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 (const char* name, const char* type,
 110                     location_t loc)
 111 {
 112   merger_list *syms;
 113   merger_list head;
 114   int n;
 115
 116   if (! glr_parser)
 117     return 0;
 118
 119   if (type == NULL)
 120     type = "";
 121
 122   head.next = merge_functions;
 123   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 124     if (strcmp (name, syms->next->name) == 0)
 125       break;
 126   if (syms->next == NULL)
 127     {
 128       syms->next = XMALLOC (merger_list, 1);
 129       syms->next->name = xstrdup (name);
 130       syms->next->type = xstrdup (type);
 131       syms->next->next = NULL;
 132       merge_functions = head.next;
 133     }
 134   else if (strcmp (type, syms->next->type) != 0)
 135     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 136              name, type, syms->next->type);
 137   return n;
 138 }
 139
 140 /*--------------------------------------.
 141 | Free all merge-function definitions.  |
 142 `--------------------------------------*/
 143
 144 void
 145 free_merger_functions (void)
 146 {
 147   merger_list *L0;
 148   if (! glr_parser)
 149     return;
 150   L0 = merge_functions;
 151   while (L0 != NULL)
 152     {
 153       merger_list *L1 = L0->next;
 154       free (L0);
 155       L0 = L1;
 156     }
 157 }
 158
 159 \f
 160 /*-------------------------------------------------------------------.
 161 | Parse the input grammar into a one symbol_list_t structure.  Each  |
 162 | rule is represented by a sequence of symbols: the left hand side   |
 163 | followed by the contents of the right hand side, followed by a     |
 164 | null pointer instead of a symbol to terminate the rule.  The next  |
 165 | symbol is the lhs of the following rule.                           |
 166 |                                                                    |
 167 | All actions are copied out, labelled by the rule number they apply |
 168 | to.                                                                |
 169 |                                                                    |
 170 | Bison used to allow some %directives in the rules sections, but    |
 171 | this is no longer consider appropriate: (i) the documented grammar |
 172 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 173 | recovery for %directives consists in skipping the junk until a `%' |
 174 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 175 | in the rules section.                                              |
 176 `-------------------------------------------------------------------*/
 177
 178 /* The (currently) last symbol of GRAMMAR. */
 179 symbol_list_t *grammar_end = NULL;
 180
 181 /* Append S to the GRAMMAR. */
 182 void
 183 grammar_symbol_append (symbol_t *symbol, location_t location)
 184 {
 185   symbol_list_t *p = symbol_list_new (symbol, location);
 186
 187   if (grammar_end)
 188     grammar_end->next = p;
 189   else
 190     grammar = p;
 191
 192   grammar_end = p;
 193 }
 194
 195 /* The rule currently being defined, and the previous rule.
 196    CURRENT_RULE points to the first LHS of the current rule, while
 197    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 198 symbol_list_t *current_rule = NULL;
 199 symbol_list_t *previous_rule_end = NULL;
 200
 201
 202 /*----------------------------------------------.
 203 | Create a new rule for LHS in to the GRAMMAR.  |
 204 `----------------------------------------------*/
 205
 206 void
 207 grammar_rule_begin (symbol_t *lhs, location_t location)
 208 {
 209   if (!start_flag)
 210     {
 211       startsymbol = lhs;
 212       startsymbol_location = location;
 213       start_flag = 1;
 214     }
 215
 216   /* Start a new rule and record its lhs.  */
 217   ++nrules;
 218   ++nritems;
 219
 220   previous_rule_end = grammar_end;
 221   grammar_symbol_append (lhs, location);
 222   current_rule = grammar_end;
 223
 224   /* Mark the rule's lhs as a nonterminal if not already so.  */
 225
 226   if (lhs->class == unknown_sym)
 227     {
 228       lhs->class = nterm_sym;
 229       lhs->number = nvars;
 230       ++nvars;
 231     }
 232   else if (lhs->class == token_sym)
 233     complain_at (location, _("rule given for %s, which is a token"), lhs->tag);
 234 }
 235
 236 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 237    instance, there should be no type clash on the default action.  */
 238
 239 static void
 240 grammar_current_rule_check (void)
 241 {
 242   symbol_t *lhs = current_rule->sym;
 243   char const *lhs_type = lhs->type_name;
 244   symbol_t *first_rhs = current_rule->next->sym;
 245
 246   /* If there is an action, then there is nothing we can do: the user
 247      is allowed to shoot herself in the foot.  */
 248   if (current_rule->action)
 249     return;
 250
 251   /* Don't worry about the default action if $$ is untyped, since $$'s
 252      value can't be used.  */
 253   if (! lhs_type)
 254     return;
 255
 256   /* If $$ is being set in default way, report if any type mismatch.  */
 257   if (first_rhs)
 258     {
 259       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 260       if (strcmp (lhs_type, rhs_type))
 261         complain_at (current_rule->location,
 262                      _("type clash on default action: <%s> != <%s>"),
 263                      lhs_type, rhs_type);
 264     }
 265   /* Warn if there is no default for $$ but we need one.  */
 266   else
 267     complain_at (current_rule->location,
 268                  _("empty rule for typed nonterminal, and no action"));
 269 }
 270
 271
 272 /*-------------------------------------.
 273 | End the currently being grown rule.  |
 274 `-------------------------------------*/
 275
 276 void
 277 grammar_rule_end (location_t location)
 278 {
 279   /* Put an empty link in the list to mark the end of this rule  */
 280   grammar_symbol_append (NULL, grammar_end->location);
 281   current_rule->location = location;
 282   grammar_current_rule_check ();
 283 }
 284
 285
 286 /*-------------------------------------------------------------------.
 287 | The previous action turns out the be a mid-rule action.  Attach it |
 288 | to the current rule, i.e., create a dummy symbol, attach it this   |
 289 | mid-rule action, and append this dummy nonterminal to the current  |
 290 | rule.                                                              |
 291 `-------------------------------------------------------------------*/
 292
 293 void
 294 grammar_midrule_action (void)
 295 {
 296   /* Since the action was written out with this rule's number, we must
 297      give the new rule this number by inserting the new rule before
 298      it.  */
 299
 300   /* Make a DUMMY nonterminal, whose location is that of the midrule
 301      action.  Create the MIDRULE.  */
 302   location_t dummy_location = current_rule->action_location;
 303   symbol_t *dummy = dummy_symbol_get (dummy_location);
 304   symbol_list_t *midrule = symbol_list_new (dummy, dummy_location);
 305
 306   /* Make a new rule, whose body is empty, before the current one, so
 307      that the action just read can belong to it.  */
 308   ++nrules;
 309   ++nritems;
 310   /* Attach its location and actions to that of the DUMMY.  */
 311   midrule->location = dummy_location;
 312   midrule->action = current_rule->action;
 313   midrule->action_location = dummy_location;
 314   current_rule->action = NULL;
 315
 316   if (previous_rule_end)
 317     previous_rule_end->next = midrule;
 318   else
 319     grammar = midrule;
 320
 321   /* End the dummy's rule.  */
 322   previous_rule_end = symbol_list_new (NULL, dummy_location);
 323   previous_rule_end->next = current_rule;
 324
 325   midrule->next = previous_rule_end;
 326
 327   /* Insert the dummy nonterminal replacing the midrule action into
 328      the current rule.  */
 329   grammar_current_rule_symbol_append (dummy, dummy_location);
 330 }
 331
 332 /* Set the precedence symbol of the current rule to PRECSYM. */
 333
 334 void
 335 grammar_current_rule_prec_set (symbol_t *precsym, location_t location)
 336 {
 337   if (current_rule->ruleprec)
 338     complain_at (location, _("only one %s allowed per rule"), "%prec");
 339   current_rule->ruleprec = precsym;
 340 }
 341
 342 /* Attach dynamic precedence DPREC to the current rule. */
 343
 344 void
 345 grammar_current_rule_dprec_set (int dprec, location_t location)
 346 {
 347   if (! glr_parser)
 348     warn_at (location, _("%s affects only GLR parsers"), "%dprec");
 349   if (dprec <= 0)
 350     complain_at (location,
 351                  _("%s must be followed by positive number"), "%dprec");
 352   else if (current_rule->dprec != 0)
 353     complain_at (location, _("only one %s allowed per rule"), "%dprec");
 354   current_rule->dprec = dprec;
 355 }
 356
 357 /* Attach a merge function NAME with argument type TYPE to current
 358    rule. */
 359
 360 void
 361 grammar_current_rule_merge_set (const char* name, location_t location)
 362 {
 363   if (! glr_parser)
 364     warn_at (location, _("%s affects only GLR parsers"), "%merge");
 365   if (current_rule->merger != 0)
 366     complain_at (location, _("only one %s allowed per rule"), "%merge");
 367   current_rule->merger =
 368     get_merge_function (name, current_rule->sym->type_name, location);
 369 }
 370
 371 /* Attach a SYMBOL to the current rule.  If needed, move the previous
 372    action as a mid-rule action.  */
 373
 374 void
 375 grammar_current_rule_symbol_append (symbol_t *symbol, location_t location)
 376 {
 377   if (current_rule->action)
 378     grammar_midrule_action ();
 379   ++nritems;
 380   grammar_symbol_append (symbol, location);
 381 }
 382
 383 /* Attach an ACTION to the current rule.  If needed, move the previous
 384    action as a mid-rule action.  */
 385
 386 void
 387 grammar_current_rule_action_append (const char *action, location_t location)
 388 {
 389   if (current_rule->action)
 390     grammar_midrule_action ();
 391   current_rule->action = action;
 392   current_rule->action_location = location;
 393 }
 394
 395 \f
 396 /*---------------------------------------------------------------.
 397 | Convert the rules into the representation using RRHS, RLHS and |
 398 | RITEM.                                                         |
 399 `---------------------------------------------------------------*/
 400
 401 static void
 402 packgram (void)
 403 {
 404   unsigned int itemno = 0;
 405   rule_number_t ruleno = 0;
 406   symbol_list_t *p = grammar;
 407
 408   ritem = XCALLOC (item_number_t, nritems);
 409   rules = XCALLOC (rule_t, nrules);
 410
 411   while (p)
 412     {
 413       symbol_t *ruleprec = p->ruleprec;
 414       rules[ruleno].user_number = ruleno;
 415       rules[ruleno].number = ruleno;
 416       rules[ruleno].lhs = p->sym;
 417       rules[ruleno].rhs = ritem + itemno;
 418       rules[ruleno].location = p->location;
 419       rules[ruleno].useful = true;
 420       rules[ruleno].action = p->action;
 421       rules[ruleno].action_location = p->action_location;
 422       rules[ruleno].dprec = p->dprec;
 423       rules[ruleno].merger = p->merger;
 424
 425       p = p->next;
 426       while (p && p->sym)
 427         {
 428           /* item_number_t = symbol_number_t.
 429              But the former needs to contain more: negative rule numbers. */
 430           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 431           /* A rule gets by default the precedence and associativity
 432              of the last token in it.  */
 433           if (p->sym->class == token_sym)
 434             rules[ruleno].prec = p->sym;
 435           if (p)
 436             p = p->next;
 437         }
 438
 439       /* If this rule has a %prec,
 440          the specified symbol's precedence replaces the default.  */
 441       if (ruleprec)
 442         {
 443           rules[ruleno].precsym = ruleprec;
 444           rules[ruleno].prec = ruleprec;
 445         }
 446       ritem[itemno++] = rule_number_as_item_number (ruleno);
 447       ++ruleno;
 448
 449       if (p)
 450         p = p->next;
 451     }
 452
 453   assert (itemno == nritems);
 454
 455   if (trace_flag & trace_sets)
 456     ritem_print (stderr);
 457 }
 458 \f
 459 /*------------------------------------------------------------------.
 460 | Read in the grammar specification and record it in the format     |
 461 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 462 | in each case forming the body of a C function (YYACTION) which    |
 463 | contains a switch statement to decide which action to execute.    |
 464 `------------------------------------------------------------------*/
 465
 466 void
 467 reader (void)
 468 {
 469   gram_control_t gram_control;
 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 (infile, "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 (&gram_control);
 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_t *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 (SYMBOL_NUMBER_MAX < nsyms)
 541     fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
 542            SYMBOL_NUMBER_MAX);
 543
 544   assert (nsyms == ntokens + nvars);
 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 }