]> git.saurik.com Git - bison.git/blob - src/reduce.c
Kill GCC warnings.
[bison.git] / src / reduce.c
1 /* Grammar reduction for Bison.
2 Copyright (C) 1988, 1989, 2000, 2001, 2002 Free Software Foundation, Inc.
3
4 This file is part of Bison, the GNU Compiler Compiler.
5
6 Bison is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2, or (at your option)
9 any later version.
10
11 Bison is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
15
16 You should have received a copy of the GNU General Public License
17 along with Bison; see the file COPYING. If not, write to
18 the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 Boston, MA 02111-1307, USA. */
20
21
22 /* Reduce the grammar: Find and eliminate unreachable terminals,
23 nonterminals, and productions. David S. Bakin. */
24
25 /* Don't eliminate unreachable terminals: They may be used by the
26 user's parser. */
27
28 #include "system.h"
29 #include "quotearg.h"
30 #include "getargs.h"
31 #include "files.h"
32 #include "symtab.h"
33 #include "gram.h"
34 #include "complain.h"
35 #include "reduce.h"
36 #include "reader.h"
37 #include "getargs.h"
38 #include "bitset.h"
39
40 /* Set of all nonterminals which are not useless. */
41 static bitset N;
42
43 /* Set of all rules which have no useless nonterminals in their RHS. */
44 static bitset P;
45
46 /* Set of all accessible symbols. */
47 static bitset V;
48
49 /* Set of symbols used to define rule precedence (so they are
50 `useless', but no warning should be issued). */
51 static bitset V1;
52
53 static int nuseful_productions;
54 int nuseless_productions;
55 static int nuseful_nonterminals;
56 int nuseless_nonterminals;
57 \f
58 /*-------------------------------------------------------------------.
59 | Another way to do this would be with a set for each production and |
60 | then do subset tests against N0, but even for the C grammar the |
61 | whole reducing process takes only 2 seconds on my 8Mhz AT. |
62 `-------------------------------------------------------------------*/
63
64 static bool
65 useful_production (int i, bitset N0)
66 {
67 item_number_t *r;
68 short n;
69
70 /* A production is useful if all of the nonterminals in its appear
71 in the set of useful nonterminals. */
72
73 for (r = rules[i].rhs; *r >= 0; r++)
74 if (ISVAR (n = *r) && !bitset_test (N0, n - ntokens))
75 return FALSE;
76 return TRUE;
77 }
78
79
80 /*---------------------------------------------------------.
81 | Remember that rules are 1-origin, symbols are 0-origin. |
82 `---------------------------------------------------------*/
83
84 static void
85 useless_nonterminals (void)
86 {
87 bitset Np, Ns;
88 int i;
89
90 /* N is set as built. Np is set being built this iteration. P is
91 set of all productions which have a RHS all in N. */
92
93 Np = bitset_create (nvars, BITSET_FIXED);
94
95
96 /* The set being computed is a set of nonterminals which can derive
97 the empty string or strings consisting of all terminals. At each
98 iteration a nonterminal is added to the set if there is a
99 production with that nonterminal as its LHS for which all the
100 nonterminals in its RHS are already in the set. Iterate until
101 the set being computed remains unchanged. Any nonterminals not
102 in the set at that point are useless in that they will never be
103 used in deriving a sentence of the language.
104
105 This iteration doesn't use any special traversal over the
106 productions. A set is kept of all productions for which all the
107 nonterminals in the RHS are in useful. Only productions not in
108 this set are scanned on each iteration. At the end, this set is
109 saved to be used when finding useful productions: only
110 productions in this set will appear in the final grammar. */
111
112 while (1)
113 {
114 bitset_copy (Np, N);
115 for (i = 1; i < nrules + 1; i++)
116 if (!bitset_test (P, i)
117 && useful_production (i, N))
118 {
119 bitset_set (Np, rules[i].lhs->number - ntokens);
120 bitset_set (P, i);
121 }
122 if (bitset_equal_p (N, Np))
123 break;
124 Ns = Np;
125 Np = N;
126 N = Ns;
127 }
128 bitset_free (N);
129 N = Np;
130 }
131
132
133 static void
134 inaccessable_symbols (void)
135 {
136 bitset Vp, Vs, Pp;
137 int i;
138 short t;
139 item_number_t *r;
140
141 /* Find out which productions are reachable and which symbols are
142 used. Starting with an empty set of productions and a set of
143 symbols which only has the start symbol in it, iterate over all
144 productions until the set of productions remains unchanged for an
145 iteration. For each production which has a LHS in the set of
146 reachable symbols, add the production to the set of reachable
147 productions, and add all of the nonterminals in the RHS of the
148 production to the set of reachable symbols.
149
150 Consider only the (partially) reduced grammar which has only
151 nonterminals in N and productions in P.
152
153 The result is the set P of productions in the reduced grammar,
154 and the set V of symbols in the reduced grammar.
155
156 Although this algorithm also computes the set of terminals which
157 are reachable, no terminal will be deleted from the grammar. Some
158 terminals might not be in the grammar but might be generated by
159 semantic routines, and so the user might want them available with
160 specified numbers. (Is this true?) However, the nonreachable
161 terminals are printed (if running in verbose mode) so that the
162 user can know. */
163
164 Vp = bitset_create (nsyms, BITSET_FIXED);
165 Pp = bitset_create (nrules + 1, BITSET_FIXED);
166
167 /* If the start symbol isn't useful, then nothing will be useful. */
168 if (bitset_test (N, start_symbol - ntokens))
169 {
170 bitset_set (V, start_symbol);
171
172 while (1)
173 {
174 bitset_copy (Vp, V);
175 for (i = 1; i < nrules + 1; i++)
176 {
177 if (!bitset_test (Pp, i)
178 && bitset_test (P, i)
179 && bitset_test (V, rules[i].lhs->number))
180 {
181 for (r = rules[i].rhs; *r >= 0; r++)
182 if (ISTOKEN (t = *r) || bitset_test (N, t - ntokens))
183 bitset_set (Vp, t);
184 bitset_set (Pp, i);
185 }
186 }
187 if (bitset_equal_p (V, Vp))
188 break;
189 Vs = Vp;
190 Vp = V;
191 V = Vs;
192 }
193 }
194
195 bitset_free (V);
196 V = Vp;
197
198 /* Tokens 0, 1, and 2 are internal to Bison. Consider them useful. */
199 bitset_set (V, eoftoken->number); /* end-of-input token */
200 bitset_set (V, errtoken->number); /* error token */
201 bitset_set (V, undeftoken->number); /* some undefined token */
202
203 bitset_free (P);
204 P = Pp;
205
206 nuseful_productions = bitset_count (P);
207 nuseless_productions = nrules - nuseful_productions;
208
209 nuseful_nonterminals = 0;
210 for (i = ntokens; i < nsyms; i++)
211 if (bitset_test (V, i))
212 nuseful_nonterminals++;
213 nuseless_nonterminals = nvars - nuseful_nonterminals;
214
215 /* A token that was used in %prec should not be warned about. */
216 for (i = 1; i < nrules + 1; i++)
217 if (rules[i].precsym != 0)
218 bitset_set (V1, rules[i].precsym->number);
219 }
220
221
222 /*-------------------------------------------------------------------.
223 | Put the useless productions at the end of RULES, and adjust NRULES |
224 | accordingly. |
225 `-------------------------------------------------------------------*/
226
227 static void
228 reduce_grammar_tables (void)
229 {
230 /* Flag useless productions. */
231 {
232 int pn;
233 for (pn = 1; pn < nrules + 1; pn++)
234 rules[pn].useful = bitset_test (P, pn);
235 }
236
237 /* Map the nonterminals to their new index: useful first, useless
238 afterwards. Kept for later report. */
239 {
240 int useful = 1;
241 int useless = nrules + 1 - nuseless_productions;
242 rule_t *rules_sorted = XMALLOC (rule_t, nrules + 1) - 1;
243 int i;
244 for (i = 1; i < nrules + 1; ++i)
245 rules_sorted[rules[i].useful ? useful++ : useless++] = rules[i];
246 free (rules + 1);
247 rules = rules_sorted;
248
249 /* Renumber the rules markers in RITEMS. */
250 for (i = 1; i < nrules + 1; ++i)
251 {
252 item_number_t *rhsp = rules[i].rhs;
253 for (/* Nothing. */; *rhsp >= 0; ++rhsp)
254 /* Nothing. */;
255 *rhsp = -i;
256 rules[i].number = i;
257 }
258 nrules -= nuseless_productions;
259 }
260
261 /* Adjust NRITEMS. */
262 {
263 int r;
264 int length;
265 for (r = nrules + 1; r < nrules + 1 + nuseless_productions; ++r)
266 {
267 length = rule_rhs_length (&rules[r]);
268 nritems -= length + 1;
269 }
270 }
271 }
272
273
274 /*------------------------------.
275 | Remove useless nonterminals. |
276 `------------------------------*/
277
278 static void
279 nonterminals_reduce (void)
280 {
281 token_number_t i, n;
282
283 /* Map the nonterminals to their new index: useful first, useless
284 afterwards. Kept for later report. */
285
286 token_number_t *nontermmap = XCALLOC (token_number_t, nvars) - ntokens;
287 n = ntokens;
288 for (i = ntokens; i < nsyms; i++)
289 if (bitset_test (V, i))
290 nontermmap[i] = n++;
291 for (i = ntokens; i < nsyms; i++)
292 if (!bitset_test (V, i))
293 nontermmap[i] = n++;
294
295
296 /* Shuffle elements of tables indexed by symbol number. */
297 {
298 symbol_t **symbols_sorted = XMALLOC (symbol_t *, nvars) - ntokens;
299
300 for (i = ntokens; i < nsyms; i++)
301 symbols[i]->number = nontermmap[i];
302 for (i = ntokens; i < nsyms; i++)
303 symbols_sorted[nontermmap[i]] = symbols[i];
304 for (i = ntokens; i < nsyms; i++)
305 symbols[i] = symbols_sorted[i];
306 free (symbols_sorted + ntokens);
307 }
308
309 {
310 int r;
311 for (r = 1; r < nrules + 1; ++r)
312 {
313 item_number_t *rhsp;
314 for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
315 if (ISVAR (*rhsp))
316 *rhsp = token_number_as_item_number (nontermmap[*rhsp]);
317 }
318 start_symbol = nontermmap[start_symbol];
319 }
320
321 nsyms -= nuseless_nonterminals;
322 nvars -= nuseless_nonterminals;
323
324 free (nontermmap + ntokens);
325 }
326
327
328 /*------------------------------------------------------------------.
329 | Output the detailed results of the reductions. For FILE.output. |
330 `------------------------------------------------------------------*/
331
332 void
333 reduce_output (FILE *out)
334 {
335 if (nuseless_nonterminals > 0)
336 {
337 int i;
338 fprintf (out, "%s\n\n", _("Useless nonterminals:"));
339 for (i = 0; i < nuseless_nonterminals; ++i)
340 fprintf (out, " %s\n", quotearg_style (escape_quoting_style,
341 symbols[nsyms + i]->tag));
342 fputs ("\n\n", out);
343 }
344
345 {
346 bool b = FALSE;
347 int i;
348 for (i = 0; i < ntokens; i++)
349 if (!bitset_test (V, i) && !bitset_test (V1, i))
350 {
351 if (!b)
352 fprintf (out, "%s\n\n", _("Terminals which are not used:"));
353 b = TRUE;
354 fprintf (out, " %s\n", quotearg_style (escape_quoting_style,
355 symbols[i]->tag));
356 }
357 if (b)
358 fputs ("\n\n", out);
359 }
360
361 if (nuseless_productions > 0)
362 {
363 int i;
364 fprintf (out, "%s\n\n", _("Useless rules:"));
365 for (i = nrules + 1; i < nuseless_productions + nrules + 1; i++)
366 {
367 item_number_t *r;
368 fprintf (out, "#%-4d ", rules[i].user_number - 1);
369 fprintf (out, "%s:", quotearg_style (escape_quoting_style,
370 rules[i].lhs->tag));
371 for (r = rules[i].rhs; *r >= 0; r++)
372 fprintf (out, " %s", quotearg_style (escape_quoting_style,
373 symbols[*r]->tag));
374 fputs (";\n", out);
375 }
376 fputs ("\n\n", out);
377 }
378 }
379 \f
380
381
382
383
384 /*-------------------------------.
385 | Report the results to STDERR. |
386 `-------------------------------*/
387
388 static void
389 reduce_print (void)
390 {
391 if (yacc_flag && nuseless_productions)
392 fprintf (stderr, ngettext ("%d rule never reduced\n",
393 "%d rules never reduced\n",
394 nuseless_productions),
395 nuseless_productions);
396
397 fprintf (stderr, _("%s contains "), infile);
398
399 if (nuseless_nonterminals > 0)
400 fprintf (stderr, ngettext ("%d useless nonterminal",
401 "%d useless nonterminals",
402 nuseless_nonterminals),
403 nuseless_nonterminals);
404
405 if (nuseless_nonterminals > 0 && nuseless_productions > 0)
406 fprintf (stderr, _(" and "));
407
408 if (nuseless_productions > 0)
409 fprintf (stderr, ngettext ("%d useless rule",
410 "%d useless rules",
411 nuseless_productions),
412 nuseless_productions);
413 fprintf (stderr, "\n");
414 fflush (stderr);
415 }
416 \f
417 void
418 reduce_grammar (void)
419 {
420 bool reduced;
421
422 /* Allocate the global sets used to compute the reduced grammar */
423
424 N = bitset_create (nvars, BITSET_FIXED);
425 P = bitset_create (nrules + 1, BITSET_FIXED);
426 V = bitset_create (nsyms, BITSET_FIXED);
427 V1 = bitset_create (nsyms, BITSET_FIXED);
428
429 useless_nonterminals ();
430 inaccessable_symbols ();
431
432 reduced = (bool) (nuseless_nonterminals + nuseless_productions > 0);
433 if (!reduced)
434 return;
435
436 reduce_print ();
437
438 if (!bitset_test (N, start_symbol - ntokens))
439 fatal (_("Start symbol %s does not derive any sentence"),
440 quotearg_style (escape_quoting_style, symbols[start_symbol]->tag));
441
442 /* First reduce the nonterminals, as they renumber themselves in the
443 whole grammar. If you change the order, nonterms would be
444 renumbered only in the reduced grammar. */
445 if (nuseless_nonterminals > 0)
446 nonterminals_reduce ();
447 if (nuseless_productions > 0)
448 reduce_grammar_tables ();
449
450 if (trace_flag)
451 {
452 grammar_dump (stderr, "Reduced Grammar");
453
454 fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\
455 , and %d productions.\n",
456 infile, ntokens, nvars, nrules);
457 }
458 }
459
460
461 /*-----------------------------------------------------------.
462 | Free the global sets used to compute the reduced grammar. |
463 `-----------------------------------------------------------*/
464
465 void
466 reduce_free (void)
467 {
468 bitset_free (N);
469 bitset_free (V);
470 bitset_free (V1);
471 bitset_free (P);
472 }