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