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 "output.h"
  35 #include "reader.h"
  36 #include "symlist.h"
  37 #include "symtab.h"
  38
  39 static void check_and_convert_grammar (void);
  40
  41 static symbol_list *grammar = NULL;
  42 static bool start_flag = false;
  43 merger_list *merge_functions;
  44
  45 /* Was %union seen?  */
  46 bool typed = false;
  47
  48 /* Should rules have a default precedence?  */
  49 bool default_prec = true;
  50 \f
  51 /*-----------------------.
  52 | Set the start symbol.  |
  53 `-----------------------*/
  54
  55 void
  56 grammar_start_symbol_set (symbol *sym, location loc)
  57 {
  58   if (start_flag)
  59     complain_at (loc, _("multiple %s declarations"), "%start");
  60   else
  61     {
  62       start_flag = true;
  63       startsymbol = sym;
  64       startsymbol_location = loc;
  65     }
  66 }
  67
  68
  69 /*----------------------------------------------------------------.
  70 | There are two prologues: one before %union, one after.  Augment |
  71 | the current one.                                                |
  72 `----------------------------------------------------------------*/
  73
  74 void
  75 prologue_augment (const char *prologue, location loc)
  76 {
  77   struct obstack *oout =
  78     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  79
  80   obstack_fgrow1 (oout, "]b4_syncline(%d, [[", loc.start.line);
  81   MUSCLE_OBSTACK_SGROW (oout,
  82                         quotearg_style (c_quoting_style, loc.start.file));
  83   obstack_sgrow (oout, "]])[\n");
  84   obstack_sgrow (oout, prologue);
  85 }
  86
  87 \f
  88
  89 /*-------------------------------------------------------------------.
  90 | Return the merger index for a merging function named NAME, whose   |
  91 | arguments have type TYPE.  Records the function, if new, in        |
  92 | MERGER_LIST.                                                       |
  93 `-------------------------------------------------------------------*/
  94
  95 static int
  96 get_merge_function (uniqstr name, uniqstr type, location loc)
  97 {
  98   merger_list *syms;
  99   merger_list head;
 100   int n;
 101
 102   if (! glr_parser)
 103     return 0;
 104
 105   if (type == NULL)
 106     type = uniqstr_new ("");
 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       syms->next->type = uniqstr_new (type);
 117       syms->next->next = NULL;
 118       merge_functions = head.next;
 119     }
 120   else if (!UNIQSTR_EQ (type, syms->next->type))
 121     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 122              name, type, syms->next->type);
 123   return n;
 124 }
 125
 126 /*--------------------------------------.
 127 | Free all merge-function definitions.  |
 128 `--------------------------------------*/
 129
 130 void
 131 free_merger_functions (void)
 132 {
 133   merger_list *L0 = merge_functions;
 134   while (L0)
 135     {
 136       merger_list *L1 = L0->next;
 137       free (L0);
 138       L0 = L1;
 139     }
 140 }
 141
 142 \f
 143 /*-------------------------------------------------------------------.
 144 | Parse the input grammar into a one symbol_list structure.  Each    |
 145 | rule is represented by a sequence of symbols: the left hand side   |
 146 | followed by the contents of the right hand side, followed by a     |
 147 | null pointer instead of a symbol to terminate the rule.  The next  |
 148 | symbol is the lhs of the following rule.                           |
 149 |                                                                    |
 150 | All actions are copied out, labelled by the rule number they apply |
 151 | to.                                                                |
 152 `-------------------------------------------------------------------*/
 153
 154 /* The (currently) last symbol of GRAMMAR. */
 155 static symbol_list *grammar_end = NULL;
 156
 157 /* Append SYM to the grammar.  */
 158 static void
 159 grammar_symbol_append (symbol *sym, location loc)
 160 {
 161   symbol_list *p = symbol_list_new (sym, loc);
 162
 163   if (grammar_end)
 164     grammar_end->next = p;
 165   else
 166     grammar = p;
 167
 168   grammar_end = p;
 169
 170   /* A null SYM stands for an end of rule; it is not an actual
 171      part of it.  */
 172   if (sym)
 173     ++nritems;
 174 }
 175
 176 /* The rule currently being defined, and the previous rule.
 177    CURRENT_RULE points to the first LHS of the current rule, while
 178    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 179 symbol_list *current_rule = NULL;
 180 static symbol_list *previous_rule_end = NULL;
 181
 182
 183 /*----------------------------------------------.
 184 | Create a new rule for LHS in to the GRAMMAR.  |
 185 `----------------------------------------------*/
 186
 187 void
 188 grammar_current_rule_begin (symbol *lhs, location loc)
 189 {
 190   if (!start_flag)
 191     {
 192       startsymbol = lhs;
 193       startsymbol_location = loc;
 194       start_flag = true;
 195     }
 196
 197   /* Start a new rule and record its lhs.  */
 198   ++nrules;
 199   previous_rule_end = grammar_end;
 200   grammar_symbol_append (lhs, loc);
 201   current_rule = grammar_end;
 202
 203   /* Mark the rule's lhs as a nonterminal if not already so.  */
 204   if (lhs->class == unknown_sym)
 205     {
 206       lhs->class = nterm_sym;
 207       lhs->number = nvars;
 208       ++nvars;
 209     }
 210   else if (lhs->class == token_sym)
 211     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 212 }
 213
 214
 215 /*-----------------------------------------------------------------.
 216 | A symbol is typed if it has a declared %type, or if it is a      |
 217 | mid-rule symbol (i.e., the generated LHS replacing a mid-rule    |
 218 | action) that was assigned to, as in `exp: { $$ = 1; } { $$ = $1; |
 219 | }'.                                                              |
 220 `-----------------------------------------------------------------*/
 221
 222 static bool
 223 symbol_typed_p (const symbol_list *s)
 224 {
 225   return (s->sym->type_name
 226           || (s->midrule && s->midrule->used));
 227 }
 228
 229 /*----------------------------------------------------------------.
 230 | Check that the rule R is properly defined.  For instance, there |
 231 | should be no type clash on the default action.                  |
 232 `----------------------------------------------------------------*/
 233
 234 static void
 235 grammar_rule_check (const symbol_list *r)
 236 {
 237   /* Type check.
 238
 239      If there is an action, then there is nothing we can do: the user
 240      is allowed to shoot herself in the foot.
 241
 242      Don't worry about the default action if $$ is untyped, since $$'s
 243      value can't be used.  */
 244   if (!r->action && r->sym->type_name)
 245     {
 246       symbol *first_rhs = r->next->sym;
 247       /* If $$ is being set in default way, report if any type mismatch.  */
 248       if (first_rhs)
 249         {
 250           char const *lhs_type = r->sym->type_name;
 251           const char *rhs_type =
 252             first_rhs->type_name ? first_rhs->type_name : "";
 253           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 254             warn_at (r->location,
 255                      _("type clash on default action: <%s> != <%s>"),
 256                      lhs_type, rhs_type);
 257         }
 258       /* Warn if there is no default for $$ but we need one.  */
 259       else
 260         warn_at (r->location,
 261                  _("empty rule for typed nonterminal, and no action"));
 262     }
 263
 264   /* Check that typed symbol values are used.  */
 265   {
 266     symbol_list const *l = r;
 267     int n = 0;
 268     for (; l && l->sym; l = l->next, ++n)
 269       if (! (l->used
 270              || !symbol_typed_p (l)
 271              /* The default action, $$ = $1, `uses' both.  */
 272              || (!r->action && (n == 0 || n == 1))))
 273         {
 274           if (n)
 275             warn_at (r->location, _("unused value: $%d"), n);
 276           else
 277             warn_at (r->location, _("unset value: $$"));
 278         }
 279   }
 280 }
 281
 282
 283 /*-------------------------------------.
 284 | End the currently being grown rule.  |
 285 `-------------------------------------*/
 286
 287 void
 288 grammar_current_rule_end (location loc)
 289 {
 290   /* Put an empty link in the list to mark the end of this rule  */
 291   grammar_symbol_append (NULL, grammar_end->location);
 292   current_rule->location = loc;
 293   grammar_rule_check (current_rule);
 294 }
 295
 296
 297 /*-------------------------------------------------------------------.
 298 | The previous action turns out the be a mid-rule action.  Attach it |
 299 | to the current rule, i.e., create a dummy symbol, attach it this   |
 300 | mid-rule action, and append this dummy nonterminal to the current  |
 301 | rule.                                                              |
 302 `-------------------------------------------------------------------*/
 303
 304 static void
 305 grammar_midrule_action (void)
 306 {
 307   /* Since the action was written out with this rule's number, we must
 308      give the new rule this number by inserting the new rule before
 309      it.  */
 310
 311   /* Make a DUMMY nonterminal, whose location is that of the midrule
 312      action.  Create the MIDRULE.  */
 313   location dummy_location = current_rule->action_location;
 314   symbol *dummy = dummy_symbol_get (dummy_location);
 315   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 316
 317   /* Make a new rule, whose body is empty, before the current one, so
 318      that the action just read can belong to it.  */
 319   ++nrules;
 320   ++nritems;
 321   /* Attach its location and actions to that of the DUMMY.  */
 322   midrule->location = dummy_location;
 323   midrule->action = current_rule->action;
 324   midrule->action_location = dummy_location;
 325   current_rule->action = NULL;
 326   /* If $$ was used in the action, the LHS of the enclosing rule was
 327      incorrectly flagged as used.  */
 328   midrule->used = current_rule->used;
 329   current_rule->used = false;
 330
 331   if (previous_rule_end)
 332     previous_rule_end->next = midrule;
 333   else
 334     grammar = midrule;
 335
 336   /* End the dummy's rule.  */
 337   midrule->next = symbol_list_new (NULL, dummy_location);
 338   grammar_rule_check (midrule);
 339   midrule->next->next = current_rule;
 340
 341   previous_rule_end = midrule->next;
 342
 343   /* Insert the dummy nonterminal replacing the midrule action into
 344      the current rule.  Bind it to its dedicated rule.  */
 345   grammar_current_rule_symbol_append (dummy, dummy_location);
 346   grammar_end->midrule = midrule;
 347 }
 348
 349 /* Set the precedence symbol of the current rule to PRECSYM. */
 350
 351 void
 352 grammar_current_rule_prec_set (symbol *precsym, location loc)
 353 {
 354   if (current_rule->ruleprec)
 355     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 356   current_rule->ruleprec = precsym;
 357 }
 358
 359 /* Attach dynamic precedence DPREC to the current rule. */
 360
 361 void
 362 grammar_current_rule_dprec_set (int dprec, location loc)
 363 {
 364   if (! glr_parser)
 365     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 366   if (dprec <= 0)
 367     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 368   else if (current_rule->dprec != 0)
 369     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 370   current_rule->dprec = dprec;
 371 }
 372
 373 /* Attach a merge function NAME with argument type TYPE to current
 374    rule. */
 375
 376 void
 377 grammar_current_rule_merge_set (uniqstr name, location loc)
 378 {
 379   if (! glr_parser)
 380     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 381   if (current_rule->merger != 0)
 382     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 383   current_rule->merger =
 384     get_merge_function (name, current_rule->sym->type_name, loc);
 385 }
 386
 387 /* Attach SYM to the current rule.  If needed, move the previous
 388    action as a mid-rule action.  */
 389
 390 void
 391 grammar_current_rule_symbol_append (symbol *sym, location loc)
 392 {
 393   if (current_rule->action)
 394     grammar_midrule_action ();
 395   grammar_symbol_append (sym, loc);
 396 }
 397
 398 /* Attach an ACTION to the current rule.  If needed, move the previous
 399    action as a mid-rule action.  */
 400
 401 void
 402 grammar_current_rule_action_append (const char *action, location loc)
 403 {
 404   if (current_rule->action)
 405     grammar_midrule_action ();
 406   current_rule->action = action;
 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       symbol *ruleprec = p->ruleprec;
 433       rules[ruleno].user_number = ruleno;
 434       rules[ruleno].number = ruleno;
 435       rules[ruleno].lhs = p->sym;
 436       rules[ruleno].rhs = ritem + itemno;
 437       rules[ruleno].prec = NULL;
 438       rules[ruleno].dprec = p->dprec;
 439       rules[ruleno].merger = p->merger;
 440       rules[ruleno].precsym = NULL;
 441       rules[ruleno].location = p->location;
 442       rules[ruleno].useful = true;
 443       rules[ruleno].action = p->action;
 444       rules[ruleno].action_location = p->action_location;
 445
 446       p = p->next;
 447       while (p && p->sym)
 448         {
 449           /* item_number = symbol_number.
 450              But the former needs to contain more: negative rule numbers. */
 451           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 452           /* A rule gets by default the precedence and associativity
 453              of the last token in it.  */
 454           if (p->sym->class == token_sym && default_prec)
 455             rules[ruleno].prec = p->sym;
 456           if (p)
 457             p = p->next;
 458         }
 459
 460       /* If this rule has a %prec,
 461          the specified symbol's precedence replaces the default.  */
 462       if (ruleprec)
 463         {
 464           rules[ruleno].precsym = ruleprec;
 465           rules[ruleno].prec = ruleprec;
 466         }
 467       ritem[itemno++] = rule_number_as_item_number (ruleno);
 468       ++ruleno;
 469
 470       if (p)
 471         p = p->next;
 472     }
 473
 474   assert (itemno == nritems);
 475
 476   if (trace_flag & trace_sets)
 477     ritem_print (stderr);
 478 }
 479 \f
 480 /*------------------------------------------------------------------.
 481 | Read in the grammar specification and record it in the format     |
 482 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 483 | in each case forming the body of a C function (YYACTION) which    |
 484 | contains a switch statement to decide which action to execute.    |
 485 `------------------------------------------------------------------*/
 486
 487 void
 488 reader (void)
 489 {
 490   /* Initialize the symbol table.  */
 491   symbols_new ();
 492
 493   /* Construct the accept symbol. */
 494   accept = symbol_get ("$accept", empty_location);
 495   accept->class = nterm_sym;
 496   accept->number = nvars++;
 497
 498   /* Construct the error token */
 499   errtoken = symbol_get ("error", empty_location);
 500   errtoken->class = token_sym;
 501   errtoken->number = ntokens++;
 502
 503   /* Construct a token that represents all undefined literal tokens.
 504      It is always token number 2.  */
 505   undeftoken = symbol_get ("$undefined", empty_location);
 506   undeftoken->class = token_sym;
 507   undeftoken->number = ntokens++;
 508
 509   /* Initialize the obstacks. */
 510   obstack_init (&pre_prologue_obstack);
 511   obstack_init (&post_prologue_obstack);
 512
 513   gram_in = xfopen (grammar_file, "r");
 514
 515   gram__flex_debug = trace_flag & trace_scan;
 516   gram_debug = trace_flag & trace_parse;
 517   scanner_initialize ();
 518   gram_parse ();
 519
 520   if (! complaint_issued)
 521     check_and_convert_grammar ();
 522
 523   xfclose (gram_in);
 524 }
 525
 526
 527 /*-------------------------------------------------------------.
 528 | Check the grammar that has just been read, and convert it to |
 529 | internal form.                                               |
 530 `-------------------------------------------------------------*/
 531
 532 static void
 533 check_and_convert_grammar (void)
 534 {
 535   /* Grammar has been read.  Do some checking.  */
 536   if (nrules == 0)
 537     fatal (_("no rules in the input grammar"));
 538
 539   /* Report any undefined symbols and consider them nonterminals.  */
 540   symbols_check_defined ();
 541
 542   /* If the user did not define her ENDTOKEN, do it now. */
 543   if (!endtoken)
 544     {
 545       endtoken = symbol_get ("$end", empty_location);
 546       endtoken->class = token_sym;
 547       endtoken->number = 0;
 548       /* Value specified by POSIX.  */
 549       endtoken->user_token_number = 0;
 550     }
 551
 552   /* Insert the initial rule, whose line is that of the first rule
 553      (not that of the start symbol):
 554
 555      accept: %start EOF.  */
 556   {
 557     symbol_list *p = symbol_list_new (accept, empty_location);
 558     p->location = grammar->location;
 559     p->next = symbol_list_new (startsymbol, empty_location);
 560     p->next->next = symbol_list_new (endtoken, empty_location);
 561     p->next->next->next = symbol_list_new (NULL, empty_location);
 562     p->next->next->next->next = grammar;
 563     nrules += 1;
 564     nritems += 3;
 565     grammar = p;
 566   }
 567
 568   assert (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 569
 570   /* Assign the symbols their symbol numbers.  Write #defines for the
 571      token symbols into FDEFINES if requested.  */
 572   symbols_pack ();
 573
 574   /* Convert the grammar into the format described in gram.h.  */
 575   packgram ();
 576
 577   /* The grammar as a symbol_list is no longer needed. */
 578   LIST_FREE (symbol_list, grammar);
 579 }