1 /* Input parser for Bison
3 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007,
4 2009-2013 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 (&loc
, complaint
, _("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
))
132 complain_indent (&declaration_loc
, complaint
, &indent
,
133 _("result type clash on merge function %s: "
135 quote (merge_function
->name
), type
,
136 merge_function
->type
);
137 indent
+= SUB_INDENT
;
138 complain_indent (&merge_function
->type_declaration_location
, complaint
,
140 _("previous declaration"));
142 merge_function
->type
= uniqstr_new (type
);
143 merge_function
->type_declaration_location
= declaration_loc
;
146 /*--------------------------------------.
147 | Free all merge-function definitions. |
148 `--------------------------------------*/
151 free_merger_functions (void)
153 merger_list
*L0
= merge_functions
;
156 merger_list
*L1
= L0
->next
;
163 /*-------------------------------------------------------------------.
164 | Parse the input grammar into a one symbol_list structure. Each |
165 | rule is represented by a sequence of symbols: the left hand side |
166 | followed by the contents of the right hand side, followed by a |
167 | null pointer instead of a symbol to terminate the rule. The next |
168 | symbol is the lhs of the following rule. |
170 | All actions are copied out, labelled by the rule number they apply |
172 `-------------------------------------------------------------------*/
174 /* The (currently) last symbol of GRAMMAR. */
175 static symbol_list
*grammar_end
= NULL
;
177 /* Append SYM to the grammar. */
179 grammar_symbol_append (symbol
*sym
, location loc
)
181 symbol_list
*p
= symbol_list_sym_new (sym
, loc
);
184 grammar_end
->next
= p
;
190 /* A null SYM stands for an end of rule; it is not an actual
199 assign_named_ref (symbol_list
*p
, named_ref
*name
)
201 symbol
*sym
= p
->content
.sym
;
203 if (name
->id
== sym
->tag
)
205 complain (&name
->loc
, Wother
,
206 _("duplicated symbol name for %s ignored"),
208 named_ref_free (name
);
215 /* The rule currently being defined, and the previous rule.
216 CURRENT_RULE points to the first LHS of the current rule, while
217 PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
218 static symbol_list
*current_rule
= NULL
;
219 static symbol_list
*previous_rule_end
= NULL
;
222 /*----------------------------------------------.
223 | Create a new rule for LHS in to the GRAMMAR. |
224 `----------------------------------------------*/
227 grammar_current_rule_begin (symbol
*lhs
, location loc
,
232 /* Start a new rule and record its lhs. */
234 previous_rule_end
= grammar_end
;
236 p
= grammar_symbol_append (lhs
, loc
);
238 assign_named_ref (p
, named_ref_copy (lhs_name
));
240 current_rule
= grammar_end
;
242 /* Mark the rule's lhs as a nonterminal if not already so. */
243 if (lhs
->class == unknown_sym
)
245 lhs
->class = nterm_sym
;
249 else if (lhs
->class == token_sym
)
250 complain (&loc
, complaint
, _("rule given for %s, which is a token"),
255 /*----------------------------------------------------------------------.
256 | A symbol should be used if either: |
257 | 1. It has a destructor. |
258 | 2. The symbol is a mid-rule symbol (i.e., the generated LHS |
259 | replacing a mid-rule action) that was assigned to or used, as in |
260 | "exp: { $$ = 1; } { $$ = $1; }". |
261 `----------------------------------------------------------------------*/
264 symbol_should_be_used (symbol_list
const *s
, bool *midrule_warning
)
266 if (symbol_code_props_get (s
->content
.sym
, destructor
)->code
)
268 if ((s
->midrule
&& s
->midrule
->action_props
.is_value_used
)
269 || (s
->midrule_parent_rule
270 && (symbol_list_n_get (s
->midrule_parent_rule
,
271 s
->midrule_parent_rhs_index
)
272 ->action_props
.is_value_used
)))
274 *midrule_warning
= true;
280 /*----------------------------------------------------------------.
281 | Check that the rule R is properly defined. For instance, there |
282 | should be no type clash on the default action. |
283 `----------------------------------------------------------------*/
286 grammar_rule_check (const symbol_list
*r
)
290 If there is an action, then there is nothing we can do: the user
291 is allowed to shoot herself in the foot.
293 Don't worry about the default action if $$ is untyped, since $$'s
294 value can't be used. */
295 if (!r
->action_props
.code
&& r
->content
.sym
->type_name
)
297 symbol
*first_rhs
= r
->next
->content
.sym
;
298 /* If $$ is being set in default way, report if any type mismatch. */
301 char const *lhs_type
= r
->content
.sym
->type_name
;
302 const char *rhs_type
=
303 first_rhs
->type_name
? first_rhs
->type_name
: "";
304 if (!UNIQSTR_EQ (lhs_type
, rhs_type
))
305 complain (&r
->location
, Wother
,
306 _("type clash on default action: <%s> != <%s>"),
309 /* Warn if there is no default for $$ but we need one. */
311 complain (&r
->location
, Wother
,
312 _("empty rule for typed nonterminal, and no action"));
315 /* Check that symbol values that should be used are in fact used. */
317 symbol_list
const *l
= r
;
319 for (; l
&& l
->content
.sym
; l
= l
->next
, ++n
)
321 bool midrule_warning
= false;
322 if (!l
->action_props
.is_value_used
323 && symbol_should_be_used (l
, &midrule_warning
)
324 /* The default action, $$ = $1, 'uses' both. */
325 && (r
->action_props
.code
|| (n
!= 0 && n
!= 1)))
327 warnings warn_flag
= midrule_warning
? Wmidrule_values
: Wother
;
329 complain (&l
->location
, warn_flag
, _("unused value: $%d"), n
);
331 complain (&l
->location
, warn_flag
, _("unset value: $$"));
336 /* Check that %empty => empty rule. */
337 if (r
->percent_empty_loc
.start
.file
338 && r
->next
&& r
->next
->content
.sym
)
339 complain (&r
->percent_empty_loc
, complaint
,
340 _("%%empty on non-empty rule"));
342 /* Check that empty rule => %empty. */
343 if (!(r
->next
&& r
->next
->content
.sym
)
344 && !r
->midrule_parent_rule
345 && !r
->percent_empty_loc
.start
.file
)
346 complain (&r
->location
, Wempty_rule
, _("empty rule without %%empty"));
348 /* See comments in grammar_current_rule_prec_set for how POSIX
349 mandates this complaint. It's only for identifiers, so skip
350 it for char literals and strings, which are always tokens. */
352 && r
->ruleprec
->tag
[0] != '\'' && r
->ruleprec
->tag
[0] != '"'
353 && r
->ruleprec
->status
!= declared
&& !r
->ruleprec
->prec
)
354 complain (&r
->location
, Wother
,
355 _("token for %%prec is not defined: %s"), r
->ruleprec
->tag
);
359 /*-------------------------------------.
360 | End the currently being grown rule. |
361 `-------------------------------------*/
364 grammar_current_rule_end (location loc
)
366 /* Put an empty link in the list to mark the end of this rule */
367 grammar_symbol_append (NULL
, grammar_end
->location
);
368 current_rule
->location
= loc
;
372 /*-------------------------------------------------------------------.
373 | The previous action turns out the be a mid-rule action. Attach it |
374 | to the current rule, i.e., create a dummy symbol, attach it this |
375 | mid-rule action, and append this dummy nonterminal to the current |
377 `-------------------------------------------------------------------*/
380 grammar_midrule_action (void)
382 /* Since the action was written out with this rule's number, we must
383 give the new rule this number by inserting the new rule before
386 /* Make a DUMMY nonterminal, whose location is that of the midrule
387 action. Create the MIDRULE. */
388 location dummy_location
= current_rule
->action_props
.location
;
389 symbol
*dummy
= dummy_symbol_get (dummy_location
);
390 symbol_list
*midrule
= symbol_list_sym_new (dummy
, dummy_location
);
392 /* Remember named_ref of previous action. */
393 named_ref
*action_name
= current_rule
->action_props
.named_ref
;
395 /* Make a new rule, whose body is empty, before the current one, so
396 that the action just read can belong to it. */
399 /* Attach its location and actions to that of the DUMMY. */
400 midrule
->location
= dummy_location
;
401 code_props_rule_action_init (&midrule
->action_props
,
402 current_rule
->action_props
.code
,
403 current_rule
->action_props
.location
,
405 current_rule
->action_props
.is_predicate
);
406 code_props_none_init (¤t_rule
->action_props
);
408 if (previous_rule_end
)
409 previous_rule_end
->next
= midrule
;
413 /* End the dummy's rule. */
414 midrule
->next
= symbol_list_sym_new (NULL
, dummy_location
);
415 midrule
->next
->next
= current_rule
;
417 previous_rule_end
= midrule
->next
;
419 /* Insert the dummy nonterminal replacing the midrule action into
420 the current rule. Bind it to its dedicated rule. */
421 grammar_current_rule_symbol_append (dummy
, dummy_location
,
423 grammar_end
->midrule
= midrule
;
424 midrule
->midrule_parent_rule
= current_rule
;
425 midrule
->midrule_parent_rhs_index
= symbol_list_length (current_rule
->next
);
428 /* Set the precedence symbol of the current rule to PRECSYM. */
431 grammar_current_rule_prec_set (symbol
*precsym
, location loc
)
433 /* POSIX says that any identifier is a nonterminal if it does not
434 appear on the LHS of a grammar rule and is not defined by %token
435 or by one of the directives that assigns precedence to a token. We
436 ignore this here because the only kind of identifier that POSIX
437 allows to follow a %prec is a token and because assuming it's a
438 token now can produce more logical error messages. Nevertheless,
439 grammar_rule_check does obey what we believe is the real intent of
440 POSIX here: that an error be reported for any identifier that
441 appears after %prec but that is not defined separately as a
443 symbol_class_set (precsym
, token_sym
, loc
, false);
444 if (current_rule
->ruleprec
)
445 duplicate_directive ("%prec",
446 current_rule
->ruleprec
->location
, loc
);
448 current_rule
->ruleprec
= precsym
;
451 /* Set %empty for the current rule. */
454 grammar_current_rule_empty_set (location loc
)
456 /* If %empty is used and -Wno-empty-rule is not, then enable
458 if (warning_is_unset (Wempty_rule
))
459 warning_argmatch ("empty-rule", 0, 0);
460 if (current_rule
->percent_empty_loc
.start
.file
)
461 duplicate_directive ("%empty",
462 current_rule
->percent_empty_loc
, loc
);
464 current_rule
->percent_empty_loc
= loc
;
467 /* Attach dynamic precedence DPREC to the current rule. */
470 grammar_current_rule_dprec_set (int dprec
, location loc
)
473 complain (&loc
, Wother
, _("%s affects only GLR parsers"),
476 complain (&loc
, complaint
, _("%s must be followed by positive number"),
478 else if (current_rule
->dprec
!= 0)
479 duplicate_directive ("%dprec",
480 current_rule
->dprec_location
, loc
);
483 current_rule
->dprec
= dprec
;
484 current_rule
->dprec_location
= loc
;
488 /* Attach a merge function NAME with argument type TYPE to current
492 grammar_current_rule_merge_set (uniqstr name
, location loc
)
495 complain (&loc
, Wother
, _("%s affects only GLR parsers"),
497 if (current_rule
->merger
!= 0)
498 duplicate_directive ("%merge",
499 current_rule
->merger_declaration_location
, loc
);
502 current_rule
->merger
= get_merge_function (name
);
503 current_rule
->merger_declaration_location
= loc
;
507 /* Attach SYM to the current rule. If needed, move the previous
508 action as a mid-rule action. */
511 grammar_current_rule_symbol_append (symbol
*sym
, location loc
,
515 if (current_rule
->action_props
.code
)
516 grammar_midrule_action ();
517 p
= grammar_symbol_append (sym
, loc
);
519 assign_named_ref(p
, name
);
520 if (sym
->status
== undeclared
|| sym
->status
== used
)
521 sym
->status
= needed
;
524 /* Attach an ACTION to the current rule. */
527 grammar_current_rule_action_append (const char *action
, location loc
,
528 named_ref
*name
, bool is_predicate
)
530 if (current_rule
->action_props
.code
)
531 grammar_midrule_action ();
532 /* After all symbol declarations have been parsed, packgram invokes
533 code_props_translate_code. */
534 code_props_rule_action_init (¤t_rule
->action_props
, action
, loc
,
535 current_rule
, name
, is_predicate
);
539 /*---------------------------------------------------------------.
540 | Convert the rules into the representation using RRHS, RLHS and |
542 `---------------------------------------------------------------*/
547 unsigned int itemno
= 0;
548 rule_number ruleno
= 0;
551 ritem
= xnmalloc (nritems
+ 1, sizeof *ritem
);
553 /* This sentinel is used by build_relations in gram.c. */
556 rules
= xnmalloc (nrules
, sizeof *rules
);
558 for (p
= grammar
; p
; p
= p
->next
)
560 symbol
*ruleprec
= p
->ruleprec
;
561 record_merge_function_type (p
->merger
, p
->content
.sym
->type_name
,
562 p
->merger_declaration_location
);
563 rules
[ruleno
].user_number
= ruleno
;
564 rules
[ruleno
].number
= ruleno
;
565 rules
[ruleno
].lhs
= p
->content
.sym
;
566 rules
[ruleno
].rhs
= ritem
+ itemno
;
567 rules
[ruleno
].prec
= NULL
;
568 rules
[ruleno
].dprec
= p
->dprec
;
569 rules
[ruleno
].merger
= p
->merger
;
570 rules
[ruleno
].precsym
= NULL
;
571 rules
[ruleno
].location
= p
->location
;
572 rules
[ruleno
].useful
= true;
573 rules
[ruleno
].action
= p
->action_props
.code
;
574 rules
[ruleno
].action_location
= p
->action_props
.location
;
575 rules
[ruleno
].is_predicate
= p
->action_props
.is_predicate
;
577 /* If the midrule's $$ is set or its $n is used, remove the '$' from the
578 symbol name so that it's a user-defined symbol so that the default
579 %destructor and %printer apply. */
580 if (p
->midrule_parent_rule
581 && (p
->action_props
.is_value_used
582 || (symbol_list_n_get (p
->midrule_parent_rule
,
583 p
->midrule_parent_rhs_index
)
584 ->action_props
.is_value_used
)))
585 p
->content
.sym
->tag
+= 1;
587 /* Don't check the generated rule 0. It has no action, so some rhs
588 symbols may appear unused, but the parsing algorithm ensures that
589 %destructor's are invoked appropriately. */
591 grammar_rule_check (p
);
594 size_t rule_length
= 0;
595 for (p
= p
->next
; p
->content
.sym
; p
= p
->next
)
599 /* Don't allow rule_length == INT_MAX, since that might
600 cause confusion with strtol if INT_MAX == LONG_MAX. */
601 if (rule_length
== INT_MAX
)
602 complain (&rules
[ruleno
].location
, fatal
, _("rule is too long"));
604 /* item_number = symbol_number.
605 But the former needs to contain more: negative rule numbers. */
607 symbol_number_as_item_number (p
->content
.sym
->number
);
608 /* A rule gets by default the precedence and associativity
609 of its last token. */
610 if (p
->content
.sym
->class == token_sym
&& default_prec
)
611 rules
[ruleno
].prec
= p
->content
.sym
;
615 /* If this rule has a %prec,
616 the specified symbol's precedence replaces the default. */
619 rules
[ruleno
].precsym
= ruleprec
;
620 rules
[ruleno
].prec
= ruleprec
;
622 /* An item ends by the rule number (negated). */
623 ritem
[itemno
++] = rule_number_as_item_number (ruleno
);
624 aver (itemno
< ITEM_NUMBER_MAX
);
626 aver (ruleno
< RULE_NUMBER_MAX
);
629 aver (itemno
== nritems
);
631 if (trace_flag
& trace_sets
)
632 ritem_print (stderr
);
635 /*------------------------------------------------------------------.
636 | Read in the grammar specification and record it in the format |
637 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
638 | in each case forming the body of a C function (YYACTION) which |
639 | contains a switch statement to decide which action to execute. |
640 `------------------------------------------------------------------*/
645 /* Initialize the symbol table. */
648 /* Construct the accept symbol. */
649 accept
= symbol_get ("$accept", empty_location
);
650 accept
->class = nterm_sym
;
651 accept
->number
= nvars
++;
653 /* Construct the error token */
654 errtoken
= symbol_get ("error", empty_location
);
655 errtoken
->class = token_sym
;
656 errtoken
->number
= ntokens
++;
658 /* Construct a token that represents all undefined literal tokens.
659 It is always token number 2. */
660 undeftoken
= symbol_get ("$undefined", empty_location
);
661 undeftoken
->class = token_sym
;
662 undeftoken
->number
= ntokens
++;
664 gram_in
= xfopen (grammar_file
, "r");
666 gram__flex_debug
= trace_flag
& trace_scan
;
667 gram_debug
= trace_flag
& trace_parse
;
668 gram_scanner_initialize ();
670 prepare_percent_define_front_end_variables ();
672 if (complaint_status
< status_complaint
)
673 check_and_convert_grammar ();
679 prepare_percent_define_front_end_variables (void)
681 /* Set %define front-end variable defaults. */
682 muscle_percent_define_default ("lr.keep-unreachable-state", "false");
685 /* IELR would be a better default, but LALR is historically the
687 muscle_percent_define_default ("lr.type", "lalr");
688 lr_type
= muscle_percent_define_get ("lr.type");
689 if (STRNEQ (lr_type
, "canonical-lr"))
690 muscle_percent_define_default ("lr.default-reduction", "most");
692 muscle_percent_define_default ("lr.default-reduction", "accepting");
696 /* Check %define front-end variables. */
698 static char const * const values
[] = {
699 "lr.type", "lalr", "ielr", "canonical-lr", NULL
,
700 "lr.default-reduction", "most", "consistent", "accepting", NULL
,
703 muscle_percent_define_check_values (values
);
708 /*-------------------------------------------------------------.
709 | Check the grammar that has just been read, and convert it to |
711 `-------------------------------------------------------------*/
714 check_and_convert_grammar (void)
716 /* Grammar has been read. Do some checking. */
718 complain (NULL
, fatal
, _("no rules in the input grammar"));
720 /* If the user did not define her ENDTOKEN, do it now. */
723 endtoken
= symbol_get ("$end", empty_location
);
724 endtoken
->class = token_sym
;
725 endtoken
->number
= 0;
726 /* Value specified by POSIX. */
727 endtoken
->user_token_number
= 0;
730 /* Report any undefined symbols and consider them nonterminals. */
731 symbols_check_defined ();
733 /* Find the start symbol if no %start. */
738 node
!= NULL
&& symbol_is_dummy (node
->content
.sym
);
741 for (node
= node
->next
;
742 node
!= NULL
&& node
->content
.sym
!= NULL
;
747 grammar_start_symbol_set (node
->content
.sym
,
748 node
->content
.sym
->location
);
751 /* Insert the initial rule, whose line is that of the first rule
752 (not that of the start symbol):
754 $accept: %start $end. */
756 symbol_list
*p
= symbol_list_sym_new (accept
, empty_location
);
757 p
->location
= grammar
->location
;
758 p
->next
= symbol_list_sym_new (startsymbol
, empty_location
);
759 p
->next
->next
= symbol_list_sym_new (endtoken
, empty_location
);
760 p
->next
->next
->next
= symbol_list_sym_new (NULL
, empty_location
);
761 p
->next
->next
->next
->next
= grammar
;
767 aver (nsyms
<= SYMBOL_NUMBER_MAXIMUM
);
768 aver (nsyms
== ntokens
+ nvars
);
770 /* Assign the symbols their symbol numbers. Write #defines for the
771 token symbols into FDEFINES if requested. */
774 /* Scan rule actions after invoking symbol_check_alias_consistency (in
775 symbols_pack above) so that token types are set correctly before the rule
776 action type checking.
778 Before invoking grammar_rule_check (in packgram below) on any rule, make
779 sure all actions have already been scanned in order to set 'used' flags.
780 Otherwise, checking that a midrule's $$ should be set will not always work
781 properly because the check must forward-reference the midrule's parent
782 rule. For the same reason, all the 'used' flags must be set before
783 checking whether to remove '$' from any midrule symbol name (also in
787 for (sym
= grammar
; sym
; sym
= sym
->next
)
788 code_props_translate_code (&sym
->action_props
);
791 /* Convert the grammar into the format described in gram.h. */
794 /* The grammar as a symbol_list is no longer needed. */
795 symbol_list_free (grammar
);