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   if (!no_lines_flag)
  73     {
  74       obstack_fgrow2 (oout, muscle_find ("linef"),
  75                       location.first_line,
  76                       quotearg_style (c_quoting_style,
  77                                       muscle_find ("filename")));
  78     }
  79   obstack_sgrow (oout, prologue);
  80 }
  81
  82
  83
  84
  85 /*----------------------.
  86 | Handle the epilogue.  |
  87 `----------------------*/
  88
  89 void
  90 epilogue_set (const char *epilogue, location_t location)
  91 {
  92   if (!no_lines_flag)
  93     {
  94       obstack_fgrow2 (&muscle_obstack, muscle_find ("linef"),
  95                       location.first_line,
  96                       quotearg_style (c_quoting_style,
  97                                       muscle_find ("filename")));
  98     }
  99   obstack_sgrow (&muscle_obstack, epilogue);
 100   obstack_1grow (&muscle_obstack, 0);
 101   muscle_insert ("epilogue", obstack_finish (&muscle_obstack));
 102 }
 103
 104
 105 \f
 106
 107  /*-------------------------------------------------------------------.
 108 | Return the merger index for a merging function named NAME, whose   |
 109 | arguments have type TYPE.  Records the function, if new, in        |
 110 | merger_list.                                                       |
 111 `-------------------------------------------------------------------*/
 112
 113 static int
 114 get_merge_function (const char* name, const char* type,
 115                     location_t loc)
 116 {
 117   merger_list *syms;
 118   merger_list head;
 119   int n;
 120
 121   if (! glr_parser)
 122     return 0;
 123
 124   if (type == NULL)
 125     type = "";
 126
 127   head.next = merge_functions;
 128   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 129     if (strcmp (name, syms->next->name) == 0)
 130       break;
 131   if (syms->next == NULL)
 132     {
 133       syms->next = XMALLOC (merger_list, 1);
 134       syms->next->name = xstrdup (name);
 135       syms->next->type = xstrdup (type);
 136       syms->next->next = NULL;
 137       merge_functions = head.next;
 138     }
 139   else if (strcmp (type, syms->next->type) != 0)
 140     warn_at (loc, _("result type clash (`%s' `%s') on merge function %s"),
 141              type, syms->next->type, name);
 142   return n;
 143 }
 144
 145 /*--------------------------------------.
 146 | Free all merge-function definitions.  |
 147 `--------------------------------------*/
 148
 149 void
 150 free_merger_functions (void)
 151 {
 152   merger_list *L0;
 153   if (! glr_parser)
 154     return;
 155   L0 = merge_functions;
 156   while (L0 != NULL)
 157     {
 158       merger_list *L1 = L0->next;
 159       free (L0);
 160       L0 = L1;
 161     }
 162 }
 163
 164 \f
 165 /*-------------------------------------------------------------------.
 166 | Parse the input grammar into a one symbol_list_t structure.  Each  |
 167 | rule is represented by a sequence of symbols: the left hand side   |
 168 | followed by the contents of the right hand side, followed by a     |
 169 | null pointer instead of a symbol to terminate the rule.  The next  |
 170 | symbol is the lhs of the following rule.                           |
 171 |                                                                    |
 172 | All actions are copied out, labelled by the rule number they apply |
 173 | to.                                                                |
 174 |                                                                    |
 175 | Bison used to allow some %directives in the rules sections, but    |
 176 | this is no longer consider appropriate: (i) the documented grammar |
 177 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 178 | recovery for %directives consists in skipping the junk until a `%' |
 179 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 180 | in the rules section.                                              |
 181 `-------------------------------------------------------------------*/
 182
 183 /* The (currently) last symbol of GRAMMAR. */
 184 symbol_list_t *grammar_end = NULL;
 185
 186 /* Append S to the GRAMMAR. */
 187 void
 188 grammar_symbol_append (symbol_t *symbol, location_t location)
 189 {
 190   symbol_list_t *p = symbol_list_new (symbol, location);
 191
 192   if (grammar_end)
 193     grammar_end->next = p;
 194   else
 195     grammar = p;
 196
 197   grammar_end = p;
 198 }
 199
 200 /* The rule currently being defined, and the previous rule.
 201    CURRENT_RULE points to the first LHS of the current rule, while
 202    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 203 symbol_list_t *current_rule = NULL;
 204 symbol_list_t *previous_rule_end = NULL;
 205
 206
 207 /*----------------------------------------------.
 208 | Create a new rule for LHS in to the GRAMMAR.  |
 209 `----------------------------------------------*/
 210
 211 void
 212 grammar_rule_begin (symbol_t *lhs, location_t location)
 213 {
 214   if (!start_flag)
 215     {
 216       startsymbol = lhs;
 217       startsymbol_location = location;
 218       start_flag = 1;
 219     }
 220
 221   /* Start a new rule and record its lhs.  */
 222   ++nrules;
 223   ++nritems;
 224
 225   previous_rule_end = grammar_end;
 226   grammar_symbol_append (lhs, location);
 227   current_rule = grammar_end;
 228
 229   /* Mark the rule's lhs as a nonterminal if not already so.  */
 230
 231   if (lhs->class == unknown_sym)
 232     {
 233       lhs->class = nterm_sym;
 234       lhs->number = nvars;
 235       ++nvars;
 236     }
 237   else if (lhs->class == token_sym)
 238     complain_at (location, _("rule given for %s, which is a token"), lhs->tag);
 239 }
 240
 241 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 242    instance, there should be no type clash on the default action.  */
 243
 244 static void
 245 grammar_current_rule_check (void)
 246 {
 247   symbol_t *lhs = current_rule->sym;
 248   char const *lhs_type = lhs->type_name;
 249   symbol_t *first_rhs = current_rule->next->sym;
 250
 251   /* If there is an action, then there is nothing we can do: the user
 252      is allowed to shoot herself in the foot.  */
 253   if (current_rule->action)
 254     return;
 255
 256   /* Don't worry about the default action if $$ is untyped, since $$'s
 257      value can't be used.  */
 258   if (! lhs_type)
 259     return;
 260
 261   /* If $$ is being set in default way, report if any type mismatch.  */
 262   if (first_rhs)
 263     {
 264       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 265       if (strcmp (lhs_type, rhs_type))
 266         complain_at (current_rule->location,
 267                      _("type clash (`%s' `%s') on default action"),
 268                      lhs_type, rhs_type);
 269     }
 270   /* Warn if there is no default for $$ but we need one.  */
 271   else
 272     complain_at (current_rule->location,
 273                  _("empty rule for typed nonterminal, and no action"));
 274 }
 275
 276
 277 /*-------------------------------------.
 278 | End the currently being grown rule.  |
 279 `-------------------------------------*/
 280
 281 void
 282 grammar_rule_end (location_t location)
 283 {
 284   /* Put an empty link in the list to mark the end of this rule  */
 285   grammar_symbol_append (NULL, grammar_end->location);
 286   current_rule->location = location;
 287   grammar_current_rule_check ();
 288 }
 289
 290
 291 /*-------------------------------------------------------------------.
 292 | The previous action turns out the be a mid-rule action.  Attach it |
 293 | to the current rule, i.e., create a dummy symbol, attach it this   |
 294 | mid-rule action, and append this dummy nonterminal to the current  |
 295 | rule.                                                              |
 296 `-------------------------------------------------------------------*/
 297
 298 void
 299 grammar_midrule_action (void)
 300 {
 301   /* Since the action was written out with this rule's number, we must
 302      give the new rule this number by inserting the new rule before
 303      it.  */
 304
 305   /* Make a DUMMY nonterminal, whose location is that of the midrule
 306      action.  Create the MIDRULE.  */
 307   location_t dummy_location = current_rule->action_location;
 308   symbol_t *dummy = dummy_symbol_get (dummy_location);
 309   symbol_list_t *midrule = symbol_list_new (dummy, dummy_location);
 310
 311   /* Make a new rule, whose body is empty, before the current one, so
 312      that the action just read can belong to it.  */
 313   ++nrules;
 314   ++nritems;
 315   /* Attach its location and actions to that of the DUMMY.  */
 316   midrule->location = dummy_location;
 317   midrule->action = current_rule->action;
 318   midrule->action_location = dummy_location;
 319   current_rule->action = NULL;
 320
 321   if (previous_rule_end)
 322     previous_rule_end->next = midrule;
 323   else
 324     grammar = midrule;
 325
 326   /* End the dummy's rule.  */
 327   previous_rule_end = symbol_list_new (NULL, dummy_location);
 328   previous_rule_end->next = current_rule;
 329
 330   midrule->next = previous_rule_end;
 331
 332   /* Insert the dummy nonterminal replacing the midrule action into
 333      the current rule.  */
 334   grammar_current_rule_symbol_append (dummy, dummy_location);
 335 }
 336
 337 /* Set the precedence symbol of the current rule to PRECSYM. */
 338
 339 void
 340 grammar_current_rule_prec_set (symbol_t *precsym, location_t location)
 341 {
 342   if (current_rule->ruleprec)
 343     complain_at (location, _("only one %s allowed per rule"), "%prec");
 344   current_rule->ruleprec = precsym;
 345 }
 346
 347 /* Attach dynamic precedence DPREC to the current rule. */
 348
 349 void
 350 grammar_current_rule_dprec_set (int dprec, location_t location)
 351 {
 352   if (! glr_parser)
 353     warn_at (location, _("%s affects only GLR parsers"), "%dprec");
 354   if (dprec <= 0)
 355     complain_at (location,
 356                  _("%s must be followed by positive number"), "%dprec");
 357   else if (current_rule->dprec != 0)
 358     complain_at (location, _("only one %s allowed per rule"), "%dprec");
 359   current_rule->dprec = dprec;
 360 }
 361
 362 /* Attach a merge function NAME with argument type TYPE to current
 363    rule. */
 364
 365 void
 366 grammar_current_rule_merge_set (const char* name, location_t location)
 367 {
 368   if (! glr_parser)
 369     warn_at (location, _("%s affects only GLR parsers"), "%merge");
 370   if (current_rule->merger != 0)
 371     complain_at (location, _("only one %s allowed per rule"), "%merge");
 372   current_rule->merger =
 373     get_merge_function (name, current_rule->sym->type_name, location);
 374 }
 375
 376 /* Attach a SYMBOL to the current rule.  If needed, move the previous
 377    action as a mid-rule action.  */
 378
 379 void
 380 grammar_current_rule_symbol_append (symbol_t *symbol, location_t location)
 381 {
 382   if (current_rule->action)
 383     grammar_midrule_action ();
 384   ++nritems;
 385   grammar_symbol_append (symbol, location);
 386 }
 387
 388 /* Attach an ACTION to the current rule.  If needed, move the previous
 389    action as a mid-rule action.  */
 390
 391 void
 392 grammar_current_rule_action_append (const char *action, location_t location)
 393 {
 394   if (current_rule->action)
 395     grammar_midrule_action ();
 396   current_rule->action = action;
 397   current_rule->action_location = location;
 398 }
 399
 400 \f
 401 /*---------------------------------------------------------------.
 402 | Convert the rules into the representation using RRHS, RLHS and |
 403 | RITEM.                                                         |
 404 `---------------------------------------------------------------*/
 405
 406 static void
 407 packgram (void)
 408 {
 409   unsigned int itemno = 0;
 410   rule_number_t ruleno = 0;
 411   symbol_list_t *p = grammar;
 412
 413   ritem = XCALLOC (item_number_t, nritems);
 414   rules = XCALLOC (rule_t, nrules);
 415
 416   while (p)
 417     {
 418       symbol_t *ruleprec = p->ruleprec;
 419       rules[ruleno].user_number = ruleno;
 420       rules[ruleno].number = ruleno;
 421       rules[ruleno].lhs = p->sym;
 422       rules[ruleno].rhs = ritem + itemno;
 423       rules[ruleno].location = p->location;
 424       rules[ruleno].useful = true;
 425       rules[ruleno].action = p->action;
 426       rules[ruleno].action_location = p->action_location;
 427       rules[ruleno].dprec = p->dprec;
 428       rules[ruleno].merger = p->merger;
 429
 430       p = p->next;
 431       while (p && p->sym)
 432         {
 433           /* item_number_t = symbol_number_t.
 434              But the former needs to contain more: negative rule numbers. */
 435           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 436           /* A rule gets by default the precedence and associativity
 437              of the last token in it.  */
 438           if (p->sym->class == token_sym)
 439             rules[ruleno].prec = p->sym;
 440           if (p)
 441             p = p->next;
 442         }
 443
 444       /* If this rule has a %prec,
 445          the specified symbol's precedence replaces the default.  */
 446       if (ruleprec)
 447         {
 448           rules[ruleno].precsym = ruleprec;
 449           rules[ruleno].prec = ruleprec;
 450         }
 451       ritem[itemno++] = rule_number_as_item_number (ruleno);
 452       ++ruleno;
 453
 454       if (p)
 455         p = p->next;
 456     }
 457
 458   assert (itemno == nritems);
 459
 460   if (trace_flag & trace_sets)
 461     ritem_print (stderr);
 462 }
 463 \f
 464 /*------------------------------------------------------------------.
 465 | Read in the grammar specification and record it in the format     |
 466 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 467 | in each case forming the body of a C function (YYACTION) which    |
 468 | contains a switch statement to decide which action to execute.    |
 469 `------------------------------------------------------------------*/
 470
 471 void
 472 reader (void)
 473 {
 474   gram_control_t gram_control;
 475
 476   /* Initialize the symbol table.  */
 477   symbols_new ();
 478
 479   /* Construct the accept symbol. */
 480   accept = symbol_get ("$accept", empty_location);
 481   accept->class = nterm_sym;
 482   accept->number = nvars++;
 483
 484   /* Construct the error token */
 485   errtoken = symbol_get ("error", empty_location);
 486   errtoken->class = token_sym;
 487   errtoken->number = ntokens++;
 488
 489   /* Construct a token that represents all undefined literal tokens.
 490      It is always token number 2.  */
 491   undeftoken = symbol_get ("$undefined", empty_location);
 492   undeftoken->class = token_sym;
 493   undeftoken->number = ntokens++;
 494
 495   /* Initialize the obstacks. */
 496   obstack_init (&pre_prologue_obstack);
 497   obstack_init (&post_prologue_obstack);
 498
 499   finput = xfopen (infile, "r");
 500   gram_in = finput;
 501
 502   gram__flex_debug = trace_flag & trace_scan;
 503   gram_debug = trace_flag & trace_parse;
 504   scanner_initialize ();
 505   gram_parse (&gram_control);
 506
 507   /* If something went wrong during the parsing, don't try to
 508      continue.  */
 509   if (complaint_issued)
 510     return;
 511
 512   /* Grammar has been read.  Do some checking */
 513   if (nrules == 0)
 514     fatal (_("no rules in the input grammar"));
 515
 516   /* Report any undefined symbols and consider them nonterminals.  */
 517   symbols_check_defined ();
 518
 519   /* If the user did not define her ENDTOKEN, do it now. */
 520   if (!endtoken)
 521     {
 522       endtoken = symbol_get ("$end", empty_location);
 523       endtoken->class = token_sym;
 524       endtoken->number = 0;
 525       /* Value specified by POSIX.  */
 526       endtoken->user_token_number = 0;
 527     }
 528
 529   /* Insert the initial rule, which line is that of the first rule
 530      (not that of the start symbol):
 531
 532      accept: %start EOF.  */
 533   {
 534     symbol_list_t *p = symbol_list_new (accept, empty_location);
 535     p->location = grammar->location;
 536     p->next = symbol_list_new (startsymbol, empty_location);
 537     p->next->next = symbol_list_new (endtoken, empty_location);
 538     p->next->next->next = symbol_list_new (NULL, empty_location);
 539     p->next->next->next->next = grammar;
 540     nrules += 1;
 541     nritems += 3;
 542     grammar = p;
 543   }
 544
 545   if (SYMBOL_NUMBER_MAX < nsyms)
 546     fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
 547            SYMBOL_NUMBER_MAX);
 548
 549   assert (nsyms == ntokens + nvars);
 550
 551   xfclose (finput);
 552
 553   /* Assign the symbols their symbol numbers.  Write #defines for the
 554      token symbols into FDEFINES if requested.  */
 555   symbols_pack ();
 556
 557   /* Convert the grammar into the format described in gram.h.  */
 558   packgram ();
 559
 560   /* The grammar as a symbol_list_t is no longer needed. */
 561   LIST_FREE (symbol_list_t, grammar);
 562 }