|
|
1.1 root 1: /* Grammar reduction for Bison.
2: Copyright (C) 1988, 1989 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, 675 Mass Ave, Cambridge, MA 02139, USA. */
19:
20:
21: /*
22: * Reduce the grammar: Find and eliminate unreachable terminals,
23: * nonterminals, and productions. David S. Bakin.
24: */
25:
26: /*
27: * Don't eliminate unreachable terminals: They may be used by the user's
28: * parser.
29: */
30:
31: #include <stdio.h>
32: #include "system.h"
33: #include "files.h"
34: #include "gram.h"
35: #include "machine.h"
36: #include "new.h"
37:
38:
39: extern char **tags; /* reader.c */
40: extern int verboseflag; /* getargs.c */
41: static int statisticsflag; /* XXXXXXX */
42:
43: #ifndef TRUE
44: #define TRUE (1)
45: #define FALSE (0)
46: #endif
47: typedef int bool;
48: typedef unsigned *BSet;
49: typedef short *rule;
50:
51:
52: /*
53: * N is set of all nonterminals which are not useless. P is set of all rules
54: * which have no useless nonterminals in their RHS. V is the set of all
55: * accessible symbols.
56: */
57:
58: static BSet N, P, V, V1;
59:
60: static int nuseful_productions, nuseless_productions,
61: nuseful_nonterminals, nuseless_nonterminals;
62:
63:
64: static void useless_nonterminals();
65: static void inaccessable_symbols();
66: static void reduce_grammar_tables();
67: static void print_results();
68: static void print_notices();
69: void dump_grammar();
70:
71: extern void fatals ();
72:
73:
74: bool
75: bits_equal (L, R, n)
76: BSet L;
77: BSet R;
78: int n;
79: {
80: int i;
81:
82: for (i = n - 1; i >= 0; i--)
83: if (L[i] != R[i])
84: return FALSE;
85: return TRUE;
86: }
87:
88:
89: int
90: nbits (i)
91: unsigned i;
92: {
93: int count = 0;
94:
95: while (i != 0) {
96: i ^= (i & -i);
97: ++count;
98: }
99: return count;
100: }
101:
102:
103: int
104: bits_size (S, n)
105: BSet S;
106: int n;
107: {
108: int i, count = 0;
109:
110: for (i = n - 1; i >= 0; i--)
111: count += nbits(S[i]);
112: return count;
113: }
114:
115: void
116: reduce_grammar ()
117: {
118: bool reduced;
119:
120: /* Allocate the global sets used to compute the reduced grammar */
121:
122: N = NEW2(WORDSIZE(nvars), unsigned);
123: P = NEW2(WORDSIZE(nrules + 1), unsigned);
124: V = NEW2(WORDSIZE(nsyms), unsigned);
125: V1 = NEW2(WORDSIZE(nsyms), unsigned);
126:
127: useless_nonterminals();
128: inaccessable_symbols();
129:
130: reduced = (bool) (nuseless_nonterminals + nuseless_productions > 0);
131:
132: if (verboseflag)
133: print_results();
134:
135: if (reduced == FALSE)
136: goto done_reducing;
137:
138: print_notices();
139:
140: if (!BITISSET(N, start_symbol - ntokens))
141: fatals("Start symbol %s does not derive any sentence.",
142: tags[start_symbol]);
143:
144: reduce_grammar_tables();
145: /* if (verboseflag) {
146: fprintf(foutput, "REDUCED GRAMMAR\n\n");
147: dump_grammar();
148: }
149: */
150:
151: /**/ statisticsflag = FALSE; /* someday getopts should handle this */
152: if (statisticsflag == TRUE)
153: fprintf(stderr,
154: "reduced %s defines %d terminal%s, %d nonterminal%s\
155: , and %d production%s.\n", infile,
156: ntokens, (ntokens == 1 ? "" : "s"),
157: nvars, (nvars == 1 ? "" : "s"),
158: nrules, (nrules == 1 ? "" : "s"));
159:
160: done_reducing:
161:
162: /* Free the global sets used to compute the reduced grammar */
163:
164: FREE(N);
165: FREE(V);
166: FREE(P);
167:
168: }
169:
170: /*
171: * Another way to do this would be with a set for each production and then do
172: * subset tests against N, but even for the C grammar the whole reducing
173: * process takes only 2 seconds on my 8Mhz AT.
174: */
175:
176: static bool
177: useful_production (i, N)
178: int i;
179: BSet N;
180: {
181: rule r;
182: short n;
183:
184: /*
185: * A production is useful if all of the nonterminals in its RHS
186: * appear in the set of useful nonterminals.
187: */
188:
189: for (r = &ritem[rrhs[i]]; *r > 0; r++)
190: if (ISVAR(n = *r))
191: if (!BITISSET(N, n - ntokens))
192: return FALSE;
193: return TRUE;
194: }
195:
196:
197: /* Remember that rules are 1-origin, symbols are 0-origin. */
198:
199: static void
200: useless_nonterminals ()
201: {
202: BSet Np, Ns;
203: int i, n;
204:
205: /*
206: * N is set as built. Np is set being built this iteration. P is set
207: * of all productions which have a RHS all in N.
208: */
209:
210: Np = NEW2(WORDSIZE(nvars), unsigned);
211:
212: /*
213: * The set being computed is a set of nonterminals which can derive
214: * the empty string or strings consisting of all terminals. At each
215: * iteration a nonterminal is added to the set if there is a
216: * production with that nonterminal as its LHS for which all the
217: * nonterminals in its RHS are already in the set. Iterate until the
218: * set being computed remains unchanged. Any nonterminals not in the
219: * set at that point are useless in that they will never be used in
220: * deriving a sentence of the language.
221: *
222: * This iteration doesn't use any special traversal over the
223: * productions. A set is kept of all productions for which all the
224: * nonterminals in the RHS are in useful. Only productions not in
225: * this set are scanned on each iteration. At the end, this set is
226: * saved to be used when finding useful productions: only productions
227: * in this set will appear in the final grammar.
228: */
229:
230: n = 0;
231: while (1)
232: {
233: for (i = WORDSIZE(nvars) - 1; i >= 0; i--)
234: Np[i] = N[i];
235: for (i = 1; i <= nrules; i++)
236: {
237: if (!BITISSET(P, i))
238: {
239: if (useful_production(i, N))
240: {
241: SETBIT(Np, rlhs[i] - ntokens);
242: SETBIT(P, i);
243: }
244: }
245: }
246: if (bits_equal(N, Np, WORDSIZE(nvars)))
247: break;
248: Ns = Np;
249: Np = N;
250: N = Ns;
251: }
252: FREE(N);
253: N = Np;
254: }
255:
256: static void
257: inaccessable_symbols ()
258: {
259: BSet Vp, Vs, Pp;
260: int i, n;
261: short t;
262: rule r;
263:
264: /*
265: * Find out which productions are reachable and which symbols are
266: * used. Starting with an empty set of productions and a set of
267: * symbols which only has the start symbol in it, iterate over all
268: * productions until the set of productions remains unchanged for an
269: * iteration. For each production which has a LHS in the set of
270: * reachable symbols, add the production to the set of reachable
271: * productions, and add all of the nonterminals in the RHS of the
272: * production to the set of reachable symbols.
273: *
274: * Consider only the (partially) reduced grammar which has only
275: * nonterminals in N and productions in P.
276: *
277: * The result is the set P of productions in the reduced grammar, and
278: * the set V of symbols in the reduced grammar.
279: *
280: * Although this algorithm also computes the set of terminals which are
281: * reachable, no terminal will be deleted from the grammar. Some
282: * terminals might not be in the grammar but might be generated by
283: * semantic routines, and so the user might want them available with
284: * specified numbers. (Is this true?) However, the nonreachable
285: * terminals are printed (if running in verbose mode) so that the user
286: * can know.
287: */
288:
289: Vp = NEW2(WORDSIZE(nsyms), unsigned);
290: Pp = NEW2(WORDSIZE(nrules + 1), unsigned);
291:
292: /* If the start symbol isn't useful, then nothing will be useful. */
293: if (!BITISSET(N, start_symbol - ntokens))
294: goto end_iteration;
295:
296: SETBIT(V, start_symbol);
297:
298: n = 0;
299: while (1)
300: {
301: for (i = WORDSIZE(nsyms) - 1; i >= 0; i--)
302: Vp[i] = V[i];
303: for (i = 1; i <= nrules; i++)
304: {
305: if (!BITISSET(Pp, i) && BITISSET(P, i) &&
306: BITISSET(V, rlhs[i]))
307: {
308: for (r = &ritem[rrhs[i]]; *r >= 0; r++)
309: {
310: if (ISTOKEN(t = *r)
311: || BITISSET(N, t - ntokens))
312: {
313: SETBIT(Vp, t);
314: }
315: }
316: SETBIT(Pp, i);
317: }
318: }
319: if (bits_equal(V, Vp, WORDSIZE(nsyms)))
320: {
321: break;
322: }
323: Vs = Vp;
324: Vp = V;
325: V = Vs;
326: }
327: end_iteration:
328:
329: FREE(V);
330: V = Vp;
331:
332: /* Tokens 0, 1, and 2 are internal to Bison. Consider them useful. */
333: SETBIT(V, 0); /* end-of-input token */
334: SETBIT(V, 1); /* error token */
335: SETBIT(V, 2); /* illegal token */
336:
337: FREE(P);
338: P = Pp;
339:
340: nuseful_productions = bits_size(P, WORDSIZE(nrules + 1));
341: nuseless_productions = nrules - nuseful_productions;
342:
343: nuseful_nonterminals = 0;
344: for (i = ntokens; i < nsyms; i++)
345: if (BITISSET(V, i))
346: nuseful_nonterminals++;
347: nuseless_nonterminals = nvars - nuseful_nonterminals;
348:
349: /* A token that was used in %prec should not be warned about. */
350: for (i = 1; i < nrules; i++)
351: if (rprecsym[i] != 0)
352: SETBIT(V1, rprecsym[i]);
353: }
354:
355: static void
356: reduce_grammar_tables ()
357: {
358: /* This is turned off because we would need to change the numbers
359: in the case statements in the actions file. */
360: #if 0
361: /* remove useless productions */
362: if (nuseless_productions > 0)
363: {
364: short np, pn, ni, pi;
365:
366: np = 0;
367: ni = 0;
368: for (pn = 1; pn <= nrules; pn++)
369: {
370: if (BITISSET(P, pn))
371: {
372: np++;
373: if (pn != np)
374: {
375: rlhs[np] = rlhs[pn];
376: rline[np] = rline[pn];
377: rprec[np] = rprec[pn];
378: rassoc[np] = rassoc[pn];
379: rrhs[np] = rrhs[pn];
380: if (rrhs[np] != ni)
381: {
382: pi = rrhs[np];
383: rrhs[np] = ni;
384: while (ritem[pi] >= 0)
385: ritem[ni++] = ritem[pi++];
386: ritem[ni++] = -np;
387: }
388: } else {
389: while (ritem[ni++] >= 0);
390: }
391: }
392: }
393: ritem[ni] = 0;
394: nrules -= nuseless_productions;
395: nitems = ni;
396:
397: /*
398: * Is it worth it to reduce the amount of memory for the
399: * grammar? Probably not.
400: */
401:
402: }
403: #endif /* 0 */
404: /* Disable useless productions,
405: since they may contain useless nonterms
406: that would get mapped below to -1 and confuse everyone. */
407: if (nuseless_productions > 0)
408: {
409: int pn;
410:
411: for (pn = 1; pn <= nrules; pn++)
412: {
413: if (!BITISSET(P, pn))
414: {
415: rlhs[pn] = -1;
416: }
417: }
418: }
419:
420: /* remove useless symbols */
421: if (nuseless_nonterminals > 0)
422: {
423:
424: int i, n;
425: /* short j; JF unused */
426: short *nontermmap;
427: rule r;
428:
429: /*
430: * create a map of nonterminal number to new nonterminal
431: * number. -1 in the map means it was useless and is being
432: * eliminated.
433: */
434:
435: nontermmap = NEW2(nvars, short) - ntokens;
436: for (i = ntokens; i < nsyms; i++)
437: nontermmap[i] = -1;
438:
439: n = ntokens;
440: for (i = ntokens; i < nsyms; i++)
441: if (BITISSET(V, i))
442: nontermmap[i] = n++;
443:
444: /* Shuffle elements of tables indexed by symbol number. */
445:
446: for (i = ntokens; i < nsyms; i++)
447: {
448: n = nontermmap[i];
449: if (n >= 0)
450: {
451: sassoc[n] = sassoc[i];
452: sprec[n] = sprec[i];
453: tags[n] = tags[i];
454: } else {
455: free(tags[i]);
456: }
457: }
458:
459: /* Replace all symbol numbers in valid data structures. */
460:
461: for (i = 1; i <= nrules; i++)
462: {
463: /* Ignore the rules disabled above. */
464: if (rlhs[i] >= 0)
465: rlhs[i] = nontermmap[rlhs[i]];
466: if (ISVAR (rprecsym[i]))
467: /* Can this happen? */
468: rprecsym[i] = nontermmap[rprecsym[i]];
469: }
470:
471: for (r = ritem; *r; r++)
472: if (ISVAR(*r))
473: *r = nontermmap[*r];
474:
475: start_symbol = nontermmap[start_symbol];
476:
477: nsyms -= nuseless_nonterminals;
478: nvars -= nuseless_nonterminals;
479:
480: free(&nontermmap[ntokens]);
481: }
482: }
483:
484: static void
485: print_results ()
486: {
487: int i;
488: /* short j; JF unused */
489: rule r;
490: bool b;
491:
492: if (nuseless_nonterminals > 0)
493: {
494: fprintf(foutput, "Useless nonterminals:\n\n");
495: for (i = ntokens; i < nsyms; i++)
496: if (!BITISSET(V, i))
497: fprintf(foutput, " %s\n", tags[i]);
498: }
499: b = FALSE;
500: for (i = 0; i < ntokens; i++)
501: {
502: if (!BITISSET(V, i) && !BITISSET(V1, i))
503: {
504: if (!b)
505: {
506: fprintf(foutput, "\n\nTerminals which are not used:\n\n");
507: b = TRUE;
508: }
509: fprintf(foutput, " %s\n", tags[i]);
510: }
511: }
512:
513: if (nuseless_productions > 0)
514: {
515: fprintf(foutput, "\n\nUseless rules:\n\n");
516: for (i = 1; i <= nrules; i++)
517: {
518: if (!BITISSET(P, i))
519: {
520: fprintf(foutput, "#%-4d ", i);
521: fprintf(foutput, "%s :\t", tags[rlhs[i]]);
522: for (r = &ritem[rrhs[i]]; *r >= 0; r++)
523: {
524: fprintf(foutput, " %s", tags[*r]);
525: }
526: fprintf(foutput, ";\n");
527: }
528: }
529: }
530: if (nuseless_nonterminals > 0 || nuseless_productions > 0 || b)
531: fprintf(foutput, "\n\n");
532: }
533:
534: void
535: dump_grammar ()
536: {
537: int i;
538: rule r;
539:
540: fprintf(foutput,
541: "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n",
542: ntokens, nvars, nsyms, nrules, nitems);
543: fprintf(foutput, "Variables\n---------\n\n");
544: fprintf(foutput, "Value Sprec Sassoc Tag\n");
545: for (i = ntokens; i < nsyms; i++)
546: fprintf(foutput, "%5d %5d %5d %s\n",
547: i, sprec[i], sassoc[i], tags[i]);
548: fprintf(foutput, "\n\n");
549: fprintf(foutput, "Rules\n-----\n\n");
550: for (i = 1; i <= nrules; i++)
551: {
552: fprintf(foutput, "%-5d(%5d%5d)%5d : (@%-5d)",
553: i, rprec[i], rassoc[i], rlhs[i], rrhs[i]);
554: for (r = &ritem[rrhs[i]]; *r > 0; r++)
555: fprintf(foutput, "%5d", *r);
556: fprintf(foutput, " [%d]\n", -(*r));
557: }
558: fprintf(foutput, "\n\n");
559: fprintf(foutput, "Rules interpreted\n-----------------\n\n");
560: for (i = 1; i <= nrules; i++)
561: {
562: fprintf(foutput, "%-5d %s :", i, tags[rlhs[i]]);
563: for (r = &ritem[rrhs[i]]; *r > 0; r++)
564: fprintf(foutput, " %s", tags[*r]);
565: fprintf(foutput, "\n");
566: }
567: fprintf(foutput, "\n\n");
568: }
569:
570:
571: static void
572: print_notices ()
573: {
574: extern int fixed_outfiles;
575:
576: if (fixed_outfiles && nuseless_productions)
577: fprintf(stderr, "%d rules never reduced\n", nuseless_productions);
578:
579: fprintf(stderr, "%s contains ", infile);
580:
581: if (nuseless_nonterminals > 0)
582: {
583: fprintf(stderr, "%d useless nonterminal%s",
584: nuseless_nonterminals,
585: (nuseless_nonterminals == 1 ? "" : "s"));
586: }
587: if (nuseless_nonterminals > 0 && nuseless_productions > 0)
588: fprintf(stderr, " and ");
589:
590: if (nuseless_productions > 0)
591: {
592: fprintf(stderr, "%d useless rule%s",
593: nuseless_productions,
594: (nuseless_productions == 1 ? "" : "s"));
595: }
596: fprintf(stderr, ".\n");
597: fflush(stderr);
598: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.