]> git.saurik.com Git - bison.git/blob - src/reader.c
14624df880b56ecd9d0d80425520e801e6cf5406
[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 prepare_percent_define_front_end_variables (void);
40 static void check_and_convert_grammar (void);
41
42 static symbol_list *grammar = NULL;
43 static bool start_flag = false;
44 merger_list *merge_functions;
45
46 /* Was %union seen? */
47 bool union_seen = false;
48
49 /* Was a tag seen? */
50 bool tag_seen = false;
51
52 /* Should rules have a default precedence? */
53 bool default_prec = true;
54 \f
55 /*-----------------------.
56 | Set the start symbol. |
57 `-----------------------*/
58
59 void
60 grammar_start_symbol_set (symbol *sym, location loc)
61 {
62 if (start_flag)
63 complain_at (loc, _("multiple %s declarations"), "%start");
64 else
65 {
66 start_flag = true;
67 startsymbol = sym;
68 startsymbol_location = loc;
69 }
70 }
71
72 \f
73
74 /*------------------------------------------------------------------------.
75 | Return the merger index for a merging function named NAME. Records the |
76 | function, if new, in MERGER_LIST. |
77 `------------------------------------------------------------------------*/
78
79 static int
80 get_merge_function (uniqstr name)
81 {
82 merger_list *syms;
83 merger_list head;
84 int n;
85
86 if (! glr_parser)
87 return 0;
88
89 head.next = merge_functions;
90 for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
91 if (UNIQSTR_EQ (name, syms->next->name))
92 break;
93 if (syms->next == NULL)
94 {
95 syms->next = xmalloc (sizeof syms->next[0]);
96 syms->next->name = uniqstr_new (name);
97 /* After all symbol type declarations have been parsed, packgram invokes
98 record_merge_function_type to set the type. */
99 syms->next->type = NULL;
100 syms->next->next = NULL;
101 merge_functions = head.next;
102 }
103 return n;
104 }
105
106 /*-------------------------------------------------------------------------.
107 | For the existing merging function with index MERGER, record the result |
108 | type as TYPE as required by the lhs of the rule whose %merge declaration |
109 | is at DECLARATION_LOC. |
110 `-------------------------------------------------------------------------*/
111
112 static void
113 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
114 {
115 int merger_find;
116 merger_list *merge_function;
117
118 if (merger <= 0)
119 return;
120
121 if (type == NULL)
122 type = uniqstr_new ("");
123
124 merger_find = 1;
125 for (merge_function = merge_functions;
126 merge_function != NULL && merger_find != merger;
127 merge_function = merge_function->next)
128 merger_find += 1;
129 aver (merge_function != NULL && merger_find == merger);
130 if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
131 {
132 complain_at (declaration_loc,
133 _("result type clash on merge function `%s': <%s> != <%s>"),
134 merge_function->name, type, merge_function->type);
135 complain_at (merge_function->type_declaration_location,
136 _("previous declaration"));
137 }
138 merge_function->type = uniqstr_new (type);
139 merge_function->type_declaration_location = declaration_loc;
140 }
141
142 /*--------------------------------------.
143 | Free all merge-function definitions. |
144 `--------------------------------------*/
145
146 void
147 free_merger_functions (void)
148 {
149 merger_list *L0 = merge_functions;
150 while (L0)
151 {
152 merger_list *L1 = L0->next;
153 free (L0);
154 L0 = L1;
155 }
156 }
157
158 \f
159 /*-------------------------------------------------------------------.
160 | Parse the input grammar into a one symbol_list structure. Each |
161 | rule is represented by a sequence of symbols: the left hand side |
162 | followed by the contents of the right hand side, followed by a |
163 | null pointer instead of a symbol to terminate the rule. The next |
164 | symbol is the lhs of the following rule. |
165 | |
166 | All actions are copied out, labelled by the rule number they apply |
167 | to. |
168 `-------------------------------------------------------------------*/
169
170 /* The (currently) last symbol of GRAMMAR. */
171 static symbol_list *grammar_end = NULL;
172
173 /* Append SYM to the grammar. */
174 static symbol_list *
175 grammar_symbol_append (symbol *sym, location loc)
176 {
177 symbol_list *p = symbol_list_sym_new (sym, loc);
178
179 if (grammar_end)
180 grammar_end->next = p;
181 else
182 grammar = p;
183
184 grammar_end = p;
185
186 /* A null SYM stands for an end of rule; it is not an actual
187 part of it. */
188 if (sym)
189 ++nritems;
190
191 return p;
192 }
193
194 static void
195 assign_named_ref (symbol_list *p, named_ref *name)
196 {
197 symbol *sym = p->content.sym;
198
199 if (name->id == sym->tag)
200 {
201 warn_at (name->loc,
202 _("duplicated symbol name for %s ignored"),
203 quote (sym->tag));
204 named_ref_free (name);
205 }
206 else
207 p->named_ref = name;
208 }
209
210
211 /* The rule currently being defined, and the previous rule.
212 CURRENT_RULE points to the first LHS of the current rule, while
213 PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
214 static symbol_list *current_rule = NULL;
215 static symbol_list *previous_rule_end = NULL;
216
217
218 /*----------------------------------------------.
219 | Create a new rule for LHS in to the GRAMMAR. |
220 `----------------------------------------------*/
221
222 void
223 grammar_current_rule_begin (symbol *lhs, location loc,
224 named_ref *lhs_name)
225 {
226 symbol_list* p;
227
228 /* Start a new rule and record its lhs. */
229 ++nrules;
230 previous_rule_end = grammar_end;
231
232 p = grammar_symbol_append (lhs, loc);
233 if (lhs_name)
234 assign_named_ref(p, lhs_name);
235
236 current_rule = grammar_end;
237
238 /* Mark the rule's lhs as a nonterminal if not already so. */
239 if (lhs->class == unknown_sym)
240 {
241 lhs->class = nterm_sym;
242 lhs->number = nvars;
243 ++nvars;
244 }
245 else if (lhs->class == token_sym)
246 complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
247 }
248
249
250 /*----------------------------------------------------------------------.
251 | A symbol should be used if either: |
252 | 1. It has a destructor. |
253 | 2. --warnings=midrule-values and the symbol is a mid-rule symbol |
254 | (i.e., the generated LHS replacing a mid-rule action) that was |
255 | assigned to or used, as in "exp: { $$ = 1; } { $$ = $1; }". |
256 `----------------------------------------------------------------------*/
257
258 static bool
259 symbol_should_be_used (symbol_list const *s)
260 {
261 if (symbol_destructor_get (s->content.sym)->code)
262 return true;
263 if (warnings_flag & warnings_midrule_values)
264 return ((s->midrule && s->midrule->action_props.is_value_used)
265 || (s->midrule_parent_rule
266 && symbol_list_n_get (s->midrule_parent_rule,
267 s->midrule_parent_rhs_index)
268 ->action_props.is_value_used));
269 return false;
270 }
271
272 /*----------------------------------------------------------------.
273 | Check that the rule R is properly defined. For instance, there |
274 | should be no type clash on the default action. |
275 `----------------------------------------------------------------*/
276
277 static void
278 grammar_rule_check (const symbol_list *r)
279 {
280 /* Type check.
281
282 If there is an action, then there is nothing we can do: the user
283 is allowed to shoot herself in the foot.
284
285 Don't worry about the default action if $$ is untyped, since $$'s
286 value can't be used. */
287 if (!r->action_props.code && r->content.sym->type_name)
288 {
289 symbol *first_rhs = r->next->content.sym;
290 /* If $$ is being set in default way, report if any type mismatch. */
291 if (first_rhs)
292 {
293 char const *lhs_type = r->content.sym->type_name;
294 const char *rhs_type =
295 first_rhs->type_name ? first_rhs->type_name : "";
296 if (!UNIQSTR_EQ (lhs_type, rhs_type))
297 warn_at (r->location,
298 _("type clash on default action: <%s> != <%s>"),
299 lhs_type, rhs_type);
300 }
301 /* Warn if there is no default for $$ but we need one. */
302 else
303 warn_at (r->location,
304 _("empty rule for typed nonterminal, and no action"));
305 }
306
307 /* Check that symbol values that should be used are in fact used. */
308 {
309 symbol_list const *l = r;
310 int n = 0;
311 for (; l && l->content.sym; l = l->next, ++n)
312 if (! (l->action_props.is_value_used
313 || !symbol_should_be_used (l)
314 /* The default action, $$ = $1, `uses' both. */
315 || (!r->action_props.code && (n == 0 || n == 1))))
316 {
317 if (n)
318 warn_at (r->location, _("unused value: $%d"), n);
319 else
320 warn_at (r->location, _("unset value: $$"));
321 }
322 }
323 }
324
325
326 /*-------------------------------------.
327 | End the currently being grown rule. |
328 `-------------------------------------*/
329
330 void
331 grammar_current_rule_end (location loc)
332 {
333 /* Put an empty link in the list to mark the end of this rule */
334 grammar_symbol_append (NULL, grammar_end->location);
335 current_rule->location = loc;
336 }
337
338
339 /*-------------------------------------------------------------------.
340 | The previous action turns out the be a mid-rule action. Attach it |
341 | to the current rule, i.e., create a dummy symbol, attach it this |
342 | mid-rule action, and append this dummy nonterminal to the current |
343 | rule. |
344 `-------------------------------------------------------------------*/
345
346 void
347 grammar_midrule_action (void)
348 {
349 /* Since the action was written out with this rule's number, we must
350 give the new rule this number by inserting the new rule before
351 it. */
352
353 /* Make a DUMMY nonterminal, whose location is that of the midrule
354 action. Create the MIDRULE. */
355 location dummy_location = current_rule->action_props.location;
356 symbol *dummy = dummy_symbol_get (dummy_location);
357 symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
358
359 /* Remember named_ref of previous action. */
360 named_ref *action_name = current_rule->action_props.named_ref;
361
362 /* Make a new rule, whose body is empty, before the current one, so
363 that the action just read can belong to it. */
364 ++nrules;
365 ++nritems;
366 /* Attach its location and actions to that of the DUMMY. */
367 midrule->location = dummy_location;
368 code_props_rule_action_init (&midrule->action_props,
369 current_rule->action_props.code,
370 current_rule->action_props.location,
371 midrule, 0);
372 code_props_none_init (&current_rule->action_props);
373
374 if (previous_rule_end)
375 previous_rule_end->next = midrule;
376 else
377 grammar = midrule;
378
379 /* End the dummy's rule. */
380 midrule->next = symbol_list_sym_new (NULL, dummy_location);
381 midrule->next->next = current_rule;
382
383 previous_rule_end = midrule->next;
384
385 /* Insert the dummy nonterminal replacing the midrule action into
386 the current rule. Bind it to its dedicated rule. */
387 grammar_current_rule_symbol_append (dummy, dummy_location,
388 action_name);
389 grammar_end->midrule = midrule;
390 midrule->midrule_parent_rule = current_rule;
391 midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
392 }
393
394 /* Set the precedence symbol of the current rule to PRECSYM. */
395
396 void
397 grammar_current_rule_prec_set (symbol *precsym, location loc)
398 {
399 symbol_class_set (precsym, token_sym, loc, false);
400 if (current_rule->ruleprec)
401 complain_at (loc, _("only one %s allowed per rule"), "%prec");
402 current_rule->ruleprec = precsym;
403 }
404
405 /* Attach dynamic precedence DPREC to the current rule. */
406
407 void
408 grammar_current_rule_dprec_set (int dprec, location loc)
409 {
410 if (! glr_parser)
411 warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
412 if (dprec <= 0)
413 complain_at (loc, _("%s must be followed by positive number"), "%dprec");
414 else if (current_rule->dprec != 0)
415 complain_at (loc, _("only one %s allowed per rule"), "%dprec");
416 current_rule->dprec = dprec;
417 }
418
419 /* Attach a merge function NAME with argument type TYPE to current
420 rule. */
421
422 void
423 grammar_current_rule_merge_set (uniqstr name, location loc)
424 {
425 if (! glr_parser)
426 warn_at (loc, _("%s affects only GLR parsers"), "%merge");
427 if (current_rule->merger != 0)
428 complain_at (loc, _("only one %s allowed per rule"), "%merge");
429 current_rule->merger = get_merge_function (name);
430 current_rule->merger_declaration_location = loc;
431 }
432
433 /* Attach SYM to the current rule. If needed, move the previous
434 action as a mid-rule action. */
435
436 void
437 grammar_current_rule_symbol_append (symbol *sym, location loc,
438 named_ref *name)
439 {
440 symbol_list *p;
441 if (current_rule->action_props.code)
442 grammar_midrule_action ();
443 p = grammar_symbol_append (sym, loc);
444 if (name)
445 assign_named_ref(p, name);
446 }
447
448 /* Attach an ACTION to the current rule. */
449
450 void
451 grammar_current_rule_action_append (const char *action, location loc,
452 named_ref *name)
453 {
454 if (current_rule->action_props.code)
455 grammar_midrule_action ();
456 /* After all symbol declarations have been parsed, packgram invokes
457 code_props_translate_code. */
458 code_props_rule_action_init (&current_rule->action_props, action, loc,
459 current_rule, name);
460 }
461
462 \f
463 /*---------------------------------------------------------------.
464 | Convert the rules into the representation using RRHS, RLHS and |
465 | RITEM. |
466 `---------------------------------------------------------------*/
467
468 static void
469 packgram (void)
470 {
471 unsigned int itemno = 0;
472 rule_number ruleno = 0;
473 symbol_list *p = grammar;
474
475 ritem = xnmalloc (nritems + 1, sizeof *ritem);
476
477 /* This sentinel is used by build_relations in gram.c. */
478 *ritem++ = 0;
479
480 rules = xnmalloc (nrules, sizeof *rules);
481
482 while (p)
483 {
484 int rule_length = 0;
485 symbol *ruleprec = p->ruleprec;
486 record_merge_function_type (p->merger, p->content.sym->type_name,
487 p->merger_declaration_location);
488 rules[ruleno].user_number = ruleno;
489 rules[ruleno].number = ruleno;
490 rules[ruleno].lhs = p->content.sym;
491 rules[ruleno].rhs = ritem + itemno;
492 rules[ruleno].prec = NULL;
493 rules[ruleno].dprec = p->dprec;
494 rules[ruleno].merger = p->merger;
495 rules[ruleno].precsym = NULL;
496 rules[ruleno].location = p->location;
497 rules[ruleno].useful = true;
498 rules[ruleno].action = p->action_props.code;
499 rules[ruleno].action_location = p->action_props.location;
500
501 /* If the midrule's $$ is set or its $n is used, remove the `$' from the
502 symbol name so that it's a user-defined symbol so that the default
503 %destructor and %printer apply. */
504 if (p->midrule_parent_rule
505 && (p->action_props.is_value_used
506 || symbol_list_n_get (p->midrule_parent_rule,
507 p->midrule_parent_rhs_index)
508 ->action_props.is_value_used))
509 p->content.sym->tag += 1;
510
511 /* Don't check the generated rule 0. It has no action, so some rhs
512 symbols may appear unused, but the parsing algorithm ensures that
513 %destructor's are invoked appropriately. */
514 if (p != grammar)
515 grammar_rule_check (p);
516
517 for (p = p->next; p && p->content.sym; p = p->next)
518 {
519 ++rule_length;
520
521 /* Don't allow rule_length == INT_MAX, since that might
522 cause confusion with strtol if INT_MAX == LONG_MAX. */
523 if (rule_length == INT_MAX)
524 fatal_at (rules[ruleno].location, _("rule is too long"));
525
526 /* item_number = symbol_number.
527 But the former needs to contain more: negative rule numbers. */
528 ritem[itemno++] =
529 symbol_number_as_item_number (p->content.sym->number);
530 /* A rule gets by default the precedence and associativity
531 of its last token. */
532 if (p->content.sym->class == token_sym && default_prec)
533 rules[ruleno].prec = p->content.sym;
534 }
535
536 /* If this rule has a %prec,
537 the specified symbol's precedence replaces the default. */
538 if (ruleprec)
539 {
540 rules[ruleno].precsym = ruleprec;
541 rules[ruleno].prec = ruleprec;
542 }
543 /* An item ends by the rule number (negated). */
544 ritem[itemno++] = rule_number_as_item_number (ruleno);
545 aver (itemno < ITEM_NUMBER_MAX);
546 ++ruleno;
547 aver (ruleno < RULE_NUMBER_MAX);
548
549 if (p)
550 p = p->next;
551 }
552
553 aver (itemno == nritems);
554
555 if (trace_flag & trace_sets)
556 ritem_print (stderr);
557 }
558 \f
559 /*------------------------------------------------------------------.
560 | Read in the grammar specification and record it in the format |
561 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
562 | in each case forming the body of a C function (YYACTION) which |
563 | contains a switch statement to decide which action to execute. |
564 `------------------------------------------------------------------*/
565
566 void
567 reader (void)
568 {
569 /* Initialize the symbol table. */
570 symbols_new ();
571
572 /* Construct the accept symbol. */
573 accept = symbol_get ("$accept", empty_location);
574 accept->class = nterm_sym;
575 accept->number = nvars++;
576
577 /* Construct the error token */
578 errtoken = symbol_get ("error", empty_location);
579 errtoken->class = token_sym;
580 errtoken->number = ntokens++;
581
582 /* Construct a token that represents all undefined literal tokens.
583 It is always token number 2. */
584 undeftoken = symbol_get ("$undefined", empty_location);
585 undeftoken->class = token_sym;
586 undeftoken->number = ntokens++;
587
588 gram_in = xfopen (grammar_file, "r");
589
590 gram__flex_debug = trace_flag & trace_scan;
591 gram_debug = trace_flag & trace_parse;
592 gram_scanner_initialize ();
593 gram_parse ();
594 prepare_percent_define_front_end_variables ();
595
596 if (! complaint_issued)
597 check_and_convert_grammar ();
598
599 xfclose (gram_in);
600 }
601
602 static void
603 prepare_percent_define_front_end_variables (void)
604 {
605 /* Set %define front-end variable defaults. */
606 muscle_percent_define_default ("lr.keep-unreachable-states", "false");
607 {
608 char *lr_type;
609 /* IELR would be a better default, but LALR is historically the
610 default. */
611 muscle_percent_define_default ("lr.type", "LALR");
612 lr_type = muscle_percent_define_get ("lr.type");
613 if (0 != strcmp (lr_type, "canonical LR"))
614 muscle_percent_define_default ("lr.default-reductions", "all");
615 else
616 muscle_percent_define_default ("lr.default-reductions", "accepting");
617 free (lr_type);
618 }
619
620 /* Check %define front-end variables. */
621 {
622 static char const * const values[] = {
623 "lr.type", "LALR", "IELR", "canonical LR", NULL,
624 "lr.default-reductions", "all", "consistent", "accepting", NULL,
625 NULL
626 };
627 muscle_percent_define_check_values (values);
628 }
629 }
630
631
632 /*-------------------------------------------------------------.
633 | Check the grammar that has just been read, and convert it to |
634 | internal form. |
635 `-------------------------------------------------------------*/
636
637 static void
638 check_and_convert_grammar (void)
639 {
640 /* Grammar has been read. Do some checking. */
641 if (nrules == 0)
642 fatal (_("no rules in the input grammar"));
643
644 /* Report any undefined symbols and consider them nonterminals. */
645 symbols_check_defined ();
646
647 /* If the user did not define her ENDTOKEN, do it now. */
648 if (!endtoken)
649 {
650 endtoken = symbol_get ("$end", empty_location);
651 endtoken->class = token_sym;
652 endtoken->number = 0;
653 /* Value specified by POSIX. */
654 endtoken->user_token_number = 0;
655 }
656
657 /* Find the start symbol if no %start. */
658 if (!start_flag)
659 {
660 symbol_list *node;
661 for (node = grammar;
662 node != NULL && symbol_is_dummy (node->content.sym);
663 node = node->next)
664 {
665 for (node = node->next;
666 node != NULL && node->content.sym != NULL;
667 node = node->next)
668 ;
669 }
670 aver (node != NULL);
671 grammar_start_symbol_set (node->content.sym,
672 node->content.sym->location);
673 }
674
675 /* Insert the initial rule, whose line is that of the first rule
676 (not that of the start symbol):
677
678 accept: %start EOF. */
679 {
680 symbol_list *p = symbol_list_sym_new (accept, empty_location);
681 p->location = grammar->location;
682 p->next = symbol_list_sym_new (startsymbol, empty_location);
683 p->next->next = symbol_list_sym_new (endtoken, empty_location);
684 p->next->next->next = symbol_list_sym_new (NULL, empty_location);
685 p->next->next->next->next = grammar;
686 nrules += 1;
687 nritems += 3;
688 grammar = p;
689 }
690
691 aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
692
693 /* Assign the symbols their symbol numbers. Write #defines for the
694 token symbols into FDEFINES if requested. */
695 symbols_pack ();
696
697 /* Scan rule actions after invoking symbol_check_alias_consistency (in
698 symbols_pack above) so that token types are set correctly before the rule
699 action type checking.
700
701 Before invoking grammar_rule_check (in packgram below) on any rule, make
702 sure all actions have already been scanned in order to set `used' flags.
703 Otherwise, checking that a midrule's $$ should be set will not always work
704 properly because the check must forward-reference the midrule's parent
705 rule. For the same reason, all the `used' flags must be set before
706 checking whether to remove `$' from any midrule symbol name (also in
707 packgram). */
708 {
709 symbol_list *sym;
710 for (sym = grammar; sym; sym = sym->next)
711 code_props_translate_code (&sym->action_props);
712 }
713
714 /* Convert the grammar into the format described in gram.h. */
715 packgram ();
716
717 /* The grammar as a symbol_list is no longer needed. */
718 symbol_list_free (grammar);
719 }