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 "quote.h"
  26 #include "getargs.h"
  27 #include "files.h"
  28 #include "symtab.h"
  29 #include "symlist.h"
  30 #include "gram.h"
  31 #include "complain.h"
  32 #include "output.h"
  33 #include "reader.h"
  34 #include "conflicts.h"
  35 #include "muscle_tab.h"
  36
  37 static symbol_list_t *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_t *s, location_t l)
  50 {
  51   if (start_flag)
  52     complain_at (l, _("multiple %s declarations"), "%start");
  53   else
  54     {
  55       start_flag = 1;
  56       startsymbol = s;
  57       startsymbol_location = l;
  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_t location)
  69 {
  70   struct obstack *oout =
  71     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  72
  73   if (!no_lines_flag)
  74     {
  75       obstack_fgrow2 (oout, muscle_find ("linef"),
  76                       location.first_line,
  77                       quotearg_style (c_quoting_style,
  78                                       muscle_find ("filename")));
  79     }
  80   obstack_sgrow (oout, prologue);
  81 }
  82
  83
  84
  85
  86 /*----------------------.
  87 | Handle the epilogue.  |
  88 `----------------------*/
  89
  90 void
  91 epilogue_set (const char *epilogue, location_t location)
  92 {
  93   if (!no_lines_flag)
  94     {
  95       obstack_fgrow2 (&muscle_obstack, muscle_find ("linef"),
  96                       location.first_line,
  97                       quotearg_style (c_quoting_style,
  98                                       muscle_find ("filename")));
  99     }
 100   obstack_sgrow (&muscle_obstack, epilogue);
 101   obstack_1grow (&muscle_obstack, 0);
 102   muscle_insert ("epilogue", obstack_finish (&muscle_obstack));
 103 }
 104
 105
 106 \f
 107
 108  /*-------------------------------------------------------------------.
 109 | Return the merger index for a merging function named NAME, whose   |
 110 | arguments have type TYPE.  Records the function, if new, in        |
 111 | merger_list.                                                       |
 112 `-------------------------------------------------------------------*/
 113
 114 static int
 115 get_merge_function (const char* name, const char* type,
 116                     location_t loc)
 117 {
 118   merger_list *syms;
 119   merger_list head;
 120   int n;
 121
 122   if (! glr_parser)
 123     return 0;
 124
 125   if (type == NULL)
 126     type = "";
 127
 128   head.next = merge_functions;
 129   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 130     if (strcmp (name, syms->next->name) == 0)
 131       break;
 132   if (syms->next == NULL)
 133     {
 134       syms->next = XMALLOC (merger_list, 1);
 135       syms->next->name = xstrdup (name);
 136       syms->next->type = xstrdup (type);
 137       syms->next->next = NULL;
 138       merge_functions = head.next;
 139     }
 140   else if (strcmp (type, syms->next->type) != 0)
 141     warn_at (loc, _("result type clash on merge function %s: `%s' vs. `%s'"),
 142              name, type, syms->next->type);
 143   return n;
 144 }
 145
 146 /*--------------------------------------.
 147 | Free all merge-function definitions.  |
 148 `--------------------------------------*/
 149
 150 void
 151 free_merger_functions (void)
 152 {
 153   merger_list *L0;
 154   if (! glr_parser)
 155     return;
 156   L0 = merge_functions;
 157   while (L0 != NULL)
 158     {
 159       merger_list *L1 = L0->next;
 160       free (L0);
 161       L0 = L1;
 162     }
 163 }
 164
 165 \f
 166 /*-------------------------------------------------------------------.
 167 | Parse the input grammar into a one symbol_list_t structure.  Each  |
 168 | rule is represented by a sequence of symbols: the left hand side   |
 169 | followed by the contents of the right hand side, followed by a     |
 170 | null pointer instead of a symbol to terminate the rule.  The next  |
 171 | symbol is the lhs of the following rule.                           |
 172 |                                                                    |
 173 | All actions are copied out, labelled by the rule number they apply |
 174 | to.                                                                |
 175 |                                                                    |
 176 | Bison used to allow some %directives in the rules sections, but    |
 177 | this is no longer consider appropriate: (i) the documented grammar |
 178 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 179 | recovery for %directives consists in skipping the junk until a `%' |
 180 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 181 | in the rules section.                                              |
 182 `-------------------------------------------------------------------*/
 183
 184 /* The (currently) last symbol of GRAMMAR. */
 185 symbol_list_t *grammar_end = NULL;
 186
 187 /* Append S to the GRAMMAR. */
 188 void
 189 grammar_symbol_append (symbol_t *symbol, location_t location)
 190 {
 191   symbol_list_t *p = symbol_list_new (symbol, location);
 192
 193   if (grammar_end)
 194     grammar_end->next = p;
 195   else
 196     grammar = p;
 197
 198   grammar_end = p;
 199 }
 200
 201 /* The rule currently being defined, and the previous rule.
 202    CURRENT_RULE points to the first LHS of the current rule, while
 203    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 204 symbol_list_t *current_rule = NULL;
 205 symbol_list_t *previous_rule_end = NULL;
 206
 207
 208 /*----------------------------------------------.
 209 | Create a new rule for LHS in to the GRAMMAR.  |
 210 `----------------------------------------------*/
 211
 212 void
 213 grammar_rule_begin (symbol_t *lhs, location_t location)
 214 {
 215   if (!start_flag)
 216     {
 217       startsymbol = lhs;
 218       startsymbol_location = location;
 219       start_flag = 1;
 220     }
 221
 222   /* Start a new rule and record its lhs.  */
 223   ++nrules;
 224   ++nritems;
 225
 226   previous_rule_end = grammar_end;
 227   grammar_symbol_append (lhs, location);
 228   current_rule = grammar_end;
 229
 230   /* Mark the rule's lhs as a nonterminal if not already so.  */
 231
 232   if (lhs->class == unknown_sym)
 233     {
 234       lhs->class = nterm_sym;
 235       lhs->number = nvars;
 236       ++nvars;
 237     }
 238   else if (lhs->class == token_sym)
 239     complain_at (location, _("rule given for %s, which is a token"), lhs->tag);
 240 }
 241
 242 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 243    instance, there should be no type clash on the default action.  */
 244
 245 static void
 246 grammar_current_rule_check (void)
 247 {
 248   symbol_t *lhs = current_rule->sym;
 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 in her foot.  */
 253   if (current_rule->action)
 254     return;
 255
 256   /* If $$ is being set in default way, report if any type mismatch.
 257      */
 258   if (first_rhs)
 259     {
 260       const char *lhs_type = lhs->type_name       ? lhs->type_name       : "";
 261       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 262       if (strcmp (lhs_type, rhs_type))
 263         complain_at (current_rule->location,
 264                      _("type clash (`%s' `%s') on default action"),
 265                      lhs_type, rhs_type);
 266     }
 267   /* Warn if there is no default for $$ but we need one.  */
 268   else
 269     {
 270       if (lhs->type_name)
 271         complain_at (current_rule->location,
 272                      _("empty rule for typed nonterminal, and no action"));
 273     }
 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, _("two @prec's in a row"));
 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, _("%%dprec affects only GLR parsers"));
 354   if (dprec <= 0)
 355     complain_at (location, _("%%dprec must be followed by positive number"));
 356   else if (current_rule->dprec != 0)
 357     complain_at (location, _("only one %%dprec allowed per rule"));
 358   current_rule->dprec = dprec;
 359 }
 360
 361 /* Attach a merge function NAME with argument type TYPE to current
 362    rule. */
 363
 364 void
 365 grammar_current_rule_merge_set (const char* name, location_t location)
 366 {
 367   if (! glr_parser)
 368     warn_at (location, _("%%merge affects only GLR parsers"));
 369   if (current_rule->merger != 0)
 370     complain_at (location, _("only one %%merge allowed per rule"));
 371   current_rule->merger =
 372     get_merge_function (name, current_rule->sym->type_name, location);
 373 }
 374
 375 /* Attach a SYMBOL to the current rule.  If needed, move the previous
 376    action as a mid-rule action.  */
 377
 378 void
 379 grammar_current_rule_symbol_append (symbol_t *symbol, location_t location)
 380 {
 381   if (current_rule->action)
 382     grammar_midrule_action ();
 383   ++nritems;
 384   grammar_symbol_append (symbol, location);
 385 }
 386
 387 /* Attach an ACTION to the current rule.  If needed, move the previous
 388    action as a mid-rule action.  */
 389
 390 void
 391 grammar_current_rule_action_append (const char *action, location_t location)
 392 {
 393   if (current_rule->action)
 394     grammar_midrule_action ();
 395   current_rule->action = action;
 396   current_rule->action_location = location;
 397 }
 398
 399 \f
 400 /*---------------------------------------------------------------.
 401 | Convert the rules into the representation using RRHS, RLHS and |
 402 | RITEM.                                                         |
 403 `---------------------------------------------------------------*/
 404
 405 static void
 406 packgram (void)
 407 {
 408   unsigned int itemno = 0;
 409   rule_number_t ruleno = 0;
 410   symbol_list_t *p = grammar;
 411
 412   ritem = XCALLOC (item_number_t, nritems);
 413   rules = XCALLOC (rule_t, nrules);
 414
 415   while (p)
 416     {
 417       symbol_t *ruleprec = p->ruleprec;
 418       rules[ruleno].user_number = ruleno;
 419       rules[ruleno].number = ruleno;
 420       rules[ruleno].lhs = p->sym;
 421       rules[ruleno].rhs = ritem + itemno;
 422       rules[ruleno].location = p->location;
 423       rules[ruleno].useful = TRUE;
 424       rules[ruleno].action = p->action;
 425       rules[ruleno].action_location = p->action_location;
 426       rules[ruleno].dprec = p->dprec;
 427       rules[ruleno].merger = p->merger;
 428
 429       p = p->next;
 430       while (p && p->sym)
 431         {
 432           /* item_number_t = symbol_number_t.
 433              But the former needs to contain more: negative rule numbers. */
 434           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 435           /* A rule gets by default the precedence and associativity
 436              of the last token in it.  */
 437           if (p->sym->class == token_sym)
 438             rules[ruleno].prec = p->sym;
 439           if (p)
 440             p = p->next;
 441         }
 442
 443       /* If this rule has a %prec,
 444          the specified symbol's precedence replaces the default.  */
 445       if (ruleprec)
 446         {
 447           rules[ruleno].precsym = ruleprec;
 448           rules[ruleno].prec = ruleprec;
 449         }
 450       ritem[itemno++] = rule_number_as_item_number (ruleno);
 451       ++ruleno;
 452
 453       if (p)
 454         p = p->next;
 455     }
 456
 457   assert (itemno == nritems);
 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   gram_control_t gram_control;
 474
 475   /* Initialize the symbol table.  */
 476   symbols_new ();
 477
 478   /* Construct the accept symbol. */
 479   accept = symbol_get ("$accept", empty_location);
 480   accept->class = nterm_sym;
 481   accept->number = nvars++;
 482
 483   /* Construct the error token */
 484   errtoken = symbol_get ("error", empty_location);
 485   errtoken->class = token_sym;
 486   errtoken->number = ntokens++;
 487
 488   /* Construct a token that represents all undefined literal tokens.
 489      It is always token number 2.  */
 490   undeftoken = symbol_get ("$undefined", empty_location);
 491   undeftoken->class = token_sym;
 492   undeftoken->number = ntokens++;
 493
 494   /* Initialize the obstacks. */
 495   obstack_init (&pre_prologue_obstack);
 496   obstack_init (&post_prologue_obstack);
 497
 498   finput = xfopen (infile, "r");
 499   gram_in = finput;
 500
 501   gram_debug = !!getenv ("parse");
 502   gram__flex_debug = !!getenv ("scan");
 503   scanner_initialize ();
 504   gram_parse (&gram_control);
 505
 506   /* If something went wrong during the parsing, don't try to
 507      continue.  */
 508   if (complain_message_count)
 509     exit (1);
 510
 511   /* Grammar has been read.  Do some checking */
 512   if (nrules == 0)
 513     fatal (_("no rules in the input grammar"));
 514
 515   /* Report any undefined symbols and consider them nonterminals.  */
 516   symbols_check_defined ();
 517
 518   /* If the user did not define her ENDTOKEN, do it now. */
 519   if (!endtoken)
 520     {
 521       endtoken = symbol_get ("$end", empty_location);
 522       endtoken->class = token_sym;
 523       endtoken->number = 0;
 524       /* Value specified by POSIX.  */
 525       endtoken->user_token_number = 0;
 526     }
 527
 528   /* Insert the initial rule, which line is that of the first rule
 529      (not that of the start symbol):
 530
 531      accept: %start EOF.  */
 532   {
 533     symbol_list_t *p = symbol_list_new (accept, empty_location);
 534     p->location = grammar->location;
 535     p->next = symbol_list_new (startsymbol, empty_location);
 536     p->next->next = symbol_list_new (endtoken, empty_location);
 537     p->next->next->next = symbol_list_new (NULL, empty_location);
 538     p->next->next->next->next = grammar;
 539     nrules += 1;
 540     nritems += 3;
 541     grammar = p;
 542   }
 543
 544   if (nsyms > SHRT_MAX)
 545     fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
 546           SHRT_MAX);
 547
 548   assert (nsyms == ntokens + nvars);
 549
 550   xfclose (finput);
 551
 552   /* Assign the symbols their symbol numbers.  Write #defines for the
 553      token symbols into FDEFINES if requested.  */
 554   symbols_pack ();
 555
 556   /* Convert the grammar into the format described in gram.h.  */
 557   packgram ();
 558
 559   /* The grammar as a symbol_list_t is no longer needed. */
 560   LIST_FREE (symbol_list_t, grammar);
 561 }