]>
Commit | Line | Data |
---|---|---|
1 | pub mod attr; | |
2 | mod attr_wrapper; | |
3 | mod diagnostics; | |
4 | mod expr; | |
5 | mod generics; | |
6 | mod item; | |
7 | mod nonterminal; | |
8 | mod pat; | |
9 | mod path; | |
10 | mod stmt; | |
11 | mod ty; | |
12 | ||
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; | |
19 | ||
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, AttrVec, 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; | |
34 | use rustc_errors::{ | |
35 | Applicability, DiagnosticBuilder, ErrorGuaranteed, FatalError, IntoDiagnostic, MultiSpan, | |
36 | }; | |
37 | use rustc_session::parse::ParseSess; | |
38 | use rustc_span::source_map::{Span, DUMMY_SP}; | |
39 | use rustc_span::symbol::{kw, sym, Ident, Symbol}; | |
40 | ||
41 | use std::ops::Range; | |
42 | use std::{cmp, mem, slice}; | |
43 | ||
44 | use crate::errors::{ | |
45 | DocCommentDoesNotDocumentAnything, IncorrectVisibilityRestriction, MismatchedClosingDelimiter, | |
46 | NonStringAbiLiteral, | |
47 | }; | |
48 | ||
49 | bitflags::bitflags! { | |
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; | |
55 | } | |
56 | } | |
57 | ||
58 | #[derive(Clone, Copy, PartialEq, Debug)] | |
59 | enum SemiColonMode { | |
60 | Break, | |
61 | Ignore, | |
62 | Comma, | |
63 | } | |
64 | ||
65 | #[derive(Clone, Copy, PartialEq, Debug)] | |
66 | enum BlockMode { | |
67 | Break, | |
68 | Ignore, | |
69 | } | |
70 | ||
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 { | |
75 | Yes, | |
76 | No, | |
77 | } | |
78 | ||
79 | #[derive(Debug, Eq, PartialEq)] | |
80 | pub enum TrailingToken { | |
81 | None, | |
82 | Semi, | |
83 | Gt, | |
84 | /// If the trailing token is a comma, then capture it | |
85 | /// Otherwise, ignore the trailing token | |
86 | MaybeComma, | |
87 | } | |
88 | ||
89 | /// Like `maybe_whole_expr`, but for things other than expressions. | |
90 | #[macro_export] | |
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 { | |
95 | let $x = x.clone(); | |
96 | $p.bump(); | |
97 | return Ok($e); | |
98 | } | |
99 | } | |
100 | }; | |
101 | } | |
102 | ||
103 | /// If the next tokens are ill-formed `$ty::` recover them as `<$ty>::`. | |
104 | #[macro_export] | |
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 | |
112 | { | |
113 | let ty = ty.clone(); | |
114 | $self.bump(); | |
115 | return $self.maybe_recover_from_bad_qpath_stage_2($self.prev_token.span, ty); | |
116 | } | |
117 | }; | |
118 | } | |
119 | ||
120 | #[derive(Clone, Copy)] | |
121 | pub enum Recovery { | |
122 | Allowed, | |
123 | Forbidden, | |
124 | } | |
125 | ||
126 | #[derive(Clone)] | |
127 | pub struct Parser<'a> { | |
128 | pub sess: &'a ParseSess, | |
129 | /// The current token. | |
130 | pub token: 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 | |
144 | /// appropriately. | |
145 | /// | |
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 | |
151 | /// error. | |
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, | |
166 | } | |
167 | ||
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); | |
172 | ||
173 | /// Stores span information about a closure. | |
174 | #[derive(Clone)] | |
175 | pub struct ClosureSpans { | |
176 | pub whole_closure: Span, | |
177 | pub closing_pipe: Span, | |
178 | pub body: Span, | |
179 | } | |
180 | ||
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: | |
184 | /// | |
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`. | |
190 | /// | |
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)>); | |
199 | ||
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)] | |
204 | pub enum Capturing { | |
205 | /// We aren't performing any capturing - this is the default mode. | |
206 | No, | |
207 | /// We are capturing tokens | |
208 | Yes, | |
209 | } | |
210 | ||
211 | #[derive(Clone)] | |
212 | struct CaptureState { | |
213 | capturing: Capturing, | |
214 | replace_ranges: Vec<ReplaceRange>, | |
215 | inner_attr_ranges: FxHashMap<AttrId, ReplaceRange>, | |
216 | } | |
217 | ||
218 | impl<'a> Drop for Parser<'a> { | |
219 | fn drop(&mut self) { | |
220 | emit_unclosed_delims(&mut self.unclosed_delims, &self.sess); | |
221 | } | |
222 | } | |
223 | ||
224 | #[derive(Clone)] | |
225 | struct TokenCursor { | |
226 | // The current (innermost) frame. `frame` and `stack` could be combined, | |
227 | // but it's faster to have them separately to 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`. | |
244 | // | |
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. | |
248 | // | |
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, | |
258 | } | |
259 | ||
260 | #[derive(Clone)] | |
261 | struct TokenCursorFrame { | |
262 | delim_sp: Option<(Delimiter, DelimSpan)>, | |
263 | tree_cursor: tokenstream::Cursor, | |
264 | } | |
265 | ||
266 | impl TokenCursorFrame { | |
267 | fn new(delim_sp: Option<(Delimiter, DelimSpan)>, tts: TokenStream) -> Self { | |
268 | TokenCursorFrame { delim_sp, tree_cursor: tts.into_trees() } | |
269 | } | |
270 | } | |
271 | ||
272 | impl TokenCursor { | |
273 | fn next(&mut self, desugar_doc_comments: bool) -> (Token, Spacing) { | |
274 | self.inlined_next(desugar_doc_comments) | |
275 | } | |
276 | ||
277 | /// This always-inlined version should only be used on hot code paths. | |
278 | #[inline(always)] | |
279 | fn inlined_next(&mut self, desugar_doc_comments: bool) -> (Token, Spacing) { | |
280 | loop { | |
281 | // FIXME: we currently don't return `Delimiter` open/close delims. To fix #67062 we will | |
282 | // need to, whereupon the `delim != Delimiter::Invisible` conditions below can be | |
283 | // removed. | |
284 | if let Some(tree) = self.frame.tree_cursor.next_ref() { | |
285 | match tree { | |
286 | &TokenTree::Token(ref token, spacing) => match (desugar_doc_comments, token) { | |
287 | (true, &Token { kind: token::DocComment(_, attr_style, data), span }) => { | |
288 | return self.desugar(attr_style, data, span); | |
289 | } | |
290 | _ => return (token.clone(), spacing), | |
291 | }, | |
292 | &TokenTree::Delimited(sp, delim, ref tts) => { | |
293 | // Set `open_delim` to true here because we deal with it immediately. | |
294 | let frame = TokenCursorFrame::new(Some((delim, sp)), tts.clone()); | |
295 | self.stack.push(mem::replace(&mut self.frame, frame)); | |
296 | if delim != Delimiter::Invisible { | |
297 | return (Token::new(token::OpenDelim(delim), sp.open), Spacing::Alone); | |
298 | } | |
299 | // No open delimiter to return; continue on to the next iteration. | |
300 | } | |
301 | }; | |
302 | } else if let Some(frame) = self.stack.pop() { | |
303 | if let Some((delim, span)) = self.frame.delim_sp && delim != Delimiter::Invisible { | |
304 | self.frame = frame; | |
305 | return (Token::new(token::CloseDelim(delim), span.close), Spacing::Alone); | |
306 | } | |
307 | self.frame = frame; | |
308 | // No close delimiter to return; continue on to the next iteration. | |
309 | } else { | |
310 | return (Token::new(token::Eof, DUMMY_SP), Spacing::Alone); | |
311 | } | |
312 | } | |
313 | } | |
314 | ||
315 | fn desugar(&mut self, attr_style: AttrStyle, data: Symbol, span: Span) -> (Token, Spacing) { | |
316 | // Searches for the occurrences of `"#*` and returns the minimum number of `#`s | |
317 | // required to wrap the text. E.g. | |
318 | // - `abc d` is wrapped as `r"abc d"` (num_of_hashes = 0) | |
319 | // - `abc "d"` is wrapped as `r#"abc "d""#` (num_of_hashes = 1) | |
320 | // - `abc "##d##"` is wrapped as `r###"abc "d""###` (num_of_hashes = 3) | |
321 | let mut num_of_hashes = 0; | |
322 | let mut count = 0; | |
323 | for ch in data.as_str().chars() { | |
324 | count = match ch { | |
325 | '"' => 1, | |
326 | '#' if count > 0 => count + 1, | |
327 | _ => 0, | |
328 | }; | |
329 | num_of_hashes = cmp::max(num_of_hashes, count); | |
330 | } | |
331 | ||
332 | // `/// foo` becomes `doc = r"foo". | |
333 | let delim_span = DelimSpan::from_single(span); | |
334 | let body = TokenTree::Delimited( | |
335 | delim_span, | |
336 | Delimiter::Bracket, | |
337 | [ | |
338 | TokenTree::token_alone(token::Ident(sym::doc, false), span), | |
339 | TokenTree::token_alone(token::Eq, span), | |
340 | TokenTree::token_alone( | |
341 | TokenKind::lit(token::StrRaw(num_of_hashes), data, None), | |
342 | span, | |
343 | ), | |
344 | ] | |
345 | .into_iter() | |
346 | .collect::<TokenStream>(), | |
347 | ); | |
348 | ||
349 | self.stack.push(mem::replace( | |
350 | &mut self.frame, | |
351 | TokenCursorFrame::new( | |
352 | None, | |
353 | if attr_style == AttrStyle::Inner { | |
354 | [ | |
355 | TokenTree::token_alone(token::Pound, span), | |
356 | TokenTree::token_alone(token::Not, span), | |
357 | body, | |
358 | ] | |
359 | .into_iter() | |
360 | .collect::<TokenStream>() | |
361 | } else { | |
362 | [TokenTree::token_alone(token::Pound, span), body] | |
363 | .into_iter() | |
364 | .collect::<TokenStream>() | |
365 | }, | |
366 | ), | |
367 | )); | |
368 | ||
369 | self.next(/* desugar_doc_comments */ false) | |
370 | } | |
371 | } | |
372 | ||
373 | #[derive(Debug, Clone, PartialEq)] | |
374 | enum TokenType { | |
375 | Token(TokenKind), | |
376 | Keyword(Symbol), | |
377 | Operator, | |
378 | Lifetime, | |
379 | Ident, | |
380 | Path, | |
381 | Type, | |
382 | Const, | |
383 | } | |
384 | ||
385 | impl TokenType { | |
386 | fn to_string(&self) -> String { | |
387 | match self { | |
388 | TokenType::Token(t) => format!("`{}`", pprust::token_kind_to_string(t)), | |
389 | TokenType::Keyword(kw) => format!("`{}`", kw), | |
390 | TokenType::Operator => "an operator".to_string(), | |
391 | TokenType::Lifetime => "lifetime".to_string(), | |
392 | TokenType::Ident => "identifier".to_string(), | |
393 | TokenType::Path => "path".to_string(), | |
394 | TokenType::Type => "type".to_string(), | |
395 | TokenType::Const => "a const expression".to_string(), | |
396 | } | |
397 | } | |
398 | } | |
399 | ||
400 | #[derive(Copy, Clone, Debug)] | |
401 | enum TokenExpectType { | |
402 | Expect, | |
403 | NoExpect, | |
404 | } | |
405 | ||
406 | /// A sequence separator. | |
407 | struct SeqSep { | |
408 | /// The separator token. | |
409 | sep: Option<TokenKind>, | |
410 | /// `true` if a trailing separator is allowed. | |
411 | trailing_sep_allowed: bool, | |
412 | } | |
413 | ||
414 | impl SeqSep { | |
415 | fn trailing_allowed(t: TokenKind) -> SeqSep { | |
416 | SeqSep { sep: Some(t), trailing_sep_allowed: true } | |
417 | } | |
418 | ||
419 | fn none() -> SeqSep { | |
420 | SeqSep { sep: None, trailing_sep_allowed: false } | |
421 | } | |
422 | } | |
423 | ||
424 | pub enum FollowedByType { | |
425 | Yes, | |
426 | No, | |
427 | } | |
428 | ||
429 | #[derive(Clone, Copy, PartialEq, Eq)] | |
430 | pub enum TokenDescription { | |
431 | ReservedIdentifier, | |
432 | Keyword, | |
433 | ReservedKeyword, | |
434 | DocComment, | |
435 | } | |
436 | ||
437 | impl TokenDescription { | |
438 | pub fn from_token(token: &Token) -> Option<Self> { | |
439 | match token.kind { | |
440 | _ if token.is_special_ident() => Some(TokenDescription::ReservedIdentifier), | |
441 | _ if token.is_used_keyword() => Some(TokenDescription::Keyword), | |
442 | _ if token.is_unused_keyword() => Some(TokenDescription::ReservedKeyword), | |
443 | token::DocComment(..) => Some(TokenDescription::DocComment), | |
444 | _ => None, | |
445 | } | |
446 | } | |
447 | } | |
448 | ||
449 | pub(super) fn token_descr(token: &Token) -> String { | |
450 | let name = pprust::token_to_string(token).to_string(); | |
451 | ||
452 | let kind = TokenDescription::from_token(token).map(|kind| match kind { | |
453 | TokenDescription::ReservedIdentifier => "reserved identifier", | |
454 | TokenDescription::Keyword => "keyword", | |
455 | TokenDescription::ReservedKeyword => "reserved keyword", | |
456 | TokenDescription::DocComment => "doc comment", | |
457 | }); | |
458 | ||
459 | if let Some(kind) = kind { format!("{} `{}`", kind, name) } else { format!("`{}`", name) } | |
460 | } | |
461 | ||
462 | impl<'a> Parser<'a> { | |
463 | pub fn new( | |
464 | sess: &'a ParseSess, | |
465 | tokens: TokenStream, | |
466 | desugar_doc_comments: bool, | |
467 | subparser_name: Option<&'static str>, | |
468 | ) -> Self { | |
469 | let mut parser = Parser { | |
470 | sess, | |
471 | token: Token::dummy(), | |
472 | token_spacing: Spacing::Alone, | |
473 | prev_token: Token::dummy(), | |
474 | capture_cfg: false, | |
475 | restrictions: Restrictions::empty(), | |
476 | expected_tokens: Vec::new(), | |
477 | token_cursor: TokenCursor { | |
478 | frame: TokenCursorFrame::new(None, tokens), | |
479 | stack: Vec::new(), | |
480 | num_next_calls: 0, | |
481 | desugar_doc_comments, | |
482 | break_last_token: false, | |
483 | }, | |
484 | desugar_doc_comments, | |
485 | unmatched_angle_bracket_count: 0, | |
486 | max_angle_bracket_count: 0, | |
487 | unclosed_delims: Vec::new(), | |
488 | last_unexpected_token_span: None, | |
489 | last_type_ascription: None, | |
490 | subparser_name, | |
491 | capture_state: CaptureState { | |
492 | capturing: Capturing::No, | |
493 | replace_ranges: Vec::new(), | |
494 | inner_attr_ranges: Default::default(), | |
495 | }, | |
496 | current_closure: None, | |
497 | recovery: Recovery::Allowed, | |
498 | }; | |
499 | ||
500 | // Make parser point to the first token. | |
501 | parser.bump(); | |
502 | ||
503 | parser | |
504 | } | |
505 | ||
506 | pub fn recovery(mut self, recovery: Recovery) -> Self { | |
507 | self.recovery = recovery; | |
508 | self | |
509 | } | |
510 | ||
511 | /// Whether the parser is allowed to recover from broken code. | |
512 | /// | |
513 | /// If this returns false, recovering broken code into valid code (especially if this recovery does lookahead) | |
514 | /// is not allowed. All recovery done by the parser must be gated behind this check. | |
515 | /// | |
516 | /// Technically, this only needs to restrict eager recovery by doing lookahead at more tokens. | |
517 | /// But making the distinction is very subtle, and simply forbidding all recovery is a lot simpler to uphold. | |
518 | fn may_recover(&self) -> bool { | |
519 | matches!(self.recovery, Recovery::Allowed) | |
520 | } | |
521 | ||
522 | pub fn unexpected<T>(&mut self) -> PResult<'a, T> { | |
523 | match self.expect_one_of(&[], &[]) { | |
524 | Err(e) => Err(e), | |
525 | // We can get `Ok(true)` from `recover_closing_delimiter` | |
526 | // which is called in `expected_one_of_not_found`. | |
527 | Ok(_) => FatalError.raise(), | |
528 | } | |
529 | } | |
530 | ||
531 | /// Expects and consumes the token `t`. Signals an error if the next token is not `t`. | |
532 | pub fn expect(&mut self, t: &TokenKind) -> PResult<'a, bool /* recovered */> { | |
533 | if self.expected_tokens.is_empty() { | |
534 | if self.token == *t { | |
535 | self.bump(); | |
536 | Ok(false) | |
537 | } else { | |
538 | self.unexpected_try_recover(t) | |
539 | } | |
540 | } else { | |
541 | self.expect_one_of(slice::from_ref(t), &[]) | |
542 | } | |
543 | } | |
544 | ||
545 | /// Expect next token to be edible or inedible token. If edible, | |
546 | /// then consume it; if inedible, then return without consuming | |
547 | /// anything. Signal a fatal error if next token is unexpected. | |
548 | pub fn expect_one_of( | |
549 | &mut self, | |
550 | edible: &[TokenKind], | |
551 | inedible: &[TokenKind], | |
552 | ) -> PResult<'a, bool /* recovered */> { | |
553 | if edible.contains(&self.token.kind) { | |
554 | self.bump(); | |
555 | Ok(false) | |
556 | } else if inedible.contains(&self.token.kind) { | |
557 | // leave it in the input | |
558 | Ok(false) | |
559 | } else if self.last_unexpected_token_span == Some(self.token.span) { | |
560 | FatalError.raise(); | |
561 | } else { | |
562 | self.expected_one_of_not_found(edible, inedible) | |
563 | } | |
564 | } | |
565 | ||
566 | // Public for rustfmt usage. | |
567 | pub fn parse_ident(&mut self) -> PResult<'a, Ident> { | |
568 | self.parse_ident_common(true) | |
569 | } | |
570 | ||
571 | fn ident_or_err(&mut self) -> PResult<'a, (Ident, /* is_raw */ bool)> { | |
572 | self.token.ident().ok_or_else(|| match self.prev_token.kind { | |
573 | TokenKind::DocComment(..) => DocCommentDoesNotDocumentAnything { | |
574 | span: self.prev_token.span, | |
575 | missing_comma: None, | |
576 | } | |
577 | .into_diagnostic(&self.sess.span_diagnostic), | |
578 | _ => self.expected_ident_found(), | |
579 | }) | |
580 | } | |
581 | ||
582 | fn parse_ident_common(&mut self, recover: bool) -> PResult<'a, Ident> { | |
583 | let (ident, is_raw) = self.ident_or_err()?; | |
584 | if !is_raw && ident.is_reserved() { | |
585 | let mut err = self.expected_ident_found(); | |
586 | if recover { | |
587 | err.emit(); | |
588 | } else { | |
589 | return Err(err); | |
590 | } | |
591 | } | |
592 | self.bump(); | |
593 | Ok(ident) | |
594 | } | |
595 | ||
596 | /// Checks if the next token is `tok`, and returns `true` if so. | |
597 | /// | |
598 | /// This method will automatically add `tok` to `expected_tokens` if `tok` is not | |
599 | /// encountered. | |
600 | fn check(&mut self, tok: &TokenKind) -> bool { | |
601 | let is_present = self.token == *tok; | |
602 | if !is_present { | |
603 | self.expected_tokens.push(TokenType::Token(tok.clone())); | |
604 | } | |
605 | is_present | |
606 | } | |
607 | ||
608 | fn check_noexpect(&self, tok: &TokenKind) -> bool { | |
609 | self.token == *tok | |
610 | } | |
611 | ||
612 | /// Consumes a token 'tok' if it exists. Returns whether the given token was present. | |
613 | /// | |
614 | /// the main purpose of this function is to reduce the cluttering of the suggestions list | |
615 | /// which using the normal eat method could introduce in some cases. | |
616 | pub fn eat_noexpect(&mut self, tok: &TokenKind) -> bool { | |
617 | let is_present = self.check_noexpect(tok); | |
618 | if is_present { | |
619 | self.bump() | |
620 | } | |
621 | is_present | |
622 | } | |
623 | ||
624 | /// Consumes a token 'tok' if it exists. Returns whether the given token was present. | |
625 | pub fn eat(&mut self, tok: &TokenKind) -> bool { | |
626 | let is_present = self.check(tok); | |
627 | if is_present { | |
628 | self.bump() | |
629 | } | |
630 | is_present | |
631 | } | |
632 | ||
633 | /// If the next token is the given keyword, returns `true` without eating it. | |
634 | /// An expectation is also added for diagnostics purposes. | |
635 | fn check_keyword(&mut self, kw: Symbol) -> bool { | |
636 | self.expected_tokens.push(TokenType::Keyword(kw)); | |
637 | self.token.is_keyword(kw) | |
638 | } | |
639 | ||
640 | fn check_keyword_case(&mut self, kw: Symbol, case: Case) -> bool { | |
641 | if self.check_keyword(kw) { | |
642 | return true; | |
643 | } | |
644 | ||
645 | if case == Case::Insensitive | |
646 | && let Some((ident, /* is_raw */ false)) = self.token.ident() | |
647 | && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() { | |
648 | true | |
649 | } else { | |
650 | false | |
651 | } | |
652 | } | |
653 | ||
654 | /// If the next token is the given keyword, eats it and returns `true`. | |
655 | /// Otherwise, returns `false`. An expectation is also added for diagnostics purposes. | |
656 | // Public for rustfmt usage. | |
657 | pub fn eat_keyword(&mut self, kw: Symbol) -> bool { | |
658 | if self.check_keyword(kw) { | |
659 | self.bump(); | |
660 | true | |
661 | } else { | |
662 | false | |
663 | } | |
664 | } | |
665 | ||
666 | /// Eats a keyword, optionally ignoring the case. | |
667 | /// If the case differs (and is ignored) an error is issued. | |
668 | /// This is useful for recovery. | |
669 | fn eat_keyword_case(&mut self, kw: Symbol, case: Case) -> bool { | |
670 | if self.eat_keyword(kw) { | |
671 | return true; | |
672 | } | |
673 | ||
674 | if case == Case::Insensitive | |
675 | && let Some((ident, /* is_raw */ false)) = self.token.ident() | |
676 | && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() { | |
677 | self | |
678 | .struct_span_err(ident.span, format!("keyword `{kw}` is written in a wrong case")) | |
679 | .span_suggestion( | |
680 | ident.span, | |
681 | "write it in the correct case", | |
682 | kw, | |
683 | Applicability::MachineApplicable | |
684 | ).emit(); | |
685 | ||
686 | self.bump(); | |
687 | return true; | |
688 | } | |
689 | ||
690 | false | |
691 | } | |
692 | ||
693 | fn eat_keyword_noexpect(&mut self, kw: Symbol) -> bool { | |
694 | if self.token.is_keyword(kw) { | |
695 | self.bump(); | |
696 | true | |
697 | } else { | |
698 | false | |
699 | } | |
700 | } | |
701 | ||
702 | /// If the given word is not a keyword, signals an error. | |
703 | /// If the next token is not the given word, signals an error. | |
704 | /// Otherwise, eats it. | |
705 | fn expect_keyword(&mut self, kw: Symbol) -> PResult<'a, ()> { | |
706 | if !self.eat_keyword(kw) { self.unexpected() } else { Ok(()) } | |
707 | } | |
708 | ||
709 | /// Is the given keyword `kw` followed by a non-reserved identifier? | |
710 | fn is_kw_followed_by_ident(&self, kw: Symbol) -> bool { | |
711 | self.token.is_keyword(kw) && self.look_ahead(1, |t| t.is_ident() && !t.is_reserved_ident()) | |
712 | } | |
713 | ||
714 | fn check_or_expected(&mut self, ok: bool, typ: TokenType) -> bool { | |
715 | if ok { | |
716 | true | |
717 | } else { | |
718 | self.expected_tokens.push(typ); | |
719 | false | |
720 | } | |
721 | } | |
722 | ||
723 | fn check_ident(&mut self) -> bool { | |
724 | self.check_or_expected(self.token.is_ident(), TokenType::Ident) | |
725 | } | |
726 | ||
727 | fn check_path(&mut self) -> bool { | |
728 | self.check_or_expected(self.token.is_path_start(), TokenType::Path) | |
729 | } | |
730 | ||
731 | fn check_type(&mut self) -> bool { | |
732 | self.check_or_expected(self.token.can_begin_type(), TokenType::Type) | |
733 | } | |
734 | ||
735 | fn check_const_arg(&mut self) -> bool { | |
736 | self.check_or_expected(self.token.can_begin_const_arg(), TokenType::Const) | |
737 | } | |
738 | ||
739 | fn check_inline_const(&self, dist: usize) -> bool { | |
740 | self.is_keyword_ahead(dist, &[kw::Const]) | |
741 | && self.look_ahead(dist + 1, |t| match &t.kind { | |
742 | token::Interpolated(nt) => matches!(**nt, token::NtBlock(..)), | |
743 | token::OpenDelim(Delimiter::Brace) => true, | |
744 | _ => false, | |
745 | }) | |
746 | } | |
747 | ||
748 | /// Checks to see if the next token is either `+` or `+=`. | |
749 | /// Otherwise returns `false`. | |
750 | fn check_plus(&mut self) -> bool { | |
751 | self.check_or_expected( | |
752 | self.token.is_like_plus(), | |
753 | TokenType::Token(token::BinOp(token::Plus)), | |
754 | ) | |
755 | } | |
756 | ||
757 | /// Eats the expected token if it's present possibly breaking | |
758 | /// compound tokens like multi-character operators in process. | |
759 | /// Returns `true` if the token was eaten. | |
760 | fn break_and_eat(&mut self, expected: TokenKind) -> bool { | |
761 | if self.token.kind == expected { | |
762 | self.bump(); | |
763 | return true; | |
764 | } | |
765 | match self.token.kind.break_two_token_op() { | |
766 | Some((first, second)) if first == expected => { | |
767 | let first_span = self.sess.source_map().start_point(self.token.span); | |
768 | let second_span = self.token.span.with_lo(first_span.hi()); | |
769 | self.token = Token::new(first, first_span); | |
770 | // Keep track of this token - if we end token capturing now, | |
771 | // we'll want to append this token to the captured stream. | |
772 | // | |
773 | // If we consume any additional tokens, then this token | |
774 | // is not needed (we'll capture the entire 'glued' token), | |
775 | // and `bump` will set this field to `None` | |
776 | self.token_cursor.break_last_token = true; | |
777 | // Use the spacing of the glued token as the spacing | |
778 | // of the unglued second token. | |
779 | self.bump_with((Token::new(second, second_span), self.token_spacing)); | |
780 | true | |
781 | } | |
782 | _ => { | |
783 | self.expected_tokens.push(TokenType::Token(expected)); | |
784 | false | |
785 | } | |
786 | } | |
787 | } | |
788 | ||
789 | /// Eats `+` possibly breaking tokens like `+=` in process. | |
790 | fn eat_plus(&mut self) -> bool { | |
791 | self.break_and_eat(token::BinOp(token::Plus)) | |
792 | } | |
793 | ||
794 | /// Eats `&` possibly breaking tokens like `&&` in process. | |
795 | /// Signals an error if `&` is not eaten. | |
796 | fn expect_and(&mut self) -> PResult<'a, ()> { | |
797 | if self.break_and_eat(token::BinOp(token::And)) { Ok(()) } else { self.unexpected() } | |
798 | } | |
799 | ||
800 | /// Eats `|` possibly breaking tokens like `||` in process. | |
801 | /// Signals an error if `|` was not eaten. | |
802 | fn expect_or(&mut self) -> PResult<'a, ()> { | |
803 | if self.break_and_eat(token::BinOp(token::Or)) { Ok(()) } else { self.unexpected() } | |
804 | } | |
805 | ||
806 | /// Eats `<` possibly breaking tokens like `<<` in process. | |
807 | fn eat_lt(&mut self) -> bool { | |
808 | let ate = self.break_and_eat(token::Lt); | |
809 | if ate { | |
810 | // See doc comment for `unmatched_angle_bracket_count`. | |
811 | self.unmatched_angle_bracket_count += 1; | |
812 | self.max_angle_bracket_count += 1; | |
813 | debug!("eat_lt: (increment) count={:?}", self.unmatched_angle_bracket_count); | |
814 | } | |
815 | ate | |
816 | } | |
817 | ||
818 | /// Eats `<` possibly breaking tokens like `<<` in process. | |
819 | /// Signals an error if `<` was not eaten. | |
820 | fn expect_lt(&mut self) -> PResult<'a, ()> { | |
821 | if self.eat_lt() { Ok(()) } else { self.unexpected() } | |
822 | } | |
823 | ||
824 | /// Eats `>` possibly breaking tokens like `>>` in process. | |
825 | /// Signals an error if `>` was not eaten. | |
826 | fn expect_gt(&mut self) -> PResult<'a, ()> { | |
827 | if self.break_and_eat(token::Gt) { | |
828 | // See doc comment for `unmatched_angle_bracket_count`. | |
829 | if self.unmatched_angle_bracket_count > 0 { | |
830 | self.unmatched_angle_bracket_count -= 1; | |
831 | debug!("expect_gt: (decrement) count={:?}", self.unmatched_angle_bracket_count); | |
832 | } | |
833 | Ok(()) | |
834 | } else { | |
835 | self.unexpected() | |
836 | } | |
837 | } | |
838 | ||
839 | fn expect_any_with_type(&mut self, kets: &[&TokenKind], expect: TokenExpectType) -> bool { | |
840 | kets.iter().any(|k| match expect { | |
841 | TokenExpectType::Expect => self.check(k), | |
842 | TokenExpectType::NoExpect => self.token == **k, | |
843 | }) | |
844 | } | |
845 | ||
846 | fn parse_seq_to_before_tokens<T>( | |
847 | &mut self, | |
848 | kets: &[&TokenKind], | |
849 | sep: SeqSep, | |
850 | expect: TokenExpectType, | |
851 | mut f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, | |
852 | ) -> PResult<'a, (Vec<T>, bool /* trailing */, bool /* recovered */)> { | |
853 | let mut first = true; | |
854 | let mut recovered = false; | |
855 | let mut trailing = false; | |
856 | let mut v = vec![]; | |
857 | let unclosed_delims = !self.unclosed_delims.is_empty(); | |
858 | ||
859 | while !self.expect_any_with_type(kets, expect) { | |
860 | if let token::CloseDelim(..) | token::Eof = self.token.kind { | |
861 | break; | |
862 | } | |
863 | if let Some(t) = &sep.sep { | |
864 | if first { | |
865 | first = false; | |
866 | } else { | |
867 | match self.expect(t) { | |
868 | Ok(false) => { | |
869 | self.current_closure.take(); | |
870 | } | |
871 | Ok(true) => { | |
872 | self.current_closure.take(); | |
873 | recovered = true; | |
874 | break; | |
875 | } | |
876 | Err(mut expect_err) => { | |
877 | let sp = self.prev_token.span.shrink_to_hi(); | |
878 | let token_str = pprust::token_kind_to_string(t); | |
879 | ||
880 | match self.current_closure.take() { | |
881 | Some(closure_spans) if self.token.kind == TokenKind::Semi => { | |
882 | // Finding a semicolon instead of a comma | |
883 | // after a closure body indicates that the | |
884 | // closure body may be a block but the user | |
885 | // forgot to put braces around its | |
886 | // statements. | |
887 | ||
888 | self.recover_missing_braces_around_closure_body( | |
889 | closure_spans, | |
890 | expect_err, | |
891 | )?; | |
892 | ||
893 | continue; | |
894 | } | |
895 | ||
896 | _ => { | |
897 | // Attempt to keep parsing if it was a similar separator. | |
898 | if let Some(tokens) = t.similar_tokens() { | |
899 | if tokens.contains(&self.token.kind) && !unclosed_delims { | |
900 | self.bump(); | |
901 | } | |
902 | } | |
903 | } | |
904 | } | |
905 | ||
906 | // If this was a missing `@` in a binding pattern | |
907 | // bail with a suggestion | |
908 | // https://github.com/rust-lang/rust/issues/72373 | |
909 | if self.prev_token.is_ident() && self.token.kind == token::DotDot { | |
910 | let msg = format!( | |
911 | "if you meant to bind the contents of \ | |
912 | the rest of the array pattern into `{}`, use `@`", | |
913 | pprust::token_to_string(&self.prev_token) | |
914 | ); | |
915 | expect_err | |
916 | .span_suggestion_verbose( | |
917 | self.prev_token.span.shrink_to_hi().until(self.token.span), | |
918 | &msg, | |
919 | " @ ", | |
920 | Applicability::MaybeIncorrect, | |
921 | ) | |
922 | .emit(); | |
923 | break; | |
924 | } | |
925 | ||
926 | // Attempt to keep parsing if it was an omitted separator. | |
927 | match f(self) { | |
928 | Ok(t) => { | |
929 | // Parsed successfully, therefore most probably the code only | |
930 | // misses a separator. | |
931 | expect_err | |
932 | .span_suggestion_short( | |
933 | sp, | |
934 | &format!("missing `{}`", token_str), | |
935 | token_str, | |
936 | Applicability::MaybeIncorrect, | |
937 | ) | |
938 | .emit(); | |
939 | ||
940 | v.push(t); | |
941 | continue; | |
942 | } | |
943 | Err(e) => { | |
944 | // Parsing failed, therefore it must be something more serious | |
945 | // than just a missing separator. | |
946 | for xx in &e.children { | |
947 | // propagate the help message from sub error 'e' to main error 'expect_err; | |
948 | expect_err.children.push(xx.clone()); | |
949 | } | |
950 | expect_err.emit(); | |
951 | ||
952 | e.cancel(); | |
953 | break; | |
954 | } | |
955 | } | |
956 | } | |
957 | } | |
958 | } | |
959 | } | |
960 | if sep.trailing_sep_allowed && self.expect_any_with_type(kets, expect) { | |
961 | trailing = true; | |
962 | break; | |
963 | } | |
964 | ||
965 | let t = f(self)?; | |
966 | v.push(t); | |
967 | } | |
968 | ||
969 | Ok((v, trailing, recovered)) | |
970 | } | |
971 | ||
972 | fn recover_missing_braces_around_closure_body( | |
973 | &mut self, | |
974 | closure_spans: ClosureSpans, | |
975 | mut expect_err: DiagnosticBuilder<'_, ErrorGuaranteed>, | |
976 | ) -> PResult<'a, ()> { | |
977 | let initial_semicolon = self.token.span; | |
978 | ||
979 | while self.eat(&TokenKind::Semi) { | |
980 | let _ = self.parse_stmt(ForceCollect::Yes)?; | |
981 | } | |
982 | ||
983 | expect_err.set_primary_message( | |
984 | "closure bodies that contain statements must be surrounded by braces", | |
985 | ); | |
986 | ||
987 | let preceding_pipe_span = closure_spans.closing_pipe; | |
988 | let following_token_span = self.token.span; | |
989 | ||
990 | let mut first_note = MultiSpan::from(vec![initial_semicolon]); | |
991 | first_note.push_span_label( | |
992 | initial_semicolon, | |
993 | "this `;` turns the preceding closure into a statement", | |
994 | ); | |
995 | first_note.push_span_label( | |
996 | closure_spans.body, | |
997 | "this expression is a statement because of the trailing semicolon", | |
998 | ); | |
999 | expect_err.span_note(first_note, "statement found outside of a block"); | |
1000 | ||
1001 | let mut second_note = MultiSpan::from(vec![closure_spans.whole_closure]); | |
1002 | second_note.push_span_label(closure_spans.whole_closure, "this is the parsed closure..."); | |
1003 | second_note.push_span_label( | |
1004 | following_token_span, | |
1005 | "...but likely you meant the closure to end here", | |
1006 | ); | |
1007 | expect_err.span_note(second_note, "the closure body may be incorrectly delimited"); | |
1008 | ||
1009 | expect_err.set_span(vec![preceding_pipe_span, following_token_span]); | |
1010 | ||
1011 | let opening_suggestion_str = " {".to_string(); | |
1012 | let closing_suggestion_str = "}".to_string(); | |
1013 | ||
1014 | expect_err.multipart_suggestion( | |
1015 | "try adding braces", | |
1016 | vec![ | |
1017 | (preceding_pipe_span.shrink_to_hi(), opening_suggestion_str), | |
1018 | (following_token_span.shrink_to_lo(), closing_suggestion_str), | |
1019 | ], | |
1020 | Applicability::MaybeIncorrect, | |
1021 | ); | |
1022 | ||
1023 | expect_err.emit(); | |
1024 | ||
1025 | Ok(()) | |
1026 | } | |
1027 | ||
1028 | /// Parses a sequence, not including the closing delimiter. The function | |
1029 | /// `f` must consume tokens until reaching the next separator or | |
1030 | /// closing bracket. | |
1031 | fn parse_seq_to_before_end<T>( | |
1032 | &mut self, | |
1033 | ket: &TokenKind, | |
1034 | sep: SeqSep, | |
1035 | f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, | |
1036 | ) -> PResult<'a, (Vec<T>, bool, bool)> { | |
1037 | self.parse_seq_to_before_tokens(&[ket], sep, TokenExpectType::Expect, f) | |
1038 | } | |
1039 | ||
1040 | /// Parses a sequence, including the closing delimiter. The function | |
1041 | /// `f` must consume tokens until reaching the next separator or | |
1042 | /// closing bracket. | |
1043 | fn parse_seq_to_end<T>( | |
1044 | &mut self, | |
1045 | ket: &TokenKind, | |
1046 | sep: SeqSep, | |
1047 | f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, | |
1048 | ) -> PResult<'a, (Vec<T>, bool /* trailing */)> { | |
1049 | let (val, trailing, recovered) = self.parse_seq_to_before_end(ket, sep, f)?; | |
1050 | if !recovered { | |
1051 | self.eat(ket); | |
1052 | } | |
1053 | Ok((val, trailing)) | |
1054 | } | |
1055 | ||
1056 | /// Parses a sequence, including the closing delimiter. The function | |
1057 | /// `f` must consume tokens until reaching the next separator or | |
1058 | /// closing bracket. | |
1059 | fn parse_unspanned_seq<T>( | |
1060 | &mut self, | |
1061 | bra: &TokenKind, | |
1062 | ket: &TokenKind, | |
1063 | sep: SeqSep, | |
1064 | f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, | |
1065 | ) -> PResult<'a, (Vec<T>, bool)> { | |
1066 | self.expect(bra)?; | |
1067 | self.parse_seq_to_end(ket, sep, f) | |
1068 | } | |
1069 | ||
1070 | fn parse_delim_comma_seq<T>( | |
1071 | &mut self, | |
1072 | delim: Delimiter, | |
1073 | f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, | |
1074 | ) -> PResult<'a, (Vec<T>, bool)> { | |
1075 | self.parse_unspanned_seq( | |
1076 | &token::OpenDelim(delim), | |
1077 | &token::CloseDelim(delim), | |
1078 | SeqSep::trailing_allowed(token::Comma), | |
1079 | f, | |
1080 | ) | |
1081 | } | |
1082 | ||
1083 | fn parse_paren_comma_seq<T>( | |
1084 | &mut self, | |
1085 | f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, | |
1086 | ) -> PResult<'a, (Vec<T>, bool)> { | |
1087 | self.parse_delim_comma_seq(Delimiter::Parenthesis, f) | |
1088 | } | |
1089 | ||
1090 | /// Advance the parser by one token using provided token as the next one. | |
1091 | fn bump_with(&mut self, next: (Token, Spacing)) { | |
1092 | self.inlined_bump_with(next) | |
1093 | } | |
1094 | ||
1095 | /// This always-inlined version should only be used on hot code paths. | |
1096 | #[inline(always)] | |
1097 | fn inlined_bump_with(&mut self, (next_token, next_spacing): (Token, Spacing)) { | |
1098 | // Update the current and previous tokens. | |
1099 | self.prev_token = mem::replace(&mut self.token, next_token); | |
1100 | self.token_spacing = next_spacing; | |
1101 | ||
1102 | // Diagnostics. | |
1103 | self.expected_tokens.clear(); | |
1104 | } | |
1105 | ||
1106 | /// Advance the parser by one token. | |
1107 | pub fn bump(&mut self) { | |
1108 | // Note: destructuring here would give nicer code, but it was found in #96210 to be slower | |
1109 | // than `.0`/`.1` access. | |
1110 | let mut next = self.token_cursor.inlined_next(self.desugar_doc_comments); | |
1111 | self.token_cursor.num_next_calls += 1; | |
1112 | // We've retrieved an token from the underlying | |
1113 | // cursor, so we no longer need to worry about | |
1114 | // an unglued token. See `break_and_eat` for more details | |
1115 | self.token_cursor.break_last_token = false; | |
1116 | if next.0.span.is_dummy() { | |
1117 | // Tweak the location for better diagnostics, but keep syntactic context intact. | |
1118 | let fallback_span = self.token.span; | |
1119 | next.0.span = fallback_span.with_ctxt(next.0.span.ctxt()); | |
1120 | } | |
1121 | debug_assert!(!matches!( | |
1122 | next.0.kind, | |
1123 | token::OpenDelim(Delimiter::Invisible) | token::CloseDelim(Delimiter::Invisible) | |
1124 | )); | |
1125 | self.inlined_bump_with(next) | |
1126 | } | |
1127 | ||
1128 | /// Look-ahead `dist` tokens of `self.token` and get access to that token there. | |
1129 | /// When `dist == 0` then the current token is looked at. | |
1130 | pub fn look_ahead<R>(&self, dist: usize, looker: impl FnOnce(&Token) -> R) -> R { | |
1131 | if dist == 0 { | |
1132 | return looker(&self.token); | |
1133 | } | |
1134 | ||
1135 | let frame = &self.token_cursor.frame; | |
1136 | if let Some((delim, span)) = frame.delim_sp && delim != Delimiter::Invisible { | |
1137 | let all_normal = (0..dist).all(|i| { | |
1138 | let token = frame.tree_cursor.look_ahead(i); | |
1139 | !matches!(token, Some(TokenTree::Delimited(_, Delimiter::Invisible, _))) | |
1140 | }); | |
1141 | if all_normal { | |
1142 | return match frame.tree_cursor.look_ahead(dist - 1) { | |
1143 | Some(tree) => match tree { | |
1144 | TokenTree::Token(token, _) => looker(token), | |
1145 | TokenTree::Delimited(dspan, delim, _) => { | |
1146 | looker(&Token::new(token::OpenDelim(*delim), dspan.open)) | |
1147 | } | |
1148 | }, | |
1149 | None => looker(&Token::new(token::CloseDelim(delim), span.close)), | |
1150 | }; | |
1151 | } | |
1152 | } | |
1153 | ||
1154 | let mut cursor = self.token_cursor.clone(); | |
1155 | let mut i = 0; | |
1156 | let mut token = Token::dummy(); | |
1157 | while i < dist { | |
1158 | token = cursor.next(/* desugar_doc_comments */ false).0; | |
1159 | if matches!( | |
1160 | token.kind, | |
1161 | token::OpenDelim(Delimiter::Invisible) | token::CloseDelim(Delimiter::Invisible) | |
1162 | ) { | |
1163 | continue; | |
1164 | } | |
1165 | i += 1; | |
1166 | } | |
1167 | return looker(&token); | |
1168 | } | |
1169 | ||
1170 | /// Returns whether any of the given keywords are `dist` tokens ahead of the current one. | |
1171 | fn is_keyword_ahead(&self, dist: usize, kws: &[Symbol]) -> bool { | |
1172 | self.look_ahead(dist, |t| kws.iter().any(|&kw| t.is_keyword(kw))) | |
1173 | } | |
1174 | ||
1175 | /// Parses asyncness: `async` or nothing. | |
1176 | fn parse_asyncness(&mut self, case: Case) -> Async { | |
1177 | if self.eat_keyword_case(kw::Async, case) { | |
1178 | let span = self.prev_token.uninterpolated_span(); | |
1179 | Async::Yes { span, closure_id: DUMMY_NODE_ID, return_impl_trait_id: DUMMY_NODE_ID } | |
1180 | } else { | |
1181 | Async::No | |
1182 | } | |
1183 | } | |
1184 | ||
1185 | /// Parses unsafety: `unsafe` or nothing. | |
1186 | fn parse_unsafety(&mut self, case: Case) -> Unsafe { | |
1187 | if self.eat_keyword_case(kw::Unsafe, case) { | |
1188 | Unsafe::Yes(self.prev_token.uninterpolated_span()) | |
1189 | } else { | |
1190 | Unsafe::No | |
1191 | } | |
1192 | } | |
1193 | ||
1194 | /// Parses constness: `const` or nothing. | |
1195 | fn parse_constness(&mut self, case: Case) -> Const { | |
1196 | // Avoid const blocks to be parsed as const items | |
1197 | if self.look_ahead(1, |t| t != &token::OpenDelim(Delimiter::Brace)) | |
1198 | && self.eat_keyword_case(kw::Const, case) | |
1199 | { | |
1200 | Const::Yes(self.prev_token.uninterpolated_span()) | |
1201 | } else { | |
1202 | Const::No | |
1203 | } | |
1204 | } | |
1205 | ||
1206 | /// Parses inline const expressions. | |
1207 | fn parse_const_block(&mut self, span: Span, pat: bool) -> PResult<'a, P<Expr>> { | |
1208 | if pat { | |
1209 | self.sess.gated_spans.gate(sym::inline_const_pat, span); | |
1210 | } else { | |
1211 | self.sess.gated_spans.gate(sym::inline_const, span); | |
1212 | } | |
1213 | self.eat_keyword(kw::Const); | |
1214 | let (attrs, blk) = self.parse_inner_attrs_and_block()?; | |
1215 | let anon_const = AnonConst { | |
1216 | id: DUMMY_NODE_ID, | |
1217 | value: self.mk_expr(blk.span, ExprKind::Block(blk, None)), | |
1218 | }; | |
1219 | let blk_span = anon_const.value.span; | |
1220 | Ok(self.mk_expr_with_attrs( | |
1221 | span.to(blk_span), | |
1222 | ExprKind::ConstBlock(anon_const), | |
1223 | AttrVec::from(attrs), | |
1224 | )) | |
1225 | } | |
1226 | ||
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 } | |
1230 | } | |
1231 | ||
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) | |
1238 | } else { | |
1239 | None | |
1240 | } | |
1241 | } | |
1242 | ||
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 | |
1245 | { | |
1246 | if let Some(suffix) = suffix { | |
1247 | self.expect_no_tuple_index_suffix(self.token.span, suffix); | |
1248 | } | |
1249 | self.bump(); | |
1250 | Ok(Ident::new(symbol, self.prev_token.span)) | |
1251 | } else { | |
1252 | self.parse_ident_common(true) | |
1253 | } | |
1254 | } | |
1255 | ||
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() } | |
1258 | } | |
1259 | ||
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) | |
1263 | } else { | |
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()?)) | |
1267 | } else { | |
1268 | AttrArgs::Empty | |
1269 | } | |
1270 | }) | |
1271 | } | |
1272 | ||
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)) | |
1277 | { | |
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 { | |
1281 | dspan, | |
1282 | delim: MacDelimiter::from_token(delim).unwrap(), | |
1283 | tokens, | |
1284 | }), | |
1285 | _ => unreachable!(), | |
1286 | } | |
1287 | } else { | |
1288 | None | |
1289 | } | |
1290 | } | |
1291 | ||
1292 | fn parse_or_use_outer_attributes( | |
1293 | &mut self, | |
1294 | already_parsed_attrs: Option<AttrWrapper>, | |
1295 | ) -> PResult<'a, AttrWrapper> { | |
1296 | if let Some(attrs) = already_parsed_attrs { | |
1297 | Ok(attrs) | |
1298 | } else { | |
1299 | self.parse_outer_attributes() | |
1300 | } | |
1301 | } | |
1302 | ||
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(); | |
1311 | ||
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; | |
1318 | loop { | |
1319 | // Advance one token at a time, so `TokenCursor::next()` | |
1320 | // can capture these tokens if necessary. | |
1321 | self.bump(); | |
1322 | if self.token_cursor.stack.len() == target_depth { | |
1323 | debug_assert!(matches!(self.token.kind, token::CloseDelim(_))); | |
1324 | break; | |
1325 | } | |
1326 | } | |
1327 | ||
1328 | // Consume close delimiter | |
1329 | self.bump(); | |
1330 | TokenTree::Delimited(span, delim, stream) | |
1331 | } | |
1332 | token::CloseDelim(_) | token::Eof => unreachable!(), | |
1333 | _ => { | |
1334 | self.bump(); | |
1335 | TokenTree::Token(self.prev_token.clone(), Spacing::Alone) | |
1336 | } | |
1337 | } | |
1338 | } | |
1339 | ||
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()); | |
1345 | } | |
1346 | Ok(tts) | |
1347 | } | |
1348 | ||
1349 | pub fn parse_tokens(&mut self) -> TokenStream { | |
1350 | let mut result = Vec::new(); | |
1351 | loop { | |
1352 | match self.token.kind { | |
1353 | token::Eof | token::CloseDelim(..) => break, | |
1354 | _ => result.push(self.parse_token_tree()), | |
1355 | } | |
1356 | } | |
1357 | TokenStream::new(result) | |
1358 | } | |
1359 | ||
1360 | /// Evaluates the closure with restrictions in place. | |
1361 | /// | |
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; | |
1366 | let res = f(self); | |
1367 | self.restrictions = old; | |
1368 | res | |
1369 | } | |
1370 | ||
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()); | |
1379 | ||
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, | |
1387 | tokens: None, | |
1388 | }); | |
1389 | } | |
1390 | let lo = self.prev_token.span; | |
1391 | ||
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)`. | |
1399 | self.bump(); // `(` | |
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 { | |
1404 | path: P(path), | |
1405 | id: ast::DUMMY_NODE_ID, | |
1406 | shorthand: false, | |
1407 | }; | |
1408 | return Ok(Visibility { | |
1409 | span: lo.to(self.prev_token.span), | |
1410 | kind: vis, | |
1411 | tokens: None, | |
1412 | }); | |
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]) | |
1415 | { | |
1416 | // Parse `pub(crate)`, `pub(self)`, or `pub(super)`. | |
1417 | self.bump(); // `(` | |
1418 | let path = self.parse_path(PathStyle::Mod)?; // `crate`/`super`/`self` | |
1419 | self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; // `)` | |
1420 | let vis = VisibilityKind::Restricted { | |
1421 | path: P(path), | |
1422 | id: ast::DUMMY_NODE_ID, | |
1423 | shorthand: true, | |
1424 | }; | |
1425 | return Ok(Visibility { | |
1426 | span: lo.to(self.prev_token.span), | |
1427 | kind: vis, | |
1428 | tokens: None, | |
1429 | }); | |
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. | |
1435 | } | |
1436 | } | |
1437 | ||
1438 | Ok(Visibility { span: lo, kind: VisibilityKind::Public, tokens: None }) | |
1439 | } | |
1440 | ||
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, ()> { | |
1443 | self.bump(); // `(` | |
1444 | let path = self.parse_path(PathStyle::Mod)?; | |
1445 | self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; // `)` | |
1446 | ||
1447 | let path_str = pprust::path_to_string(&path); | |
1448 | self.sess.emit_err(IncorrectVisibilityRestriction { span: path.span, inner_str: path_str }); | |
1449 | ||
1450 | Ok(()) | |
1451 | } | |
1452 | ||
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); | |
1460 | } | |
1461 | Extern::from_abi(abi, extern_span) | |
1462 | } else { | |
1463 | Extern::None | |
1464 | } | |
1465 | } | |
1466 | ||
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, | |
1473 | _ => { | |
1474 | self.sess.emit_err(NonStringAbiLiteral { span: lit.span }); | |
1475 | None | |
1476 | } | |
1477 | }, | |
1478 | Err(None) => None, | |
1479 | } | |
1480 | } | |
1481 | ||
1482 | pub fn collect_tokens_no_attrs<R: HasAttrs + HasTokens>( | |
1483 | &mut self, | |
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(), | |
1490 | ForceCollect::Yes, | |
1491 | |this, _attrs| Ok((f(this)?, TrailingToken::None)), | |
1492 | ) | |
1493 | } | |
1494 | ||
1495 | /// `::{` or `::*` | |
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) | |
1500 | }) | |
1501 | } | |
1502 | ||
1503 | pub fn clear_expected_tokens(&mut self) { | |
1504 | self.expected_tokens.clear(); | |
1505 | } | |
1506 | } | |
1507 | ||
1508 | pub(crate) fn make_unclosed_delims_error( | |
1509 | unmatched: UnmatchedBrace, | |
1510 | sess: &ParseSess, | |
1511 | ) -> Option<DiagnosticBuilder<'_, ErrorGuaranteed>> { | |
1512 | // `None` here means an `Eof` was found. We already emit those errors elsewhere, we add them to | |
1513 | // `unmatched_braces` only for error recovery in the `Parser`. | |
1514 | let found_delim = unmatched.found_delim?; | |
1515 | let mut spans = vec![unmatched.found_span]; | |
1516 | if let Some(sp) = unmatched.unclosed_span { | |
1517 | spans.push(sp); | |
1518 | }; | |
1519 | let err = MismatchedClosingDelimiter { | |
1520 | spans, | |
1521 | delimiter: pprust::token_kind_to_string(&token::CloseDelim(found_delim)).to_string(), | |
1522 | unmatched: unmatched.found_span, | |
1523 | opening_candidate: unmatched.candidate_span, | |
1524 | unclosed: unmatched.unclosed_span, | |
1525 | } | |
1526 | .into_diagnostic(&sess.span_diagnostic); | |
1527 | Some(err) | |
1528 | } | |
1529 | ||
1530 | pub fn emit_unclosed_delims(unclosed_delims: &mut Vec<UnmatchedBrace>, sess: &ParseSess) { | |
1531 | *sess.reached_eof.borrow_mut() |= | |
1532 | unclosed_delims.iter().any(|unmatched_delim| unmatched_delim.found_delim.is_none()); | |
1533 | for unmatched in unclosed_delims.drain(..) { | |
1534 | if let Some(mut e) = make_unclosed_delims_error(unmatched, sess) { | |
1535 | e.emit(); | |
1536 | } | |
1537 | } | |
1538 | } | |
1539 | ||
1540 | /// A helper struct used when building an `AttrTokenStream` from | |
1541 | /// a `LazyAttrTokenStream`. Both delimiter and non-delimited tokens | |
1542 | /// are stored as `FlatToken::Token`. A vector of `FlatToken`s | |
1543 | /// is then 'parsed' to build up an `AttrTokenStream` with nested | |
1544 | /// `AttrTokenTree::Delimited` tokens. | |
1545 | #[derive(Debug, Clone)] | |
1546 | pub enum FlatToken { | |
1547 | /// A token - this holds both delimiter (e.g. '{' and '}') | |
1548 | /// and non-delimiter tokens | |
1549 | Token(Token), | |
1550 | /// Holds the `AttributesData` for an AST node. The | |
1551 | /// `AttributesData` is inserted directly into the | |
1552 | /// constructed `AttrTokenStream` as | |
1553 | /// an `AttrTokenTree::Attributes`. | |
1554 | AttrTarget(AttributesData), | |
1555 | /// A special 'empty' token that is ignored during the conversion | |
1556 | /// to an `AttrTokenStream`. This is used to simplify the | |
1557 | /// handling of replace ranges. | |
1558 | Empty, | |
1559 | } | |
1560 | ||
1561 | #[derive(Debug)] | |
1562 | pub enum NtOrTt { | |
1563 | Nt(Nonterminal), | |
1564 | Tt(TokenTree), | |
1565 | } |