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
  45 /* Should rules have a default precedence?  */
  46 bool default_prec = true;
  47 \f
  48 /*-----------------------.
  49 | Set the start symbol.  |
  50 `-----------------------*/
  51
  52 void
  53 grammar_start_symbol_set (symbol *sym, location loc)
  54 {
  55   if (start_flag)
  56     complain_at (loc, _("multiple %s declarations"), "%start");
  57   else
  58     {
  59       start_flag = true;
  60       startsymbol = sym;
  61       startsymbol_location = loc;
  62     }
  63 }
  64
  65
  66 /*----------------------------------------------------------------.
  67 | There are two prologues: one before %union, one after.  Augment |
  68 | the current one.                                                |
  69 `----------------------------------------------------------------*/
  70
  71 void
  72 prologue_augment (const char *prologue, location loc)
  73 {
  74   struct obstack *oout =
  75     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  76
  77   obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[", loc.start.line);
  78   MUSCLE_OBSTACK_SGROW (oout,
  79                         quotearg_style (c_quoting_style, loc.start.file));
  80   obstack_sgrow (oout, "]])[\n");
  81   obstack_sgrow (oout, prologue);
  82 }
  83
  84 \f
  85
  86 /*-------------------------------------------------------------------.
  87 | Return the merger index for a merging function named NAME, whose   |
  88 | arguments have type TYPE.  Records the function, if new, in        |
  89 | MERGER_LIST.                                                       |
  90 `-------------------------------------------------------------------*/
  91
  92 static int
  93 get_merge_function (uniqstr name, uniqstr type, location loc)
  94 {
  95   merger_list *syms;
  96   merger_list head;
  97   int n;
  98
  99   if (! glr_parser)
 100     return 0;
 101
 102   if (type == NULL)
 103     type = uniqstr_new ("");
 104
 105   head.next = merge_functions;
 106   for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
 107     if (UNIQSTR_EQ (name, syms->next->name))
 108       break;
 109   if (syms->next == NULL)
 110     {
 111       MALLOC (syms->next, 1);
 112       syms->next->name = uniqstr_new (name);
 113       syms->next->type = uniqstr_new (type);
 114       syms->next->next = NULL;
 115       merge_functions = head.next;
 116     }
 117   else if (!UNIQSTR_EQ (type, syms->next->type))
 118     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 119              name, type, syms->next->type);
 120   return n;
 121 }
 122
 123 /*--------------------------------------.
 124 | Free all merge-function definitions.  |
 125 `--------------------------------------*/
 126
 127 void
 128 free_merger_functions (void)
 129 {
 130   merger_list *L0;
 131   if (! glr_parser)
 132     return;
 133   L0 = merge_functions;
 134   while (L0 != NULL)
 135     {
 136       merger_list *L1 = L0->next;
 137       free (L0);
 138       L0 = L1;
 139     }
 140 }
 141
 142 \f
 143 /*-------------------------------------------------------------------.
 144 | Parse the input grammar into a one symbol_list structure.  Each    |
 145 | rule is represented by a sequence of symbols: the left hand side   |
 146 | followed by the contents of the right hand side, followed by a     |
 147 | null pointer instead of a symbol to terminate the rule.  The next  |
 148 | symbol is the lhs of the following rule.                           |
 149 |                                                                    |
 150 | All actions are copied out, labelled by the rule number they apply |
 151 | to.                                                                |
 152 |                                                                    |
 153 | Bison used to allow some %directives in the rules sections, but    |
 154 | this is no longer consider appropriate: (i) the documented grammar |
 155 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
 156 | recovery for %directives consists in skipping the junk until a `%' |
 157 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
 158 | in the rules section.                                              |
 159 `-------------------------------------------------------------------*/
 160
 161 /* The (currently) last symbol of GRAMMAR. */
 162 symbol_list *grammar_end = NULL;
 163
 164 /* Append SYM to the grammar.  */
 165 void
 166 grammar_symbol_append (symbol *sym, location loc)
 167 {
 168   symbol_list *p = symbol_list_new (sym, loc);
 169
 170   if (grammar_end)
 171     grammar_end->next = p;
 172   else
 173     grammar = p;
 174
 175   grammar_end = p;
 176 }
 177
 178 /* The rule currently being defined, and the previous rule.
 179    CURRENT_RULE points to the first LHS of the current rule, while
 180    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 181 symbol_list *current_rule = NULL;
 182 symbol_list *previous_rule_end = NULL;
 183
 184
 185 /*----------------------------------------------.
 186 | Create a new rule for LHS in to the GRAMMAR.  |
 187 `----------------------------------------------*/
 188
 189 void
 190 grammar_rule_begin (symbol *lhs, location loc)
 191 {
 192   if (!start_flag)
 193     {
 194       startsymbol = lhs;
 195       startsymbol_location = loc;
 196       start_flag = true;
 197     }
 198
 199   /* Start a new rule and record its lhs.  */
 200   ++nrules;
 201   ++nritems;
 202
 203   previous_rule_end = grammar_end;
 204   grammar_symbol_append (lhs, loc);
 205   current_rule = grammar_end;
 206
 207   /* Mark the rule's lhs as a nonterminal if not already so.  */
 208
 209   if (lhs->class == unknown_sym)
 210     {
 211       lhs->class = nterm_sym;
 212       lhs->number = nvars;
 213       ++nvars;
 214     }
 215   else if (lhs->class == token_sym)
 216     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 217 }
 218
 219 /* Check that the last rule (CURRENT_RULE) is properly defined.  For
 220    instance, there should be no type clash on the default action.  */
 221
 222 static void
 223 grammar_current_rule_check (void)
 224 {
 225   symbol *lhs = current_rule->sym;
 226   char const *lhs_type = lhs->type_name;
 227   symbol *first_rhs = current_rule->next->sym;
 228
 229   /* If there is an action, then there is nothing we can do: the user
 230      is allowed to shoot herself in the foot.  */
 231   if (current_rule->action)
 232     return;
 233
 234   /* Don't worry about the default action if $$ is untyped, since $$'s
 235      value can't be used.  */
 236   if (! lhs_type)
 237     return;
 238
 239   /* If $$ is being set in default way, report if any type mismatch.  */
 240   if (first_rhs)
 241     {
 242       const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
 243       if (!UNIQSTR_EQ (lhs_type, rhs_type))
 244         warn_at (current_rule->location,
 245                  _("type clash on default action: <%s> != <%s>"),
 246                  lhs_type, rhs_type);
 247     }
 248   /* Warn if there is no default for $$ but we need one.  */
 249   else
 250     warn_at (current_rule->location,
 251              _("empty rule for typed nonterminal, and no action"));
 252 }
 253
 254
 255 /*-------------------------------------.
 256 | End the currently being grown rule.  |
 257 `-------------------------------------*/
 258
 259 void
 260 grammar_rule_end (location loc)
 261 {
 262   /* Put an empty link in the list to mark the end of this rule  */
 263   grammar_symbol_append (NULL, grammar_end->location);
 264   current_rule->location = loc;
 265   grammar_current_rule_check ();
 266 }
 267
 268
 269 /*-------------------------------------------------------------------.
 270 | The previous action turns out the be a mid-rule action.  Attach it |
 271 | to the current rule, i.e., create a dummy symbol, attach it this   |
 272 | mid-rule action, and append this dummy nonterminal to the current  |
 273 | rule.                                                              |
 274 `-------------------------------------------------------------------*/
 275
 276 void
 277 grammar_midrule_action (void)
 278 {
 279   /* Since the action was written out with this rule's number, we must
 280      give the new rule this number by inserting the new rule before
 281      it.  */
 282
 283   /* Make a DUMMY nonterminal, whose location is that of the midrule
 284      action.  Create the MIDRULE.  */
 285   location dummy_location = current_rule->action_location;
 286   symbol *dummy = dummy_symbol_get (dummy_location);
 287   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 288
 289   /* Make a new rule, whose body is empty, before the current one, so
 290      that the action just read can belong to it.  */
 291   ++nrules;
 292   ++nritems;
 293   /* Attach its location and actions to that of the DUMMY.  */
 294   midrule->location = dummy_location;
 295   midrule->action = current_rule->action;
 296   midrule->action_location = dummy_location;
 297   current_rule->action = NULL;
 298
 299   if (previous_rule_end)
 300     previous_rule_end->next = midrule;
 301   else
 302     grammar = midrule;
 303
 304   /* End the dummy's rule.  */
 305   previous_rule_end = symbol_list_new (NULL, dummy_location);
 306   previous_rule_end->next = current_rule;
 307
 308   midrule->next = previous_rule_end;
 309
 310   /* Insert the dummy nonterminal replacing the midrule action into
 311      the current rule.  */
 312   grammar_current_rule_symbol_append (dummy, dummy_location);
 313 }
 314
 315 /* Set the precedence symbol of the current rule to PRECSYM. */
 316
 317 void
 318 grammar_current_rule_prec_set (symbol *precsym, location loc)
 319 {
 320   if (current_rule->ruleprec)
 321     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 322   current_rule->ruleprec = precsym;
 323 }
 324
 325 /* Attach dynamic precedence DPREC to the current rule. */
 326
 327 void
 328 grammar_current_rule_dprec_set (int dprec, location loc)
 329 {
 330   if (! glr_parser)
 331     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 332   if (dprec <= 0)
 333     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 334   else if (current_rule->dprec != 0)
 335     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 336   current_rule->dprec = dprec;
 337 }
 338
 339 /* Attach a merge function NAME with argument type TYPE to current
 340    rule. */
 341
 342 void
 343 grammar_current_rule_merge_set (uniqstr name, location loc)
 344 {
 345   if (! glr_parser)
 346     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 347   if (current_rule->merger != 0)
 348     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 349   current_rule->merger =
 350     get_merge_function (name, current_rule->sym->type_name, loc);
 351 }
 352
 353 /* Attach SYM to the current rule.  If needed, move the previous
 354    action as a mid-rule action.  */
 355
 356 void
 357 grammar_current_rule_symbol_append (symbol *sym, location loc)
 358 {
 359   if (current_rule->action)
 360     grammar_midrule_action ();
 361   ++nritems;
 362   grammar_symbol_append (sym, loc);
 363 }
 364
 365 /* Attach an ACTION to the current rule.  If needed, move the previous
 366    action as a mid-rule action.  */
 367
 368 void
 369 grammar_current_rule_action_append (const char *action, location loc)
 370 {
 371   if (current_rule->action)
 372     grammar_midrule_action ();
 373   current_rule->action = action;
 374   current_rule->action_location = loc;
 375 }
 376
 377 \f
 378 /*---------------------------------------------------------------.
 379 | Convert the rules into the representation using RRHS, RLHS and |
 380 | RITEM.                                                         |
 381 `---------------------------------------------------------------*/
 382
 383 static void
 384 packgram (void)
 385 {
 386   unsigned int itemno = 0;
 387   rule_number ruleno = 0;
 388   symbol_list *p = grammar;
 389
 390   CALLOC (ritem, nritems);
 391   CALLOC (rules, nrules);
 392
 393   while (p)
 394     {
 395       symbol *ruleprec = p->ruleprec;
 396       rules[ruleno].user_number = ruleno;
 397       rules[ruleno].number = ruleno;
 398       rules[ruleno].lhs = p->sym;
 399       rules[ruleno].rhs = ritem + itemno;
 400       rules[ruleno].location = p->location;
 401       rules[ruleno].useful = true;
 402       rules[ruleno].action = p->action;
 403       rules[ruleno].action_location = p->action_location;
 404       rules[ruleno].dprec = p->dprec;
 405       rules[ruleno].merger = p->merger;
 406
 407       p = p->next;
 408       while (p && p->sym)
 409         {
 410           /* item_number = symbol_number.
 411              But the former needs to contain more: negative rule numbers. */
 412           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 413           /* A rule gets by default the precedence and associativity
 414              of the last token in it.  */
 415           if (p->sym->class == token_sym && default_prec)
 416             rules[ruleno].prec = p->sym;
 417           if (p)
 418             p = p->next;
 419         }
 420
 421       /* If this rule has a %prec,
 422          the specified symbol's precedence replaces the default.  */
 423       if (ruleprec)
 424         {
 425           rules[ruleno].precsym = ruleprec;
 426           rules[ruleno].prec = ruleprec;
 427         }
 428       ritem[itemno++] = rule_number_as_item_number (ruleno);
 429       ++ruleno;
 430
 431       if (p)
 432         p = p->next;
 433     }
 434
 435   if (itemno != nritems)
 436     abort ();
 437
 438   if (trace_flag & trace_sets)
 439     ritem_print (stderr);
 440 }
 441 \f
 442 /*------------------------------------------------------------------.
 443 | Read in the grammar specification and record it in the format     |
 444 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 445 | in each case forming the body of a C function (YYACTION) which    |
 446 | contains a switch statement to decide which action to execute.    |
 447 `------------------------------------------------------------------*/
 448
 449 void
 450 reader (void)
 451 {
 452   /* Initialize the symbol table.  */
 453   symbols_new ();
 454
 455   /* Construct the accept symbol. */
 456   accept = symbol_get ("$accept", empty_location);
 457   accept->class = nterm_sym;
 458   accept->number = nvars++;
 459
 460   /* Construct the error token */
 461   errtoken = symbol_get ("error", empty_location);
 462   errtoken->class = token_sym;
 463   errtoken->number = ntokens++;
 464
 465   /* Construct a token that represents all undefined literal tokens.
 466      It is always token number 2.  */
 467   undeftoken = symbol_get ("$undefined", empty_location);
 468   undeftoken->class = token_sym;
 469   undeftoken->number = ntokens++;
 470
 471   /* Initialize the obstacks. */
 472   obstack_init (&pre_prologue_obstack);
 473   obstack_init (&post_prologue_obstack);
 474
 475   finput = xfopen (grammar_file, "r");
 476   gram_in = finput;
 477
 478   gram__flex_debug = trace_flag & trace_scan;
 479   gram_debug = trace_flag & trace_parse;
 480   scanner_initialize ();
 481   gram_parse ();
 482
 483   /* If something went wrong during the parsing, don't try to
 484      continue.  */
 485   if (complaint_issued)
 486     return;
 487
 488   /* Grammar has been read.  Do some checking */
 489   if (nrules == 0)
 490     fatal (_("no rules in the input grammar"));
 491
 492   /* Report any undefined symbols and consider them nonterminals.  */
 493   symbols_check_defined ();
 494
 495   /* If the user did not define her ENDTOKEN, do it now. */
 496   if (!endtoken)
 497     {
 498       endtoken = symbol_get ("$end", empty_location);
 499       endtoken->class = token_sym;
 500       endtoken->number = 0;
 501       /* Value specified by POSIX.  */
 502       endtoken->user_token_number = 0;
 503     }
 504
 505   /* Insert the initial rule, which line is that of the first rule
 506      (not that of the start symbol):
 507
 508      accept: %start EOF.  */
 509   {
 510     symbol_list *p = symbol_list_new (accept, empty_location);
 511     p->location = grammar->location;
 512     p->next = symbol_list_new (startsymbol, empty_location);
 513     p->next->next = symbol_list_new (endtoken, empty_location);
 514     p->next->next->next = symbol_list_new (NULL, empty_location);
 515     p->next->next->next->next = grammar;
 516     nrules += 1;
 517     nritems += 3;
 518     grammar = p;
 519   }
 520
 521   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 522     abort ();
 523
 524   xfclose (finput);
 525
 526   /* Assign the symbols their symbol numbers.  Write #defines for the
 527      token symbols into FDEFINES if requested.  */
 528   symbols_pack ();
 529
 530   /* Convert the grammar into the format described in gram.h.  */
 531   packgram ();
 532
 533   /* The grammar as a symbol_list is no longer needed. */
 534   LIST_FREE (symbol_list, grammar);
 535 }