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 "options.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 *grammar = NULL;
  39 static int start_flag = 0;
  40
  41 /* Nonzero if %union has been seen.  */
  42 int typed = 0;
  43
  44 static symbol_list *
  45 symbol_list_new (symbol_t *sym, location_t location)
  46 {
  47   symbol_list *res = XMALLOC (symbol_list, 1);
  48   res->next = NULL;
  49   res->sym = sym;
  50   res->location = location;
  51   res->action = NULL;
  52   res->ruleprec = NULL;
  53   return res;
  54 }
  55
  56 \f
  57 /*--------------------------------------------------------------.
  58 | Get the data type (alternative in the union) of the value for |
  59 | symbol N in rule RULE.                                        |
  60 `--------------------------------------------------------------*/
  61
  62 char *
  63 get_type_name (int n, symbol_list *rule)
  64 {
  65   int i;
  66   symbol_list *rp;
  67
  68   if (n < 0)
  69     {
  70       complain (_("invalid $ value"));
  71       return NULL;
  72     }
  73
  74   rp = rule;
  75   i = 0;
  76
  77   while (i < n)
  78     {
  79       rp = rp->next;
  80       if (rp == NULL || rp->sym == NULL)
  81         {
  82           complain (_("invalid $ value"));
  83           return NULL;
  84         }
  85       ++i;
  86     }
  87
  88   return rp->sym->type_name;
  89 }
  90
  91
  92 /*-----------------------.
  93 | Set the start symbol.  |
  94 `-----------------------*/
  95
  96 void
  97 grammar_start_symbol_set (symbol_t *s, location_t l)
  98 {
  99   if (start_flag)
 100     complain (_("multiple %s declarations"), "%start");
 101   else
 102     {
 103       start_flag = 1;
 104       startsymbol = s;
 105       startsymbol_location = l;
 106     }
 107 }
 108
 109
 110 /*----------------------------------------------------------------.
 111 | There are two prologues: one before %union, one after.  Augment |
 112 | the current one.                                                |
 113 `----------------------------------------------------------------*/
 114
 115 void
 116 prologue_augment (const char *prologue, location_t location)
 117 {
 118   struct obstack *oout =
 119     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
 120
 121   if (!no_lines_flag)
 122     {
 123       obstack_fgrow2 (oout, muscle_find ("linef"),
 124                       location.first_line,
 125                       quotearg_style (c_quoting_style,
 126                                       muscle_find ("filename")));
 127     }
 128   obstack_sgrow (oout, prologue);
 129 }
 130
 131
 132
 133
 134 /*----------------------.
 135 | Handle the epilogue.  |
 136 `----------------------*/
 137
 138 void
 139 epilogue_set (const char *epilogue, location_t location)
 140 {
 141   if (!no_lines_flag)
 142     {
 143       obstack_fgrow2 (&muscle_obstack, muscle_find ("linef"),
 144                       location.first_line,
 145                       quotearg_style (c_quoting_style,
 146                                       muscle_find ("filename")));
 147     }
 148   obstack_sgrow (&muscle_obstack, epilogue);
 149   obstack_1grow (&muscle_obstack, 0);
 150   muscle_insert ("epilogue", obstack_finish (&muscle_obstack));
 151 }
 152
 153
 154 \f
 155
 156 /*-------------------------------------------------------------------.
 157 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
 158 | with the user's names.                                             |
 159 `-------------------------------------------------------------------*/
 160
 161 static symbol_t *
 162 gensym (location_t location)
 163 {
 164   /* Incremented for each generated symbol */
 165   static int gensym_count = 0;
 166   static char buf[256];
 167
 168   symbol_t *sym;
 169
 170   sprintf (buf, "@%d", ++gensym_count);
 171   sym = getsym (buf, location);
 172   sym->class = nterm_sym;
 173   sym->number = nvars++;
 174   return sym;
 175 }
 176 \f
 177 /*-------------------------------------------------------------------.
 178 | Parse the input grammar into a one symbol_list structure.  Each    |
 179 | rule is represented by a sequence of symbols: the left hand side   |
 180 | followed by the contents of the right hand side, followed by a     |
 181 | null pointer instead of a symbol to terminate the rule.  The next  |
 182 | symbol is the lhs of the following rule.                           |
 183 |                                                                    |
 184 | All actions are copied out, labelled by the rule number they apply |
 185 | to.                                                                |
 186 |                                                                    |
 187 | Bison used to allow some %directives in the rules sections, but    |
 188 | this is no longer consider appropriate: (i) the documented grammar |
 189 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 190 | recovery for %directives consists in skipping the junk until a `%' |
 191 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 192 | in the rules section.                                              |
 193 `-------------------------------------------------------------------*/
 194
 195 /* The (currently) last symbol of GRAMMAR. */
 196 symbol_list *grammar_end = NULL;
 197
 198 /* Append S to the GRAMMAR. */
 199 void
 200 grammar_symbol_append (symbol_t *symbol, location_t location)
 201 {
 202   symbol_list *p = symbol_list_new (symbol, location);
 203
 204   if (grammar_end)
 205     grammar_end->next = p;
 206   else
 207     grammar = p;
 208
 209   grammar_end = p;
 210 }
 211
 212 /* The rule currently being defined, and the previous rule.
 213    CURRENT_RULE points to the first LHS of the current rule, while
 214    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 215 symbol_list *current_rule = NULL;
 216 symbol_list *previous_rule_end = NULL;
 217
 218
 219 /*----------------------------------------------.
 220 | Create a new rule for LHS in to the GRAMMAR.  |
 221 `----------------------------------------------*/
 222
 223 void
 224 grammar_rule_begin (symbol_t *lhs, location_t location)
 225 {
 226   if (!start_flag)
 227     {
 228       startsymbol = lhs;
 229       startsymbol_location = location;
 230       start_flag = 1;
 231     }
 232
 233   /* Start a new rule and record its lhs.  */
 234   ++nrules;
 235   ++nritems;
 236
 237   previous_rule_end = grammar_end;
 238   grammar_symbol_append (lhs, location);
 239   current_rule = grammar_end;
 240
 241   /* Mark the rule's lhs as a nonterminal if not already so.  */
 242
 243   if (lhs->class == unknown_sym)
 244     {
 245       lhs->class = nterm_sym;
 246       lhs->number = nvars;
 247       ++nvars;
 248     }
 249   else if (lhs->class == token_sym)
 250     complain (_("rule given for %s, which is a token"), lhs->tag);
 251 }
 252
 253 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 254    instance, there should be no type clash on the default action.  */
 255
 256 static void
 257 grammar_current_rule_check (void)
 258 {
 259   symbol_t *lhs = current_rule->sym;
 260   symbol_t *first_rhs = current_rule->next->sym;
 261
 262   /* If there is an action, then there is nothing we can do: the user
 263      is allowed to shoot in her foot.  */
 264   if (current_rule->action)
 265     return;
 266
 267   /* If $$ is being set in default way, report if any type mismatch.
 268      */
 269   if (first_rhs)
 270     {
 271       const char *lhs_type = lhs->type_name       ? lhs->type_name       : "";
 272       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 273       if (strcmp (lhs_type, rhs_type))
 274         complain (_("type clash (`%s' `%s') on default action"),
 275                   lhs_type, rhs_type);
 276     }
 277   /* Warn if there is no default for $$ but we need one.  */
 278   else
 279     {
 280       if (lhs->type_name)
 281         complain (_("empty rule for typed nonterminal, and no action"));
 282     }
 283 }
 284
 285
 286 /*-------------------------------------.
 287 | End the currently being grown rule.  |
 288 `-------------------------------------*/
 289
 290 void
 291 grammar_rule_end (location_t location)
 292 {
 293   /* Put an empty link in the list to mark the end of this rule  */
 294   grammar_symbol_append (NULL, grammar_end->location);
 295   current_rule->location = location;
 296   grammar_current_rule_check ();
 297 }
 298
 299
 300 /*-------------------------------------------------------------------.
 301 | The previous action turns out the be a mid-rule action.  Attach it |
 302 | to the current rule, i.e., create a dummy symbol, attach it this   |
 303 | mid-rule action, and append this dummy nonterminal to the current  |
 304 | rule.                                                              |
 305 `-------------------------------------------------------------------*/
 306
 307 void
 308 grammar_midrule_action (void)
 309 {
 310   /* Since the action was written out with this rule's number, we must
 311      give the new rule this number by inserting the new rule before
 312      it.  */
 313
 314   /* Make a DUMMY nonterminal, whose location is that of the midrule
 315      action.  Create the MIDRULE.  */
 316   location_t dummy_location = current_rule->action_location;
 317   symbol_t *dummy = gensym (dummy_location);
 318   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 319
 320   /* Make a new rule, whose body is empty, before the current one, so
 321      that the action just read can belong to it.  */
 322   ++nrules;
 323   ++nritems;
 324   /* Attach its location and actions to that of the DUMMY.  */
 325   midrule->location = dummy_location;
 326   midrule->action = current_rule->action;
 327   midrule->action_location = dummy_location;
 328   current_rule->action = NULL;
 329
 330   if (previous_rule_end)
 331     previous_rule_end->next = midrule;
 332   else
 333     grammar = midrule;
 334
 335   /* End the dummy's rule.  */
 336   previous_rule_end = symbol_list_new (NULL, dummy_location);
 337   previous_rule_end->next = current_rule;
 338
 339   midrule->next = previous_rule_end;
 340
 341   /* Insert the dummy nonterminal replacing the midrule action into
 342      the current rule.  */
 343   grammar_current_rule_symbol_append (dummy, dummy_location);
 344 }
 345
 346 /* Set the precedence symbol of the current rule to PRECSYM. */
 347
 348 void
 349 grammar_current_rule_prec_set (symbol_t *precsym)
 350 {
 351   if (current_rule->ruleprec)
 352     complain (_("two @prec's in a row"));
 353   current_rule->ruleprec = precsym;
 354 }
 355
 356 /* Attach a SYMBOL to the current rule.  If needed, move the previous
 357    action as a mid-rule action.  */
 358
 359 void
 360 grammar_current_rule_symbol_append (symbol_t *symbol, location_t location)
 361 {
 362   if (current_rule->action)
 363     grammar_midrule_action ();
 364   ++nritems;
 365   grammar_symbol_append (symbol, location);
 366 }
 367
 368
 369 /* Attach an ACTION to the current rule.  If needed, move the previous
 370    action as a mid-rule action.  */
 371
 372 void
 373 grammar_current_rule_action_append (const char *action, location_t location)
 374 {
 375   if (current_rule->action)
 376     grammar_midrule_action ();
 377   current_rule->action = action;
 378   current_rule->action_location = location;
 379 }
 380
 381 \f
 382 /*---------------------------------------------------------------.
 383 | Convert the rules into the representation using RRHS, RLHS and |
 384 | RITEM.                                                         |
 385 `---------------------------------------------------------------*/
 386
 387 static void
 388 packgram (void)
 389 {
 390   unsigned int itemno;
 391   int ruleno;
 392   symbol_list *p;
 393
 394   ritem = XCALLOC (item_number_t, nritems);
 395   rules = XCALLOC (rule_t, nrules) - 1;
 396
 397   itemno = 0;
 398   ruleno = 1;
 399
 400   p = grammar;
 401   while (p)
 402     {
 403       symbol_t *ruleprec = p->ruleprec;
 404       rules[ruleno].user_number = ruleno;
 405       rules[ruleno].number = ruleno;
 406       rules[ruleno].lhs = p->sym;
 407       rules[ruleno].rhs = ritem + itemno;
 408       rules[ruleno].location = p->location;
 409       rules[ruleno].useful = TRUE;
 410       rules[ruleno].action = p->action;
 411       rules[ruleno].action_location = p->action_location;
 412
 413       p = p->next;
 414       while (p && p->sym)
 415         {
 416           /* item_number_t = symbol_number_t.
 417              But the former needs to contain more: negative rule numbers. */
 418           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 419           /* A rule gets by default the precedence and associativity
 420              of the last token in it.  */
 421           if (p->sym->class == token_sym)
 422             rules[ruleno].prec = p->sym;
 423           if (p)
 424             p = p->next;
 425         }
 426
 427       /* If this rule has a %prec,
 428          the specified symbol's precedence replaces the default.  */
 429       if (ruleprec)
 430         {
 431           rules[ruleno].precsym = ruleprec;
 432           rules[ruleno].prec = ruleprec;
 433         }
 434       ritem[itemno++] = -ruleno;
 435       ++ruleno;
 436
 437       if (p)
 438         p = p->next;
 439     }
 440
 441   assert (itemno == nritems);
 442
 443   if (trace_flag)
 444     ritem_print (stderr);
 445 }
 446 \f
 447 /*------------------------------------------------------------------.
 448 | Read in the grammar specification and record it in the format     |
 449 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 450 | in each case forming the body of a C function (YYACTION) which    |
 451 | contains a switch statement to decide which action to execute.    |
 452 `------------------------------------------------------------------*/
 453
 454 void
 455 reader (void)
 456 {
 457   gram_control_t gram_control;
 458   lineno = 1;
 459
 460   /* Initialize the symbol table.  */
 461   symbols_new ();
 462
 463   /* Construct the axiom symbol. */
 464   axiom = getsym ("$axiom", empty_location);
 465   axiom->class = nterm_sym;
 466   axiom->number = nvars++;
 467
 468   /* Construct the error token */
 469   errtoken = getsym ("error", empty_location);
 470   errtoken->class = token_sym;
 471   errtoken->number = ntokens++;
 472
 473   /* Construct a token that represents all undefined literal tokens.
 474      It is always token number 2.  */
 475   undeftoken = getsym ("$undefined.", empty_location);
 476   undeftoken->class = token_sym;
 477   undeftoken->number = ntokens++;
 478
 479   /* Initialize the obstacks. */
 480   obstack_init (&pre_prologue_obstack);
 481   obstack_init (&post_prologue_obstack);
 482
 483   finput = xfopen (infile, "r");
 484   gram_in = finput;
 485
 486   gram_debug = !!getenv ("parse");
 487   gram__flex_debug = !!getenv ("scan");
 488   scanner_initialize ();
 489   gram_parse (&gram_control);
 490
 491   /* Grammar has been read.  Do some checking */
 492   if (nrules == 0)
 493     fatal (_("no rules in the input grammar"));
 494
 495   /* Report any undefined symbols and consider them nonterminals.  */
 496   symbols_check_defined ();
 497
 498   /* If the user did not define her EOFTOKEN, do it now. */
 499   if (!eoftoken)
 500     {
 501       eoftoken = getsym ("$", empty_location);
 502       eoftoken->class = token_sym;
 503       eoftoken->number = 0;
 504       /* Value specified by POSIX.  */
 505       eoftoken->user_token_number = 0;
 506     }
 507
 508   /* Insert the initial rule, which line is that of the first rule
 509      (not that of the start symbol):
 510
 511      axiom: %start EOF.  */
 512   {
 513     symbol_list *p = symbol_list_new (axiom, empty_location);
 514     p->location = grammar->location;
 515     p->next = symbol_list_new (startsymbol, empty_location);
 516     p->next->next = symbol_list_new (eoftoken, empty_location);
 517     p->next->next->next = symbol_list_new (NULL, empty_location);
 518     p->next->next->next->next = grammar;
 519     nrules += 1;
 520     nritems += 3;
 521     grammar = p;
 522   }
 523
 524   if (nsyms > SHRT_MAX)
 525     fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
 526           SHRT_MAX);
 527
 528   assert (nsyms == ntokens + nvars);
 529
 530   xfclose (finput);
 531
 532   /* Assign the symbols their symbol numbers.  Write #defines for the
 533      token symbols into FDEFINES if requested.  */
 534   symbols_pack ();
 535
 536   /* Convert the grammar into the format described in gram.h.  */
 537   packgram ();
 538
 539   /* The grammar as a symbol_list is no longer needed. */
 540   LIST_FREE (symbol_list, grammar);
 541 }