1/* Grammar reduction for Bison.
2
3   Copyright (C) 1988-1989, 2000-2003, 2005-2012 Free Software
4   Foundation, Inc.
5
6   This file is part of Bison, the GNU Compiler Compiler.
7
8   This program is free software: you can redistribute it and/or modify
9   it under the terms of the GNU General Public License as published by
10   the Free Software Foundation, either version 3 of the License, or
11   (at your option) any later version.
12
13   This program is distributed in the hope that it will be useful,
14   but WITHOUT ANY WARRANTY; without even the implied warranty of
15   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16   GNU General Public License for more details.
17
18   You should have received a copy of the GNU General Public License
19   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
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 <config.h>
29#include "system.h"
30
31#include <bitset.h>
32
33#include "complain.h"
34#include "files.h"
35#include "getargs.h"
36#include "gram.h"
37#include "print-xml.h"
38#include "reader.h"
39#include "reduce.h"
40#include "symtab.h"
41
42/* Set of all nonterminals which are not useless.  */
43static bitset N;
44
45/* Set of all rules which have no useless nonterminals in their RHS.  */
46static bitset P;
47
48/* Set of all accessible symbols.  */
49static bitset V;
50
51/* Set of symbols used to define rule precedence (so they are
52   `useless', but no warning should be issued).  */
53static bitset V1;
54
55static rule_number nuseful_productions;
56rule_number nuseless_productions;
57static int nuseful_nonterminals;
58symbol_number nuseless_nonterminals;
59
60/*-------------------------------------------------------------------.
61| Another way to do this would be with a set for each production and |
62| then do subset tests against N0, but even for the C grammar the    |
63| whole reducing process takes only 2 seconds on my 8Mhz AT.         |
64`-------------------------------------------------------------------*/
65
66static bool
67useful_production (rule_number r, bitset N0)
68{
69  item_number *rhsp;
70
71  /* A production is useful if all of the nonterminals in its appear
72     in the set of useful nonterminals.  */
73
74  for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
75    if (ISVAR (*rhsp) && !bitset_test (N0, *rhsp - ntokens))
76      return false;
77  return true;
78}
79
80
81/*---------------------------------------------------------.
82| Remember that rules are 1-origin, symbols are 0-origin.  |
83`---------------------------------------------------------*/
84
85static void
86useless_nonterminals (void)
87{
88  bitset Np, Ns;
89  rule_number r;
90
91  /* N is set as built.  Np is set being built this iteration. P is
92     set of all productions which have a RHS all in N.  */
93
94  Np = bitset_create (nvars, BITSET_FIXED);
95
96
97  /* The set being computed is a set of nonterminals which can derive
98     the empty string or strings consisting of all terminals. At each
99     iteration a nonterminal is added to the set if there is a
100     production with that nonterminal as its LHS for which all the
101     nonterminals in its RHS are already in the set.  Iterate until
102     the set being computed remains unchanged.  Any nonterminals not
103     in the set at that point are useless in that they will never be
104     used in deriving a sentence of the language.
105
106     This iteration doesn't use any special traversal over the
107     productions.  A set is kept of all productions for which all the
108     nonterminals in the RHS are in useful.  Only productions not in
109     this set are scanned on each iteration.  At the end, this set is
110     saved to be used when finding useful productions: only
111     productions in this set will appear in the final grammar.  */
112
113  while (1)
114    {
115      bitset_copy (Np, N);
116      for (r = 0; r < nrules; r++)
117	if (!bitset_test (P, r)
118	    && useful_production (r, N))
119	  {
120	    bitset_set (Np, rules[r].lhs->number - ntokens);
121	    bitset_set (P, r);
122	  }
123      if (bitset_equal_p (N, Np))
124	break;
125      Ns = Np;
126      Np = N;
127      N = Ns;
128    }
129  bitset_free (N);
130  N = Np;
131}
132
133
134static void
135inaccessable_symbols (void)
136{
137  bitset Vp, Vs, Pp;
138
139  /* Find out which productions are reachable and which symbols are
140     used.  Starting with an empty set of productions and a set of
141     symbols which only has the start symbol in it, iterate over all
142     productions until the set of productions remains unchanged for an
143     iteration.  For each production which has a LHS in the set of
144     reachable symbols, add the production to the set of reachable
145     productions, and add all of the nonterminals in the RHS of the
146     production to the set of reachable symbols.
147
148     Consider only the (partially) reduced grammar which has only
149     nonterminals in N and productions in P.
150
151     The result is the set P of productions in the reduced grammar,
152     and the set V of symbols in the reduced grammar.
153
154     Although this algorithm also computes the set of terminals which
155     are reachable, no terminal will be deleted from the grammar. Some
156     terminals might not be in the grammar but might be generated by
157     semantic routines, and so the user might want them available with
158     specified numbers.  (Is this true?)  However, the nonreachable
159     terminals are printed (if running in verbose mode) so that the
160     user can know.  */
161
162  Vp = bitset_create (nsyms, BITSET_FIXED);
163  Pp = bitset_create (nrules, BITSET_FIXED);
164
165  /* If the start symbol isn't useful, then nothing will be useful. */
166  if (bitset_test (N, accept->number - ntokens))
167    {
168      bitset_set (V, accept->number);
169
170      while (1)
171	{
172	  rule_number r;
173	  bitset_copy (Vp, V);
174	  for (r = 0; r < nrules; r++)
175	    {
176	      if (!bitset_test (Pp, r)
177		  && bitset_test (P, r)
178		  && bitset_test (V, rules[r].lhs->number))
179		{
180		  item_number *rhsp;
181		  for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++)
182		    if (ISTOKEN (*rhsp) || bitset_test (N, *rhsp - ntokens))
183		      bitset_set (Vp, *rhsp);
184		  bitset_set (Pp, r);
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, endtoken->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  {
211    symbol_number i;
212    for (i = ntokens; i < nsyms; i++)
213      if (bitset_test (V, i))
214	nuseful_nonterminals++;
215  }
216  nuseless_nonterminals = nvars - nuseful_nonterminals;
217
218  /* A token that was used in %prec should not be warned about.  */
219  {
220    rule_number r;
221    for (r = 0; r < nrules; ++r)
222      if (rules[r].precsym != 0)
223	bitset_set (V1, rules[r].precsym->number);
224  }
225}
226
227
228/*-------------------------------------------------------------------.
229| Put the useless productions at the end of RULES, and adjust NRULES |
230| accordingly.                                                       |
231`-------------------------------------------------------------------*/
232
233static void
234reduce_grammar_tables (void)
235{
236  /* Report and flag useless productions.  */
237  {
238    rule_number r;
239    for (r = 0; r < nrules; r++)
240      rules[r].useful = bitset_test (P, r);
241    grammar_rules_useless_report (_("rule useless in grammar"));
242  }
243
244  /* Map the nonterminals to their new index: useful first, useless
245     afterwards.  Kept for later report.  */
246  {
247    int useful = 0;
248    int useless = nrules - nuseless_productions;
249    rule *rules_sorted = xnmalloc (nrules, sizeof *rules_sorted);
250    rule_number r;
251    for (r = 0; r < nrules; ++r)
252      rules_sorted[rules[r].useful ? useful++ : useless++] = rules[r];
253    free (rules);
254    rules = rules_sorted;
255
256    /* Renumber the rules markers in RITEMS.  */
257    for (r = 0; r < nrules; ++r)
258      {
259	item_number *rhsp = rules[r].rhs;
260	for (/* Nothing. */; *rhsp >= 0; ++rhsp)
261	  /* Nothing. */;
262	*rhsp = rule_number_as_item_number (r);
263	rules[r].number = r;
264      }
265    nrules -= nuseless_productions;
266  }
267
268  /* Adjust NRITEMS.  */
269  {
270    rule_number r;
271    int length;
272    for (r = nrules; r < nrules + nuseless_productions; ++r)
273      {
274	length = rule_rhs_length (&rules[r]);
275	nritems -= length + 1;
276      }
277  }
278}
279
280
281/*------------------------------.
282| Remove useless nonterminals.  |
283`------------------------------*/
284
285static void
286nonterminals_reduce (void)
287{
288  symbol_number i, n;
289
290  /* Map the nonterminals to their new index: useful first, useless
291     afterwards.  Kept for later report.  */
292
293  symbol_number *nontermmap = xnmalloc (nvars, sizeof *nontermmap);
294  n = ntokens;
295  for (i = ntokens; i < nsyms; i++)
296    if (bitset_test (V, i))
297      nontermmap[i - ntokens] = n++;
298  for (i = ntokens; i < nsyms; i++)
299    if (!bitset_test (V, i))
300      {
301	nontermmap[i - ntokens] = n++;
302	warn_at (symbols[i]->location, _("nonterminal useless in grammar: %s"),
303		 symbols[i]->tag);
304      }
305
306
307  /* Shuffle elements of tables indexed by symbol number.  */
308  {
309    symbol **symbols_sorted = xnmalloc (nvars, sizeof *symbols_sorted);
310
311    for (i = ntokens; i < nsyms; i++)
312      symbols[i]->number = nontermmap[i - ntokens];
313    for (i = ntokens; i < nsyms; i++)
314      symbols_sorted[nontermmap[i - ntokens] - ntokens] = symbols[i];
315    for (i = ntokens; i < nsyms; i++)
316      symbols[i] = symbols_sorted[i - ntokens];
317    free (symbols_sorted);
318  }
319
320  {
321    rule_number r;
322    for (r = 0; r < nrules; ++r)
323      {
324	item_number *rhsp;
325	for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
326	  if (ISVAR (*rhsp))
327	    *rhsp =  symbol_number_as_item_number (nontermmap[*rhsp
328							      - ntokens]);
329      }
330    accept->number = nontermmap[accept->number - ntokens];
331  }
332
333  nsyms -= nuseless_nonterminals;
334  nvars -= nuseless_nonterminals;
335
336  free (nontermmap);
337}
338
339
340/*------------------------------------------------------------------.
341| Output the detailed results of the reductions.  For FILE.output.  |
342`------------------------------------------------------------------*/
343
344void
345reduce_output (FILE *out)
346{
347  if (nuseless_nonterminals > 0)
348    {
349      int i;
350      fprintf (out, "%s\n\n", _("Nonterminals useless in grammar"));
351      for (i = 0; i < nuseless_nonterminals; ++i)
352	fprintf (out, "   %s\n", symbols[nsyms + i]->tag);
353      fputs ("\n\n", out);
354    }
355
356  {
357    bool b = false;
358    int i;
359    for (i = 0; i < ntokens; i++)
360      if (reduce_token_unused_in_grammar (i))
361	{
362	  if (!b)
363	    fprintf (out, "%s\n\n", _("Terminals unused in grammar"));
364	  b = true;
365	  fprintf (out, "   %s\n", symbols[i]->tag);
366	}
367    if (b)
368      fputs ("\n\n", out);
369  }
370
371  if (nuseless_productions > 0)
372    grammar_rules_partial_print (out, _("Rules useless in grammar"),
373				 rule_useless_in_grammar_p);
374}
375
376
377/*-------------------------------.
378| Report the results to STDERR.  |
379`-------------------------------*/
380
381static void
382reduce_print (void)
383{
384  if (nuseless_nonterminals > 0)
385    warn (ngettext ("%d nonterminal useless in grammar",
386                    "%d nonterminals useless in grammar",
387                    nuseless_nonterminals),
388          nuseless_nonterminals);
389  if (nuseless_productions > 0)
390    warn (ngettext ("%d rule useless in grammar",
391                    "%d rules useless in grammar",
392                    nuseless_productions),
393          nuseless_productions);
394}
395
396void
397reduce_grammar (void)
398{
399  bool reduced;
400
401  /* Allocate the global sets used to compute the reduced grammar */
402
403  N = bitset_create (nvars, BITSET_FIXED);
404  P =  bitset_create (nrules, BITSET_FIXED);
405  V = bitset_create (nsyms, BITSET_FIXED);
406  V1 = bitset_create (nsyms, BITSET_FIXED);
407
408  useless_nonterminals ();
409  inaccessable_symbols ();
410
411  reduced = (nuseless_nonterminals + nuseless_productions > 0);
412  if (!reduced)
413    return;
414
415  reduce_print ();
416
417  if (!bitset_test (N, accept->number - ntokens))
418    fatal_at (startsymbol_location,
419	      _("start symbol %s does not derive any sentence"),
420	      startsymbol->tag);
421
422  /* First reduce the nonterminals, as they renumber themselves in the
423     whole grammar.  If you change the order, nonterms would be
424     renumbered only in the reduced grammar.  */
425  if (nuseless_nonterminals > 0)
426    nonterminals_reduce ();
427  if (nuseless_productions > 0)
428    reduce_grammar_tables ();
429
430  if (trace_flag & trace_grammar)
431    {
432      grammar_dump (stderr, "Reduced Grammar");
433
434      fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\
435, and %d productions.\n",
436	       grammar_file, ntokens, nvars, nrules);
437    }
438}
439
440bool
441reduce_token_unused_in_grammar (symbol_number i)
442{
443  aver (i < ntokens);
444  return !bitset_test (V, i) && !bitset_test (V1, i);
445}
446
447bool
448reduce_nonterminal_useless_in_grammar (symbol_number i)
449{
450  aver (ntokens <= i && i < nsyms + nuseless_nonterminals);
451  return nsyms <= i;
452}
453
454/*-----------------------------------------------------------.
455| Free the global sets used to compute the reduced grammar.  |
456`-----------------------------------------------------------*/
457
458void
459reduce_free (void)
460{
461  bitset_free (N);
462  bitset_free (V);
463  bitset_free (V1);
464  bitset_free (P);
465}
466