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 bool start_flag = false;
  40 merger_list *merge_functions;
  41
  42 /* Has %union been seen?  */
  43 bool typed = false;
  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 = true;
  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 \f
  82
  83 /*-------------------------------------------------------------------.
  84 | Return the merger index for a merging function named NAME, whose   |
  85 | arguments have type TYPE.  Records the function, if new, in        |
  86 | MERGER_LIST.                                                       |
  87 `-------------------------------------------------------------------*/
  88
  89 static int
  90 get_merge_function (uniqstr name, uniqstr type, location loc)
  91 {
  92   merger_list *syms;
  93   merger_list head;
  94   int n;
  95
  96   if (! glr_parser)
  97     return 0;
  98
  99   if (type == NULL)
 100     type = uniqstr_new ("");
 101
 102   head.next = merge_functions;
 103   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 104     if (UNIQSTR_EQ (name, syms->next->name))
 105       break;
 106   if (syms->next == NULL)
 107     {
 108       MALLOC (syms->next, 1);
 109       syms->next->name = uniqstr_new (name);
 110       syms->next->type = uniqstr_new (type);
 111       syms->next->next = NULL;
 112       merge_functions = head.next;
 113     }
 114   else if (!UNIQSTR_EQ (type, syms->next->type))
 115     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 116              name, type, syms->next->type);
 117   return n;
 118 }
 119
 120 /*--------------------------------------.
 121 | Free all merge-function definitions.  |
 122 `--------------------------------------*/
 123
 124 void
 125 free_merger_functions (void)
 126 {
 127   merger_list *L0;
 128   if (! glr_parser)
 129     return;
 130   L0 = merge_functions;
 131   while (L0 != NULL)
 132     {
 133       merger_list *L1 = L0->next;
 134       free (L0);
 135       L0 = L1;
 136     }
 137 }
 138
 139 \f
 140 /*-------------------------------------------------------------------.
 141 | Parse the input grammar into a one symbol_list structure.  Each    |
 142 | rule is represented by a sequence of symbols: the left hand side   |
 143 | followed by the contents of the right hand side, followed by a     |
 144 | null pointer instead of a symbol to terminate the rule.  The next  |
 145 | symbol is the lhs of the following rule.                           |
 146 |                                                                    |
 147 | All actions are copied out, labelled by the rule number they apply |
 148 | to.                                                                |
 149 |                                                                    |
 150 | Bison used to allow some %directives in the rules sections, but    |
 151 | this is no longer consider appropriate: (i) the documented grammar |
 152 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 153 | recovery for %directives consists in skipping the junk until a `%' |
 154 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 155 | in the rules section.                                              |
 156 `-------------------------------------------------------------------*/
 157
 158 /* The (currently) last symbol of GRAMMAR. */
 159 symbol_list *grammar_end = NULL;
 160
 161 /* Append SYM to the grammar.  */
 162 void
 163 grammar_symbol_append (symbol *sym, location loc)
 164 {
 165   symbol_list *p = symbol_list_new (sym, loc);
 166
 167   if (grammar_end)
 168     grammar_end->next = p;
 169   else
 170     grammar = p;
 171
 172   grammar_end = p;
 173 }
 174
 175 /* The rule currently being defined, and the previous rule.
 176    CURRENT_RULE points to the first LHS of the current rule, while
 177    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 178 symbol_list *current_rule = NULL;
 179 symbol_list *previous_rule_end = NULL;
 180
 181
 182 /*----------------------------------------------.
 183 | Create a new rule for LHS in to the GRAMMAR.  |
 184 `----------------------------------------------*/
 185
 186 void
 187 grammar_rule_begin (symbol *lhs, location loc)
 188 {
 189   if (!start_flag)
 190     {
 191       startsymbol = lhs;
 192       startsymbol_location = loc;
 193       start_flag = true;
 194     }
 195
 196   /* Start a new rule and record its lhs.  */
 197   ++nrules;
 198   ++nritems;
 199
 200   previous_rule_end = grammar_end;
 201   grammar_symbol_append (lhs, loc);
 202   current_rule = grammar_end;
 203
 204   /* Mark the rule's lhs as a nonterminal if not already so.  */
 205
 206   if (lhs->class == unknown_sym)
 207     {
 208       lhs->class = nterm_sym;
 209       lhs->number = nvars;
 210       ++nvars;
 211     }
 212   else if (lhs->class == token_sym)
 213     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 214 }
 215
 216 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 217    instance, there should be no type clash on the default action.  */
 218
 219 static void
 220 grammar_current_rule_check (void)
 221 {
 222   symbol *lhs = current_rule->sym;
 223   char const *lhs_type = lhs->type_name;
 224   symbol *first_rhs = current_rule->next->sym;
 225
 226   /* If there is an action, then there is nothing we can do: the user
 227      is allowed to shoot herself in the foot.  */
 228   if (current_rule->action)
 229     return;
 230
 231   /* Don't worry about the default action if $$ is untyped, since $$'s
 232      value can't be used.  */
 233   if (! lhs_type)
 234     return;
 235
 236   /* If $$ is being set in default way, report if any type mismatch.  */
 237   if (first_rhs)
 238     {
 239       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 240       if (!UNIQSTR_EQ (lhs_type, rhs_type))
 241         warn_at (current_rule->location,
 242                  _("type clash on default action: <%s> != <%s>"),
 243                  lhs_type, rhs_type);
 244     }
 245   /* Warn if there is no default for $$ but we need one.  */
 246   else
 247     warn_at (current_rule->location,
 248              _("empty rule for typed nonterminal, and no action"));
 249 }
 250
 251
 252 /*-------------------------------------.
 253 | End the currently being grown rule.  |
 254 `-------------------------------------*/
 255
 256 void
 257 grammar_rule_end (location loc)
 258 {
 259   /* Put an empty link in the list to mark the end of this rule  */
 260   grammar_symbol_append (NULL, grammar_end->location);
 261   current_rule->location = loc;
 262   grammar_current_rule_check ();
 263 }
 264
 265
 266 /*-------------------------------------------------------------------.
 267 | The previous action turns out the be a mid-rule action.  Attach it |
 268 | to the current rule, i.e., create a dummy symbol, attach it this   |
 269 | mid-rule action, and append this dummy nonterminal to the current  |
 270 | rule.                                                              |
 271 `-------------------------------------------------------------------*/
 272
 273 void
 274 grammar_midrule_action (void)
 275 {
 276   /* Since the action was written out with this rule's number, we must
 277      give the new rule this number by inserting the new rule before
 278      it.  */
 279
 280   /* Make a DUMMY nonterminal, whose location is that of the midrule
 281      action.  Create the MIDRULE.  */
 282   location dummy_location = current_rule->action_location;
 283   symbol *dummy = dummy_symbol_get (dummy_location);
 284   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 285
 286   /* Make a new rule, whose body is empty, before the current one, so
 287      that the action just read can belong to it.  */
 288   ++nrules;
 289   ++nritems;
 290   /* Attach its location and actions to that of the DUMMY.  */
 291   midrule->location = dummy_location;
 292   midrule->action = current_rule->action;
 293   midrule->action_location = dummy_location;
 294   current_rule->action = NULL;
 295
 296   if (previous_rule_end)
 297     previous_rule_end->next = midrule;
 298   else
 299     grammar = midrule;
 300
 301   /* End the dummy's rule.  */
 302   previous_rule_end = symbol_list_new (NULL, dummy_location);
 303   previous_rule_end->next = current_rule;
 304
 305   midrule->next = previous_rule_end;
 306
 307   /* Insert the dummy nonterminal replacing the midrule action into
 308      the current rule.  */
 309   grammar_current_rule_symbol_append (dummy, dummy_location);
 310 }
 311
 312 /* Set the precedence symbol of the current rule to PRECSYM. */
 313
 314 void
 315 grammar_current_rule_prec_set (symbol *precsym, location loc)
 316 {
 317   if (current_rule->ruleprec)
 318     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 319   current_rule->ruleprec = precsym;
 320 }
 321
 322 /* Attach dynamic precedence DPREC to the current rule. */
 323
 324 void
 325 grammar_current_rule_dprec_set (int dprec, location loc)
 326 {
 327   if (! glr_parser)
 328     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 329   if (dprec <= 0)
 330     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 331   else if (current_rule->dprec != 0)
 332     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 333   current_rule->dprec = dprec;
 334 }
 335
 336 /* Attach a merge function NAME with argument type TYPE to current
 337    rule. */
 338
 339 void
 340 grammar_current_rule_merge_set (uniqstr name, location loc)
 341 {
 342   if (! glr_parser)
 343     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 344   if (current_rule->merger != 0)
 345     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 346   current_rule->merger =
 347     get_merge_function (name, current_rule->sym->type_name, loc);
 348 }
 349
 350 /* Attach SYM to the current rule.  If needed, move the previous
 351    action as a mid-rule action.  */
 352
 353 void
 354 grammar_current_rule_symbol_append (symbol *sym, location loc)
 355 {
 356   if (current_rule->action)
 357     grammar_midrule_action ();
 358   ++nritems;
 359   grammar_symbol_append (sym, loc);
 360 }
 361
 362 /* Attach an ACTION to the current rule.  If needed, move the previous
 363    action as a mid-rule action.  */
 364
 365 void
 366 grammar_current_rule_action_append (const char *action, location loc)
 367 {
 368   if (current_rule->action)
 369     grammar_midrule_action ();
 370   current_rule->action = action;
 371   current_rule->action_location = loc;
 372 }
 373
 374 \f
 375 /*---------------------------------------------------------------.
 376 | Convert the rules into the representation using RRHS, RLHS and |
 377 | RITEM.                                                         |
 378 `---------------------------------------------------------------*/
 379
 380 static void
 381 packgram (void)
 382 {
 383   unsigned int itemno = 0;
 384   rule_number ruleno = 0;
 385   symbol_list *p = grammar;
 386
 387   CALLOC (ritem, nritems);
 388   CALLOC (rules, nrules);
 389
 390   while (p)
 391     {
 392       symbol *ruleprec = p->ruleprec;
 393       rules[ruleno].user_number = ruleno;
 394       rules[ruleno].number = ruleno;
 395       rules[ruleno].lhs = p->sym;
 396       rules[ruleno].rhs = ritem + itemno;
 397       rules[ruleno].location = p->location;
 398       rules[ruleno].useful = true;
 399       rules[ruleno].action = p->action;
 400       rules[ruleno].action_location = p->action_location;
 401       rules[ruleno].dprec = p->dprec;
 402       rules[ruleno].merger = p->merger;
 403
 404       p = p->next;
 405       while (p && p->sym)
 406         {
 407           /* item_number = symbol_number.
 408              But the former needs to contain more: negative rule numbers. */
 409           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 410           /* A rule gets by default the precedence and associativity
 411              of the last token in it.  */
 412           if (p->sym->class == token_sym)
 413             rules[ruleno].prec = p->sym;
 414           if (p)
 415             p = p->next;
 416         }
 417
 418       /* If this rule has a %prec,
 419          the specified symbol's precedence replaces the default.  */
 420       if (ruleprec)
 421         {
 422           rules[ruleno].precsym = ruleprec;
 423           rules[ruleno].prec = ruleprec;
 424         }
 425       ritem[itemno++] = rule_number_as_item_number (ruleno);
 426       ++ruleno;
 427
 428       if (p)
 429         p = p->next;
 430     }
 431
 432   if (itemno != nritems)
 433     abort ();
 434
 435   if (trace_flag & trace_sets)
 436     ritem_print (stderr);
 437 }
 438 \f
 439 /*------------------------------------------------------------------.
 440 | Read in the grammar specification and record it in the format     |
 441 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 442 | in each case forming the body of a C function (YYACTION) which    |
 443 | contains a switch statement to decide which action to execute.    |
 444 `------------------------------------------------------------------*/
 445
 446 void
 447 reader (void)
 448 {
 449   /* Initialize the symbol table.  */
 450   symbols_new ();
 451
 452   /* Construct the accept symbol. */
 453   accept = symbol_get ("$accept", empty_location);
 454   accept->class = nterm_sym;
 455   accept->number = nvars++;
 456
 457   /* Construct the error token */
 458   errtoken = symbol_get ("error", empty_location);
 459   errtoken->class = token_sym;
 460   errtoken->number = ntokens++;
 461
 462   /* Construct a token that represents all undefined literal tokens.
 463      It is always token number 2.  */
 464   undeftoken = symbol_get ("$undefined", empty_location);
 465   undeftoken->class = token_sym;
 466   undeftoken->number = ntokens++;
 467
 468   /* Initialize the obstacks. */
 469   obstack_init (&pre_prologue_obstack);
 470   obstack_init (&post_prologue_obstack);
 471
 472   finput = xfopen (grammar_file, "r");
 473   gram_in = finput;
 474
 475   gram__flex_debug = trace_flag & trace_scan;
 476   gram_debug = trace_flag & trace_parse;
 477   scanner_initialize ();
 478   gram_parse ();
 479
 480   /* If something went wrong during the parsing, don't try to
 481      continue.  */
 482   if (complaint_issued)
 483     return;
 484
 485   /* Grammar has been read.  Do some checking */
 486   if (nrules == 0)
 487     fatal (_("no rules in the input grammar"));
 488
 489   /* Report any undefined symbols and consider them nonterminals.  */
 490   symbols_check_defined ();
 491
 492   /* If the user did not define her ENDTOKEN, do it now. */
 493   if (!endtoken)
 494     {
 495       endtoken = symbol_get ("$end", empty_location);
 496       endtoken->class = token_sym;
 497       endtoken->number = 0;
 498       /* Value specified by POSIX.  */
 499       endtoken->user_token_number = 0;
 500     }
 501
 502   /* Insert the initial rule, which line is that of the first rule
 503      (not that of the start symbol):
 504
 505      accept: %start EOF.  */
 506   {
 507     symbol_list *p = symbol_list_new (accept, empty_location);
 508     p->location = grammar->location;
 509     p->next = symbol_list_new (startsymbol, empty_location);
 510     p->next->next = symbol_list_new (endtoken, empty_location);
 511     p->next->next->next = symbol_list_new (NULL, empty_location);
 512     p->next->next->next->next = grammar;
 513     nrules += 1;
 514     nritems += 3;
 515     grammar = p;
 516   }
 517
 518   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 519     abort ();
 520
 521   xfclose (finput);
 522
 523   /* Assign the symbols their symbol numbers.  Write #defines for the
 524      token symbols into FDEFINES if requested.  */
 525   symbols_pack ();
 526
 527   /* Convert the grammar into the format described in gram.h.  */
 528   packgram ();
 529
 530   /* The grammar as a symbol_list is no longer needed. */
 531   LIST_FREE (symbol_list, grammar);
 532 }