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 #include "system.h"
  23
  24 #include <quotearg.h>
  25
  26 #include "complain.h"
  27 #include "conflicts.h"
  28 #include "files.h"
  29 #include "getargs.h"
  30 #include "gram.h"
  31 #include "muscle_tab.h"
  32 #include "output.h"
  33 #include "reader.h"
  34 #include "symlist.h"
  35 #include "symtab.h"
  36
  37 static symbol_list *grammar = NULL;
  38 static int start_flag = 0;
  39 merger_list *merge_functions;
  40
  41 /* Nonzero if %union has been seen.  */
  42 int typed = 0;
  43 \f
  44 /*-----------------------.
  45 | Set the start symbol.  |
  46 `-----------------------*/
  47
  48 void
  49 grammar_start_symbol_set (symbol *s, location loc)
  50 {
  51   if (start_flag)
  52     complain_at (loc, _("multiple %s declarations"), "%start");
  53   else
  54     {
  55       start_flag = 1;
  56       startsymbol = s;
  57       startsymbol_location = loc;
  58     }
  59 }
  60
  61
  62 /*----------------------------------------------------------------.
  63 | There are two prologues: one before %union, one after.  Augment |
  64 | the current one.                                                |
  65 `----------------------------------------------------------------*/
  66
  67 void
  68 prologue_augment (const char *prologue, location loc)
  69 {
  70   struct obstack *oout =
  71     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  72
  73   obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[", loc.start.line);
  74   MUSCLE_OBSTACK_SGROW (oout,
  75                         quotearg_style (c_quoting_style, loc.start.file));
  76   obstack_sgrow (oout, "]])[\n");
  77   obstack_sgrow (oout, prologue);
  78 }
  79
  80
  81
  82
  83 /*----------------------.
  84 | Handle the epilogue.  |
  85 `----------------------*/
  86
  87 void
  88 epilogue_augment (const char *epilogue, location loc)
  89 {
  90   char *extension = NULL;
  91   obstack_fgrow1 (&muscle_obstack, "]b4_syncline([[%d]], [[", loc.start.line);
  92   MUSCLE_OBSTACK_SGROW (&muscle_obstack,
  93                         quotearg_style (c_quoting_style, loc.start.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 (uniqstr name, uniqstr type, location 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 = uniqstr_new ("");
 123
 124   head.next = merge_functions;
 125   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 126     if (UNIQSTR_EQ (name, syms->next->name))
 127       break;
 128   if (syms->next == NULL)
 129     {
 130       syms->next = XMALLOC (merger_list, 1);
 131       syms->next->name = uniqstr_new (name);
 132       syms->next->type = uniqstr_new (type);
 133       syms->next->next = NULL;
 134       merge_functions = head.next;
 135     }
 136   else if (!UNIQSTR_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 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 *grammar_end = NULL;
 182
 183 /* Append S to the GRAMMAR. */
 184 void
 185 grammar_symbol_append (symbol *sym, location loc)
 186 {
 187   symbol_list *p = symbol_list_new (sym, loc);
 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 *current_rule = NULL;
 201 symbol_list *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 *lhs, location loc)
 210 {
 211   if (!start_flag)
 212     {
 213       startsymbol = lhs;
 214       startsymbol_location = loc;
 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, loc);
 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 (loc, _("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 *lhs = current_rule->sym;
 245   char const *lhs_type = lhs->type_name;
 246   symbol *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 $$ is being set in default way, report if any type mismatch.  */
 259   if (first_rhs)
 260     {
 261       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 262       if (!UNIQSTR_EQ (lhs_type, rhs_type))
 263         warn_at (current_rule->location,
 264                  _("type clash on default action: <%s> != <%s>"),
 265                  lhs_type, rhs_type);
 266     }
 267   /* Warn if there is no default for $$ but we need one.  */
 268   else
 269     warn_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 loc)
 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 = loc;
 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 dummy_location = current_rule->action_location;
 305   symbol *dummy = dummy_symbol_get (dummy_location);
 306   symbol_list *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 *precsym, location loc)
 338 {
 339   if (current_rule->ruleprec)
 340     complain_at (loc, _("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 loc)
 348 {
 349   if (! glr_parser)
 350     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 351   if (dprec <= 0)
 352     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 353   else if (current_rule->dprec != 0)
 354     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 355   current_rule->dprec = dprec;
 356 }
 357
 358 /* Attach a merge function NAME with argument type TYPE to current
 359    rule. */
 360
 361 void
 362 grammar_current_rule_merge_set (uniqstr name, location loc)
 363 {
 364   if (! glr_parser)
 365     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 366   if (current_rule->merger != 0)
 367     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 368   current_rule->merger =
 369     get_merge_function (name, current_rule->sym->type_name, loc);
 370 }
 371
 372 /* Attach SYM to the current rule.  If needed, move the previous
 373    action as a mid-rule action.  */
 374
 375 void
 376 grammar_current_rule_symbol_append (symbol *sym, location loc)
 377 {
 378   if (current_rule->action)
 379     grammar_midrule_action ();
 380   ++nritems;
 381   grammar_symbol_append (sym, loc);
 382 }
 383
 384 /* Attach an ACTION to the current rule.  If needed, move the previous
 385    action as a mid-rule action.  */
 386
 387 void
 388 grammar_current_rule_action_append (const char *action, location loc)
 389 {
 390   if (current_rule->action)
 391     grammar_midrule_action ();
 392   current_rule->action = action;
 393   current_rule->action_location = loc;
 394 }
 395
 396 \f
 397 /*---------------------------------------------------------------.
 398 | Convert the rules into the representation using RRHS, RLHS and |
 399 | RITEM.                                                         |
 400 `---------------------------------------------------------------*/
 401
 402 static void
 403 packgram (void)
 404 {
 405   unsigned int itemno = 0;
 406   rule_number ruleno = 0;
 407   symbol_list *p = grammar;
 408
 409   ritem = XCALLOC (item_number, nritems);
 410   rules = XCALLOC (rule, nrules);
 411
 412   while (p)
 413     {
 414       symbol *ruleprec = p->ruleprec;
 415       rules[ruleno].user_number = ruleno;
 416       rules[ruleno].number = ruleno;
 417       rules[ruleno].lhs = p->sym;
 418       rules[ruleno].rhs = ritem + itemno;
 419       rules[ruleno].location = p->location;
 420       rules[ruleno].useful = true;
 421       rules[ruleno].action = p->action;
 422       rules[ruleno].action_location = p->action_location;
 423       rules[ruleno].dprec = p->dprec;
 424       rules[ruleno].merger = p->merger;
 425
 426       p = p->next;
 427       while (p && p->sym)
 428         {
 429           /* item_number = symbol_number.
 430              But the former needs to contain more: negative rule numbers. */
 431           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 432           /* A rule gets by default the precedence and associativity
 433              of the last token in it.  */
 434           if (p->sym->class == token_sym)
 435             rules[ruleno].prec = p->sym;
 436           if (p)
 437             p = p->next;
 438         }
 439
 440       /* If this rule has a %prec,
 441          the specified symbol's precedence replaces the default.  */
 442       if (ruleprec)
 443         {
 444           rules[ruleno].precsym = ruleprec;
 445           rules[ruleno].prec = ruleprec;
 446         }
 447       ritem[itemno++] = rule_number_as_item_number (ruleno);
 448       ++ruleno;
 449
 450       if (p)
 451         p = p->next;
 452     }
 453
 454   if (itemno != nritems)
 455     abort ();
 456
 457   if (trace_flag & trace_sets)
 458     ritem_print (stderr);
 459 }
 460 \f
 461 /*------------------------------------------------------------------.
 462 | Read in the grammar specification and record it in the format     |
 463 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 464 | in each case forming the body of a C function (YYACTION) which    |
 465 | contains a switch statement to decide which action to execute.    |
 466 `------------------------------------------------------------------*/
 467
 468 void
 469 reader (void)
 470 {
 471   /* Initialize the symbol table.  */
 472   symbols_new ();
 473
 474   /* Construct the accept symbol. */
 475   accept = symbol_get ("$accept", empty_location);
 476   accept->class = nterm_sym;
 477   accept->number = nvars++;
 478
 479   /* Construct the error token */
 480   errtoken = symbol_get ("error", empty_location);
 481   errtoken->class = token_sym;
 482   errtoken->number = ntokens++;
 483
 484   /* Construct a token that represents all undefined literal tokens.
 485      It is always token number 2.  */
 486   undeftoken = symbol_get ("$undefined", empty_location);
 487   undeftoken->class = token_sym;
 488   undeftoken->number = ntokens++;
 489
 490   /* Initialize the obstacks. */
 491   obstack_init (&pre_prologue_obstack);
 492   obstack_init (&post_prologue_obstack);
 493
 494   finput = xfopen (grammar_file, "r");
 495   gram_in = finput;
 496
 497   gram__flex_debug = trace_flag & trace_scan;
 498   gram_debug = trace_flag & trace_parse;
 499   scanner_initialize ();
 500   gram_parse ();
 501
 502   /* If something went wrong during the parsing, don't try to
 503      continue.  */
 504   if (complaint_issued)
 505     return;
 506
 507   /* Grammar has been read.  Do some checking */
 508   if (nrules == 0)
 509     fatal (_("no rules in the input grammar"));
 510
 511   /* Report any undefined symbols and consider them nonterminals.  */
 512   symbols_check_defined ();
 513
 514   /* If the user did not define her ENDTOKEN, do it now. */
 515   if (!endtoken)
 516     {
 517       endtoken = symbol_get ("$end", empty_location);
 518       endtoken->class = token_sym;
 519       endtoken->number = 0;
 520       /* Value specified by POSIX.  */
 521       endtoken->user_token_number = 0;
 522     }
 523
 524   /* Insert the initial rule, which line is that of the first rule
 525      (not that of the start symbol):
 526
 527      accept: %start EOF.  */
 528   {
 529     symbol_list *p = symbol_list_new (accept, empty_location);
 530     p->location = grammar->location;
 531     p->next = symbol_list_new (startsymbol, empty_location);
 532     p->next->next = symbol_list_new (endtoken, empty_location);
 533     p->next->next->next = symbol_list_new (NULL, empty_location);
 534     p->next->next->next->next = grammar;
 535     nrules += 1;
 536     nritems += 3;
 537     grammar = p;
 538   }
 539
 540   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 541     abort ();
 542
 543   xfclose (finput);
 544
 545   /* Assign the symbols their symbol numbers.  Write #defines for the
 546      token symbols into FDEFINES if requested.  */
 547   symbols_pack ();
 548
 549   /* Convert the grammar into the format described in gram.h.  */
 550   packgram ();
 551
 552   /* The grammar as a symbol_list is no longer needed. */
 553   LIST_FREE (symbol_list, grammar);
 554 }