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