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 "getargs.h"
  26 #include "files.h"
  27 #include "symtab.h"
  28 #include "symlist.h"
  29 #include "gram.h"
  30 #include "complain.h"
  31 #include "output.h"
  32 #include "reader.h"
  33 #include "conflicts.h"
  34 #include "muscle_tab.h"
  35
  36 static symbol_list_t *grammar = NULL;
  37 static int start_flag = 0;
  38 merger_list *merge_functions;
  39
  40 /* Nonzero if %union has been seen.  */
  41 int typed = 0;
  42 \f
  43 /*-----------------------.
  44 | Set the start symbol.  |
  45 `-----------------------*/
  46
  47 void
  48 grammar_start_symbol_set (symbol_t *s, location_t l)
  49 {
  50   if (start_flag)
  51     complain_at (l, _("multiple %s declarations"), "%start");
  52   else
  53     {
  54       start_flag = 1;
  55       startsymbol = s;
  56       startsymbol_location = l;
  57     }
  58 }
  59
  60
  61 /*----------------------------------------------------------------.
  62 | There are two prologues: one before %union, one after.  Augment |
  63 | the current one.                                                |
  64 `----------------------------------------------------------------*/
  65
  66 void
  67 prologue_augment (const char *prologue, location_t location)
  68 {
  69   struct obstack *oout =
  70     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  71
  72   obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[",
  73                   location.first_line);
  74   MUSCLE_OBSTACK_SGROW (oout, quotearg_style (c_quoting_style, location.file));
  75   obstack_sgrow (oout, "]])[\n");
  76   obstack_sgrow (oout, prologue);
  77 }
  78
  79
  80
  81
  82 /*----------------------.
  83 | Handle the epilogue.  |
  84 `----------------------*/
  85
  86 void
  87 epilogue_augment (const char *epilogue, location_t location)
  88 {
  89   char *extension = NULL;
  90   obstack_fgrow1 (&muscle_obstack, "]b4_syncline([[%d]], [[",
  91                   location.first_line);
  92   MUSCLE_OBSTACK_SGROW (&muscle_obstack,
  93                         quotearg_style (c_quoting_style, location.file));
  94   obstack_sgrow (&muscle_obstack, "]])[\n");
  95   obstack_sgrow (&muscle_obstack, epilogue);
  96   obstack_1grow (&muscle_obstack, 0);
  97   extension = obstack_finish (&muscle_obstack);
  98   muscle_grow ("epilogue", extension, "");
  99   obstack_free (&muscle_obstack, extension);
 100 }
 101
 102
 103 \f
 104
 105 /*-------------------------------------------------------------------.
 106 | Return the merger index for a merging function named NAME, whose   |
 107 | arguments have type TYPE.  Records the function, if new, in        |
 108 | MERGER_LIST.                                                       |
 109 `-------------------------------------------------------------------*/
 110
 111 static int
 112 get_merge_function (struniq_t name, struniq_t type, location_t loc)
 113 {
 114   merger_list *syms;
 115   merger_list head;
 116   int n;
 117
 118   if (! glr_parser)
 119     return 0;
 120
 121   if (type == NULL)
 122     type = struniq_new ("");
 123
 124   head.next = merge_functions;
 125   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 126     if (STRUNIQ_EQ (name, syms->next->name))
 127       break;
 128   if (syms->next == NULL)
 129     {
 130       syms->next = XMALLOC (merger_list, 1);
 131       syms->next->name = struniq_new (name);
 132       syms->next->type = struniq_new (type);
 133       syms->next->next = NULL;
 134       merge_functions = head.next;
 135     }
 136   else if (!STRUNIQ_EQ (type, syms->next->type))
 137     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 138              name, type, syms->next->type);
 139   return n;
 140 }
 141
 142 /*--------------------------------------.
 143 | Free all merge-function definitions.  |
 144 `--------------------------------------*/
 145
 146 void
 147 free_merger_functions (void)
 148 {
 149   merger_list *L0;
 150   if (! glr_parser)
 151     return;
 152   L0 = merge_functions;
 153   while (L0 != NULL)
 154     {
 155       merger_list *L1 = L0->next;
 156       free (L0);
 157       L0 = L1;
 158     }
 159 }
 160
 161 \f
 162 /*-------------------------------------------------------------------.
 163 | Parse the input grammar into a one symbol_list_t structure.  Each  |
 164 | rule is represented by a sequence of symbols: the left hand side   |
 165 | followed by the contents of the right hand side, followed by a     |
 166 | null pointer instead of a symbol to terminate the rule.  The next  |
 167 | symbol is the lhs of the following rule.                           |
 168 |                                                                    |
 169 | All actions are copied out, labelled by the rule number they apply |
 170 | to.                                                                |
 171 |                                                                    |
 172 | Bison used to allow some %directives in the rules sections, but    |
 173 | this is no longer consider appropriate: (i) the documented grammar |
 174 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 175 | recovery for %directives consists in skipping the junk until a `%' |
 176 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 177 | in the rules section.                                              |
 178 `-------------------------------------------------------------------*/
 179
 180 /* The (currently) last symbol of GRAMMAR. */
 181 symbol_list_t *grammar_end = NULL;
 182
 183 /* Append S to the GRAMMAR. */
 184 void
 185 grammar_symbol_append (symbol_t *symbol, location_t location)
 186 {
 187   symbol_list_t *p = symbol_list_new (symbol, location);
 188
 189   if (grammar_end)
 190     grammar_end->next = p;
 191   else
 192     grammar = p;
 193
 194   grammar_end = p;
 195 }
 196
 197 /* The rule currently being defined, and the previous rule.
 198    CURRENT_RULE points to the first LHS of the current rule, while
 199    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 200 symbol_list_t *current_rule = NULL;
 201 symbol_list_t *previous_rule_end = NULL;
 202
 203
 204 /*----------------------------------------------.
 205 | Create a new rule for LHS in to the GRAMMAR.  |
 206 `----------------------------------------------*/
 207
 208 void
 209 grammar_rule_begin (symbol_t *lhs, location_t location)
 210 {
 211   if (!start_flag)
 212     {
 213       startsymbol = lhs;
 214       startsymbol_location = location;
 215       start_flag = 1;
 216     }
 217
 218   /* Start a new rule and record its lhs.  */
 219   ++nrules;
 220   ++nritems;
 221
 222   previous_rule_end = grammar_end;
 223   grammar_symbol_append (lhs, location);
 224   current_rule = grammar_end;
 225
 226   /* Mark the rule's lhs as a nonterminal if not already so.  */
 227
 228   if (lhs->class == unknown_sym)
 229     {
 230       lhs->class = nterm_sym;
 231       lhs->number = nvars;
 232       ++nvars;
 233     }
 234   else if (lhs->class == token_sym)
 235     complain_at (location, _("rule given for %s, which is a token"), lhs->tag);
 236 }
 237
 238 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 239    instance, there should be no type clash on the default action.  */
 240
 241 static void
 242 grammar_current_rule_check (void)
 243 {
 244   symbol_t *lhs = current_rule->sym;
 245   char const *lhs_type = lhs->type_name;
 246   symbol_t *first_rhs = current_rule->next->sym;
 247
 248   /* If there is an action, then there is nothing we can do: the user
 249      is allowed to shoot herself in the foot.  */
 250   if (current_rule->action)
 251     return;
 252
 253   /* Don't worry about the default action if $$ is untyped, since $$'s
 254      value can't be used.  */
 255   if (! lhs_type)
 256     return;
 257
 258   /* If $$ is being set in default way, report if any type mismatch.  */
 259   if (first_rhs)
 260     {
 261       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 262       if (!STRUNIQ_EQ (lhs_type, rhs_type))
 263         complain_at (current_rule->location,
 264                      _("type clash on default action: <%s> != <%s>"),
 265                      lhs_type, rhs_type);
 266     }
 267   /* Warn if there is no default for $$ but we need one.  */
 268   else
 269     complain_at (current_rule->location,
 270                  _("empty rule for typed nonterminal, and no action"));
 271 }
 272
 273
 274 /*-------------------------------------.
 275 | End the currently being grown rule.  |
 276 `-------------------------------------*/
 277
 278 void
 279 grammar_rule_end (location_t location)
 280 {
 281   /* Put an empty link in the list to mark the end of this rule  */
 282   grammar_symbol_append (NULL, grammar_end->location);
 283   current_rule->location = location;
 284   grammar_current_rule_check ();
 285 }
 286
 287
 288 /*-------------------------------------------------------------------.
 289 | The previous action turns out the be a mid-rule action.  Attach it |
 290 | to the current rule, i.e., create a dummy symbol, attach it this   |
 291 | mid-rule action, and append this dummy nonterminal to the current  |
 292 | rule.                                                              |
 293 `-------------------------------------------------------------------*/
 294
 295 void
 296 grammar_midrule_action (void)
 297 {
 298   /* Since the action was written out with this rule's number, we must
 299      give the new rule this number by inserting the new rule before
 300      it.  */
 301
 302   /* Make a DUMMY nonterminal, whose location is that of the midrule
 303      action.  Create the MIDRULE.  */
 304   location_t dummy_location = current_rule->action_location;
 305   symbol_t *dummy = dummy_symbol_get (dummy_location);
 306   symbol_list_t *midrule = symbol_list_new (dummy, dummy_location);
 307
 308   /* Make a new rule, whose body is empty, before the current one, so
 309      that the action just read can belong to it.  */
 310   ++nrules;
 311   ++nritems;
 312   /* Attach its location and actions to that of the DUMMY.  */
 313   midrule->location = dummy_location;
 314   midrule->action = current_rule->action;
 315   midrule->action_location = dummy_location;
 316   current_rule->action = NULL;
 317
 318   if (previous_rule_end)
 319     previous_rule_end->next = midrule;
 320   else
 321     grammar = midrule;
 322
 323   /* End the dummy's rule.  */
 324   previous_rule_end = symbol_list_new (NULL, dummy_location);
 325   previous_rule_end->next = current_rule;
 326
 327   midrule->next = previous_rule_end;
 328
 329   /* Insert the dummy nonterminal replacing the midrule action into
 330      the current rule.  */
 331   grammar_current_rule_symbol_append (dummy, dummy_location);
 332 }
 333
 334 /* Set the precedence symbol of the current rule to PRECSYM. */
 335
 336 void
 337 grammar_current_rule_prec_set (symbol_t *precsym, location_t location)
 338 {
 339   if (current_rule->ruleprec)
 340     complain_at (location, _("only one %s allowed per rule"), "%prec");
 341   current_rule->ruleprec = precsym;
 342 }
 343
 344 /* Attach dynamic precedence DPREC to the current rule. */
 345
 346 void
 347 grammar_current_rule_dprec_set (int dprec, location_t location)
 348 {
 349   if (! glr_parser)
 350     warn_at (location, _("%s affects only GLR parsers"), "%dprec");
 351   if (dprec <= 0)
 352     complain_at (location,
 353                  _("%s must be followed by positive number"), "%dprec");
 354   else if (current_rule->dprec != 0)
 355     complain_at (location, _("only one %s allowed per rule"), "%dprec");
 356   current_rule->dprec = dprec;
 357 }
 358
 359 /* Attach a merge function NAME with argument type TYPE to current
 360    rule. */
 361
 362 void
 363 grammar_current_rule_merge_set (struniq_t name, location_t location)
 364 {
 365   if (! glr_parser)
 366     warn_at (location, _("%s affects only GLR parsers"), "%merge");
 367   if (current_rule->merger != 0)
 368     complain_at (location, _("only one %s allowed per rule"), "%merge");
 369   current_rule->merger =
 370     get_merge_function (name, current_rule->sym->type_name, location);
 371 }
 372
 373 /* Attach a SYMBOL to the current rule.  If needed, move the previous
 374    action as a mid-rule action.  */
 375
 376 void
 377 grammar_current_rule_symbol_append (symbol_t *symbol, location_t location)
 378 {
 379   if (current_rule->action)
 380     grammar_midrule_action ();
 381   ++nritems;
 382   grammar_symbol_append (symbol, location);
 383 }
 384
 385 /* Attach an ACTION to the current rule.  If needed, move the previous
 386    action as a mid-rule action.  */
 387
 388 void
 389 grammar_current_rule_action_append (const char *action, location_t location)
 390 {
 391   if (current_rule->action)
 392     grammar_midrule_action ();
 393   current_rule->action = action;
 394   current_rule->action_location = location;
 395 }
 396
 397 \f
 398 /*---------------------------------------------------------------.
 399 | Convert the rules into the representation using RRHS, RLHS and |
 400 | RITEM.                                                         |
 401 `---------------------------------------------------------------*/
 402
 403 static void
 404 packgram (void)
 405 {
 406   unsigned int itemno = 0;
 407   rule_number_t ruleno = 0;
 408   symbol_list_t *p = grammar;
 409
 410   ritem = XCALLOC (item_number_t, nritems);
 411   rules = XCALLOC (rule_t, nrules);
 412
 413   while (p)
 414     {
 415       symbol_t *ruleprec = p->ruleprec;
 416       rules[ruleno].user_number = ruleno;
 417       rules[ruleno].number = ruleno;
 418       rules[ruleno].lhs = p->sym;
 419       rules[ruleno].rhs = ritem + itemno;
 420       rules[ruleno].location = p->location;
 421       rules[ruleno].useful = true;
 422       rules[ruleno].action = p->action;
 423       rules[ruleno].action_location = p->action_location;
 424       rules[ruleno].dprec = p->dprec;
 425       rules[ruleno].merger = p->merger;
 426
 427       p = p->next;
 428       while (p && p->sym)
 429         {
 430           /* item_number_t = symbol_number_t.
 431              But the former needs to contain more: negative rule numbers. */
 432           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 433           /* A rule gets by default the precedence and associativity
 434              of the last token in it.  */
 435           if (p->sym->class == token_sym)
 436             rules[ruleno].prec = p->sym;
 437           if (p)
 438             p = p->next;
 439         }
 440
 441       /* If this rule has a %prec,
 442          the specified symbol's precedence replaces the default.  */
 443       if (ruleprec)
 444         {
 445           rules[ruleno].precsym = ruleprec;
 446           rules[ruleno].prec = ruleprec;
 447         }
 448       ritem[itemno++] = rule_number_as_item_number (ruleno);
 449       ++ruleno;
 450
 451       if (p)
 452         p = p->next;
 453     }
 454
 455   if (itemno != nritems)
 456     abort ();
 457
 458   if (trace_flag & trace_sets)
 459     ritem_print (stderr);
 460 }
 461 \f
 462 /*------------------------------------------------------------------.
 463 | Read in the grammar specification and record it in the format     |
 464 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 465 | in each case forming the body of a C function (YYACTION) which    |
 466 | contains a switch statement to decide which action to execute.    |
 467 `------------------------------------------------------------------*/
 468
 469 void
 470 reader (void)
 471 {
 472   gram_control_t gram_control;
 473
 474   /* Initialize the symbol table.  */
 475   symbols_new ();
 476
 477   /* Construct the accept symbol. */
 478   accept = symbol_get ("$accept", empty_location);
 479   accept->class = nterm_sym;
 480   accept->number = nvars++;
 481
 482   /* Construct the error token */
 483   errtoken = symbol_get ("error", empty_location);
 484   errtoken->class = token_sym;
 485   errtoken->number = ntokens++;
 486
 487   /* Construct a token that represents all undefined literal tokens.
 488      It is always token number 2.  */
 489   undeftoken = symbol_get ("$undefined", empty_location);
 490   undeftoken->class = token_sym;
 491   undeftoken->number = ntokens++;
 492
 493   /* Initialize the obstacks. */
 494   obstack_init (&pre_prologue_obstack);
 495   obstack_init (&post_prologue_obstack);
 496
 497   finput = xfopen (grammar_file, "r");
 498   gram_in = finput;
 499
 500   gram__flex_debug = trace_flag & trace_scan;
 501   gram_debug = trace_flag & trace_parse;
 502   scanner_initialize ();
 503   gram_parse (&gram_control);
 504
 505   /* If something went wrong during the parsing, don't try to
 506      continue.  */
 507   if (complaint_issued)
 508     return;
 509
 510   /* Grammar has been read.  Do some checking */
 511   if (nrules == 0)
 512     fatal (_("no rules in the input grammar"));
 513
 514   /* Report any undefined symbols and consider them nonterminals.  */
 515   symbols_check_defined ();
 516
 517   /* If the user did not define her ENDTOKEN, do it now. */
 518   if (!endtoken)
 519     {
 520       endtoken = symbol_get ("$end", empty_location);
 521       endtoken->class = token_sym;
 522       endtoken->number = 0;
 523       /* Value specified by POSIX.  */
 524       endtoken->user_token_number = 0;
 525     }
 526
 527   /* Insert the initial rule, which line is that of the first rule
 528      (not that of the start symbol):
 529
 530      accept: %start EOF.  */
 531   {
 532     symbol_list_t *p = symbol_list_new (accept, empty_location);
 533     p->location = grammar->location;
 534     p->next = symbol_list_new (startsymbol, empty_location);
 535     p->next->next = symbol_list_new (endtoken, empty_location);
 536     p->next->next->next = symbol_list_new (NULL, empty_location);
 537     p->next->next->next->next = grammar;
 538     nrules += 1;
 539     nritems += 3;
 540     grammar = p;
 541   }
 542
 543   if (! (nsyms <= SYMBOL_NUMBER_MAX && nsyms == ntokens + nvars))
 544     abort ();
 545
 546   xfclose (finput);
 547
 548   /* Assign the symbols their symbol numbers.  Write #defines for the
 549      token symbols into FDEFINES if requested.  */
 550   symbols_pack ();
 551
 552   /* Convert the grammar into the format described in gram.h.  */
 553   packgram ();
 554
 555   /* The grammar as a symbol_list_t is no longer needed. */
 556   LIST_FREE (symbol_list_t, grammar);
 557 }