]>
Commit | Line | Data |
---|---|---|
fc512014 | 1 | //! This is an NFA-based parser, which calls out to the main Rust parser for named non-terminals |
3b2f2976 XL |
2 | //! (which it commits to fully when it hits one in a grammar). There's a set of current NFA threads |
3 | //! and a set of next ones. Instead of NTs, we have a special case for Kleene star. The big-O, in | |
4 | //! pathological cases, is worse than traditional use of NFA or Earley parsing, but it's an easier | |
5 | //! fit for Macro-by-Example-style rules. | |
6 | //! | |
7 | //! (In order to prevent the pathological case, we'd need to lazily construct the resulting | |
8 | //! `NamedMatch`es at the very end. It'd be a pain, and require more memory to keep around old | |
5e7ed085 | 9 | //! matcher positions, but it would also save overhead) |
3b2f2976 | 10 | //! |
94b46f34 | 11 | //! We don't say this parser uses the Earley algorithm, because it's unnecessarily inaccurate. |
3b2f2976 XL |
12 | //! The macro parser restricts itself to the features of finite state automata. Earley parsers |
13 | //! can be described as an extension of NFAs with completion rules, prediction rules, and recursion. | |
1a4d82fc JJ |
14 | //! |
15 | //! Quick intro to how the parser works: | |
16 | //! | |
5e7ed085 FG |
17 | //! A "matcher position" (a.k.a. "position" or "mp") is a dot in the middle of a matcher, usually |
18 | //! written as a `·`. For example `· a $( a )* a b` is one, as is `a $( · a )* a b`. | |
1a4d82fc | 19 | //! |
04454e1e | 20 | //! The parser walks through the input a token at a time, maintaining a list |
5e7ed085 | 21 | //! of threads consistent with the current position in the input string: `cur_mps`. |
1a4d82fc | 22 | //! |
5e7ed085 FG |
23 | //! As it processes them, it fills up `eof_mps` with threads that would be valid if |
24 | //! the macro invocation is now over, `bb_mps` with threads that are waiting on | |
25 | //! a Rust non-terminal like `$e:expr`, and `next_mps` with threads that are waiting | |
b039eaaf | 26 | //! on a particular token. Most of the logic concerns moving the · through the |
3b2f2976 XL |
27 | //! repetitions indicated by Kleene stars. The rules for moving the · without |
28 | //! consuming any input are called epsilon transitions. It only advances or calls | |
5e7ed085 | 29 | //! out to the real Rust parser when no `cur_mps` threads remain. |
1a4d82fc | 30 | //! |
7cac9316 | 31 | //! Example: |
1a4d82fc | 32 | //! |
7cac9316 XL |
33 | //! ```text, ignore |
34 | //! Start parsing a a a a b against [· a $( a )* a b]. | |
35 | //! | |
36 | //! Remaining input: a a a a b | |
3b2f2976 | 37 | //! next: [· a $( a )* a b] |
1a4d82fc | 38 | //! |
7cac9316 | 39 | //! - - - Advance over an a. - - - |
1a4d82fc | 40 | //! |
7cac9316 | 41 | //! Remaining input: a a a b |
1a4d82fc | 42 | //! cur: [a · $( a )* a b] |
5e7ed085 | 43 | //! Descend/Skip (first position). |
1a4d82fc JJ |
44 | //! next: [a $( · a )* a b] [a $( a )* · a b]. |
45 | //! | |
7cac9316 | 46 | //! - - - Advance over an a. - - - |
1a4d82fc | 47 | //! |
7cac9316 | 48 | //! Remaining input: a a b |
3b2f2976 | 49 | //! cur: [a $( a · )* a b] [a $( a )* a · b] |
5e7ed085 | 50 | //! Follow epsilon transition: Finish/Repeat (first position) |
1a4d82fc JJ |
51 | //! next: [a $( a )* · a b] [a $( · a )* a b] [a $( a )* a · b] |
52 | //! | |
7cac9316 | 53 | //! - - - Advance over an a. - - - (this looks exactly like the last step) |
1a4d82fc | 54 | //! |
7cac9316 | 55 | //! Remaining input: a b |
3b2f2976 | 56 | //! cur: [a $( a · )* a b] [a $( a )* a · b] |
5e7ed085 | 57 | //! Follow epsilon transition: Finish/Repeat (first position) |
1a4d82fc JJ |
58 | //! next: [a $( a )* · a b] [a $( · a )* a b] [a $( a )* a · b] |
59 | //! | |
7cac9316 | 60 | //! - - - Advance over an a. - - - (this looks exactly like the last step) |
1a4d82fc | 61 | //! |
7cac9316 | 62 | //! Remaining input: b |
3b2f2976 | 63 | //! cur: [a $( a · )* a b] [a $( a )* a · b] |
5e7ed085 | 64 | //! Follow epsilon transition: Finish/Repeat (first position) |
3b2f2976 | 65 | //! next: [a $( a )* · a b] [a $( · a )* a b] [a $( a )* a · b] |
1a4d82fc | 66 | //! |
7cac9316 | 67 | //! - - - Advance over a b. - - - |
1a4d82fc | 68 | //! |
7cac9316 | 69 | //! Remaining input: '' |
1a4d82fc | 70 | //! eof: [a $( a )* a b ·] |
7cac9316 | 71 | //! ``` |
1a4d82fc | 72 | |
923072b8 FG |
73 | pub(crate) use NamedMatch::*; |
74 | pub(crate) use ParseResult::*; | |
9fa01778 | 75 | |
04454e1e | 76 | use crate::mbe::{KleeneOp, TokenTree}; |
e74abb32 | 77 | |
04454e1e FG |
78 | use rustc_ast::token::{self, DocComment, Nonterminal, NonterminalKind, Token}; |
79 | use rustc_lint_defs::pluralize; | |
5e7ed085 | 80 | use rustc_parse::parser::{NtOrTt, Parser}; |
5869c6ff | 81 | use rustc_span::symbol::MacroRulesNormalizedIdent; |
04454e1e | 82 | use rustc_span::Span; |
223e47cc | 83 | |
b7449926 | 84 | use rustc_data_structures::fx::FxHashMap; |
9fa01778 | 85 | use rustc_data_structures::sync::Lrc; |
17df50a5 | 86 | use rustc_span::symbol::Ident; |
74b04a01 | 87 | use std::borrow::Cow; |
b7449926 | 88 | use std::collections::hash_map::Entry::{Occupied, Vacant}; |
223e47cc | 89 | |
04454e1e FG |
90 | /// A unit within a matcher that a `MatcherPos` can refer to. Similar to (and derived from) |
91 | /// `mbe::TokenTree`, but designed specifically for fast and easy traversal during matching. | |
92 | /// Notable differences to `mbe::TokenTree`: | |
93 | /// - It is non-recursive, i.e. there is no nesting. | |
94 | /// - The end pieces of each sequence (the separator, if present, and the Kleene op) are | |
95 | /// represented explicitly, as is the very end of the matcher. | |
96 | /// | |
97 | /// This means a matcher can be represented by `&[MatcherLoc]`, and traversal mostly involves | |
98 | /// simply incrementing the current matcher position index by one. | |
99 | pub(super) enum MatcherLoc { | |
100 | Token { | |
101 | token: Token, | |
102 | }, | |
103 | Delimited, | |
104 | Sequence { | |
105 | op: KleeneOp, | |
106 | num_metavar_decls: usize, | |
107 | idx_first_after: usize, | |
108 | next_metavar: usize, | |
109 | seq_depth: usize, | |
110 | }, | |
111 | SequenceKleeneOpNoSep { | |
112 | op: KleeneOp, | |
113 | idx_first: usize, | |
114 | }, | |
115 | SequenceSep { | |
116 | separator: Token, | |
117 | }, | |
118 | SequenceKleeneOpAfterSep { | |
119 | idx_first: usize, | |
120 | }, | |
121 | MetaVarDecl { | |
122 | span: Span, | |
123 | bind: Ident, | |
124 | kind: Option<NonterminalKind>, | |
125 | next_metavar: usize, | |
126 | seq_depth: usize, | |
127 | }, | |
128 | Eof, | |
129 | } | |
5e7ed085 | 130 | |
04454e1e FG |
131 | pub(super) fn compute_locs(matcher: &[TokenTree]) -> Vec<MatcherLoc> { |
132 | fn inner( | |
133 | tts: &[TokenTree], | |
134 | locs: &mut Vec<MatcherLoc>, | |
135 | next_metavar: &mut usize, | |
136 | seq_depth: usize, | |
137 | ) { | |
138 | for tt in tts { | |
139 | match tt { | |
140 | TokenTree::Token(token) => { | |
141 | locs.push(MatcherLoc::Token { token: token.clone() }); | |
142 | } | |
143 | TokenTree::Delimited(span, delimited) => { | |
144 | let open_token = Token::new(token::OpenDelim(delimited.delim), span.open); | |
145 | let close_token = Token::new(token::CloseDelim(delimited.delim), span.close); | |
146 | ||
147 | locs.push(MatcherLoc::Delimited); | |
148 | locs.push(MatcherLoc::Token { token: open_token }); | |
149 | inner(&delimited.tts, locs, next_metavar, seq_depth); | |
150 | locs.push(MatcherLoc::Token { token: close_token }); | |
151 | } | |
152 | TokenTree::Sequence(_, seq) => { | |
153 | // We can't determine `idx_first_after` and construct the final | |
154 | // `MatcherLoc::Sequence` until after `inner()` is called and the sequence end | |
155 | // pieces are processed. So we push a dummy value (`Eof` is cheapest to | |
156 | // construct) now, and overwrite it with the proper value below. | |
157 | let dummy = MatcherLoc::Eof; | |
158 | locs.push(dummy); | |
159 | ||
160 | let next_metavar_orig = *next_metavar; | |
161 | let op = seq.kleene.op; | |
162 | let idx_first = locs.len(); | |
163 | let idx_seq = idx_first - 1; | |
164 | inner(&seq.tts, locs, next_metavar, seq_depth + 1); | |
165 | ||
166 | if let Some(separator) = &seq.separator { | |
167 | locs.push(MatcherLoc::SequenceSep { separator: separator.clone() }); | |
168 | locs.push(MatcherLoc::SequenceKleeneOpAfterSep { idx_first }); | |
169 | } else { | |
170 | locs.push(MatcherLoc::SequenceKleeneOpNoSep { op, idx_first }); | |
171 | } | |
223e47cc | 172 | |
04454e1e FG |
173 | // Overwrite the dummy value pushed above with the proper value. |
174 | locs[idx_seq] = MatcherLoc::Sequence { | |
175 | op, | |
176 | num_metavar_decls: seq.num_captures, | |
177 | idx_first_after: locs.len(), | |
178 | next_metavar: next_metavar_orig, | |
179 | seq_depth, | |
180 | }; | |
181 | } | |
182 | &TokenTree::MetaVarDecl(span, bind, kind) => { | |
183 | locs.push(MatcherLoc::MetaVarDecl { | |
184 | span, | |
185 | bind, | |
186 | kind, | |
187 | next_metavar: *next_metavar, | |
188 | seq_depth, | |
189 | }); | |
190 | *next_metavar += 1; | |
191 | } | |
192 | TokenTree::MetaVar(..) | TokenTree::MetaVarExpr(..) => unreachable!(), | |
193 | } | |
194 | } | |
195 | } | |
223e47cc | 196 | |
04454e1e FG |
197 | let mut locs = vec![]; |
198 | let mut next_metavar = 0; | |
199 | inner(matcher, &mut locs, &mut next_metavar, /* seq_depth */ 0); | |
223e47cc | 200 | |
04454e1e FG |
201 | // A final entry is needed for eof. |
202 | locs.push(MatcherLoc::Eof); | |
223e47cc | 203 | |
04454e1e | 204 | locs |
5e7ed085 | 205 | } |
a1dfa0c6 | 206 | |
04454e1e FG |
207 | /// A single matcher position, representing the state of matching. |
208 | struct MatcherPos { | |
209 | /// The index into `TtParser::locs`, which represents the "dot". | |
5e7ed085 | 210 | idx: usize, |
2c00a5a8 | 211 | |
04454e1e FG |
212 | /// The matches made against metavar decls so far. On a successful match, this vector ends up |
213 | /// with one element per metavar decl in the matcher. Each element records token trees matched | |
214 | /// against the relevant metavar by the black box parser. An element will be a `MatchedSeq` if | |
215 | /// the corresponding metavar decl is within a sequence. | |
216 | /// | |
217 | /// It is critical to performance that this is an `Lrc`, because it gets cloned frequently when | |
218 | /// processing sequences. Mostly for sequence-ending possibilities that must be tried but end | |
219 | /// up failing. | |
220 | matches: Lrc<Vec<NamedMatch>>, | |
223e47cc LB |
221 | } |
222 | ||
5e7ed085 FG |
223 | // This type is used a lot. Make sure it doesn't unintentionally get bigger. |
224 | #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))] | |
04454e1e | 225 | rustc_data_structures::static_assert_size!(MatcherPos, 16); |
94b46f34 | 226 | |
04454e1e FG |
227 | impl MatcherPos { |
228 | /// Adds `m` as a named match for the `metavar_idx`-th metavar. There are only two call sites, | |
229 | /// and both are hot enough to be always worth inlining. | |
230 | #[inline(always)] | |
231 | fn push_match(&mut self, metavar_idx: usize, seq_depth: usize, m: NamedMatch) { | |
5e7ed085 | 232 | let matches = Lrc::make_mut(&mut self.matches); |
04454e1e | 233 | match seq_depth { |
5e7ed085 FG |
234 | 0 => { |
235 | // We are not within a sequence. Just append `m`. | |
04454e1e | 236 | assert_eq!(metavar_idx, matches.len()); |
5e7ed085 FG |
237 | matches.push(m); |
238 | } | |
239 | _ => { | |
240 | // We are within a sequence. Find the final `MatchedSeq` at the appropriate depth | |
241 | // and append `m` to its vector. | |
04454e1e FG |
242 | let mut curr = &mut matches[metavar_idx]; |
243 | for _ in 0..seq_depth - 1 { | |
5e7ed085 | 244 | match curr { |
04454e1e | 245 | MatchedSeq(seq) => curr = seq.last_mut().unwrap(), |
5e7ed085 FG |
246 | _ => unreachable!(), |
247 | } | |
248 | } | |
249 | match curr { | |
04454e1e | 250 | MatchedSeq(seq) => seq.push(m), |
5e7ed085 FG |
251 | _ => unreachable!(), |
252 | } | |
253 | } | |
94b46f34 XL |
254 | } |
255 | } | |
256 | } | |
257 | ||
04454e1e | 258 | enum EofMatcherPositions { |
5e7ed085 | 259 | None, |
04454e1e | 260 | One(MatcherPos), |
5e7ed085 FG |
261 | Multiple, |
262 | } | |
263 | ||
2c00a5a8 | 264 | /// Represents the possible results of an attempted parse. |
923072b8 | 265 | pub(crate) enum ParseResult<T> { |
2c00a5a8 XL |
266 | /// Parsed successfully. |
267 | Success(T), | |
268 | /// Arm failed to match. If the second parameter is `token::Eof`, it indicates an unexpected | |
269 | /// end of macro invocation. Otherwise, it indicates that no rules expected the given token. | |
dc9dc135 | 270 | Failure(Token, &'static str), |
2c00a5a8 | 271 | /// Fatal error (malformed macro?). Abort compilation. |
dfeec247 | 272 | Error(rustc_span::Span, String), |
ba9703b0 | 273 | ErrorReported, |
2c00a5a8 XL |
274 | } |
275 | ||
ba9703b0 XL |
276 | /// A `ParseResult` where the `Success` variant contains a mapping of |
277 | /// `MacroRulesNormalizedIdent`s to `NamedMatch`es. This represents the mapping | |
278 | /// of metavars to the token trees they bind to. | |
923072b8 | 279 | pub(crate) type NamedParseResult = ParseResult<FxHashMap<MacroRulesNormalizedIdent, NamedMatch>>; |
476ff2be | 280 | |
5e7ed085 FG |
281 | /// Count how many metavars declarations are in `matcher`. |
282 | pub(super) fn count_metavar_decls(matcher: &[TokenTree]) -> usize { | |
283 | matcher | |
284 | .iter() | |
285 | .map(|tt| match tt { | |
286 | TokenTree::MetaVarDecl(..) => 1, | |
287 | TokenTree::Sequence(_, seq) => seq.num_captures, | |
04454e1e | 288 | TokenTree::Delimited(_, delim) => count_metavar_decls(&delim.tts), |
5e7ed085 FG |
289 | TokenTree::Token(..) => 0, |
290 | TokenTree::MetaVar(..) | TokenTree::MetaVarExpr(..) => unreachable!(), | |
291 | }) | |
292 | .sum() | |
223e47cc LB |
293 | } |
294 | ||
5e7ed085 | 295 | /// `NamedMatch` is a pattern-match result for a single metavar. All |
9fa01778 | 296 | /// `MatchedNonterminal`s in the `NamedMatch` have the same non-terminal type |
5e7ed085 | 297 | /// (expr, item, etc). |
1a4d82fc | 298 | /// |
7cac9316 | 299 | /// The in-memory structure of a particular `NamedMatch` represents the match |
1a4d82fc JJ |
300 | /// that occurred when a particular subset of a matcher was applied to a |
301 | /// particular token tree. | |
302 | /// | |
7cac9316 | 303 | /// The width of each `MatchedSeq` in the `NamedMatch`, and the identity of |
5e7ed085 FG |
304 | /// the `MatchedNtNonTts`s, will depend on the token tree it was applied |
305 | /// to: each `MatchedSeq` corresponds to a single repetition in the originating | |
7cac9316 | 306 | /// token tree. The depth of the `NamedMatch` structure will therefore depend |
5e7ed085 FG |
307 | /// only on the nesting depth of repetitions in the originating token tree it |
308 | /// was derived from. | |
309 | /// | |
04454e1e | 310 | /// In layperson's terms: `NamedMatch` will form a tree representing nested matches of a particular |
5e7ed085 FG |
311 | /// meta variable. For example, if we are matching the following macro against the following |
312 | /// invocation... | |
313 | /// | |
314 | /// ```rust | |
315 | /// macro_rules! foo { | |
316 | /// ($($($x:ident),+);+) => {} | |
317 | /// } | |
318 | /// | |
319 | /// foo!(a, b, c, d; a, b, c, d, e); | |
320 | /// ``` | |
321 | /// | |
322 | /// Then, the tree will have the following shape: | |
323 | /// | |
04454e1e FG |
324 | /// ```ignore (private-internal) |
325 | /// # use NamedMatch::*; | |
5e7ed085 FG |
326 | /// MatchedSeq([ |
327 | /// MatchedSeq([ | |
328 | /// MatchedNonterminal(a), | |
329 | /// MatchedNonterminal(b), | |
330 | /// MatchedNonterminal(c), | |
331 | /// MatchedNonterminal(d), | |
332 | /// ]), | |
333 | /// MatchedSeq([ | |
334 | /// MatchedNonterminal(a), | |
335 | /// MatchedNonterminal(b), | |
336 | /// MatchedNonterminal(c), | |
337 | /// MatchedNonterminal(d), | |
338 | /// MatchedNonterminal(e), | |
339 | /// ]) | |
340 | /// ]) | |
341 | /// ``` | |
041b39d2 | 342 | #[derive(Debug, Clone)] |
923072b8 | 343 | pub(crate) enum NamedMatch { |
04454e1e | 344 | MatchedSeq(Vec<NamedMatch>), |
5e7ed085 FG |
345 | |
346 | // A metavar match of type `tt`. | |
347 | MatchedTokenTree(rustc_ast::tokenstream::TokenTree), | |
348 | ||
349 | // A metavar match of any type other than `tt`. | |
9fa01778 | 350 | MatchedNonterminal(Lrc<Nonterminal>), |
223e47cc LB |
351 | } |
352 | ||
9fa01778 | 353 | /// Performs a token equality check, ignoring syntax context (that is, an unhygienic comparison) |
2c00a5a8 | 354 | fn token_name_eq(t1: &Token, t2: &Token) -> bool { |
dc9dc135 XL |
355 | if let (Some((ident1, is_raw1)), Some((ident2, is_raw2))) = (t1.ident(), t2.ident()) { |
356 | ident1.name == ident2.name && is_raw1 == is_raw2 | |
357 | } else if let (Some(ident1), Some(ident2)) = (t1.lifetime(), t2.lifetime()) { | |
358 | ident1.name == ident2.name | |
cc61c64b | 359 | } else { |
dc9dc135 | 360 | t1.kind == t2.kind |
1a4d82fc JJ |
361 | } |
362 | } | |
363 | ||
04454e1e | 364 | // Note: the vectors could be created and dropped within `parse_tt`, but to avoid excess |
923072b8 | 365 | // allocations we have a single vector for each kind that is cleared and reused repeatedly. |
04454e1e | 366 | pub struct TtParser { |
5e7ed085 | 367 | macro_name: Ident, |
223e47cc | 368 | |
5e7ed085 FG |
369 | /// The set of current mps to be processed. This should be empty by the end of a successful |
370 | /// execution of `parse_tt_inner`. | |
04454e1e | 371 | cur_mps: Vec<MatcherPos>, |
476ff2be | 372 | |
5e7ed085 FG |
373 | /// The set of newly generated mps. These are used to replenish `cur_mps` in the function |
374 | /// `parse_tt`. | |
04454e1e | 375 | next_mps: Vec<MatcherPos>, |
223e47cc | 376 | |
5e7ed085 | 377 | /// The set of mps that are waiting for the black-box parser. |
04454e1e | 378 | bb_mps: Vec<MatcherPos>, |
223e47cc | 379 | |
5e7ed085 FG |
380 | /// Pre-allocate an empty match array, so it can be cloned cheaply for macros with many rules |
381 | /// that have no metavars. | |
04454e1e | 382 | empty_matches: Lrc<Vec<NamedMatch>>, |
5e7ed085 FG |
383 | } |
384 | ||
04454e1e FG |
385 | impl TtParser { |
386 | pub(super) fn new(macro_name: Ident) -> TtParser { | |
5e7ed085 FG |
387 | TtParser { |
388 | macro_name, | |
389 | cur_mps: vec![], | |
390 | next_mps: vec![], | |
391 | bb_mps: vec![], | |
04454e1e | 392 | empty_matches: Lrc::new(vec![]), |
2c00a5a8 | 393 | } |
5e7ed085 FG |
394 | } |
395 | ||
396 | /// Process the matcher positions of `cur_mps` until it is empty. In the process, this will | |
397 | /// produce more mps in `next_mps` and `bb_mps`. | |
398 | /// | |
399 | /// # Returns | |
400 | /// | |
401 | /// `Some(result)` if everything is finished, `None` otherwise. Note that matches are kept | |
402 | /// track of through the mps generated. | |
403 | fn parse_tt_inner( | |
404 | &mut self, | |
04454e1e | 405 | matcher: &[MatcherLoc], |
5e7ed085 FG |
406 | token: &Token, |
407 | ) -> Option<NamedParseResult> { | |
408 | // Matcher positions that would be valid if the macro invocation was over now. Only | |
409 | // modified if `token == Eof`. | |
410 | let mut eof_mps = EofMatcherPositions::None; | |
411 | ||
412 | while let Some(mut mp) = self.cur_mps.pop() { | |
04454e1e FG |
413 | match &matcher[mp.idx] { |
414 | MatcherLoc::Token { token: t } => { | |
415 | // If it's a doc comment, we just ignore it and move on to the next tt in the | |
416 | // matcher. This is a bug, but #95267 showed that existing programs rely on | |
417 | // this behaviour, and changing it would require some care and a transition | |
418 | // period. | |
419 | // | |
420 | // If the token matches, we can just advance the parser. | |
421 | // | |
422 | // Otherwise, this match has failed, there is nothing to do, and hopefully | |
423 | // another mp in `cur_mps` will match. | |
424 | if matches!(t, Token { kind: DocComment(..), .. }) { | |
425 | mp.idx += 1; | |
426 | self.cur_mps.push(mp); | |
427 | } else if token_name_eq(&t, token) { | |
428 | mp.idx += 1; | |
429 | self.next_mps.push(mp); | |
5e7ed085 | 430 | } |
04454e1e FG |
431 | } |
432 | MatcherLoc::Delimited => { | |
f2b60f7d | 433 | // Entering the delimiter is trivial. |
04454e1e FG |
434 | mp.idx += 1; |
435 | self.cur_mps.push(mp); | |
436 | } | |
437 | &MatcherLoc::Sequence { | |
438 | op, | |
439 | num_metavar_decls, | |
440 | idx_first_after, | |
441 | next_metavar, | |
442 | seq_depth, | |
443 | } => { | |
444 | // Install an empty vec for each metavar within the sequence. | |
445 | for metavar_idx in next_metavar..next_metavar + num_metavar_decls { | |
446 | mp.push_match(metavar_idx, seq_depth, MatchedSeq(vec![])); | |
b9856134 | 447 | } |
b9856134 | 448 | |
04454e1e FG |
449 | if op == KleeneOp::ZeroOrMore || op == KleeneOp::ZeroOrOne { |
450 | // Try zero matches of this sequence, by skipping over it. | |
451 | self.cur_mps.push(MatcherPos { | |
452 | idx: idx_first_after, | |
453 | matches: mp.matches.clone(), // a cheap clone | |
5e7ed085 | 454 | }); |
5e7ed085 FG |
455 | } |
456 | ||
04454e1e FG |
457 | // Try one or more matches of this sequence, by entering it. |
458 | mp.idx += 1; | |
459 | self.cur_mps.push(mp); | |
476ff2be | 460 | } |
04454e1e FG |
461 | &MatcherLoc::SequenceKleeneOpNoSep { op, idx_first } => { |
462 | // We are past the end of a sequence with no separator. Try ending the | |
463 | // sequence. If that's not possible, `ending_mp` will fail quietly when it is | |
5e7ed085 | 464 | // processed next time around the loop. |
04454e1e FG |
465 | let ending_mp = MatcherPos { |
466 | idx: mp.idx + 1, // +1 skips the Kleene op | |
5e7ed085 | 467 | matches: mp.matches.clone(), // a cheap clone |
5e7ed085 | 468 | }; |
04454e1e FG |
469 | self.cur_mps.push(ending_mp); |
470 | ||
471 | if op != KleeneOp::ZeroOrOne { | |
472 | // Try another repetition. | |
473 | mp.idx = idx_first; | |
474 | self.cur_mps.push(mp); | |
475 | } | |
223e47cc | 476 | } |
04454e1e FG |
477 | MatcherLoc::SequenceSep { separator } => { |
478 | // We are past the end of a sequence with a separator but we haven't seen the | |
479 | // separator yet. Try ending the sequence. If that's not possible, `ending_mp` | |
480 | // will fail quietly when it is processed next time around the loop. | |
481 | let ending_mp = MatcherPos { | |
482 | idx: mp.idx + 2, // +2 skips the separator and the Kleene op | |
483 | matches: mp.matches.clone(), // a cheap clone | |
484 | }; | |
485 | self.cur_mps.push(ending_mp); | |
2c00a5a8 | 486 | |
04454e1e FG |
487 | if token_name_eq(token, separator) { |
488 | // The separator matches the current token. Advance past it. | |
5e7ed085 FG |
489 | mp.idx += 1; |
490 | self.next_mps.push(mp); | |
491 | } | |
04454e1e FG |
492 | } |
493 | &MatcherLoc::SequenceKleeneOpAfterSep { idx_first } => { | |
494 | // We are past the sequence separator. This can't be a `?` Kleene op, because | |
495 | // they don't permit separators. Try another repetition. | |
496 | mp.idx = idx_first; | |
5e7ed085 FG |
497 | self.cur_mps.push(mp); |
498 | } | |
04454e1e FG |
499 | &MatcherLoc::MetaVarDecl { span, kind, .. } => { |
500 | // Built-in nonterminals never start with these tokens, so we can eliminate | |
501 | // them from consideration. We use the span of the metavariable declaration | |
502 | // to determine any edition-specific matching behavior for non-terminals. | |
503 | if let Some(kind) = kind { | |
504 | if Parser::nonterminal_may_begin_with(kind, token) { | |
505 | self.bb_mps.push(mp); | |
506 | } | |
507 | } else { | |
508 | // E.g. `$e` instead of `$e:expr`, reported as a hard error if actually used. | |
509 | // Both this check and the one in `nameize` are necessary, surprisingly. | |
510 | return Some(Error(span, "missing fragment specifier".to_string())); | |
511 | } | |
512 | } | |
513 | MatcherLoc::Eof => { | |
514 | // We are past the matcher's end, and not in a sequence. Try to end things. | |
515 | debug_assert_eq!(mp.idx, matcher.len() - 1); | |
516 | if *token == token::Eof { | |
517 | eof_mps = match eof_mps { | |
518 | EofMatcherPositions::None => EofMatcherPositions::One(mp), | |
519 | EofMatcherPositions::One(_) | EofMatcherPositions::Multiple => { | |
520 | EofMatcherPositions::Multiple | |
521 | } | |
5e7ed085 FG |
522 | } |
523 | } | |
524 | } | |
223e47cc LB |
525 | } |
526 | } | |
476ff2be | 527 | |
5e7ed085 FG |
528 | // If we reached the end of input, check that there is EXACTLY ONE possible matcher. |
529 | // Otherwise, either the parse is ambiguous (which is an error) or there is a syntax error. | |
530 | if *token == token::Eof { | |
531 | Some(match eof_mps { | |
532 | EofMatcherPositions::One(mut eof_mp) => { | |
5e7ed085 FG |
533 | // Need to take ownership of the matches from within the `Lrc`. |
534 | Lrc::make_mut(&mut eof_mp.matches); | |
535 | let matches = Lrc::try_unwrap(eof_mp.matches).unwrap().into_iter(); | |
04454e1e | 536 | self.nameize(matcher, matches) |
5e7ed085 FG |
537 | } |
538 | EofMatcherPositions::Multiple => { | |
539 | Error(token.span, "ambiguity: multiple successful parses".to_string()) | |
540 | } | |
541 | EofMatcherPositions::None => Failure( | |
dfeec247 XL |
542 | Token::new( |
543 | token::Eof, | |
5e7ed085 | 544 | if token.span.is_dummy() { token.span } else { token.span.shrink_to_hi() }, |
dfeec247 | 545 | ), |
0731742a | 546 | "missing tokens in macro arguments", |
2c00a5a8 | 547 | ), |
5e7ed085 FG |
548 | }) |
549 | } else { | |
550 | None | |
2c00a5a8 | 551 | } |
5e7ed085 FG |
552 | } |
553 | ||
554 | /// Match the token stream from `parser` against `matcher`. | |
555 | pub(super) fn parse_tt( | |
556 | &mut self, | |
557 | parser: &mut Cow<'_, Parser<'_>>, | |
04454e1e | 558 | matcher: &[MatcherLoc], |
5e7ed085 FG |
559 | ) -> NamedParseResult { |
560 | // A queue of possible matcher positions. We initialize it with the matcher position in | |
561 | // which the "dot" is before the first token of the first token tree in `matcher`. | |
562 | // `parse_tt_inner` then processes all of these possible matcher positions and produces | |
563 | // possible next positions into `next_mps`. After some post-processing, the contents of | |
564 | // `next_mps` replenish `cur_mps` and we start over again. | |
565 | self.cur_mps.clear(); | |
04454e1e | 566 | self.cur_mps.push(MatcherPos { idx: 0, matches: self.empty_matches.clone() }); |
5e7ed085 FG |
567 | |
568 | loop { | |
569 | self.next_mps.clear(); | |
570 | self.bb_mps.clear(); | |
571 | ||
572 | // Process `cur_mps` until either we have finished the input or we need to get some | |
573 | // parsing from the black-box parser done. | |
04454e1e FG |
574 | if let Some(res) = self.parse_tt_inner(matcher, &parser.token) { |
575 | return res; | |
5e7ed085 FG |
576 | } |
577 | ||
578 | // `parse_tt_inner` handled all of `cur_mps`, so it's empty. | |
579 | assert!(self.cur_mps.is_empty()); | |
580 | ||
581 | // Error messages here could be improved with links to original rules. | |
582 | match (self.next_mps.len(), self.bb_mps.len()) { | |
583 | (0, 0) => { | |
584 | // There are no possible next positions AND we aren't waiting for the black-box | |
585 | // parser: syntax error. | |
586 | return Failure( | |
587 | parser.token.clone(), | |
588 | "no rules expected this token in macro call", | |
589 | ); | |
590 | } | |
591 | ||
592 | (_, 0) => { | |
593 | // Dump all possible `next_mps` into `cur_mps` for the next iteration. Then | |
594 | // process the next token. | |
04454e1e | 595 | self.cur_mps.append(&mut self.next_mps); |
5e7ed085 FG |
596 | parser.to_mut().bump(); |
597 | } | |
598 | ||
599 | (0, 1) => { | |
600 | // We need to call the black-box parser to get some nonterminal. | |
601 | let mut mp = self.bb_mps.pop().unwrap(); | |
04454e1e FG |
602 | let loc = &matcher[mp.idx]; |
603 | if let &MatcherLoc::MetaVarDecl { | |
604 | span, | |
605 | kind: Some(kind), | |
606 | next_metavar, | |
607 | seq_depth, | |
608 | .. | |
609 | } = loc | |
610 | { | |
5e7ed085 FG |
611 | // We use the span of the metavariable declaration to determine any |
612 | // edition-specific matching behavior for non-terminals. | |
613 | let nt = match parser.to_mut().parse_nonterminal(kind) { | |
614 | Err(mut err) => { | |
615 | err.span_label( | |
616 | span, | |
617 | format!( | |
618 | "while parsing argument for this `{kind}` macro fragment" | |
619 | ), | |
620 | ) | |
621 | .emit(); | |
622 | return ErrorReported; | |
623 | } | |
624 | Ok(nt) => nt, | |
625 | }; | |
626 | let m = match nt { | |
627 | NtOrTt::Nt(nt) => MatchedNonterminal(Lrc::new(nt)), | |
628 | NtOrTt::Tt(tt) => MatchedTokenTree(tt), | |
629 | }; | |
04454e1e | 630 | mp.push_match(next_metavar, seq_depth, m); |
5e7ed085 | 631 | mp.idx += 1; |
5e7ed085 FG |
632 | } else { |
633 | unreachable!() | |
3dfed10e | 634 | } |
5e7ed085 FG |
635 | self.cur_mps.push(mp); |
636 | } | |
637 | ||
638 | (_, _) => { | |
639 | // Too many possibilities! | |
04454e1e | 640 | return self.ambiguity_error(matcher, parser.token.span); |
5e7ed085 | 641 | } |
223e47cc | 642 | } |
5e7ed085 FG |
643 | |
644 | assert!(!self.cur_mps.is_empty()); | |
223e47cc | 645 | } |
5e7ed085 | 646 | } |
223e47cc | 647 | |
04454e1e FG |
648 | fn ambiguity_error( |
649 | &self, | |
650 | matcher: &[MatcherLoc], | |
651 | token_span: rustc_span::Span, | |
652 | ) -> NamedParseResult { | |
5e7ed085 FG |
653 | let nts = self |
654 | .bb_mps | |
655 | .iter() | |
04454e1e FG |
656 | .map(|mp| match &matcher[mp.idx] { |
657 | MatcherLoc::MetaVarDecl { bind, kind: Some(kind), .. } => { | |
5e7ed085 FG |
658 | format!("{} ('{}')", kind, bind) |
659 | } | |
04454e1e | 660 | _ => unreachable!(), |
5e7ed085 FG |
661 | }) |
662 | .collect::<Vec<String>>() | |
663 | .join(" or "); | |
664 | ||
665 | Error( | |
666 | token_span, | |
667 | format!( | |
668 | "local ambiguity when calling macro `{}`: multiple parsing options: {}", | |
669 | self.macro_name, | |
670 | match self.next_mps.len() { | |
671 | 0 => format!("built-in NTs {}.", nts), | |
04454e1e | 672 | n => format!("built-in NTs {} or {n} other option{s}.", nts, s = pluralize!(n)), |
5e7ed085 FG |
673 | } |
674 | ), | |
675 | ) | |
223e47cc | 676 | } |
04454e1e FG |
677 | |
678 | fn nameize<I: Iterator<Item = NamedMatch>>( | |
679 | &self, | |
680 | matcher: &[MatcherLoc], | |
681 | mut res: I, | |
682 | ) -> NamedParseResult { | |
683 | // Make that each metavar has _exactly one_ binding. If so, insert the binding into the | |
684 | // `NamedParseResult`. Otherwise, it's an error. | |
685 | let mut ret_val = FxHashMap::default(); | |
686 | for loc in matcher { | |
687 | if let &MatcherLoc::MetaVarDecl { span, bind, kind, .. } = loc { | |
688 | if kind.is_some() { | |
689 | match ret_val.entry(MacroRulesNormalizedIdent::new(bind)) { | |
690 | Vacant(spot) => spot.insert(res.next().unwrap()), | |
691 | Occupied(..) => { | |
692 | return Error(span, format!("duplicated bind name: {}", bind)); | |
693 | } | |
694 | }; | |
695 | } else { | |
696 | // E.g. `$e` instead of `$e:expr`, reported as a hard error if actually used. | |
697 | // Both this check and the one in `parse_tt_inner` are necessary, surprisingly. | |
698 | return Error(span, "missing fragment specifier".to_string()); | |
699 | } | |
700 | } | |
701 | } | |
702 | Success(ret_val) | |
703 | } | |
223e47cc | 704 | } |