]>
Commit | Line | Data |
---|---|---|
1 | /* Symbol table manager for Bison. | |
2 | ||
3 | Copyright (C) 1984, 1989, 2000-2002, 2004-2013 Free Software | |
4 | Foundation, Inc. | |
5 | ||
6 | This file is part of Bison, the GNU Compiler Compiler. | |
7 | ||
8 | This program is free software: you can redistribute it and/or modify | |
9 | it under the terms of the GNU General Public License as published by | |
10 | the Free Software Foundation, either version 3 of the License, or | |
11 | (at your option) any later version. | |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
19 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ | |
20 | ||
21 | #include <config.h> | |
22 | #include "system.h" | |
23 | ||
24 | #include <hash.h> | |
25 | ||
26 | #include "complain.h" | |
27 | #include "gram.h" | |
28 | #include "symtab.h" | |
29 | ||
30 | /*-------------------------------------------------------------------. | |
31 | | Symbols sorted by tag. Allocated by the first invocation of | | |
32 | | symbols_do, after which no more symbols should be created. | | |
33 | `-------------------------------------------------------------------*/ | |
34 | ||
35 | static symbol **symbols_sorted = NULL; | |
36 | static symbol **semantic_types_sorted = NULL; | |
37 | ||
38 | /*------------------------. | |
39 | | Distinguished symbols. | | |
40 | `------------------------*/ | |
41 | ||
42 | symbol *errtoken = NULL; | |
43 | symbol *undeftoken = NULL; | |
44 | symbol *endtoken = NULL; | |
45 | symbol *accept = NULL; | |
46 | symbol *startsymbol = NULL; | |
47 | location startsymbol_location; | |
48 | ||
49 | /*---------------------------. | |
50 | | Precedence relation graph. | | |
51 | `---------------------------*/ | |
52 | ||
53 | static symgraph **prec_nodes; | |
54 | ||
55 | /*-----------------------------------. | |
56 | | Store which associativity is used. | | |
57 | `-----------------------------------*/ | |
58 | ||
59 | bool *used_assoc = NULL; | |
60 | ||
61 | /*---------------------------------. | |
62 | | Create a new symbol, named TAG. | | |
63 | `---------------------------------*/ | |
64 | ||
65 | static symbol * | |
66 | symbol_new (uniqstr tag, location loc) | |
67 | { | |
68 | symbol *res = xmalloc (sizeof *res); | |
69 | uniqstr_assert (tag); | |
70 | ||
71 | /* If the tag is not a string (starts with a double quote), check | |
72 | that it is valid for Yacc. */ | |
73 | if (tag[0] != '\"' && tag[0] != '\'' && strchr (tag, '-')) | |
74 | complain (&loc, Wyacc, | |
75 | _("POSIX Yacc forbids dashes in symbol names: %s"), tag); | |
76 | ||
77 | res->tag = tag; | |
78 | res->location = loc; | |
79 | ||
80 | res->type_name = NULL; | |
81 | { | |
82 | int i; | |
83 | for (i = 0; i < CODE_PROPS_SIZE; ++i) | |
84 | code_props_none_init (&res->props[i]); | |
85 | } | |
86 | ||
87 | res->number = NUMBER_UNDEFINED; | |
88 | res->prec = 0; | |
89 | res->assoc = undef_assoc; | |
90 | res->user_token_number = USER_NUMBER_UNDEFINED; | |
91 | ||
92 | res->alias = NULL; | |
93 | res->class = unknown_sym; | |
94 | res->status = undeclared; | |
95 | ||
96 | if (nsyms == SYMBOL_NUMBER_MAXIMUM) | |
97 | complain (NULL, fatal, _("too many symbols in input grammar (limit is %d)"), | |
98 | SYMBOL_NUMBER_MAXIMUM); | |
99 | nsyms++; | |
100 | return res; | |
101 | } | |
102 | ||
103 | char const * | |
104 | code_props_type_string (code_props_type kind) | |
105 | { | |
106 | switch (kind) | |
107 | { | |
108 | case destructor: | |
109 | return "%destructor"; | |
110 | case printer: | |
111 | return "%printer"; | |
112 | } | |
113 | assert (0); | |
114 | } | |
115 | ||
116 | /*----------------------------------------. | |
117 | | Create a new semantic type, named TAG. | | |
118 | `----------------------------------------*/ | |
119 | ||
120 | static semantic_type * | |
121 | semantic_type_new (uniqstr tag, const location *loc) | |
122 | { | |
123 | semantic_type *res = xmalloc (sizeof *res); | |
124 | ||
125 | uniqstr_assert (tag); | |
126 | res->tag = tag; | |
127 | res->location = loc ? *loc : empty_location; | |
128 | res->status = undeclared; | |
129 | { | |
130 | int i; | |
131 | for (i = 0; i < CODE_PROPS_SIZE; ++i) | |
132 | code_props_none_init (&res->props[i]); | |
133 | } | |
134 | ||
135 | return res; | |
136 | } | |
137 | ||
138 | ||
139 | /*-----------------. | |
140 | | Print a symbol. | | |
141 | `-----------------*/ | |
142 | ||
143 | #define SYMBOL_ATTR_PRINT(Attr) \ | |
144 | if (s->Attr) \ | |
145 | fprintf (f, " %s { %s }", #Attr, s->Attr) | |
146 | ||
147 | #define SYMBOL_CODE_PRINT(Attr) \ | |
148 | if (s->props[Attr].code) \ | |
149 | fprintf (f, " %s { %s }", #Attr, s->props[Attr].code) | |
150 | ||
151 | void | |
152 | symbol_print (symbol const *s, FILE *f) | |
153 | { | |
154 | if (s) | |
155 | { | |
156 | fprintf (f, "\"%s\"", s->tag); | |
157 | SYMBOL_ATTR_PRINT (type_name); | |
158 | SYMBOL_CODE_PRINT (destructor); | |
159 | SYMBOL_CODE_PRINT (printer); | |
160 | } | |
161 | else | |
162 | fprintf (f, "<NULL>"); | |
163 | } | |
164 | ||
165 | #undef SYMBOL_ATTR_PRINT | |
166 | #undef SYMBOL_CODE_PRINT | |
167 | ||
168 | ||
169 | /*----------------------------------. | |
170 | | Whether S is a valid identifier. | | |
171 | `----------------------------------*/ | |
172 | ||
173 | static bool | |
174 | is_identifier (uniqstr s) | |
175 | { | |
176 | static char const alphanum[26 + 26 + 1 + 10] = | |
177 | "abcdefghijklmnopqrstuvwxyz" | |
178 | "ABCDEFGHIJKLMNOPQRSTUVWXYZ" | |
179 | "_" | |
180 | "0123456789"; | |
181 | if (!s || ! memchr (alphanum, *s, sizeof alphanum - 10)) | |
182 | return false; | |
183 | for (++s; *s; ++s) | |
184 | if (! memchr (alphanum, *s, sizeof alphanum)) | |
185 | return false; | |
186 | return true; | |
187 | } | |
188 | ||
189 | ||
190 | /*-----------------------------------------------. | |
191 | | Get the identifier associated to this symbol. | | |
192 | `-----------------------------------------------*/ | |
193 | uniqstr | |
194 | symbol_id_get (symbol const *sym) | |
195 | { | |
196 | aver (sym->user_token_number != USER_NUMBER_HAS_STRING_ALIAS); | |
197 | if (sym->alias) | |
198 | sym = sym->alias; | |
199 | return is_identifier (sym->tag) ? sym->tag : 0; | |
200 | } | |
201 | ||
202 | ||
203 | /*------------------------------------------------------------------. | |
204 | | Complain that S's WHAT is redeclared at SECOND, and was first set | | |
205 | | at FIRST. | | |
206 | `------------------------------------------------------------------*/ | |
207 | ||
208 | static void | |
209 | symbol_redeclaration (symbol *s, const char *what, location first, | |
210 | location second) | |
211 | { | |
212 | unsigned i = 0; | |
213 | complain_indent (&second, complaint, &i, | |
214 | _("%s redeclaration for %s"), what, s->tag); | |
215 | i += SUB_INDENT; | |
216 | complain_indent (&first, complaint, &i, | |
217 | _("previous declaration")); | |
218 | } | |
219 | ||
220 | static void | |
221 | semantic_type_redeclaration (semantic_type *s, const char *what, location first, | |
222 | location second) | |
223 | { | |
224 | unsigned i = 0; | |
225 | complain_indent (&second, complaint, &i, | |
226 | _("%s redeclaration for <%s>"), what, s->tag); | |
227 | i += SUB_INDENT; | |
228 | complain_indent (&first, complaint, &i, | |
229 | _("previous declaration")); | |
230 | } | |
231 | ||
232 | ||
233 | ||
234 | /*-----------------------------------------------------------------. | |
235 | | Set the TYPE_NAME associated with SYM. Does nothing if passed 0 | | |
236 | | as TYPE_NAME. | | |
237 | `-----------------------------------------------------------------*/ | |
238 | ||
239 | void | |
240 | symbol_type_set (symbol *sym, uniqstr type_name, location loc) | |
241 | { | |
242 | if (type_name) | |
243 | { | |
244 | if (sym->type_name) | |
245 | symbol_redeclaration (sym, "%type", sym->type_location, loc); | |
246 | uniqstr_assert (type_name); | |
247 | sym->type_name = type_name; | |
248 | sym->type_location = loc; | |
249 | } | |
250 | } | |
251 | ||
252 | /*--------------------------------------------------------. | |
253 | | Set the DESTRUCTOR or PRINTER associated with the SYM. | | |
254 | `--------------------------------------------------------*/ | |
255 | ||
256 | void | |
257 | symbol_code_props_set (symbol *sym, code_props_type kind, | |
258 | code_props const *code) | |
259 | { | |
260 | if (sym->props[kind].code) | |
261 | symbol_redeclaration (sym, code_props_type_string (kind), | |
262 | sym->props[kind].location, | |
263 | code->location); | |
264 | sym->props[kind] = *code; | |
265 | } | |
266 | ||
267 | /*-----------------------------------------------------. | |
268 | | Set the DESTRUCTOR or PRINTER associated with TYPE. | | |
269 | `-----------------------------------------------------*/ | |
270 | ||
271 | void | |
272 | semantic_type_code_props_set (semantic_type *type, | |
273 | code_props_type kind, | |
274 | code_props const *code) | |
275 | { | |
276 | if (type->props[kind].code) | |
277 | semantic_type_redeclaration (type, code_props_type_string (kind), | |
278 | type->props[kind].location, | |
279 | code->location); | |
280 | type->props[kind] = *code; | |
281 | } | |
282 | ||
283 | /*---------------------------------------------------. | |
284 | | Get the computed %destructor or %printer for SYM. | | |
285 | `---------------------------------------------------*/ | |
286 | ||
287 | code_props * | |
288 | symbol_code_props_get (symbol *sym, code_props_type kind) | |
289 | { | |
290 | /* Per-symbol code props. */ | |
291 | if (sym->props[kind].code) | |
292 | return &sym->props[kind]; | |
293 | ||
294 | /* Per-type code props. */ | |
295 | if (sym->type_name) | |
296 | { | |
297 | code_props *code = | |
298 | &semantic_type_get (sym->type_name, NULL)->props[kind]; | |
299 | if (code->code) | |
300 | return code; | |
301 | } | |
302 | ||
303 | /* Apply default code props's only to user-defined symbols. */ | |
304 | if (sym->tag[0] != '$' && sym != errtoken) | |
305 | { | |
306 | code_props *code = | |
307 | &semantic_type_get (sym->type_name ? "*" : "", NULL)->props[kind]; | |
308 | if (code->code) | |
309 | return code; | |
310 | } | |
311 | return &code_props_none; | |
312 | } | |
313 | ||
314 | /*-----------------------------------------------------------------. | |
315 | | Set the PRECEDENCE associated with SYM. Does nothing if invoked | | |
316 | | with UNDEF_ASSOC as ASSOC. | | |
317 | `-----------------------------------------------------------------*/ | |
318 | ||
319 | void | |
320 | symbol_precedence_set (symbol *sym, int prec, assoc a, location loc) | |
321 | { | |
322 | if (a != undef_assoc) | |
323 | { | |
324 | if (sym->prec != 0) | |
325 | symbol_redeclaration (sym, assoc_to_string (a), sym->prec_location, | |
326 | loc); | |
327 | sym->prec = prec; | |
328 | sym->assoc = a; | |
329 | sym->prec_location = loc; | |
330 | } | |
331 | ||
332 | /* Only terminals have a precedence. */ | |
333 | symbol_class_set (sym, token_sym, loc, false); | |
334 | } | |
335 | ||
336 | ||
337 | /*------------------------------------. | |
338 | | Set the CLASS associated with SYM. | | |
339 | `------------------------------------*/ | |
340 | ||
341 | void | |
342 | symbol_class_set (symbol *sym, symbol_class class, location loc, bool declaring) | |
343 | { | |
344 | bool warned = false; | |
345 | if (sym->class != unknown_sym && sym->class != class) | |
346 | { | |
347 | complain (&loc, complaint, _("symbol %s redefined"), sym->tag); | |
348 | /* Don't report both "redefined" and "redeclared". */ | |
349 | warned = true; | |
350 | } | |
351 | ||
352 | if (class == nterm_sym && sym->class != nterm_sym) | |
353 | sym->number = nvars++; | |
354 | else if (class == token_sym && sym->number == NUMBER_UNDEFINED) | |
355 | sym->number = ntokens++; | |
356 | ||
357 | sym->class = class; | |
358 | ||
359 | if (declaring) | |
360 | { | |
361 | if (sym->status == declared && !warned) | |
362 | complain (&loc, Wother, _("symbol %s redeclared"), sym->tag); | |
363 | sym->status = declared; | |
364 | } | |
365 | } | |
366 | ||
367 | ||
368 | /*------------------------------------------------. | |
369 | | Set the USER_TOKEN_NUMBER associated with SYM. | | |
370 | `------------------------------------------------*/ | |
371 | ||
372 | void | |
373 | symbol_user_token_number_set (symbol *sym, int user_token_number, location loc) | |
374 | { | |
375 | int *user_token_numberp; | |
376 | ||
377 | if (sym->user_token_number != USER_NUMBER_HAS_STRING_ALIAS) | |
378 | user_token_numberp = &sym->user_token_number; | |
379 | else | |
380 | user_token_numberp = &sym->alias->user_token_number; | |
381 | if (*user_token_numberp != USER_NUMBER_UNDEFINED | |
382 | && *user_token_numberp != user_token_number) | |
383 | complain (&loc, complaint, _("redefining user token number of %s"), | |
384 | sym->tag); | |
385 | ||
386 | *user_token_numberp = user_token_number; | |
387 | /* User defined $end token? */ | |
388 | if (user_token_number == 0) | |
389 | { | |
390 | endtoken = sym; | |
391 | /* It is always mapped to 0, so it was already counted in | |
392 | NTOKENS. */ | |
393 | if (endtoken->number != NUMBER_UNDEFINED) | |
394 | --ntokens; | |
395 | endtoken->number = 0; | |
396 | } | |
397 | } | |
398 | ||
399 | ||
400 | /*----------------------------------------------------------. | |
401 | | If SYM is not defined, report an error, and consider it a | | |
402 | | nonterminal. | | |
403 | `----------------------------------------------------------*/ | |
404 | ||
405 | static inline bool | |
406 | symbol_check_defined (symbol *sym) | |
407 | { | |
408 | if (sym->class == unknown_sym) | |
409 | { | |
410 | assert (sym->status != declared); | |
411 | complain (&sym->location, | |
412 | sym->status == needed ? complaint : Wother, | |
413 | _("symbol %s is used, but is not defined as a token" | |
414 | " and has no rules"), | |
415 | sym->tag); | |
416 | sym->class = nterm_sym; | |
417 | sym->number = nvars++; | |
418 | } | |
419 | ||
420 | { | |
421 | int i; | |
422 | for (i = 0; i < 2; ++i) | |
423 | symbol_code_props_get (sym, i)->is_used = true; | |
424 | } | |
425 | ||
426 | /* Set the semantic type status associated to the current symbol to | |
427 | 'declared' so that we could check semantic types unnecessary uses. */ | |
428 | if (sym->type_name) | |
429 | { | |
430 | semantic_type *sem_type = semantic_type_get (sym->type_name, NULL); | |
431 | if (sem_type) | |
432 | sem_type->status = declared; | |
433 | } | |
434 | ||
435 | return true; | |
436 | } | |
437 | ||
438 | static inline bool | |
439 | semantic_type_check_defined (semantic_type *sem_type) | |
440 | { | |
441 | /* <*> and <> do not have to be "declared". */ | |
442 | if (sem_type->status == declared | |
443 | || !*sem_type->tag | |
444 | || STREQ(sem_type->tag, "*")) | |
445 | { | |
446 | int i; | |
447 | for (i = 0; i < 2; ++i) | |
448 | if (sem_type->props[i].kind != CODE_PROPS_NONE | |
449 | && ! sem_type->props[i].is_used) | |
450 | complain (&sem_type->location, Wother, | |
451 | _("useless %s for type <%s>"), | |
452 | code_props_type_string (i), sem_type->tag); | |
453 | } | |
454 | else | |
455 | complain (&sem_type->location, Wother, | |
456 | _("type <%s> is used, but is not associated to any symbol"), | |
457 | sem_type->tag); | |
458 | ||
459 | return true; | |
460 | } | |
461 | ||
462 | static bool | |
463 | symbol_check_defined_processor (void *sym, void *null ATTRIBUTE_UNUSED) | |
464 | { | |
465 | return symbol_check_defined (sym); | |
466 | } | |
467 | ||
468 | static bool | |
469 | semantic_type_check_defined_processor (void *sem_type, | |
470 | void *null ATTRIBUTE_UNUSED) | |
471 | { | |
472 | return semantic_type_check_defined (sem_type); | |
473 | } | |
474 | ||
475 | ||
476 | void | |
477 | symbol_make_alias (symbol *sym, symbol *str, location loc) | |
478 | { | |
479 | if (str->alias) | |
480 | complain (&loc, Wother, | |
481 | _("symbol %s used more than once as a literal string"), str->tag); | |
482 | else if (sym->alias) | |
483 | complain (&loc, Wother, | |
484 | _("symbol %s given more than one literal string"), sym->tag); | |
485 | else | |
486 | { | |
487 | str->class = token_sym; | |
488 | str->user_token_number = sym->user_token_number; | |
489 | sym->user_token_number = USER_NUMBER_HAS_STRING_ALIAS; | |
490 | str->alias = sym; | |
491 | sym->alias = str; | |
492 | str->number = sym->number; | |
493 | symbol_type_set (str, sym->type_name, loc); | |
494 | } | |
495 | } | |
496 | ||
497 | ||
498 | /*---------------------------------------------------------. | |
499 | | Check that THIS, and its alias, have same precedence and | | |
500 | | associativity. | | |
501 | `---------------------------------------------------------*/ | |
502 | ||
503 | static inline void | |
504 | symbol_check_alias_consistency (symbol *this) | |
505 | { | |
506 | symbol *sym = this; | |
507 | symbol *str = this->alias; | |
508 | ||
509 | /* Check only the symbol in the symbol-string pair. */ | |
510 | if (!(this->alias | |
511 | && this->user_token_number == USER_NUMBER_HAS_STRING_ALIAS)) | |
512 | return; | |
513 | ||
514 | if (str->type_name != sym->type_name) | |
515 | { | |
516 | if (str->type_name) | |
517 | symbol_type_set (sym, str->type_name, str->type_location); | |
518 | else | |
519 | symbol_type_set (str, sym->type_name, sym->type_location); | |
520 | } | |
521 | ||
522 | ||
523 | { | |
524 | int i; | |
525 | for (i = 0; i < CODE_PROPS_SIZE; ++i) | |
526 | if (str->props[i].code) | |
527 | symbol_code_props_set (sym, i, &str->props[i]); | |
528 | else if (sym->props[i].code) | |
529 | symbol_code_props_set (str, i, &sym->props[i]); | |
530 | } | |
531 | ||
532 | if (sym->prec || str->prec) | |
533 | { | |
534 | if (str->prec) | |
535 | symbol_precedence_set (sym, str->prec, str->assoc, | |
536 | str->prec_location); | |
537 | else | |
538 | symbol_precedence_set (str, sym->prec, sym->assoc, | |
539 | sym->prec_location); | |
540 | } | |
541 | } | |
542 | ||
543 | static bool | |
544 | symbol_check_alias_consistency_processor (void *this, | |
545 | void *null ATTRIBUTE_UNUSED) | |
546 | { | |
547 | symbol_check_alias_consistency (this); | |
548 | return true; | |
549 | } | |
550 | ||
551 | ||
552 | /*-------------------------------------------------------------------. | |
553 | | Assign a symbol number, and write the definition of the token name | | |
554 | | into FDEFINES. Put in SYMBOLS. | | |
555 | `-------------------------------------------------------------------*/ | |
556 | ||
557 | static inline bool | |
558 | symbol_pack (symbol *this) | |
559 | { | |
560 | aver (this->number != NUMBER_UNDEFINED); | |
561 | if (this->class == nterm_sym) | |
562 | this->number += ntokens; | |
563 | else if (this->user_token_number == USER_NUMBER_HAS_STRING_ALIAS) | |
564 | return true; | |
565 | ||
566 | symbols[this->number] = this; | |
567 | return true; | |
568 | } | |
569 | ||
570 | static bool | |
571 | symbol_pack_processor (void *this, void *null ATTRIBUTE_UNUSED) | |
572 | { | |
573 | return symbol_pack (this); | |
574 | } | |
575 | ||
576 | ||
577 | static void | |
578 | user_token_number_redeclaration (int num, symbol *first, symbol *second) | |
579 | { | |
580 | unsigned i = 0; | |
581 | /* User token numbers are not assigned during the parsing, but in a | |
582 | second step, via a traversal of the symbol table sorted on tag. | |
583 | ||
584 | However, error messages make more sense if we keep the first | |
585 | declaration first. */ | |
586 | if (location_cmp (first->location, second->location) > 0) | |
587 | { | |
588 | symbol* tmp = first; | |
589 | first = second; | |
590 | second = tmp; | |
591 | } | |
592 | complain_indent (&second->location, complaint, &i, | |
593 | _("user token number %d redeclaration for %s"), | |
594 | num, second->tag); | |
595 | i += SUB_INDENT; | |
596 | complain_indent (&first->location, complaint, &i, | |
597 | _("previous declaration for %s"), | |
598 | first->tag); | |
599 | } | |
600 | ||
601 | /*--------------------------------------------------. | |
602 | | Put THIS in TOKEN_TRANSLATIONS if it is a token. | | |
603 | `--------------------------------------------------*/ | |
604 | ||
605 | static inline bool | |
606 | symbol_translation (symbol *this) | |
607 | { | |
608 | /* Non-terminal? */ | |
609 | if (this->class == token_sym | |
610 | && this->user_token_number != USER_NUMBER_HAS_STRING_ALIAS) | |
611 | { | |
612 | /* A token which translation has already been set? */ | |
613 | if (token_translations[this->user_token_number] != undeftoken->number) | |
614 | user_token_number_redeclaration | |
615 | (this->user_token_number, | |
616 | symbols[token_translations[this->user_token_number]], | |
617 | this); | |
618 | ||
619 | token_translations[this->user_token_number] = this->number; | |
620 | } | |
621 | ||
622 | return true; | |
623 | } | |
624 | ||
625 | static bool | |
626 | symbol_translation_processor (void *this, void *null ATTRIBUTE_UNUSED) | |
627 | { | |
628 | return symbol_translation (this); | |
629 | } | |
630 | ||
631 | ||
632 | /*---------------------------------------. | |
633 | | Symbol and semantic type hash tables. | | |
634 | `---------------------------------------*/ | |
635 | ||
636 | /* Initial capacity of symbol and semantic type hash table. */ | |
637 | #define HT_INITIAL_CAPACITY 257 | |
638 | ||
639 | static struct hash_table *symbol_table = NULL; | |
640 | static struct hash_table *semantic_type_table = NULL; | |
641 | ||
642 | static inline bool | |
643 | hash_compare_symbol (const symbol *m1, const symbol *m2) | |
644 | { | |
645 | /* Since tags are unique, we can compare the pointers themselves. */ | |
646 | return UNIQSTR_EQ (m1->tag, m2->tag); | |
647 | } | |
648 | ||
649 | static inline bool | |
650 | hash_compare_semantic_type (const semantic_type *m1, const semantic_type *m2) | |
651 | { | |
652 | /* Since names are unique, we can compare the pointers themselves. */ | |
653 | return UNIQSTR_EQ (m1->tag, m2->tag); | |
654 | } | |
655 | ||
656 | static bool | |
657 | hash_symbol_comparator (void const *m1, void const *m2) | |
658 | { | |
659 | return hash_compare_symbol (m1, m2); | |
660 | } | |
661 | ||
662 | static bool | |
663 | hash_semantic_type_comparator (void const *m1, void const *m2) | |
664 | { | |
665 | return hash_compare_semantic_type (m1, m2); | |
666 | } | |
667 | ||
668 | static inline size_t | |
669 | hash_symbol (const symbol *m, size_t tablesize) | |
670 | { | |
671 | /* Since tags are unique, we can hash the pointer itself. */ | |
672 | return ((uintptr_t) m->tag) % tablesize; | |
673 | } | |
674 | ||
675 | static inline size_t | |
676 | hash_semantic_type (const semantic_type *m, size_t tablesize) | |
677 | { | |
678 | /* Since names are unique, we can hash the pointer itself. */ | |
679 | return ((uintptr_t) m->tag) % tablesize; | |
680 | } | |
681 | ||
682 | static size_t | |
683 | hash_symbol_hasher (void const *m, size_t tablesize) | |
684 | { | |
685 | return hash_symbol (m, tablesize); | |
686 | } | |
687 | ||
688 | static size_t | |
689 | hash_semantic_type_hasher (void const *m, size_t tablesize) | |
690 | { | |
691 | return hash_semantic_type (m, tablesize); | |
692 | } | |
693 | ||
694 | /*-------------------------------. | |
695 | | Create the symbol hash table. | | |
696 | `-------------------------------*/ | |
697 | ||
698 | void | |
699 | symbols_new (void) | |
700 | { | |
701 | symbol_table = hash_initialize (HT_INITIAL_CAPACITY, | |
702 | NULL, | |
703 | hash_symbol_hasher, | |
704 | hash_symbol_comparator, | |
705 | free); | |
706 | semantic_type_table = hash_initialize (HT_INITIAL_CAPACITY, | |
707 | NULL, | |
708 | hash_semantic_type_hasher, | |
709 | hash_semantic_type_comparator, | |
710 | free); | |
711 | } | |
712 | ||
713 | ||
714 | /*----------------------------------------------------------------. | |
715 | | Find the symbol named KEY, and return it. If it does not exist | | |
716 | | yet, create it. | | |
717 | `----------------------------------------------------------------*/ | |
718 | ||
719 | symbol * | |
720 | symbol_from_uniqstr (const uniqstr key, location loc) | |
721 | { | |
722 | symbol probe; | |
723 | symbol *entry; | |
724 | ||
725 | probe.tag = key; | |
726 | entry = hash_lookup (symbol_table, &probe); | |
727 | ||
728 | if (!entry) | |
729 | { | |
730 | /* First insertion in the hash. */ | |
731 | aver (!symbols_sorted); | |
732 | entry = symbol_new (key, loc); | |
733 | if (!hash_insert (symbol_table, entry)) | |
734 | xalloc_die (); | |
735 | } | |
736 | return entry; | |
737 | } | |
738 | ||
739 | ||
740 | /*-----------------------------------------------------------------------. | |
741 | | Find the semantic type named KEY, and return it. If it does not exist | | |
742 | | yet, create it. | | |
743 | `-----------------------------------------------------------------------*/ | |
744 | ||
745 | semantic_type * | |
746 | semantic_type_from_uniqstr (const uniqstr key, const location *loc) | |
747 | { | |
748 | semantic_type probe; | |
749 | semantic_type *entry; | |
750 | ||
751 | probe.tag = key; | |
752 | entry = hash_lookup (semantic_type_table, &probe); | |
753 | ||
754 | if (!entry) | |
755 | { | |
756 | /* First insertion in the hash. */ | |
757 | entry = semantic_type_new (key, loc); | |
758 | if (!hash_insert (semantic_type_table, entry)) | |
759 | xalloc_die (); | |
760 | } | |
761 | return entry; | |
762 | } | |
763 | ||
764 | ||
765 | /*----------------------------------------------------------------. | |
766 | | Find the symbol named KEY, and return it. If it does not exist | | |
767 | | yet, create it. | | |
768 | `----------------------------------------------------------------*/ | |
769 | ||
770 | symbol * | |
771 | symbol_get (const char *key, location loc) | |
772 | { | |
773 | return symbol_from_uniqstr (uniqstr_new (key), loc); | |
774 | } | |
775 | ||
776 | ||
777 | /*-----------------------------------------------------------------------. | |
778 | | Find the semantic type named KEY, and return it. If it does not exist | | |
779 | | yet, create it. | | |
780 | `-----------------------------------------------------------------------*/ | |
781 | ||
782 | semantic_type * | |
783 | semantic_type_get (const char *key, const location *loc) | |
784 | { | |
785 | return semantic_type_from_uniqstr (uniqstr_new (key), loc); | |
786 | } | |
787 | ||
788 | ||
789 | /*------------------------------------------------------------------. | |
790 | | Generate a dummy nonterminal, whose name cannot conflict with the | | |
791 | | user's names. | | |
792 | `------------------------------------------------------------------*/ | |
793 | ||
794 | symbol * | |
795 | dummy_symbol_get (location loc) | |
796 | { | |
797 | /* Incremented for each generated symbol. */ | |
798 | static int dummy_count = 0; | |
799 | static char buf[256]; | |
800 | ||
801 | symbol *sym; | |
802 | ||
803 | sprintf (buf, "$@%d", ++dummy_count); | |
804 | sym = symbol_get (buf, loc); | |
805 | sym->class = nterm_sym; | |
806 | sym->number = nvars++; | |
807 | return sym; | |
808 | } | |
809 | ||
810 | bool | |
811 | symbol_is_dummy (const symbol *sym) | |
812 | { | |
813 | return sym->tag[0] == '@' || (sym->tag[0] == '$' && sym->tag[1] == '@'); | |
814 | } | |
815 | ||
816 | /*-------------------. | |
817 | | Free the symbols. | | |
818 | `-------------------*/ | |
819 | ||
820 | void | |
821 | symbols_free (void) | |
822 | { | |
823 | hash_free (symbol_table); | |
824 | hash_free (semantic_type_table); | |
825 | free (symbols); | |
826 | free (symbols_sorted); | |
827 | free (semantic_types_sorted); | |
828 | } | |
829 | ||
830 | ||
831 | /*---------------------------------------------------------------. | |
832 | | Look for undefined symbols, report an error, and consider them | | |
833 | | terminals. | | |
834 | `---------------------------------------------------------------*/ | |
835 | ||
836 | static int | |
837 | symbols_cmp (symbol const *a, symbol const *b) | |
838 | { | |
839 | return strcmp (a->tag, b->tag); | |
840 | } | |
841 | ||
842 | static int | |
843 | symbols_cmp_qsort (void const *a, void const *b) | |
844 | { | |
845 | return symbols_cmp (*(symbol * const *)a, *(symbol * const *)b); | |
846 | } | |
847 | ||
848 | static void | |
849 | symbols_do (Hash_processor processor, void *processor_data, | |
850 | struct hash_table *table, symbol ***sorted) | |
851 | { | |
852 | size_t count = hash_get_n_entries (table); | |
853 | if (!*sorted) | |
854 | { | |
855 | *sorted = xnmalloc (count, sizeof **sorted); | |
856 | hash_get_entries (table, (void**)*sorted, count); | |
857 | qsort (*sorted, count, sizeof **sorted, symbols_cmp_qsort); | |
858 | } | |
859 | { | |
860 | size_t i; | |
861 | for (i = 0; i < count; ++i) | |
862 | processor ((*sorted)[i], processor_data); | |
863 | } | |
864 | } | |
865 | ||
866 | /*--------------------------------------------------------------. | |
867 | | Check that all the symbols are defined. Report any undefined | | |
868 | | symbols and consider them nonterminals. | | |
869 | `--------------------------------------------------------------*/ | |
870 | ||
871 | void | |
872 | symbols_check_defined (void) | |
873 | { | |
874 | symbols_do (symbol_check_defined_processor, NULL, | |
875 | symbol_table, &symbols_sorted); | |
876 | symbols_do (semantic_type_check_defined_processor, NULL, | |
877 | semantic_type_table, &semantic_types_sorted); | |
878 | } | |
879 | ||
880 | /*------------------------------------------------------------------. | |
881 | | Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | | |
882 | | number. | | |
883 | `------------------------------------------------------------------*/ | |
884 | ||
885 | static void | |
886 | symbols_token_translations_init (void) | |
887 | { | |
888 | bool num_256_available_p = true; | |
889 | int i; | |
890 | ||
891 | /* Find the highest user token number, and whether 256, the POSIX | |
892 | preferred user token number for the error token, is used. */ | |
893 | max_user_token_number = 0; | |
894 | for (i = 0; i < ntokens; ++i) | |
895 | { | |
896 | symbol *this = symbols[i]; | |
897 | if (this->user_token_number != USER_NUMBER_UNDEFINED) | |
898 | { | |
899 | if (this->user_token_number > max_user_token_number) | |
900 | max_user_token_number = this->user_token_number; | |
901 | if (this->user_token_number == 256) | |
902 | num_256_available_p = false; | |
903 | } | |
904 | } | |
905 | ||
906 | /* If 256 is not used, assign it to error, to follow POSIX. */ | |
907 | if (num_256_available_p | |
908 | && errtoken->user_token_number == USER_NUMBER_UNDEFINED) | |
909 | errtoken->user_token_number = 256; | |
910 | ||
911 | /* Set the missing user numbers. */ | |
912 | if (max_user_token_number < 256) | |
913 | max_user_token_number = 256; | |
914 | ||
915 | for (i = 0; i < ntokens; ++i) | |
916 | { | |
917 | symbol *this = symbols[i]; | |
918 | if (this->user_token_number == USER_NUMBER_UNDEFINED) | |
919 | this->user_token_number = ++max_user_token_number; | |
920 | if (this->user_token_number > max_user_token_number) | |
921 | max_user_token_number = this->user_token_number; | |
922 | } | |
923 | ||
924 | token_translations = xnmalloc (max_user_token_number + 1, | |
925 | sizeof *token_translations); | |
926 | ||
927 | /* Initialize all entries for literal tokens to the internal token | |
928 | number for $undefined, which represents all invalid inputs. */ | |
929 | for (i = 0; i < max_user_token_number + 1; i++) | |
930 | token_translations[i] = undeftoken->number; | |
931 | symbols_do (symbol_translation_processor, NULL, | |
932 | symbol_table, &symbols_sorted); | |
933 | } | |
934 | ||
935 | ||
936 | /*----------------------------------------------------------------. | |
937 | | Assign symbol numbers, and write definition of token names into | | |
938 | | FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | | |
939 | `----------------------------------------------------------------*/ | |
940 | ||
941 | void | |
942 | symbols_pack (void) | |
943 | { | |
944 | symbols_do (symbol_check_alias_consistency_processor, NULL, | |
945 | symbol_table, &symbols_sorted); | |
946 | ||
947 | symbols = xcalloc (nsyms, sizeof *symbols); | |
948 | symbols_do (symbol_pack_processor, NULL, symbol_table, &symbols_sorted); | |
949 | ||
950 | /* Aliases leave empty slots in symbols, so remove them. */ | |
951 | { | |
952 | int writei; | |
953 | int readi; | |
954 | int nsyms_old = nsyms; | |
955 | for (writei = 0, readi = 0; readi < nsyms_old; readi += 1) | |
956 | { | |
957 | if (symbols[readi] == NULL) | |
958 | { | |
959 | nsyms -= 1; | |
960 | ntokens -= 1; | |
961 | } | |
962 | else | |
963 | { | |
964 | symbols[writei] = symbols[readi]; | |
965 | symbols[writei]->number = writei; | |
966 | if (symbols[writei]->alias) | |
967 | symbols[writei]->alias->number = writei; | |
968 | writei += 1; | |
969 | } | |
970 | } | |
971 | } | |
972 | symbols = xnrealloc (symbols, nsyms, sizeof *symbols); | |
973 | ||
974 | symbols_token_translations_init (); | |
975 | ||
976 | if (startsymbol->class == unknown_sym) | |
977 | complain (&startsymbol_location, fatal, | |
978 | _("the start symbol %s is undefined"), | |
979 | startsymbol->tag); | |
980 | else if (startsymbol->class == token_sym) | |
981 | complain (&startsymbol_location, fatal, | |
982 | _("the start symbol %s is a token"), | |
983 | startsymbol->tag); | |
984 | } | |
985 | ||
986 | /*---------------------------------. | |
987 | | Initialize relation graph nodes. | | |
988 | `---------------------------------*/ | |
989 | ||
990 | static void | |
991 | init_prec_nodes (void) | |
992 | { | |
993 | int i; | |
994 | prec_nodes = xcalloc (nsyms, sizeof *prec_nodes); | |
995 | for (i = 0; i < nsyms; ++i) | |
996 | { | |
997 | prec_nodes[i] = xmalloc (sizeof *prec_nodes[i]); | |
998 | symgraph *s = prec_nodes[i]; | |
999 | s->id = i; | |
1000 | s->succ = 0; | |
1001 | s->pred = 0; | |
1002 | } | |
1003 | } | |
1004 | ||
1005 | /*----------------. | |
1006 | | Create a link. | | |
1007 | `----------------*/ | |
1008 | ||
1009 | static symgraphlink * | |
1010 | symgraphlink_new (graphid id, symgraphlink *next) | |
1011 | { | |
1012 | symgraphlink *l = xmalloc (sizeof *l); | |
1013 | l->id = id; | |
1014 | l->next = next; | |
1015 | return l; | |
1016 | } | |
1017 | ||
1018 | ||
1019 | /*------------------------------------------------------------------. | |
1020 | | Register the second symbol of the precedence relation, and return | | |
1021 | | whether this relation is new. Use only in register_precedence. | | |
1022 | `------------------------------------------------------------------*/ | |
1023 | ||
1024 | static bool | |
1025 | register_precedence_second_symbol (symgraphlink **first, graphid sym) | |
1026 | { | |
1027 | if (!*first || sym < (*first)->id) | |
1028 | *first = symgraphlink_new (sym, *first); | |
1029 | else | |
1030 | { | |
1031 | symgraphlink *slist = *first; | |
1032 | ||
1033 | while (slist->next && slist->next->id <= sym) | |
1034 | slist = slist->next; | |
1035 | ||
1036 | if (slist->id == sym) | |
1037 | /* Relation already present. */ | |
1038 | return false; | |
1039 | ||
1040 | slist->next = symgraphlink_new (sym, slist->next); | |
1041 | } | |
1042 | return true; | |
1043 | } | |
1044 | ||
1045 | /*------------------------------------------------------------------. | |
1046 | | Register a new relation between symbols as used. The first symbol | | |
1047 | | has a greater precedence than the second one. | | |
1048 | `------------------------------------------------------------------*/ | |
1049 | ||
1050 | void | |
1051 | register_precedence (graphid first, graphid snd) | |
1052 | { | |
1053 | if (!prec_nodes) | |
1054 | init_prec_nodes (); | |
1055 | register_precedence_second_symbol (&(prec_nodes[first]->succ), snd); | |
1056 | register_precedence_second_symbol (&(prec_nodes[snd]->pred), first); | |
1057 | } | |
1058 | ||
1059 | ||
1060 | /*---------------------------------------. | |
1061 | | Initialize association tracking table. | | |
1062 | `---------------------------------------*/ | |
1063 | ||
1064 | static void | |
1065 | init_assoc (void) | |
1066 | { | |
1067 | graphid i; | |
1068 | used_assoc = xcalloc(nsyms, sizeof(*used_assoc)); | |
1069 | for (i = 0; i < nsyms; ++i) | |
1070 | used_assoc[i] = false; | |
1071 | } | |
1072 | ||
1073 | /*------------------------------------------------------------------. | |
1074 | | Test if the associativity for the symbols is defined and useless. | | |
1075 | `------------------------------------------------------------------*/ | |
1076 | ||
1077 | static inline bool | |
1078 | is_assoc_useless (symbol *s) | |
1079 | { | |
1080 | return s | |
1081 | && s->assoc != undef_assoc | |
1082 | && s->assoc != precedence_assoc | |
1083 | && !used_assoc[s->number]; | |
1084 | } | |
1085 | ||
1086 | /*-------------------------------. | |
1087 | | Register a used associativity. | | |
1088 | `-------------------------------*/ | |
1089 | ||
1090 | void | |
1091 | register_assoc (graphid i, graphid j) | |
1092 | { | |
1093 | if (!used_assoc) | |
1094 | init_assoc (); | |
1095 | used_assoc[i] = true; | |
1096 | used_assoc[j] = true; | |
1097 | } | |
1098 | ||
1099 | /*--------------------------------------------------. | |
1100 | | Print a warning for unused precedence relations. | | |
1101 | `--------------------------------------------------*/ | |
1102 | ||
1103 | void | |
1104 | print_precedence_warnings (void) | |
1105 | { | |
1106 | int i; | |
1107 | if (!prec_nodes) | |
1108 | init_prec_nodes (); | |
1109 | if (!used_assoc) | |
1110 | init_assoc (); | |
1111 | for (i = 0; i < nsyms; ++i) | |
1112 | { | |
1113 | symbol *s = symbols[i]; | |
1114 | if (s | |
1115 | && s->prec != 0 | |
1116 | && !prec_nodes[i]->pred | |
1117 | && !prec_nodes[i]->succ) | |
1118 | { | |
1119 | if (is_assoc_useless (s)) | |
1120 | complain (&s->location, Wprecedence, | |
1121 | _("useless precedence and associativity for %s"), s->tag); | |
1122 | else if (s->assoc == precedence_assoc) | |
1123 | complain (&s->location, Wprecedence, | |
1124 | _("useless precedence for %s"), s->tag); | |
1125 | } | |
1126 | else if (is_assoc_useless (s)) | |
1127 | complain (&s->location, Wprecedence, | |
1128 | _("useless associativity for %s, use %%precedence"), s->tag); | |
1129 | } | |
1130 | } |