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 "options.h"
  31 #include "gram.h"
  32 #include "complain.h"
  33 #include "output.h"
  34 #include "reader.h"
  35 #include "conflicts.h"
  36 #include "muscle_tab.h"
  37
  38 int lineno;
  39 static symbol_list_t *grammar = NULL;
  40 static int start_flag = 0;
  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_t *s, location_t l)
  51 {
  52   if (start_flag)
  53     complain_at (l, _("multiple %s declarations"), "%start");
  54   else
  55     {
  56       start_flag = 1;
  57       startsymbol = s;
  58       startsymbol_location = l;
  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_t location)
  70 {
  71   struct obstack *oout =
  72     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  73
  74   if (!no_lines_flag)
  75     {
  76       obstack_fgrow2 (oout, muscle_find ("linef"),
  77                       location.first_line,
  78                       quotearg_style (c_quoting_style,
  79                                       muscle_find ("filename")));
  80     }
  81   obstack_sgrow (oout, prologue);
  82 }
  83
  84
  85
  86
  87 /*----------------------.
  88 | Handle the epilogue.  |
  89 `----------------------*/
  90
  91 void
  92 epilogue_set (const char *epilogue, location_t location)
  93 {
  94   if (!no_lines_flag)
  95     {
  96       obstack_fgrow2 (&muscle_obstack, muscle_find ("linef"),
  97                       location.first_line,
  98                       quotearg_style (c_quoting_style,
  99                                       muscle_find ("filename")));
 100     }
 101   obstack_sgrow (&muscle_obstack, epilogue);
 102   obstack_1grow (&muscle_obstack, 0);
 103   muscle_insert ("epilogue", obstack_finish (&muscle_obstack));
 104 }
 105
 106
 107 \f
 108
 109 /*-------------------------------------------------------------------.
 110 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
 111 | with the user's names.                                             |
 112 `-------------------------------------------------------------------*/
 113
 114 static symbol_t *
 115 gensym (location_t location)
 116 {
 117   /* Incremented for each generated symbol */
 118   static int gensym_count = 0;
 119   static char buf[256];
 120
 121   symbol_t *sym;
 122
 123   sprintf (buf, "@%d", ++gensym_count);
 124   sym = getsym (buf, location);
 125   sym->class = nterm_sym;
 126   sym->number = nvars++;
 127   return sym;
 128 }
 129 \f
 130 /*-------------------------------------------------------------------.
 131 | Parse the input grammar into a one symbol_list_t structure.  Each    |
 132 | rule is represented by a sequence of symbols: the left hand side   |
 133 | followed by the contents of the right hand side, followed by a     |
 134 | null pointer instead of a symbol to terminate the rule.  The next  |
 135 | symbol is the lhs of the following rule.                           |
 136 |                                                                    |
 137 | All actions are copied out, labelled by the rule number they apply |
 138 | to.                                                                |
 139 |                                                                    |
 140 | Bison used to allow some %directives in the rules sections, but    |
 141 | this is no longer consider appropriate: (i) the documented grammar |
 142 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 143 | recovery for %directives consists in skipping the junk until a `%' |
 144 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 145 | in the rules section.                                              |
 146 `-------------------------------------------------------------------*/
 147
 148 /* The (currently) last symbol of GRAMMAR. */
 149 symbol_list_t *grammar_end = NULL;
 150
 151 /* Append S to the GRAMMAR. */
 152 void
 153 grammar_symbol_append (symbol_t *symbol, location_t location)
 154 {
 155   symbol_list_t *p = symbol_list_new (symbol, location);
 156
 157   if (grammar_end)
 158     grammar_end->next = p;
 159   else
 160     grammar = p;
 161
 162   grammar_end = p;
 163 }
 164
 165 /* The rule currently being defined, and the previous rule.
 166    CURRENT_RULE points to the first LHS of the current rule, while
 167    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 168 symbol_list_t *current_rule = NULL;
 169 symbol_list_t *previous_rule_end = NULL;
 170
 171
 172 /*----------------------------------------------.
 173 | Create a new rule for LHS in to the GRAMMAR.  |
 174 `----------------------------------------------*/
 175
 176 void
 177 grammar_rule_begin (symbol_t *lhs, location_t location)
 178 {
 179   if (!start_flag)
 180     {
 181       startsymbol = lhs;
 182       startsymbol_location = location;
 183       start_flag = 1;
 184     }
 185
 186   /* Start a new rule and record its lhs.  */
 187   ++nrules;
 188   ++nritems;
 189
 190   previous_rule_end = grammar_end;
 191   grammar_symbol_append (lhs, location);
 192   current_rule = grammar_end;
 193
 194   /* Mark the rule's lhs as a nonterminal if not already so.  */
 195
 196   if (lhs->class == unknown_sym)
 197     {
 198       lhs->class = nterm_sym;
 199       lhs->number = nvars;
 200       ++nvars;
 201     }
 202   else if (lhs->class == token_sym)
 203     complain_at (location, _("rule given for %s, which is a token"), lhs->tag);
 204 }
 205
 206 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 207    instance, there should be no type clash on the default action.  */
 208
 209 static void
 210 grammar_current_rule_check (void)
 211 {
 212   symbol_t *lhs = current_rule->sym;
 213   symbol_t *first_rhs = current_rule->next->sym;
 214
 215   /* If there is an action, then there is nothing we can do: the user
 216      is allowed to shoot in her foot.  */
 217   if (current_rule->action)
 218     return;
 219
 220   /* If $$ is being set in default way, report if any type mismatch.
 221      */
 222   if (first_rhs)
 223     {
 224       const char *lhs_type = lhs->type_name       ? lhs->type_name       : "";
 225       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 226       if (strcmp (lhs_type, rhs_type))
 227         complain_at (current_rule->location,
 228                      _("type clash (`%s' `%s') on default action"),
 229                      lhs_type, rhs_type);
 230     }
 231   /* Warn if there is no default for $$ but we need one.  */
 232   else
 233     {
 234       if (lhs->type_name)
 235         complain_at (current_rule->location,
 236                      _("empty rule for typed nonterminal, and no action"));
 237     }
 238 }
 239
 240
 241 /*-------------------------------------.
 242 | End the currently being grown rule.  |
 243 `-------------------------------------*/
 244
 245 void
 246 grammar_rule_end (location_t location)
 247 {
 248   /* Put an empty link in the list to mark the end of this rule  */
 249   grammar_symbol_append (NULL, grammar_end->location);
 250   current_rule->location = location;
 251   grammar_current_rule_check ();
 252 }
 253
 254
 255 /*-------------------------------------------------------------------.
 256 | The previous action turns out the be a mid-rule action.  Attach it |
 257 | to the current rule, i.e., create a dummy symbol, attach it this   |
 258 | mid-rule action, and append this dummy nonterminal to the current  |
 259 | rule.                                                              |
 260 `-------------------------------------------------------------------*/
 261
 262 void
 263 grammar_midrule_action (void)
 264 {
 265   /* Since the action was written out with this rule's number, we must
 266      give the new rule this number by inserting the new rule before
 267      it.  */
 268
 269   /* Make a DUMMY nonterminal, whose location is that of the midrule
 270      action.  Create the MIDRULE.  */
 271   location_t dummy_location = current_rule->action_location;
 272   symbol_t *dummy = gensym (dummy_location);
 273   symbol_list_t *midrule = symbol_list_new (dummy, dummy_location);
 274
 275   /* Make a new rule, whose body is empty, before the current one, so
 276      that the action just read can belong to it.  */
 277   ++nrules;
 278   ++nritems;
 279   /* Attach its location and actions to that of the DUMMY.  */
 280   midrule->location = dummy_location;
 281   midrule->action = current_rule->action;
 282   midrule->action_location = dummy_location;
 283   current_rule->action = NULL;
 284
 285   if (previous_rule_end)
 286     previous_rule_end->next = midrule;
 287   else
 288     grammar = midrule;
 289
 290   /* End the dummy's rule.  */
 291   previous_rule_end = symbol_list_new (NULL, dummy_location);
 292   previous_rule_end->next = current_rule;
 293
 294   midrule->next = previous_rule_end;
 295
 296   /* Insert the dummy nonterminal replacing the midrule action into
 297      the current rule.  */
 298   grammar_current_rule_symbol_append (dummy, dummy_location);
 299 }
 300
 301 /* Set the precedence symbol of the current rule to PRECSYM. */
 302
 303 void
 304 grammar_current_rule_prec_set (symbol_t *precsym, location_t location)
 305 {
 306   if (current_rule->ruleprec)
 307     complain_at (location, _("two @prec's in a row"));
 308   current_rule->ruleprec = precsym;
 309 }
 310
 311 /* Attach a SYMBOL to the current rule.  If needed, move the previous
 312    action as a mid-rule action.  */
 313
 314 void
 315 grammar_current_rule_symbol_append (symbol_t *symbol, location_t location)
 316 {
 317   if (current_rule->action)
 318     grammar_midrule_action ();
 319   ++nritems;
 320   grammar_symbol_append (symbol, location);
 321 }
 322
 323
 324 /* Attach an ACTION to the current rule.  If needed, move the previous
 325    action as a mid-rule action.  */
 326
 327 void
 328 grammar_current_rule_action_append (const char *action, location_t location)
 329 {
 330   if (current_rule->action)
 331     grammar_midrule_action ();
 332   current_rule->action = action;
 333   current_rule->action_location = location;
 334 }
 335
 336 \f
 337 /*---------------------------------------------------------------.
 338 | Convert the rules into the representation using RRHS, RLHS and |
 339 | RITEM.                                                         |
 340 `---------------------------------------------------------------*/
 341
 342 static void
 343 packgram (void)
 344 {
 345   unsigned int itemno;
 346   int ruleno;
 347   symbol_list_t *p;
 348
 349   ritem = XCALLOC (item_number_t, nritems);
 350   rules = XCALLOC (rule_t, nrules) - 1;
 351
 352   itemno = 0;
 353   ruleno = 1;
 354
 355   p = grammar;
 356   while (p)
 357     {
 358       symbol_t *ruleprec = p->ruleprec;
 359       rules[ruleno].user_number = ruleno;
 360       rules[ruleno].number = ruleno;
 361       rules[ruleno].lhs = p->sym;
 362       rules[ruleno].rhs = ritem + itemno;
 363       rules[ruleno].location = p->location;
 364       rules[ruleno].useful = TRUE;
 365       rules[ruleno].action = p->action;
 366       rules[ruleno].action_location = p->action_location;
 367
 368       p = p->next;
 369       while (p && p->sym)
 370         {
 371           /* item_number_t = symbol_number_t.
 372              But the former needs to contain more: negative rule numbers. */
 373           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 374           /* A rule gets by default the precedence and associativity
 375              of the last token in it.  */
 376           if (p->sym->class == token_sym)
 377             rules[ruleno].prec = p->sym;
 378           if (p)
 379             p = p->next;
 380         }
 381
 382       /* If this rule has a %prec,
 383          the specified symbol's precedence replaces the default.  */
 384       if (ruleprec)
 385         {
 386           rules[ruleno].precsym = ruleprec;
 387           rules[ruleno].prec = ruleprec;
 388         }
 389       ritem[itemno++] = -ruleno;
 390       ++ruleno;
 391
 392       if (p)
 393         p = p->next;
 394     }
 395
 396   assert (itemno == nritems);
 397
 398   if (trace_flag)
 399     ritem_print (stderr);
 400 }
 401 \f
 402 /*------------------------------------------------------------------.
 403 | Read in the grammar specification and record it in the format     |
 404 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 405 | in each case forming the body of a C function (YYACTION) which    |
 406 | contains a switch statement to decide which action to execute.    |
 407 `------------------------------------------------------------------*/
 408
 409 void
 410 reader (void)
 411 {
 412   gram_control_t gram_control;
 413   lineno = 1;
 414
 415   /* Initialize the symbol table.  */
 416   symbols_new ();
 417
 418   /* Construct the axiom symbol. */
 419   axiom = getsym ("$axiom", empty_location);
 420   axiom->class = nterm_sym;
 421   axiom->number = nvars++;
 422
 423   /* Construct the error token */
 424   errtoken = getsym ("error", empty_location);
 425   errtoken->class = token_sym;
 426   errtoken->number = ntokens++;
 427
 428   /* Construct a token that represents all undefined literal tokens.
 429      It is always token number 2.  */
 430   undeftoken = getsym ("$undefined.", empty_location);
 431   undeftoken->class = token_sym;
 432   undeftoken->number = ntokens++;
 433
 434   /* Initialize the obstacks. */
 435   obstack_init (&pre_prologue_obstack);
 436   obstack_init (&post_prologue_obstack);
 437
 438   finput = xfopen (infile, "r");
 439   gram_in = finput;
 440
 441   gram_debug = !!getenv ("parse");
 442   gram__flex_debug = !!getenv ("scan");
 443   scanner_initialize ();
 444   gram_parse (&gram_control);
 445
 446   /* Grammar has been read.  Do some checking */
 447   if (nrules == 0)
 448     fatal (_("no rules in the input grammar"));
 449
 450   /* Report any undefined symbols and consider them nonterminals.  */
 451   symbols_check_defined ();
 452
 453   /* If the user did not define her EOFTOKEN, do it now. */
 454   if (!eoftoken)
 455     {
 456       eoftoken = getsym ("$", empty_location);
 457       eoftoken->class = token_sym;
 458       eoftoken->number = 0;
 459       /* Value specified by POSIX.  */
 460       eoftoken->user_token_number = 0;
 461     }
 462
 463   /* Insert the initial rule, which line is that of the first rule
 464      (not that of the start symbol):
 465
 466      axiom: %start EOF.  */
 467   {
 468     symbol_list_t *p = symbol_list_new (axiom, empty_location);
 469     p->location = grammar->location;
 470     p->next = symbol_list_new (startsymbol, empty_location);
 471     p->next->next = symbol_list_new (eoftoken, empty_location);
 472     p->next->next->next = symbol_list_new (NULL, empty_location);
 473     p->next->next->next->next = grammar;
 474     nrules += 1;
 475     nritems += 3;
 476     grammar = p;
 477   }
 478
 479   if (nsyms > SHRT_MAX)
 480     fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
 481           SHRT_MAX);
 482
 483   assert (nsyms == ntokens + nvars);
 484
 485   xfclose (finput);
 486
 487   /* Assign the symbols their symbol numbers.  Write #defines for the
 488      token symbols into FDEFINES if requested.  */
 489   symbols_pack ();
 490
 491   /* Convert the grammar into the format described in gram.h.  */
 492   packgram ();
 493
 494   /* The grammar as a symbol_list_t is no longer needed. */
 495   LIST_FREE (symbol_list_t, grammar);
 496 }