]> git.saurik.com Git - bison.git/blob - src/reader.c
Regen.
[bison.git] / src / reader.c
1 /* Input parser for Bison
2
3 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003,
4 2005, 2006, 2007, 2009 Free Software 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 <quote.h>
25 #include <quotearg.h>
26
27 #include "complain.h"
28 #include "conflicts.h"
29 #include "files.h"
30 #include "getargs.h"
31 #include "gram.h"
32 #include "muscle-tab.h"
33 #include "reader.h"
34 #include "symlist.h"
35 #include "symtab.h"
36 #include "scan-gram.h"
37 #include "scan-code.h"
38
39 static void check_and_convert_grammar (void);
40
41 static symbol_list *grammar = NULL;
42 static bool start_flag = false;
43 merger_list *merge_functions;
44
45 /* Was %union seen? */
46 bool union_seen = false;
47
48 /* Was a tag seen? */
49 bool tag_seen = false;
50
51 /* Should rules have a default precedence? */
52 bool default_prec = true;
53 \f
54 /*-----------------------.
55 | Set the start symbol. |
56 `-----------------------*/
57
58 void
59 grammar_start_symbol_set (symbol *sym, location loc)
60 {
61 if (start_flag)
62 complain_at (loc, _("multiple %s declarations"), "%start");
63 else
64 {
65 start_flag = true;
66 startsymbol = sym;
67 startsymbol_location = loc;
68 }
69 }
70
71 \f
72
73 /*------------------------------------------------------------------------.
74 | Return the merger index for a merging function named NAME. Records the |
75 | function, if new, in MERGER_LIST. |
76 `------------------------------------------------------------------------*/
77
78 static int
79 get_merge_function (uniqstr name)
80 {
81 merger_list *syms;
82 merger_list head;
83 int n;
84
85 if (! glr_parser)
86 return 0;
87
88 head.next = merge_functions;
89 for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
90 if (UNIQSTR_EQ (name, syms->next->name))
91 break;
92 if (syms->next == NULL)
93 {
94 syms->next = xmalloc (sizeof syms->next[0]);
95 syms->next->name = uniqstr_new (name);
96 /* After all symbol type declarations have been parsed, packgram invokes
97 record_merge_function_type to set the type. */
98 syms->next->type = NULL;
99 syms->next->next = NULL;
100 merge_functions = head.next;
101 }
102 return n;
103 }
104
105 /*-------------------------------------------------------------------------.
106 | For the existing merging function with index MERGER, record the result |
107 | type as TYPE as required by the lhs of the rule whose %merge declaration |
108 | is at DECLARATION_LOC. |
109 `-------------------------------------------------------------------------*/
110
111 static void
112 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
113 {
114 int merger_find;
115 merger_list *merge_function;
116
117 if (merger <= 0)
118 return;
119
120 if (type == NULL)
121 type = uniqstr_new ("");
122
123 merger_find = 1;
124 for (merge_function = merge_functions;
125 merge_function != NULL && merger_find != merger;
126 merge_function = merge_function->next)
127 merger_find += 1;
128 aver (merge_function != NULL && merger_find == merger);
129 if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
130 {
131 complain_at (declaration_loc,
132 _("result type clash on merge function `%s': <%s> != <%s>"),
133 merge_function->name, type, merge_function->type);
134 complain_at (merge_function->type_declaration_location,
135 _("previous declaration"));
136 }
137 merge_function->type = uniqstr_new (type);
138 merge_function->type_declaration_location = declaration_loc;
139 }
140
141 /*--------------------------------------.
142 | Free all merge-function definitions. |
143 `--------------------------------------*/
144
145 void
146 free_merger_functions (void)
147 {
148 merger_list *L0 = merge_functions;
149 while (L0)
150 {
151 merger_list *L1 = L0->next;
152 free (L0);
153 L0 = L1;
154 }
155 }
156
157 \f
158 /*-------------------------------------------------------------------.
159 | Parse the input grammar into a one symbol_list structure. Each |
160 | rule is represented by a sequence of symbols: the left hand side |
161 | followed by the contents of the right hand side, followed by a |
162 | null pointer instead of a symbol to terminate the rule. The next |
163 | symbol is the lhs of the following rule. |
164 | |
165 | All actions are copied out, labelled by the rule number they apply |
166 | to. |
167 `-------------------------------------------------------------------*/
168
169 /* The (currently) last symbol of GRAMMAR. */
170 static symbol_list *grammar_end = NULL;
171
172 /* Append SYM to the grammar. */
173 static symbol_list *
174 grammar_symbol_append (symbol *sym, location loc)
175 {
176 symbol_list *p = symbol_list_sym_new (sym, loc);
177
178 if (grammar_end)
179 grammar_end->next = p;
180 else
181 grammar = p;
182
183 grammar_end = p;
184
185 /* A null SYM stands for an end of rule; it is not an actual
186 part of it. */
187 if (sym)
188 ++nritems;
189
190 return p;
191 }
192
193 /* The rule currently being defined, and the previous rule.
194 CURRENT_RULE points to the first LHS of the current rule, while
195 PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
196 static symbol_list *current_rule = NULL;
197 static symbol_list *previous_rule_end = NULL;
198
199
200 /*----------------------------------------------.
201 | Create a new rule for LHS in to the GRAMMAR. |
202 `----------------------------------------------*/
203
204 void
205 grammar_current_rule_begin (symbol *lhs, location loc,
206 named_ref *lhs_named_ref)
207 {
208 symbol_list* p;
209
210 /* Start a new rule and record its lhs. */
211 ++nrules;
212 previous_rule_end = grammar_end;
213
214 p = grammar_symbol_append (lhs, loc);
215 if (lhs_named_ref)
216 {
217 if (lhs_named_ref->id == lhs->tag)
218 {
219 warn_at (lhs_named_ref->loc,
220 _("duplicated symbol name for %s ignored"),
221 quote (lhs->tag));
222 named_ref_free (lhs_named_ref);
223 }
224 else
225 p->named_ref = lhs_named_ref;
226 }
227
228 current_rule = grammar_end;
229
230 /* Mark the rule's lhs as a nonterminal if not already so. */
231 if (lhs->class == unknown_sym)
232 {
233 lhs->class = nterm_sym;
234 lhs->number = nvars;
235 ++nvars;
236 }
237 else if (lhs->class == token_sym)
238 complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
239 }
240
241
242 /*----------------------------------------------------------------------.
243 | A symbol should be used if either: |
244 | 1. It has a destructor. |
245 | 2. --warnings=midrule-values and the symbol is a mid-rule symbol |
246 | (i.e., the generated LHS replacing a mid-rule action) that was |
247 | assigned to or used, as in "exp: { $$ = 1; } { $$ = $1; }". |
248 `----------------------------------------------------------------------*/
249
250 static bool
251 symbol_should_be_used (symbol_list const *s)
252 {
253 if (symbol_destructor_get (s->content.sym)->code)
254 return true;
255 if (warnings_flag & warnings_midrule_values)
256 return ((s->midrule && s->midrule->action_props.is_value_used)
257 || (s->midrule_parent_rule
258 && symbol_list_n_get (s->midrule_parent_rule,
259 s->midrule_parent_rhs_index)
260 ->action_props.is_value_used));
261 return false;
262 }
263
264 /*----------------------------------------------------------------.
265 | Check that the rule R is properly defined. For instance, there |
266 | should be no type clash on the default action. |
267 `----------------------------------------------------------------*/
268
269 static void
270 grammar_rule_check (const symbol_list *r)
271 {
272 /* Type check.
273
274 If there is an action, then there is nothing we can do: the user
275 is allowed to shoot herself in the foot.
276
277 Don't worry about the default action if $$ is untyped, since $$'s
278 value can't be used. */
279 if (!r->action_props.code && r->content.sym->type_name)
280 {
281 symbol *first_rhs = r->next->content.sym;
282 /* If $$ is being set in default way, report if any type mismatch. */
283 if (first_rhs)
284 {
285 char const *lhs_type = r->content.sym->type_name;
286 const char *rhs_type =
287 first_rhs->type_name ? first_rhs->type_name : "";
288 if (!UNIQSTR_EQ (lhs_type, rhs_type))
289 warn_at (r->location,
290 _("type clash on default action: <%s> != <%s>"),
291 lhs_type, rhs_type);
292 }
293 /* Warn if there is no default for $$ but we need one. */
294 else
295 warn_at (r->location,
296 _("empty rule for typed nonterminal, and no action"));
297 }
298
299 /* Check that symbol values that should be used are in fact used. */
300 {
301 symbol_list const *l = r;
302 int n = 0;
303 for (; l && l->content.sym; l = l->next, ++n)
304 if (! (l->action_props.is_value_used
305 || !symbol_should_be_used (l)
306 /* The default action, $$ = $1, `uses' both. */
307 || (!r->action_props.code && (n == 0 || n == 1))))
308 {
309 if (n)
310 warn_at (r->location, _("unused value: $%d"), n);
311 else
312 warn_at (r->location, _("unset value: $$"));
313 }
314 }
315 }
316
317
318 /*-------------------------------------.
319 | End the currently being grown rule. |
320 `-------------------------------------*/
321
322 void
323 grammar_current_rule_end (location loc)
324 {
325 /* Put an empty link in the list to mark the end of this rule */
326 grammar_symbol_append (NULL, grammar_end->location);
327 current_rule->location = loc;
328 }
329
330
331 /*-------------------------------------------------------------------.
332 | The previous action turns out the be a mid-rule action. Attach it |
333 | to the current rule, i.e., create a dummy symbol, attach it this |
334 | mid-rule action, and append this dummy nonterminal to the current |
335 | rule. |
336 `-------------------------------------------------------------------*/
337
338 void
339 grammar_midrule_action (void)
340 {
341 /* Since the action was written out with this rule's number, we must
342 give the new rule this number by inserting the new rule before
343 it. */
344
345 /* Make a DUMMY nonterminal, whose location is that of the midrule
346 action. Create the MIDRULE. */
347 location dummy_location = current_rule->action_props.location;
348 symbol *dummy = dummy_symbol_get (dummy_location);
349 symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
350
351 /* Remember named_ref of previous action */
352 named_ref *named_ref = current_rule->action_props.named_ref;
353
354 /* Make a new rule, whose body is empty, before the current one, so
355 that the action just read can belong to it. */
356 ++nrules;
357 ++nritems;
358 /* Attach its location and actions to that of the DUMMY. */
359 midrule->location = dummy_location;
360 code_props_rule_action_init (&midrule->action_props,
361 current_rule->action_props.code,
362 current_rule->action_props.location,
363 midrule, 0);
364 code_props_none_init (&current_rule->action_props);
365
366 if (previous_rule_end)
367 previous_rule_end->next = midrule;
368 else
369 grammar = midrule;
370
371 /* End the dummy's rule. */
372 midrule->next = symbol_list_sym_new (NULL, dummy_location);
373 midrule->next->next = current_rule;
374
375 previous_rule_end = midrule->next;
376
377 /* Insert the dummy nonterminal replacing the midrule action into
378 the current rule. Bind it to its dedicated rule. */
379 grammar_current_rule_symbol_append (dummy, dummy_location, named_ref);
380 grammar_end->midrule = midrule;
381 midrule->midrule_parent_rule = current_rule;
382 midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
383 }
384
385 /* Set the precedence symbol of the current rule to PRECSYM. */
386
387 void
388 grammar_current_rule_prec_set (symbol *precsym, location loc)
389 {
390 symbol_class_set (precsym, token_sym, loc, false);
391 if (current_rule->ruleprec)
392 complain_at (loc, _("only one %s allowed per rule"), "%prec");
393 current_rule->ruleprec = precsym;
394 }
395
396 /* Attach dynamic precedence DPREC to the current rule. */
397
398 void
399 grammar_current_rule_dprec_set (int dprec, location loc)
400 {
401 if (! glr_parser)
402 warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
403 if (dprec <= 0)
404 complain_at (loc, _("%s must be followed by positive number"), "%dprec");
405 else if (current_rule->dprec != 0)
406 complain_at (loc, _("only one %s allowed per rule"), "%dprec");
407 current_rule->dprec = dprec;
408 }
409
410 /* Attach a merge function NAME with argument type TYPE to current
411 rule. */
412
413 void
414 grammar_current_rule_merge_set (uniqstr name, location loc)
415 {
416 if (! glr_parser)
417 warn_at (loc, _("%s affects only GLR parsers"), "%merge");
418 if (current_rule->merger != 0)
419 complain_at (loc, _("only one %s allowed per rule"), "%merge");
420 current_rule->merger = get_merge_function (name);
421 current_rule->merger_declaration_location = loc;
422 }
423
424 /* Attach SYM to the current rule. If needed, move the previous
425 action as a mid-rule action. */
426
427 void
428 grammar_current_rule_symbol_append (symbol *sym, location loc,
429 named_ref *named_ref)
430 {
431 symbol_list *p;
432 if (current_rule->action_props.code)
433 grammar_midrule_action ();
434 p = grammar_symbol_append (sym, loc);
435
436 if (named_ref)
437 {
438 if (named_ref->id == sym->tag)
439 {
440 warn_at (named_ref->loc,
441 _("duplicated symbol name for %s ignored"),
442 quote (sym->tag));
443 named_ref_free (named_ref);
444 }
445 else
446 p->named_ref = named_ref;
447 }
448 }
449
450 /* Attach an ACTION to the current rule. */
451
452 void
453 grammar_current_rule_action_append (const char *action, location loc,
454 named_ref *named_ref)
455 {
456 if (current_rule->action_props.code)
457 grammar_midrule_action ();
458 /* After all symbol declarations have been parsed, packgram invokes
459 code_props_translate_code. */
460 code_props_rule_action_init (&current_rule->action_props, action, loc,
461 current_rule, named_ref);
462 }
463
464 \f
465 /*---------------------------------------------------------------.
466 | Convert the rules into the representation using RRHS, RLHS and |
467 | RITEM. |
468 `---------------------------------------------------------------*/
469
470 static void
471 packgram (void)
472 {
473 unsigned int itemno = 0;
474 rule_number ruleno = 0;
475 symbol_list *p = grammar;
476
477 ritem = xnmalloc (nritems + 1, sizeof *ritem);
478
479 /* This sentinel is used by build_relations in gram.c. */
480 *ritem++ = 0;
481
482 rules = xnmalloc (nrules, sizeof *rules);
483
484 while (p)
485 {
486 int rule_length = 0;
487 symbol *ruleprec = p->ruleprec;
488 record_merge_function_type (p->merger, p->content.sym->type_name,
489 p->merger_declaration_location);
490 rules[ruleno].user_number = ruleno;
491 rules[ruleno].number = ruleno;
492 rules[ruleno].lhs = p->content.sym;
493 rules[ruleno].rhs = ritem + itemno;
494 rules[ruleno].prec = NULL;
495 rules[ruleno].dprec = p->dprec;
496 rules[ruleno].merger = p->merger;
497 rules[ruleno].precsym = NULL;
498 rules[ruleno].location = p->location;
499 rules[ruleno].useful = true;
500 rules[ruleno].action = p->action_props.code;
501 rules[ruleno].action_location = p->action_props.location;
502
503 /* If the midrule's $$ is set or its $n is used, remove the `$' from the
504 symbol name so that it's a user-defined symbol so that the default
505 %destructor and %printer apply. */
506 if (p->midrule_parent_rule
507 && (p->action_props.is_value_used
508 || symbol_list_n_get (p->midrule_parent_rule,
509 p->midrule_parent_rhs_index)
510 ->action_props.is_value_used))
511 p->content.sym->tag += 1;
512
513 /* Don't check the generated rule 0. It has no action, so some rhs
514 symbols may appear unused, but the parsing algorithm ensures that
515 %destructor's are invoked appropriately. */
516 if (p != grammar)
517 grammar_rule_check (p);
518
519 for (p = p->next; p && p->content.sym; p = p->next)
520 {
521 ++rule_length;
522
523 /* Don't allow rule_length == INT_MAX, since that might
524 cause confusion with strtol if INT_MAX == LONG_MAX. */
525 if (rule_length == INT_MAX)
526 fatal_at (rules[ruleno].location, _("rule is too long"));
527
528 /* item_number = symbol_number.
529 But the former needs to contain more: negative rule numbers. */
530 ritem[itemno++] =
531 symbol_number_as_item_number (p->content.sym->number);
532 /* A rule gets by default the precedence and associativity
533 of its last token. */
534 if (p->content.sym->class == token_sym && default_prec)
535 rules[ruleno].prec = p->content.sym;
536 }
537
538 /* If this rule has a %prec,
539 the specified symbol's precedence replaces the default. */
540 if (ruleprec)
541 {
542 rules[ruleno].precsym = ruleprec;
543 rules[ruleno].prec = ruleprec;
544 }
545 /* An item ends by the rule number (negated). */
546 ritem[itemno++] = rule_number_as_item_number (ruleno);
547 aver (itemno < ITEM_NUMBER_MAX);
548 ++ruleno;
549 aver (ruleno < RULE_NUMBER_MAX);
550
551 if (p)
552 p = p->next;
553 }
554
555 aver (itemno == nritems);
556
557 if (trace_flag & trace_sets)
558 ritem_print (stderr);
559 }
560 \f
561 /*------------------------------------------------------------------.
562 | Read in the grammar specification and record it in the format |
563 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
564 | in each case forming the body of a C function (YYACTION) which |
565 | contains a switch statement to decide which action to execute. |
566 `------------------------------------------------------------------*/
567
568 void
569 reader (void)
570 {
571 /* Initialize the symbol table. */
572 symbols_new ();
573
574 /* Construct the accept symbol. */
575 accept = symbol_get ("$accept", empty_location);
576 accept->class = nterm_sym;
577 accept->number = nvars++;
578
579 /* Construct the error token */
580 errtoken = symbol_get ("error", empty_location);
581 errtoken->class = token_sym;
582 errtoken->number = ntokens++;
583
584 /* Construct a token that represents all undefined literal tokens.
585 It is always token number 2. */
586 undeftoken = symbol_get ("$undefined", empty_location);
587 undeftoken->class = token_sym;
588 undeftoken->number = ntokens++;
589
590 gram_in = xfopen (grammar_file, "r");
591
592 gram__flex_debug = trace_flag & trace_scan;
593 gram_debug = trace_flag & trace_parse;
594 gram_scanner_initialize ();
595 gram_parse ();
596
597 /* Set front-end %define variable defaults. */
598 muscle_percent_define_default ("lr.keep-unreachable-states", "false");
599 {
600 char *lr_type;
601 /* IELR would be a better default, but LALR is historically the
602 default. */
603 muscle_percent_define_default ("lr.type", "LALR");
604 lr_type = muscle_percent_define_get ("lr.type");
605 if (0 != strcmp (lr_type, "canonical LR"))
606 muscle_percent_define_default ("lr.default-reductions", "all");
607 else
608 muscle_percent_define_default ("lr.default-reductions", "accepting");
609 free (lr_type);
610 }
611
612 /* Check front-end %define variables. */
613 {
614 static char const * const values[] = {
615 "lr.type", "LALR", "IELR", "canonical LR", NULL,
616 "lr.default-reductions", "all", "consistent", "accepting", NULL,
617 NULL
618 };
619 muscle_percent_define_check_values (values);
620 }
621
622 if (! complaint_issued)
623 check_and_convert_grammar ();
624
625 xfclose (gram_in);
626 }
627
628
629 /*-------------------------------------------------------------.
630 | Check the grammar that has just been read, and convert it to |
631 | internal form. |
632 `-------------------------------------------------------------*/
633
634 static void
635 check_and_convert_grammar (void)
636 {
637 /* Grammar has been read. Do some checking. */
638 if (nrules == 0)
639 fatal (_("no rules in the input grammar"));
640
641 /* Report any undefined symbols and consider them nonterminals. */
642 symbols_check_defined ();
643
644 /* If the user did not define her ENDTOKEN, do it now. */
645 if (!endtoken)
646 {
647 endtoken = symbol_get ("$end", empty_location);
648 endtoken->class = token_sym;
649 endtoken->number = 0;
650 /* Value specified by POSIX. */
651 endtoken->user_token_number = 0;
652 }
653
654 /* Find the start symbol if no %start. */
655 if (!start_flag)
656 {
657 symbol_list *node;
658 for (node = grammar;
659 node != NULL && symbol_is_dummy (node->content.sym);
660 node = node->next)
661 {
662 for (node = node->next;
663 node != NULL && node->content.sym != NULL;
664 node = node->next)
665 ;
666 }
667 aver (node != NULL);
668 grammar_start_symbol_set (node->content.sym,
669 node->content.sym->location);
670 }
671
672 /* Insert the initial rule, whose line is that of the first rule
673 (not that of the start symbol):
674
675 accept: %start EOF. */
676 {
677 symbol_list *p = symbol_list_sym_new (accept, empty_location);
678 p->location = grammar->location;
679 p->next = symbol_list_sym_new (startsymbol, empty_location);
680 p->next->next = symbol_list_sym_new (endtoken, empty_location);
681 p->next->next->next = symbol_list_sym_new (NULL, empty_location);
682 p->next->next->next->next = grammar;
683 nrules += 1;
684 nritems += 3;
685 grammar = p;
686 }
687
688 aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
689
690 /* Assign the symbols their symbol numbers. Write #defines for the
691 token symbols into FDEFINES if requested. */
692 symbols_pack ();
693
694 /* Scan rule actions after invoking symbol_check_alias_consistency (in
695 symbols_pack above) so that token types are set correctly before the rule
696 action type checking.
697
698 Before invoking grammar_rule_check (in packgram below) on any rule, make
699 sure all actions have already been scanned in order to set `used' flags.
700 Otherwise, checking that a midrule's $$ should be set will not always work
701 properly because the check must forward-reference the midrule's parent
702 rule. For the same reason, all the `used' flags must be set before
703 checking whether to remove `$' from any midrule symbol name (also in
704 packgram). */
705 {
706 symbol_list *sym;
707 for (sym = grammar; sym; sym = sym->next)
708 code_props_translate_code (&sym->action_props);
709 }
710
711 /* Convert the grammar into the format described in gram.h. */
712 packgram ();
713
714 /* The grammar as a symbol_list is no longer needed. */
715 symbol_list_free (grammar);
716 }