]> git.saurik.com Git - bison.git/blob - src/reader.c
* src/scan-gram.l (YY_OBS_INIT): Remove, replace with...
[bison.git] / src / reader.c
1 /* Input parser for bison
2 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002
3 Free Software Foundation, Inc.
4
5 This file is part of Bison, the GNU Compiler Compiler.
6
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)
10 any later version.
11
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.
16
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. */
21
22
23 #include "system.h"
24 #include "quotearg.h"
25 #include "quote.h"
26 #include "getargs.h"
27 #include "files.h"
28 #include "symtab.h"
29 #include "options.h"
30 #include "gram.h"
31 #include "complain.h"
32 #include "output.h"
33 #include "reader.h"
34 #include "conflicts.h"
35 #include "muscle_tab.h"
36
37 int lineno;
38 static symbol_list *grammar = NULL;
39 static int start_flag = 0;
40
41 /* Nonzero if %union has been seen. */
42 int typed = 0;
43
44 static symbol_list *
45 symbol_list_new (symbol_t *sym)
46 {
47 symbol_list *res = XMALLOC (symbol_list, 1);
48 res->next = NULL;
49 res->sym = sym;
50 res->line = lineno;
51 res->action = NULL;
52 res->action_line = 0;
53 res->ruleprec = NULL;
54 return res;
55 }
56
57 \f
58 /*--------------------------------------------------------------.
59 | Get the data type (alternative in the union) of the value for |
60 | symbol N in rule RULE. |
61 `--------------------------------------------------------------*/
62
63 char *
64 get_type_name (int n, symbol_list *rule)
65 {
66 int i;
67 symbol_list *rp;
68
69 if (n < 0)
70 {
71 complain (_("invalid $ value"));
72 return NULL;
73 }
74
75 rp = rule;
76 i = 0;
77
78 while (i < n)
79 {
80 rp = rp->next;
81 if (rp == NULL || rp->sym == NULL)
82 {
83 complain (_("invalid $ value"));
84 return NULL;
85 }
86 ++i;
87 }
88
89 return rp->sym->type_name;
90 }
91
92
93 /*-----------------------.
94 | Set the start symbol. |
95 `-----------------------*/
96
97 void
98 grammar_start_symbol_set (symbol_t *s)
99 {
100 if (start_flag)
101 complain (_("multiple %s declarations"), "%start");
102 else
103 {
104 start_flag = 1;
105 startsymbol = s;
106 }
107 }
108
109
110 /*----------------------------------------------------------------.
111 | There are two prologues: one before %union, one after. Augment |
112 | the current one. |
113 `----------------------------------------------------------------*/
114
115 void
116 prologue_augment (const char *prologue, int location)
117 {
118 struct obstack *oout =
119 !typed ? &pre_prologue_obstack : &post_prologue_obstack;
120
121 if (!no_lines_flag)
122 {
123 obstack_fgrow2 (oout, muscle_find ("linef"),
124 location, quotearg_style (c_quoting_style,
125 muscle_find ("filename")));
126 }
127 obstack_sgrow (oout, prologue);
128 }
129
130
131
132
133 /*----------------------.
134 | Handle the epilogue. |
135 `----------------------*/
136
137 void
138 epilogue_set (const char *epilogue, int location)
139 {
140 struct obstack el_obstack;
141 obstack_init (&el_obstack);
142
143 if (!no_lines_flag)
144 {
145 obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
146 location, quotearg_style (c_quoting_style,
147 muscle_find ("filename")));
148 }
149 obstack_sgrow (&el_obstack, epilogue);
150 obstack_1grow (&el_obstack, 0);
151 muscle_insert ("epilogue", obstack_finish (&el_obstack));
152 }
153
154
155 \f
156
157 /*-------------------------------------------------------------------.
158 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
159 | with the user's names. |
160 `-------------------------------------------------------------------*/
161
162 static symbol_t *
163 gensym (void)
164 {
165 /* Incremented for each generated symbol */
166 static int gensym_count = 0;
167 static char buf[256];
168
169 symbol_t *sym;
170
171 sprintf (buf, "@%d", ++gensym_count);
172 sym = getsym (buf);
173 sym->class = nterm_sym;
174 sym->number = nvars++;
175 return sym;
176 }
177 \f
178 /*-------------------------------------------------------------------.
179 | Parse the input grammar into a one symbol_list structure. Each |
180 | rule is represented by a sequence of symbols: the left hand side |
181 | followed by the contents of the right hand side, followed by a |
182 | null pointer instead of a symbol to terminate the rule. The next |
183 | symbol is the lhs of the following rule. |
184 | |
185 | All actions are copied out, labelled by the rule number they apply |
186 | to. |
187 | |
188 | Bison used to allow some %directives in the rules sections, but |
189 | this is no longer consider appropriate: (i) the documented grammar |
190 | doesn't claim it, (ii), it would promote bad style, (iii), error |
191 | recovery for %directives consists in skipping the junk until a `%' |
192 | is seen and helrp synchronizing. This scheme is definitely wrong |
193 | in the rules section. |
194 `-------------------------------------------------------------------*/
195
196 /* The (currently) last symbol of GRAMMAR. */
197 symbol_list *grammar_end = NULL;
198
199 /* Append S to the GRAMMAR. */
200 void
201 grammar_symbol_append (symbol_t *s)
202 {
203 symbol_list *p = symbol_list_new (s);
204
205 if (grammar_end)
206 grammar_end->next = p;
207 else
208 grammar = p;
209
210 grammar_end = p;
211 }
212
213 /* The rule currently being defined, and the previous rule. Point to
214 the first symbol of each list: their lhs. */
215 symbol_list *current_rule = NULL;
216 symbol_list *previous_rule = NULL;
217
218
219 /* Create a new rule for LHS in to the GRAMMAR. */
220
221 void
222 grammar_rule_begin (symbol_t *lhs)
223 {
224 if (!start_flag)
225 {
226 startsymbol = lhs;
227 start_flag = 1;
228 }
229
230 /* Start a new rule and record its lhs. */
231 ++nrules;
232 ++nritems;
233
234 previous_rule = grammar_end;
235 grammar_symbol_append (lhs);
236 current_rule = grammar_end;
237
238 /* Mark the rule's lhs as a nonterminal if not already so. */
239
240 if (lhs->class == unknown_sym)
241 {
242 lhs->class = nterm_sym;
243 lhs->number = nvars;
244 ++nvars;
245 }
246 else if (lhs->class == token_sym)
247 complain (_("rule given for %s, which is a token"), lhs->tag);
248 }
249
250 /* Check that the last rule (CURRENT_RULE) is properly defined. For
251 instance, there should be no type clash on the default action. */
252
253 static void
254 grammar_current_rule_check (void)
255 {
256 symbol_t *lhs = current_rule->sym;
257 symbol_t *first_rhs = current_rule->next->sym;
258
259 /* If there is an action, then there is nothing we can do: the user
260 is allowed to shoot in her foot. */
261 if (current_rule->action)
262 return;
263
264 /* If $$ is being set in default way, report if any type mismatch.
265 */
266 if (first_rhs)
267 {
268 const char *lhs_type = lhs->type_name ? lhs->type_name : "";
269 const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
270 if (strcmp (lhs_type, rhs_type))
271 complain (_("type clash (`%s' `%s') on default action"),
272 lhs_type, rhs_type);
273 }
274 /* Warn if there is no default for $$ but we need one. */
275 else
276 {
277 if (lhs->type_name)
278 complain (_("empty rule for typed nonterminal, and no action"));
279 }
280 }
281
282
283 /* End the currently being grown rule. */
284
285 void
286 grammar_rule_end (void)
287 {
288 /* Put an empty link in the list to mark the end of this rule */
289 grammar_symbol_append (NULL);
290 grammar_current_rule_check ();
291 }
292
293
294 /* The previous action turns out the be a mid-rule action. Attach it
295 to the current rule, i.e., create a dummy symbol, attach it this
296 mid-rule action, and append this dummy nonterminal to the current
297 rule. */
298
299 void
300 grammar_midrule_action (void)
301 {
302 /* Since the action was written out with this rule's number, we must
303 give the new rule this number by inserting the new rule before
304 it. */
305
306 /* Make a dummy nonterminal, a gensym. */
307 symbol_t *sdummy = gensym ();
308 symbol_list *midrule_action = symbol_list_new (sdummy);
309
310 /* Make a new rule, whose body is empty, before the current one, so
311 that the action just read can belong to it. */
312 ++nrules;
313 ++nritems;
314 /* Attach its lineno to that of the host rule. */
315 midrule_action->line = current_rule->line;
316 /* Move the action from the host rule to this one. */
317 midrule_action->action = current_rule->action;
318 midrule_action->action_line = current_rule->action_line;
319 current_rule->action = NULL;
320
321 if (previous_rule)
322 previous_rule->next = midrule_action;
323 else
324 grammar = midrule_action;
325
326 /* End of the rule. */
327 previous_rule = symbol_list_new (NULL);
328 previous_rule->next = current_rule;
329
330 midrule_action->next = previous_rule;
331
332 /* Insert the dummy generated by that rule into this rule. */
333 ++nritems;
334 grammar_symbol_append (sdummy);
335 }
336
337 /* Set the precedence symbol of the current rule to PRECSYM. */
338
339 void
340 grammar_current_rule_prec_set (symbol_t *precsym)
341 {
342 if (current_rule->ruleprec)
343 complain (_("two @prec's in a row"));
344 current_rule->ruleprec = precsym;
345 }
346
347 /* Attach a SYMBOL to the current rule. If needed, move the previous
348 action as a mid-rule action. */
349
350 void
351 grammar_current_rule_symbol_append (symbol_t *symbol)
352 {
353 if (current_rule->action)
354 grammar_midrule_action ();
355 ++nritems;
356 grammar_symbol_append (symbol);
357 }
358
359
360 /* Attach an ACTION to the current rule. If needed, move the previous
361 action as a mid-rule action. */
362
363 void
364 grammar_current_rule_action_append (const char *action, int action_line)
365 {
366 if (current_rule->action)
367 grammar_midrule_action ();
368 current_rule->action = action;
369 current_rule->action_line = action_line;
370 }
371
372 \f
373 /*---------------------------------------------------------------.
374 | Convert the rules into the representation using RRHS, RLHS and |
375 | RITEM. |
376 `---------------------------------------------------------------*/
377
378 static void
379 packgram (void)
380 {
381 unsigned int itemno;
382 int ruleno;
383 symbol_list *p;
384
385 ritem = XCALLOC (item_number_t, nritems);
386 rules = XCALLOC (rule_t, nrules) - 1;
387
388 itemno = 0;
389 ruleno = 1;
390
391 p = grammar;
392 while (p)
393 {
394 symbol_t *ruleprec = p->ruleprec;
395 rules[ruleno].user_number = ruleno;
396 rules[ruleno].number = ruleno;
397 rules[ruleno].lhs = p->sym;
398 rules[ruleno].rhs = ritem + itemno;
399 rules[ruleno].line = p->line;
400 rules[ruleno].useful = TRUE;
401 rules[ruleno].action = p->action;
402 rules[ruleno].action_line = p->action_line;
403
404 p = p->next;
405 while (p && p->sym)
406 {
407 /* item_number_t = symbol_number_t.
408 But the former needs to contain more: negative rule numbers. */
409 ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
410 /* A rule gets by default the precedence and associativity
411 of the last token in it. */
412 if (p->sym->class == token_sym)
413 rules[ruleno].prec = p->sym;
414 if (p)
415 p = p->next;
416 }
417
418 /* If this rule has a %prec,
419 the specified symbol's precedence replaces the default. */
420 if (ruleprec)
421 {
422 rules[ruleno].precsym = ruleprec;
423 rules[ruleno].prec = ruleprec;
424 }
425 ritem[itemno++] = -ruleno;
426 ++ruleno;
427
428 if (p)
429 p = p->next;
430 }
431
432 assert (itemno == nritems);
433
434 if (trace_flag)
435 ritem_print (stderr);
436 }
437 \f
438 /*------------------------------------------------------------------.
439 | Read in the grammar specification and record it in the format |
440 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
441 | in each case forming the body of a C function (YYACTION) which |
442 | contains a switch statement to decide which action to execute. |
443 `------------------------------------------------------------------*/
444
445 void
446 reader (void)
447 {
448 gram_control_t gram_control;
449 lineno = 1;
450
451 /* Initialize the muscle obstack. */
452 obstack_init (&muscle_obstack);
453
454 /* Initialize the symbol table. */
455 symbols_new ();
456
457 /* Construct the axiom symbol. */
458 axiom = getsym ("$axiom");
459 axiom->class = nterm_sym;
460 axiom->number = nvars++;
461
462 /* Construct the error token */
463 errtoken = getsym ("error");
464 errtoken->class = token_sym;
465 errtoken->number = ntokens++;
466
467 /* Construct a token that represents all undefined literal tokens.
468 It is always token number 2. */
469 undeftoken = getsym ("$undefined.");
470 undeftoken->class = token_sym;
471 undeftoken->number = ntokens++;
472
473 /* Initialize the obstacks. */
474 obstack_init (&action_obstack);
475 obstack_init (&output_obstack);
476 obstack_init (&pre_prologue_obstack);
477 obstack_init (&post_prologue_obstack);
478
479 finput = xfopen (infile, "r");
480 gram_in = finput;
481
482 gram_debug = !!getenv ("parse");
483 gram__flex_debug = !!getenv ("scan");
484 scanner_initialize ();
485 gram_parse (&gram_control);
486
487 /* Grammar has been read. Do some checking */
488 if (nrules == 0)
489 fatal (_("no rules in the input grammar"));
490
491 /* Report any undefined symbols and consider them nonterminals. */
492 symbols_check_defined ();
493
494 /* If the user did not define her EOFTOKEN, do it now. */
495 if (!eoftoken)
496 {
497 eoftoken = getsym ("$");
498 eoftoken->class = token_sym;
499 eoftoken->number = 0;
500 /* Value specified by POSIX. */
501 eoftoken->user_token_number = 0;
502 }
503
504 /* Insert the initial rule, which line is that of the first rule
505 (not that of the start symbol):
506
507 axiom: %start EOF. */
508 {
509 symbol_list *p = symbol_list_new (axiom);
510 p->line = grammar->line;
511 p->next = symbol_list_new (startsymbol);
512 p->next->next = symbol_list_new (eoftoken);
513 p->next->next->next = symbol_list_new (NULL);
514 p->next->next->next->next = grammar;
515 nrules += 1;
516 nritems += 3;
517 grammar = p;
518 }
519
520 if (nsyms > SHRT_MAX)
521 fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
522 SHRT_MAX);
523
524 assert (nsyms == ntokens + nvars);
525
526 xfclose (finput);
527
528 /* Assign the symbols their symbol numbers. Write #defines for the
529 token symbols into FDEFINES if requested. */
530 symbols_pack ();
531
532 /* Convert the grammar into the format described in gram.h. */
533 packgram ();
534
535 /* The grammar as a symbol_list is no longer needed. */
536 LIST_FREE (symbol_list, grammar);
537 }