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 it has a destructor, or if it is a         |
 247 | mid-rule symbol (i.e., the generated LHS replacing a mid-rule         |
 248 | action) that was assigned to, as in "exp: { $$ = 1; } { $$ = $1; }".  |
 249 `----------------------------------------------------------------------*/
 250
 251 static bool
 252 symbol_should_be_used (symbol_list const *s)
 253 {
 254   return (symbol_destructor_get (s->content.sym)
 255           || (s->midrule && s->midrule->used));
 256 }
 257
 258 /*----------------------------------------------------------------.
 259 | Check that the rule R is properly defined.  For instance, there |
 260 | should be no type clash on the default action.                  |
 261 `----------------------------------------------------------------*/
 262
 263 static void
 264 grammar_rule_check (const symbol_list *r)
 265 {
 266   /* Type check.
 267
 268      If there is an action, then there is nothing we can do: the user
 269      is allowed to shoot herself in the foot.
 270
 271      Don't worry about the default action if $$ is untyped, since $$'s
 272      value can't be used.  */
 273   if (!r->action && r->content.sym->type_name)
 274     {
 275       symbol *first_rhs = r->next->content.sym;
 276       /* If $$ is being set in default way, report if any type mismatch.  */
 277       if (first_rhs)
 278         {
 279           char const *lhs_type = r->content.sym->type_name;
 280           const char *rhs_type =
 281             first_rhs->type_name ? first_rhs->type_name : "";
 282           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 283             warn_at (r->location,
 284                      _("type clash on default action: <%s> != <%s>"),
 285                      lhs_type, rhs_type);
 286         }
 287       /* Warn if there is no default for $$ but we need one.  */
 288       else
 289         warn_at (r->location,
 290                  _("empty rule for typed nonterminal, and no action"));
 291     }
 292
 293   /* Check that symbol values that should be used are in fact used.  */
 294   {
 295     symbol_list const *l = r;
 296     int n = 0;
 297     for (; l && l->content.sym; l = l->next, ++n)
 298       if (! (l->used
 299              || !symbol_should_be_used (l)
 300              /* The default action, $$ = $1, `uses' both.  */
 301              || (!r->action && (n == 0 || n == 1))))
 302         {
 303           if (n)
 304             warn_at (r->location, _("unused value: $%d"), n);
 305           else
 306             warn_at (r->location, _("unset value: $$"));
 307         }
 308   }
 309 }
 310
 311
 312 /*-------------------------------------.
 313 | End the currently being grown rule.  |
 314 `-------------------------------------*/
 315
 316 void
 317 grammar_current_rule_end (location loc)
 318 {
 319   /* Put an empty link in the list to mark the end of this rule  */
 320   grammar_symbol_append (NULL, grammar_end->location);
 321   current_rule->location = loc;
 322 }
 323
 324
 325 /*-------------------------------------------------------------------.
 326 | The previous action turns out the be a mid-rule action.  Attach it |
 327 | to the current rule, i.e., create a dummy symbol, attach it this   |
 328 | mid-rule action, and append this dummy nonterminal to the current  |
 329 | rule.                                                              |
 330 `-------------------------------------------------------------------*/
 331
 332 void
 333 grammar_midrule_action (void)
 334 {
 335   /* Since the action was written out with this rule's number, we must
 336      give the new rule this number by inserting the new rule before
 337      it.  */
 338
 339   /* Make a DUMMY nonterminal, whose location is that of the midrule
 340      action.  Create the MIDRULE.  */
 341   location dummy_location = current_rule->action_location;
 342   symbol *dummy = dummy_symbol_get (dummy_location);
 343   symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
 344
 345   /* Make a new rule, whose body is empty, before the current one, so
 346      that the action just read can belong to it.  */
 347   ++nrules;
 348   ++nritems;
 349   /* Attach its location and actions to that of the DUMMY.  */
 350   midrule->location = dummy_location;
 351   midrule->action = current_rule->action;
 352   midrule->action_location = dummy_location;
 353   current_rule->action = NULL;
 354   /* The action has not been translated yet, so $$ use hasn't been
 355      detected yet.  */
 356   midrule->used = false;
 357
 358   if (previous_rule_end)
 359     previous_rule_end->next = midrule;
 360   else
 361     grammar = midrule;
 362
 363   /* End the dummy's rule.  */
 364   midrule->next = symbol_list_sym_new (NULL, dummy_location);
 365   midrule->next->next = current_rule;
 366
 367   previous_rule_end = midrule->next;
 368
 369   /* Insert the dummy nonterminal replacing the midrule action into
 370      the current rule.  Bind it to its dedicated rule.  */
 371   grammar_current_rule_symbol_append (dummy, dummy_location);
 372   grammar_end->midrule = midrule;
 373   midrule->midrule_parent_rule = current_rule;
 374   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 375 }
 376
 377 /* Set the precedence symbol of the current rule to PRECSYM. */
 378
 379 void
 380 grammar_current_rule_prec_set (symbol *precsym, location loc)
 381 {
 382   if (current_rule->ruleprec)
 383     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 384   current_rule->ruleprec = precsym;
 385 }
 386
 387 /* Attach dynamic precedence DPREC to the current rule. */
 388
 389 void
 390 grammar_current_rule_dprec_set (int dprec, location loc)
 391 {
 392   if (! glr_parser)
 393     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 394   if (dprec <= 0)
 395     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 396   else if (current_rule->dprec != 0)
 397     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 398   current_rule->dprec = dprec;
 399 }
 400
 401 /* Attach a merge function NAME with argument type TYPE to current
 402    rule. */
 403
 404 void
 405 grammar_current_rule_merge_set (uniqstr name, location loc)
 406 {
 407   if (! glr_parser)
 408     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 409   if (current_rule->merger != 0)
 410     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 411   current_rule->merger = get_merge_function (name);
 412   current_rule->merger_declaration_location = loc;
 413 }
 414
 415 /* Attach SYM to the current rule.  If needed, move the previous
 416    action as a mid-rule action.  */
 417
 418 void
 419 grammar_current_rule_symbol_append (symbol *sym, location loc)
 420 {
 421   if (current_rule->action)
 422     grammar_midrule_action ();
 423   grammar_symbol_append (sym, loc);
 424 }
 425
 426 /* Attach an ACTION to the current rule.  */
 427
 428 void
 429 grammar_current_rule_action_append (const char *action, location loc)
 430 {
 431   if (current_rule->action)
 432     grammar_midrule_action ();
 433   /* After all symbol declarations have been parsed, packgram invokes
 434      translate_rule_action.  */
 435   current_rule->action = action;
 436   current_rule->action_location = loc;
 437 }
 438
 439 \f
 440 /*---------------------------------------------------------------.
 441 | Convert the rules into the representation using RRHS, RLHS and |
 442 | RITEM.                                                         |
 443 `---------------------------------------------------------------*/
 444
 445 static void
 446 packgram (void)
 447 {
 448   unsigned int itemno = 0;
 449   rule_number ruleno = 0;
 450   symbol_list *p = grammar;
 451
 452   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 453
 454   /* This sentinel is used by build_relations in gram.c.  */
 455   *ritem++ = 0;
 456
 457   rules = xnmalloc (nrules, sizeof *rules);
 458
 459   while (p)
 460     {
 461       int rule_length = 0;
 462       symbol *ruleprec = p->ruleprec;
 463       record_merge_function_type (p->merger, p->content.sym->type_name,
 464                                   p->merger_declaration_location);
 465       rules[ruleno].user_number = ruleno;
 466       rules[ruleno].number = ruleno;
 467       rules[ruleno].lhs = p->content.sym;
 468       rules[ruleno].rhs = ritem + itemno;
 469       rules[ruleno].prec = NULL;
 470       rules[ruleno].dprec = p->dprec;
 471       rules[ruleno].merger = p->merger;
 472       rules[ruleno].precsym = NULL;
 473       rules[ruleno].location = p->location;
 474       rules[ruleno].useful = true;
 475       rules[ruleno].action = p->action ? translate_rule_action (p) : NULL;
 476       rules[ruleno].action_location = p->action_location;
 477
 478       /* If this rule contains midrules, rest assured that
 479          grammar_midrule_action inserted the midrules into grammar before this
 480          rule.  Thus, the midrule actions have already been scanned in order to
 481          set `used' flags for this rule's rhs, so grammar_rule_check will work
 482          properly.  */
 483       /* Don't check the generated rule 0.  It has no action, so some rhs
 484          symbols may appear unused, but the parsing algorithm ensures that
 485          %destructor's are invoked appropriately.  */
 486       if (p != grammar)
 487         grammar_rule_check (p);
 488
 489       for (p = p->next; p && p->content.sym; p = p->next)
 490         {
 491           ++rule_length;
 492
 493           /* Don't allow rule_length == INT_MAX, since that might
 494              cause confusion with strtol if INT_MAX == LONG_MAX.  */
 495           if (rule_length == INT_MAX)
 496               fatal_at (rules[ruleno].location, _("rule is too long"));
 497
 498           /* item_number = symbol_number.
 499              But the former needs to contain more: negative rule numbers. */
 500           ritem[itemno++] =
 501             symbol_number_as_item_number (p->content.sym->number);
 502           /* A rule gets by default the precedence and associativity
 503              of its last token.  */
 504           if (p->content.sym->class == token_sym && default_prec)
 505             rules[ruleno].prec = p->content.sym;
 506         }
 507
 508       /* If this rule has a %prec,
 509          the specified symbol's precedence replaces the default.  */
 510       if (ruleprec)
 511         {
 512           rules[ruleno].precsym = ruleprec;
 513           rules[ruleno].prec = ruleprec;
 514         }
 515       /* An item ends by the rule number (negated).  */
 516       ritem[itemno++] = rule_number_as_item_number (ruleno);
 517       aver (itemno < ITEM_NUMBER_MAX);
 518       ++ruleno;
 519       aver (ruleno < RULE_NUMBER_MAX);
 520
 521       if (p)
 522         p = p->next;
 523     }
 524
 525   aver (itemno == nritems);
 526
 527   if (trace_flag & trace_sets)
 528     ritem_print (stderr);
 529 }
 530 \f
 531 /*------------------------------------------------------------------.
 532 | Read in the grammar specification and record it in the format     |
 533 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 534 | in each case forming the body of a C function (YYACTION) which    |
 535 | contains a switch statement to decide which action to execute.    |
 536 `------------------------------------------------------------------*/
 537
 538 void
 539 reader (void)
 540 {
 541   /* Initialize the symbol table.  */
 542   symbols_new ();
 543
 544   /* Construct the accept symbol. */
 545   accept = symbol_get ("$accept", empty_location);
 546   accept->class = nterm_sym;
 547   accept->number = nvars++;
 548
 549   /* Construct the error token */
 550   errtoken = symbol_get ("error", empty_location);
 551   errtoken->class = token_sym;
 552   errtoken->number = ntokens++;
 553
 554   /* Construct a token that represents all undefined literal tokens.
 555      It is always token number 2.  */
 556   undeftoken = symbol_get ("$undefined", empty_location);
 557   undeftoken->class = token_sym;
 558   undeftoken->number = ntokens++;
 559
 560   /* Initialize the obstacks. */
 561   obstack_init (&pre_prologue_obstack);
 562   obstack_init (&post_prologue_obstack);
 563
 564   gram_in = xfopen (grammar_file, "r");
 565
 566   gram__flex_debug = trace_flag & trace_scan;
 567   gram_debug = trace_flag & trace_parse;
 568   gram_scanner_initialize ();
 569   gram_parse ();
 570
 571   if (! complaint_issued)
 572     check_and_convert_grammar ();
 573
 574   xfclose (gram_in);
 575 }
 576
 577
 578 /*-------------------------------------------------------------.
 579 | Check the grammar that has just been read, and convert it to |
 580 | internal form.                                               |
 581 `-------------------------------------------------------------*/
 582
 583 static void
 584 check_and_convert_grammar (void)
 585 {
 586   /* Grammar has been read.  Do some checking.  */
 587   if (nrules == 0)
 588     fatal (_("no rules in the input grammar"));
 589
 590   /* Report any undefined symbols and consider them nonterminals.  */
 591   symbols_check_defined ();
 592
 593   /* If the user did not define her ENDTOKEN, do it now. */
 594   if (!endtoken)
 595     {
 596       endtoken = symbol_get ("$end", empty_location);
 597       endtoken->class = token_sym;
 598       endtoken->number = 0;
 599       /* Value specified by POSIX.  */
 600       endtoken->user_token_number = 0;
 601     }
 602
 603   /* Find the start symbol if no %start.  */
 604   if (!start_flag)
 605     {
 606       symbol_list *node;
 607       for (node = grammar;
 608            node != NULL && symbol_is_dummy (node->content.sym);
 609            node = node->next)
 610         {
 611           for (node = node->next;
 612                node != NULL && node->content.sym != NULL;
 613                node = node->next)
 614             ;
 615         }
 616       aver (node != NULL);
 617       grammar_start_symbol_set (node->content.sym,
 618                                 node->content.sym->location);
 619     }
 620
 621   /* Insert the initial rule, whose line is that of the first rule
 622      (not that of the start symbol):
 623
 624      accept: %start EOF.  */
 625   {
 626     symbol_list *p = symbol_list_sym_new (accept, empty_location);
 627     p->location = grammar->location;
 628     p->next = symbol_list_sym_new (startsymbol, empty_location);
 629     p->next->next = symbol_list_sym_new (endtoken, empty_location);
 630     p->next->next->next = symbol_list_sym_new (NULL, empty_location);
 631     p->next->next->next->next = grammar;
 632     nrules += 1;
 633     nritems += 3;
 634     grammar = p;
 635   }
 636
 637   aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 638
 639   /* Assign the symbols their symbol numbers.  Write #defines for the
 640      token symbols into FDEFINES if requested.  */
 641   symbols_pack ();
 642
 643   /* Convert the grammar into the format described in gram.h.  */
 644   packgram ();
 645
 646   /* The grammar as a symbol_list is no longer needed. */
 647   LIST_FREE (symbol_list, grammar);
 648 }