src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003,
   4    2005 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., 51 Franklin Street, Fifth Floor,
  21    Boston, MA 02110-1301, USA.  */
  22
  23 #include <config.h>
  24 #include "system.h"
  25
  26 #include <quotearg.h>
  27
  28 #include "complain.h"
  29 #include "conflicts.h"
  30 #include "files.h"
  31 #include "getargs.h"
  32 #include "gram.h"
  33 #include "muscle_tab.h"
  34 #include "output.h"
  35 #include "reader.h"
  36 #include "symlist.h"
  37 #include "symtab.h"
  38
  39 static void grammar_midrule_action (void);
  40 static void grammar_symbol_append (symbol *sym, location loc);
  41
  42 static symbol_list *grammar = NULL;
  43 static bool start_flag = false;
  44 merger_list *merge_functions;
  45
  46 /* Was %union seen?  */
  47 bool typed = false;
  48
  49 /* Should rules have a default precedence?  */
  50 bool default_prec = true;
  51 \f
  52 /*-----------------------.
  53 | Set the start symbol.  |
  54 `-----------------------*/
  55
  56 void
  57 grammar_start_symbol_set (symbol *sym, location loc)
  58 {
  59   if (start_flag)
  60     complain_at (loc, _("multiple %s declarations"), "%start");
  61   else
  62     {
  63       start_flag = true;
  64       startsymbol = sym;
  65       startsymbol_location = loc;
  66     }
  67 }
  68
  69
  70 /*----------------------------------------------------------------.
  71 | There are two prologues: one before %union, one after.  Augment |
  72 | the current one.                                                |
  73 `----------------------------------------------------------------*/
  74
  75 void
  76 prologue_augment (const char *prologue, location loc)
  77 {
  78   struct obstack *oout =
  79     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  80
  81   obstack_fgrow1 (oout, "]b4_syncline(%d, [[", loc.start.line);
  82   MUSCLE_OBSTACK_SGROW (oout,
  83                         quotearg_style (c_quoting_style, loc.start.file));
  84   obstack_sgrow (oout, "]])[\n");
  85   obstack_sgrow (oout, prologue);
  86 }
  87
  88 \f
  89
  90 /*-------------------------------------------------------------------.
  91 | Return the merger index for a merging function named NAME, whose   |
  92 | arguments have type TYPE.  Records the function, if new, in        |
  93 | MERGER_LIST.                                                       |
  94 `-------------------------------------------------------------------*/
  95
  96 static int
  97 get_merge_function (uniqstr name, uniqstr type, location loc)
  98 {
  99   merger_list *syms;
 100   merger_list head;
 101   int n;
 102
 103   if (! glr_parser)
 104     return 0;
 105
 106   if (type == NULL)
 107     type = uniqstr_new ("");
 108
 109   head.next = merge_functions;
 110   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
 111     if (UNIQSTR_EQ (name, syms->next->name))
 112       break;
 113   if (syms->next == NULL)
 114     {
 115       syms->next = xmalloc (sizeof syms->next[0]);
 116       syms->next->name = uniqstr_new (name);
 117       syms->next->type = uniqstr_new (type);
 118       syms->next->next = NULL;
 119       merge_functions = head.next;
 120     }
 121   else if (!UNIQSTR_EQ (type, syms->next->type))
 122     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 123              name, type, syms->next->type);
 124   return n;
 125 }
 126
 127 /*--------------------------------------.
 128 | Free all merge-function definitions.  |
 129 `--------------------------------------*/
 130
 131 void
 132 free_merger_functions (void)
 133 {
 134   merger_list *L0 = merge_functions;
 135   while (L0)
 136     {
 137       merger_list *L1 = L0->next;
 138       free (L0);
 139       L0 = L1;
 140     }
 141 }
 142
 143 \f
 144 /*-------------------------------------------------------------------.
 145 | Parse the input grammar into a one symbol_list structure.  Each    |
 146 | rule is represented by a sequence of symbols: the left hand side   |
 147 | followed by the contents of the right hand side, followed by a     |
 148 | null pointer instead of a symbol to terminate the rule.  The next  |
 149 | symbol is the lhs of the following rule.                           |
 150 |                                                                    |
 151 | All actions are copied out, labelled by the rule number they apply |
 152 | to.                                                                |
 153 `-------------------------------------------------------------------*/
 154
 155 /* The (currently) last symbol of GRAMMAR. */
 156 static symbol_list *grammar_end = NULL;
 157
 158 /* Append SYM to the grammar.  */
 159 static void
 160 grammar_symbol_append (symbol *sym, location loc)
 161 {
 162   symbol_list *p = symbol_list_new (sym, loc);
 163
 164   if (grammar_end)
 165     grammar_end->next = p;
 166   else
 167     grammar = p;
 168
 169   grammar_end = p;
 170
 171   /* SYM = 0 stands for an end of rule, it is not an actual
 172      part of it.  */
 173   if (sym)
 174     ++nritems;
 175 }
 176
 177 /* The rule currently being defined, and the previous rule.
 178    CURRENT_RULE points to the first LHS of the current rule, while
 179    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 180 symbol_list *current_rule = NULL;
 181 static symbol_list *previous_rule_end = NULL;
 182
 183
 184 /*----------------------------------------------.
 185 | Create a new rule for LHS in to the GRAMMAR.  |
 186 `----------------------------------------------*/
 187
 188 void
 189 grammar_current_rule_begin (symbol *lhs, location loc)
 190 {
 191   if (!start_flag)
 192     {
 193       startsymbol = lhs;
 194       startsymbol_location = loc;
 195       start_flag = true;
 196     }
 197
 198   /* Start a new rule and record its lhs.  */
 199   ++nrules;
 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   if (lhs->class == unknown_sym)
 206     {
 207       lhs->class = nterm_sym;
 208       lhs->number = nvars;
 209       ++nvars;
 210     }
 211   else if (lhs->class == token_sym)
 212     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 213 }
 214
 215
 216 /*----------------------------------------------------------------.
 217 | Check that the rule R is properly defined.  For instance, there |
 218 | should be no type clash on the default action.                  |
 219 `----------------------------------------------------------------*/
 220
 221 static void
 222 grammar_rule_check (const symbol_list *r)
 223 {
 224   /* Type check.
 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
 229      Don't worry about the default action if $$ is untyped, since $$'s
 230      value can't be used.  */
 231   if (!r->action && r->sym->type_name)
 232     {
 233       symbol *first_rhs = r->next->sym;
 234       /* If $$ is being set in default way, report if any type mismatch.  */
 235       if (first_rhs)
 236         {
 237           char const *lhs_type = r->sym->type_name;
 238           const char *rhs_type =
 239             first_rhs->type_name ? first_rhs->type_name : "";
 240           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 241             warn_at (r->location,
 242                      _("type clash on default action: <%s> != <%s>"),
 243                      lhs_type, rhs_type);
 244         }
 245       /* Warn if there is no default for $$ but we need one.  */
 246       else
 247         warn_at (r->location,
 248                  _("empty rule for typed nonterminal, and no action"));
 249     }
 250
 251   /* Check that typed symbol values are used.  */
 252   {
 253     symbol_list *l = r;
 254     int n = 0;
 255     for (; l && l->sym; l = l->next, ++n)
 256       if (! (l->used
 257              || !l->sym->type_name
 258              /* The default action, $$ = $1, `uses' both.  */
 259              || !r->action && (n == 0 || n == 1)))
 260         if (n)
 261           warn_at (r->location, _("unused value: $%d"), n);
 262         else
 263           warn_at (r->location, _("unset value: $$"));
 264   }
 265 }
 266
 267
 268 /*-------------------------------------.
 269 | End the currently being grown rule.  |
 270 `-------------------------------------*/
 271
 272 void
 273 grammar_current_rule_end (location loc)
 274 {
 275   /* Put an empty link in the list to mark the end of this rule  */
 276   grammar_symbol_append (NULL, grammar_end->location);
 277   current_rule->location = loc;
 278   grammar_rule_check (current_rule);
 279 }
 280
 281
 282 /*-------------------------------------------------------------------.
 283 | The previous action turns out the be a mid-rule action.  Attach it |
 284 | to the current rule, i.e., create a dummy symbol, attach it this   |
 285 | mid-rule action, and append this dummy nonterminal to the current  |
 286 | rule.                                                              |
 287 `-------------------------------------------------------------------*/
 288
 289 static void
 290 grammar_midrule_action (void)
 291 {
 292   /* Since the action was written out with this rule's number, we must
 293      give the new rule this number by inserting the new rule before
 294      it.  */
 295
 296   /* Make a DUMMY nonterminal, whose location is that of the midrule
 297      action.  Create the MIDRULE.  */
 298   location dummy_location = current_rule->action_location;
 299   symbol *dummy = dummy_symbol_get (dummy_location);
 300   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 301
 302   /* Make a new rule, whose body is empty, before the current one, so
 303      that the action just read can belong to it.  */
 304   ++nrules;
 305   ++nritems;
 306   /* Attach its location and actions to that of the DUMMY.  */
 307   midrule->location = dummy_location;
 308   midrule->action = current_rule->action;
 309   midrule->action_location = dummy_location;
 310   current_rule->action = NULL;
 311
 312   if (previous_rule_end)
 313     previous_rule_end->next = midrule;
 314   else
 315     grammar = midrule;
 316
 317   /* End the dummy's rule.  */
 318   previous_rule_end = symbol_list_new (NULL, dummy_location);
 319   previous_rule_end->next = current_rule;
 320
 321   midrule->next = previous_rule_end;
 322
 323   /* Insert the dummy nonterminal replacing the midrule action into
 324      the current rule.  */
 325   grammar_current_rule_symbol_append (dummy, dummy_location);
 326 }
 327
 328 /* Set the precedence symbol of the current rule to PRECSYM. */
 329
 330 void
 331 grammar_current_rule_prec_set (symbol *precsym, location loc)
 332 {
 333   if (current_rule->ruleprec)
 334     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 335   current_rule->ruleprec = precsym;
 336 }
 337
 338 /* Attach dynamic precedence DPREC to the current rule. */
 339
 340 void
 341 grammar_current_rule_dprec_set (int dprec, location loc)
 342 {
 343   if (! glr_parser)
 344     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 345   if (dprec <= 0)
 346     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 347   else if (current_rule->dprec != 0)
 348     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 349   current_rule->dprec = dprec;
 350 }
 351
 352 /* Attach a merge function NAME with argument type TYPE to current
 353    rule. */
 354
 355 void
 356 grammar_current_rule_merge_set (uniqstr name, location loc)
 357 {
 358   if (! glr_parser)
 359     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 360   if (current_rule->merger != 0)
 361     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 362   current_rule->merger =
 363     get_merge_function (name, current_rule->sym->type_name, loc);
 364 }
 365
 366 /* Attach SYM to the current rule.  If needed, move the previous
 367    action as a mid-rule action.  */
 368
 369 void
 370 grammar_current_rule_symbol_append (symbol *sym, location loc)
 371 {
 372   if (current_rule->action)
 373     grammar_midrule_action ();
 374   grammar_symbol_append (sym, loc);
 375 }
 376
 377 /* Attach an ACTION to the current rule.  If needed, move the previous
 378    action as a mid-rule action.  */
 379
 380 void
 381 grammar_current_rule_action_append (const char *action, location loc)
 382 {
 383   if (current_rule->action)
 384     grammar_midrule_action ();
 385   current_rule->action = action;
 386   current_rule->action_location = loc;
 387 }
 388
 389 \f
 390 /*---------------------------------------------------------------.
 391 | Convert the rules into the representation using RRHS, RLHS and |
 392 | RITEM.                                                         |
 393 `---------------------------------------------------------------*/
 394
 395 static void
 396 packgram (void)
 397 {
 398   unsigned int itemno = 0;
 399   rule_number ruleno = 0;
 400   symbol_list *p = grammar;
 401
 402   ritem = xnmalloc (nritems, sizeof *ritem);
 403   rules = xnmalloc (nrules, sizeof *rules);
 404
 405   while (p)
 406     {
 407       symbol *ruleprec = p->ruleprec;
 408       rules[ruleno].user_number = ruleno;
 409       rules[ruleno].number = ruleno;
 410       rules[ruleno].lhs = p->sym;
 411       rules[ruleno].rhs = ritem + itemno;
 412       rules[ruleno].prec = NULL;
 413       rules[ruleno].dprec = p->dprec;
 414       rules[ruleno].merger = p->merger;
 415       rules[ruleno].precsym = NULL;
 416       rules[ruleno].location = p->location;
 417       rules[ruleno].useful = true;
 418       rules[ruleno].action = p->action;
 419       rules[ruleno].action_location = p->action_location;
 420
 421       p = p->next;
 422       while (p && p->sym)
 423         {
 424           /* item_number = symbol_number.
 425              But the former needs to contain more: negative rule numbers. */
 426           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 427           /* A rule gets by default the precedence and associativity
 428              of the last token in it.  */
 429           if (p->sym->class == token_sym && default_prec)
 430             rules[ruleno].prec = p->sym;
 431           if (p)
 432             p = p->next;
 433         }
 434
 435       /* If this rule has a %prec,
 436          the specified symbol's precedence replaces the default.  */
 437       if (ruleprec)
 438         {
 439           rules[ruleno].precsym = ruleprec;
 440           rules[ruleno].prec = ruleprec;
 441         }
 442       ritem[itemno++] = rule_number_as_item_number (ruleno);
 443       ++ruleno;
 444
 445       if (p)
 446         p = p->next;
 447     }
 448
 449   if (itemno != nritems)
 450     abort ();
 451
 452   if (trace_flag & trace_sets)
 453     ritem_print (stderr);
 454 }
 455 \f
 456 /*------------------------------------------------------------------.
 457 | Read in the grammar specification and record it in the format     |
 458 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 459 | in each case forming the body of a C function (YYACTION) which    |
 460 | contains a switch statement to decide which action to execute.    |
 461 `------------------------------------------------------------------*/
 462
 463 void
 464 reader (void)
 465 {
 466   /* Initialize the symbol table.  */
 467   symbols_new ();
 468
 469   /* Construct the accept symbol. */
 470   accept = symbol_get ("$accept", empty_location);
 471   accept->class = nterm_sym;
 472   accept->number = nvars++;
 473
 474   /* Construct the error token */
 475   errtoken = symbol_get ("error", empty_location);
 476   errtoken->class = token_sym;
 477   errtoken->number = ntokens++;
 478
 479   /* Construct a token that represents all undefined literal tokens.
 480      It is always token number 2.  */
 481   undeftoken = symbol_get ("$undefined", empty_location);
 482   undeftoken->class = token_sym;
 483   undeftoken->number = ntokens++;
 484
 485   /* Initialize the obstacks. */
 486   obstack_init (&pre_prologue_obstack);
 487   obstack_init (&post_prologue_obstack);
 488
 489   gram_in = xfopen (grammar_file, "r");
 490
 491   gram__flex_debug = trace_flag & trace_scan;
 492   gram_debug = trace_flag & trace_parse;
 493   scanner_initialize ();
 494   gram_parse ();
 495
 496   /* If something went wrong during the parsing, don't try to
 497      continue.  */
 498   if (complaint_issued)
 499     return;
 500
 501   /* Grammar has been read.  Do some checking */
 502   if (nrules == 0)
 503     fatal (_("no rules in the input grammar"));
 504
 505   /* Report any undefined symbols and consider them nonterminals.  */
 506   symbols_check_defined ();
 507
 508   /* If the user did not define her ENDTOKEN, do it now. */
 509   if (!endtoken)
 510     {
 511       endtoken = symbol_get ("$end", empty_location);
 512       endtoken->class = token_sym;
 513       endtoken->number = 0;
 514       /* Value specified by POSIX.  */
 515       endtoken->user_token_number = 0;
 516     }
 517
 518   /* Insert the initial rule, which line is that of the first rule
 519      (not that of the start symbol):
 520
 521      accept: %start EOF.  */
 522   {
 523     symbol_list *p = symbol_list_new (accept, empty_location);
 524     p->location = grammar->location;
 525     p->next = symbol_list_new (startsymbol, empty_location);
 526     p->next->next = symbol_list_new (endtoken, empty_location);
 527     p->next->next->next = symbol_list_new (NULL, empty_location);
 528     p->next->next->next->next = grammar;
 529     nrules += 1;
 530     nritems += 3;
 531     grammar = p;
 532   }
 533
 534   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 535     abort ();
 536
 537   xfclose (gram_in);
 538
 539   /* Assign the symbols their symbol numbers.  Write #defines for the
 540      token symbols into FDEFINES if requested.  */
 541   symbols_pack ();
 542
 543   /* Convert the grammar into the format described in gram.h.  */
 544   packgram ();
 545
 546   /* The grammar as a symbol_list is no longer needed. */
 547   LIST_FREE (symbol_list, grammar);
 548 }