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