1 /* Input parser for Bison
3 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007,
4 2009-2012 Free Software Foundation, Inc.
6 This file is part of Bison, the GNU Compiler Compiler.
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.
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.
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/>. */
27 #include "conflicts.h"
31 #include "muscle-tab.h"
35 #include "scan-gram.h"
36 #include "scan-code.h"
38 static void prepare_percent_define_front_end_variables (void);
39 static void check_and_convert_grammar (void);
41 static symbol_list
*grammar
= NULL
;
42 static bool start_flag
= false;
43 merger_list
*merge_functions
;
45 /* Was %union seen? */
46 bool union_seen
= false;
49 bool tag_seen
= false;
51 /* Should rules have a default precedence? */
52 bool default_prec
= true;
54 /*-----------------------.
55 | Set the start symbol. |
56 `-----------------------*/
59 grammar_start_symbol_set (symbol
*sym
, location loc
)
62 complain_at (loc
, _("multiple %s declarations"), "%start");
67 startsymbol_location
= loc
;
73 /*------------------------------------------------------------------------.
74 | Return the merger index for a merging function named NAME. Records the |
75 | function, if new, in MERGER_LIST. |
76 `------------------------------------------------------------------------*/
79 get_merge_function (uniqstr name
)
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
))
92 if (syms
->next
== NULL
)
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
;
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 `-------------------------------------------------------------------------*/
112 record_merge_function_type (int merger
, uniqstr type
, location declaration_loc
)
115 merger_list
*merge_function
;
121 type
= uniqstr_new ("");
124 for (merge_function
= merge_functions
;
125 merge_function
!= NULL
&& merger_find
!= merger
;
126 merge_function
= merge_function
->next
)
128 aver (merge_function
!= NULL
&& merger_find
== merger
);
129 if (merge_function
->type
!= NULL
&& !UNIQSTR_EQ (merge_function
->type
, type
))
131 complain_at (declaration_loc
,
132 _("result type clash on merge function %s: <%s> != <%s>"),
133 quote (merge_function
->name
), type
, merge_function
->type
);
134 complain_at (merge_function
->type_declaration_location
,
135 _("previous declaration"));
137 merge_function
->type
= uniqstr_new (type
);
138 merge_function
->type_declaration_location
= declaration_loc
;
141 /*--------------------------------------.
142 | Free all merge-function definitions. |
143 `--------------------------------------*/
146 free_merger_functions (void)
148 merger_list
*L0
= merge_functions
;
151 merger_list
*L1
= L0
->next
;
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. |
165 | All actions are copied out, labelled by the rule number they apply |
167 `-------------------------------------------------------------------*/
169 /* The (currently) last symbol of GRAMMAR. */
170 static symbol_list
*grammar_end
= NULL
;
172 /* Append SYM to the grammar. */
174 grammar_symbol_append (symbol
*sym
, location loc
)
176 symbol_list
*p
= symbol_list_sym_new (sym
, loc
);
179 grammar_end
->next
= p
;
185 /* A null SYM stands for an end of rule; it is not an actual
194 assign_named_ref (symbol_list
*p
, named_ref
*name
)
196 symbol
*sym
= p
->content
.sym
;
198 if (name
->id
== sym
->tag
)
201 _("duplicated symbol name for %s ignored"),
203 named_ref_free (name
);
210 /* The rule currently being defined, and the previous rule.
211 CURRENT_RULE points to the first LHS of the current rule, while
212 PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
213 static symbol_list
*current_rule
= NULL
;
214 static symbol_list
*previous_rule_end
= NULL
;
217 /*----------------------------------------------.
218 | Create a new rule for LHS in to the GRAMMAR. |
219 `----------------------------------------------*/
222 grammar_current_rule_begin (symbol
*lhs
, location loc
,
227 /* Start a new rule and record its lhs. */
229 previous_rule_end
= grammar_end
;
231 p
= grammar_symbol_append (lhs
, loc
);
233 assign_named_ref (p
, named_ref_copy (lhs_name
));
235 current_rule
= grammar_end
;
237 /* Mark the rule's lhs as a nonterminal if not already so. */
238 if (lhs
->class == unknown_sym
)
240 lhs
->class = nterm_sym
;
244 else if (lhs
->class == token_sym
)
245 complain_at (loc
, _("rule given for %s, which is a token"), lhs
->tag
);
249 /*----------------------------------------------------------------------.
250 | A symbol should be used if either: |
251 | 1. It has a destructor. |
252 | 2. The symbol is a mid-rule symbol (i.e., the generated LHS |
253 | replacing a mid-rule action) that was assigned to or used, as in |
254 | "exp: { $$ = 1; } { $$ = $1; }". |
255 `----------------------------------------------------------------------*/
258 symbol_should_be_used (symbol_list
const *s
, bool *midrule_warning
)
260 if (symbol_destructor_get (s
->content
.sym
)->code
)
262 if ((s
->midrule
&& s
->midrule
->action_props
.is_value_used
)
263 || (s
->midrule_parent_rule
264 && symbol_list_n_get (s
->midrule_parent_rule
,
265 s
->midrule_parent_rhs_index
)
266 ->action_props
.is_value_used
))
268 *midrule_warning
= true;
274 /*----------------------------------------------------------------.
275 | Check that the rule R is properly defined. For instance, there |
276 | should be no type clash on the default action. |
277 `----------------------------------------------------------------*/
280 grammar_rule_check (const symbol_list
*r
)
284 If there is an action, then there is nothing we can do: the user
285 is allowed to shoot herself in the foot.
287 Don't worry about the default action if $$ is untyped, since $$'s
288 value can't be used. */
289 if (!r
->action_props
.code
&& r
->content
.sym
->type_name
)
291 symbol
*first_rhs
= r
->next
->content
.sym
;
292 /* If $$ is being set in default way, report if any type mismatch. */
295 char const *lhs_type
= r
->content
.sym
->type_name
;
296 const char *rhs_type
=
297 first_rhs
->type_name
? first_rhs
->type_name
: "";
298 if (!UNIQSTR_EQ (lhs_type
, rhs_type
))
299 warn_at (r
->location
,
300 _("type clash on default action: <%s> != <%s>"),
303 /* Warn if there is no default for $$ but we need one. */
305 warn_at (r
->location
,
306 _("empty rule for typed nonterminal, and no action"));
309 /* Check that symbol values that should be used are in fact used. */
311 symbol_list
const *l
= r
;
313 for (; l
&& l
->content
.sym
; l
= l
->next
, ++n
)
315 bool midrule_warning
= false;
316 if (!l
->action_props
.is_value_used
317 && symbol_should_be_used (l
, &midrule_warning
)
318 /* The default action, $$ = $1, `uses' both. */
319 && (r
->action_props
.code
|| (n
!= 0 && n
!= 1)))
321 void (*warn_at_ptr
)(location
, char const*, ...) =
322 midrule_warning
? midrule_value_at
: warn_at
;
324 warn_at_ptr (r
->location
, _("unused value: $%d"), n
);
326 warn_at_ptr (r
->location
, _("unset value: $$"));
331 /* See comments in grammar_current_rule_prec_set for how POSIX
332 mandates this complaint. It's only for identifiers, so skip
333 it for char literals and strings, which are always tokens. */
335 && r
->ruleprec
->tag
[0] != '\'' && r
->ruleprec
->tag
[0] != '"'
336 && !r
->ruleprec
->declared
&& !r
->ruleprec
->prec
)
337 warn_at (r
->location
, _("token for %%prec is not defined: %s"),
342 /*-------------------------------------.
343 | End the currently being grown rule. |
344 `-------------------------------------*/
347 grammar_current_rule_end (location loc
)
349 /* Put an empty link in the list to mark the end of this rule */
350 grammar_symbol_append (NULL
, grammar_end
->location
);
351 current_rule
->location
= loc
;
355 /*-------------------------------------------------------------------.
356 | The previous action turns out the be a mid-rule action. Attach it |
357 | to the current rule, i.e., create a dummy symbol, attach it this |
358 | mid-rule action, and append this dummy nonterminal to the current |
360 `-------------------------------------------------------------------*/
363 grammar_midrule_action (void)
365 /* Since the action was written out with this rule's number, we must
366 give the new rule this number by inserting the new rule before
369 /* Make a DUMMY nonterminal, whose location is that of the midrule
370 action. Create the MIDRULE. */
371 location dummy_location
= current_rule
->action_props
.location
;
372 symbol
*dummy
= dummy_symbol_get (dummy_location
);
373 symbol_list
*midrule
= symbol_list_sym_new (dummy
, dummy_location
);
375 /* Remember named_ref of previous action. */
376 named_ref
*action_name
= current_rule
->action_props
.named_ref
;
378 /* Make a new rule, whose body is empty, before the current one, so
379 that the action just read can belong to it. */
382 /* Attach its location and actions to that of the DUMMY. */
383 midrule
->location
= dummy_location
;
384 code_props_rule_action_init (&midrule
->action_props
,
385 current_rule
->action_props
.code
,
386 current_rule
->action_props
.location
,
388 current_rule
->action_props
.is_predicate
);
389 code_props_none_init (¤t_rule
->action_props
);
391 if (previous_rule_end
)
392 previous_rule_end
->next
= midrule
;
396 /* End the dummy's rule. */
397 midrule
->next
= symbol_list_sym_new (NULL
, dummy_location
);
398 midrule
->next
->next
= current_rule
;
400 previous_rule_end
= midrule
->next
;
402 /* Insert the dummy nonterminal replacing the midrule action into
403 the current rule. Bind it to its dedicated rule. */
404 grammar_current_rule_symbol_append (dummy
, dummy_location
,
406 grammar_end
->midrule
= midrule
;
407 midrule
->midrule_parent_rule
= current_rule
;
408 midrule
->midrule_parent_rhs_index
= symbol_list_length (current_rule
->next
);
411 /* Set the precedence symbol of the current rule to PRECSYM. */
414 grammar_current_rule_prec_set (symbol
*precsym
, location loc
)
416 /* POSIX says that any identifier is a nonterminal if it does not
417 appear on the LHS of a grammar rule and is not defined by %token
418 or by one of the directives that assigns precedence to a token. We
419 ignore this here because the only kind of identifier that POSIX
420 allows to follow a %prec is a token and because assuming it's a
421 token now can produce more logical error messages. Nevertheless,
422 grammar_rule_check does obey what we believe is the real intent of
423 POSIX here: that an error be reported for any identifier that
424 appears after %prec but that is not defined separately as a
426 symbol_class_set (precsym
, token_sym
, loc
, false);
427 if (current_rule
->ruleprec
)
428 complain_at (loc
, _("only one %s allowed per rule"), "%prec");
429 current_rule
->ruleprec
= precsym
;
432 /* Attach dynamic precedence DPREC to the current rule. */
435 grammar_current_rule_dprec_set (int dprec
, location loc
)
438 warn_at (loc
, _("%s affects only GLR parsers"), "%dprec");
440 complain_at (loc
, _("%s must be followed by positive number"), "%dprec");
441 else if (current_rule
->dprec
!= 0)
442 complain_at (loc
, _("only one %s allowed per rule"), "%dprec");
443 current_rule
->dprec
= dprec
;
446 /* Attach a merge function NAME with argument type TYPE to current
450 grammar_current_rule_merge_set (uniqstr name
, location loc
)
453 warn_at (loc
, _("%s affects only GLR parsers"), "%merge");
454 if (current_rule
->merger
!= 0)
455 complain_at (loc
, _("only one %s allowed per rule"), "%merge");
456 current_rule
->merger
= get_merge_function (name
);
457 current_rule
->merger_declaration_location
= loc
;
460 /* Attach SYM to the current rule. If needed, move the previous
461 action as a mid-rule action. */
464 grammar_current_rule_symbol_append (symbol
*sym
, location loc
,
468 if (current_rule
->action_props
.code
)
469 grammar_midrule_action ();
470 p
= grammar_symbol_append (sym
, loc
);
472 assign_named_ref(p
, name
);
475 /* Attach an ACTION to the current rule. */
478 grammar_current_rule_action_append (const char *action
, location loc
,
479 named_ref
*name
, bool is_predicate
)
481 if (current_rule
->action_props
.code
)
482 grammar_midrule_action ();
483 /* After all symbol declarations have been parsed, packgram invokes
484 code_props_translate_code. */
485 code_props_rule_action_init (¤t_rule
->action_props
, action
, loc
,
486 current_rule
, name
, is_predicate
);
490 /*---------------------------------------------------------------.
491 | Convert the rules into the representation using RRHS, RLHS and |
493 `---------------------------------------------------------------*/
498 unsigned int itemno
= 0;
499 rule_number ruleno
= 0;
500 symbol_list
*p
= grammar
;
502 ritem
= xnmalloc (nritems
+ 1, sizeof *ritem
);
504 /* This sentinel is used by build_relations in gram.c. */
507 rules
= xnmalloc (nrules
, sizeof *rules
);
512 symbol
*ruleprec
= p
->ruleprec
;
513 record_merge_function_type (p
->merger
, p
->content
.sym
->type_name
,
514 p
->merger_declaration_location
);
515 rules
[ruleno
].user_number
= ruleno
;
516 rules
[ruleno
].number
= ruleno
;
517 rules
[ruleno
].lhs
= p
->content
.sym
;
518 rules
[ruleno
].rhs
= ritem
+ itemno
;
519 rules
[ruleno
].prec
= NULL
;
520 rules
[ruleno
].dprec
= p
->dprec
;
521 rules
[ruleno
].merger
= p
->merger
;
522 rules
[ruleno
].precsym
= NULL
;
523 rules
[ruleno
].location
= p
->location
;
524 rules
[ruleno
].useful
= true;
525 rules
[ruleno
].action
= p
->action_props
.code
;
526 rules
[ruleno
].action_location
= p
->action_props
.location
;
527 rules
[ruleno
].is_predicate
= p
->action_props
.is_predicate
;
529 /* If the midrule's $$ is set or its $n is used, remove the `$' from the
530 symbol name so that it's a user-defined symbol so that the default
531 %destructor and %printer apply. */
532 if (p
->midrule_parent_rule
533 && (p
->action_props
.is_value_used
534 || symbol_list_n_get (p
->midrule_parent_rule
,
535 p
->midrule_parent_rhs_index
)
536 ->action_props
.is_value_used
))
537 p
->content
.sym
->tag
+= 1;
539 /* Don't check the generated rule 0. It has no action, so some rhs
540 symbols may appear unused, but the parsing algorithm ensures that
541 %destructor's are invoked appropriately. */
543 grammar_rule_check (p
);
545 for (p
= p
->next
; p
&& p
->content
.sym
; p
= p
->next
)
549 /* Don't allow rule_length == INT_MAX, since that might
550 cause confusion with strtol if INT_MAX == LONG_MAX. */
551 if (rule_length
== INT_MAX
)
552 fatal_at (rules
[ruleno
].location
, _("rule is too long"));
554 /* item_number = symbol_number.
555 But the former needs to contain more: negative rule numbers. */
557 symbol_number_as_item_number (p
->content
.sym
->number
);
558 /* A rule gets by default the precedence and associativity
559 of its last token. */
560 if (p
->content
.sym
->class == token_sym
&& default_prec
)
561 rules
[ruleno
].prec
= p
->content
.sym
;
564 /* If this rule has a %prec,
565 the specified symbol's precedence replaces the default. */
568 rules
[ruleno
].precsym
= ruleprec
;
569 rules
[ruleno
].prec
= ruleprec
;
571 /* An item ends by the rule number (negated). */
572 ritem
[itemno
++] = rule_number_as_item_number (ruleno
);
573 aver (itemno
< ITEM_NUMBER_MAX
);
575 aver (ruleno
< RULE_NUMBER_MAX
);
581 aver (itemno
== nritems
);
583 if (trace_flag
& trace_sets
)
584 ritem_print (stderr
);
587 /*------------------------------------------------------------------.
588 | Read in the grammar specification and record it in the format |
589 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
590 | in each case forming the body of a C function (YYACTION) which |
591 | contains a switch statement to decide which action to execute. |
592 `------------------------------------------------------------------*/
597 /* Initialize the symbol table. */
600 /* Construct the accept symbol. */
601 accept
= symbol_get ("$accept", empty_location
);
602 accept
->class = nterm_sym
;
603 accept
->number
= nvars
++;
605 /* Construct the error token */
606 errtoken
= symbol_get ("error", empty_location
);
607 errtoken
->class = token_sym
;
608 errtoken
->number
= ntokens
++;
610 /* Construct a token that represents all undefined literal tokens.
611 It is always token number 2. */
612 undeftoken
= symbol_get ("$undefined", empty_location
);
613 undeftoken
->class = token_sym
;
614 undeftoken
->number
= ntokens
++;
616 gram_in
= xfopen (grammar_file
, "r");
618 gram__flex_debug
= trace_flag
& trace_scan
;
619 gram_debug
= trace_flag
& trace_parse
;
620 gram_scanner_initialize ();
622 prepare_percent_define_front_end_variables ();
624 if (! complaint_issued
)
625 check_and_convert_grammar ();
631 prepare_percent_define_front_end_variables (void)
633 /* Set %define front-end variable defaults. */
634 muscle_percent_define_default ("lr.keep-unreachable-states", "false");
637 /* IELR would be a better default, but LALR is historically the
639 muscle_percent_define_default ("lr.type", "lalr");
640 lr_type
= muscle_percent_define_get ("lr.type");
641 if (STRNEQ (lr_type
, "canonical-lr"))
642 muscle_percent_define_default ("lr.default-reductions", "most");
644 muscle_percent_define_default ("lr.default-reductions", "accepting");
648 /* Check %define front-end variables. */
650 static char const * const values
[] = {
651 "lr.type", "lalr", "ielr", "canonical-lr", NULL
,
652 "lr.default-reductions", "most", "consistent", "accepting", NULL
,
655 muscle_percent_define_check_values (values
);
660 /*-------------------------------------------------------------.
661 | Check the grammar that has just been read, and convert it to |
663 `-------------------------------------------------------------*/
666 check_and_convert_grammar (void)
668 /* Grammar has been read. Do some checking. */
670 fatal (_("no rules in the input grammar"));
672 /* If the user did not define her ENDTOKEN, do it now. */
675 endtoken
= symbol_get ("$end", empty_location
);
676 endtoken
->class = token_sym
;
677 endtoken
->number
= 0;
678 /* Value specified by POSIX. */
679 endtoken
->user_token_number
= 0;
682 /* Report any undefined symbols and consider them nonterminals. */
683 symbols_check_defined ();
685 /* Find the start symbol if no %start. */
690 node
!= NULL
&& symbol_is_dummy (node
->content
.sym
);
693 for (node
= node
->next
;
694 node
!= NULL
&& node
->content
.sym
!= NULL
;
699 grammar_start_symbol_set (node
->content
.sym
,
700 node
->content
.sym
->location
);
703 /* Insert the initial rule, whose line is that of the first rule
704 (not that of the start symbol):
706 accept: %start EOF. */
708 symbol_list
*p
= symbol_list_sym_new (accept
, empty_location
);
709 p
->location
= grammar
->location
;
710 p
->next
= symbol_list_sym_new (startsymbol
, empty_location
);
711 p
->next
->next
= symbol_list_sym_new (endtoken
, empty_location
);
712 p
->next
->next
->next
= symbol_list_sym_new (NULL
, empty_location
);
713 p
->next
->next
->next
->next
= grammar
;
719 aver (nsyms
<= SYMBOL_NUMBER_MAXIMUM
&& nsyms
== ntokens
+ nvars
);
721 /* Assign the symbols their symbol numbers. Write #defines for the
722 token symbols into FDEFINES if requested. */
725 /* Scan rule actions after invoking symbol_check_alias_consistency (in
726 symbols_pack above) so that token types are set correctly before the rule
727 action type checking.
729 Before invoking grammar_rule_check (in packgram below) on any rule, make
730 sure all actions have already been scanned in order to set `used' flags.
731 Otherwise, checking that a midrule's $$ should be set will not always work
732 properly because the check must forward-reference the midrule's parent
733 rule. For the same reason, all the `used' flags must be set before
734 checking whether to remove `$' from any midrule symbol name (also in
738 for (sym
= grammar
; sym
; sym
= sym
->next
)
739 code_props_translate_code (&sym
->action_props
);
742 /* Convert the grammar into the format described in gram.h. */
745 /* The grammar as a symbol_list is no longer needed. */
746 symbol_list_free (grammar
);