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 the first %union and one after.  |
  73 |  Augment the one specified by POST.                                  |
  74 `---------------------------------------------------------------------*/
  75
  76 void
  77 prologue_augment (const char *prologue, location loc, bool post)
  78 {
  79   struct obstack *oout =
  80     !post ? &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.  Records the |
  95 | function, if new, in MERGER_LIST.                                       |
  96 `------------------------------------------------------------------------*/
  97
  98 static int
  99 get_merge_function (uniqstr name)
 100 {
 101   merger_list *syms;
 102   merger_list head;
 103   int n;
 104
 105   if (! glr_parser)
 106     return 0;
 107
 108   head.next = merge_functions;
 109   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
 110     if (UNIQSTR_EQ (name, syms->next->name))
 111       break;
 112   if (syms->next == NULL)
 113     {
 114       syms->next = xmalloc (sizeof syms->next[0]);
 115       syms->next->name = uniqstr_new (name);
 116       /* After all symbol type declarations have been parsed, packgram invokes
 117          record_merge_function_type to set the type.  */
 118       syms->next->type = NULL;
 119       syms->next->next = NULL;
 120       merge_functions = head.next;
 121     }
 122   return n;
 123 }
 124
 125 /*-------------------------------------------------------------------------.
 126 | For the existing merging function with index MERGER, record the result   |
 127 | type as TYPE as required by the lhs of the rule whose %merge declaration |
 128 | is at DECLARATION_LOC.                                                   |
 129 `-------------------------------------------------------------------------*/
 130
 131 static void
 132 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
 133 {
 134   int merger_find;
 135   merger_list *merge_function;
 136
 137   if (merger <= 0)
 138     return;
 139
 140   if (type == NULL)
 141     type = uniqstr_new ("");
 142
 143   merger_find = 1;
 144   for (merge_function = merge_functions;
 145        merge_function != NULL && merger_find != merger;
 146        merge_function = merge_function->next)
 147     merger_find += 1;
 148   assert (merge_function != NULL && merger_find == merger);
 149   if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
 150     {
 151       complain_at (declaration_loc,
 152                    _("result type clash on merge function `%s': <%s> != <%s>"),
 153                    merge_function->name, type, merge_function->type);
 154       complain_at (merge_function->type_declaration_location,
 155                    _("previous declaration"));
 156     }
 157   merge_function->type = uniqstr_new (type);
 158   merge_function->type_declaration_location = declaration_loc;
 159 }
 160
 161 /*--------------------------------------.
 162 | Free all merge-function definitions.  |
 163 `--------------------------------------*/
 164
 165 void
 166 free_merger_functions (void)
 167 {
 168   merger_list *L0 = merge_functions;
 169   while (L0)
 170     {
 171       merger_list *L1 = L0->next;
 172       free (L0);
 173       L0 = L1;
 174     }
 175 }
 176
 177 \f
 178 /*-------------------------------------------------------------------.
 179 | Parse the input grammar into a one symbol_list structure.  Each    |
 180 | rule is represented by a sequence of symbols: the left hand side   |
 181 | followed by the contents of the right hand side, followed by a     |
 182 | null pointer instead of a symbol to terminate the rule.  The next  |
 183 | symbol is the lhs of the following rule.                           |
 184 |                                                                    |
 185 | All actions are copied out, labelled by the rule number they apply |
 186 | to.                                                                |
 187 `-------------------------------------------------------------------*/
 188
 189 /* The (currently) last symbol of GRAMMAR. */
 190 static symbol_list *grammar_end = NULL;
 191
 192 /* Append SYM to the grammar.  */
 193 static void
 194 grammar_symbol_append (symbol *sym, location loc)
 195 {
 196   symbol_list *p = symbol_list_new (sym, loc);
 197
 198   if (grammar_end)
 199     grammar_end->next = p;
 200   else
 201     grammar = p;
 202
 203   grammar_end = p;
 204
 205   /* A null SYM stands for an end of rule; it is not an actual
 206      part of it.  */
 207   if (sym)
 208     ++nritems;
 209 }
 210
 211 /* The rule currently being defined, and the previous rule.
 212    CURRENT_RULE points to the first LHS of the current rule, while
 213    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 214 static symbol_list *current_rule = NULL;
 215 static symbol_list *previous_rule_end = NULL;
 216
 217
 218 /*----------------------------------------------.
 219 | Create a new rule for LHS in to the GRAMMAR.  |
 220 `----------------------------------------------*/
 221
 222 void
 223 grammar_current_rule_begin (symbol *lhs, location loc)
 224 {
 225   if (!start_flag)
 226     {
 227       startsymbol = lhs;
 228       startsymbol_location = loc;
 229       start_flag = true;
 230     }
 231
 232   /* Start a new rule and record its lhs.  */
 233   ++nrules;
 234   previous_rule_end = grammar_end;
 235   grammar_symbol_append (lhs, loc);
 236   current_rule = grammar_end;
 237
 238   /* Mark the rule's lhs as a nonterminal if not already so.  */
 239   if (lhs->class == unknown_sym)
 240     {
 241       lhs->class = nterm_sym;
 242       lhs->number = nvars;
 243       ++nvars;
 244     }
 245   else if (lhs->class == token_sym)
 246     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 247 }
 248
 249
 250 /*----------------------------------------------------------------------.
 251 | A symbol should be used if it has a destructor, or if it is a         |
 252 | mid-rule symbol (i.e., the generated LHS replacing a mid-rule         |
 253 | action) that was assigned to, as in "exp: { $$ = 1; } { $$ = $1; }".  |
 254 `----------------------------------------------------------------------*/
 255
 256 static bool
 257 symbol_should_be_used (symbol_list const *s)
 258 {
 259   return (s->sym->destructor
 260           || (s->midrule && s->midrule->used));
 261 }
 262
 263 /*----------------------------------------------------------------.
 264 | Check that the rule R is properly defined.  For instance, there |
 265 | should be no type clash on the default action.                  |
 266 `----------------------------------------------------------------*/
 267
 268 static void
 269 grammar_rule_check (const symbol_list *r)
 270 {
 271   /* Type check.
 272
 273      If there is an action, then there is nothing we can do: the user
 274      is allowed to shoot herself in the foot.
 275
 276      Don't worry about the default action if $$ is untyped, since $$'s
 277      value can't be used.  */
 278   if (!r->action && r->sym->type_name)
 279     {
 280       symbol *first_rhs = r->next->sym;
 281       /* If $$ is being set in default way, report if any type mismatch.  */
 282       if (first_rhs)
 283         {
 284           char const *lhs_type = r->sym->type_name;
 285           const char *rhs_type =
 286             first_rhs->type_name ? first_rhs->type_name : "";
 287           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 288             warn_at (r->location,
 289                      _("type clash on default action: <%s> != <%s>"),
 290                      lhs_type, rhs_type);
 291         }
 292       /* Warn if there is no default for $$ but we need one.  */
 293       else
 294         warn_at (r->location,
 295                  _("empty rule for typed nonterminal, and no action"));
 296     }
 297
 298   /* Check that symbol values that should be used are in fact used.  */
 299   {
 300     symbol_list const *l = r;
 301     int n = 0;
 302     for (; l && l->sym; l = l->next, ++n)
 303       if (! (l->used
 304              || !symbol_should_be_used (l)
 305              /* The default action, $$ = $1, `uses' both.  */
 306              || (!r->action && (n == 0 || n == 1))))
 307         {
 308           if (n)
 309             warn_at (r->location, _("unused value: $%d"), n);
 310           else
 311             warn_at (r->location, _("unset value: $$"));
 312         }
 313   }
 314 }
 315
 316
 317 /*-------------------------------------.
 318 | End the currently being grown rule.  |
 319 `-------------------------------------*/
 320
 321 void
 322 grammar_current_rule_end (location loc)
 323 {
 324   /* Put an empty link in the list to mark the end of this rule  */
 325   grammar_symbol_append (NULL, grammar_end->location);
 326   current_rule->location = loc;
 327 }
 328
 329
 330 /*-------------------------------------------------------------------.
 331 | The previous action turns out the be a mid-rule action.  Attach it |
 332 | to the current rule, i.e., create a dummy symbol, attach it this   |
 333 | mid-rule action, and append this dummy nonterminal to the current  |
 334 | rule.                                                              |
 335 `-------------------------------------------------------------------*/
 336
 337 void
 338 grammar_midrule_action (void)
 339 {
 340   /* Since the action was written out with this rule's number, we must
 341      give the new rule this number by inserting the new rule before
 342      it.  */
 343
 344   /* Make a DUMMY nonterminal, whose location is that of the midrule
 345      action.  Create the MIDRULE.  */
 346   location dummy_location = current_rule->action_location;
 347   symbol *dummy = dummy_symbol_get (dummy_location);
 348   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 349
 350   /* Make a new rule, whose body is empty, before the current one, so
 351      that the action just read can belong to it.  */
 352   ++nrules;
 353   ++nritems;
 354   /* Attach its location and actions to that of the DUMMY.  */
 355   midrule->location = dummy_location;
 356   midrule->action = current_rule->action;
 357   midrule->action_location = dummy_location;
 358   current_rule->action = NULL;
 359   /* The action has not been translated yet, so $$ use hasn't been
 360      detected yet.  */
 361   midrule->used = false;
 362
 363   if (previous_rule_end)
 364     previous_rule_end->next = midrule;
 365   else
 366     grammar = midrule;
 367
 368   /* End the dummy's rule.  */
 369   midrule->next = symbol_list_new (NULL, dummy_location);
 370   midrule->next->next = current_rule;
 371
 372   previous_rule_end = midrule->next;
 373
 374   /* Insert the dummy nonterminal replacing the midrule action into
 375      the current rule.  Bind it to its dedicated rule.  */
 376   grammar_current_rule_symbol_append (dummy, dummy_location);
 377   grammar_end->midrule = midrule;
 378   midrule->midrule_parent_rule = current_rule;
 379   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 380 }
 381
 382 /* Set the precedence symbol of the current rule to PRECSYM. */
 383
 384 void
 385 grammar_current_rule_prec_set (symbol *precsym, location loc)
 386 {
 387   if (current_rule->ruleprec)
 388     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 389   current_rule->ruleprec = precsym;
 390 }
 391
 392 /* Attach dynamic precedence DPREC to the current rule. */
 393
 394 void
 395 grammar_current_rule_dprec_set (int dprec, location loc)
 396 {
 397   if (! glr_parser)
 398     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 399   if (dprec <= 0)
 400     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 401   else if (current_rule->dprec != 0)
 402     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 403   current_rule->dprec = dprec;
 404 }
 405
 406 /* Attach a merge function NAME with argument type TYPE to current
 407    rule. */
 408
 409 void
 410 grammar_current_rule_merge_set (uniqstr name, location loc)
 411 {
 412   if (! glr_parser)
 413     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 414   if (current_rule->merger != 0)
 415     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 416   current_rule->merger = get_merge_function (name);
 417   current_rule->merger_declaration_location = loc;
 418 }
 419
 420 /* Attach SYM to the current rule.  If needed, move the previous
 421    action as a mid-rule action.  */
 422
 423 void
 424 grammar_current_rule_symbol_append (symbol *sym, location loc)
 425 {
 426   if (current_rule->action)
 427     grammar_midrule_action ();
 428   grammar_symbol_append (sym, loc);
 429 }
 430
 431 /* Attach an ACTION to the current rule.  */
 432
 433 void
 434 grammar_current_rule_action_append (const char *action, location loc)
 435 {
 436   if (current_rule->action)
 437     grammar_midrule_action ();
 438   /* After all symbol declarations have been parsed, packgram invokes
 439      translate_rule_action.  */
 440   current_rule->action = action;
 441   current_rule->action_location = loc;
 442 }
 443
 444 \f
 445 /*---------------------------------------------------------------.
 446 | Convert the rules into the representation using RRHS, RLHS and |
 447 | RITEM.                                                         |
 448 `---------------------------------------------------------------*/
 449
 450 static void
 451 packgram (void)
 452 {
 453   unsigned int itemno = 0;
 454   rule_number ruleno = 0;
 455   symbol_list *p = grammar;
 456
 457   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 458
 459   /* This sentinel is used by build_relations in gram.c.  */
 460   *ritem++ = 0;
 461
 462   rules = xnmalloc (nrules, sizeof *rules);
 463
 464   while (p)
 465     {
 466       int rule_length = 0;
 467       symbol *ruleprec = p->ruleprec;
 468       record_merge_function_type (p->merger, p->sym->type_name,
 469                                   p->merger_declaration_location);
 470       rules[ruleno].user_number = ruleno;
 471       rules[ruleno].number = ruleno;
 472       rules[ruleno].lhs = p->sym;
 473       rules[ruleno].rhs = ritem + itemno;
 474       rules[ruleno].prec = NULL;
 475       rules[ruleno].dprec = p->dprec;
 476       rules[ruleno].merger = p->merger;
 477       rules[ruleno].precsym = NULL;
 478       rules[ruleno].location = p->location;
 479       rules[ruleno].useful = true;
 480       rules[ruleno].action = p->action ? translate_rule_action (p) : NULL;
 481       rules[ruleno].action_location = p->action_location;
 482
 483       /* If this rule contains midrules, rest assured that
 484          grammar_midrule_action inserted the midrules into grammar before this
 485          rule.  Thus, the midrule actions have already been scanned in order to
 486          set `used' flags for this rule's rhs, so grammar_rule_check will work
 487          properly.  */
 488       grammar_rule_check (p);
 489
 490       for (p = p->next; p && p->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++] = symbol_number_as_item_number (p->sym->number);
 502           /* A rule gets by default the precedence and associativity
 503              of its last token.  */
 504           if (p->sym->class == token_sym && default_prec)
 505             rules[ruleno].prec = p->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       assert (itemno < ITEM_NUMBER_MAX);
 518       ++ruleno;
 519       assert (ruleno < RULE_NUMBER_MAX);
 520
 521       if (p)
 522         p = p->next;
 523     }
 524
 525   assert (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   /* Insert the initial rule, whose line is that of the first rule
 604      (not that of the start symbol):
 605
 606      accept: %start EOF.  */
 607   {
 608     symbol_list *p = symbol_list_new (accept, empty_location);
 609     p->location = grammar->location;
 610     p->next = symbol_list_new (startsymbol, empty_location);
 611     p->next->next = symbol_list_new (endtoken, empty_location);
 612     p->next->next->next = symbol_list_new (NULL, empty_location);
 613     p->next->next->next->next = grammar;
 614     nrules += 1;
 615     nritems += 3;
 616     grammar = p;
 617   }
 618
 619   assert (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 620
 621   /* Assign the symbols their symbol numbers.  Write #defines for the
 622      token symbols into FDEFINES if requested.  */
 623   symbols_pack ();
 624
 625   /* Convert the grammar into the format described in gram.h.  */
 626   packgram ();
 627
 628   /* The grammar as a symbol_list is no longer needed. */
 629   LIST_FREE (symbol_list, grammar);
 630 }