1 /* Input parser for Bison
3 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007,
4 2009-2010 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/>. */
28 #include "conflicts.h"
32 #include "muscle-tab.h"
36 #include "scan-gram.h"
37 #include "scan-code.h"
39 static void prepare_percent_define_front_end_variables (void);
40 static void check_and_convert_grammar (void);
42 static symbol_list
*grammar
= NULL
;
43 static bool start_flag
= false;
44 merger_list
*merge_functions
;
46 /* Was %union seen? */
47 bool union_seen
= false;
50 bool tag_seen
= false;
52 /* Should rules have a default precedence? */
53 bool default_prec
= true;
55 /*-----------------------.
56 | Set the start symbol. |
57 `-----------------------*/
60 grammar_start_symbol_set (symbol
*sym
, location loc
)
63 complain_at (loc
, _("multiple %s declarations"), "%start");
68 startsymbol_location
= loc
;
74 /*------------------------------------------------------------------------.
75 | Return the merger index for a merging function named NAME. Records the |
76 | function, if new, in MERGER_LIST. |
77 `------------------------------------------------------------------------*/
80 get_merge_function (uniqstr name
)
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
))
93 if (syms
->next
== NULL
)
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
;
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 `-------------------------------------------------------------------------*/
113 record_merge_function_type (int merger
, uniqstr type
, location declaration_loc
)
116 merger_list
*merge_function
;
122 type
= uniqstr_new ("");
125 for (merge_function
= merge_functions
;
126 merge_function
!= NULL
&& merger_find
!= merger
;
127 merge_function
= merge_function
->next
)
129 aver (merge_function
!= NULL
&& merger_find
== merger
);
130 if (merge_function
->type
!= NULL
&& !UNIQSTR_EQ (merge_function
->type
, type
))
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"));
138 merge_function
->type
= uniqstr_new (type
);
139 merge_function
->type_declaration_location
= declaration_loc
;
142 /*--------------------------------------.
143 | Free all merge-function definitions. |
144 `--------------------------------------*/
147 free_merger_functions (void)
149 merger_list
*L0
= merge_functions
;
152 merger_list
*L1
= L0
->next
;
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. |
166 | All actions are copied out, labelled by the rule number they apply |
168 `-------------------------------------------------------------------*/
170 /* The (currently) last symbol of GRAMMAR. */
171 static symbol_list
*grammar_end
= NULL
;
173 /* Append SYM to the grammar. */
175 grammar_symbol_append (symbol
*sym
, location loc
)
177 symbol_list
*p
= symbol_list_sym_new (sym
, loc
);
180 grammar_end
->next
= p
;
186 /* A null SYM stands for an end of rule; it is not an actual
195 assign_named_ref (symbol_list
*p
, named_ref
*name
)
197 symbol
*sym
= p
->content
.sym
;
199 if (name
->id
== sym
->tag
)
202 _("duplicated symbol name for %s ignored"),
204 named_ref_free (name
);
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
;
218 /*----------------------------------------------.
219 | Create a new rule for LHS in to the GRAMMAR. |
220 `----------------------------------------------*/
223 grammar_current_rule_begin (symbol
*lhs
, location loc
,
228 /* Start a new rule and record its lhs. */
230 previous_rule_end
= grammar_end
;
232 p
= grammar_symbol_append (lhs
, loc
);
234 assign_named_ref(p
, lhs_name
);
236 current_rule
= grammar_end
;
238 /* Mark the rule's lhs as a nonterminal if not already so. */
239 if (lhs
->class == unknown_sym
)
241 lhs
->class = nterm_sym
;
245 else if (lhs
->class == token_sym
)
246 complain_at (loc
, _("rule given for %s, which is a token"), lhs
->tag
);
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 `----------------------------------------------------------------------*/
259 symbol_should_be_used (symbol_list
const *s
)
261 if (symbol_destructor_get (s
->content
.sym
)->code
)
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
));
272 /*----------------------------------------------------------------.
273 | Check that the rule R is properly defined. For instance, there |
274 | should be no type clash on the default action. |
275 `----------------------------------------------------------------*/
278 grammar_rule_check (const symbol_list
*r
)
282 If there is an action, then there is nothing we can do: the user
283 is allowed to shoot herself in the foot.
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
)
289 symbol
*first_rhs
= r
->next
->content
.sym
;
290 /* If $$ is being set in default way, report if any type mismatch. */
293 char const *lhs_type
= r
->content
.sym
->type_name
;
294 const char *rhs_type
=
295 first_rhs
->type_name
? first_rhs
->type_name
: "";
296 if (!UNIQSTR_EQ (lhs_type
, rhs_type
))
297 warn_at (r
->location
,
298 _("type clash on default action: <%s> != <%s>"),
301 /* Warn if there is no default for $$ but we need one. */
303 warn_at (r
->location
,
304 _("empty rule for typed nonterminal, and no action"));
307 /* Check that symbol values that should be used are in fact used. */
309 symbol_list
const *l
= r
;
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, $$ = $1, `uses' both. */
315 || (!r
->action_props
.code
&& (n
== 0 || n
== 1))))
318 warn_at (r
->location
, _("unused value: $%d"), n
);
320 warn_at (r
->location
, _("unset value: $$"));
324 /* See comments in grammar_current_rule_prec_set for how POSIX
325 mandates this complaint. It's only for identifiers, so skip
326 it for char literals and strings, which are always tokens. */
328 && r
->ruleprec
->tag
[0] != '\'' && r
->ruleprec
->tag
[0] != '"'
329 && !r
->ruleprec
->declared
&& !r
->ruleprec
->prec
)
330 complain_at (r
->location
, _("token for %%prec is not defined: %s"),
335 /*-------------------------------------.
336 | End the currently being grown rule. |
337 `-------------------------------------*/
340 grammar_current_rule_end (location loc
)
342 /* Put an empty link in the list to mark the end of this rule */
343 grammar_symbol_append (NULL
, grammar_end
->location
);
344 current_rule
->location
= loc
;
348 /*-------------------------------------------------------------------.
349 | The previous action turns out the be a mid-rule action. Attach it |
350 | to the current rule, i.e., create a dummy symbol, attach it this |
351 | mid-rule action, and append this dummy nonterminal to the current |
353 `-------------------------------------------------------------------*/
356 grammar_midrule_action (void)
358 /* Since the action was written out with this rule's number, we must
359 give the new rule this number by inserting the new rule before
362 /* Make a DUMMY nonterminal, whose location is that of the midrule
363 action. Create the MIDRULE. */
364 location dummy_location
= current_rule
->action_props
.location
;
365 symbol
*dummy
= dummy_symbol_get (dummy_location
);
366 symbol_list
*midrule
= symbol_list_sym_new (dummy
, dummy_location
);
368 /* Remember named_ref of previous action. */
369 named_ref
*action_name
= current_rule
->action_props
.named_ref
;
371 /* Make a new rule, whose body is empty, before the current one, so
372 that the action just read can belong to it. */
375 /* Attach its location and actions to that of the DUMMY. */
376 midrule
->location
= dummy_location
;
377 code_props_rule_action_init (&midrule
->action_props
,
378 current_rule
->action_props
.code
,
379 current_rule
->action_props
.location
,
381 code_props_none_init (¤t_rule
->action_props
);
383 if (previous_rule_end
)
384 previous_rule_end
->next
= midrule
;
388 /* End the dummy's rule. */
389 midrule
->next
= symbol_list_sym_new (NULL
, dummy_location
);
390 midrule
->next
->next
= current_rule
;
392 previous_rule_end
= midrule
->next
;
394 /* Insert the dummy nonterminal replacing the midrule action into
395 the current rule. Bind it to its dedicated rule. */
396 grammar_current_rule_symbol_append (dummy
, dummy_location
,
398 grammar_end
->midrule
= midrule
;
399 midrule
->midrule_parent_rule
= current_rule
;
400 midrule
->midrule_parent_rhs_index
= symbol_list_length (current_rule
->next
);
403 /* Set the precedence symbol of the current rule to PRECSYM. */
406 grammar_current_rule_prec_set (symbol
*precsym
, location loc
)
408 /* POSIX says that any identifier is a nonterminal if it does not
409 appear on the LHS of a grammar rule and is not defined by %token
410 or by one of the directives that assigns precedence to a token. We
411 ignore this here because the only kind of identifier that POSIX
412 allows to follow a %prec is a token and because assuming it's a
413 token now can produce more logical error messages. Nevertheless,
414 grammar_rule_check does obey what we believe is the real intent of
415 POSIX here: that an error be reported for any identifier that
416 appears after %prec but that is not defined separately as a
418 symbol_class_set (precsym
, token_sym
, loc
, false);
419 if (current_rule
->ruleprec
)
420 complain_at (loc
, _("only one %s allowed per rule"), "%prec");
421 current_rule
->ruleprec
= precsym
;
424 /* Attach dynamic precedence DPREC to the current rule. */
427 grammar_current_rule_dprec_set (int dprec
, location loc
)
430 warn_at (loc
, _("%s affects only GLR parsers"), "%dprec");
432 complain_at (loc
, _("%s must be followed by positive number"), "%dprec");
433 else if (current_rule
->dprec
!= 0)
434 complain_at (loc
, _("only one %s allowed per rule"), "%dprec");
435 current_rule
->dprec
= dprec
;
438 /* Attach a merge function NAME with argument type TYPE to current
442 grammar_current_rule_merge_set (uniqstr name
, location loc
)
445 warn_at (loc
, _("%s affects only GLR parsers"), "%merge");
446 if (current_rule
->merger
!= 0)
447 complain_at (loc
, _("only one %s allowed per rule"), "%merge");
448 current_rule
->merger
= get_merge_function (name
);
449 current_rule
->merger_declaration_location
= loc
;
452 /* Attach SYM to the current rule. If needed, move the previous
453 action as a mid-rule action. */
456 grammar_current_rule_symbol_append (symbol
*sym
, location loc
,
460 if (current_rule
->action_props
.code
)
461 grammar_midrule_action ();
462 p
= grammar_symbol_append (sym
, loc
);
464 assign_named_ref(p
, name
);
467 /* Attach an ACTION to the current rule. */
470 grammar_current_rule_action_append (const char *action
, location loc
,
473 if (current_rule
->action_props
.code
)
474 grammar_midrule_action ();
475 /* After all symbol declarations have been parsed, packgram invokes
476 code_props_translate_code. */
477 code_props_rule_action_init (¤t_rule
->action_props
, action
, loc
,
482 /*---------------------------------------------------------------.
483 | Convert the rules into the representation using RRHS, RLHS and |
485 `---------------------------------------------------------------*/
490 unsigned int itemno
= 0;
491 rule_number ruleno
= 0;
492 symbol_list
*p
= grammar
;
494 ritem
= xnmalloc (nritems
+ 1, sizeof *ritem
);
496 /* This sentinel is used by build_relations in gram.c. */
499 rules
= xnmalloc (nrules
, sizeof *rules
);
504 symbol
*ruleprec
= p
->ruleprec
;
505 record_merge_function_type (p
->merger
, p
->content
.sym
->type_name
,
506 p
->merger_declaration_location
);
507 rules
[ruleno
].user_number
= ruleno
;
508 rules
[ruleno
].number
= ruleno
;
509 rules
[ruleno
].lhs
= p
->content
.sym
;
510 rules
[ruleno
].rhs
= ritem
+ itemno
;
511 rules
[ruleno
].prec
= NULL
;
512 rules
[ruleno
].dprec
= p
->dprec
;
513 rules
[ruleno
].merger
= p
->merger
;
514 rules
[ruleno
].precsym
= NULL
;
515 rules
[ruleno
].location
= p
->location
;
516 rules
[ruleno
].useful
= true;
517 rules
[ruleno
].action
= p
->action_props
.code
;
518 rules
[ruleno
].action_location
= p
->action_props
.location
;
520 /* If the midrule's $$ is set or its $n is used, remove the `$' from the
521 symbol name so that it's a user-defined symbol so that the default
522 %destructor and %printer apply. */
523 if (p
->midrule_parent_rule
524 && (p
->action_props
.is_value_used
525 || symbol_list_n_get (p
->midrule_parent_rule
,
526 p
->midrule_parent_rhs_index
)
527 ->action_props
.is_value_used
))
528 p
->content
.sym
->tag
+= 1;
530 /* Don't check the generated rule 0. It has no action, so some rhs
531 symbols may appear unused, but the parsing algorithm ensures that
532 %destructor's are invoked appropriately. */
534 grammar_rule_check (p
);
536 for (p
= p
->next
; p
&& p
->content
.sym
; p
= p
->next
)
540 /* Don't allow rule_length == INT_MAX, since that might
541 cause confusion with strtol if INT_MAX == LONG_MAX. */
542 if (rule_length
== INT_MAX
)
543 fatal_at (rules
[ruleno
].location
, _("rule is too long"));
545 /* item_number = symbol_number.
546 But the former needs to contain more: negative rule numbers. */
548 symbol_number_as_item_number (p
->content
.sym
->number
);
549 /* A rule gets by default the precedence and associativity
550 of its last token. */
551 if (p
->content
.sym
->class == token_sym
&& default_prec
)
552 rules
[ruleno
].prec
= p
->content
.sym
;
555 /* If this rule has a %prec,
556 the specified symbol's precedence replaces the default. */
559 rules
[ruleno
].precsym
= ruleprec
;
560 rules
[ruleno
].prec
= ruleprec
;
562 /* An item ends by the rule number (negated). */
563 ritem
[itemno
++] = rule_number_as_item_number (ruleno
);
564 aver (itemno
< ITEM_NUMBER_MAX
);
566 aver (ruleno
< RULE_NUMBER_MAX
);
572 aver (itemno
== nritems
);
574 if (trace_flag
& trace_sets
)
575 ritem_print (stderr
);
578 /*------------------------------------------------------------------.
579 | Read in the grammar specification and record it in the format |
580 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
581 | in each case forming the body of a C function (YYACTION) which |
582 | contains a switch statement to decide which action to execute. |
583 `------------------------------------------------------------------*/
588 /* Initialize the symbol table. */
591 /* Construct the accept symbol. */
592 accept
= symbol_get ("$accept", empty_location
);
593 accept
->class = nterm_sym
;
594 accept
->number
= nvars
++;
596 /* Construct the error token */
597 errtoken
= symbol_get ("error", empty_location
);
598 errtoken
->class = token_sym
;
599 errtoken
->number
= ntokens
++;
601 /* Construct a token that represents all undefined literal tokens.
602 It is always token number 2. */
603 undeftoken
= symbol_get ("$undefined", empty_location
);
604 undeftoken
->class = token_sym
;
605 undeftoken
->number
= ntokens
++;
607 gram_in
= xfopen (grammar_file
, "r");
609 gram__flex_debug
= trace_flag
& trace_scan
;
610 gram_debug
= trace_flag
& trace_parse
;
611 gram_scanner_initialize ();
613 prepare_percent_define_front_end_variables ();
615 if (! complaint_issued
)
616 check_and_convert_grammar ();
622 prepare_percent_define_front_end_variables (void)
624 /* Set %define front-end variable defaults. */
625 muscle_percent_define_default ("lr.keep-unreachable-states", "false");
628 /* IELR would be a better default, but LALR is historically the
630 muscle_percent_define_default ("lr.type", "lalr");
631 lr_type
= muscle_percent_define_get ("lr.type");
632 if (0 != strcmp (lr_type
, "canonical-lr"))
633 muscle_percent_define_default ("lr.default-reductions", "all");
635 muscle_percent_define_default ("lr.default-reductions", "accepting");
639 /* Check %define front-end variables. */
641 static char const * const values
[] = {
642 "lr.type", "lalr", "ielr", "canonical-lr", NULL
,
643 "lr.default-reductions", "all", "consistent", "accepting", NULL
,
646 muscle_percent_define_check_values (values
);
651 /*-------------------------------------------------------------.
652 | Check the grammar that has just been read, and convert it to |
654 `-------------------------------------------------------------*/
657 check_and_convert_grammar (void)
659 /* Grammar has been read. Do some checking. */
661 fatal (_("no rules in the input grammar"));
663 /* If the user did not define her ENDTOKEN, do it now. */
666 endtoken
= symbol_get ("$end", empty_location
);
667 endtoken
->class = token_sym
;
668 endtoken
->number
= 0;
669 /* Value specified by POSIX. */
670 endtoken
->user_token_number
= 0;
673 /* Report any undefined symbols and consider them nonterminals. */
674 symbols_check_defined ();
676 /* Find the start symbol if no %start. */
681 node
!= NULL
&& symbol_is_dummy (node
->content
.sym
);
684 for (node
= node
->next
;
685 node
!= NULL
&& node
->content
.sym
!= NULL
;
690 grammar_start_symbol_set (node
->content
.sym
,
691 node
->content
.sym
->location
);
694 /* Insert the initial rule, whose line is that of the first rule
695 (not that of the start symbol):
697 accept: %start EOF. */
699 symbol_list
*p
= symbol_list_sym_new (accept
, empty_location
);
700 p
->location
= grammar
->location
;
701 p
->next
= symbol_list_sym_new (startsymbol
, empty_location
);
702 p
->next
->next
= symbol_list_sym_new (endtoken
, empty_location
);
703 p
->next
->next
->next
= symbol_list_sym_new (NULL
, empty_location
);
704 p
->next
->next
->next
->next
= grammar
;
710 aver (nsyms
<= SYMBOL_NUMBER_MAXIMUM
&& nsyms
== ntokens
+ nvars
);
712 /* Assign the symbols their symbol numbers. Write #defines for the
713 token symbols into FDEFINES if requested. */
716 /* Scan rule actions after invoking symbol_check_alias_consistency (in
717 symbols_pack above) so that token types are set correctly before the rule
718 action type checking.
720 Before invoking grammar_rule_check (in packgram below) on any rule, make
721 sure all actions have already been scanned in order to set `used' flags.
722 Otherwise, checking that a midrule's $$ should be set will not always work
723 properly because the check must forward-reference the midrule's parent
724 rule. For the same reason, all the `used' flags must be set before
725 checking whether to remove `$' from any midrule symbol name (also in
729 for (sym
= grammar
; sym
; sym
= sym
->next
)
730 code_props_translate_code (&sym
->action_props
);
733 /* Convert the grammar into the format described in gram.h. */
736 /* The grammar as a symbol_list is no longer needed. */
737 symbol_list_free (grammar
);