1 /* Input parser for bison
2 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002
3 Free Software Foundation, Inc.
5 This file is part of Bison, the GNU Compiler Compiler.
7 Bison is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2, or (at your option)
12 Bison is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Bison; see the file COPYING. If not, write to
19 the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 Boston, MA 02111-1307, USA. */
34 #include "conflicts.h"
35 #include "muscle_tab.h"
38 static symbol_list
*grammar
= NULL
;
39 static int start_flag
= 0;
41 /* Nonzero if %union has been seen. */
45 symbol_list_new (symbol_t
*sym
)
47 symbol_list
*res
= XMALLOC (symbol_list
, 1);
58 /*--------------------------------------------------------------.
59 | Get the data type (alternative in the union) of the value for |
60 | symbol N in rule RULE. |
61 `--------------------------------------------------------------*/
64 get_type_name (int n
, symbol_list
*rule
)
71 complain (_("invalid $ value"));
81 if (rp
== NULL
|| rp
->sym
== NULL
)
83 complain (_("invalid $ value"));
89 return rp
->sym
->type_name
;
93 /*-----------------------.
94 | Set the start symbol. |
95 `-----------------------*/
98 grammar_start_symbol_set (symbol_t
*s
)
101 complain (_("multiple %s declarations"), "%start");
110 /*----------------------------------------------------------------.
111 | There are two prologues: one before %union, one after. Augment |
113 `----------------------------------------------------------------*/
116 prologue_augment (const char *prologue
, location_t location
)
118 struct obstack
*oout
=
119 !typed
? &pre_prologue_obstack
: &post_prologue_obstack
;
123 obstack_fgrow2 (oout
, muscle_find ("linef"),
125 quotearg_style (c_quoting_style
,
126 muscle_find ("filename")));
128 obstack_sgrow (oout
, prologue
);
134 /*----------------------.
135 | Handle the epilogue. |
136 `----------------------*/
139 epilogue_set (const char *epilogue
, location_t location
)
143 obstack_fgrow2 (&muscle_obstack
, muscle_find ("linef"),
145 quotearg_style (c_quoting_style
,
146 muscle_find ("filename")));
148 obstack_sgrow (&muscle_obstack
, epilogue
);
149 obstack_1grow (&muscle_obstack
, 0);
150 muscle_insert ("epilogue", obstack_finish (&muscle_obstack
));
156 /*-------------------------------------------------------------------.
157 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
158 | with the user's names. |
159 `-------------------------------------------------------------------*/
164 /* Incremented for each generated symbol */
165 static int gensym_count
= 0;
166 static char buf
[256];
170 sprintf (buf
, "@%d", ++gensym_count
);
172 sym
->class = nterm_sym
;
173 sym
->number
= nvars
++;
177 /*-------------------------------------------------------------------.
178 | Parse the input grammar into a one symbol_list structure. Each |
179 | rule is represented by a sequence of symbols: the left hand side |
180 | followed by the contents of the right hand side, followed by a |
181 | null pointer instead of a symbol to terminate the rule. The next |
182 | symbol is the lhs of the following rule. |
184 | All actions are copied out, labelled by the rule number they apply |
187 | Bison used to allow some %directives in the rules sections, but |
188 | this is no longer consider appropriate: (i) the documented grammar |
189 | doesn't claim it, (ii), it would promote bad style, (iii), error |
190 | recovery for %directives consists in skipping the junk until a `%' |
191 | is seen and helrp synchronizing. This scheme is definitely wrong |
192 | in the rules section. |
193 `-------------------------------------------------------------------*/
195 /* The (currently) last symbol of GRAMMAR. */
196 symbol_list
*grammar_end
= NULL
;
198 /* Append S to the GRAMMAR. */
200 grammar_symbol_append (symbol_t
*s
)
202 symbol_list
*p
= symbol_list_new (s
);
205 grammar_end
->next
= p
;
212 /* The rule currently being defined, and the previous rule. Point to
213 the first symbol of each list: their lhs. */
214 symbol_list
*current_rule
= NULL
;
215 symbol_list
*previous_rule
= NULL
;
218 /* Create a new rule for LHS in to the GRAMMAR. */
221 grammar_rule_begin (symbol_t
*lhs
)
229 /* Start a new rule and record its lhs. */
233 previous_rule
= grammar_end
;
234 grammar_symbol_append (lhs
);
235 current_rule
= grammar_end
;
237 /* 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 (_("rule given for %s, which is a token"), lhs
->tag
);
249 /* Check that the last rule (CURRENT_RULE) is properly defined. For
250 instance, there should be no type clash on the default action. */
253 grammar_current_rule_check (void)
255 symbol_t
*lhs
= current_rule
->sym
;
256 symbol_t
*first_rhs
= current_rule
->next
->sym
;
258 /* If there is an action, then there is nothing we can do: the user
259 is allowed to shoot in her foot. */
260 if (current_rule
->action
)
263 /* If $$ is being set in default way, report if any type mismatch.
267 const char *lhs_type
= lhs
->type_name
? lhs
->type_name
: "";
268 const char *rhs_type
= first_rhs
->type_name
? first_rhs
->type_name
: "";
269 if (strcmp (lhs_type
, rhs_type
))
270 complain (_("type clash (`%s' `%s') on default action"),
273 /* Warn if there is no default for $$ but we need one. */
277 complain (_("empty rule for typed nonterminal, and no action"));
282 /* End the currently being grown rule. */
285 grammar_rule_end (void)
287 /* Put an empty link in the list to mark the end of this rule */
288 grammar_symbol_append (NULL
);
289 grammar_current_rule_check ();
293 /* The previous action turns out the be a mid-rule action. Attach it
294 to the current rule, i.e., create a dummy symbol, attach it this
295 mid-rule action, and append this dummy nonterminal to the current
299 grammar_midrule_action (void)
301 /* Since the action was written out with this rule's number, we must
302 give the new rule this number by inserting the new rule before
305 /* Make a dummy nonterminal, a gensym. */
306 symbol_t
*sdummy
= gensym ();
307 symbol_list
*midrule_action
= symbol_list_new (sdummy
);
309 /* Make a new rule, whose body is empty, before the current one, so
310 that the action just read can belong to it. */
313 /* Attach its lineno to that of the host rule. */
314 midrule_action
->line
= current_rule
->line
;
315 /* Move the action from the host rule to this one. */
316 midrule_action
->action
= current_rule
->action
;
317 midrule_action
->action_line
= current_rule
->action_line
;
318 current_rule
->action
= NULL
;
321 previous_rule
->next
= midrule_action
;
323 grammar
= midrule_action
;
325 /* End of the rule. */
326 previous_rule
= symbol_list_new (NULL
);
327 previous_rule
->next
= current_rule
;
329 midrule_action
->next
= previous_rule
;
331 /* Insert the dummy generated by that rule into this rule. */
333 grammar_symbol_append (sdummy
);
336 /* Set the precedence symbol of the current rule to PRECSYM. */
339 grammar_current_rule_prec_set (symbol_t
*precsym
)
341 if (current_rule
->ruleprec
)
342 complain (_("two @prec's in a row"));
343 current_rule
->ruleprec
= precsym
;
346 /* Attach a SYMBOL to the current rule. If needed, move the previous
347 action as a mid-rule action. */
350 grammar_current_rule_symbol_append (symbol_t
*symbol
)
352 if (current_rule
->action
)
353 grammar_midrule_action ();
355 grammar_symbol_append (symbol
);
359 /* Attach an ACTION to the current rule. If needed, move the previous
360 action as a mid-rule action. */
363 grammar_current_rule_action_append (const char *action
, int action_line
)
365 if (current_rule
->action
)
366 grammar_midrule_action ();
367 current_rule
->action
= action
;
368 current_rule
->action_line
= action_line
;
372 /*---------------------------------------------------------------.
373 | Convert the rules into the representation using RRHS, RLHS and |
375 `---------------------------------------------------------------*/
384 ritem
= XCALLOC (item_number_t
, nritems
);
385 rules
= XCALLOC (rule_t
, nrules
) - 1;
393 symbol_t
*ruleprec
= p
->ruleprec
;
394 rules
[ruleno
].user_number
= ruleno
;
395 rules
[ruleno
].number
= ruleno
;
396 rules
[ruleno
].lhs
= p
->sym
;
397 rules
[ruleno
].rhs
= ritem
+ itemno
;
398 rules
[ruleno
].line
= p
->line
;
399 rules
[ruleno
].useful
= TRUE
;
400 rules
[ruleno
].action
= p
->action
;
401 rules
[ruleno
].action_line
= p
->action_line
;
406 /* item_number_t = symbol_number_t.
407 But the former needs to contain more: negative rule numbers. */
408 ritem
[itemno
++] = symbol_number_as_item_number (p
->sym
->number
);
409 /* A rule gets by default the precedence and associativity
410 of the last token in it. */
411 if (p
->sym
->class == token_sym
)
412 rules
[ruleno
].prec
= p
->sym
;
417 /* If this rule has a %prec,
418 the specified symbol's precedence replaces the default. */
421 rules
[ruleno
].precsym
= ruleprec
;
422 rules
[ruleno
].prec
= ruleprec
;
424 ritem
[itemno
++] = -ruleno
;
431 assert (itemno
== nritems
);
434 ritem_print (stderr
);
437 /*------------------------------------------------------------------.
438 | Read in the grammar specification and record it in the format |
439 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
440 | in each case forming the body of a C function (YYACTION) which |
441 | contains a switch statement to decide which action to execute. |
442 `------------------------------------------------------------------*/
447 gram_control_t gram_control
;
450 /* Initialize the symbol table. */
453 /* Construct the axiom symbol. */
454 axiom
= getsym ("$axiom");
455 axiom
->class = nterm_sym
;
456 axiom
->number
= nvars
++;
458 /* Construct the error token */
459 errtoken
= getsym ("error");
460 errtoken
->class = token_sym
;
461 errtoken
->number
= ntokens
++;
463 /* Construct a token that represents all undefined literal tokens.
464 It is always token number 2. */
465 undeftoken
= getsym ("$undefined.");
466 undeftoken
->class = token_sym
;
467 undeftoken
->number
= ntokens
++;
469 /* Initialize the obstacks. */
470 obstack_init (&pre_prologue_obstack
);
471 obstack_init (&post_prologue_obstack
);
473 finput
= xfopen (infile
, "r");
476 gram_debug
= !!getenv ("parse");
477 gram__flex_debug
= !!getenv ("scan");
478 scanner_initialize ();
479 gram_parse (&gram_control
);
481 /* Grammar has been read. Do some checking */
483 fatal (_("no rules in the input grammar"));
485 /* Report any undefined symbols and consider them nonterminals. */
486 symbols_check_defined ();
488 /* If the user did not define her EOFTOKEN, do it now. */
491 eoftoken
= getsym ("$");
492 eoftoken
->class = token_sym
;
493 eoftoken
->number
= 0;
494 /* Value specified by POSIX. */
495 eoftoken
->user_token_number
= 0;
498 /* Insert the initial rule, which line is that of the first rule
499 (not that of the start symbol):
501 axiom: %start EOF. */
503 symbol_list
*p
= symbol_list_new (axiom
);
504 p
->line
= grammar
->line
;
505 p
->next
= symbol_list_new (startsymbol
);
506 p
->next
->next
= symbol_list_new (eoftoken
);
507 p
->next
->next
->next
= symbol_list_new (NULL
);
508 p
->next
->next
->next
->next
= grammar
;
514 if (nsyms
> SHRT_MAX
)
515 fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
518 assert (nsyms
== ntokens
+ nvars
);
522 /* Assign the symbols their symbol numbers. Write #defines for the
523 token symbols into FDEFINES if requested. */
526 /* Convert the grammar into the format described in gram.h. */
529 /* The grammar as a symbol_list is no longer needed. */
530 LIST_FREE (symbol_list
, grammar
);