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