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 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 static 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   /* A null SYM stands for an end of rule; it is not an actual
 169      part of it.  */
 170   if (sym)
 171     ++nritems;
 172 }
 173
 174 /* The rule currently being defined, and the previous rule.
 175    CURRENT_RULE points to the first LHS of the current rule, while
 176    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 177 symbol_list *current_rule = NULL;
 178 static symbol_list *previous_rule_end = NULL;
 179
 180
 181 /*----------------------------------------------.
 182 | Create a new rule for LHS in to the GRAMMAR.  |
 183 `----------------------------------------------*/
 184
 185 void
 186 grammar_current_rule_begin (symbol *lhs, location loc)
 187 {
 188   if (!start_flag)
 189     {
 190       startsymbol = lhs;
 191       startsymbol_location = loc;
 192       start_flag = true;
 193     }
 194
 195   /* Start a new rule and record its lhs.  */
 196   ++nrules;
 197   previous_rule_end = grammar_end;
 198   grammar_symbol_append (lhs, loc);
 199   current_rule = grammar_end;
 200
 201   /* Mark the rule's lhs as a nonterminal if not already so.  */
 202   if (lhs->class == unknown_sym)
 203     {
 204       lhs->class = nterm_sym;
 205       lhs->number = nvars;
 206       ++nvars;
 207     }
 208   else if (lhs->class == token_sym)
 209     complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
 210 }
 211
 212
 213 /*-----------------------------------------------------------------.
 214 | A symbol is typed if it has a declared %type, or if it is a      |
 215 | mid-rule symbol (i.e., the generated LHS replacing a mid-rule    |
 216 | action) that was assigned to, as in `exp: { $$ = 1; } { $$ = $1; |
 217 | }'.                                                              |
 218 `-----------------------------------------------------------------*/
 219
 220 static bool
 221 symbol_typed_p (const symbol_list *s)
 222 {
 223   return (s->sym->type_name
 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 typed symbol values are 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_typed_p (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 static 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.  If needed, move the previous
 397    action as a mid-rule action.  */
 398
 399 void
 400 grammar_current_rule_action_append (const char *action, location loc)
 401 {
 402   if (current_rule->action)
 403     grammar_midrule_action ();
 404   current_rule->action = action;
 405   current_rule->action_location = loc;
 406 }
 407
 408 \f
 409 /*---------------------------------------------------------------.
 410 | Convert the rules into the representation using RRHS, RLHS and |
 411 | RITEM.                                                         |
 412 `---------------------------------------------------------------*/
 413
 414 static void
 415 packgram (void)
 416 {
 417   unsigned int itemno = 0;
 418   rule_number ruleno = 0;
 419   symbol_list *p = grammar;
 420
 421   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 422
 423   /* This sentinel is used by build_relations in gram.c.  */
 424   *ritem++ = 0;
 425
 426   rules = xnmalloc (nrules, sizeof *rules);
 427
 428   while (p)
 429     {
 430       symbol *ruleprec = p->ruleprec;
 431       rules[ruleno].user_number = ruleno;
 432       rules[ruleno].number = ruleno;
 433       rules[ruleno].lhs = p->sym;
 434       rules[ruleno].rhs = ritem + itemno;
 435       rules[ruleno].prec = NULL;
 436       rules[ruleno].dprec = p->dprec;
 437       rules[ruleno].merger = p->merger;
 438       rules[ruleno].precsym = NULL;
 439       rules[ruleno].location = p->location;
 440       rules[ruleno].useful = true;
 441       rules[ruleno].action = p->action;
 442       rules[ruleno].action_location = p->action_location;
 443
 444       p = p->next;
 445       while (p && p->sym)
 446         {
 447           /* item_number = symbol_number.
 448              But the former needs to contain more: negative rule numbers. */
 449           ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
 450           /* A rule gets by default the precedence and associativity
 451              of the last token in it.  */
 452           if (p->sym->class == token_sym && default_prec)
 453             rules[ruleno].prec = p->sym;
 454           if (p)
 455             p = p->next;
 456         }
 457
 458       /* If this rule has a %prec,
 459          the specified symbol's precedence replaces the default.  */
 460       if (ruleprec)
 461         {
 462           rules[ruleno].precsym = ruleprec;
 463           rules[ruleno].prec = ruleprec;
 464         }
 465       ritem[itemno++] = rule_number_as_item_number (ruleno);
 466       ++ruleno;
 467
 468       if (p)
 469         p = p->next;
 470     }
 471
 472   if (itemno != nritems)
 473     abort ();
 474
 475   if (trace_flag & trace_sets)
 476     ritem_print (stderr);
 477 }
 478 \f
 479 /*------------------------------------------------------------------.
 480 | Read in the grammar specification and record it in the format     |
 481 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 482 | in each case forming the body of a C function (YYACTION) which    |
 483 | contains a switch statement to decide which action to execute.    |
 484 `------------------------------------------------------------------*/
 485
 486 void
 487 reader (void)
 488 {
 489   /* Initialize the symbol table.  */
 490   symbols_new ();
 491
 492   /* Construct the accept symbol. */
 493   accept = symbol_get ("$accept", empty_location);
 494   accept->class = nterm_sym;
 495   accept->number = nvars++;
 496
 497   /* Construct the error token */
 498   errtoken = symbol_get ("error", empty_location);
 499   errtoken->class = token_sym;
 500   errtoken->number = ntokens++;
 501
 502   /* Construct a token that represents all undefined literal tokens.
 503      It is always token number 2.  */
 504   undeftoken = symbol_get ("$undefined", empty_location);
 505   undeftoken->class = token_sym;
 506   undeftoken->number = ntokens++;
 507
 508   /* Initialize the obstacks. */
 509   obstack_init (&pre_prologue_obstack);
 510   obstack_init (&post_prologue_obstack);
 511
 512   gram_in = xfopen (grammar_file, "r");
 513
 514   gram__flex_debug = trace_flag & trace_scan;
 515   gram_debug = trace_flag & trace_parse;
 516   scanner_initialize ();
 517   gram_parse ();
 518
 519   /* If something went wrong during the parsing, don't try to
 520      continue.  */
 521   if (complaint_issued)
 522     return;
 523
 524   /* Grammar has been read.  Do some checking */
 525   if (nrules == 0)
 526     fatal (_("no rules in the input grammar"));
 527
 528   /* Report any undefined symbols and consider them nonterminals.  */
 529   symbols_check_defined ();
 530
 531   /* If the user did not define her ENDTOKEN, do it now. */
 532   if (!endtoken)
 533     {
 534       endtoken = symbol_get ("$end", empty_location);
 535       endtoken->class = token_sym;
 536       endtoken->number = 0;
 537       /* Value specified by POSIX.  */
 538       endtoken->user_token_number = 0;
 539     }
 540
 541   /* Insert the initial rule, which line is that of the first rule
 542      (not that of the start symbol):
 543
 544      accept: %start EOF.  */
 545   {
 546     symbol_list *p = symbol_list_new (accept, empty_location);
 547     p->location = grammar->location;
 548     p->next = symbol_list_new (startsymbol, empty_location);
 549     p->next->next = symbol_list_new (endtoken, empty_location);
 550     p->next->next->next = symbol_list_new (NULL, empty_location);
 551     p->next->next->next->next = grammar;
 552     nrules += 1;
 553     nritems += 3;
 554     grammar = p;
 555   }
 556
 557   if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
 558     abort ();
 559
 560   xfclose (gram_in);
 561
 562   /* Assign the symbols their symbol numbers.  Write #defines for the
 563      token symbols into FDEFINES if requested.  */
 564   symbols_pack ();
 565
 566   /* Convert the grammar into the format described in gram.h.  */
 567   packgram ();
 568
 569   /* The grammar as a symbol_list is no longer needed. */
 570   LIST_FREE (symbol_list, grammar);
 571 }