src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003,
   4    2005 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 symbol_list *grammar = NULL;
  40 static bool start_flag = false;
  41 merger_list *merge_functions;
  42
  43 /* Was %union seen?  */
  44 bool typed = false;
  45
  46 /* Should rules have a default precedence?  */
  47 bool default_prec = true;
  48 \f
  49 /*-----------------------.
  50 | Set the start symbol.  |
  51 `-----------------------*/
  52
  53 void
  54 grammar_start_symbol_set (symbol *sym, location loc)
  55 {
  56   if (start_flag)
  57     complain_at (loc, _("multiple %s declarations"), "%start");
  58   else
  59     {
  60       start_flag = true;
  61       startsymbol = sym;
  62       startsymbol_location = loc;
  63     }
  64 }
  65
  66
  67 /*----------------------------------------------------------------.
  68 | There are two prologues: one before %union, one after.  Augment |
  69 | the current one.                                                |
  70 `----------------------------------------------------------------*/
  71
  72 void
  73 prologue_augment (const char *prologue, location loc)
  74 {
  75   struct obstack *oout =
  76     !typed ? &pre_prologue_obstack : &post_prologue_obstack;
  77
  78   obstack_fgrow1 (oout, "]b4_syncline(%d, [[", loc.start.line);
  79   MUSCLE_OBSTACK_SGROW (oout,
  80                         quotearg_style (c_quoting_style, loc.start.file));
  81   obstack_sgrow (oout, "]])[\n");
  82   obstack_sgrow (oout, prologue);
  83 }
  84
  85 \f
  86
  87 /*-------------------------------------------------------------------.
  88 | Return the merger index for a merging function named NAME, whose   |
  89 | arguments have type TYPE.  Records the function, if new, in        |
  90 | MERGER_LIST.                                                       |
  91 `-------------------------------------------------------------------*/
  92
  93 static int
  94 get_merge_function (uniqstr name, uniqstr type, location loc)
  95 {
  96   merger_list *syms;
  97   merger_list head;
  98   int n;
  99
 100   if (! glr_parser)
 101     return 0;
 102
 103   if (type == NULL)
 104     type = uniqstr_new ("");
 105
 106   head.next = merge_functions;
 107   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
 108     if (UNIQSTR_EQ (name, syms->next->name))
 109       break;
 110   if (syms->next == NULL)
 111     {
 112       syms->next = xmalloc (sizeof syms->next[0]);
 113       syms->next->name = uniqstr_new (name);
 114       syms->next->type = uniqstr_new (type);
 115       syms->next->next = NULL;
 116       merge_functions = head.next;
 117     }
 118   else if (!UNIQSTR_EQ (type, syms->next->type))
 119     warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
 120              name, type, syms->next->type);
 121   return n;
 122 }
 123
 124 /*--------------------------------------.
 125 | Free all merge-function definitions.  |
 126 `--------------------------------------*/
 127
 128 void
 129 free_merger_functions (void)
 130 {
 131   merger_list *L0 = merge_functions;
 132   while (L0)
 133     {
 134       merger_list *L1 = L0->next;
 135       free (L0);
 136       L0 = L1;
 137     }
 138 }
 139
 140 \f
 141 /*-------------------------------------------------------------------.
 142 | Parse the input grammar into a one symbol_list structure.  Each    |
 143 | rule is represented by a sequence of symbols: the left hand side   |
 144 | followed by the contents of the right hand side, followed by a     |
 145 | null pointer instead of a symbol to terminate the rule.  The next  |
 146 | symbol is the lhs of the following rule.                           |
 147 |                                                                    |
 148 | All actions are copied out, labelled by the rule number they apply |
 149 | to.                                                                |
 150 `-------------------------------------------------------------------*/
 151
 152 /* The (currently) last symbol of GRAMMAR. */
 153 static symbol_list *grammar_end = NULL;
 154
 155 /* Append SYM to the grammar.  */
 156 void
 157 grammar_symbol_append (symbol *sym, location loc)
 158 {
 159   symbol_list *p = symbol_list_new (sym, loc);
 160
 161   if (grammar_end)
 162     grammar_end->next = p;
 163   else
 164     grammar = p;
 165
 166   grammar_end = p;
 167 }
 168
 169 /* The rule currently being defined, and the previous rule.
 170    CURRENT_RULE points to the first LHS of the current rule, while
 171    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 172 symbol_list *current_rule = NULL;
 173 static symbol_list *previous_rule_end = NULL;
 174
 175
 176 /*----------------------------------------------.
 177 | Create a new rule for LHS in to the GRAMMAR.  |
 178 `----------------------------------------------*/
 179
 180 void
 181 grammar_rule_begin (symbol *lhs, location loc)
 182 {
 183   if (!start_flag)
 184     {
 185       startsymbol = lhs;
 186       startsymbol_location = loc;
 187       start_flag = true;
 188     }
 189
 190   /* Start a new rule and record its lhs.  */
 191   ++nrules;
 192   ++nritems;
 193
 194   previous_rule_end = grammar_end;
 195   grammar_symbol_append (lhs, loc);
 196   current_rule = grammar_end;
 197
 198   /* Mark the rule's lhs as a nonterminal if not already so.  */
 199   if (lhs->class == unknown_sym)
 200     {
 201       lhs->class = nterm_sym;
 202       lhs->number = nvars;
 203       ++nvars;
 204     }
 205   else if (lhs->class == token_sym)
 206     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 207 }
 208
 209
 210 /*------------------------------------------------------------------.
 211 | Check that the last rule (CURRENT_RULE) is properly defined.  For |
 212 | instance, there should be no type clash on the default action.    |
 213 `------------------------------------------------------------------*/
 214
 215 static void
 216 grammar_current_rule_check (void)
 217 {
 218   symbol *lhs = current_rule->sym;
 219   char const *lhs_type = lhs->type_name;
 220
 221   /* Type check.
 222
 223      If there is an action, then there is nothing we can do: the user
 224      is allowed to shoot herself in the foot.
 225
 226      Don't worry about the default action if $$ is untyped, since $$'s
 227      value can't be used.  */
 228   if (!current_rule->action && lhs_type)
 229     {
 230       symbol *first_rhs = current_rule->next->sym;
 231       /* If $$ is being set in default way, report if any type mismatch.  */
 232       if (first_rhs)
 233         {
 234           const char *rhs_type =
 235             first_rhs->type_name ? first_rhs->type_name : "";
 236           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 237             warn_at (current_rule->location,
 238                      _("type clash on default action: <%s> != <%s>"),
 239                      lhs_type, rhs_type);
 240         }
 241       /* Warn if there is no default for $$ but we need one.  */
 242       else
 243         warn_at (current_rule->location,
 244                  _("empty rule for typed nonterminal, and no action"));
 245     }
 246
 247   /* Check that all the symbol values are used. */
 248   if (typed)
 249     {
 250       symbol_list *l = current_rule;
 251       int n = 1;
 252       for (l = current_rule->next; l && l->sym; l = l->next, ++n)
 253         /* The default action `uses' $1. */
 254         if (! (!current_rule->action && n == 1)
 255             && l->sym->type_name && !l->used)
 256           warn_at (current_rule->location, _("unused value: $%d"), n);
 257     }
 258 }
 259
 260
 261 /*-------------------------------------.
 262 | End the currently being grown rule.  |
 263 `-------------------------------------*/
 264
 265 void
 266 grammar_rule_end (location loc)
 267 {
 268   /* Put an empty link in the list to mark the end of this rule  */
 269   grammar_symbol_append (NULL, grammar_end->location);
 270   current_rule->location = loc;
 271   grammar_current_rule_check ();
 272 }
 273
 274
 275 /*-------------------------------------------------------------------.
 276 | The previous action turns out the be a mid-rule action.  Attach it |
 277 | to the current rule, i.e., create a dummy symbol, attach it this   |
 278 | mid-rule action, and append this dummy nonterminal to the current  |
 279 | rule.                                                              |
 280 `-------------------------------------------------------------------*/
 281
 282 void
 283 grammar_midrule_action (void)
 284 {
 285   /* Since the action was written out with this rule's number, we must
 286      give the new rule this number by inserting the new rule before
 287      it.  */
 288
 289   /* Make a DUMMY nonterminal, whose location is that of the midrule
 290      action.  Create the MIDRULE.  */
 291   location dummy_location = current_rule->action_location;
 292   symbol *dummy = dummy_symbol_get (dummy_location);
 293   symbol_list *midrule = symbol_list_new (dummy, dummy_location);
 294
 295   /* Make a new rule, whose body is empty, before the current one, so
 296      that the action just read can belong to it.  */
 297   ++nrules;
 298   ++nritems;
 299   /* Attach its location and actions to that of the DUMMY.  */
 300   midrule->location = dummy_location;
 301   midrule->action = current_rule->action;
 302   midrule->action_location = dummy_location;
 303   current_rule->action = NULL;
 304
 305   if (previous_rule_end)
 306     previous_rule_end->next = midrule;
 307   else
 308     grammar = midrule;
 309
 310   /* End the dummy's rule.  */
 311   previous_rule_end = symbol_list_new (NULL, dummy_location);
 312   previous_rule_end->next = current_rule;
 313
 314   midrule->next = previous_rule_end;
 315
 316   /* Insert the dummy nonterminal replacing the midrule action into
 317      the current rule.  */
 318   grammar_current_rule_symbol_append (dummy, dummy_location);
 319 }
 320
 321 /* Set the precedence symbol of the current rule to PRECSYM. */
 322
 323 void
 324 grammar_current_rule_prec_set (symbol *precsym, location loc)
 325 {
 326   if (current_rule->ruleprec)
 327     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 328   current_rule->ruleprec = precsym;
 329 }
 330
 331 /* Attach dynamic precedence DPREC to the current rule. */
 332
 333 void
 334 grammar_current_rule_dprec_set (int dprec, location loc)
 335 {
 336   if (! glr_parser)
 337     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 338   if (dprec <= 0)
 339     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 340   else if (current_rule->dprec != 0)
 341     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 342   current_rule->dprec = dprec;
 343 }
 344
 345 /* Attach a merge function NAME with argument type TYPE to current
 346    rule. */
 347
 348 void
 349 grammar_current_rule_merge_set (uniqstr name, location loc)
 350 {
 351   if (! glr_parser)
 352     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 353   if (current_rule->merger != 0)
 354     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 355   current_rule->merger =
 356     get_merge_function (name, current_rule->sym->type_name, loc);
 357 }
 358
 359 /* Attach SYM to the current rule.  If needed, move the previous
 360    action as a mid-rule action.  */
 361
 362 void
 363 grammar_current_rule_symbol_append (symbol *sym, location loc)
 364 {
 365   if (current_rule->action)
 366     grammar_midrule_action ();
 367   ++nritems;
 368   grammar_symbol_append (sym, loc);
 369 }
 370
 371 /* Attach an ACTION to the current rule.  If needed, move the previous
 372    action as a mid-rule action.  */
 373
 374 void
 375 grammar_current_rule_action_append (const char *action, location loc)
 376 {
 377   if (current_rule->action)
 378     grammar_midrule_action ();
 379   current_rule->action = action;
 380   current_rule->action_location = loc;
 381 }
 382
 383 \f
 384 /*---------------------------------------------------------------.
 385 | Convert the rules into the representation using RRHS, RLHS and |
 386 | RITEM.                                                         |
 387 `---------------------------------------------------------------*/
 388
 389 static void
 390 packgram (void)
 391 {
 392   unsigned int itemno = 0;
 393   rule_number ruleno = 0;
 394   symbol_list *p = grammar;
 395
 396   ritem = xnmalloc (nritems, sizeof *ritem);
 397   rules = xnmalloc (nrules, sizeof *rules);
 398
 399   while (p)
 400     {
 401       symbol *ruleprec = p->ruleprec;
 402       rules[ruleno].user_number = ruleno;
 403       rules[ruleno].number = ruleno;
 404       rules[ruleno].lhs = p->sym;
 405       rules[ruleno].rhs = ritem + itemno;
 406       rules[ruleno].prec = NULL;
 407       rules[ruleno].dprec = p->dprec;
 408       rules[ruleno].merger = p->merger;
 409       rules[ruleno].precsym = NULL;
 410       rules[ruleno].location = p->location;
 411       rules[ruleno].useful = true;
 412       rules[ruleno].action = p->action;
 413       rules[ruleno].action_location = p->action_location;
 414
 415       p = p->next;
 416       while (p && p->sym)
 417         {
 418           /* item_number = symbol_number.
 419              But the former needs to contain more: negative rule numbers. */
 420           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 421           /* A rule gets by default the precedence and associativity
 422              of the last token in it.  */
 423           if (p->sym->class == token_sym && default_prec)
 424             rules[ruleno].prec = p->sym;
 425           if (p)
 426             p = p->next;
 427         }
 428
 429       /* If this rule has a %prec,
 430          the specified symbol's precedence replaces the default.  */
 431       if (ruleprec)
 432         {
 433           rules[ruleno].precsym = ruleprec;
 434           rules[ruleno].prec = ruleprec;
 435         }
 436       ritem[itemno++] = rule_number_as_item_number (ruleno);
 437       ++ruleno;
 438
 439       if (p)
 440         p = p->next;
 441     }
 442
 443   if (itemno != nritems)
 444     abort ();
 445
 446   if (trace_flag & trace_sets)
 447     ritem_print (stderr);
 448 }
 449 \f
 450 /*------------------------------------------------------------------.
 451 | Read in the grammar specification and record it in the format     |
 452 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 453 | in each case forming the body of a C function (YYACTION) which    |
 454 | contains a switch statement to decide which action to execute.    |
 455 `------------------------------------------------------------------*/
 456
 457 void
 458 reader (void)
 459 {
 460   /* Initialize the symbol table.  */
 461   symbols_new ();
 462
 463   /* Construct the accept symbol. */
 464   accept = symbol_get ("$accept", empty_location);
 465   accept->class = nterm_sym;
 466   accept->number = nvars++;
 467
 468   /* Construct the error token */
 469   errtoken = symbol_get ("error", empty_location);
 470   errtoken->class = token_sym;
 471   errtoken->number = ntokens++;
 472
 473   /* Construct a token that represents all undefined literal tokens.
 474      It is always token number 2.  */
 475   undeftoken = symbol_get ("$undefined", empty_location);
 476   undeftoken->class = token_sym;
 477   undeftoken->number = ntokens++;
 478
 479   /* Initialize the obstacks. */
 480   obstack_init (&pre_prologue_obstack);
 481   obstack_init (&post_prologue_obstack);
 482
 483   gram_in = xfopen (grammar_file, "r");
 484
 485   gram__flex_debug = trace_flag & trace_scan;
 486   gram_debug = trace_flag & trace_parse;
 487   scanner_initialize ();
 488   gram_parse ();
 489
 490   /* If something went wrong during the parsing, don't try to
 491      continue.  */
 492   if (complaint_issued)
 493     return;
 494
 495   /* Grammar has been read.  Do some checking */
 496   if (nrules == 0)
 497     fatal (_("no rules in the input grammar"));
 498
 499   /* Report any undefined symbols and consider them nonterminals.  */
 500   symbols_check_defined ();
 501
 502   /* If the user did not define her ENDTOKEN, do it now. */
 503   if (!endtoken)
 504     {
 505       endtoken = symbol_get ("$end", empty_location);
 506       endtoken->class = token_sym;
 507       endtoken->number = 0;
 508       /* Value specified by POSIX.  */
 509       endtoken->user_token_number = 0;
 510     }
 511
 512   /* Insert the initial rule, which line is that of the first rule
 513      (not that of the start symbol):
 514
 515      accept: %start EOF.  */
 516   {
 517     symbol_list *p = symbol_list_new (accept, empty_location);
 518     p->location = grammar->location;
 519     p->next = symbol_list_new (startsymbol, empty_location);
 520     p->next->next = symbol_list_new (endtoken, empty_location);
 521     p->next->next->next = symbol_list_new (NULL, empty_location);
 522     p->next->next->next->next = grammar;
 523     nrules += 1;
 524     nritems += 3;
 525     grammar = p;
 526   }
 527
 528   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 529     abort ();
 530
 531   xfclose (gram_in);
 532
 533   /* Assign the symbols their symbol numbers.  Write #defines for the
 534      token symbols into FDEFINES if requested.  */
 535   symbols_pack ();
 536
 537   /* Convert the grammar into the format described in gram.h.  */
 538   packgram ();
 539
 540   /* The grammar as a symbol_list is no longer needed. */
 541   LIST_FREE (symbol_list, grammar);
 542 }