src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003,
   4    2005, 2006 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 "reader.h"
  35 #include "symlist.h"
  36 #include "symtab.h"
  37 #include "scan-gram.h"
  38 #include "scan-code.h"
  39
  40 static void check_and_convert_grammar (void);
  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 union_seen = false;
  48
  49 /* Was a tag seen?  */
  50 bool tag_seen = false;
  51
  52 /* Should rules have a default precedence?  */
  53 bool default_prec = true;
  54 \f
  55 /*-----------------------.
  56 | Set the start symbol.  |
  57 `-----------------------*/
  58
  59 void
  60 grammar_start_symbol_set (symbol *sym, location loc)
  61 {
  62   if (start_flag)
  63     complain_at (loc, _("multiple %s declarations"), "%start");
  64   else
  65     {
  66       start_flag = true;
  67       startsymbol = sym;
  68       startsymbol_location = loc;
  69     }
  70 }
  71
  72
  73 /*---------------------------------------------------------------------.
  74 | There are two prologues: one before the first %union and one after.  |
  75 | Augment the one specified by POST.                                   |
  76 `---------------------------------------------------------------------*/
  77
  78 void
  79 prologue_augment (const char *prologue, location loc, bool post)
  80 {
  81   struct obstack *oout =
  82     !post ? &pre_prologue_obstack : &post_prologue_obstack;
  83
  84   obstack_fgrow1 (oout, "]b4_syncline(%d, [[", loc.start.line);
  85   /* FIXME: Protection of M4 characters missing here.  See
  86      output.c:escaped_output.  */
  87   MUSCLE_OBSTACK_SGROW (oout,
  88                         quotearg_style (c_quoting_style, loc.start.file));
  89   obstack_sgrow (oout, "]])[\n");
  90   obstack_sgrow (oout, prologue);
  91 }
  92
  93 \f
  94
  95 /*------------------------------------------------------------------------.
  96 | Return the merger index for a merging function named NAME.  Records the |
  97 | function, if new, in MERGER_LIST.                                       |
  98 `------------------------------------------------------------------------*/
  99
 100 static int
 101 get_merge_function (uniqstr name)
 102 {
 103   merger_list *syms;
 104   merger_list head;
 105   int n;
 106
 107   if (! glr_parser)
 108     return 0;
 109
 110   head.next = merge_functions;
 111   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
 112     if (UNIQSTR_EQ (name, syms->next->name))
 113       break;
 114   if (syms->next == NULL)
 115     {
 116       syms->next = xmalloc (sizeof syms->next[0]);
 117       syms->next->name = uniqstr_new (name);
 118       /* After all symbol type declarations have been parsed, packgram invokes
 119          record_merge_function_type to set the type.  */
 120       syms->next->type = NULL;
 121       syms->next->next = NULL;
 122       merge_functions = head.next;
 123     }
 124   return n;
 125 }
 126
 127 /*-------------------------------------------------------------------------.
 128 | For the existing merging function with index MERGER, record the result   |
 129 | type as TYPE as required by the lhs of the rule whose %merge declaration |
 130 | is at DECLARATION_LOC.                                                   |
 131 `-------------------------------------------------------------------------*/
 132
 133 static void
 134 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
 135 {
 136   int merger_find;
 137   merger_list *merge_function;
 138
 139   if (merger <= 0)
 140     return;
 141
 142   if (type == NULL)
 143     type = uniqstr_new ("");
 144
 145   merger_find = 1;
 146   for (merge_function = merge_functions;
 147        merge_function != NULL && merger_find != merger;
 148        merge_function = merge_function->next)
 149     merger_find += 1;
 150   aver (merge_function != NULL && merger_find == merger);
 151   if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
 152     {
 153       complain_at (declaration_loc,
 154                    _("result type clash on merge function `%s': <%s> != <%s>"),
 155                    merge_function->name, type, merge_function->type);
 156       complain_at (merge_function->type_declaration_location,
 157                    _("previous declaration"));
 158     }
 159   merge_function->type = uniqstr_new (type);
 160   merge_function->type_declaration_location = declaration_loc;
 161 }
 162
 163 /*--------------------------------------.
 164 | Free all merge-function definitions.  |
 165 `--------------------------------------*/
 166
 167 void
 168 free_merger_functions (void)
 169 {
 170   merger_list *L0 = merge_functions;
 171   while (L0)
 172     {
 173       merger_list *L1 = L0->next;
 174       free (L0);
 175       L0 = L1;
 176     }
 177 }
 178
 179 \f
 180 /*-------------------------------------------------------------------.
 181 | Parse the input grammar into a one symbol_list structure.  Each    |
 182 | rule is represented by a sequence of symbols: the left hand side   |
 183 | followed by the contents of the right hand side, followed by a     |
 184 | null pointer instead of a symbol to terminate the rule.  The next  |
 185 | symbol is the lhs of the following rule.                           |
 186 |                                                                    |
 187 | All actions are copied out, labelled by the rule number they apply |
 188 | to.                                                                |
 189 `-------------------------------------------------------------------*/
 190
 191 /* The (currently) last symbol of GRAMMAR. */
 192 static symbol_list *grammar_end = NULL;
 193
 194 /* Append SYM to the grammar.  */
 195 static void
 196 grammar_symbol_append (symbol *sym, location loc)
 197 {
 198   symbol_list *p = symbol_list_sym_new (sym, loc);
 199
 200   if (grammar_end)
 201     grammar_end->next = p;
 202   else
 203     grammar = p;
 204
 205   grammar_end = p;
 206
 207   /* A null SYM stands for an end of rule; it is not an actual
 208      part of it.  */
 209   if (sym)
 210     ++nritems;
 211 }
 212
 213 /* The rule currently being defined, and the previous rule.
 214    CURRENT_RULE points to the first LHS of the current rule, while
 215    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 216 static symbol_list *current_rule = NULL;
 217 static symbol_list *previous_rule_end = NULL;
 218
 219
 220 /*----------------------------------------------.
 221 | Create a new rule for LHS in to the GRAMMAR.  |
 222 `----------------------------------------------*/
 223
 224 void
 225 grammar_current_rule_begin (symbol *lhs, location loc)
 226 {
 227   /* Start a new rule and record its lhs.  */
 228   ++nrules;
 229   previous_rule_end = grammar_end;
 230   grammar_symbol_append (lhs, loc);
 231   current_rule = grammar_end;
 232
 233   /* Mark the rule's lhs as a nonterminal if not already so.  */
 234   if (lhs->class == unknown_sym)
 235     {
 236       lhs->class = nterm_sym;
 237       lhs->number = nvars;
 238       ++nvars;
 239     }
 240   else if (lhs->class == token_sym)
 241     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 242 }
 243
 244
 245 /*----------------------------------------------------------------------.
 246 | A symbol should be used if either:                                    |
 247 |   1. It has a destructor.                                             |
 248 |   2. --warnings=midrule-values and the symbol is a mid-rule symbol    |
 249 |      (i.e., the generated LHS replacing a mid-rule action) that was   |
 250 |      assigned to or used, as in "exp: { $$ = 1; } { $$ = $1; }".      |
 251 `----------------------------------------------------------------------*/
 252
 253 static bool
 254 symbol_should_be_used (symbol_list const *s)
 255 {
 256   if (code_props_code_get (symbol_destructor_get (s->content.sym)))
 257     return true;
 258   if (warnings_flag & warnings_midrule_values)
 259     {
 260       if (s->midrule && code_props_is_value_used (s->midrule->action_props))
 261         return true;
 262       if (s->midrule_parent_rule)
 263         {
 264           symbol_list *rhs_node =
 265             symbol_list_n_get (s->midrule_parent_rule,
 266                                s->midrule_parent_rhs_index);
 267           if (code_props_is_value_used (rhs_node->action_props))
 268             return true;
 269         }
 270     }
 271   return false;
 272 }
 273
 274 /*----------------------------------------------------------------.
 275 | Check that the rule R is properly defined.  For instance, there |
 276 | should be no type clash on the default action.                  |
 277 `----------------------------------------------------------------*/
 278
 279 static void
 280 grammar_rule_check (const symbol_list *r)
 281 {
 282   /* Type check.
 283
 284      If there is an action, then there is nothing we can do: the user
 285      is allowed to shoot herself in the foot.
 286
 287      Don't worry about the default action if $$ is untyped, since $$'s
 288      value can't be used.  */
 289   if (!code_props_code_get (r->action_props) && r->content.sym->type_name)
 290     {
 291       symbol *first_rhs = r->next->content.sym;
 292       /* If $$ is being set in default way, report if any type mismatch.  */
 293       if (first_rhs)
 294         {
 295           char const *lhs_type = r->content.sym->type_name;
 296           const char *rhs_type =
 297             first_rhs->type_name ? first_rhs->type_name : "";
 298           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 299             warn_at (r->location,
 300                      _("type clash on default action: <%s> != <%s>"),
 301                      lhs_type, rhs_type);
 302         }
 303       /* Warn if there is no default for $$ but we need one.  */
 304       else
 305         warn_at (r->location,
 306                  _("empty rule for typed nonterminal, and no action"));
 307     }
 308
 309   /* Check that symbol values that should be used are in fact used.  */
 310   {
 311     symbol_list const *l = r;
 312     int n = 0;
 313     for (; l && l->content.sym; l = l->next, ++n)
 314       if (! (code_props_is_value_used (l->action_props)
 315              || !symbol_should_be_used (l)
 316              /* The default action, $$ = $1, `uses' both.  */
 317              || (!code_props_code_get (r->action_props)
 318                  && (n == 0 || n == 1))))
 319         {
 320           if (n)
 321             warn_at (r->location, _("unused value: $%d"), n);
 322           else
 323             warn_at (r->location, _("unset value: $$"));
 324         }
 325   }
 326 }
 327
 328
 329 /*-------------------------------------.
 330 | End the currently being grown rule.  |
 331 `-------------------------------------*/
 332
 333 void
 334 grammar_current_rule_end (location loc)
 335 {
 336   /* Put an empty link in the list to mark the end of this rule  */
 337   grammar_symbol_append (NULL, grammar_end->location);
 338   current_rule->location = loc;
 339 }
 340
 341
 342 /*-------------------------------------------------------------------.
 343 | The previous action turns out the be a mid-rule action.  Attach it |
 344 | to the current rule, i.e., create a dummy symbol, attach it this   |
 345 | mid-rule action, and append this dummy nonterminal to the current  |
 346 | rule.                                                              |
 347 `-------------------------------------------------------------------*/
 348
 349 void
 350 grammar_midrule_action (void)
 351 {
 352   /* Since the action was written out with this rule's number, we must
 353      give the new rule this number by inserting the new rule before
 354      it.  */
 355
 356   /* Make a DUMMY nonterminal, whose location is that of the midrule
 357      action.  Create the MIDRULE.  */
 358   location dummy_location =
 359     code_props_location_get (current_rule->action_props);
 360   symbol *dummy = dummy_symbol_get (dummy_location);
 361   symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
 362
 363   /* Make a new rule, whose body is empty, before the current one, so
 364      that the action just read can belong to it.  */
 365   ++nrules;
 366   ++nritems;
 367   /* Attach its location and actions to that of the DUMMY.  */
 368   midrule->location = dummy_location;
 369   code_props_rule_action_init (
 370     &midrule->action_props,
 371     code_props_code_get (current_rule->action_props),
 372     code_props_location_get (current_rule->action_props),
 373     midrule);
 374   code_props_none_init (&current_rule->action_props);
 375
 376   if (previous_rule_end)
 377     previous_rule_end->next = midrule;
 378   else
 379     grammar = midrule;
 380
 381   /* End the dummy's rule.  */
 382   midrule->next = symbol_list_sym_new (NULL, dummy_location);
 383   midrule->next->next = current_rule;
 384
 385   previous_rule_end = midrule->next;
 386
 387   /* Insert the dummy nonterminal replacing the midrule action into
 388      the current rule.  Bind it to its dedicated rule.  */
 389   grammar_current_rule_symbol_append (dummy, dummy_location);
 390   grammar_end->midrule = midrule;
 391   midrule->midrule_parent_rule = current_rule;
 392   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 393 }
 394
 395 /* Set the precedence symbol of the current rule to PRECSYM. */
 396
 397 void
 398 grammar_current_rule_prec_set (symbol *precsym, location loc)
 399 {
 400   if (current_rule->ruleprec)
 401     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 402   current_rule->ruleprec = precsym;
 403 }
 404
 405 /* Attach dynamic precedence DPREC to the current rule. */
 406
 407 void
 408 grammar_current_rule_dprec_set (int dprec, location loc)
 409 {
 410   if (! glr_parser)
 411     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 412   if (dprec <= 0)
 413     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 414   else if (current_rule->dprec != 0)
 415     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 416   current_rule->dprec = dprec;
 417 }
 418
 419 /* Attach a merge function NAME with argument type TYPE to current
 420    rule. */
 421
 422 void
 423 grammar_current_rule_merge_set (uniqstr name, location loc)
 424 {
 425   if (! glr_parser)
 426     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 427   if (current_rule->merger != 0)
 428     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 429   current_rule->merger = get_merge_function (name);
 430   current_rule->merger_declaration_location = loc;
 431 }
 432
 433 /* Attach SYM to the current rule.  If needed, move the previous
 434    action as a mid-rule action.  */
 435
 436 void
 437 grammar_current_rule_symbol_append (symbol *sym, location loc)
 438 {
 439   if (code_props_code_get (current_rule->action_props))
 440     grammar_midrule_action ();
 441   grammar_symbol_append (sym, loc);
 442 }
 443
 444 /* Attach an ACTION to the current rule.  */
 445
 446 void
 447 grammar_current_rule_action_append (const char *action, location loc)
 448 {
 449   if (code_props_code_get (current_rule->action_props))
 450     grammar_midrule_action ();
 451   /* After all symbol declarations have been parsed, packgram invokes
 452      code_props_translate_code.  */
 453   code_props_rule_action_init (&current_rule->action_props, action, loc,
 454                                current_rule);
 455 }
 456
 457 \f
 458 /*---------------------------------------------------------------.
 459 | Convert the rules into the representation using RRHS, RLHS and |
 460 | RITEM.                                                         |
 461 `---------------------------------------------------------------*/
 462
 463 static void
 464 packgram (void)
 465 {
 466   unsigned int itemno = 0;
 467   rule_number ruleno = 0;
 468   symbol_list *p = grammar;
 469
 470   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 471
 472   /* This sentinel is used by build_relations in gram.c.  */
 473   *ritem++ = 0;
 474
 475   rules = xnmalloc (nrules, sizeof *rules);
 476
 477   /* Before invoking grammar_rule_check on any rule, make sure all actions have
 478      already been scanned in order to set `used' flags.  Otherwise, checking
 479      that a midrule's $$ should be set will not always work properly because
 480      the check must forward-reference the midrule's parent rule.  For the same
 481      reason, all the `used' flags must be set before checking whether to remove
 482      `$' from any midrule symbol name.  */
 483   while (p)
 484     {
 485       code_props_translate_code (&p->action_props);
 486       if (p)
 487         p = p->next;
 488     }
 489
 490   p = grammar;
 491   while (p)
 492     {
 493       int rule_length = 0;
 494       symbol *ruleprec = p->ruleprec;
 495       record_merge_function_type (p->merger, p->content.sym->type_name,
 496                                   p->merger_declaration_location);
 497       rules[ruleno].user_number = ruleno;
 498       rules[ruleno].number = ruleno;
 499       rules[ruleno].lhs = p->content.sym;
 500       rules[ruleno].rhs = ritem + itemno;
 501       rules[ruleno].prec = NULL;
 502       rules[ruleno].dprec = p->dprec;
 503       rules[ruleno].merger = p->merger;
 504       rules[ruleno].precsym = NULL;
 505       rules[ruleno].location = p->location;
 506       rules[ruleno].useful = true;
 507       rules[ruleno].action = code_props_code_get (p->action_props);
 508       if (rules[ruleno].action)
 509         rules[ruleno].action_location =
 510           code_props_location_get (p->action_props);
 511
 512       /* If the midrule's $$ is set or its $n is used, remove the `$' from the
 513          symbol name so that it's a user-defined symbol so that the default
 514          %destructor and %printer apply.  */
 515       if (p->midrule_parent_rule
 516           && (code_props_is_value_used (p->action_props)
 517               || code_props_is_value_used (
 518                    symbol_list_n_get (
 519                      p->midrule_parent_rule,
 520                      p->midrule_parent_rhs_index)->action_props)))
 521         p->content.sym->tag += 1;
 522
 523       /* Don't check the generated rule 0.  It has no action, so some rhs
 524          symbols may appear unused, but the parsing algorithm ensures that
 525          %destructor's are invoked appropriately.  */
 526       if (p != grammar)
 527         grammar_rule_check (p);
 528
 529       for (p = p->next; p && p->content.sym; p = p->next)
 530         {
 531           ++rule_length;
 532
 533           /* Don't allow rule_length == INT_MAX, since that might
 534              cause confusion with strtol if INT_MAX == LONG_MAX.  */
 535           if (rule_length == INT_MAX)
 536               fatal_at (rules[ruleno].location, _("rule is too long"));
 537
 538           /* item_number = symbol_number.
 539              But the former needs to contain more: negative rule numbers. */
 540           ritem[itemno++] =
 541             symbol_number_as_item_number (p->content.sym->number);
 542           /* A rule gets by default the precedence and associativity
 543              of its last token.  */
 544           if (p->content.sym->class == token_sym && default_prec)
 545             rules[ruleno].prec = p->content.sym;
 546         }
 547
 548       /* If this rule has a %prec,
 549          the specified symbol's precedence replaces the default.  */
 550       if (ruleprec)
 551         {
 552           rules[ruleno].precsym = ruleprec;
 553           rules[ruleno].prec = ruleprec;
 554         }
 555       /* An item ends by the rule number (negated).  */
 556       ritem[itemno++] = rule_number_as_item_number (ruleno);
 557       aver (itemno < ITEM_NUMBER_MAX);
 558       ++ruleno;
 559       aver (ruleno < RULE_NUMBER_MAX);
 560
 561       if (p)
 562         p = p->next;
 563     }
 564
 565   aver (itemno == nritems);
 566
 567   if (trace_flag & trace_sets)
 568     ritem_print (stderr);
 569 }
 570 \f
 571 /*------------------------------------------------------------------.
 572 | Read in the grammar specification and record it in the format     |
 573 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 574 | in each case forming the body of a C function (YYACTION) which    |
 575 | contains a switch statement to decide which action to execute.    |
 576 `------------------------------------------------------------------*/
 577
 578 void
 579 reader (void)
 580 {
 581   /* Initialize the symbol table.  */
 582   symbols_new ();
 583
 584   /* Construct the accept symbol. */
 585   accept = symbol_get ("$accept", empty_location);
 586   accept->class = nterm_sym;
 587   accept->number = nvars++;
 588
 589   /* Construct the error token */
 590   errtoken = symbol_get ("error", empty_location);
 591   errtoken->class = token_sym;
 592   errtoken->number = ntokens++;
 593
 594   /* Construct a token that represents all undefined literal tokens.
 595      It is always token number 2.  */
 596   undeftoken = symbol_get ("$undefined", empty_location);
 597   undeftoken->class = token_sym;
 598   undeftoken->number = ntokens++;
 599
 600   /* Initialize the obstacks. */
 601   obstack_init (&pre_prologue_obstack);
 602   obstack_init (&post_prologue_obstack);
 603
 604   gram_in = xfopen (grammar_file, "r");
 605
 606   gram__flex_debug = trace_flag & trace_scan;
 607   gram_debug = trace_flag & trace_parse;
 608   gram_scanner_initialize ();
 609   gram_parse ();
 610
 611   if (! complaint_issued)
 612     check_and_convert_grammar ();
 613
 614   xfclose (gram_in);
 615 }
 616
 617
 618 /*-------------------------------------------------------------.
 619 | Check the grammar that has just been read, and convert it to |
 620 | internal form.                                               |
 621 `-------------------------------------------------------------*/
 622
 623 static void
 624 check_and_convert_grammar (void)
 625 {
 626   /* Grammar has been read.  Do some checking.  */
 627   if (nrules == 0)
 628     fatal (_("no rules in the input grammar"));
 629
 630   /* Report any undefined symbols and consider them nonterminals.  */
 631   symbols_check_defined ();
 632
 633   /* If the user did not define her ENDTOKEN, do it now. */
 634   if (!endtoken)
 635     {
 636       endtoken = symbol_get ("$end", empty_location);
 637       endtoken->class = token_sym;
 638       endtoken->number = 0;
 639       /* Value specified by POSIX.  */
 640       endtoken->user_token_number = 0;
 641     }
 642
 643   /* Find the start symbol if no %start.  */
 644   if (!start_flag)
 645     {
 646       symbol_list *node;
 647       for (node = grammar;
 648            node != NULL && symbol_is_dummy (node->content.sym);
 649            node = node->next)
 650         {
 651           for (node = node->next;
 652                node != NULL && node->content.sym != NULL;
 653                node = node->next)
 654             ;
 655         }
 656       aver (node != NULL);
 657       grammar_start_symbol_set (node->content.sym,
 658                                 node->content.sym->location);
 659     }
 660
 661   /* Insert the initial rule, whose line is that of the first rule
 662      (not that of the start symbol):
 663
 664      accept: %start EOF.  */
 665   {
 666     symbol_list *p = symbol_list_sym_new (accept, empty_location);
 667     p->location = grammar->location;
 668     p->next = symbol_list_sym_new (startsymbol, empty_location);
 669     p->next->next = symbol_list_sym_new (endtoken, empty_location);
 670     p->next->next->next = symbol_list_sym_new (NULL, empty_location);
 671     p->next->next->next->next = grammar;
 672     nrules += 1;
 673     nritems += 3;
 674     grammar = p;
 675   }
 676
 677   aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 678
 679   /* Assign the symbols their symbol numbers.  Write #defines for the
 680      token symbols into FDEFINES if requested.  */
 681   symbols_pack ();
 682
 683   /* Convert the grammar into the format described in gram.h.  */
 684   packgram ();
 685
 686   /* The grammar as a symbol_list is no longer needed. */
 687   symbol_list_free (grammar);
 688 }