::`.
105 macro_rules! maybe_recover_from_interpolated_ty_qpath {
106 ($self: expr, $allow_qpath_recovery: expr) => {
107 if $allow_qpath_recovery
108 && $self.may_recover()
109 && $self.look_ahead(1, |t| t == &token::ModSep)
110 && let token::Interpolated(nt) = &$self.token.kind
111 && let token::NtTy(ty) = &**nt
115 return $self.maybe_recover_from_bad_qpath_stage_2($self.prev_token.span, ty);
120 #[derive(Clone, Copy)]
127 pub struct Parser<'a> {
128 pub sess: &'a ParseSess,
129 /// The current token.
131 /// The spacing for the current token
132 pub token_spacing: Spacing,
133 /// The previous token.
134 pub prev_token: Token,
135 pub capture_cfg: bool,
136 restrictions: Restrictions,
137 expected_tokens: Vec<TokenType>,
138 // Important: This must only be advanced from `bump` to ensure that
139 // `token_cursor.num_next_calls` is updated properly.
140 token_cursor: TokenCursor,
141 desugar_doc_comments: bool,
142 /// This field is used to keep track of how many left angle brackets we have seen. This is
143 /// required in order to detect extra leading left angle brackets (`<` characters) and error
146 /// See the comments in the `parse_path_segment` function for more details.
147 unmatched_angle_bracket_count: u32,
148 max_angle_bracket_count: u32,
149 /// A list of all unclosed delimiters found by the lexer. If an entry is used for error recovery
150 /// it gets removed from here. Every entry left at the end gets emitted as an independent
152 pub(super) unclosed_delims: Vec<UnmatchedBrace>,
153 last_unexpected_token_span: Option<Span>,
154 /// Span pointing at the `:` for the last type ascription the parser has seen, and whether it
155 /// looked like it could have been a mistyped path or literal `Option:Some(42)`).
156 pub last_type_ascription: Option<(Span, bool /* likely path typo */)>,
157 /// If present, this `Parser` is not parsing Rust code but rather a macro call.
158 subparser_name: Option<&'static str>,
159 capture_state: CaptureState,
160 /// This allows us to recover when the user forget to add braces around
161 /// multiple statements in the closure body.
162 pub current_closure: Option<ClosureSpans>,
163 /// Whether the parser is allowed to do recovery.
164 /// This is disabled when parsing macro arguments, see #103534
165 pub recovery: Recovery,
168 // This type is used a lot, e.g. it's cloned when matching many declarative macro rules with nonterminals. Make sure
169 // it doesn't unintentionally get bigger.
170 #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))]
171 rustc_data_structures::static_assert_size!(Parser<'_>, 336);
173 /// Stores span information about a closure.
175 pub struct ClosureSpans {
176 pub whole_closure: Span,
177 pub closing_pipe: Span,
181 /// Indicates a range of tokens that should be replaced by
182 /// the tokens in the provided vector. This is used in two
183 /// places during token collection:
185 /// 1. During the parsing of an AST node that may have a `#[derive]`
186 /// attribute, we parse a nested AST node that has `#[cfg]` or `#[cfg_attr]`
187 /// In this case, we use a `ReplaceRange` to replace the entire inner AST node
188 /// with `FlatToken::AttrTarget`, allowing us to perform eager cfg-expansion
189 /// on an `AttrTokenStream`.
191 /// 2. When we parse an inner attribute while collecting tokens. We
192 /// remove inner attributes from the token stream entirely, and
193 /// instead track them through the `attrs` field on the AST node.
194 /// This allows us to easily manipulate them (for example, removing
195 /// the first macro inner attribute to invoke a proc-macro).
196 /// When create a `TokenStream`, the inner attributes get inserted
197 /// into the proper place in the token stream.
198 pub type ReplaceRange = (Range<u32>, Vec<(FlatToken, Spacing)>);
200 /// Controls how we capture tokens. Capturing can be expensive,
201 /// so we try to avoid performing capturing in cases where
202 /// we will never need an `AttrTokenStream`.
203 #[derive(Copy, Clone)]
205 /// We aren't performing any capturing - this is the default mode.
207 /// We are capturing tokens
212 struct CaptureState {
213 capturing: Capturing,
214 replace_ranges: Vec<ReplaceRange>,
215 inner_attr_ranges: FxHashMap<AttrId, ReplaceRange>,
218 impl<'a> Drop for Parser<'a> {
220 emit_unclosed_delims(&mut self.unclosed_delims, &self.sess);
226 // The current (innermost) frame. `frame` and `stack` could be combined,
227 // but it's faster to keep them separate and access `frame` directly
228 // rather than via something like `stack.last().unwrap()` or
229 // `stack[stack.len() - 1]`.
230 frame: TokenCursorFrame,
231 // Additional frames that enclose `frame`.
232 stack: Vec<TokenCursorFrame>,
233 desugar_doc_comments: bool,
234 // Counts the number of calls to `{,inlined_}next`.
235 num_next_calls: usize,
236 // During parsing, we may sometimes need to 'unglue' a
237 // glued token into two component tokens
238 // (e.g. '>>' into '>' and '>), so that the parser
239 // can consume them one at a time. This process
240 // bypasses the normal capturing mechanism
241 // (e.g. `num_next_calls` will not be incremented),
242 // since the 'unglued' tokens due not exist in
243 // the original `TokenStream`.
245 // If we end up consuming both unglued tokens,
246 // then this is not an issue - we'll end up
247 // capturing the single 'glued' token.
249 // However, in certain circumstances, we may
250 // want to capture just the first 'unglued' token.
251 // For example, capturing the `Vec<u8>`
252 // in `Option<Vec<u8>>` requires us to unglue
253 // the trailing `>>` token. The `break_last_token`
254 // field is used to track this token - it gets
255 // appended to the captured stream when
256 // we evaluate a `LazyAttrTokenStream`.
257 break_last_token: bool,
261 struct TokenCursorFrame {
262 // This is `None` only for the outermost frame.
263 delim_sp: Option<(Delimiter, DelimSpan)>,
264 tree_cursor: tokenstream::Cursor,
267 impl TokenCursorFrame {
268 fn new(delim_sp: Option<(Delimiter, DelimSpan)>, tts: TokenStream) -> Self {
269 TokenCursorFrame { delim_sp, tree_cursor: tts.into_trees() }
274 fn next(&mut self, desugar_doc_comments: bool) -> (Token, Spacing) {
275 self.inlined_next(desugar_doc_comments)
278 /// This always-inlined version should only be used on hot code paths.
280 fn inlined_next(&mut self, desugar_doc_comments: bool) -> (Token, Spacing) {
282 // FIXME: we currently don't return `Delimiter` open/close delims. To fix #67062 we will
283 // need to, whereupon the `delim != Delimiter::Invisible` conditions below can be
285 if let Some(tree) = self.frame.tree_cursor.next_ref() {
287 &TokenTree::Token(ref token, spacing) => match (desugar_doc_comments, token) {
288 (true, &Token { kind: token::DocComment(_, attr_style, data), span }) => {
289 let desugared = self.desugar(attr_style, data, span);
290 self.frame.tree_cursor.replace_prev_and_rewind(desugared);
291 // Continue to get the first token of the desugared doc comment.
293 _ => return (token.clone(), spacing),
295 &TokenTree::Delimited(sp, delim, ref tts) => {
296 // Set `open_delim` to true here because we deal with it immediately.
297 let frame = TokenCursorFrame::new(Some((delim, sp)), tts.clone());
298 self.stack.push(mem::replace(&mut self.frame, frame));
299 if delim != Delimiter::Invisible {
300 return (Token::new(token::OpenDelim(delim), sp.open), Spacing::Alone);
302 // No open delimiter to return; continue on to the next iteration.
305 } else if let Some(frame) = self.stack.pop() {
306 // We have exhausted this frame. Move back to its parent frame.
307 let (delim, span) = self.frame.delim_sp.unwrap();
309 if delim != Delimiter::Invisible {
310 return (Token::new(token::CloseDelim(delim), span.close), Spacing::Alone);
312 // No close delimiter to return; continue on to the next iteration.
314 // We have exhausted the outermost frame.
315 return (Token::new(token::Eof, DUMMY_SP), Spacing::Alone);
320 // Desugar a doc comment into something like `#[doc = r"foo"]`.
321 fn desugar(&mut self, attr_style: AttrStyle, data: Symbol, span: Span) -> Vec<TokenTree> {
322 // Searches for the occurrences of `"#*` and returns the minimum number of `#`s
323 // required to wrap the text. E.g.
324 // - `abc d` is wrapped as `r"abc d"` (num_of_hashes = 0)
325 // - `abc "d"` is wrapped as `r#"abc "d""#` (num_of_hashes = 1)
326 // - `abc "##d##"` is wrapped as `r###"abc ##"d"##"###` (num_of_hashes = 3)
327 let mut num_of_hashes = 0;
329 for ch in data.as_str().chars() {
332 '#' if count > 0 => count + 1,
335 num_of_hashes = cmp::max(num_of_hashes, count);
338 // `/// foo` becomes `doc = r"foo".
339 let delim_span = DelimSpan::from_single(span);
340 let body = TokenTree::Delimited(
344 TokenTree::token_alone(token::Ident(sym::doc, false), span),
345 TokenTree::token_alone(token::Eq, span),
346 TokenTree::token_alone(
347 TokenKind::lit(token::StrRaw(num_of_hashes), data, None),
352 .collect::<TokenStream>(),
355 if attr_style == AttrStyle::Inner {
357 TokenTree::token_alone(token::Pound, span),
358 TokenTree::token_alone(token::Not, span),
362 vec![TokenTree::token_alone(token::Pound, span), body]
367 #[derive(Debug, Clone, PartialEq)]
380 fn to_string(&self) -> String {
382 TokenType::Token(t) => format!("`{}`", pprust::token_kind_to_string(t)),
383 TokenType::Keyword(kw) => format!("`{}`", kw),
384 TokenType::Operator => "an operator".to_string(),
385 TokenType::Lifetime => "lifetime".to_string(),
386 TokenType::Ident => "identifier".to_string(),
387 TokenType::Path => "path".to_string(),
388 TokenType::Type => "type".to_string(),
389 TokenType::Const => "a const expression".to_string(),
394 #[derive(Copy, Clone, Debug)]
395 enum TokenExpectType {
400 /// A sequence separator.
402 /// The separator token.
403 sep: Option<TokenKind>,
404 /// `true` if a trailing separator is allowed.
405 trailing_sep_allowed: bool,
409 fn trailing_allowed(t: TokenKind) -> SeqSep {
410 SeqSep { sep: Some(t), trailing_sep_allowed: true }
413 fn none() -> SeqSep {
414 SeqSep { sep: None, trailing_sep_allowed: false }
418 pub enum FollowedByType {
423 #[derive(Clone, Copy, PartialEq, Eq)]
424 pub enum TokenDescription {
431 impl TokenDescription {
432 pub fn from_token(token: &Token) -> Option<Self> {
434 _ if token.is_special_ident() => Some(TokenDescription::ReservedIdentifier),
435 _ if token.is_used_keyword() => Some(TokenDescription::Keyword),
436 _ if token.is_unused_keyword() => Some(TokenDescription::ReservedKeyword),
437 token::DocComment(..) => Some(TokenDescription::DocComment),
443 pub(super) fn token_descr(token: &Token) -> String {
444 let name = pprust::token_to_string(token).to_string();
446 let kind = TokenDescription::from_token(token).map(|kind| match kind {
447 TokenDescription::ReservedIdentifier => "reserved identifier",
448 TokenDescription::Keyword => "keyword",
449 TokenDescription::ReservedKeyword => "reserved keyword",
450 TokenDescription::DocComment => "doc comment",
453 if let Some(kind) = kind { format!("{} `{}`", kind, name) } else { format!("`{}`", name) }
456 impl<'a> Parser<'a> {
460 desugar_doc_comments: bool,
461 subparser_name: Option<&'static str>,
463 let mut parser = Parser {
465 token: Token::dummy(),
466 token_spacing: Spacing::Alone,
467 prev_token: Token::dummy(),
469 restrictions: Restrictions::empty(),
470 expected_tokens: Vec::new(),
471 token_cursor: TokenCursor {
472 frame: TokenCursorFrame::new(None, tokens),
475 desugar_doc_comments,
476 break_last_token: false,
478 desugar_doc_comments,
479 unmatched_angle_bracket_count: 0,
480 max_angle_bracket_count: 0,
481 unclosed_delims: Vec::new(),
482 last_unexpected_token_span: None,
483 last_type_ascription: None,
485 capture_state: CaptureState {
486 capturing: Capturing::No,
487 replace_ranges: Vec::new(),
488 inner_attr_ranges: Default::default(),
490 current_closure: None,
491 recovery: Recovery::Allowed,
494 // Make parser point to the first token.
500 pub fn recovery(mut self, recovery: Recovery) -> Self {
501 self.recovery = recovery;
505 /// Whether the parser is allowed to recover from broken code.
507 /// If this returns false, recovering broken code into valid code (especially if this recovery does lookahead)
508 /// is not allowed. All recovery done by the parser must be gated behind this check.
510 /// Technically, this only needs to restrict eager recovery by doing lookahead at more tokens.
511 /// But making the distinction is very subtle, and simply forbidding all recovery is a lot simpler to uphold.
512 fn may_recover(&self) -> bool {
513 matches!(self.recovery, Recovery::Allowed)
516 pub fn unexpected<T>(&mut self) -> PResult<'a, T> {
517 match self.expect_one_of(&[], &[]) {
519 // We can get `Ok(true)` from `recover_closing_delimiter`
520 // which is called in `expected_one_of_not_found`.
521 Ok(_) => FatalError.raise(),
525 /// Expects and consumes the token `t`. Signals an error if the next token is not `t`.
526 pub fn expect(&mut self, t: &TokenKind) -> PResult<'a, bool /* recovered */> {
527 if self.expected_tokens.is_empty() {
528 if self.token == *t {
532 self.unexpected_try_recover(t)
535 self.expect_one_of(slice::from_ref(t), &[])
539 /// Expect next token to be edible or inedible token. If edible,
540 /// then consume it; if inedible, then return without consuming
541 /// anything. Signal a fatal error if next token is unexpected.
542 pub fn expect_one_of(
544 edible: &[TokenKind],
545 inedible: &[TokenKind],
546 ) -> PResult<'a, bool /* recovered */> {
547 if edible.contains(&self.token.kind) {
550 } else if inedible.contains(&self.token.kind) {
551 // leave it in the input
553 } else if self.last_unexpected_token_span == Some(self.token.span) {
556 self.expected_one_of_not_found(edible, inedible)
560 // Public for rustfmt usage.
561 pub fn parse_ident(&mut self) -> PResult<'a, Ident> {
562 self.parse_ident_common(true)
565 fn ident_or_err(&mut self) -> PResult<'a, (Ident, /* is_raw */ bool)> {
566 self.token.ident().ok_or_else(|| match self.prev_token.kind {
567 TokenKind::DocComment(..) => DocCommentDoesNotDocumentAnything {
568 span: self.prev_token.span,
571 .into_diagnostic(&self.sess.span_diagnostic),
572 _ => self.expected_ident_found(),
576 fn parse_ident_common(&mut self, recover: bool) -> PResult<'a, Ident> {
577 let (ident, is_raw) = self.ident_or_err()?;
578 if !is_raw && ident.is_reserved() {
579 let mut err = self.expected_ident_found();
590 /// Checks if the next token is `tok`, and returns `true` if so.
592 /// This method will automatically add `tok` to `expected_tokens` if `tok` is not
594 fn check(&mut self, tok: &TokenKind) -> bool {
595 let is_present = self.token == *tok;
597 self.expected_tokens.push(TokenType::Token(tok.clone()));
602 fn check_noexpect(&self, tok: &TokenKind) -> bool {
606 /// Consumes a token 'tok' if it exists. Returns whether the given token was present.
608 /// the main purpose of this function is to reduce the cluttering of the suggestions list
609 /// which using the normal eat method could introduce in some cases.
610 pub fn eat_noexpect(&mut self, tok: &TokenKind) -> bool {
611 let is_present = self.check_noexpect(tok);
618 /// Consumes a token 'tok' if it exists. Returns whether the given token was present.
619 pub fn eat(&mut self, tok: &TokenKind) -> bool {
620 let is_present = self.check(tok);
627 /// If the next token is the given keyword, returns `true` without eating it.
628 /// An expectation is also added for diagnostics purposes.
629 fn check_keyword(&mut self, kw: Symbol) -> bool {
630 self.expected_tokens.push(TokenType::Keyword(kw));
631 self.token.is_keyword(kw)
634 fn check_keyword_case(&mut self, kw: Symbol, case: Case) -> bool {
635 if self.check_keyword(kw) {
639 if case == Case::Insensitive
640 && let Some((ident, /* is_raw */ false)) = self.token.ident()
641 && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() {
648 /// If the next token is the given keyword, eats it and returns `true`.
649 /// Otherwise, returns `false`. An expectation is also added for diagnostics purposes.
650 // Public for rustfmt usage.
651 pub fn eat_keyword(&mut self, kw: Symbol) -> bool {
652 if self.check_keyword(kw) {
660 /// Eats a keyword, optionally ignoring the case.
661 /// If the case differs (and is ignored) an error is issued.
662 /// This is useful for recovery.
663 fn eat_keyword_case(&mut self, kw: Symbol, case: Case) -> bool {
664 if self.eat_keyword(kw) {
668 if case == Case::Insensitive
669 && let Some((ident, /* is_raw */ false)) = self.token.ident()
670 && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() {
672 .struct_span_err(ident.span, format!("keyword `{kw}` is written in a wrong case"))
675 "write it in the correct case",
677 Applicability::MachineApplicable
687 fn eat_keyword_noexpect(&mut self, kw: Symbol) -> bool {
688 if self.token.is_keyword(kw) {
696 /// If the given word is not a keyword, signals an error.
697 /// If the next token is not the given word, signals an error.
698 /// Otherwise, eats it.
699 fn expect_keyword(&mut self, kw: Symbol) -> PResult<'a, ()> {
700 if !self.eat_keyword(kw) { self.unexpected() } else { Ok(()) }
703 /// Is the given keyword `kw` followed by a non-reserved identifier?
704 fn is_kw_followed_by_ident(&self, kw: Symbol) -> bool {
705 self.token.is_keyword(kw) && self.look_ahead(1, |t| t.is_ident() && !t.is_reserved_ident())
708 fn check_or_expected(&mut self, ok: bool, typ: TokenType) -> bool {
712 self.expected_tokens.push(typ);
717 fn check_ident(&mut self) -> bool {
718 self.check_or_expected(self.token.is_ident(), TokenType::Ident)
721 fn check_path(&mut self) -> bool {
722 self.check_or_expected(self.token.is_path_start(), TokenType::Path)
725 fn check_type(&mut self) -> bool {
726 self.check_or_expected(self.token.can_begin_type(), TokenType::Type)
729 fn check_const_arg(&mut self) -> bool {
730 self.check_or_expected(self.token.can_begin_const_arg(), TokenType::Const)
733 fn check_const_closure(&self) -> bool {
734 self.is_keyword_ahead(0, &[kw::Const])
735 && self.look_ahead(1, |t| match &t.kind {
736 token::Ident(kw::Move | kw::Static | kw::Async, _)
738 | token::BinOp(token::Or) => true,
743 fn check_inline_const(&self, dist: usize) -> bool {
744 self.is_keyword_ahead(dist, &[kw::Const])
745 && self.look_ahead(dist + 1, |t| match &t.kind {
746 token::Interpolated(nt) => matches!(**nt, token::NtBlock(..)),
747 token::OpenDelim(Delimiter::Brace) => true,
752 /// Checks to see if the next token is either `+` or `+=`.
753 /// Otherwise returns `false`.
754 fn check_plus(&mut self) -> bool {
755 self.check_or_expected(
756 self.token.is_like_plus(),
757 TokenType::Token(token::BinOp(token::Plus)),
761 /// Eats the expected token if it's present possibly breaking
762 /// compound tokens like multi-character operators in process.
763 /// Returns `true` if the token was eaten.
764 fn break_and_eat(&mut self, expected: TokenKind) -> bool {
765 if self.token.kind == expected {
769 match self.token.kind.break_two_token_op() {
770 Some((first, second)) if first == expected => {
771 let first_span = self.sess.source_map().start_point(self.token.span);
772 let second_span = self.token.span.with_lo(first_span.hi());
773 self.token = Token::new(first, first_span);
774 // Keep track of this token - if we end token capturing now,
775 // we'll want to append this token to the captured stream.
777 // If we consume any additional tokens, then this token
778 // is not needed (we'll capture the entire 'glued' token),
779 // and `bump` will set this field to `None`
780 self.token_cursor.break_last_token = true;
781 // Use the spacing of the glued token as the spacing
782 // of the unglued second token.
783 self.bump_with((Token::new(second, second_span), self.token_spacing));
787 self.expected_tokens.push(TokenType::Token(expected));
793 /// Eats `+` possibly breaking tokens like `+=` in process.
794 fn eat_plus(&mut self) -> bool {
795 self.break_and_eat(token::BinOp(token::Plus))
798 /// Eats `&` possibly breaking tokens like `&&` in process.
799 /// Signals an error if `&` is not eaten.
800 fn expect_and(&mut self) -> PResult<'a, ()> {
801 if self.break_and_eat(token::BinOp(token::And)) { Ok(()) } else { self.unexpected() }
804 /// Eats `|` possibly breaking tokens like `||` in process.
805 /// Signals an error if `|` was not eaten.
806 fn expect_or(&mut self) -> PResult<'a, ()> {
807 if self.break_and_eat(token::BinOp(token::Or)) { Ok(()) } else { self.unexpected() }
810 /// Eats `<` possibly breaking tokens like `<<` in process.
811 fn eat_lt(&mut self) -> bool {
812 let ate = self.break_and_eat(token::Lt);
814 // See doc comment for `unmatched_angle_bracket_count`.
815 self.unmatched_angle_bracket_count += 1;
816 self.max_angle_bracket_count += 1;
817 debug!("eat_lt: (increment) count={:?}", self.unmatched_angle_bracket_count);
822 /// Eats `<` possibly breaking tokens like `<<` in process.
823 /// Signals an error if `<` was not eaten.
824 fn expect_lt(&mut self) -> PResult<'a, ()> {
825 if self.eat_lt() { Ok(()) } else { self.unexpected() }
828 /// Eats `>` possibly breaking tokens like `>>` in process.
829 /// Signals an error if `>` was not eaten.
830 fn expect_gt(&mut self) -> PResult<'a, ()> {
831 if self.break_and_eat(token::Gt) {
832 // See doc comment for `unmatched_angle_bracket_count`.
833 if self.unmatched_angle_bracket_count > 0 {
834 self.unmatched_angle_bracket_count -= 1;
835 debug!("expect_gt: (decrement) count={:?}", self.unmatched_angle_bracket_count);
843 fn expect_any_with_type(&mut self, kets: &[&TokenKind], expect: TokenExpectType) -> bool {
844 kets.iter().any(|k| match expect {
845 TokenExpectType::Expect => self.check(k),
846 TokenExpectType::NoExpect => self.token == **k,
850 fn parse_seq_to_before_tokens<T>(
854 expect: TokenExpectType,
855 mut f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
856 ) -> PResult<'a, (Vec<T>, bool /* trailing */, bool /* recovered */)> {
857 let mut first = true;
858 let mut recovered = false;
859 let mut trailing = false;
861 let unclosed_delims = !self.unclosed_delims.is_empty();
863 while !self.expect_any_with_type(kets, expect) {
864 if let token::CloseDelim(..) | token::Eof = self.token.kind {
867 if let Some(t) = &sep.sep {
871 match self.expect(t) {
873 self.current_closure.take();
876 self.current_closure.take();
880 Err(mut expect_err) => {
881 let sp = self.prev_token.span.shrink_to_hi();
882 let token_str = pprust::token_kind_to_string(t);
884 match self.current_closure.take() {
885 Some(closure_spans) if self.token.kind == TokenKind::Semi => {
886 // Finding a semicolon instead of a comma
887 // after a closure body indicates that the
888 // closure body may be a block but the user
889 // forgot to put braces around its
892 self.recover_missing_braces_around_closure_body(
901 // Attempt to keep parsing if it was a similar separator.
902 if let Some(tokens) = t.similar_tokens() {
903 if tokens.contains(&self.token.kind) && !unclosed_delims {
910 // If this was a missing `@` in a binding pattern
911 // bail with a suggestion
912 // https://github.com/rust-lang/rust/issues/72373
913 if self.prev_token.is_ident() && self.token.kind == token::DotDot {
915 "if you meant to bind the contents of \
916 the rest of the array pattern into `{}`, use `@`",
917 pprust::token_to_string(&self.prev_token)
920 .span_suggestion_verbose(
921 self.prev_token.span.shrink_to_hi().until(self.token.span),
924 Applicability::MaybeIncorrect,
930 // Attempt to keep parsing if it was an omitted separator.
933 // Parsed successfully, therefore most probably the code only
934 // misses a separator.
936 .span_suggestion_short(
938 &format!("missing `{}`", token_str),
940 Applicability::MaybeIncorrect,
948 // Parsing failed, therefore it must be something more serious
949 // than just a missing separator.
950 for xx in &e.children {
951 // propagate the help message from sub error 'e' to main error 'expect_err;
952 expect_err.children.push(xx.clone());
964 if sep.trailing_sep_allowed && self.expect_any_with_type(kets, expect) {
973 Ok((v, trailing, recovered))
976 fn recover_missing_braces_around_closure_body(
978 closure_spans: ClosureSpans,
979 mut expect_err: DiagnosticBuilder<'_, ErrorGuaranteed>,
980 ) -> PResult<'a, ()> {
981 let initial_semicolon = self.token.span;
983 while self.eat(&TokenKind::Semi) {
984 let _ = self.parse_stmt(ForceCollect::Yes)?;
987 expect_err.set_primary_message(
988 "closure bodies that contain statements must be surrounded by braces",
991 let preceding_pipe_span = closure_spans.closing_pipe;
992 let following_token_span = self.token.span;
994 let mut first_note = MultiSpan::from(vec![initial_semicolon]);
995 first_note.push_span_label(
997 "this `;` turns the preceding closure into a statement",
999 first_note.push_span_label(
1001 "this expression is a statement because of the trailing semicolon",
1003 expect_err.span_note(first_note, "statement found outside of a block");
1005 let mut second_note = MultiSpan::from(vec![closure_spans.whole_closure]);
1006 second_note.push_span_label(closure_spans.whole_closure, "this is the parsed closure...");
1007 second_note.push_span_label(
1008 following_token_span,
1009 "...but likely you meant the closure to end here",
1011 expect_err.span_note(second_note, "the closure body may be incorrectly delimited");
1013 expect_err.set_span(vec![preceding_pipe_span, following_token_span]);
1015 let opening_suggestion_str = " {".to_string();
1016 let closing_suggestion_str = "}".to_string();
1018 expect_err.multipart_suggestion(
1019 "try adding braces",
1021 (preceding_pipe_span.shrink_to_hi(), opening_suggestion_str),
1022 (following_token_span.shrink_to_lo(), closing_suggestion_str),
1024 Applicability::MaybeIncorrect,
1032 /// Parses a sequence, not including the closing delimiter. The function
1033 /// `f` must consume tokens until reaching the next separator or
1034 /// closing bracket.
1035 fn parse_seq_to_before_end<T>(
1039 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
1040 ) -> PResult<'a, (Vec<T>, bool, bool)> {
1041 self.parse_seq_to_before_tokens(&[ket], sep, TokenExpectType::Expect, f)
1044 /// Parses a sequence, including the closing delimiter. The function
1045 /// `f` must consume tokens until reaching the next separator or
1046 /// closing bracket.
1047 fn parse_seq_to_end<T>(
1051 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
1052 ) -> PResult<'a, (Vec<T>, bool /* trailing */)> {
1053 let (val, trailing, recovered) = self.parse_seq_to_before_end(ket, sep, f)?;
1060 /// Parses a sequence, including the closing delimiter. The function
1061 /// `f` must consume tokens until reaching the next separator or
1062 /// closing bracket.
1063 fn parse_unspanned_seq<T>(
1068 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
1069 ) -> PResult<'a, (Vec<T>, bool)> {
1071 self.parse_seq_to_end(ket, sep, f)
1074 fn parse_delim_comma_seq<T>(
1077 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
1078 ) -> PResult<'a, (Vec<T>, bool)> {
1079 self.parse_unspanned_seq(
1080 &token::OpenDelim(delim),
1081 &token::CloseDelim(delim),
1082 SeqSep::trailing_allowed(token::Comma),
1087 fn parse_paren_comma_seq<T>(
1089 f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>,
1090 ) -> PResult<'a, (Vec<T>, bool)> {
1091 self.parse_delim_comma_seq(Delimiter::Parenthesis, f)
1094 /// Advance the parser by one token using provided token as the next one.
1095 fn bump_with(&mut self, next: (Token, Spacing)) {
1096 self.inlined_bump_with(next)
1099 /// This always-inlined version should only be used on hot code paths.
1101 fn inlined_bump_with(&mut self, (next_token, next_spacing): (Token, Spacing)) {
1102 // Update the current and previous tokens.
1103 self.prev_token = mem::replace(&mut self.token, next_token);
1104 self.token_spacing = next_spacing;
1107 self.expected_tokens.clear();
1110 /// Advance the parser by one token.
1111 pub fn bump(&mut self) {
1112 // Note: destructuring here would give nicer code, but it was found in #96210 to be slower
1113 // than `.0`/`.1` access.
1114 let mut next = self.token_cursor.inlined_next(self.desugar_doc_comments);
1115 self.token_cursor.num_next_calls += 1;
1116 // We've retrieved an token from the underlying
1117 // cursor, so we no longer need to worry about
1118 // an unglued token. See `break_and_eat` for more details
1119 self.token_cursor.break_last_token = false;
1120 if next.0.span.is_dummy() {
1121 // Tweak the location for better diagnostics, but keep syntactic context intact.
1122 let fallback_span = self.token.span;
1123 next.0.span = fallback_span.with_ctxt(next.0.span.ctxt());
1125 debug_assert!(!matches!(
1127 token::OpenDelim(Delimiter::Invisible) | token::CloseDelim(Delimiter::Invisible)
1129 self.inlined_bump_with(next)
1132 /// Look-ahead `dist` tokens of `self.token` and get access to that token there.
1133 /// When `dist == 0` then the current token is looked at.
1134 pub fn look_ahead<R>(&self, dist: usize, looker: impl FnOnce(&Token) -> R) -> R {
1136 return looker(&self.token);
1139 let frame = &self.token_cursor.frame;
1140 if let Some((delim, span)) = frame.delim_sp && delim != Delimiter::Invisible {
1141 let all_normal = (0..dist).all(|i| {
1142 let token = frame.tree_cursor.look_ahead(i);
1143 !matches!(token, Some(TokenTree::Delimited(_, Delimiter::Invisible, _)))
1146 return match frame.tree_cursor.look_ahead(dist - 1) {
1147 Some(tree) => match tree {
1148 TokenTree::Token(token, _) => looker(token),
1149 TokenTree::Delimited(dspan, delim, _) => {
1150 looker(&Token::new(token::OpenDelim(*delim), dspan.open))
1153 None => looker(&Token::new(token::CloseDelim(delim), span.close)),
1158 let mut cursor = self.token_cursor.clone();
1160 let mut token = Token::dummy();
1162 token = cursor.next(/* desugar_doc_comments */ false).0;
1165 token::OpenDelim(Delimiter::Invisible) | token::CloseDelim(Delimiter::Invisible)
1171 return looker(&token);
1174 /// Returns whether any of the given keywords are `dist` tokens ahead of the current one.
1175 fn is_keyword_ahead(&self, dist: usize, kws: &[Symbol]) -> bool {
1176 self.look_ahead(dist, |t| kws.iter().any(|&kw| t.is_keyword(kw)))
1179 /// Parses asyncness: `async` or nothing.
1180 fn parse_asyncness(&mut self, case: Case) -> Async {
1181 if self.eat_keyword_case(kw::Async, case) {
1182 let span = self.prev_token.uninterpolated_span();
1183 Async::Yes { span, closure_id: DUMMY_NODE_ID, return_impl_trait_id: DUMMY_NODE_ID }
1189 /// Parses unsafety: `unsafe` or nothing.
1190 fn parse_unsafety(&mut self, case: Case) -> Unsafe {
1191 if self.eat_keyword_case(kw::Unsafe, case) {
1192 Unsafe::Yes(self.prev_token.uninterpolated_span())
1198 /// Parses constness: `const` or nothing.
1199 fn parse_constness(&mut self, case: Case) -> Const {
1200 // Avoid const blocks to be parsed as const items
1201 if self.look_ahead(1, |t| t != &token::OpenDelim(Delimiter::Brace))
1202 && self.eat_keyword_case(kw::Const, case)
1204 Const::Yes(self.prev_token.uninterpolated_span())
1210 /// Parses inline const expressions.
1211 fn parse_const_block(&mut self, span: Span, pat: bool) -> PResult<'a, P<Expr>> {
1213 self.sess.gated_spans.gate(sym::inline_const_pat, span);
1215 self.sess.gated_spans.gate(sym::inline_const, span);
1217 self.eat_keyword(kw::Const);
1218 let (attrs, blk) = self.parse_inner_attrs_and_block()?;
1219 let anon_const = AnonConst {
1221 value: self.mk_expr(blk.span, ExprKind::Block(blk, None)),
1223 let blk_span = anon_const.value.span;
1224 Ok(self.mk_expr_with_attrs(span.to(blk_span), ExprKind::ConstBlock(anon_const), attrs))
1227 /// Parses mutability (`mut` or nothing).
1228 fn parse_mutability(&mut self) -> Mutability {
1229 if self.eat_keyword(kw::Mut) { Mutability::Mut } else { Mutability::Not }
1232 /// Possibly parses mutability (`const` or `mut`).
1233 fn parse_const_or_mut(&mut self) -> Option<Mutability> {
1234 if self.eat_keyword(kw::Mut) {
1235 Some(Mutability::Mut)
1236 } else if self.eat_keyword(kw::Const) {
1237 Some(Mutability::Not)
1243 fn parse_field_name(&mut self) -> PResult<'a, Ident> {
1244 if let token::Literal(token::Lit { kind: token::Integer, symbol, suffix }) = self.token.kind
1246 if let Some(suffix) = suffix {
1247 self.expect_no_tuple_index_suffix(self.token.span, suffix);
1250 Ok(Ident::new(symbol, self.prev_token.span))
1252 self.parse_ident_common(true)
1256 fn parse_delim_args(&mut self) -> PResult<'a, P<DelimArgs>> {
1257 if let Some(args) = self.parse_delim_args_inner() { Ok(P(args)) } else { self.unexpected() }
1260 fn parse_attr_args(&mut self) -> PResult<'a, AttrArgs> {
1261 Ok(if let Some(args) = self.parse_delim_args_inner() {
1262 AttrArgs::Delimited(args)
1264 if self.eat(&token::Eq) {
1265 let eq_span = self.prev_token.span;
1266 AttrArgs::Eq(eq_span, AttrArgsEq::Ast(self.parse_expr_force_collect()?))
1273 fn parse_delim_args_inner(&mut self) -> Option<DelimArgs> {
1274 if self.check(&token::OpenDelim(Delimiter::Parenthesis))
1275 || self.check(&token::OpenDelim(Delimiter::Bracket))
1276 || self.check(&token::OpenDelim(Delimiter::Brace))
1278 match self.parse_token_tree() {
1279 // We've confirmed above that there is a delimiter so unwrapping is OK.
1280 TokenTree::Delimited(dspan, delim, tokens) => Some(DelimArgs {
1282 delim: MacDelimiter::from_token(delim).unwrap(),
1285 _ => unreachable!(),
1292 fn parse_or_use_outer_attributes(
1294 already_parsed_attrs: Option<AttrWrapper>,
1295 ) -> PResult<'a, AttrWrapper> {
1296 if let Some(attrs) = already_parsed_attrs {
1299 self.parse_outer_attributes()
1303 /// Parses a single token tree from the input.
1304 pub(crate) fn parse_token_tree(&mut self) -> TokenTree {
1305 match self.token.kind {
1306 token::OpenDelim(..) => {
1307 // Grab the tokens from this frame.
1308 let frame = &self.token_cursor.frame;
1309 let stream = frame.tree_cursor.stream.clone();
1310 let (delim, span) = frame.delim_sp.unwrap();
1312 // Advance the token cursor through the entire delimited
1313 // sequence. After getting the `OpenDelim` we are *within* the
1314 // delimited sequence, i.e. at depth `d`. After getting the
1315 // matching `CloseDelim` we are *after* the delimited sequence,
1316 // i.e. at depth `d - 1`.
1317 let target_depth = self.token_cursor.stack.len() - 1;
1319 // Advance one token at a time, so `TokenCursor::next()`
1320 // can capture these tokens if necessary.
1322 if self.token_cursor.stack.len() == target_depth {
1323 debug_assert!(matches!(self.token.kind, token::CloseDelim(_)));
1328 // Consume close delimiter
1330 TokenTree::Delimited(span, delim, stream)
1332 token::CloseDelim(_) | token::Eof => unreachable!(),
1335 TokenTree::Token(self.prev_token.clone(), Spacing::Alone)
1340 /// Parses a stream of tokens into a list of `TokenTree`s, up to EOF.
1341 pub fn parse_all_token_trees(&mut self) -> PResult<'a, Vec<TokenTree>> {
1342 let mut tts = Vec::new();
1343 while self.token != token::Eof {
1344 tts.push(self.parse_token_tree());
1349 pub fn parse_tokens(&mut self) -> TokenStream {
1350 let mut result = Vec::new();
1352 match self.token.kind {
1353 token::Eof | token::CloseDelim(..) => break,
1354 _ => result.push(self.parse_token_tree()),
1357 TokenStream::new(result)
1360 /// Evaluates the closure with restrictions in place.
1362 /// Afters the closure is evaluated, restrictions are reset.
1363 fn with_res<T>(&mut self, res: Restrictions, f: impl FnOnce(&mut Self) -> T) -> T {
1364 let old = self.restrictions;
1365 self.restrictions = res;
1367 self.restrictions = old;
1371 /// Parses `pub` and `pub(in path)` plus shortcuts `pub(crate)` for `pub(in crate)`, `pub(self)`
1372 /// for `pub(in self)` and `pub(super)` for `pub(in super)`.
1373 /// If the following element can't be a tuple (i.e., it's a function definition), then
1374 /// it's not a tuple struct field), and the contents within the parentheses aren't valid,
1375 /// so emit a proper diagnostic.
1376 // Public for rustfmt usage.
1377 pub fn parse_visibility(&mut self, fbt: FollowedByType) -> PResult<'a, Visibility> {
1378 maybe_whole!(self, NtVis, |x| x.into_inner());
1380 if !self.eat_keyword(kw::Pub) {
1381 // We need a span for our `Spanned<VisibilityKind>`, but there's inherently no
1382 // keyword to grab a span from for inherited visibility; an empty span at the
1383 // beginning of the current token would seem to be the "Schelling span".
1384 return Ok(Visibility {
1385 span: self.token.span.shrink_to_lo(),
1386 kind: VisibilityKind::Inherited,
1390 let lo = self.prev_token.span;
1392 if self.check(&token::OpenDelim(Delimiter::Parenthesis)) {
1393 // We don't `self.bump()` the `(` yet because this might be a struct definition where
1394 // `()` or a tuple might be allowed. For example, `struct Struct(pub (), pub (usize));`.
1395 // Because of this, we only `bump` the `(` if we're assured it is appropriate to do so
1396 // by the following tokens.
1397 if self.is_keyword_ahead(1, &[kw::In]) {
1398 // Parse `pub(in path)`.
1400 self.bump(); // `in`
1401 let path = self.parse_path(PathStyle::Mod)?; // `path`
1402 self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; // `)`
1403 let vis = VisibilityKind::Restricted {
1405 id: ast::DUMMY_NODE_ID,
1408 return Ok(Visibility {
1409 span: lo.to(self.prev_token.span),
1413 } else if self.look_ahead(2, |t| t == &token::CloseDelim(Delimiter::Parenthesis))
1414 && self.is_keyword_ahead(1, &[kw::Crate, kw::Super, kw::SelfLower])
1416 // Parse `pub(crate)`, `pub(self)`, or `pub(super)`.
1418 let path = self.parse_path(PathStyle::Mod)?; // `crate`/`super`/`self`
1419 self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; // `)`
1420 let vis = VisibilityKind::Restricted {
1422 id: ast::DUMMY_NODE_ID,
1425 return Ok(Visibility {
1426 span: lo.to(self.prev_token.span),
1430 } else if let FollowedByType::No = fbt {
1431 // Provide this diagnostic if a type cannot follow;
1432 // in particular, if this is not a tuple struct.
1433 self.recover_incorrect_vis_restriction()?;
1434 // Emit diagnostic, but continue with public visibility.
1438 Ok(Visibility { span: lo, kind: VisibilityKind::Public, tokens: None })
1441 /// Recovery for e.g. `pub(something) fn ...` or `struct X { pub(something) y: Z }`
1442 fn recover_incorrect_vis_restriction(&mut self) -> PResult<'a, ()> {
1444 let path = self.parse_path(PathStyle::Mod)?;
1445 self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; // `)`
1447 let path_str = pprust::path_to_string(&path);
1448 self.sess.emit_err(IncorrectVisibilityRestriction { span: path.span, inner_str: path_str });
1453 /// Parses `extern string_literal?`.
1454 fn parse_extern(&mut self, case: Case) -> Extern {
1455 if self.eat_keyword_case(kw::Extern, case) {
1456 let mut extern_span = self.prev_token.span;
1457 let abi = self.parse_abi();
1458 if let Some(abi) = abi {
1459 extern_span = extern_span.to(abi.span);
1461 Extern::from_abi(abi, extern_span)
1467 /// Parses a string literal as an ABI spec.
1468 fn parse_abi(&mut self) -> Option<StrLit> {
1469 match self.parse_str_lit() {
1470 Ok(str_lit) => Some(str_lit),
1471 Err(Some(lit)) => match lit.kind {
1472 ast::LitKind::Err => None,
1474 self.sess.emit_err(NonStringAbiLiteral { span: lit.span });
1482 pub fn collect_tokens_no_attrs<R: HasAttrs + HasTokens>(
1484 f: impl FnOnce(&mut Self) -> PResult<'a, R>,
1485 ) -> PResult<'a, R> {
1486 // The only reason to call `collect_tokens_no_attrs` is if you want tokens, so use
1487 // `ForceCollect::Yes`
1488 self.collect_tokens_trailing_token(
1489 AttrWrapper::empty(),
1491 |this, _attrs| Ok((f(this)?, TrailingToken::None)),
1496 fn is_import_coupler(&mut self) -> bool {
1497 self.check(&token::ModSep)
1498 && self.look_ahead(1, |t| {
1499 *t == token::OpenDelim(Delimiter::Brace) || *t == token::BinOp(token::Star)
1503 pub fn clear_expected_tokens(&mut self) {
1504 self.expected_tokens.clear();
1507 pub fn approx_token_stream_pos(&self) -> usize {
1508 self.token_cursor.num_next_calls
1512 pub(crate) fn make_unclosed_delims_error(
1513 unmatched: UnmatchedBrace,
1515 ) -> Option<DiagnosticBuilder<'_, ErrorGuaranteed>> {
1516 // `None` here means an `Eof` was found. We already emit those errors elsewhere, we add them to
1517 // `unmatched_braces` only for error recovery in the `Parser`.
1518 let found_delim = unmatched.found_delim?;
1519 let mut spans = vec![unmatched.found_span];
1520 if let Some(sp) = unmatched.unclosed_span {
1523 let err = MismatchedClosingDelimiter {
1525 delimiter: pprust::token_kind_to_string(&token::CloseDelim(found_delim)).to_string(),
1526 unmatched: unmatched.found_span,
1527 opening_candidate: unmatched.candidate_span,
1528 unclosed: unmatched.unclosed_span,
1530 .into_diagnostic(&sess.span_diagnostic);
1534 pub fn emit_unclosed_delims(unclosed_delims: &mut Vec<UnmatchedBrace>, sess: &ParseSess) {
1535 *sess.reached_eof.borrow_mut() |=
1536 unclosed_delims.iter().any(|unmatched_delim| unmatched_delim.found_delim.is_none());
1537 for unmatched in unclosed_delims.drain(..) {
1538 if let Some(mut e) = make_unclosed_delims_error(unmatched, sess) {
1544 /// A helper struct used when building an `AttrTokenStream` from
1545 /// a `LazyAttrTokenStream`. Both delimiter and non-delimited tokens
1546 /// are stored as `FlatToken::Token`. A vector of `FlatToken`s
1547 /// is then 'parsed' to build up an `AttrTokenStream` with nested
1548 /// `AttrTokenTree::Delimited` tokens.
1549 #[derive(Debug, Clone)]
1550 pub enum FlatToken {
1551 /// A token - this holds both delimiter (e.g. '{' and '}')
1552 /// and non-delimiter tokens
1554 /// Holds the `AttributesData` for an AST node. The
1555 /// `AttributesData` is inserted directly into the
1556 /// constructed `AttrTokenStream` as
1557 /// an `AttrTokenTree::Attributes`.
1558 AttrTarget(AttributesData),
1559 /// A special 'empty' token that is ignored during the conversion
1560 /// to an `AttrTokenStream`. This is used to simplify the
1561 /// handling of replace ranges.