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