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 #include <assert.h>
  26
  27 #include <quotearg.h>
  28
  29 #include "complain.h"
  30 #include "conflicts.h"
  31 #include "files.h"
  32 #include "getargs.h"
  33 #include "gram.h"
  34 #include "muscle_tab.h"
  35 #include "reader.h"
  36 #include "symlist.h"
  37 #include "symtab.h"
  38 #include "scan-gram.h"
  39 #include "scan-code.h"
  40
  41 static void check_and_convert_grammar (void);
  42
  43 static symbol_list *grammar = NULL;
  44 static bool start_flag = false;
  45 merger_list *merge_functions;
  46
  47 /* Was %union seen?  */
  48 bool union_seen = false;
  49
  50 /* Was a tag seen?  */
  51 bool tag_seen = false;
  52
  53 /* Should rules have a default precedence?  */
  54 bool default_prec = true;
  55 \f
  56 /*-----------------------.
  57 | Set the start symbol.  |
  58 `-----------------------*/
  59
  60 void
  61 grammar_start_symbol_set (symbol *sym, location loc)
  62 {
  63   if (start_flag)
  64     complain_at (loc, _("multiple %s declarations"), "%start");
  65   else
  66     {
  67       start_flag = true;
  68       startsymbol = sym;
  69       startsymbol_location = loc;
  70     }
  71 }
  72
  73
  74 /*---------------------------------------------------------------------.
  75 | There are two prologues: one before the first %union and one after.  |
  76 |  Augment the one specified by POST.                                  |
  77 `---------------------------------------------------------------------*/
  78
  79 void
  80 prologue_augment (const char *prologue, location loc, bool post)
  81 {
  82   struct obstack *oout =
  83     !post ? &pre_prologue_obstack : &post_prologue_obstack;
  84
  85   obstack_fgrow1 (oout, "]b4_syncline(%d, [[", loc.start.line);
  86   /* FIXME: Protection of M4 characters missing here.  See
  87      output.c:escaped_output.  */
  88   MUSCLE_OBSTACK_SGROW (oout,
  89                         quotearg_style (c_quoting_style, loc.start.file));
  90   obstack_sgrow (oout, "]])[\n");
  91   obstack_sgrow (oout, prologue);
  92 }
  93
  94 \f
  95
  96 /*------------------------------------------------------------------------.
  97 | Return the merger index for a merging function named NAME.  Records the |
  98 | function, if new, in MERGER_LIST.                                       |
  99 `------------------------------------------------------------------------*/
 100
 101 static int
 102 get_merge_function (uniqstr name)
 103 {
 104   merger_list *syms;
 105   merger_list head;
 106   int n;
 107
 108   if (! glr_parser)
 109     return 0;
 110
 111   head.next = merge_functions;
 112   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
 113     if (UNIQSTR_EQ (name, syms->next->name))
 114       break;
 115   if (syms->next == NULL)
 116     {
 117       syms->next = xmalloc (sizeof syms->next[0]);
 118       syms->next->name = uniqstr_new (name);
 119       /* After all symbol type declarations have been parsed, packgram invokes
 120          record_merge_function_type to set the type.  */
 121       syms->next->type = NULL;
 122       syms->next->next = NULL;
 123       merge_functions = head.next;
 124     }
 125   return n;
 126 }
 127
 128 /*-------------------------------------------------------------------------.
 129 | For the existing merging function with index MERGER, record the result   |
 130 | type as TYPE as required by the lhs of the rule whose %merge declaration |
 131 | is at DECLARATION_LOC.                                                   |
 132 `-------------------------------------------------------------------------*/
 133
 134 static void
 135 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
 136 {
 137   int merger_find;
 138   merger_list *merge_function;
 139
 140   if (merger <= 0)
 141     return;
 142
 143   if (type == NULL)
 144     type = uniqstr_new ("");
 145
 146   merger_find = 1;
 147   for (merge_function = merge_functions;
 148        merge_function != NULL && merger_find != merger;
 149        merge_function = merge_function->next)
 150     merger_find += 1;
 151   assert (merge_function != NULL && merger_find == merger);
 152   if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
 153     {
 154       complain_at (declaration_loc,
 155                    _("result type clash on merge function `%s': <%s> != <%s>"),
 156                    merge_function->name, type, merge_function->type);
 157       complain_at (merge_function->type_declaration_location,
 158                    _("previous declaration"));
 159     }
 160   merge_function->type = uniqstr_new (type);
 161   merge_function->type_declaration_location = declaration_loc;
 162 }
 163
 164 /*--------------------------------------.
 165 | Free all merge-function definitions.  |
 166 `--------------------------------------*/
 167
 168 void
 169 free_merger_functions (void)
 170 {
 171   merger_list *L0 = merge_functions;
 172   while (L0)
 173     {
 174       merger_list *L1 = L0->next;
 175       free (L0);
 176       L0 = L1;
 177     }
 178 }
 179
 180 \f
 181 /*-------------------------------------------------------------------.
 182 | Parse the input grammar into a one symbol_list structure.  Each    |
 183 | rule is represented by a sequence of symbols: the left hand side   |
 184 | followed by the contents of the right hand side, followed by a     |
 185 | null pointer instead of a symbol to terminate the rule.  The next  |
 186 | symbol is the lhs of the following rule.                           |
 187 |                                                                    |
 188 | All actions are copied out, labelled by the rule number they apply |
 189 | to.                                                                |
 190 `-------------------------------------------------------------------*/
 191
 192 /* The (currently) last symbol of GRAMMAR. */
 193 static symbol_list *grammar_end = NULL;
 194
 195 /* Append SYM to the grammar.  */
 196 static void
 197 grammar_symbol_append (symbol *sym, location loc)
 198 {
 199   symbol_list *p = symbol_list_sym_new (sym, loc);
 200
 201   if (grammar_end)
 202     grammar_end->next = p;
 203   else
 204     grammar = p;
 205
 206   grammar_end = p;
 207
 208   /* A null SYM stands for an end of rule; it is not an actual
 209      part of it.  */
 210   if (sym)
 211     ++nritems;
 212 }
 213
 214 /* The rule currently being defined, and the previous rule.
 215    CURRENT_RULE points to the first LHS of the current rule, while
 216    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 217 static symbol_list *current_rule = NULL;
 218 static symbol_list *previous_rule_end = NULL;
 219
 220
 221 /*----------------------------------------------.
 222 | Create a new rule for LHS in to the GRAMMAR.  |
 223 `----------------------------------------------*/
 224
 225 void
 226 grammar_current_rule_begin (symbol *lhs, location loc)
 227 {
 228   /* Start a new rule and record its lhs.  */
 229   ++nrules;
 230   previous_rule_end = grammar_end;
 231   grammar_symbol_append (lhs, loc);
 232   current_rule = grammar_end;
 233
 234   /* Mark the rule's lhs as a nonterminal if not already so.  */
 235   if (lhs->class == unknown_sym)
 236     {
 237       lhs->class = nterm_sym;
 238       lhs->number = nvars;
 239       ++nvars;
 240     }
 241   else if (lhs->class == token_sym)
 242     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 243 }
 244
 245
 246 /*----------------------------------------------------------------------.
 247 | A symbol should be used if it has a destructor, or if it is a         |
 248 | mid-rule symbol (i.e., the generated LHS replacing a mid-rule         |
 249 | action) that was assigned to, as in "exp: { $$ = 1; } { $$ = $1; }".  |
 250 `----------------------------------------------------------------------*/
 251
 252 static bool
 253 symbol_should_be_used (symbol_list const *s)
 254 {
 255   return (symbol_destructor_get (s->content.sym)
 256           || (s->midrule && s->midrule->used));
 257 }
 258
 259 /*----------------------------------------------------------------.
 260 | Check that the rule R is properly defined.  For instance, there |
 261 | should be no type clash on the default action.                  |
 262 `----------------------------------------------------------------*/
 263
 264 static void
 265 grammar_rule_check (const symbol_list *r)
 266 {
 267   /* Type check.
 268
 269      If there is an action, then there is nothing we can do: the user
 270      is allowed to shoot herself in the foot.
 271
 272      Don't worry about the default action if $$ is untyped, since $$'s
 273      value can't be used.  */
 274   if (!r->action && r->content.sym->type_name)
 275     {
 276       symbol *first_rhs = r->next->content.sym;
 277       /* If $$ is being set in default way, report if any type mismatch.  */
 278       if (first_rhs)
 279         {
 280           char const *lhs_type = r->content.sym->type_name;
 281           const char *rhs_type =
 282             first_rhs->type_name ? first_rhs->type_name : "";
 283           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 284             warn_at (r->location,
 285                      _("type clash on default action: <%s> != <%s>"),
 286                      lhs_type, rhs_type);
 287         }
 288       /* Warn if there is no default for $$ but we need one.  */
 289       else
 290         warn_at (r->location,
 291                  _("empty rule for typed nonterminal, and no action"));
 292     }
 293
 294   /* Check that symbol values that should be used are in fact used.  */
 295   {
 296     symbol_list const *l = r;
 297     int n = 0;
 298     for (; l && l->content.sym; l = l->next, ++n)
 299       if (! (l->used
 300              || !symbol_should_be_used (l)
 301              /* The default action, $$ = $1, `uses' both.  */
 302              || (!r->action && (n == 0 || n == 1))))
 303         {
 304           if (n)
 305             warn_at (r->location, _("unused value: $%d"), n);
 306           else
 307             warn_at (r->location, _("unset value: $$"));
 308         }
 309   }
 310 }
 311
 312
 313 /*-------------------------------------.
 314 | End the currently being grown rule.  |
 315 `-------------------------------------*/
 316
 317 void
 318 grammar_current_rule_end (location loc)
 319 {
 320   /* Put an empty link in the list to mark the end of this rule  */
 321   grammar_symbol_append (NULL, grammar_end->location);
 322   current_rule->location = loc;
 323 }
 324
 325
 326 /*-------------------------------------------------------------------.
 327 | The previous action turns out the be a mid-rule action.  Attach it |
 328 | to the current rule, i.e., create a dummy symbol, attach it this   |
 329 | mid-rule action, and append this dummy nonterminal to the current  |
 330 | rule.                                                              |
 331 `-------------------------------------------------------------------*/
 332
 333 void
 334 grammar_midrule_action (void)
 335 {
 336   /* Since the action was written out with this rule's number, we must
 337      give the new rule this number by inserting the new rule before
 338      it.  */
 339
 340   /* Make a DUMMY nonterminal, whose location is that of the midrule
 341      action.  Create the MIDRULE.  */
 342   location dummy_location = current_rule->action_location;
 343   symbol *dummy = dummy_symbol_get (dummy_location);
 344   symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
 345
 346   /* Make a new rule, whose body is empty, before the current one, so
 347      that the action just read can belong to it.  */
 348   ++nrules;
 349   ++nritems;
 350   /* Attach its location and actions to that of the DUMMY.  */
 351   midrule->location = dummy_location;
 352   midrule->action = current_rule->action;
 353   midrule->action_location = dummy_location;
 354   current_rule->action = NULL;
 355   /* The action has not been translated yet, so $$ use hasn't been
 356      detected yet.  */
 357   midrule->used = false;
 358
 359   if (previous_rule_end)
 360     previous_rule_end->next = midrule;
 361   else
 362     grammar = midrule;
 363
 364   /* End the dummy's rule.  */
 365   midrule->next = symbol_list_sym_new (NULL, dummy_location);
 366   midrule->next->next = current_rule;
 367
 368   previous_rule_end = midrule->next;
 369
 370   /* Insert the dummy nonterminal replacing the midrule action into
 371      the current rule.  Bind it to its dedicated rule.  */
 372   grammar_current_rule_symbol_append (dummy, dummy_location);
 373   grammar_end->midrule = midrule;
 374   midrule->midrule_parent_rule = current_rule;
 375   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 376 }
 377
 378 /* Set the precedence symbol of the current rule to PRECSYM. */
 379
 380 void
 381 grammar_current_rule_prec_set (symbol *precsym, location loc)
 382 {
 383   if (current_rule->ruleprec)
 384     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 385   current_rule->ruleprec = precsym;
 386 }
 387
 388 /* Attach dynamic precedence DPREC to the current rule. */
 389
 390 void
 391 grammar_current_rule_dprec_set (int dprec, location loc)
 392 {
 393   if (! glr_parser)
 394     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 395   if (dprec <= 0)
 396     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 397   else if (current_rule->dprec != 0)
 398     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 399   current_rule->dprec = dprec;
 400 }
 401
 402 /* Attach a merge function NAME with argument type TYPE to current
 403    rule. */
 404
 405 void
 406 grammar_current_rule_merge_set (uniqstr name, location loc)
 407 {
 408   if (! glr_parser)
 409     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 410   if (current_rule->merger != 0)
 411     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 412   current_rule->merger = get_merge_function (name);
 413   current_rule->merger_declaration_location = loc;
 414 }
 415
 416 /* Attach SYM to the current rule.  If needed, move the previous
 417    action as a mid-rule action.  */
 418
 419 void
 420 grammar_current_rule_symbol_append (symbol *sym, location loc)
 421 {
 422   if (current_rule->action)
 423     grammar_midrule_action ();
 424   grammar_symbol_append (sym, loc);
 425 }
 426
 427 /* Attach an ACTION to the current rule.  */
 428
 429 void
 430 grammar_current_rule_action_append (const char *action, location loc)
 431 {
 432   if (current_rule->action)
 433     grammar_midrule_action ();
 434   /* After all symbol declarations have been parsed, packgram invokes
 435      translate_rule_action.  */
 436   current_rule->action = action;
 437   current_rule->action_location = loc;
 438 }
 439
 440 \f
 441 /*---------------------------------------------------------------.
 442 | Convert the rules into the representation using RRHS, RLHS and |
 443 | RITEM.                                                         |
 444 `---------------------------------------------------------------*/
 445
 446 static void
 447 packgram (void)
 448 {
 449   unsigned int itemno = 0;
 450   rule_number ruleno = 0;
 451   symbol_list *p = grammar;
 452
 453   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 454
 455   /* This sentinel is used by build_relations in gram.c.  */
 456   *ritem++ = 0;
 457
 458   rules = xnmalloc (nrules, sizeof *rules);
 459
 460   while (p)
 461     {
 462       int rule_length = 0;
 463       symbol *ruleprec = p->ruleprec;
 464       record_merge_function_type (p->merger, p->content.sym->type_name,
 465                                   p->merger_declaration_location);
 466       rules[ruleno].user_number = ruleno;
 467       rules[ruleno].number = ruleno;
 468       rules[ruleno].lhs = p->content.sym;
 469       rules[ruleno].rhs = ritem + itemno;
 470       rules[ruleno].prec = NULL;
 471       rules[ruleno].dprec = p->dprec;
 472       rules[ruleno].merger = p->merger;
 473       rules[ruleno].precsym = NULL;
 474       rules[ruleno].location = p->location;
 475       rules[ruleno].useful = true;
 476       rules[ruleno].action = p->action ? translate_rule_action (p) : NULL;
 477       rules[ruleno].action_location = p->action_location;
 478
 479       /* If this rule contains midrules, rest assured that
 480          grammar_midrule_action inserted the midrules into grammar before this
 481          rule.  Thus, the midrule actions have already been scanned in order to
 482          set `used' flags for this rule's rhs, so grammar_rule_check will work
 483          properly.  */
 484       /* Don't check the generated rule 0.  It has no action, so some rhs
 485          symbols may appear unused, but the parsing algorithm ensures that
 486          %destructor's are invoked appropriately.  */
 487       if (p != grammar)
 488         grammar_rule_check (p);
 489
 490       for (p = p->next; p && p->content.sym; p = p->next)
 491         {
 492           ++rule_length;
 493
 494           /* Don't allow rule_length == INT_MAX, since that might
 495              cause confusion with strtol if INT_MAX == LONG_MAX.  */
 496           if (rule_length == INT_MAX)
 497               fatal_at (rules[ruleno].location, _("rule is too long"));
 498
 499           /* item_number = symbol_number.
 500              But the former needs to contain more: negative rule numbers. */
 501           ritem[itemno++] =
 502             symbol_number_as_item_number (p->content.sym->number);
 503           /* A rule gets by default the precedence and associativity
 504              of its last token.  */
 505           if (p->content.sym->class == token_sym && default_prec)
 506             rules[ruleno].prec = p->content.sym;
 507         }
 508
 509       /* If this rule has a %prec,
 510          the specified symbol's precedence replaces the default.  */
 511       if (ruleprec)
 512         {
 513           rules[ruleno].precsym = ruleprec;
 514           rules[ruleno].prec = ruleprec;
 515         }
 516       /* An item ends by the rule number (negated).  */
 517       ritem[itemno++] = rule_number_as_item_number (ruleno);
 518       assert (itemno < ITEM_NUMBER_MAX);
 519       ++ruleno;
 520       assert (ruleno < RULE_NUMBER_MAX);
 521
 522       if (p)
 523         p = p->next;
 524     }
 525
 526   assert (itemno == nritems);
 527
 528   if (trace_flag & trace_sets)
 529     ritem_print (stderr);
 530 }
 531 \f
 532 /*------------------------------------------------------------------.
 533 | Read in the grammar specification and record it in the format     |
 534 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 535 | in each case forming the body of a C function (YYACTION) which    |
 536 | contains a switch statement to decide which action to execute.    |
 537 `------------------------------------------------------------------*/
 538
 539 void
 540 reader (void)
 541 {
 542   /* Initialize the symbol table.  */
 543   symbols_new ();
 544
 545   /* Construct the accept symbol. */
 546   accept = symbol_get ("$accept", empty_location);
 547   accept->class = nterm_sym;
 548   accept->number = nvars++;
 549
 550   /* Construct the error token */
 551   errtoken = symbol_get ("error", empty_location);
 552   errtoken->class = token_sym;
 553   errtoken->number = ntokens++;
 554
 555   /* Construct a token that represents all undefined literal tokens.
 556      It is always token number 2.  */
 557   undeftoken = symbol_get ("$undefined", empty_location);
 558   undeftoken->class = token_sym;
 559   undeftoken->number = ntokens++;
 560
 561   /* Initialize the obstacks. */
 562   obstack_init (&pre_prologue_obstack);
 563   obstack_init (&post_prologue_obstack);
 564
 565   gram_in = xfopen (grammar_file, "r");
 566
 567   gram__flex_debug = trace_flag & trace_scan;
 568   gram_debug = trace_flag & trace_parse;
 569   gram_scanner_initialize ();
 570   gram_parse ();
 571
 572   if (! complaint_issued)
 573     check_and_convert_grammar ();
 574
 575   xfclose (gram_in);
 576 }
 577
 578
 579 /*-------------------------------------------------------------.
 580 | Check the grammar that has just been read, and convert it to |
 581 | internal form.                                               |
 582 `-------------------------------------------------------------*/
 583
 584 static void
 585 check_and_convert_grammar (void)
 586 {
 587   /* Grammar has been read.  Do some checking.  */
 588   if (nrules == 0)
 589     fatal (_("no rules in the input grammar"));
 590
 591   /* Report any undefined symbols and consider them nonterminals.  */
 592   symbols_check_defined ();
 593
 594   /* If the user did not define her ENDTOKEN, do it now. */
 595   if (!endtoken)
 596     {
 597       endtoken = symbol_get ("$end", empty_location);
 598       endtoken->class = token_sym;
 599       endtoken->number = 0;
 600       /* Value specified by POSIX.  */
 601       endtoken->user_token_number = 0;
 602     }
 603
 604   /* Find the start symbol if no %start.  */
 605   if (!start_flag)
 606     {
 607       symbol_list *node;
 608       for (node = grammar;
 609            node != NULL && symbol_is_dummy (node->content.sym);
 610            node = node->next)
 611         {
 612           for (node = node->next;
 613                node != NULL && node->content.sym != NULL;
 614                node = node->next)
 615             ;
 616         }
 617       assert (node != NULL);
 618       grammar_start_symbol_set (node->content.sym,
 619                                 node->content.sym->location);
 620     }
 621
 622   /* Insert the initial rule, whose line is that of the first rule
 623      (not that of the start symbol):
 624
 625      accept: %start EOF.  */
 626   {
 627     symbol_list *p = symbol_list_sym_new (accept, empty_location);
 628     p->location = grammar->location;
 629     p->next = symbol_list_sym_new (startsymbol, empty_location);
 630     p->next->next = symbol_list_sym_new (endtoken, empty_location);
 631     p->next->next->next = symbol_list_sym_new (NULL, empty_location);
 632     p->next->next->next->next = grammar;
 633     nrules += 1;
 634     nritems += 3;
 635     grammar = p;
 636   }
 637
 638   assert (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 639
 640   /* Assign the symbols their symbol numbers.  Write #defines for the
 641      token symbols into FDEFINES if requested.  */
 642   symbols_pack ();
 643
 644   /* Convert the grammar into the format described in gram.h.  */
 645   packgram ();
 646
 647   /* The grammar as a symbol_list is no longer needed. */
 648   LIST_FREE (symbol_list, grammar);
 649 }