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_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   if (!start_flag)
 229     {
 230       startsymbol = lhs;
 231       startsymbol_location = loc;
 232       start_flag = true;
 233     }
 234
 235   /* Start a new rule and record its lhs.  */
 236   ++nrules;
 237   previous_rule_end = grammar_end;
 238   grammar_symbol_append (lhs, loc);
 239   current_rule = grammar_end;
 240
 241   /* Mark the rule's lhs as a nonterminal if not already so.  */
 242   if (lhs->class == unknown_sym)
 243     {
 244       lhs->class = nterm_sym;
 245       lhs->number = nvars;
 246       ++nvars;
 247     }
 248   else if (lhs->class == token_sym)
 249     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 250 }
 251
 252
 253 /*----------------------------------------------------------------------.
 254 | A symbol should be used if it has a destructor, or if it is a         |
 255 | mid-rule symbol (i.e., the generated LHS replacing a mid-rule         |
 256 | action) that was assigned to, as in "exp: { $$ = 1; } { $$ = $1; }".  |
 257 `----------------------------------------------------------------------*/
 258
 259 static bool
 260 symbol_should_be_used (symbol_list const *s)
 261 {
 262   return (symbol_destructor_get (s->sym)
 263           || (s->midrule && s->midrule->used));
 264 }
 265
 266 /*----------------------------------------------------------------.
 267 | Check that the rule R is properly defined.  For instance, there |
 268 | should be no type clash on the default action.                  |
 269 `----------------------------------------------------------------*/
 270
 271 static void
 272 grammar_rule_check (const symbol_list *r)
 273 {
 274   /* Type check.
 275
 276      If there is an action, then there is nothing we can do: the user
 277      is allowed to shoot herself in the foot.
 278
 279      Don't worry about the default action if $$ is untyped, since $$'s
 280      value can't be used.  */
 281   if (!r->action && r->sym->type_name)
 282     {
 283       symbol *first_rhs = r->next->sym;
 284       /* If $$ is being set in default way, report if any type mismatch.  */
 285       if (first_rhs)
 286         {
 287           char const *lhs_type = r->sym->type_name;
 288           const char *rhs_type =
 289             first_rhs->type_name ? first_rhs->type_name : "";
 290           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 291             warn_at (r->location,
 292                      _("type clash on default action: <%s> != <%s>"),
 293                      lhs_type, rhs_type);
 294         }
 295       /* Warn if there is no default for $$ but we need one.  */
 296       else
 297         warn_at (r->location,
 298                  _("empty rule for typed nonterminal, and no action"));
 299     }
 300
 301   /* Check that symbol values that should be used are in fact used.  */
 302   {
 303     symbol_list const *l = r;
 304     int n = 0;
 305     for (; l && l->sym; l = l->next, ++n)
 306       if (! (l->used
 307              || !symbol_should_be_used (l)
 308              /* The default action, $$ = $1, `uses' both.  */
 309              || (!r->action && (n == 0 || n == 1))))
 310         {
 311           if (n)
 312             warn_at (r->location, _("unused value: $%d"), n);
 313           else
 314             warn_at (r->location, _("unset value: $$"));
 315         }
 316   }
 317 }
 318
 319
 320 /*-------------------------------------.
 321 | End the currently being grown rule.  |
 322 `-------------------------------------*/
 323
 324 void
 325 grammar_current_rule_end (location loc)
 326 {
 327   /* Put an empty link in the list to mark the end of this rule  */
 328   grammar_symbol_append (NULL, grammar_end->location);
 329   current_rule->location = loc;
 330 }
 331
 332
 333 /*-------------------------------------------------------------------.
 334 | The previous action turns out the be a mid-rule action.  Attach it |
 335 | to the current rule, i.e., create a dummy symbol, attach it this   |
 336 | mid-rule action, and append this dummy nonterminal to the current  |
 337 | rule.                                                              |
 338 `-------------------------------------------------------------------*/
 339
 340 void
 341 grammar_midrule_action (void)
 342 {
 343   /* Since the action was written out with this rule's number, we must
 344      give the new rule this number by inserting the new rule before
 345      it.  */
 346
 347   /* Make a DUMMY nonterminal, whose location is that of the midrule
 348      action.  Create the MIDRULE.  */
 349   location dummy_location = current_rule->action_location;
 350   symbol *dummy = dummy_symbol_get (dummy_location);
 351   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 352
 353   /* Make a new rule, whose body is empty, before the current one, so
 354      that the action just read can belong to it.  */
 355   ++nrules;
 356   ++nritems;
 357   /* Attach its location and actions to that of the DUMMY.  */
 358   midrule->location = dummy_location;
 359   midrule->action = current_rule->action;
 360   midrule->action_location = dummy_location;
 361   current_rule->action = NULL;
 362   /* The action has not been translated yet, so $$ use hasn't been
 363      detected yet.  */
 364   midrule->used = false;
 365
 366   if (previous_rule_end)
 367     previous_rule_end->next = midrule;
 368   else
 369     grammar = midrule;
 370
 371   /* End the dummy's rule.  */
 372   midrule->next = symbol_list_new (NULL, dummy_location);
 373   midrule->next->next = current_rule;
 374
 375   previous_rule_end = midrule->next;
 376
 377   /* Insert the dummy nonterminal replacing the midrule action into
 378      the current rule.  Bind it to its dedicated rule.  */
 379   grammar_current_rule_symbol_append (dummy, dummy_location);
 380   grammar_end->midrule = midrule;
 381   midrule->midrule_parent_rule = current_rule;
 382   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 383 }
 384
 385 /* Set the precedence symbol of the current rule to PRECSYM. */
 386
 387 void
 388 grammar_current_rule_prec_set (symbol *precsym, location loc)
 389 {
 390   if (current_rule->ruleprec)
 391     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 392   current_rule->ruleprec = precsym;
 393 }
 394
 395 /* Attach dynamic precedence DPREC to the current rule. */
 396
 397 void
 398 grammar_current_rule_dprec_set (int dprec, location loc)
 399 {
 400   if (! glr_parser)
 401     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 402   if (dprec <= 0)
 403     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 404   else if (current_rule->dprec != 0)
 405     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 406   current_rule->dprec = dprec;
 407 }
 408
 409 /* Attach a merge function NAME with argument type TYPE to current
 410    rule. */
 411
 412 void
 413 grammar_current_rule_merge_set (uniqstr name, location loc)
 414 {
 415   if (! glr_parser)
 416     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 417   if (current_rule->merger != 0)
 418     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 419   current_rule->merger = get_merge_function (name);
 420   current_rule->merger_declaration_location = loc;
 421 }
 422
 423 /* Attach SYM to the current rule.  If needed, move the previous
 424    action as a mid-rule action.  */
 425
 426 void
 427 grammar_current_rule_symbol_append (symbol *sym, location loc)
 428 {
 429   if (current_rule->action)
 430     grammar_midrule_action ();
 431   grammar_symbol_append (sym, loc);
 432 }
 433
 434 /* Attach an ACTION to the current rule.  */
 435
 436 void
 437 grammar_current_rule_action_append (const char *action, location loc)
 438 {
 439   if (current_rule->action)
 440     grammar_midrule_action ();
 441   /* After all symbol declarations have been parsed, packgram invokes
 442      translate_rule_action.  */
 443   current_rule->action = action;
 444   current_rule->action_location = loc;
 445 }
 446
 447 \f
 448 /*---------------------------------------------------------------.
 449 | Convert the rules into the representation using RRHS, RLHS and |
 450 | RITEM.                                                         |
 451 `---------------------------------------------------------------*/
 452
 453 static void
 454 packgram (void)
 455 {
 456   unsigned int itemno = 0;
 457   rule_number ruleno = 0;
 458   symbol_list *p = grammar;
 459
 460   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 461
 462   /* This sentinel is used by build_relations in gram.c.  */
 463   *ritem++ = 0;
 464
 465   rules = xnmalloc (nrules, sizeof *rules);
 466
 467   while (p)
 468     {
 469       int rule_length = 0;
 470       symbol *ruleprec = p->ruleprec;
 471       record_merge_function_type (p->merger, p->sym->type_name,
 472                                   p->merger_declaration_location);
 473       rules[ruleno].user_number = ruleno;
 474       rules[ruleno].number = ruleno;
 475       rules[ruleno].lhs = p->sym;
 476       rules[ruleno].rhs = ritem + itemno;
 477       rules[ruleno].prec = NULL;
 478       rules[ruleno].dprec = p->dprec;
 479       rules[ruleno].merger = p->merger;
 480       rules[ruleno].precsym = NULL;
 481       rules[ruleno].location = p->location;
 482       rules[ruleno].useful = true;
 483       rules[ruleno].action = p->action ? translate_rule_action (p) : NULL;
 484       rules[ruleno].action_location = p->action_location;
 485
 486       /* If this rule contains midrules, rest assured that
 487          grammar_midrule_action inserted the midrules into grammar before this
 488          rule.  Thus, the midrule actions have already been scanned in order to
 489          set `used' flags for this rule's rhs, so grammar_rule_check will work
 490          properly.  */
 491       /* Don't check the generated rule 0.  It has no action, so some rhs
 492          symbols may appear unused, but the parsing algorithm ensures that
 493          %destructor's are invoked appropriately.  */
 494       if (p != grammar)
 495         grammar_rule_check (p);
 496
 497       for (p = p->next; p && p->sym; p = p->next)
 498         {
 499           ++rule_length;
 500
 501           /* Don't allow rule_length == INT_MAX, since that might
 502              cause confusion with strtol if INT_MAX == LONG_MAX.  */
 503           if (rule_length == INT_MAX)
 504               fatal_at (rules[ruleno].location, _("rule is too long"));
 505
 506           /* item_number = symbol_number.
 507              But the former needs to contain more: negative rule numbers. */
 508           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 509           /* A rule gets by default the precedence and associativity
 510              of its last token.  */
 511           if (p->sym->class == token_sym && default_prec)
 512             rules[ruleno].prec = p->sym;
 513         }
 514
 515       /* If this rule has a %prec,
 516          the specified symbol's precedence replaces the default.  */
 517       if (ruleprec)
 518         {
 519           rules[ruleno].precsym = ruleprec;
 520           rules[ruleno].prec = ruleprec;
 521         }
 522       /* An item ends by the rule number (negated).  */
 523       ritem[itemno++] = rule_number_as_item_number (ruleno);
 524       assert (itemno < ITEM_NUMBER_MAX);
 525       ++ruleno;
 526       assert (ruleno < RULE_NUMBER_MAX);
 527
 528       if (p)
 529         p = p->next;
 530     }
 531
 532   assert (itemno == nritems);
 533
 534   if (trace_flag & trace_sets)
 535     ritem_print (stderr);
 536 }
 537 \f
 538 /*------------------------------------------------------------------.
 539 | Read in the grammar specification and record it in the format     |
 540 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 541 | in each case forming the body of a C function (YYACTION) which    |
 542 | contains a switch statement to decide which action to execute.    |
 543 `------------------------------------------------------------------*/
 544
 545 void
 546 reader (void)
 547 {
 548   /* Initialize the symbol table.  */
 549   symbols_new ();
 550
 551   /* Construct the accept symbol. */
 552   accept = symbol_get ("$accept", empty_location);
 553   accept->class = nterm_sym;
 554   accept->number = nvars++;
 555
 556   /* Construct the error token */
 557   errtoken = symbol_get ("error", empty_location);
 558   errtoken->class = token_sym;
 559   errtoken->number = ntokens++;
 560
 561   /* Construct a token that represents all undefined literal tokens.
 562      It is always token number 2.  */
 563   undeftoken = symbol_get ("$undefined", empty_location);
 564   undeftoken->class = token_sym;
 565   undeftoken->number = ntokens++;
 566
 567   /* Initialize the obstacks. */
 568   obstack_init (&pre_prologue_obstack);
 569   obstack_init (&post_prologue_obstack);
 570
 571   gram_in = xfopen (grammar_file, "r");
 572
 573   gram__flex_debug = trace_flag & trace_scan;
 574   gram_debug = trace_flag & trace_parse;
 575   gram_scanner_initialize ();
 576   gram_parse ();
 577
 578   if (! complaint_issued)
 579     check_and_convert_grammar ();
 580
 581   xfclose (gram_in);
 582 }
 583
 584
 585 /*-------------------------------------------------------------.
 586 | Check the grammar that has just been read, and convert it to |
 587 | internal form.                                               |
 588 `-------------------------------------------------------------*/
 589
 590 static void
 591 check_and_convert_grammar (void)
 592 {
 593   /* Grammar has been read.  Do some checking.  */
 594   if (nrules == 0)
 595     fatal (_("no rules in the input grammar"));
 596
 597   /* Report any undefined symbols and consider them nonterminals.  */
 598   symbols_check_defined ();
 599
 600   /* If the user did not define her ENDTOKEN, do it now. */
 601   if (!endtoken)
 602     {
 603       endtoken = symbol_get ("$end", empty_location);
 604       endtoken->class = token_sym;
 605       endtoken->number = 0;
 606       /* Value specified by POSIX.  */
 607       endtoken->user_token_number = 0;
 608     }
 609
 610   /* Insert the initial rule, whose line is that of the first rule
 611      (not that of the start symbol):
 612
 613      accept: %start EOF.  */
 614   {
 615     symbol_list *p = symbol_list_new (accept, empty_location);
 616     p->location = grammar->location;
 617     p->next = symbol_list_new (startsymbol, empty_location);
 618     p->next->next = symbol_list_new (endtoken, empty_location);
 619     p->next->next->next = symbol_list_new (NULL, empty_location);
 620     p->next->next->next->next = grammar;
 621     nrules += 1;
 622     nritems += 3;
 623     grammar = p;
 624   }
 625
 626   assert (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 627
 628   /* Assign the symbols their symbol numbers.  Write #defines for the
 629      token symbols into FDEFINES if requested.  */
 630   symbols_pack ();
 631
 632   /* Convert the grammar into the format described in gram.h.  */
 633   packgram ();
 634
 635   /* The grammar as a symbol_list is no longer needed. */
 636   LIST_FREE (symbol_list, grammar);
 637 }