]>
Commit | Line | Data |
---|---|---|
1a4d82fc | 1 | // Copyright 2015 The Rust Project Developers. See the COPYRIGHT |
223e47cc LB |
2 | // file at the top-level directory of this distribution and at |
3 | // http://rust-lang.org/COPYRIGHT. | |
4 | // | |
5 | // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or | |
6 | // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license | |
7 | // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your | |
8 | // option. This file may not be copied, modified, or distributed | |
9 | // except according to those terms. | |
10 | ||
85aaf69f | 11 | use ast::{self, TokenTree, TtDelimited, TtSequence, TtToken}; |
1a4d82fc JJ |
12 | use codemap::{Span, DUMMY_SP}; |
13 | use ext::base::{ExtCtxt, MacResult, SyntaxExtension}; | |
14 | use ext::base::{NormalTT, TTMacroExpander}; | |
15 | use ext::tt::macro_parser::{Success, Error, Failure}; | |
16 | use ext::tt::macro_parser::{NamedMatch, MatchedSeq, MatchedNonterminal}; | |
17 | use ext::tt::macro_parser::{parse, parse_or_else}; | |
c34b1796 | 18 | use parse::lexer::new_tt_reader; |
223e47cc | 19 | use parse::parser::Parser; |
85aaf69f | 20 | use parse::token::{self, special_idents, gensym_ident, NtTT, Token}; |
1a4d82fc | 21 | use parse::token::Token::*; |
223e47cc | 22 | use print; |
1a4d82fc | 23 | use ptr::P; |
223e47cc | 24 | |
1a4d82fc JJ |
25 | use util::small_vector::SmallVector; |
26 | ||
27 | use std::cell::RefCell; | |
28 | use std::rc::Rc; | |
29 | ||
30 | struct ParserAnyMacro<'a> { | |
31 | parser: RefCell<Parser<'a>>, | |
9346a6ac AL |
32 | |
33 | /// Span of the expansion site of the macro this parser is for | |
34 | site_span: Span, | |
35 | /// The ident of the macro we're parsing | |
36 | macro_ident: ast::Ident | |
1a4d82fc JJ |
37 | } |
38 | ||
39 | impl<'a> ParserAnyMacro<'a> { | |
40 | /// Make sure we don't have any tokens left to parse, so we don't | |
41 | /// silently drop anything. `allow_semi` is so that "optional" | |
42 | /// semicolons at the end of normal expressions aren't complained | |
43 | /// about e.g. the semicolon in `macro_rules! kapow { () => { | |
44 | /// panic!(); } }` doesn't get picked up by .parse_expr(), but it's | |
45 | /// allowed to be there. | |
46 | fn ensure_complete_parse(&self, allow_semi: bool) { | |
47 | let mut parser = self.parser.borrow_mut(); | |
48 | if allow_semi && parser.token == token::Semi { | |
9346a6ac | 49 | panictry!(parser.bump()) |
1a4d82fc JJ |
50 | } |
51 | if parser.token != token::Eof { | |
52 | let token_str = parser.this_token_to_string(); | |
53 | let msg = format!("macro expansion ignores token `{}` and any \ | |
54 | following", | |
55 | token_str); | |
56 | let span = parser.span; | |
85aaf69f | 57 | parser.span_err(span, &msg[..]); |
9346a6ac AL |
58 | |
59 | let name = token::get_ident(self.macro_ident); | |
60 | let msg = format!("caused by the macro expansion here; the usage \ | |
61 | of `{}` is likely invalid in this context", | |
62 | name); | |
63 | parser.span_note(self.site_span, &msg[..]); | |
1a4d82fc JJ |
64 | } |
65 | } | |
66 | } | |
67 | ||
68 | impl<'a> MacResult for ParserAnyMacro<'a> { | |
69 | fn make_expr(self: Box<ParserAnyMacro<'a>>) -> Option<P<ast::Expr>> { | |
70 | let ret = self.parser.borrow_mut().parse_expr(); | |
71 | self.ensure_complete_parse(true); | |
72 | Some(ret) | |
73 | } | |
74 | fn make_pat(self: Box<ParserAnyMacro<'a>>) -> Option<P<ast::Pat>> { | |
75 | let ret = self.parser.borrow_mut().parse_pat(); | |
76 | self.ensure_complete_parse(false); | |
77 | Some(ret) | |
78 | } | |
79 | fn make_items(self: Box<ParserAnyMacro<'a>>) -> Option<SmallVector<P<ast::Item>>> { | |
80 | let mut ret = SmallVector::zero(); | |
c34b1796 AL |
81 | while let Some(item) = self.parser.borrow_mut().parse_item() { |
82 | ret.push(item); | |
1a4d82fc JJ |
83 | } |
84 | self.ensure_complete_parse(false); | |
85 | Some(ret) | |
86 | } | |
87 | ||
c34b1796 AL |
88 | fn make_impl_items(self: Box<ParserAnyMacro<'a>>) |
89 | -> Option<SmallVector<P<ast::ImplItem>>> { | |
1a4d82fc JJ |
90 | let mut ret = SmallVector::zero(); |
91 | loop { | |
92 | let mut parser = self.parser.borrow_mut(); | |
93 | match parser.token { | |
94 | token::Eof => break, | |
9346a6ac | 95 | _ => ret.push(panictry!(parser.parse_impl_item())) |
1a4d82fc JJ |
96 | } |
97 | } | |
98 | self.ensure_complete_parse(false); | |
99 | Some(ret) | |
100 | } | |
101 | ||
9346a6ac AL |
102 | fn make_stmts(self: Box<ParserAnyMacro<'a>>) |
103 | -> Option<SmallVector<P<ast::Stmt>>> { | |
104 | let mut ret = SmallVector::zero(); | |
105 | loop { | |
106 | let mut parser = self.parser.borrow_mut(); | |
107 | match parser.token { | |
108 | token::Eof => break, | |
109 | _ => match parser.parse_stmt_nopanic() { | |
110 | Ok(maybe_stmt) => match maybe_stmt { | |
111 | Some(stmt) => ret.push(stmt), | |
112 | None => (), | |
113 | }, | |
114 | Err(_) => break, | |
115 | } | |
116 | } | |
117 | } | |
118 | self.ensure_complete_parse(false); | |
119 | Some(ret) | |
223e47cc | 120 | } |
1a4d82fc JJ |
121 | } |
122 | ||
123 | struct MacroRulesMacroExpander { | |
124 | name: ast::Ident, | |
125 | imported_from: Option<ast::Ident>, | |
126 | lhses: Vec<Rc<NamedMatch>>, | |
127 | rhses: Vec<Rc<NamedMatch>>, | |
128 | } | |
129 | ||
130 | impl TTMacroExpander for MacroRulesMacroExpander { | |
131 | fn expand<'cx>(&self, | |
132 | cx: &'cx mut ExtCtxt, | |
133 | sp: Span, | |
134 | arg: &[ast::TokenTree]) | |
135 | -> Box<MacResult+'cx> { | |
136 | generic_extension(cx, | |
137 | sp, | |
138 | self.name, | |
139 | self.imported_from, | |
140 | arg, | |
85aaf69f SL |
141 | &self.lhses, |
142 | &self.rhses) | |
1a4d82fc JJ |
143 | } |
144 | } | |
145 | ||
146 | /// Given `lhses` and `rhses`, this is the new macro we create | |
147 | fn generic_extension<'cx>(cx: &'cx ExtCtxt, | |
148 | sp: Span, | |
149 | name: ast::Ident, | |
150 | imported_from: Option<ast::Ident>, | |
151 | arg: &[ast::TokenTree], | |
152 | lhses: &[Rc<NamedMatch>], | |
153 | rhses: &[Rc<NamedMatch>]) | |
154 | -> Box<MacResult+'cx> { | |
155 | if cx.trace_macros() { | |
156 | println!("{}! {{ {} }}", | |
157 | token::get_ident(name), | |
158 | print::pprust::tts_to_string(arg)); | |
159 | } | |
160 | ||
161 | // Which arm's failure should we report? (the one furthest along) | |
162 | let mut best_fail_spot = DUMMY_SP; | |
163 | let mut best_fail_msg = "internal error: ran no matchers".to_string(); | |
164 | ||
165 | for (i, lhs) in lhses.iter().enumerate() { // try each arm's matchers | |
166 | match **lhs { | |
167 | MatchedNonterminal(NtTT(ref lhs_tt)) => { | |
168 | let lhs_tt = match **lhs_tt { | |
85aaf69f | 169 | TtDelimited(_, ref delim) => &delim.tts[..], |
9346a6ac | 170 | _ => panic!(cx.span_fatal(sp, "malformed macro lhs")) |
1a4d82fc | 171 | }; |
c34b1796 AL |
172 | |
173 | match TokenTree::parse(cx, lhs_tt, arg) { | |
1a4d82fc JJ |
174 | Success(named_matches) => { |
175 | let rhs = match *rhses[i] { | |
176 | // okay, what's your transcriber? | |
177 | MatchedNonterminal(NtTT(ref tt)) => { | |
178 | match **tt { | |
179 | // ignore delimiters | |
180 | TtDelimited(_, ref delimed) => delimed.tts.clone(), | |
9346a6ac | 181 | _ => panic!(cx.span_fatal(sp, "macro rhs must be delimited")), |
1a4d82fc JJ |
182 | } |
183 | }, | |
184 | _ => cx.span_bug(sp, "bad thing in rhs") | |
185 | }; | |
186 | // rhs has holes ( `$id` and `$(...)` that need filled) | |
187 | let trncbr = new_tt_reader(&cx.parse_sess().span_diagnostic, | |
188 | Some(named_matches), | |
189 | imported_from, | |
190 | rhs); | |
c34b1796 | 191 | let mut p = Parser::new(cx.parse_sess(), cx.cfg(), Box::new(trncbr)); |
9346a6ac | 192 | panictry!(p.check_unknown_macro_variable()); |
1a4d82fc JJ |
193 | // Let the context choose how to interpret the result. |
194 | // Weird, but useful for X-macros. | |
d9579d0f | 195 | return Box::new(ParserAnyMacro { |
1a4d82fc | 196 | parser: RefCell::new(p), |
9346a6ac AL |
197 | |
198 | // Pass along the original expansion site and the name of the macro | |
199 | // so we can print a useful error message if the parse of the expanded | |
200 | // macro leaves unparsed tokens. | |
201 | site_span: sp, | |
202 | macro_ident: name | |
d9579d0f | 203 | }) |
1a4d82fc JJ |
204 | } |
205 | Failure(sp, ref msg) => if sp.lo >= best_fail_spot.lo { | |
206 | best_fail_spot = sp; | |
207 | best_fail_msg = (*msg).clone(); | |
208 | }, | |
9346a6ac | 209 | Error(sp, ref msg) => panic!(cx.span_fatal(sp, &msg[..])) |
1a4d82fc JJ |
210 | } |
211 | } | |
212 | _ => cx.bug("non-matcher found in parsed lhses") | |
213 | } | |
214 | } | |
9346a6ac | 215 | panic!(cx.span_fatal(best_fail_spot, &best_fail_msg[..])); |
1a4d82fc JJ |
216 | } |
217 | ||
218 | // Note that macro-by-example's input is also matched against a token tree: | |
219 | // $( $lhs:tt => $rhs:tt );+ | |
220 | // | |
221 | // Holy self-referential! | |
222 | ||
223 | /// Converts a `macro_rules!` invocation into a syntax extension. | |
224 | pub fn compile<'cx>(cx: &'cx mut ExtCtxt, | |
225 | def: &ast::MacroDef) -> SyntaxExtension { | |
223e47cc | 226 | |
970d7e83 LB |
227 | let lhs_nm = gensym_ident("lhs"); |
228 | let rhs_nm = gensym_ident("rhs"); | |
223e47cc | 229 | |
1a4d82fc | 230 | // The pattern that macro_rules matches. |
223e47cc | 231 | // The grammar for macro_rules! is: |
1a4d82fc | 232 | // $( $lhs:tt => $rhs:tt );+ |
223e47cc | 233 | // ...quasiquoting this would be nice. |
1a4d82fc JJ |
234 | // These spans won't matter, anyways |
235 | let match_lhs_tok = MatchNt(lhs_nm, special_idents::tt, token::Plain, token::Plain); | |
236 | let match_rhs_tok = MatchNt(rhs_nm, special_idents::tt, token::Plain, token::Plain); | |
237 | let argument_gram = vec!( | |
238 | TtSequence(DUMMY_SP, | |
239 | Rc::new(ast::SequenceRepetition { | |
240 | tts: vec![ | |
241 | TtToken(DUMMY_SP, match_lhs_tok), | |
242 | TtToken(DUMMY_SP, token::FatArrow), | |
243 | TtToken(DUMMY_SP, match_rhs_tok)], | |
244 | separator: Some(token::Semi), | |
245 | op: ast::OneOrMore, | |
246 | num_captures: 2 | |
247 | })), | |
223e47cc LB |
248 | //to phase into semicolon-termination instead of |
249 | //semicolon-separation | |
1a4d82fc JJ |
250 | TtSequence(DUMMY_SP, |
251 | Rc::new(ast::SequenceRepetition { | |
252 | tts: vec![TtToken(DUMMY_SP, token::Semi)], | |
253 | separator: None, | |
254 | op: ast::ZeroOrMore, | |
255 | num_captures: 0 | |
256 | }))); | |
223e47cc LB |
257 | |
258 | ||
259 | // Parse the macro_rules! invocation (`none` is for no interpolations): | |
1a4d82fc JJ |
260 | let arg_reader = new_tt_reader(&cx.parse_sess().span_diagnostic, |
261 | None, | |
262 | None, | |
263 | def.body.clone()); | |
223e47cc LB |
264 | let argument_map = parse_or_else(cx.parse_sess(), |
265 | cx.cfg(), | |
1a4d82fc | 266 | arg_reader, |
223e47cc LB |
267 | argument_gram); |
268 | ||
269 | // Extract the arguments: | |
c34b1796 | 270 | let lhses = match **argument_map.get(&lhs_nm).unwrap() { |
1a4d82fc JJ |
271 | MatchedSeq(ref s, _) => /* FIXME (#2543) */ (*s).clone(), |
272 | _ => cx.span_bug(def.span, "wrong-structured lhs") | |
223e47cc LB |
273 | }; |
274 | ||
85aaf69f | 275 | for lhs in &lhses { |
1a4d82fc JJ |
276 | check_lhs_nt_follows(cx, &**lhs, def.span); |
277 | } | |
278 | ||
c34b1796 | 279 | let rhses = match **argument_map.get(&rhs_nm).unwrap() { |
1a4d82fc JJ |
280 | MatchedSeq(ref s, _) => /* FIXME (#2543) */ (*s).clone(), |
281 | _ => cx.span_bug(def.span, "wrong-structured rhs") | |
223e47cc LB |
282 | }; |
283 | ||
d9579d0f | 284 | let exp: Box<_> = Box::new(MacroRulesMacroExpander { |
1a4d82fc JJ |
285 | name: def.ident, |
286 | imported_from: def.imported_from, | |
287 | lhses: lhses, | |
288 | rhses: rhses, | |
d9579d0f | 289 | }); |
223e47cc | 290 | |
c34b1796 | 291 | NormalTT(exp, Some(def.span), def.allow_internal_unstable) |
1a4d82fc JJ |
292 | } |
293 | ||
294 | fn check_lhs_nt_follows(cx: &mut ExtCtxt, lhs: &NamedMatch, sp: Span) { | |
85aaf69f SL |
295 | // lhs is going to be like MatchedNonterminal(NtTT(TtDelimited(...))), where the entire lhs is |
296 | // those tts. Or, it can be a "bare sequence", not wrapped in parens. | |
297 | match lhs { | |
1a4d82fc | 298 | &MatchedNonterminal(NtTT(ref inner)) => match &**inner { |
85aaf69f SL |
299 | &TtDelimited(_, ref tts) => { |
300 | check_matcher(cx, tts.tts.iter(), &Eof); | |
301 | }, | |
302 | tt @ &TtSequence(..) => { | |
303 | check_matcher(cx, Some(tt).into_iter(), &Eof); | |
304 | }, | |
305 | _ => cx.span_bug(sp, "wrong-structured lhs for follow check (didn't find \ | |
306 | a TtDelimited or TtSequence)") | |
1a4d82fc | 307 | }, |
85aaf69f SL |
308 | _ => cx.span_bug(sp, "wrong-structured lhs for follow check (didn't find a \ |
309 | MatchedNonterminal)") | |
1a4d82fc | 310 | }; |
1a4d82fc JJ |
311 | // we don't abort on errors on rejection, the driver will do that for us |
312 | // after parsing/expansion. we can report every error in every macro this way. | |
313 | } | |
314 | ||
315 | // returns the last token that was checked, for TtSequence. this gets used later on. | |
316 | fn check_matcher<'a, I>(cx: &mut ExtCtxt, matcher: I, follow: &Token) | |
317 | -> Option<(Span, Token)> where I: Iterator<Item=&'a TokenTree> { | |
318 | use print::pprust::token_to_string; | |
319 | ||
320 | let mut last = None; | |
321 | ||
322 | // 2. For each token T in M: | |
323 | let mut tokens = matcher.peekable(); | |
324 | while let Some(token) = tokens.next() { | |
325 | last = match *token { | |
326 | TtToken(sp, MatchNt(ref name, ref frag_spec, _, _)) => { | |
327 | // ii. If T is a simple NT, look ahead to the next token T' in | |
328 | // M. | |
329 | let next_token = match tokens.peek() { | |
330 | // If T' closes a complex NT, replace T' with F | |
331 | Some(&&TtToken(_, CloseDelim(_))) => follow.clone(), | |
332 | Some(&&TtToken(_, ref tok)) => tok.clone(), | |
333 | Some(&&TtSequence(sp, _)) => { | |
334 | cx.span_err(sp, | |
85aaf69f SL |
335 | &format!("`${0}:{1}` is followed by a \ |
336 | sequence repetition, which is not \ | |
337 | allowed for `{1}` fragments", | |
338 | name.as_str(), frag_spec.as_str()) | |
339 | ); | |
1a4d82fc JJ |
340 | Eof |
341 | }, | |
342 | // die next iteration | |
343 | Some(&&TtDelimited(_, ref delim)) => delim.close_token(), | |
344 | // else, we're at the end of the macro or sequence | |
345 | None => follow.clone() | |
346 | }; | |
347 | ||
348 | let tok = if let TtToken(_, ref tok) = *token { tok } else { unreachable!() }; | |
349 | // If T' is in the set FOLLOW(NT), continue. Else, reject. | |
85aaf69f | 350 | match (&next_token, is_in_follow(cx, &next_token, frag_spec.as_str())) { |
c34b1796 AL |
351 | (_, Err(msg)) => { |
352 | cx.span_err(sp, &msg); | |
353 | continue | |
354 | } | |
85aaf69f SL |
355 | (&Eof, _) => return Some((sp, tok.clone())), |
356 | (_, Ok(true)) => continue, | |
357 | (next, Ok(false)) => { | |
358 | cx.span_err(sp, &format!("`${0}:{1}` is followed by `{2}`, which \ | |
359 | is not allowed for `{1}` fragments", | |
1a4d82fc | 360 | name.as_str(), frag_spec.as_str(), |
85aaf69f | 361 | token_to_string(next))); |
1a4d82fc JJ |
362 | continue |
363 | }, | |
223e47cc | 364 | } |
1a4d82fc JJ |
365 | }, |
366 | TtSequence(sp, ref seq) => { | |
367 | // iii. Else, T is a complex NT. | |
368 | match seq.separator { | |
369 | // If T has the form $(...)U+ or $(...)U* for some token U, | |
370 | // run the algorithm on the contents with F set to U. If it | |
371 | // accepts, continue, else, reject. | |
372 | Some(ref u) => { | |
373 | let last = check_matcher(cx, seq.tts.iter(), u); | |
374 | match last { | |
375 | // Since the delimiter isn't required after the last | |
376 | // repetition, make sure that the *next* token is | |
377 | // sane. This doesn't actually compute the FIRST of | |
378 | // the rest of the matcher yet, it only considers | |
379 | // single tokens and simple NTs. This is imprecise, | |
380 | // but conservatively correct. | |
381 | Some((span, tok)) => { | |
382 | let fol = match tokens.peek() { | |
383 | Some(&&TtToken(_, ref tok)) => tok.clone(), | |
384 | Some(&&TtDelimited(_, ref delim)) => delim.close_token(), | |
385 | Some(_) => { | |
386 | cx.span_err(sp, "sequence repetition followed by \ | |
387 | another sequence repetition, which is not allowed"); | |
388 | Eof | |
389 | }, | |
390 | None => Eof | |
391 | }; | |
392 | check_matcher(cx, Some(&TtToken(span, tok.clone())).into_iter(), | |
393 | &fol) | |
394 | }, | |
395 | None => last, | |
396 | } | |
397 | }, | |
398 | // If T has the form $(...)+ or $(...)*, run the algorithm | |
399 | // on the contents with F set to the token following the | |
400 | // sequence. If it accepts, continue, else, reject. | |
401 | None => { | |
402 | let fol = match tokens.peek() { | |
403 | Some(&&TtToken(_, ref tok)) => tok.clone(), | |
404 | Some(&&TtDelimited(_, ref delim)) => delim.close_token(), | |
405 | Some(_) => { | |
406 | cx.span_err(sp, "sequence repetition followed by another \ | |
407 | sequence repetition, which is not allowed"); | |
408 | Eof | |
409 | }, | |
410 | None => Eof | |
411 | }; | |
412 | check_matcher(cx, seq.tts.iter(), &fol) | |
413 | } | |
414 | } | |
415 | }, | |
416 | TtToken(..) => { | |
417 | // i. If T is not an NT, continue. | |
418 | continue | |
419 | }, | |
420 | TtDelimited(_, ref tts) => { | |
421 | // if we don't pass in that close delimiter, we'll incorrectly consider the matcher | |
422 | // `{ $foo:ty }` as having a follow that isn't `RBrace` | |
423 | check_matcher(cx, tts.tts.iter(), &tts.close_token()) | |
223e47cc LB |
424 | } |
425 | } | |
223e47cc | 426 | } |
1a4d82fc JJ |
427 | last |
428 | } | |
223e47cc | 429 | |
85aaf69f | 430 | fn is_in_follow(_: &ExtCtxt, tok: &Token, frag: &str) -> Result<bool, String> { |
1a4d82fc | 431 | if let &CloseDelim(_) = tok { |
85aaf69f SL |
432 | Ok(true) |
433 | } else { | |
434 | match frag { | |
435 | "item" => { | |
436 | // since items *must* be followed by either a `;` or a `}`, we can | |
437 | // accept anything after them | |
438 | Ok(true) | |
439 | }, | |
440 | "block" => { | |
441 | // anything can follow block, the braces provide a easy boundary to | |
442 | // maintain | |
443 | Ok(true) | |
444 | }, | |
445 | "stmt" | "expr" => { | |
446 | match *tok { | |
447 | FatArrow | Comma | Semi => Ok(true), | |
448 | _ => Ok(false) | |
449 | } | |
450 | }, | |
451 | "pat" => { | |
452 | match *tok { | |
453 | FatArrow | Comma | Eq => Ok(true), | |
454 | _ => Ok(false) | |
455 | } | |
456 | }, | |
457 | "path" | "ty" => { | |
458 | match *tok { | |
459 | Comma | FatArrow | Colon | Eq | Gt => Ok(true), | |
460 | Ident(i, _) if i.as_str() == "as" => Ok(true), | |
461 | _ => Ok(false) | |
462 | } | |
463 | }, | |
464 | "ident" => { | |
465 | // being a single token, idents are harmless | |
466 | Ok(true) | |
467 | }, | |
468 | "meta" | "tt" => { | |
469 | // being either a single token or a delimited sequence, tt is | |
470 | // harmless | |
471 | Ok(true) | |
472 | }, | |
473 | _ => Err(format!("invalid fragment specifier `{}`", frag)) | |
474 | } | |
1a4d82fc | 475 | } |
223e47cc | 476 | } |