src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003
   4    Free Software Foundation, Inc.
   5
   6    This file is part of Bison, the GNU Compiler Compiler.
   7
   8    Bison is free software; you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation; either version 2, or (at your option)
  11    any later version.
  12
  13    Bison is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License
  19    along with Bison; see the file COPYING.  If not, write to
  20    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  21    Boston, MA 02111-1307, USA.  */
  22
  23 #include "system.h"
  24
  25 #include <quotearg.h>
  26
  27 #include "complain.h"
  28 #include "conflicts.h"
  29 #include "files.h"
  30 #include "getargs.h"
  31 #include "gram.h"
  32 #include "muscle_tab.h"
  33 #include "output.h"
  34 #include "reader.h"
  35 #include "symlist.h"
  36 #include "symtab.h"
  37
  38 static symbol_list *grammar = NULL;
  39 static int start_flag = 0;
  40 merger_list *merge_functions;
  41
  42 /* Nonzero if %union has been seen.  */
  43 int typed = 0;
  44 \f
  45 /*-----------------------.
  46 | Set the start symbol.  |
  47 `-----------------------*/
  48
  49 void
  50 grammar_start_symbol_set (symbol *sym, location loc)
  51 {
  52   if (start_flag)
  53     complain_at (loc, _("multiple %s declarations"), "%start");
  54   else
  55     {
  56       start_flag = 1;
  57       startsymbol = sym;
  58       startsymbol_location = loc;
  59     }
  60 }
  61
  62
  63 /*----------------------------------------------------------------.
  64 | There are two prologues: one before %union, one after.  Augment |
  65 | the current one.                                                |
  66 `----------------------------------------------------------------*/
  67
  68 void
  69 prologue_augment (const char *prologue, location loc)
  70 {
  71   struct obstack *oout =
  72     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  73
  74   obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[", loc.start.line);
  75   MUSCLE_OBSTACK_SGROW (oout,
  76                         quotearg_style (c_quoting_style, loc.start.file));
  77   obstack_sgrow (oout, "]])[\n");
  78   obstack_sgrow (oout, prologue);
  79 }
  80
  81
  82
  83
  84 /*----------------------.
  85 | Handle the epilogue.  |
  86 `----------------------*/
  87
  88 void
  89 epilogue_augment (const char *epilogue, location loc)
  90 {
  91   char *extension = NULL;
  92   obstack_fgrow1 (&muscle_obstack, "]b4_syncline([[%d]], [[", loc.start.line);
  93   MUSCLE_OBSTACK_SGROW (&muscle_obstack,
  94                         quotearg_style (c_quoting_style, loc.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 (uniqstr name, uniqstr type, location 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 = uniqstr_new ("");
 124
 125   head.next = merge_functions;
 126   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 127     if (UNIQSTR_EQ (name, syms->next->name))
 128       break;
 129   if (syms->next == NULL)
 130     {
 131       MALLOC (syms->next, 1);
 132       syms->next->name = uniqstr_new (name);
 133       syms->next->type = uniqstr_new (type);
 134       syms->next->next = NULL;
 135       merge_functions = head.next;
 136     }
 137   else if (!UNIQSTR_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 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 *grammar_end = NULL;
 183
 184 /* Append SYM to the grammar.  */
 185 void
 186 grammar_symbol_append (symbol *sym, location loc)
 187 {
 188   symbol_list *p = symbol_list_new (sym, loc);
 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 *current_rule = NULL;
 202 symbol_list *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 *lhs, location loc)
 211 {
 212   if (!start_flag)
 213     {
 214       startsymbol = lhs;
 215       startsymbol_location = loc;
 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, loc);
 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 (loc, _("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 *lhs = current_rule->sym;
 246   char const *lhs_type = lhs->type_name;
 247   symbol *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 $$ is being set in default way, report if any type mismatch.  */
 260   if (first_rhs)
 261     {
 262       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 263       if (!UNIQSTR_EQ (lhs_type, rhs_type))
 264         warn_at (current_rule->location,
 265                  _("type clash on default action: <%s> != <%s>"),
 266                  lhs_type, rhs_type);
 267     }
 268   /* Warn if there is no default for $$ 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 loc)
 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 = loc;
 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 dummy_location = current_rule->action_location;
 306   symbol *dummy = dummy_symbol_get (dummy_location);
 307   symbol_list *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 *precsym, location loc)
 339 {
 340   if (current_rule->ruleprec)
 341     complain_at (loc, _("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 loc)
 349 {
 350   if (! glr_parser)
 351     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 352   if (dprec <= 0)
 353     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 354   else if (current_rule->dprec != 0)
 355     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 356   current_rule->dprec = dprec;
 357 }
 358
 359 /* Attach a merge function NAME with argument type TYPE to current
 360    rule. */
 361
 362 void
 363 grammar_current_rule_merge_set (uniqstr name, location loc)
 364 {
 365   if (! glr_parser)
 366     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 367   if (current_rule->merger != 0)
 368     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 369   current_rule->merger =
 370     get_merge_function (name, current_rule->sym->type_name, loc);
 371 }
 372
 373 /* Attach SYM to the current rule.  If needed, move the previous
 374    action as a mid-rule action.  */
 375
 376 void
 377 grammar_current_rule_symbol_append (symbol *sym, location loc)
 378 {
 379   if (current_rule->action)
 380     grammar_midrule_action ();
 381   ++nritems;
 382   grammar_symbol_append (sym, loc);
 383 }
 384
 385 /* Attach an ACTION to the current rule.  If needed, move the previous
 386    action as a mid-rule action.  */
 387
 388 void
 389 grammar_current_rule_action_append (const char *action, location loc)
 390 {
 391   if (current_rule->action)
 392     grammar_midrule_action ();
 393   current_rule->action = action;
 394   current_rule->action_location = loc;
 395 }
 396
 397 \f
 398 /*---------------------------------------------------------------.
 399 | Convert the rules into the representation using RRHS, RLHS and |
 400 | RITEM.                                                         |
 401 `---------------------------------------------------------------*/
 402
 403 static void
 404 packgram (void)
 405 {
 406   unsigned int itemno = 0;
 407   rule_number ruleno = 0;
 408   symbol_list *p = grammar;
 409
 410   CALLOC (ritem, nritems);
 411   CALLOC (rules, nrules);
 412
 413   while (p)
 414     {
 415       symbol *ruleprec = p->ruleprec;
 416       rules[ruleno].user_number = ruleno;
 417       rules[ruleno].number = ruleno;
 418       rules[ruleno].lhs = p->sym;
 419       rules[ruleno].rhs = ritem + itemno;
 420       rules[ruleno].location = p->location;
 421       rules[ruleno].useful = true;
 422       rules[ruleno].action = p->action;
 423       rules[ruleno].action_location = p->action_location;
 424       rules[ruleno].dprec = p->dprec;
 425       rules[ruleno].merger = p->merger;
 426
 427       p = p->next;
 428       while (p && p->sym)
 429         {
 430           /* item_number = symbol_number.
 431              But the former needs to contain more: negative rule numbers. */
 432           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 433           /* A rule gets by default the precedence and associativity
 434              of the last token in it.  */
 435           if (p->sym->class == token_sym)
 436             rules[ruleno].prec = p->sym;
 437           if (p)
 438             p = p->next;
 439         }
 440
 441       /* If this rule has a %prec,
 442          the specified symbol's precedence replaces the default.  */
 443       if (ruleprec)
 444         {
 445           rules[ruleno].precsym = ruleprec;
 446           rules[ruleno].prec = ruleprec;
 447         }
 448       ritem[itemno++] = rule_number_as_item_number (ruleno);
 449       ++ruleno;
 450
 451       if (p)
 452         p = p->next;
 453     }
 454
 455   if (itemno != nritems)
 456     abort ();
 457
 458   if (trace_flag & trace_sets)
 459     ritem_print (stderr);
 460 }
 461 \f
 462 /*------------------------------------------------------------------.
 463 | Read in the grammar specification and record it in the format     |
 464 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 465 | in each case forming the body of a C function (YYACTION) which    |
 466 | contains a switch statement to decide which action to execute.    |
 467 `------------------------------------------------------------------*/
 468
 469 void
 470 reader (void)
 471 {
 472   /* Initialize the symbol table.  */
 473   symbols_new ();
 474
 475   /* Construct the accept symbol. */
 476   accept = symbol_get ("$accept", empty_location);
 477   accept->class = nterm_sym;
 478   accept->number = nvars++;
 479
 480   /* Construct the error token */
 481   errtoken = symbol_get ("error", empty_location);
 482   errtoken->class = token_sym;
 483   errtoken->number = ntokens++;
 484
 485   /* Construct a token that represents all undefined literal tokens.
 486      It is always token number 2.  */
 487   undeftoken = symbol_get ("$undefined", empty_location);
 488   undeftoken->class = token_sym;
 489   undeftoken->number = ntokens++;
 490
 491   /* Initialize the obstacks. */
 492   obstack_init (&pre_prologue_obstack);
 493   obstack_init (&post_prologue_obstack);
 494
 495   finput = xfopen (grammar_file, "r");
 496   gram_in = finput;
 497
 498   gram__flex_debug = trace_flag & trace_scan;
 499   gram_debug = trace_flag & trace_parse;
 500   scanner_initialize ();
 501   gram_parse ();
 502
 503   /* If something went wrong during the parsing, don't try to
 504      continue.  */
 505   if (complaint_issued)
 506     return;
 507
 508   /* Grammar has been read.  Do some checking */
 509   if (nrules == 0)
 510     fatal (_("no rules in the input grammar"));
 511
 512   /* Report any undefined symbols and consider them nonterminals.  */
 513   symbols_check_defined ();
 514
 515   /* If the user did not define her ENDTOKEN, do it now. */
 516   if (!endtoken)
 517     {
 518       endtoken = symbol_get ("$end", empty_location);
 519       endtoken->class = token_sym;
 520       endtoken->number = 0;
 521       /* Value specified by POSIX.  */
 522       endtoken->user_token_number = 0;
 523     }
 524
 525   /* Insert the initial rule, which line is that of the first rule
 526      (not that of the start symbol):
 527
 528      accept: %start EOF.  */
 529   {
 530     symbol_list *p = symbol_list_new (accept, empty_location);
 531     p->location = grammar->location;
 532     p->next = symbol_list_new (startsymbol, empty_location);
 533     p->next->next = symbol_list_new (endtoken, empty_location);
 534     p->next->next->next = symbol_list_new (NULL, empty_location);
 535     p->next->next->next->next = grammar;
 536     nrules += 1;
 537     nritems += 3;
 538     grammar = p;
 539   }
 540
 541   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 542     abort ();
 543
 544   xfclose (finput);
 545
 546   /* Assign the symbols their symbol numbers.  Write #defines for the
 547      token symbols into FDEFINES if requested.  */
 548   symbols_pack ();
 549
 550   /* Convert the grammar into the format described in gram.h.  */
 551   packgram ();
 552
 553   /* The grammar as a symbol_list is no longer needed. */
 554   LIST_FREE (symbol_list, grammar);
 555 }