]> git.saurik.com Git - bison.git/blob - src/reader.c
12bb2d67f82048d1e1f321c22ae8be32a44cd8ab
[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, location_t 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.first_line,
125 quotearg_style (c_quoting_style,
126 muscle_find ("filename")));
127 }
128 obstack_sgrow (oout, prologue);
129 }
130
131
132
133
134 /*----------------------.
135 | Handle the epilogue. |
136 `----------------------*/
137
138 void
139 epilogue_set (const char *epilogue, location_t location)
140 {
141 struct obstack el_obstack;
142 obstack_init (&el_obstack);
143
144 if (!no_lines_flag)
145 {
146 obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
147 location.first_line,
148 quotearg_style (c_quoting_style,
149 muscle_find ("filename")));
150 }
151 obstack_sgrow (&el_obstack, epilogue);
152 obstack_1grow (&el_obstack, 0);
153 muscle_insert ("epilogue", obstack_finish (&el_obstack));
154 }
155
156
157 \f
158
159 /*-------------------------------------------------------------------.
160 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
161 | with the user's names. |
162 `-------------------------------------------------------------------*/
163
164 static symbol_t *
165 gensym (void)
166 {
167 /* Incremented for each generated symbol */
168 static int gensym_count = 0;
169 static char buf[256];
170
171 symbol_t *sym;
172
173 sprintf (buf, "@%d", ++gensym_count);
174 sym = getsym (buf);
175 sym->class = nterm_sym;
176 sym->number = nvars++;
177 return sym;
178 }
179 \f
180 /*-------------------------------------------------------------------.
181 | Parse the input grammar into a one symbol_list structure. Each |
182 | rule is represented by a sequence of symbols: the left hand side |
183 | followed by the contents of the right hand side, followed by a |
184 | null pointer instead of a symbol to terminate the rule. The next |
185 | symbol is the lhs of the following rule. |
186 | |
187 | All actions are copied out, labelled by the rule number they apply |
188 | to. |
189 | |
190 | Bison used to allow some %directives in the rules sections, but |
191 | this is no longer consider appropriate: (i) the documented grammar |
192 | doesn't claim it, (ii), it would promote bad style, (iii), error |
193 | recovery for %directives consists in skipping the junk until a `%' |
194 | is seen and helrp synchronizing. This scheme is definitely wrong |
195 | in the rules section. |
196 `-------------------------------------------------------------------*/
197
198 /* The (currently) last symbol of GRAMMAR. */
199 symbol_list *grammar_end = NULL;
200
201 /* Append S to the GRAMMAR. */
202 void
203 grammar_symbol_append (symbol_t *s)
204 {
205 symbol_list *p = symbol_list_new (s);
206
207 if (grammar_end)
208 grammar_end->next = p;
209 else
210 grammar = p;
211
212 grammar_end = p;
213 }
214
215 /* The rule currently being defined, and the previous rule. Point to
216 the first symbol of each list: their lhs. */
217 symbol_list *current_rule = NULL;
218 symbol_list *previous_rule = NULL;
219
220
221 /* Create a new rule for LHS in to the GRAMMAR. */
222
223 void
224 grammar_rule_begin (symbol_t *lhs)
225 {
226 if (!start_flag)
227 {
228 startsymbol = lhs;
229 start_flag = 1;
230 }
231
232 /* Start a new rule and record its lhs. */
233 ++nrules;
234 ++nritems;
235
236 previous_rule = grammar_end;
237 grammar_symbol_append (lhs);
238 current_rule = grammar_end;
239
240 /* Mark the rule's lhs as a nonterminal if not already so. */
241
242 if (lhs->class == unknown_sym)
243 {
244 lhs->class = nterm_sym;
245 lhs->number = nvars;
246 ++nvars;
247 }
248 else if (lhs->class == token_sym)
249 complain (_("rule given for %s, which is a token"), lhs->tag);
250 }
251
252 /* Check that the last rule (CURRENT_RULE) is properly defined. For
253 instance, there should be no type clash on the default action. */
254
255 static void
256 grammar_current_rule_check (void)
257 {
258 symbol_t *lhs = current_rule->sym;
259 symbol_t *first_rhs = current_rule->next->sym;
260
261 /* If there is an action, then there is nothing we can do: the user
262 is allowed to shoot in her foot. */
263 if (current_rule->action)
264 return;
265
266 /* If $$ is being set in default way, report if any type mismatch.
267 */
268 if (first_rhs)
269 {
270 const char *lhs_type = lhs->type_name ? lhs->type_name : "";
271 const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
272 if (strcmp (lhs_type, rhs_type))
273 complain (_("type clash (`%s' `%s') on default action"),
274 lhs_type, rhs_type);
275 }
276 /* Warn if there is no default for $$ but we need one. */
277 else
278 {
279 if (lhs->type_name)
280 complain (_("empty rule for typed nonterminal, and no action"));
281 }
282 }
283
284
285 /* End the currently being grown rule. */
286
287 void
288 grammar_rule_end (void)
289 {
290 /* Put an empty link in the list to mark the end of this rule */
291 grammar_symbol_append (NULL);
292 grammar_current_rule_check ();
293 }
294
295
296 /* The previous action turns out the be a mid-rule action. Attach it
297 to the current rule, i.e., create a dummy symbol, attach it this
298 mid-rule action, and append this dummy nonterminal to the current
299 rule. */
300
301 void
302 grammar_midrule_action (void)
303 {
304 /* Since the action was written out with this rule's number, we must
305 give the new rule this number by inserting the new rule before
306 it. */
307
308 /* Make a dummy nonterminal, a gensym. */
309 symbol_t *sdummy = gensym ();
310 symbol_list *midrule_action = symbol_list_new (sdummy);
311
312 /* Make a new rule, whose body is empty, before the current one, so
313 that the action just read can belong to it. */
314 ++nrules;
315 ++nritems;
316 /* Attach its lineno to that of the host rule. */
317 midrule_action->line = current_rule->line;
318 /* Move the action from the host rule to this one. */
319 midrule_action->action = current_rule->action;
320 midrule_action->action_line = current_rule->action_line;
321 current_rule->action = NULL;
322
323 if (previous_rule)
324 previous_rule->next = midrule_action;
325 else
326 grammar = midrule_action;
327
328 /* End of the rule. */
329 previous_rule = symbol_list_new (NULL);
330 previous_rule->next = current_rule;
331
332 midrule_action->next = previous_rule;
333
334 /* Insert the dummy generated by that rule into this rule. */
335 ++nritems;
336 grammar_symbol_append (sdummy);
337 }
338
339 /* Set the precedence symbol of the current rule to PRECSYM. */
340
341 void
342 grammar_current_rule_prec_set (symbol_t *precsym)
343 {
344 if (current_rule->ruleprec)
345 complain (_("two @prec's in a row"));
346 current_rule->ruleprec = precsym;
347 }
348
349 /* Attach a SYMBOL to the current rule. If needed, move the previous
350 action as a mid-rule action. */
351
352 void
353 grammar_current_rule_symbol_append (symbol_t *symbol)
354 {
355 if (current_rule->action)
356 grammar_midrule_action ();
357 ++nritems;
358 grammar_symbol_append (symbol);
359 }
360
361
362 /* Attach an ACTION to the current rule. If needed, move the previous
363 action as a mid-rule action. */
364
365 void
366 grammar_current_rule_action_append (const char *action, int action_line)
367 {
368 if (current_rule->action)
369 grammar_midrule_action ();
370 current_rule->action = action;
371 current_rule->action_line = action_line;
372 }
373
374 \f
375 /*---------------------------------------------------------------.
376 | Convert the rules into the representation using RRHS, RLHS and |
377 | RITEM. |
378 `---------------------------------------------------------------*/
379
380 static void
381 packgram (void)
382 {
383 unsigned int itemno;
384 int ruleno;
385 symbol_list *p;
386
387 ritem = XCALLOC (item_number_t, nritems);
388 rules = XCALLOC (rule_t, nrules) - 1;
389
390 itemno = 0;
391 ruleno = 1;
392
393 p = grammar;
394 while (p)
395 {
396 symbol_t *ruleprec = p->ruleprec;
397 rules[ruleno].user_number = ruleno;
398 rules[ruleno].number = ruleno;
399 rules[ruleno].lhs = p->sym;
400 rules[ruleno].rhs = ritem + itemno;
401 rules[ruleno].line = p->line;
402 rules[ruleno].useful = TRUE;
403 rules[ruleno].action = p->action;
404 rules[ruleno].action_line = p->action_line;
405
406 p = p->next;
407 while (p && p->sym)
408 {
409 /* item_number_t = symbol_number_t.
410 But the former needs to contain more: negative rule numbers. */
411 ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
412 /* A rule gets by default the precedence and associativity
413 of the last token in it. */
414 if (p->sym->class == token_sym)
415 rules[ruleno].prec = p->sym;
416 if (p)
417 p = p->next;
418 }
419
420 /* If this rule has a %prec,
421 the specified symbol's precedence replaces the default. */
422 if (ruleprec)
423 {
424 rules[ruleno].precsym = ruleprec;
425 rules[ruleno].prec = ruleprec;
426 }
427 ritem[itemno++] = -ruleno;
428 ++ruleno;
429
430 if (p)
431 p = p->next;
432 }
433
434 assert (itemno == nritems);
435
436 if (trace_flag)
437 ritem_print (stderr);
438 }
439 \f
440 /*------------------------------------------------------------------.
441 | Read in the grammar specification and record it in the format |
442 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
443 | in each case forming the body of a C function (YYACTION) which |
444 | contains a switch statement to decide which action to execute. |
445 `------------------------------------------------------------------*/
446
447 void
448 reader (void)
449 {
450 gram_control_t gram_control;
451 lineno = 1;
452
453 /* Initialize the muscle obstack. */
454 obstack_init (&muscle_obstack);
455
456 /* Initialize the symbol table. */
457 symbols_new ();
458
459 /* Construct the axiom symbol. */
460 axiom = getsym ("$axiom");
461 axiom->class = nterm_sym;
462 axiom->number = nvars++;
463
464 /* Construct the error token */
465 errtoken = getsym ("error");
466 errtoken->class = token_sym;
467 errtoken->number = ntokens++;
468
469 /* Construct a token that represents all undefined literal tokens.
470 It is always token number 2. */
471 undeftoken = getsym ("$undefined.");
472 undeftoken->class = token_sym;
473 undeftoken->number = ntokens++;
474
475 /* Initialize the obstacks. */
476 obstack_init (&action_obstack);
477 obstack_init (&output_obstack);
478 obstack_init (&pre_prologue_obstack);
479 obstack_init (&post_prologue_obstack);
480
481 finput = xfopen (infile, "r");
482 gram_in = finput;
483
484 gram_debug = !!getenv ("parse");
485 gram__flex_debug = !!getenv ("scan");
486 scanner_initialize ();
487 gram_parse (&gram_control);
488
489 /* Grammar has been read. Do some checking */
490 if (nrules == 0)
491 fatal (_("no rules in the input grammar"));
492
493 /* Report any undefined symbols and consider them nonterminals. */
494 symbols_check_defined ();
495
496 /* If the user did not define her EOFTOKEN, do it now. */
497 if (!eoftoken)
498 {
499 eoftoken = getsym ("$");
500 eoftoken->class = token_sym;
501 eoftoken->number = 0;
502 /* Value specified by POSIX. */
503 eoftoken->user_token_number = 0;
504 }
505
506 /* Insert the initial rule, which line is that of the first rule
507 (not that of the start symbol):
508
509 axiom: %start EOF. */
510 {
511 symbol_list *p = symbol_list_new (axiom);
512 p->line = grammar->line;
513 p->next = symbol_list_new (startsymbol);
514 p->next->next = symbol_list_new (eoftoken);
515 p->next->next->next = symbol_list_new (NULL);
516 p->next->next->next->next = grammar;
517 nrules += 1;
518 nritems += 3;
519 grammar = p;
520 }
521
522 if (nsyms > SHRT_MAX)
523 fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
524 SHRT_MAX);
525
526 assert (nsyms == ntokens + nvars);
527
528 xfclose (finput);
529
530 /* Assign the symbols their symbol numbers. Write #defines for the
531 token symbols into FDEFINES if requested. */
532 symbols_pack ();
533
534 /* Convert the grammar into the format described in gram.h. */
535 packgram ();
536
537 /* The grammar as a symbol_list is no longer needed. */
538 LIST_FREE (symbol_list, grammar);
539 }