| OLD | NEW |
| (Empty) |
| 1 /* | |
| 2 ** This file contains all sources (including headers) to the LEMON | |
| 3 ** LALR(1) parser generator. The sources have been combined into a | |
| 4 ** single file to make it easy to include LEMON in the source tree | |
| 5 ** and Makefile of another program. | |
| 6 ** | |
| 7 ** The author of this program disclaims copyright. | |
| 8 */ | |
| 9 #include <stdio.h> | |
| 10 #include <stdarg.h> | |
| 11 #include <string.h> | |
| 12 #include <ctype.h> | |
| 13 #include <stdlib.h> | |
| 14 #include <assert.h> | |
| 15 | |
| 16 #define ISSPACE(X) isspace((unsigned char)(X)) | |
| 17 #define ISDIGIT(X) isdigit((unsigned char)(X)) | |
| 18 #define ISALNUM(X) isalnum((unsigned char)(X)) | |
| 19 #define ISALPHA(X) isalpha((unsigned char)(X)) | |
| 20 #define ISUPPER(X) isupper((unsigned char)(X)) | |
| 21 #define ISLOWER(X) islower((unsigned char)(X)) | |
| 22 | |
| 23 | |
| 24 #ifndef __WIN32__ | |
| 25 # if defined(_WIN32) || defined(WIN32) | |
| 26 # define __WIN32__ | |
| 27 # endif | |
| 28 #endif | |
| 29 | |
| 30 #ifdef __WIN32__ | |
| 31 #ifdef __cplusplus | |
| 32 extern "C" { | |
| 33 #endif | |
| 34 extern int access(const char *path, int mode); | |
| 35 #ifdef __cplusplus | |
| 36 } | |
| 37 #endif | |
| 38 #else | |
| 39 #include <unistd.h> | |
| 40 #endif | |
| 41 | |
| 42 /* #define PRIVATE static */ | |
| 43 #define PRIVATE | |
| 44 | |
| 45 #ifdef TEST | |
| 46 #define MAXRHS 5 /* Set low to exercise exception code */ | |
| 47 #else | |
| 48 #define MAXRHS 1000 | |
| 49 #endif | |
| 50 | |
| 51 static int showPrecedenceConflict = 0; | |
| 52 static char *msort(char*,char**,int(*)(const char*,const char*)); | |
| 53 | |
| 54 /* | |
| 55 ** Compilers are getting increasingly pedantic about type conversions | |
| 56 ** as C evolves ever closer to Ada.... To work around the latest problems | |
| 57 ** we have to define the following variant of strlen(). | |
| 58 */ | |
| 59 #define lemonStrlen(X) ((int)strlen(X)) | |
| 60 | |
| 61 /* | |
| 62 ** Compilers are starting to complain about the use of sprintf() and strcpy(), | |
| 63 ** saying they are unsafe. So we define our own versions of those routines too. | |
| 64 ** | |
| 65 ** There are three routines here: lemon_sprintf(), lemon_vsprintf(), and | |
| 66 ** lemon_addtext(). The first two are replacements for sprintf() and vsprintf(). | |
| 67 ** The third is a helper routine for vsnprintf() that adds texts to the end of a | |
| 68 ** buffer, making sure the buffer is always zero-terminated. | |
| 69 ** | |
| 70 ** The string formatter is a minimal subset of stdlib sprintf() supporting only | |
| 71 ** a few simply conversions: | |
| 72 ** | |
| 73 ** %d | |
| 74 ** %s | |
| 75 ** %.*s | |
| 76 ** | |
| 77 */ | |
| 78 static void lemon_addtext( | |
| 79 char *zBuf, /* The buffer to which text is added */ | |
| 80 int *pnUsed, /* Slots of the buffer used so far */ | |
| 81 const char *zIn, /* Text to add */ | |
| 82 int nIn, /* Bytes of text to add. -1 to use strlen() */ | |
| 83 int iWidth /* Field width. Negative to left justify */ | |
| 84 ){ | |
| 85 if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){} | |
| 86 while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; } | |
| 87 if( nIn==0 ) return; | |
| 88 memcpy(&zBuf[*pnUsed], zIn, nIn); | |
| 89 *pnUsed += nIn; | |
| 90 while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; } | |
| 91 zBuf[*pnUsed] = 0; | |
| 92 } | |
| 93 static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){ | |
| 94 int i, j, k, c; | |
| 95 int nUsed = 0; | |
| 96 const char *z; | |
| 97 char zTemp[50]; | |
| 98 str[0] = 0; | |
| 99 for(i=j=0; (c = zFormat[i])!=0; i++){ | |
| 100 if( c=='%' ){ | |
| 101 int iWidth = 0; | |
| 102 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0); | |
| 103 c = zFormat[++i]; | |
| 104 if( ISDIGIT(c) || (c=='-' && ISDIGIT(zFormat[i+1])) ){ | |
| 105 if( c=='-' ) i++; | |
| 106 while( ISDIGIT(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0'; | |
| 107 if( c=='-' ) iWidth = -iWidth; | |
| 108 c = zFormat[i]; | |
| 109 } | |
| 110 if( c=='d' ){ | |
| 111 int v = va_arg(ap, int); | |
| 112 if( v<0 ){ | |
| 113 lemon_addtext(str, &nUsed, "-", 1, iWidth); | |
| 114 v = -v; | |
| 115 }else if( v==0 ){ | |
| 116 lemon_addtext(str, &nUsed, "0", 1, iWidth); | |
| 117 } | |
| 118 k = 0; | |
| 119 while( v>0 ){ | |
| 120 k++; | |
| 121 zTemp[sizeof(zTemp)-k] = (v%10) + '0'; | |
| 122 v /= 10; | |
| 123 } | |
| 124 lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth); | |
| 125 }else if( c=='s' ){ | |
| 126 z = va_arg(ap, const char*); | |
| 127 lemon_addtext(str, &nUsed, z, -1, iWidth); | |
| 128 }else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){ | |
| 129 i += 2; | |
| 130 k = va_arg(ap, int); | |
| 131 z = va_arg(ap, const char*); | |
| 132 lemon_addtext(str, &nUsed, z, k, iWidth); | |
| 133 }else if( c=='%' ){ | |
| 134 lemon_addtext(str, &nUsed, "%", 1, 0); | |
| 135 }else{ | |
| 136 fprintf(stderr, "illegal format\n"); | |
| 137 exit(1); | |
| 138 } | |
| 139 j = i+1; | |
| 140 } | |
| 141 } | |
| 142 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0); | |
| 143 return nUsed; | |
| 144 } | |
| 145 static int lemon_sprintf(char *str, const char *format, ...){ | |
| 146 va_list ap; | |
| 147 int rc; | |
| 148 va_start(ap, format); | |
| 149 rc = lemon_vsprintf(str, format, ap); | |
| 150 va_end(ap); | |
| 151 return rc; | |
| 152 } | |
| 153 static void lemon_strcpy(char *dest, const char *src){ | |
| 154 while( (*(dest++) = *(src++))!=0 ){} | |
| 155 } | |
| 156 static void lemon_strcat(char *dest, const char *src){ | |
| 157 while( *dest ) dest++; | |
| 158 lemon_strcpy(dest, src); | |
| 159 } | |
| 160 | |
| 161 | |
| 162 /* a few forward declarations... */ | |
| 163 struct rule; | |
| 164 struct lemon; | |
| 165 struct action; | |
| 166 | |
| 167 static struct action *Action_new(void); | |
| 168 static struct action *Action_sort(struct action *); | |
| 169 | |
| 170 /********** From the file "build.h" ************************************/ | |
| 171 void FindRulePrecedences(); | |
| 172 void FindFirstSets(); | |
| 173 void FindStates(); | |
| 174 void FindLinks(); | |
| 175 void FindFollowSets(); | |
| 176 void FindActions(); | |
| 177 | |
| 178 /********* From the file "configlist.h" *********************************/ | |
| 179 void Configlist_init(void); | |
| 180 struct config *Configlist_add(struct rule *, int); | |
| 181 struct config *Configlist_addbasis(struct rule *, int); | |
| 182 void Configlist_closure(struct lemon *); | |
| 183 void Configlist_sort(void); | |
| 184 void Configlist_sortbasis(void); | |
| 185 struct config *Configlist_return(void); | |
| 186 struct config *Configlist_basis(void); | |
| 187 void Configlist_eat(struct config *); | |
| 188 void Configlist_reset(void); | |
| 189 | |
| 190 /********* From the file "error.h" ***************************************/ | |
| 191 void ErrorMsg(const char *, int,const char *, ...); | |
| 192 | |
| 193 /****** From the file "option.h" ******************************************/ | |
| 194 enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR, | |
| 195 OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR}; | |
| 196 struct s_options { | |
| 197 enum option_type type; | |
| 198 const char *label; | |
| 199 char *arg; | |
| 200 const char *message; | |
| 201 }; | |
| 202 int OptInit(char**,struct s_options*,FILE*); | |
| 203 int OptNArgs(void); | |
| 204 char *OptArg(int); | |
| 205 void OptErr(int); | |
| 206 void OptPrint(void); | |
| 207 | |
| 208 /******** From the file "parse.h" *****************************************/ | |
| 209 void Parse(struct lemon *lemp); | |
| 210 | |
| 211 /********* From the file "plink.h" ***************************************/ | |
| 212 struct plink *Plink_new(void); | |
| 213 void Plink_add(struct plink **, struct config *); | |
| 214 void Plink_copy(struct plink **, struct plink *); | |
| 215 void Plink_delete(struct plink *); | |
| 216 | |
| 217 /********** From the file "report.h" *************************************/ | |
| 218 void Reprint(struct lemon *); | |
| 219 void ReportOutput(struct lemon *); | |
| 220 void ReportTable(struct lemon *, int); | |
| 221 void ReportHeader(struct lemon *); | |
| 222 void CompressTables(struct lemon *); | |
| 223 void ResortStates(struct lemon *); | |
| 224 | |
| 225 /********** From the file "set.h" ****************************************/ | |
| 226 void SetSize(int); /* All sets will be of size N */ | |
| 227 char *SetNew(void); /* A new set for element 0..N */ | |
| 228 void SetFree(char*); /* Deallocate a set */ | |
| 229 int SetAdd(char*,int); /* Add element to a set */ | |
| 230 int SetUnion(char *,char *); /* A <- A U B, thru element N */ | |
| 231 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */ | |
| 232 | |
| 233 /********** From the file "struct.h" *************************************/ | |
| 234 /* | |
| 235 ** Principal data structures for the LEMON parser generator. | |
| 236 */ | |
| 237 | |
| 238 typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean; | |
| 239 | |
| 240 /* Symbols (terminals and nonterminals) of the grammar are stored | |
| 241 ** in the following: */ | |
| 242 enum symbol_type { | |
| 243 TERMINAL, | |
| 244 NONTERMINAL, | |
| 245 MULTITERMINAL | |
| 246 }; | |
| 247 enum e_assoc { | |
| 248 LEFT, | |
| 249 RIGHT, | |
| 250 NONE, | |
| 251 UNK | |
| 252 }; | |
| 253 struct symbol { | |
| 254 const char *name; /* Name of the symbol */ | |
| 255 int index; /* Index number for this symbol */ | |
| 256 enum symbol_type type; /* Symbols are all either TERMINALS or NTs */ | |
| 257 struct rule *rule; /* Linked list of rules of this (if an NT) */ | |
| 258 struct symbol *fallback; /* fallback token in case this token doesn't parse */ | |
| 259 int prec; /* Precedence if defined (-1 otherwise) */ | |
| 260 enum e_assoc assoc; /* Associativity if precedence is defined */ | |
| 261 char *firstset; /* First-set for all rules of this symbol */ | |
| 262 Boolean lambda; /* True if NT and can generate an empty string */ | |
| 263 int useCnt; /* Number of times used */ | |
| 264 char *destructor; /* Code which executes whenever this symbol is | |
| 265 ** popped from the stack during error processing */ | |
| 266 int destLineno; /* Line number for start of destructor */ | |
| 267 char *datatype; /* The data type of information held by this | |
| 268 ** object. Only used if type==NONTERMINAL */ | |
| 269 int dtnum; /* The data type number. In the parser, the value | |
| 270 ** stack is a union. The .yy%d element of this | |
| 271 ** union is the correct data type for this object */ | |
| 272 /* The following fields are used by MULTITERMINALs only */ | |
| 273 int nsubsym; /* Number of constituent symbols in the MULTI */ | |
| 274 struct symbol **subsym; /* Array of constituent symbols */ | |
| 275 }; | |
| 276 | |
| 277 /* Each production rule in the grammar is stored in the following | |
| 278 ** structure. */ | |
| 279 struct rule { | |
| 280 struct symbol *lhs; /* Left-hand side of the rule */ | |
| 281 const char *lhsalias; /* Alias for the LHS (NULL if none) */ | |
| 282 int lhsStart; /* True if left-hand side is the start symbol */ | |
| 283 int ruleline; /* Line number for the rule */ | |
| 284 int nrhs; /* Number of RHS symbols */ | |
| 285 struct symbol **rhs; /* The RHS symbols */ | |
| 286 const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */ | |
| 287 int line; /* Line number at which code begins */ | |
| 288 const char *code; /* The code executed when this rule is reduced */ | |
| 289 struct symbol *precsym; /* Precedence symbol for this rule */ | |
| 290 int index; /* An index number for this rule */ | |
| 291 Boolean canReduce; /* True if this rule is ever reduced */ | |
| 292 struct rule *nextlhs; /* Next rule with the same LHS */ | |
| 293 struct rule *next; /* Next rule in the global list */ | |
| 294 }; | |
| 295 | |
| 296 /* A configuration is a production rule of the grammar together with | |
| 297 ** a mark (dot) showing how much of that rule has been processed so far. | |
| 298 ** Configurations also contain a follow-set which is a list of terminal | |
| 299 ** symbols which are allowed to immediately follow the end of the rule. | |
| 300 ** Every configuration is recorded as an instance of the following: */ | |
| 301 enum cfgstatus { | |
| 302 COMPLETE, | |
| 303 INCOMPLETE | |
| 304 }; | |
| 305 struct config { | |
| 306 struct rule *rp; /* The rule upon which the configuration is based */ | |
| 307 int dot; /* The parse point */ | |
| 308 char *fws; /* Follow-set for this configuration only */ | |
| 309 struct plink *fplp; /* Follow-set forward propagation links */ | |
| 310 struct plink *bplp; /* Follow-set backwards propagation links */ | |
| 311 struct state *stp; /* Pointer to state which contains this */ | |
| 312 enum cfgstatus status; /* used during followset and shift computations */ | |
| 313 struct config *next; /* Next configuration in the state */ | |
| 314 struct config *bp; /* The next basis configuration */ | |
| 315 }; | |
| 316 | |
| 317 enum e_action { | |
| 318 SHIFT, | |
| 319 ACCEPT, | |
| 320 REDUCE, | |
| 321 ERROR, | |
| 322 SSCONFLICT, /* A shift/shift conflict */ | |
| 323 SRCONFLICT, /* Was a reduce, but part of a conflict */ | |
| 324 RRCONFLICT, /* Was a reduce, but part of a conflict */ | |
| 325 SH_RESOLVED, /* Was a shift. Precedence resolved conflict */ | |
| 326 RD_RESOLVED, /* Was reduce. Precedence resolved conflict */ | |
| 327 NOT_USED, /* Deleted by compression */ | |
| 328 SHIFTREDUCE /* Shift first, then reduce */ | |
| 329 }; | |
| 330 | |
| 331 /* Every shift or reduce operation is stored as one of the following */ | |
| 332 struct action { | |
| 333 struct symbol *sp; /* The look-ahead symbol */ | |
| 334 enum e_action type; | |
| 335 union { | |
| 336 struct state *stp; /* The new state, if a shift */ | |
| 337 struct rule *rp; /* The rule, if a reduce */ | |
| 338 } x; | |
| 339 struct action *next; /* Next action for this state */ | |
| 340 struct action *collide; /* Next action with the same hash */ | |
| 341 }; | |
| 342 | |
| 343 /* Each state of the generated parser's finite state machine | |
| 344 ** is encoded as an instance of the following structure. */ | |
| 345 struct state { | |
| 346 struct config *bp; /* The basis configurations for this state */ | |
| 347 struct config *cfp; /* All configurations in this set */ | |
| 348 int statenum; /* Sequential number for this state */ | |
| 349 struct action *ap; /* Array of actions for this state */ | |
| 350 int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */ | |
| 351 int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */ | |
| 352 int iDfltReduce; /* Default action is to REDUCE by this rule */ | |
| 353 struct rule *pDfltReduce;/* The default REDUCE rule. */ | |
| 354 int autoReduce; /* True if this is an auto-reduce state */ | |
| 355 }; | |
| 356 #define NO_OFFSET (-2147483647) | |
| 357 | |
| 358 /* A followset propagation link indicates that the contents of one | |
| 359 ** configuration followset should be propagated to another whenever | |
| 360 ** the first changes. */ | |
| 361 struct plink { | |
| 362 struct config *cfp; /* The configuration to which linked */ | |
| 363 struct plink *next; /* The next propagate link */ | |
| 364 }; | |
| 365 | |
| 366 /* The state vector for the entire parser generator is recorded as | |
| 367 ** follows. (LEMON uses no global variables and makes little use of | |
| 368 ** static variables. Fields in the following structure can be thought | |
| 369 ** of as begin global variables in the program.) */ | |
| 370 struct lemon { | |
| 371 struct state **sorted; /* Table of states sorted by state number */ | |
| 372 struct rule *rule; /* List of all rules */ | |
| 373 int nstate; /* Number of states */ | |
| 374 int nxstate; /* nstate with tail degenerate states removed */ | |
| 375 int nrule; /* Number of rules */ | |
| 376 int nsymbol; /* Number of terminal and nonterminal symbols */ | |
| 377 int nterminal; /* Number of terminal symbols */ | |
| 378 struct symbol **symbols; /* Sorted array of pointers to symbols */ | |
| 379 int errorcnt; /* Number of errors */ | |
| 380 struct symbol *errsym; /* The error symbol */ | |
| 381 struct symbol *wildcard; /* Token that matches anything */ | |
| 382 char *name; /* Name of the generated parser */ | |
| 383 char *arg; /* Declaration of the 3th argument to parser */ | |
| 384 char *tokentype; /* Type of terminal symbols in the parser stack */ | |
| 385 char *vartype; /* The default type of non-terminal symbols */ | |
| 386 char *start; /* Name of the start symbol for the grammar */ | |
| 387 char *stacksize; /* Size of the parser stack */ | |
| 388 char *include; /* Code to put at the start of the C file */ | |
| 389 char *error; /* Code to execute when an error is seen */ | |
| 390 char *overflow; /* Code to execute on a stack overflow */ | |
| 391 char *failure; /* Code to execute on parser failure */ | |
| 392 char *accept; /* Code to execute when the parser excepts */ | |
| 393 char *extracode; /* Code appended to the generated file */ | |
| 394 char *tokendest; /* Code to execute to destroy token data */ | |
| 395 char *vardest; /* Code for the default non-terminal destructor */ | |
| 396 char *filename; /* Name of the input file */ | |
| 397 char *outname; /* Name of the current output file */ | |
| 398 char *tokenprefix; /* A prefix added to token names in the .h file */ | |
| 399 int nconflict; /* Number of parsing conflicts */ | |
| 400 int nactiontab; /* Number of entries in the yy_action[] table */ | |
| 401 int tablesize; /* Total table size of all tables in bytes */ | |
| 402 int basisflag; /* Print only basis configurations */ | |
| 403 int has_fallback; /* True if any %fallback is seen in the grammar */ | |
| 404 int nolinenosflag; /* True if #line statements should not be printed */ | |
| 405 char *argv0; /* Name of the program */ | |
| 406 }; | |
| 407 | |
| 408 #define MemoryCheck(X) if((X)==0){ \ | |
| 409 extern void memory_error(); \ | |
| 410 memory_error(); \ | |
| 411 } | |
| 412 | |
| 413 /**************** From the file "table.h" *********************************/ | |
| 414 /* | |
| 415 ** All code in this file has been automatically generated | |
| 416 ** from a specification in the file | |
| 417 ** "table.q" | |
| 418 ** by the associative array code building program "aagen". | |
| 419 ** Do not edit this file! Instead, edit the specification | |
| 420 ** file, then rerun aagen. | |
| 421 */ | |
| 422 /* | |
| 423 ** Code for processing tables in the LEMON parser generator. | |
| 424 */ | |
| 425 /* Routines for handling a strings */ | |
| 426 | |
| 427 const char *Strsafe(const char *); | |
| 428 | |
| 429 void Strsafe_init(void); | |
| 430 int Strsafe_insert(const char *); | |
| 431 const char *Strsafe_find(const char *); | |
| 432 | |
| 433 /* Routines for handling symbols of the grammar */ | |
| 434 | |
| 435 struct symbol *Symbol_new(const char *); | |
| 436 int Symbolcmpp(const void *, const void *); | |
| 437 void Symbol_init(void); | |
| 438 int Symbol_insert(struct symbol *, const char *); | |
| 439 struct symbol *Symbol_find(const char *); | |
| 440 struct symbol *Symbol_Nth(int); | |
| 441 int Symbol_count(void); | |
| 442 struct symbol **Symbol_arrayof(void); | |
| 443 | |
| 444 /* Routines to manage the state table */ | |
| 445 | |
| 446 int Configcmp(const char *, const char *); | |
| 447 struct state *State_new(void); | |
| 448 void State_init(void); | |
| 449 int State_insert(struct state *, struct config *); | |
| 450 struct state *State_find(struct config *); | |
| 451 struct state **State_arrayof(/* */); | |
| 452 | |
| 453 /* Routines used for efficiency in Configlist_add */ | |
| 454 | |
| 455 void Configtable_init(void); | |
| 456 int Configtable_insert(struct config *); | |
| 457 struct config *Configtable_find(struct config *); | |
| 458 void Configtable_clear(int(*)(struct config *)); | |
| 459 | |
| 460 /****************** From the file "action.c" *******************************/ | |
| 461 /* | |
| 462 ** Routines processing parser actions in the LEMON parser generator. | |
| 463 */ | |
| 464 | |
| 465 /* Allocate a new parser action */ | |
| 466 static struct action *Action_new(void){ | |
| 467 static struct action *freelist = 0; | |
| 468 struct action *newaction; | |
| 469 | |
| 470 if( freelist==0 ){ | |
| 471 int i; | |
| 472 int amt = 100; | |
| 473 freelist = (struct action *)calloc(amt, sizeof(struct action)); | |
| 474 if( freelist==0 ){ | |
| 475 fprintf(stderr,"Unable to allocate memory for a new parser action."); | |
| 476 exit(1); | |
| 477 } | |
| 478 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1]; | |
| 479 freelist[amt-1].next = 0; | |
| 480 } | |
| 481 newaction = freelist; | |
| 482 freelist = freelist->next; | |
| 483 return newaction; | |
| 484 } | |
| 485 | |
| 486 /* Compare two actions for sorting purposes. Return negative, zero, or | |
| 487 ** positive if the first action is less than, equal to, or greater than | |
| 488 ** the first | |
| 489 */ | |
| 490 static int actioncmp( | |
| 491 struct action *ap1, | |
| 492 struct action *ap2 | |
| 493 ){ | |
| 494 int rc; | |
| 495 rc = ap1->sp->index - ap2->sp->index; | |
| 496 if( rc==0 ){ | |
| 497 rc = (int)ap1->type - (int)ap2->type; | |
| 498 } | |
| 499 if( rc==0 && (ap1->type==REDUCE || ap1->type==SHIFTREDUCE) ){ | |
| 500 rc = ap1->x.rp->index - ap2->x.rp->index; | |
| 501 } | |
| 502 if( rc==0 ){ | |
| 503 rc = (int) (ap2 - ap1); | |
| 504 } | |
| 505 return rc; | |
| 506 } | |
| 507 | |
| 508 /* Sort parser actions */ | |
| 509 static struct action *Action_sort( | |
| 510 struct action *ap | |
| 511 ){ | |
| 512 ap = (struct action *)msort((char *)ap,(char **)&ap->next, | |
| 513 (int(*)(const char*,const char*))actioncmp); | |
| 514 return ap; | |
| 515 } | |
| 516 | |
| 517 void Action_add( | |
| 518 struct action **app, | |
| 519 enum e_action type, | |
| 520 struct symbol *sp, | |
| 521 char *arg | |
| 522 ){ | |
| 523 struct action *newaction; | |
| 524 newaction = Action_new(); | |
| 525 newaction->next = *app; | |
| 526 *app = newaction; | |
| 527 newaction->type = type; | |
| 528 newaction->sp = sp; | |
| 529 if( type==SHIFT ){ | |
| 530 newaction->x.stp = (struct state *)arg; | |
| 531 }else{ | |
| 532 newaction->x.rp = (struct rule *)arg; | |
| 533 } | |
| 534 } | |
| 535 /********************** New code to implement the "acttab" module ***********/ | |
| 536 /* | |
| 537 ** This module implements routines use to construct the yy_action[] table. | |
| 538 */ | |
| 539 | |
| 540 /* | |
| 541 ** The state of the yy_action table under construction is an instance of | |
| 542 ** the following structure. | |
| 543 ** | |
| 544 ** The yy_action table maps the pair (state_number, lookahead) into an | |
| 545 ** action_number. The table is an array of integers pairs. The state_number | |
| 546 ** determines an initial offset into the yy_action array. The lookahead | |
| 547 ** value is then added to this initial offset to get an index X into the | |
| 548 ** yy_action array. If the aAction[X].lookahead equals the value of the | |
| 549 ** of the lookahead input, then the value of the action_number output is | |
| 550 ** aAction[X].action. If the lookaheads do not match then the | |
| 551 ** default action for the state_number is returned. | |
| 552 ** | |
| 553 ** All actions associated with a single state_number are first entered | |
| 554 ** into aLookahead[] using multiple calls to acttab_action(). Then the | |
| 555 ** actions for that single state_number are placed into the aAction[] | |
| 556 ** array with a single call to acttab_insert(). The acttab_insert() call | |
| 557 ** also resets the aLookahead[] array in preparation for the next | |
| 558 ** state number. | |
| 559 */ | |
| 560 struct lookahead_action { | |
| 561 int lookahead; /* Value of the lookahead token */ | |
| 562 int action; /* Action to take on the given lookahead */ | |
| 563 }; | |
| 564 typedef struct acttab acttab; | |
| 565 struct acttab { | |
| 566 int nAction; /* Number of used slots in aAction[] */ | |
| 567 int nActionAlloc; /* Slots allocated for aAction[] */ | |
| 568 struct lookahead_action | |
| 569 *aAction, /* The yy_action[] table under construction */ | |
| 570 *aLookahead; /* A single new transaction set */ | |
| 571 int mnLookahead; /* Minimum aLookahead[].lookahead */ | |
| 572 int mnAction; /* Action associated with mnLookahead */ | |
| 573 int mxLookahead; /* Maximum aLookahead[].lookahead */ | |
| 574 int nLookahead; /* Used slots in aLookahead[] */ | |
| 575 int nLookaheadAlloc; /* Slots allocated in aLookahead[] */ | |
| 576 }; | |
| 577 | |
| 578 /* Return the number of entries in the yy_action table */ | |
| 579 #define acttab_size(X) ((X)->nAction) | |
| 580 | |
| 581 /* The value for the N-th entry in yy_action */ | |
| 582 #define acttab_yyaction(X,N) ((X)->aAction[N].action) | |
| 583 | |
| 584 /* The value for the N-th entry in yy_lookahead */ | |
| 585 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead) | |
| 586 | |
| 587 /* Free all memory associated with the given acttab */ | |
| 588 void acttab_free(acttab *p){ | |
| 589 free( p->aAction ); | |
| 590 free( p->aLookahead ); | |
| 591 free( p ); | |
| 592 } | |
| 593 | |
| 594 /* Allocate a new acttab structure */ | |
| 595 acttab *acttab_alloc(void){ | |
| 596 acttab *p = (acttab *) calloc( 1, sizeof(*p) ); | |
| 597 if( p==0 ){ | |
| 598 fprintf(stderr,"Unable to allocate memory for a new acttab."); | |
| 599 exit(1); | |
| 600 } | |
| 601 memset(p, 0, sizeof(*p)); | |
| 602 return p; | |
| 603 } | |
| 604 | |
| 605 /* Add a new action to the current transaction set. | |
| 606 ** | |
| 607 ** This routine is called once for each lookahead for a particular | |
| 608 ** state. | |
| 609 */ | |
| 610 void acttab_action(acttab *p, int lookahead, int action){ | |
| 611 if( p->nLookahead>=p->nLookaheadAlloc ){ | |
| 612 p->nLookaheadAlloc += 25; | |
| 613 p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead, | |
| 614 sizeof(p->aLookahead[0])*p->nLookaheadAlloc ); | |
| 615 if( p->aLookahead==0 ){ | |
| 616 fprintf(stderr,"malloc failed\n"); | |
| 617 exit(1); | |
| 618 } | |
| 619 } | |
| 620 if( p->nLookahead==0 ){ | |
| 621 p->mxLookahead = lookahead; | |
| 622 p->mnLookahead = lookahead; | |
| 623 p->mnAction = action; | |
| 624 }else{ | |
| 625 if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead; | |
| 626 if( p->mnLookahead>lookahead ){ | |
| 627 p->mnLookahead = lookahead; | |
| 628 p->mnAction = action; | |
| 629 } | |
| 630 } | |
| 631 p->aLookahead[p->nLookahead].lookahead = lookahead; | |
| 632 p->aLookahead[p->nLookahead].action = action; | |
| 633 p->nLookahead++; | |
| 634 } | |
| 635 | |
| 636 /* | |
| 637 ** Add the transaction set built up with prior calls to acttab_action() | |
| 638 ** into the current action table. Then reset the transaction set back | |
| 639 ** to an empty set in preparation for a new round of acttab_action() calls. | |
| 640 ** | |
| 641 ** Return the offset into the action table of the new transaction. | |
| 642 */ | |
| 643 int acttab_insert(acttab *p){ | |
| 644 int i, j, k, n; | |
| 645 assert( p->nLookahead>0 ); | |
| 646 | |
| 647 /* Make sure we have enough space to hold the expanded action table | |
| 648 ** in the worst case. The worst case occurs if the transaction set | |
| 649 ** must be appended to the current action table | |
| 650 */ | |
| 651 n = p->mxLookahead + 1; | |
| 652 if( p->nAction + n >= p->nActionAlloc ){ | |
| 653 int oldAlloc = p->nActionAlloc; | |
| 654 p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20; | |
| 655 p->aAction = (struct lookahead_action *) realloc( p->aAction, | |
| 656 sizeof(p->aAction[0])*p->nActionAlloc); | |
| 657 if( p->aAction==0 ){ | |
| 658 fprintf(stderr,"malloc failed\n"); | |
| 659 exit(1); | |
| 660 } | |
| 661 for(i=oldAlloc; i<p->nActionAlloc; i++){ | |
| 662 p->aAction[i].lookahead = -1; | |
| 663 p->aAction[i].action = -1; | |
| 664 } | |
| 665 } | |
| 666 | |
| 667 /* Scan the existing action table looking for an offset that is a | |
| 668 ** duplicate of the current transaction set. Fall out of the loop | |
| 669 ** if and when the duplicate is found. | |
| 670 ** | |
| 671 ** i is the index in p->aAction[] where p->mnLookahead is inserted. | |
| 672 */ | |
| 673 for(i=p->nAction-1; i>=0; i--){ | |
| 674 if( p->aAction[i].lookahead==p->mnLookahead ){ | |
| 675 /* All lookaheads and actions in the aLookahead[] transaction | |
| 676 ** must match against the candidate aAction[i] entry. */ | |
| 677 if( p->aAction[i].action!=p->mnAction ) continue; | |
| 678 for(j=0; j<p->nLookahead; j++){ | |
| 679 k = p->aLookahead[j].lookahead - p->mnLookahead + i; | |
| 680 if( k<0 || k>=p->nAction ) break; | |
| 681 if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break; | |
| 682 if( p->aLookahead[j].action!=p->aAction[k].action ) break; | |
| 683 } | |
| 684 if( j<p->nLookahead ) continue; | |
| 685 | |
| 686 /* No possible lookahead value that is not in the aLookahead[] | |
| 687 ** transaction is allowed to match aAction[i] */ | |
| 688 n = 0; | |
| 689 for(j=0; j<p->nAction; j++){ | |
| 690 if( p->aAction[j].lookahead<0 ) continue; | |
| 691 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++; | |
| 692 } | |
| 693 if( n==p->nLookahead ){ | |
| 694 break; /* An exact match is found at offset i */ | |
| 695 } | |
| 696 } | |
| 697 } | |
| 698 | |
| 699 /* If no existing offsets exactly match the current transaction, find an | |
| 700 ** an empty offset in the aAction[] table in which we can add the | |
| 701 ** aLookahead[] transaction. | |
| 702 */ | |
| 703 if( i<0 ){ | |
| 704 /* Look for holes in the aAction[] table that fit the current | |
| 705 ** aLookahead[] transaction. Leave i set to the offset of the hole. | |
| 706 ** If no holes are found, i is left at p->nAction, which means the | |
| 707 ** transaction will be appended. */ | |
| 708 for(i=0; i<p->nActionAlloc - p->mxLookahead; i++){ | |
| 709 if( p->aAction[i].lookahead<0 ){ | |
| 710 for(j=0; j<p->nLookahead; j++){ | |
| 711 k = p->aLookahead[j].lookahead - p->mnLookahead + i; | |
| 712 if( k<0 ) break; | |
| 713 if( p->aAction[k].lookahead>=0 ) break; | |
| 714 } | |
| 715 if( j<p->nLookahead ) continue; | |
| 716 for(j=0; j<p->nAction; j++){ | |
| 717 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break; | |
| 718 } | |
| 719 if( j==p->nAction ){ | |
| 720 break; /* Fits in empty slots */ | |
| 721 } | |
| 722 } | |
| 723 } | |
| 724 } | |
| 725 /* Insert transaction set at index i. */ | |
| 726 for(j=0; j<p->nLookahead; j++){ | |
| 727 k = p->aLookahead[j].lookahead - p->mnLookahead + i; | |
| 728 p->aAction[k] = p->aLookahead[j]; | |
| 729 if( k>=p->nAction ) p->nAction = k+1; | |
| 730 } | |
| 731 p->nLookahead = 0; | |
| 732 | |
| 733 /* Return the offset that is added to the lookahead in order to get the | |
| 734 ** index into yy_action of the action */ | |
| 735 return i - p->mnLookahead; | |
| 736 } | |
| 737 | |
| 738 /********************** From the file "build.c" *****************************/ | |
| 739 /* | |
| 740 ** Routines to construction the finite state machine for the LEMON | |
| 741 ** parser generator. | |
| 742 */ | |
| 743 | |
| 744 /* Find a precedence symbol of every rule in the grammar. | |
| 745 ** | |
| 746 ** Those rules which have a precedence symbol coded in the input | |
| 747 ** grammar using the "[symbol]" construct will already have the | |
| 748 ** rp->precsym field filled. Other rules take as their precedence | |
| 749 ** symbol the first RHS symbol with a defined precedence. If there | |
| 750 ** are not RHS symbols with a defined precedence, the precedence | |
| 751 ** symbol field is left blank. | |
| 752 */ | |
| 753 void FindRulePrecedences(struct lemon *xp) | |
| 754 { | |
| 755 struct rule *rp; | |
| 756 for(rp=xp->rule; rp; rp=rp->next){ | |
| 757 if( rp->precsym==0 ){ | |
| 758 int i, j; | |
| 759 for(i=0; i<rp->nrhs && rp->precsym==0; i++){ | |
| 760 struct symbol *sp = rp->rhs[i]; | |
| 761 if( sp->type==MULTITERMINAL ){ | |
| 762 for(j=0; j<sp->nsubsym; j++){ | |
| 763 if( sp->subsym[j]->prec>=0 ){ | |
| 764 rp->precsym = sp->subsym[j]; | |
| 765 break; | |
| 766 } | |
| 767 } | |
| 768 }else if( sp->prec>=0 ){ | |
| 769 rp->precsym = rp->rhs[i]; | |
| 770 } | |
| 771 } | |
| 772 } | |
| 773 } | |
| 774 return; | |
| 775 } | |
| 776 | |
| 777 /* Find all nonterminals which will generate the empty string. | |
| 778 ** Then go back and compute the first sets of every nonterminal. | |
| 779 ** The first set is the set of all terminal symbols which can begin | |
| 780 ** a string generated by that nonterminal. | |
| 781 */ | |
| 782 void FindFirstSets(struct lemon *lemp) | |
| 783 { | |
| 784 int i, j; | |
| 785 struct rule *rp; | |
| 786 int progress; | |
| 787 | |
| 788 for(i=0; i<lemp->nsymbol; i++){ | |
| 789 lemp->symbols[i]->lambda = LEMON_FALSE; | |
| 790 } | |
| 791 for(i=lemp->nterminal; i<lemp->nsymbol; i++){ | |
| 792 lemp->symbols[i]->firstset = SetNew(); | |
| 793 } | |
| 794 | |
| 795 /* First compute all lambdas */ | |
| 796 do{ | |
| 797 progress = 0; | |
| 798 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 799 if( rp->lhs->lambda ) continue; | |
| 800 for(i=0; i<rp->nrhs; i++){ | |
| 801 struct symbol *sp = rp->rhs[i]; | |
| 802 assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE ); | |
| 803 if( sp->lambda==LEMON_FALSE ) break; | |
| 804 } | |
| 805 if( i==rp->nrhs ){ | |
| 806 rp->lhs->lambda = LEMON_TRUE; | |
| 807 progress = 1; | |
| 808 } | |
| 809 } | |
| 810 }while( progress ); | |
| 811 | |
| 812 /* Now compute all first sets */ | |
| 813 do{ | |
| 814 struct symbol *s1, *s2; | |
| 815 progress = 0; | |
| 816 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 817 s1 = rp->lhs; | |
| 818 for(i=0; i<rp->nrhs; i++){ | |
| 819 s2 = rp->rhs[i]; | |
| 820 if( s2->type==TERMINAL ){ | |
| 821 progress += SetAdd(s1->firstset,s2->index); | |
| 822 break; | |
| 823 }else if( s2->type==MULTITERMINAL ){ | |
| 824 for(j=0; j<s2->nsubsym; j++){ | |
| 825 progress += SetAdd(s1->firstset,s2->subsym[j]->index); | |
| 826 } | |
| 827 break; | |
| 828 }else if( s1==s2 ){ | |
| 829 if( s1->lambda==LEMON_FALSE ) break; | |
| 830 }else{ | |
| 831 progress += SetUnion(s1->firstset,s2->firstset); | |
| 832 if( s2->lambda==LEMON_FALSE ) break; | |
| 833 } | |
| 834 } | |
| 835 } | |
| 836 }while( progress ); | |
| 837 return; | |
| 838 } | |
| 839 | |
| 840 /* Compute all LR(0) states for the grammar. Links | |
| 841 ** are added to between some states so that the LR(1) follow sets | |
| 842 ** can be computed later. | |
| 843 */ | |
| 844 PRIVATE struct state *getstate(struct lemon *); /* forward reference */ | |
| 845 void FindStates(struct lemon *lemp) | |
| 846 { | |
| 847 struct symbol *sp; | |
| 848 struct rule *rp; | |
| 849 | |
| 850 Configlist_init(); | |
| 851 | |
| 852 /* Find the start symbol */ | |
| 853 if( lemp->start ){ | |
| 854 sp = Symbol_find(lemp->start); | |
| 855 if( sp==0 ){ | |
| 856 ErrorMsg(lemp->filename,0, | |
| 857 "The specified start symbol \"%s\" is not \ | |
| 858 in a nonterminal of the grammar. \"%s\" will be used as the start \ | |
| 859 symbol instead.",lemp->start,lemp->rule->lhs->name); | |
| 860 lemp->errorcnt++; | |
| 861 sp = lemp->rule->lhs; | |
| 862 } | |
| 863 }else{ | |
| 864 sp = lemp->rule->lhs; | |
| 865 } | |
| 866 | |
| 867 /* Make sure the start symbol doesn't occur on the right-hand side of | |
| 868 ** any rule. Report an error if it does. (YACC would generate a new | |
| 869 ** start symbol in this case.) */ | |
| 870 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 871 int i; | |
| 872 for(i=0; i<rp->nrhs; i++){ | |
| 873 if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */ | |
| 874 ErrorMsg(lemp->filename,0, | |
| 875 "The start symbol \"%s\" occurs on the \ | |
| 876 right-hand side of a rule. This will result in a parser which \ | |
| 877 does not work properly.",sp->name); | |
| 878 lemp->errorcnt++; | |
| 879 } | |
| 880 } | |
| 881 } | |
| 882 | |
| 883 /* The basis configuration set for the first state | |
| 884 ** is all rules which have the start symbol as their | |
| 885 ** left-hand side */ | |
| 886 for(rp=sp->rule; rp; rp=rp->nextlhs){ | |
| 887 struct config *newcfp; | |
| 888 rp->lhsStart = 1; | |
| 889 newcfp = Configlist_addbasis(rp,0); | |
| 890 SetAdd(newcfp->fws,0); | |
| 891 } | |
| 892 | |
| 893 /* Compute the first state. All other states will be | |
| 894 ** computed automatically during the computation of the first one. | |
| 895 ** The returned pointer to the first state is not used. */ | |
| 896 (void)getstate(lemp); | |
| 897 return; | |
| 898 } | |
| 899 | |
| 900 /* Return a pointer to a state which is described by the configuration | |
| 901 ** list which has been built from calls to Configlist_add. | |
| 902 */ | |
| 903 PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */ | |
| 904 PRIVATE struct state *getstate(struct lemon *lemp) | |
| 905 { | |
| 906 struct config *cfp, *bp; | |
| 907 struct state *stp; | |
| 908 | |
| 909 /* Extract the sorted basis of the new state. The basis was constructed | |
| 910 ** by prior calls to "Configlist_addbasis()". */ | |
| 911 Configlist_sortbasis(); | |
| 912 bp = Configlist_basis(); | |
| 913 | |
| 914 /* Get a state with the same basis */ | |
| 915 stp = State_find(bp); | |
| 916 if( stp ){ | |
| 917 /* A state with the same basis already exists! Copy all the follow-set | |
| 918 ** propagation links from the state under construction into the | |
| 919 ** preexisting state, then return a pointer to the preexisting state */ | |
| 920 struct config *x, *y; | |
| 921 for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){ | |
| 922 Plink_copy(&y->bplp,x->bplp); | |
| 923 Plink_delete(x->fplp); | |
| 924 x->fplp = x->bplp = 0; | |
| 925 } | |
| 926 cfp = Configlist_return(); | |
| 927 Configlist_eat(cfp); | |
| 928 }else{ | |
| 929 /* This really is a new state. Construct all the details */ | |
| 930 Configlist_closure(lemp); /* Compute the configuration closure */ | |
| 931 Configlist_sort(); /* Sort the configuration closure */ | |
| 932 cfp = Configlist_return(); /* Get a pointer to the config list */ | |
| 933 stp = State_new(); /* A new state structure */ | |
| 934 MemoryCheck(stp); | |
| 935 stp->bp = bp; /* Remember the configuration basis */ | |
| 936 stp->cfp = cfp; /* Remember the configuration closure */ | |
| 937 stp->statenum = lemp->nstate++; /* Every state gets a sequence number */ | |
| 938 stp->ap = 0; /* No actions, yet. */ | |
| 939 State_insert(stp,stp->bp); /* Add to the state table */ | |
| 940 buildshifts(lemp,stp); /* Recursively compute successor states */ | |
| 941 } | |
| 942 return stp; | |
| 943 } | |
| 944 | |
| 945 /* | |
| 946 ** Return true if two symbols are the same. | |
| 947 */ | |
| 948 int same_symbol(struct symbol *a, struct symbol *b) | |
| 949 { | |
| 950 int i; | |
| 951 if( a==b ) return 1; | |
| 952 if( a->type!=MULTITERMINAL ) return 0; | |
| 953 if( b->type!=MULTITERMINAL ) return 0; | |
| 954 if( a->nsubsym!=b->nsubsym ) return 0; | |
| 955 for(i=0; i<a->nsubsym; i++){ | |
| 956 if( a->subsym[i]!=b->subsym[i] ) return 0; | |
| 957 } | |
| 958 return 1; | |
| 959 } | |
| 960 | |
| 961 /* Construct all successor states to the given state. A "successor" | |
| 962 ** state is any state which can be reached by a shift action. | |
| 963 */ | |
| 964 PRIVATE void buildshifts(struct lemon *lemp, struct state *stp) | |
| 965 { | |
| 966 struct config *cfp; /* For looping thru the config closure of "stp" */ | |
| 967 struct config *bcfp; /* For the inner loop on config closure of "stp" */ | |
| 968 struct config *newcfg; /* */ | |
| 969 struct symbol *sp; /* Symbol following the dot in configuration "cfp" */ | |
| 970 struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */ | |
| 971 struct state *newstp; /* A pointer to a successor state */ | |
| 972 | |
| 973 /* Each configuration becomes complete after it contibutes to a successor | |
| 974 ** state. Initially, all configurations are incomplete */ | |
| 975 for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE; | |
| 976 | |
| 977 /* Loop through all configurations of the state "stp" */ | |
| 978 for(cfp=stp->cfp; cfp; cfp=cfp->next){ | |
| 979 if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */ | |
| 980 if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */ | |
| 981 Configlist_reset(); /* Reset the new config set */ | |
| 982 sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */ | |
| 983 | |
| 984 /* For every configuration in the state "stp" which has the symbol "sp" | |
| 985 ** following its dot, add the same configuration to the basis set under | |
| 986 ** construction but with the dot shifted one symbol to the right. */ | |
| 987 for(bcfp=cfp; bcfp; bcfp=bcfp->next){ | |
| 988 if( bcfp->status==COMPLETE ) continue; /* Already used */ | |
| 989 if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */ | |
| 990 bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */ | |
| 991 if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */ | |
| 992 bcfp->status = COMPLETE; /* Mark this config as used */ | |
| 993 newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1); | |
| 994 Plink_add(&newcfg->bplp,bcfp); | |
| 995 } | |
| 996 | |
| 997 /* Get a pointer to the state described by the basis configuration set | |
| 998 ** constructed in the preceding loop */ | |
| 999 newstp = getstate(lemp); | |
| 1000 | |
| 1001 /* The state "newstp" is reached from the state "stp" by a shift action | |
| 1002 ** on the symbol "sp" */ | |
| 1003 if( sp->type==MULTITERMINAL ){ | |
| 1004 int i; | |
| 1005 for(i=0; i<sp->nsubsym; i++){ | |
| 1006 Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp); | |
| 1007 } | |
| 1008 }else{ | |
| 1009 Action_add(&stp->ap,SHIFT,sp,(char *)newstp); | |
| 1010 } | |
| 1011 } | |
| 1012 } | |
| 1013 | |
| 1014 /* | |
| 1015 ** Construct the propagation links | |
| 1016 */ | |
| 1017 void FindLinks(struct lemon *lemp) | |
| 1018 { | |
| 1019 int i; | |
| 1020 struct config *cfp, *other; | |
| 1021 struct state *stp; | |
| 1022 struct plink *plp; | |
| 1023 | |
| 1024 /* Housekeeping detail: | |
| 1025 ** Add to every propagate link a pointer back to the state to | |
| 1026 ** which the link is attached. */ | |
| 1027 for(i=0; i<lemp->nstate; i++){ | |
| 1028 stp = lemp->sorted[i]; | |
| 1029 for(cfp=stp->cfp; cfp; cfp=cfp->next){ | |
| 1030 cfp->stp = stp; | |
| 1031 } | |
| 1032 } | |
| 1033 | |
| 1034 /* Convert all backlinks into forward links. Only the forward | |
| 1035 ** links are used in the follow-set computation. */ | |
| 1036 for(i=0; i<lemp->nstate; i++){ | |
| 1037 stp = lemp->sorted[i]; | |
| 1038 for(cfp=stp->cfp; cfp; cfp=cfp->next){ | |
| 1039 for(plp=cfp->bplp; plp; plp=plp->next){ | |
| 1040 other = plp->cfp; | |
| 1041 Plink_add(&other->fplp,cfp); | |
| 1042 } | |
| 1043 } | |
| 1044 } | |
| 1045 } | |
| 1046 | |
| 1047 /* Compute all followsets. | |
| 1048 ** | |
| 1049 ** A followset is the set of all symbols which can come immediately | |
| 1050 ** after a configuration. | |
| 1051 */ | |
| 1052 void FindFollowSets(struct lemon *lemp) | |
| 1053 { | |
| 1054 int i; | |
| 1055 struct config *cfp; | |
| 1056 struct plink *plp; | |
| 1057 int progress; | |
| 1058 int change; | |
| 1059 | |
| 1060 for(i=0; i<lemp->nstate; i++){ | |
| 1061 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){ | |
| 1062 cfp->status = INCOMPLETE; | |
| 1063 } | |
| 1064 } | |
| 1065 | |
| 1066 do{ | |
| 1067 progress = 0; | |
| 1068 for(i=0; i<lemp->nstate; i++){ | |
| 1069 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){ | |
| 1070 if( cfp->status==COMPLETE ) continue; | |
| 1071 for(plp=cfp->fplp; plp; plp=plp->next){ | |
| 1072 change = SetUnion(plp->cfp->fws,cfp->fws); | |
| 1073 if( change ){ | |
| 1074 plp->cfp->status = INCOMPLETE; | |
| 1075 progress = 1; | |
| 1076 } | |
| 1077 } | |
| 1078 cfp->status = COMPLETE; | |
| 1079 } | |
| 1080 } | |
| 1081 }while( progress ); | |
| 1082 } | |
| 1083 | |
| 1084 static int resolve_conflict(struct action *,struct action *); | |
| 1085 | |
| 1086 /* Compute the reduce actions, and resolve conflicts. | |
| 1087 */ | |
| 1088 void FindActions(struct lemon *lemp) | |
| 1089 { | |
| 1090 int i,j; | |
| 1091 struct config *cfp; | |
| 1092 struct state *stp; | |
| 1093 struct symbol *sp; | |
| 1094 struct rule *rp; | |
| 1095 | |
| 1096 /* Add all of the reduce actions | |
| 1097 ** A reduce action is added for each element of the followset of | |
| 1098 ** a configuration which has its dot at the extreme right. | |
| 1099 */ | |
| 1100 for(i=0; i<lemp->nstate; i++){ /* Loop over all states */ | |
| 1101 stp = lemp->sorted[i]; | |
| 1102 for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */ | |
| 1103 if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */ | |
| 1104 for(j=0; j<lemp->nterminal; j++){ | |
| 1105 if( SetFind(cfp->fws,j) ){ | |
| 1106 /* Add a reduce action to the state "stp" which will reduce by the | |
| 1107 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */ | |
| 1108 Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp); | |
| 1109 } | |
| 1110 } | |
| 1111 } | |
| 1112 } | |
| 1113 } | |
| 1114 | |
| 1115 /* Add the accepting token */ | |
| 1116 if( lemp->start ){ | |
| 1117 sp = Symbol_find(lemp->start); | |
| 1118 if( sp==0 ) sp = lemp->rule->lhs; | |
| 1119 }else{ | |
| 1120 sp = lemp->rule->lhs; | |
| 1121 } | |
| 1122 /* Add to the first state (which is always the starting state of the | |
| 1123 ** finite state machine) an action to ACCEPT if the lookahead is the | |
| 1124 ** start nonterminal. */ | |
| 1125 Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0); | |
| 1126 | |
| 1127 /* Resolve conflicts */ | |
| 1128 for(i=0; i<lemp->nstate; i++){ | |
| 1129 struct action *ap, *nap; | |
| 1130 stp = lemp->sorted[i]; | |
| 1131 /* assert( stp->ap ); */ | |
| 1132 stp->ap = Action_sort(stp->ap); | |
| 1133 for(ap=stp->ap; ap && ap->next; ap=ap->next){ | |
| 1134 for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){ | |
| 1135 /* The two actions "ap" and "nap" have the same lookahead. | |
| 1136 ** Figure out which one should be used */ | |
| 1137 lemp->nconflict += resolve_conflict(ap,nap); | |
| 1138 } | |
| 1139 } | |
| 1140 } | |
| 1141 | |
| 1142 /* Report an error for each rule that can never be reduced. */ | |
| 1143 for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE; | |
| 1144 for(i=0; i<lemp->nstate; i++){ | |
| 1145 struct action *ap; | |
| 1146 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){ | |
| 1147 if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE; | |
| 1148 } | |
| 1149 } | |
| 1150 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 1151 if( rp->canReduce ) continue; | |
| 1152 ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n"); | |
| 1153 lemp->errorcnt++; | |
| 1154 } | |
| 1155 } | |
| 1156 | |
| 1157 /* Resolve a conflict between the two given actions. If the | |
| 1158 ** conflict can't be resolved, return non-zero. | |
| 1159 ** | |
| 1160 ** NO LONGER TRUE: | |
| 1161 ** To resolve a conflict, first look to see if either action | |
| 1162 ** is on an error rule. In that case, take the action which | |
| 1163 ** is not associated with the error rule. If neither or both | |
| 1164 ** actions are associated with an error rule, then try to | |
| 1165 ** use precedence to resolve the conflict. | |
| 1166 ** | |
| 1167 ** If either action is a SHIFT, then it must be apx. This | |
| 1168 ** function won't work if apx->type==REDUCE and apy->type==SHIFT. | |
| 1169 */ | |
| 1170 static int resolve_conflict( | |
| 1171 struct action *apx, | |
| 1172 struct action *apy | |
| 1173 ){ | |
| 1174 struct symbol *spx, *spy; | |
| 1175 int errcnt = 0; | |
| 1176 assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */ | |
| 1177 if( apx->type==SHIFT && apy->type==SHIFT ){ | |
| 1178 apy->type = SSCONFLICT; | |
| 1179 errcnt++; | |
| 1180 } | |
| 1181 if( apx->type==SHIFT && apy->type==REDUCE ){ | |
| 1182 spx = apx->sp; | |
| 1183 spy = apy->x.rp->precsym; | |
| 1184 if( spy==0 || spx->prec<0 || spy->prec<0 ){ | |
| 1185 /* Not enough precedence information. */ | |
| 1186 apy->type = SRCONFLICT; | |
| 1187 errcnt++; | |
| 1188 }else if( spx->prec>spy->prec ){ /* higher precedence wins */ | |
| 1189 apy->type = RD_RESOLVED; | |
| 1190 }else if( spx->prec<spy->prec ){ | |
| 1191 apx->type = SH_RESOLVED; | |
| 1192 }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */ | |
| 1193 apy->type = RD_RESOLVED; /* associativity */ | |
| 1194 }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */ | |
| 1195 apx->type = SH_RESOLVED; | |
| 1196 }else{ | |
| 1197 assert( spx->prec==spy->prec && spx->assoc==NONE ); | |
| 1198 apx->type = ERROR; | |
| 1199 } | |
| 1200 }else if( apx->type==REDUCE && apy->type==REDUCE ){ | |
| 1201 spx = apx->x.rp->precsym; | |
| 1202 spy = apy->x.rp->precsym; | |
| 1203 if( spx==0 || spy==0 || spx->prec<0 || | |
| 1204 spy->prec<0 || spx->prec==spy->prec ){ | |
| 1205 apy->type = RRCONFLICT; | |
| 1206 errcnt++; | |
| 1207 }else if( spx->prec>spy->prec ){ | |
| 1208 apy->type = RD_RESOLVED; | |
| 1209 }else if( spx->prec<spy->prec ){ | |
| 1210 apx->type = RD_RESOLVED; | |
| 1211 } | |
| 1212 }else{ | |
| 1213 assert( | |
| 1214 apx->type==SH_RESOLVED || | |
| 1215 apx->type==RD_RESOLVED || | |
| 1216 apx->type==SSCONFLICT || | |
| 1217 apx->type==SRCONFLICT || | |
| 1218 apx->type==RRCONFLICT || | |
| 1219 apy->type==SH_RESOLVED || | |
| 1220 apy->type==RD_RESOLVED || | |
| 1221 apy->type==SSCONFLICT || | |
| 1222 apy->type==SRCONFLICT || | |
| 1223 apy->type==RRCONFLICT | |
| 1224 ); | |
| 1225 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before | |
| 1226 ** REDUCEs on the list. If we reach this point it must be because | |
| 1227 ** the parser conflict had already been resolved. */ | |
| 1228 } | |
| 1229 return errcnt; | |
| 1230 } | |
| 1231 /********************* From the file "configlist.c" *************************/ | |
| 1232 /* | |
| 1233 ** Routines to processing a configuration list and building a state | |
| 1234 ** in the LEMON parser generator. | |
| 1235 */ | |
| 1236 | |
| 1237 static struct config *freelist = 0; /* List of free configurations */ | |
| 1238 static struct config *current = 0; /* Top of list of configurations */ | |
| 1239 static struct config **currentend = 0; /* Last on list of configs */ | |
| 1240 static struct config *basis = 0; /* Top of list of basis configs */ | |
| 1241 static struct config **basisend = 0; /* End of list of basis configs */ | |
| 1242 | |
| 1243 /* Return a pointer to a new configuration */ | |
| 1244 PRIVATE struct config *newconfig(){ | |
| 1245 struct config *newcfg; | |
| 1246 if( freelist==0 ){ | |
| 1247 int i; | |
| 1248 int amt = 3; | |
| 1249 freelist = (struct config *)calloc( amt, sizeof(struct config) ); | |
| 1250 if( freelist==0 ){ | |
| 1251 fprintf(stderr,"Unable to allocate memory for a new configuration."); | |
| 1252 exit(1); | |
| 1253 } | |
| 1254 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1]; | |
| 1255 freelist[amt-1].next = 0; | |
| 1256 } | |
| 1257 newcfg = freelist; | |
| 1258 freelist = freelist->next; | |
| 1259 return newcfg; | |
| 1260 } | |
| 1261 | |
| 1262 /* The configuration "old" is no longer used */ | |
| 1263 PRIVATE void deleteconfig(struct config *old) | |
| 1264 { | |
| 1265 old->next = freelist; | |
| 1266 freelist = old; | |
| 1267 } | |
| 1268 | |
| 1269 /* Initialized the configuration list builder */ | |
| 1270 void Configlist_init(){ | |
| 1271 current = 0; | |
| 1272 currentend = ¤t; | |
| 1273 basis = 0; | |
| 1274 basisend = &basis; | |
| 1275 Configtable_init(); | |
| 1276 return; | |
| 1277 } | |
| 1278 | |
| 1279 /* Initialized the configuration list builder */ | |
| 1280 void Configlist_reset(){ | |
| 1281 current = 0; | |
| 1282 currentend = ¤t; | |
| 1283 basis = 0; | |
| 1284 basisend = &basis; | |
| 1285 Configtable_clear(0); | |
| 1286 return; | |
| 1287 } | |
| 1288 | |
| 1289 /* Add another configuration to the configuration list */ | |
| 1290 struct config *Configlist_add( | |
| 1291 struct rule *rp, /* The rule */ | |
| 1292 int dot /* Index into the RHS of the rule where the dot goes */ | |
| 1293 ){ | |
| 1294 struct config *cfp, model; | |
| 1295 | |
| 1296 assert( currentend!=0 ); | |
| 1297 model.rp = rp; | |
| 1298 model.dot = dot; | |
| 1299 cfp = Configtable_find(&model); | |
| 1300 if( cfp==0 ){ | |
| 1301 cfp = newconfig(); | |
| 1302 cfp->rp = rp; | |
| 1303 cfp->dot = dot; | |
| 1304 cfp->fws = SetNew(); | |
| 1305 cfp->stp = 0; | |
| 1306 cfp->fplp = cfp->bplp = 0; | |
| 1307 cfp->next = 0; | |
| 1308 cfp->bp = 0; | |
| 1309 *currentend = cfp; | |
| 1310 currentend = &cfp->next; | |
| 1311 Configtable_insert(cfp); | |
| 1312 } | |
| 1313 return cfp; | |
| 1314 } | |
| 1315 | |
| 1316 /* Add a basis configuration to the configuration list */ | |
| 1317 struct config *Configlist_addbasis(struct rule *rp, int dot) | |
| 1318 { | |
| 1319 struct config *cfp, model; | |
| 1320 | |
| 1321 assert( basisend!=0 ); | |
| 1322 assert( currentend!=0 ); | |
| 1323 model.rp = rp; | |
| 1324 model.dot = dot; | |
| 1325 cfp = Configtable_find(&model); | |
| 1326 if( cfp==0 ){ | |
| 1327 cfp = newconfig(); | |
| 1328 cfp->rp = rp; | |
| 1329 cfp->dot = dot; | |
| 1330 cfp->fws = SetNew(); | |
| 1331 cfp->stp = 0; | |
| 1332 cfp->fplp = cfp->bplp = 0; | |
| 1333 cfp->next = 0; | |
| 1334 cfp->bp = 0; | |
| 1335 *currentend = cfp; | |
| 1336 currentend = &cfp->next; | |
| 1337 *basisend = cfp; | |
| 1338 basisend = &cfp->bp; | |
| 1339 Configtable_insert(cfp); | |
| 1340 } | |
| 1341 return cfp; | |
| 1342 } | |
| 1343 | |
| 1344 /* Compute the closure of the configuration list */ | |
| 1345 void Configlist_closure(struct lemon *lemp) | |
| 1346 { | |
| 1347 struct config *cfp, *newcfp; | |
| 1348 struct rule *rp, *newrp; | |
| 1349 struct symbol *sp, *xsp; | |
| 1350 int i, dot; | |
| 1351 | |
| 1352 assert( currentend!=0 ); | |
| 1353 for(cfp=current; cfp; cfp=cfp->next){ | |
| 1354 rp = cfp->rp; | |
| 1355 dot = cfp->dot; | |
| 1356 if( dot>=rp->nrhs ) continue; | |
| 1357 sp = rp->rhs[dot]; | |
| 1358 if( sp->type==NONTERMINAL ){ | |
| 1359 if( sp->rule==0 && sp!=lemp->errsym ){ | |
| 1360 ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.", | |
| 1361 sp->name); | |
| 1362 lemp->errorcnt++; | |
| 1363 } | |
| 1364 for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){ | |
| 1365 newcfp = Configlist_add(newrp,0); | |
| 1366 for(i=dot+1; i<rp->nrhs; i++){ | |
| 1367 xsp = rp->rhs[i]; | |
| 1368 if( xsp->type==TERMINAL ){ | |
| 1369 SetAdd(newcfp->fws,xsp->index); | |
| 1370 break; | |
| 1371 }else if( xsp->type==MULTITERMINAL ){ | |
| 1372 int k; | |
| 1373 for(k=0; k<xsp->nsubsym; k++){ | |
| 1374 SetAdd(newcfp->fws, xsp->subsym[k]->index); | |
| 1375 } | |
| 1376 break; | |
| 1377 }else{ | |
| 1378 SetUnion(newcfp->fws,xsp->firstset); | |
| 1379 if( xsp->lambda==LEMON_FALSE ) break; | |
| 1380 } | |
| 1381 } | |
| 1382 if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp); | |
| 1383 } | |
| 1384 } | |
| 1385 } | |
| 1386 return; | |
| 1387 } | |
| 1388 | |
| 1389 /* Sort the configuration list */ | |
| 1390 void Configlist_sort(){ | |
| 1391 current = (struct config*)msort((char*)current,(char**)&(current->next), | |
| 1392 Configcmp); | |
| 1393 currentend = 0; | |
| 1394 return; | |
| 1395 } | |
| 1396 | |
| 1397 /* Sort the basis configuration list */ | |
| 1398 void Configlist_sortbasis(){ | |
| 1399 basis = (struct config*)msort((char*)current,(char**)&(current->bp), | |
| 1400 Configcmp); | |
| 1401 basisend = 0; | |
| 1402 return; | |
| 1403 } | |
| 1404 | |
| 1405 /* Return a pointer to the head of the configuration list and | |
| 1406 ** reset the list */ | |
| 1407 struct config *Configlist_return(){ | |
| 1408 struct config *old; | |
| 1409 old = current; | |
| 1410 current = 0; | |
| 1411 currentend = 0; | |
| 1412 return old; | |
| 1413 } | |
| 1414 | |
| 1415 /* Return a pointer to the head of the configuration list and | |
| 1416 ** reset the list */ | |
| 1417 struct config *Configlist_basis(){ | |
| 1418 struct config *old; | |
| 1419 old = basis; | |
| 1420 basis = 0; | |
| 1421 basisend = 0; | |
| 1422 return old; | |
| 1423 } | |
| 1424 | |
| 1425 /* Free all elements of the given configuration list */ | |
| 1426 void Configlist_eat(struct config *cfp) | |
| 1427 { | |
| 1428 struct config *nextcfp; | |
| 1429 for(; cfp; cfp=nextcfp){ | |
| 1430 nextcfp = cfp->next; | |
| 1431 assert( cfp->fplp==0 ); | |
| 1432 assert( cfp->bplp==0 ); | |
| 1433 if( cfp->fws ) SetFree(cfp->fws); | |
| 1434 deleteconfig(cfp); | |
| 1435 } | |
| 1436 return; | |
| 1437 } | |
| 1438 /***************** From the file "error.c" *********************************/ | |
| 1439 /* | |
| 1440 ** Code for printing error message. | |
| 1441 */ | |
| 1442 | |
| 1443 void ErrorMsg(const char *filename, int lineno, const char *format, ...){ | |
| 1444 va_list ap; | |
| 1445 fprintf(stderr, "%s:%d: ", filename, lineno); | |
| 1446 va_start(ap, format); | |
| 1447 vfprintf(stderr,format,ap); | |
| 1448 va_end(ap); | |
| 1449 fprintf(stderr, "\n"); | |
| 1450 } | |
| 1451 /**************** From the file "main.c" ************************************/ | |
| 1452 /* | |
| 1453 ** Main program file for the LEMON parser generator. | |
| 1454 */ | |
| 1455 | |
| 1456 /* Report an out-of-memory condition and abort. This function | |
| 1457 ** is used mostly by the "MemoryCheck" macro in struct.h | |
| 1458 */ | |
| 1459 void memory_error(){ | |
| 1460 fprintf(stderr,"Out of memory. Aborting...\n"); | |
| 1461 exit(1); | |
| 1462 } | |
| 1463 | |
| 1464 static int nDefine = 0; /* Number of -D options on the command line */ | |
| 1465 static char **azDefine = 0; /* Name of the -D macros */ | |
| 1466 | |
| 1467 /* This routine is called with the argument to each -D command-line option. | |
| 1468 ** Add the macro defined to the azDefine array. | |
| 1469 */ | |
| 1470 static void handle_D_option(char *z){ | |
| 1471 char **paz; | |
| 1472 nDefine++; | |
| 1473 azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine); | |
| 1474 if( azDefine==0 ){ | |
| 1475 fprintf(stderr,"out of memory\n"); | |
| 1476 exit(1); | |
| 1477 } | |
| 1478 paz = &azDefine[nDefine-1]; | |
| 1479 *paz = (char *) malloc( lemonStrlen(z)+1 ); | |
| 1480 if( *paz==0 ){ | |
| 1481 fprintf(stderr,"out of memory\n"); | |
| 1482 exit(1); | |
| 1483 } | |
| 1484 lemon_strcpy(*paz, z); | |
| 1485 for(z=*paz; *z && *z!='='; z++){} | |
| 1486 *z = 0; | |
| 1487 } | |
| 1488 | |
| 1489 static char *user_templatename = NULL; | |
| 1490 static void handle_T_option(char *z){ | |
| 1491 user_templatename = (char *) malloc( lemonStrlen(z)+1 ); | |
| 1492 if( user_templatename==0 ){ | |
| 1493 memory_error(); | |
| 1494 } | |
| 1495 lemon_strcpy(user_templatename, z); | |
| 1496 } | |
| 1497 | |
| 1498 /* forward reference */ | |
| 1499 static const char *minimum_size_type(int lwr, int upr, int *pnByte); | |
| 1500 | |
| 1501 /* Print a single line of the "Parser Stats" output | |
| 1502 */ | |
| 1503 static void stats_line(const char *zLabel, int iValue){ | |
| 1504 int nLabel = lemonStrlen(zLabel); | |
| 1505 printf(" %s%.*s %5d\n", zLabel, | |
| 1506 35-nLabel, "................................", | |
| 1507 iValue); | |
| 1508 } | |
| 1509 | |
| 1510 /* The main program. Parse the command line and do it... */ | |
| 1511 int main(int argc, char **argv) | |
| 1512 { | |
| 1513 static int version = 0; | |
| 1514 static int rpflag = 0; | |
| 1515 static int basisflag = 0; | |
| 1516 static int compress = 0; | |
| 1517 static int quiet = 0; | |
| 1518 static int statistics = 0; | |
| 1519 static int mhflag = 0; | |
| 1520 static int nolinenosflag = 0; | |
| 1521 static int noResort = 0; | |
| 1522 static struct s_options options[] = { | |
| 1523 {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."}, | |
| 1524 {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."}, | |
| 1525 {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."}, | |
| 1526 {OPT_FSTR, "f", 0, "Ignored. (Placeholder for -f compiler options.)"}, | |
| 1527 {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."}, | |
| 1528 {OPT_FSTR, "I", 0, "Ignored. (Placeholder for '-I' compiler options.)"}, | |
| 1529 {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."}, | |
| 1530 {OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."}, | |
| 1531 {OPT_FSTR, "O", 0, "Ignored. (Placeholder for '-O' compiler options.)"}, | |
| 1532 {OPT_FLAG, "p", (char*)&showPrecedenceConflict, | |
| 1533 "Show conflicts resolved by precedence rules"}, | |
| 1534 {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."}, | |
| 1535 {OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"}, | |
| 1536 {OPT_FLAG, "s", (char*)&statistics, | |
| 1537 "Print parser stats to standard output."}, | |
| 1538 {OPT_FLAG, "x", (char*)&version, "Print the version number."}, | |
| 1539 {OPT_FSTR, "T", (char*)handle_T_option, "Specify a template file."}, | |
| 1540 {OPT_FSTR, "W", 0, "Ignored. (Placeholder for '-W' compiler options.)"}, | |
| 1541 {OPT_FLAG,0,0,0} | |
| 1542 }; | |
| 1543 int i; | |
| 1544 int exitcode; | |
| 1545 struct lemon lem; | |
| 1546 | |
| 1547 OptInit(argv,options,stderr); | |
| 1548 if( version ){ | |
| 1549 printf("Lemon version 1.0\n"); | |
| 1550 exit(0); | |
| 1551 } | |
| 1552 if( OptNArgs()!=1 ){ | |
| 1553 fprintf(stderr,"Exactly one filename argument is required.\n"); | |
| 1554 exit(1); | |
| 1555 } | |
| 1556 memset(&lem, 0, sizeof(lem)); | |
| 1557 lem.errorcnt = 0; | |
| 1558 | |
| 1559 /* Initialize the machine */ | |
| 1560 Strsafe_init(); | |
| 1561 Symbol_init(); | |
| 1562 State_init(); | |
| 1563 lem.argv0 = argv[0]; | |
| 1564 lem.filename = OptArg(0); | |
| 1565 lem.basisflag = basisflag; | |
| 1566 lem.nolinenosflag = nolinenosflag; | |
| 1567 Symbol_new("$"); | |
| 1568 lem.errsym = Symbol_new("error"); | |
| 1569 lem.errsym->useCnt = 0; | |
| 1570 | |
| 1571 /* Parse the input file */ | |
| 1572 Parse(&lem); | |
| 1573 if( lem.errorcnt ) exit(lem.errorcnt); | |
| 1574 if( lem.nrule==0 ){ | |
| 1575 fprintf(stderr,"Empty grammar.\n"); | |
| 1576 exit(1); | |
| 1577 } | |
| 1578 | |
| 1579 /* Count and index the symbols of the grammar */ | |
| 1580 Symbol_new("{default}"); | |
| 1581 lem.nsymbol = Symbol_count(); | |
| 1582 lem.symbols = Symbol_arrayof(); | |
| 1583 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i; | |
| 1584 qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp); | |
| 1585 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i; | |
| 1586 while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; } | |
| 1587 assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 ); | |
| 1588 lem.nsymbol = i - 1; | |
| 1589 for(i=1; ISUPPER(lem.symbols[i]->name[0]); i++); | |
| 1590 lem.nterminal = i; | |
| 1591 | |
| 1592 /* Generate a reprint of the grammar, if requested on the command line */ | |
| 1593 if( rpflag ){ | |
| 1594 Reprint(&lem); | |
| 1595 }else{ | |
| 1596 /* Initialize the size for all follow and first sets */ | |
| 1597 SetSize(lem.nterminal+1); | |
| 1598 | |
| 1599 /* Find the precedence for every production rule (that has one) */ | |
| 1600 FindRulePrecedences(&lem); | |
| 1601 | |
| 1602 /* Compute the lambda-nonterminals and the first-sets for every | |
| 1603 ** nonterminal */ | |
| 1604 FindFirstSets(&lem); | |
| 1605 | |
| 1606 /* Compute all LR(0) states. Also record follow-set propagation | |
| 1607 ** links so that the follow-set can be computed later */ | |
| 1608 lem.nstate = 0; | |
| 1609 FindStates(&lem); | |
| 1610 lem.sorted = State_arrayof(); | |
| 1611 | |
| 1612 /* Tie up loose ends on the propagation links */ | |
| 1613 FindLinks(&lem); | |
| 1614 | |
| 1615 /* Compute the follow set of every reducible configuration */ | |
| 1616 FindFollowSets(&lem); | |
| 1617 | |
| 1618 /* Compute the action tables */ | |
| 1619 FindActions(&lem); | |
| 1620 | |
| 1621 /* Compress the action tables */ | |
| 1622 if( compress==0 ) CompressTables(&lem); | |
| 1623 | |
| 1624 /* Reorder and renumber the states so that states with fewer choices | |
| 1625 ** occur at the end. This is an optimization that helps make the | |
| 1626 ** generated parser tables smaller. */ | |
| 1627 if( noResort==0 ) ResortStates(&lem); | |
| 1628 | |
| 1629 /* Generate a report of the parser generated. (the "y.output" file) */ | |
| 1630 if( !quiet ) ReportOutput(&lem); | |
| 1631 | |
| 1632 /* Generate the source code for the parser */ | |
| 1633 ReportTable(&lem, mhflag); | |
| 1634 | |
| 1635 /* Produce a header file for use by the scanner. (This step is | |
| 1636 ** omitted if the "-m" option is used because makeheaders will | |
| 1637 ** generate the file for us.) */ | |
| 1638 if( !mhflag ) ReportHeader(&lem); | |
| 1639 } | |
| 1640 if( statistics ){ | |
| 1641 printf("Parser statistics:\n"); | |
| 1642 stats_line("terminal symbols", lem.nterminal); | |
| 1643 stats_line("non-terminal symbols", lem.nsymbol - lem.nterminal); | |
| 1644 stats_line("total symbols", lem.nsymbol); | |
| 1645 stats_line("rules", lem.nrule); | |
| 1646 stats_line("states", lem.nxstate); | |
| 1647 stats_line("conflicts", lem.nconflict); | |
| 1648 stats_line("action table entries", lem.nactiontab); | |
| 1649 stats_line("total table size (bytes)", lem.tablesize); | |
| 1650 } | |
| 1651 if( lem.nconflict > 0 ){ | |
| 1652 fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict); | |
| 1653 } | |
| 1654 | |
| 1655 /* return 0 on success, 1 on failure. */ | |
| 1656 exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0; | |
| 1657 exit(exitcode); | |
| 1658 return (exitcode); | |
| 1659 } | |
| 1660 /******************** From the file "msort.c" *******************************/ | |
| 1661 /* | |
| 1662 ** A generic merge-sort program. | |
| 1663 ** | |
| 1664 ** USAGE: | |
| 1665 ** Let "ptr" be a pointer to some structure which is at the head of | |
| 1666 ** a null-terminated list. Then to sort the list call: | |
| 1667 ** | |
| 1668 ** ptr = msort(ptr,&(ptr->next),cmpfnc); | |
| 1669 ** | |
| 1670 ** In the above, "cmpfnc" is a pointer to a function which compares | |
| 1671 ** two instances of the structure and returns an integer, as in | |
| 1672 ** strcmp. The second argument is a pointer to the pointer to the | |
| 1673 ** second element of the linked list. This address is used to compute | |
| 1674 ** the offset to the "next" field within the structure. The offset to | |
| 1675 ** the "next" field must be constant for all structures in the list. | |
| 1676 ** | |
| 1677 ** The function returns a new pointer which is the head of the list | |
| 1678 ** after sorting. | |
| 1679 ** | |
| 1680 ** ALGORITHM: | |
| 1681 ** Merge-sort. | |
| 1682 */ | |
| 1683 | |
| 1684 /* | |
| 1685 ** Return a pointer to the next structure in the linked list. | |
| 1686 */ | |
| 1687 #define NEXT(A) (*(char**)(((char*)A)+offset)) | |
| 1688 | |
| 1689 /* | |
| 1690 ** Inputs: | |
| 1691 ** a: A sorted, null-terminated linked list. (May be null). | |
| 1692 ** b: A sorted, null-terminated linked list. (May be null). | |
| 1693 ** cmp: A pointer to the comparison function. | |
| 1694 ** offset: Offset in the structure to the "next" field. | |
| 1695 ** | |
| 1696 ** Return Value: | |
| 1697 ** A pointer to the head of a sorted list containing the elements | |
| 1698 ** of both a and b. | |
| 1699 ** | |
| 1700 ** Side effects: | |
| 1701 ** The "next" pointers for elements in the lists a and b are | |
| 1702 ** changed. | |
| 1703 */ | |
| 1704 static char *merge( | |
| 1705 char *a, | |
| 1706 char *b, | |
| 1707 int (*cmp)(const char*,const char*), | |
| 1708 int offset | |
| 1709 ){ | |
| 1710 char *ptr, *head; | |
| 1711 | |
| 1712 if( a==0 ){ | |
| 1713 head = b; | |
| 1714 }else if( b==0 ){ | |
| 1715 head = a; | |
| 1716 }else{ | |
| 1717 if( (*cmp)(a,b)<=0 ){ | |
| 1718 ptr = a; | |
| 1719 a = NEXT(a); | |
| 1720 }else{ | |
| 1721 ptr = b; | |
| 1722 b = NEXT(b); | |
| 1723 } | |
| 1724 head = ptr; | |
| 1725 while( a && b ){ | |
| 1726 if( (*cmp)(a,b)<=0 ){ | |
| 1727 NEXT(ptr) = a; | |
| 1728 ptr = a; | |
| 1729 a = NEXT(a); | |
| 1730 }else{ | |
| 1731 NEXT(ptr) = b; | |
| 1732 ptr = b; | |
| 1733 b = NEXT(b); | |
| 1734 } | |
| 1735 } | |
| 1736 if( a ) NEXT(ptr) = a; | |
| 1737 else NEXT(ptr) = b; | |
| 1738 } | |
| 1739 return head; | |
| 1740 } | |
| 1741 | |
| 1742 /* | |
| 1743 ** Inputs: | |
| 1744 ** list: Pointer to a singly-linked list of structures. | |
| 1745 ** next: Pointer to pointer to the second element of the list. | |
| 1746 ** cmp: A comparison function. | |
| 1747 ** | |
| 1748 ** Return Value: | |
| 1749 ** A pointer to the head of a sorted list containing the elements | |
| 1750 ** orginally in list. | |
| 1751 ** | |
| 1752 ** Side effects: | |
| 1753 ** The "next" pointers for elements in list are changed. | |
| 1754 */ | |
| 1755 #define LISTSIZE 30 | |
| 1756 static char *msort( | |
| 1757 char *list, | |
| 1758 char **next, | |
| 1759 int (*cmp)(const char*,const char*) | |
| 1760 ){ | |
| 1761 unsigned long offset; | |
| 1762 char *ep; | |
| 1763 char *set[LISTSIZE]; | |
| 1764 int i; | |
| 1765 offset = (unsigned long)((char*)next - (char*)list); | |
| 1766 for(i=0; i<LISTSIZE; i++) set[i] = 0; | |
| 1767 while( list ){ | |
| 1768 ep = list; | |
| 1769 list = NEXT(list); | |
| 1770 NEXT(ep) = 0; | |
| 1771 for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){ | |
| 1772 ep = merge(ep,set[i],cmp,offset); | |
| 1773 set[i] = 0; | |
| 1774 } | |
| 1775 set[i] = ep; | |
| 1776 } | |
| 1777 ep = 0; | |
| 1778 for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset); | |
| 1779 return ep; | |
| 1780 } | |
| 1781 /************************ From the file "option.c" **************************/ | |
| 1782 static char **argv; | |
| 1783 static struct s_options *op; | |
| 1784 static FILE *errstream; | |
| 1785 | |
| 1786 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0) | |
| 1787 | |
| 1788 /* | |
| 1789 ** Print the command line with a carrot pointing to the k-th character | |
| 1790 ** of the n-th field. | |
| 1791 */ | |
| 1792 static void errline(int n, int k, FILE *err) | |
| 1793 { | |
| 1794 int spcnt, i; | |
| 1795 if( argv[0] ) fprintf(err,"%s",argv[0]); | |
| 1796 spcnt = lemonStrlen(argv[0]) + 1; | |
| 1797 for(i=1; i<n && argv[i]; i++){ | |
| 1798 fprintf(err," %s",argv[i]); | |
| 1799 spcnt += lemonStrlen(argv[i])+1; | |
| 1800 } | |
| 1801 spcnt += k; | |
| 1802 for(; argv[i]; i++) fprintf(err," %s",argv[i]); | |
| 1803 if( spcnt<20 ){ | |
| 1804 fprintf(err,"\n%*s^-- here\n",spcnt,""); | |
| 1805 }else{ | |
| 1806 fprintf(err,"\n%*shere --^\n",spcnt-7,""); | |
| 1807 } | |
| 1808 } | |
| 1809 | |
| 1810 /* | |
| 1811 ** Return the index of the N-th non-switch argument. Return -1 | |
| 1812 ** if N is out of range. | |
| 1813 */ | |
| 1814 static int argindex(int n) | |
| 1815 { | |
| 1816 int i; | |
| 1817 int dashdash = 0; | |
| 1818 if( argv!=0 && *argv!=0 ){ | |
| 1819 for(i=1; argv[i]; i++){ | |
| 1820 if( dashdash || !ISOPT(argv[i]) ){ | |
| 1821 if( n==0 ) return i; | |
| 1822 n--; | |
| 1823 } | |
| 1824 if( strcmp(argv[i],"--")==0 ) dashdash = 1; | |
| 1825 } | |
| 1826 } | |
| 1827 return -1; | |
| 1828 } | |
| 1829 | |
| 1830 static char emsg[] = "Command line syntax error: "; | |
| 1831 | |
| 1832 /* | |
| 1833 ** Process a flag command line argument. | |
| 1834 */ | |
| 1835 static int handleflags(int i, FILE *err) | |
| 1836 { | |
| 1837 int v; | |
| 1838 int errcnt = 0; | |
| 1839 int j; | |
| 1840 for(j=0; op[j].label; j++){ | |
| 1841 if( strncmp(&argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break; | |
| 1842 } | |
| 1843 v = argv[i][0]=='-' ? 1 : 0; | |
| 1844 if( op[j].label==0 ){ | |
| 1845 if( err ){ | |
| 1846 fprintf(err,"%sundefined option.\n",emsg); | |
| 1847 errline(i,1,err); | |
| 1848 } | |
| 1849 errcnt++; | |
| 1850 }else if( op[j].arg==0 ){ | |
| 1851 /* Ignore this option */ | |
| 1852 }else if( op[j].type==OPT_FLAG ){ | |
| 1853 *((int*)op[j].arg) = v; | |
| 1854 }else if( op[j].type==OPT_FFLAG ){ | |
| 1855 (*(void(*)(int))(op[j].arg))(v); | |
| 1856 }else if( op[j].type==OPT_FSTR ){ | |
| 1857 (*(void(*)(char *))(op[j].arg))(&argv[i][2]); | |
| 1858 }else{ | |
| 1859 if( err ){ | |
| 1860 fprintf(err,"%smissing argument on switch.\n",emsg); | |
| 1861 errline(i,1,err); | |
| 1862 } | |
| 1863 errcnt++; | |
| 1864 } | |
| 1865 return errcnt; | |
| 1866 } | |
| 1867 | |
| 1868 /* | |
| 1869 ** Process a command line switch which has an argument. | |
| 1870 */ | |
| 1871 static int handleswitch(int i, FILE *err) | |
| 1872 { | |
| 1873 int lv = 0; | |
| 1874 double dv = 0.0; | |
| 1875 char *sv = 0, *end; | |
| 1876 char *cp; | |
| 1877 int j; | |
| 1878 int errcnt = 0; | |
| 1879 cp = strchr(argv[i],'='); | |
| 1880 assert( cp!=0 ); | |
| 1881 *cp = 0; | |
| 1882 for(j=0; op[j].label; j++){ | |
| 1883 if( strcmp(argv[i],op[j].label)==0 ) break; | |
| 1884 } | |
| 1885 *cp = '='; | |
| 1886 if( op[j].label==0 ){ | |
| 1887 if( err ){ | |
| 1888 fprintf(err,"%sundefined option.\n",emsg); | |
| 1889 errline(i,0,err); | |
| 1890 } | |
| 1891 errcnt++; | |
| 1892 }else{ | |
| 1893 cp++; | |
| 1894 switch( op[j].type ){ | |
| 1895 case OPT_FLAG: | |
| 1896 case OPT_FFLAG: | |
| 1897 if( err ){ | |
| 1898 fprintf(err,"%soption requires an argument.\n",emsg); | |
| 1899 errline(i,0,err); | |
| 1900 } | |
| 1901 errcnt++; | |
| 1902 break; | |
| 1903 case OPT_DBL: | |
| 1904 case OPT_FDBL: | |
| 1905 dv = strtod(cp,&end); | |
| 1906 if( *end ){ | |
| 1907 if( err ){ | |
| 1908 fprintf(err, | |
| 1909 "%sillegal character in floating-point argument.\n",emsg); | |
| 1910 errline(i,(int)((char*)end-(char*)argv[i]),err); | |
| 1911 } | |
| 1912 errcnt++; | |
| 1913 } | |
| 1914 break; | |
| 1915 case OPT_INT: | |
| 1916 case OPT_FINT: | |
| 1917 lv = strtol(cp,&end,0); | |
| 1918 if( *end ){ | |
| 1919 if( err ){ | |
| 1920 fprintf(err,"%sillegal character in integer argument.\n",emsg); | |
| 1921 errline(i,(int)((char*)end-(char*)argv[i]),err); | |
| 1922 } | |
| 1923 errcnt++; | |
| 1924 } | |
| 1925 break; | |
| 1926 case OPT_STR: | |
| 1927 case OPT_FSTR: | |
| 1928 sv = cp; | |
| 1929 break; | |
| 1930 } | |
| 1931 switch( op[j].type ){ | |
| 1932 case OPT_FLAG: | |
| 1933 case OPT_FFLAG: | |
| 1934 break; | |
| 1935 case OPT_DBL: | |
| 1936 *(double*)(op[j].arg) = dv; | |
| 1937 break; | |
| 1938 case OPT_FDBL: | |
| 1939 (*(void(*)(double))(op[j].arg))(dv); | |
| 1940 break; | |
| 1941 case OPT_INT: | |
| 1942 *(int*)(op[j].arg) = lv; | |
| 1943 break; | |
| 1944 case OPT_FINT: | |
| 1945 (*(void(*)(int))(op[j].arg))((int)lv); | |
| 1946 break; | |
| 1947 case OPT_STR: | |
| 1948 *(char**)(op[j].arg) = sv; | |
| 1949 break; | |
| 1950 case OPT_FSTR: | |
| 1951 (*(void(*)(char *))(op[j].arg))(sv); | |
| 1952 break; | |
| 1953 } | |
| 1954 } | |
| 1955 return errcnt; | |
| 1956 } | |
| 1957 | |
| 1958 int OptInit(char **a, struct s_options *o, FILE *err) | |
| 1959 { | |
| 1960 int errcnt = 0; | |
| 1961 argv = a; | |
| 1962 op = o; | |
| 1963 errstream = err; | |
| 1964 if( argv && *argv && op ){ | |
| 1965 int i; | |
| 1966 for(i=1; argv[i]; i++){ | |
| 1967 if( argv[i][0]=='+' || argv[i][0]=='-' ){ | |
| 1968 errcnt += handleflags(i,err); | |
| 1969 }else if( strchr(argv[i],'=') ){ | |
| 1970 errcnt += handleswitch(i,err); | |
| 1971 } | |
| 1972 } | |
| 1973 } | |
| 1974 if( errcnt>0 ){ | |
| 1975 fprintf(err,"Valid command line options for \"%s\" are:\n",*a); | |
| 1976 OptPrint(); | |
| 1977 exit(1); | |
| 1978 } | |
| 1979 return 0; | |
| 1980 } | |
| 1981 | |
| 1982 int OptNArgs(){ | |
| 1983 int cnt = 0; | |
| 1984 int dashdash = 0; | |
| 1985 int i; | |
| 1986 if( argv!=0 && argv[0]!=0 ){ | |
| 1987 for(i=1; argv[i]; i++){ | |
| 1988 if( dashdash || !ISOPT(argv[i]) ) cnt++; | |
| 1989 if( strcmp(argv[i],"--")==0 ) dashdash = 1; | |
| 1990 } | |
| 1991 } | |
| 1992 return cnt; | |
| 1993 } | |
| 1994 | |
| 1995 char *OptArg(int n) | |
| 1996 { | |
| 1997 int i; | |
| 1998 i = argindex(n); | |
| 1999 return i>=0 ? argv[i] : 0; | |
| 2000 } | |
| 2001 | |
| 2002 void OptErr(int n) | |
| 2003 { | |
| 2004 int i; | |
| 2005 i = argindex(n); | |
| 2006 if( i>=0 ) errline(i,0,errstream); | |
| 2007 } | |
| 2008 | |
| 2009 void OptPrint(){ | |
| 2010 int i; | |
| 2011 int max, len; | |
| 2012 max = 0; | |
| 2013 for(i=0; op[i].label; i++){ | |
| 2014 len = lemonStrlen(op[i].label) + 1; | |
| 2015 switch( op[i].type ){ | |
| 2016 case OPT_FLAG: | |
| 2017 case OPT_FFLAG: | |
| 2018 break; | |
| 2019 case OPT_INT: | |
| 2020 case OPT_FINT: | |
| 2021 len += 9; /* length of "<integer>" */ | |
| 2022 break; | |
| 2023 case OPT_DBL: | |
| 2024 case OPT_FDBL: | |
| 2025 len += 6; /* length of "<real>" */ | |
| 2026 break; | |
| 2027 case OPT_STR: | |
| 2028 case OPT_FSTR: | |
| 2029 len += 8; /* length of "<string>" */ | |
| 2030 break; | |
| 2031 } | |
| 2032 if( len>max ) max = len; | |
| 2033 } | |
| 2034 for(i=0; op[i].label; i++){ | |
| 2035 switch( op[i].type ){ | |
| 2036 case OPT_FLAG: | |
| 2037 case OPT_FFLAG: | |
| 2038 fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message); | |
| 2039 break; | |
| 2040 case OPT_INT: | |
| 2041 case OPT_FINT: | |
| 2042 fprintf(errstream," -%s<integer>%*s %s\n",op[i].label, | |
| 2043 (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message); | |
| 2044 break; | |
| 2045 case OPT_DBL: | |
| 2046 case OPT_FDBL: | |
| 2047 fprintf(errstream," -%s<real>%*s %s\n",op[i].label, | |
| 2048 (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message); | |
| 2049 break; | |
| 2050 case OPT_STR: | |
| 2051 case OPT_FSTR: | |
| 2052 fprintf(errstream," -%s<string>%*s %s\n",op[i].label, | |
| 2053 (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message); | |
| 2054 break; | |
| 2055 } | |
| 2056 } | |
| 2057 } | |
| 2058 /*********************** From the file "parse.c" ****************************/ | |
| 2059 /* | |
| 2060 ** Input file parser for the LEMON parser generator. | |
| 2061 */ | |
| 2062 | |
| 2063 /* The state of the parser */ | |
| 2064 enum e_state { | |
| 2065 INITIALIZE, | |
| 2066 WAITING_FOR_DECL_OR_RULE, | |
| 2067 WAITING_FOR_DECL_KEYWORD, | |
| 2068 WAITING_FOR_DECL_ARG, | |
| 2069 WAITING_FOR_PRECEDENCE_SYMBOL, | |
| 2070 WAITING_FOR_ARROW, | |
| 2071 IN_RHS, | |
| 2072 LHS_ALIAS_1, | |
| 2073 LHS_ALIAS_2, | |
| 2074 LHS_ALIAS_3, | |
| 2075 RHS_ALIAS_1, | |
| 2076 RHS_ALIAS_2, | |
| 2077 PRECEDENCE_MARK_1, | |
| 2078 PRECEDENCE_MARK_2, | |
| 2079 RESYNC_AFTER_RULE_ERROR, | |
| 2080 RESYNC_AFTER_DECL_ERROR, | |
| 2081 WAITING_FOR_DESTRUCTOR_SYMBOL, | |
| 2082 WAITING_FOR_DATATYPE_SYMBOL, | |
| 2083 WAITING_FOR_FALLBACK_ID, | |
| 2084 WAITING_FOR_WILDCARD_ID, | |
| 2085 WAITING_FOR_CLASS_ID, | |
| 2086 WAITING_FOR_CLASS_TOKEN | |
| 2087 }; | |
| 2088 struct pstate { | |
| 2089 char *filename; /* Name of the input file */ | |
| 2090 int tokenlineno; /* Linenumber at which current token starts */ | |
| 2091 int errorcnt; /* Number of errors so far */ | |
| 2092 char *tokenstart; /* Text of current token */ | |
| 2093 struct lemon *gp; /* Global state vector */ | |
| 2094 enum e_state state; /* The state of the parser */ | |
| 2095 struct symbol *fallback; /* The fallback token */ | |
| 2096 struct symbol *tkclass; /* Token class symbol */ | |
| 2097 struct symbol *lhs; /* Left-hand side of current rule */ | |
| 2098 const char *lhsalias; /* Alias for the LHS */ | |
| 2099 int nrhs; /* Number of right-hand side symbols seen */ | |
| 2100 struct symbol *rhs[MAXRHS]; /* RHS symbols */ | |
| 2101 const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */ | |
| 2102 struct rule *prevrule; /* Previous rule parsed */ | |
| 2103 const char *declkeyword; /* Keyword of a declaration */ | |
| 2104 char **declargslot; /* Where the declaration argument should be put */ | |
| 2105 int insertLineMacro; /* Add #line before declaration insert */ | |
| 2106 int *decllinenoslot; /* Where to write declaration line number */ | |
| 2107 enum e_assoc declassoc; /* Assign this association to decl arguments */ | |
| 2108 int preccounter; /* Assign this precedence to decl arguments */ | |
| 2109 struct rule *firstrule; /* Pointer to first rule in the grammar */ | |
| 2110 struct rule *lastrule; /* Pointer to the most recently parsed rule */ | |
| 2111 }; | |
| 2112 | |
| 2113 /* Parse a single token */ | |
| 2114 static void parseonetoken(struct pstate *psp) | |
| 2115 { | |
| 2116 const char *x; | |
| 2117 x = Strsafe(psp->tokenstart); /* Save the token permanently */ | |
| 2118 #if 0 | |
| 2119 printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno, | |
| 2120 x,psp->state); | |
| 2121 #endif | |
| 2122 switch( psp->state ){ | |
| 2123 case INITIALIZE: | |
| 2124 psp->prevrule = 0; | |
| 2125 psp->preccounter = 0; | |
| 2126 psp->firstrule = psp->lastrule = 0; | |
| 2127 psp->gp->nrule = 0; | |
| 2128 /* Fall thru to next case */ | |
| 2129 case WAITING_FOR_DECL_OR_RULE: | |
| 2130 if( x[0]=='%' ){ | |
| 2131 psp->state = WAITING_FOR_DECL_KEYWORD; | |
| 2132 }else if( ISLOWER(x[0]) ){ | |
| 2133 psp->lhs = Symbol_new(x); | |
| 2134 psp->nrhs = 0; | |
| 2135 psp->lhsalias = 0; | |
| 2136 psp->state = WAITING_FOR_ARROW; | |
| 2137 }else if( x[0]=='{' ){ | |
| 2138 if( psp->prevrule==0 ){ | |
| 2139 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2140 "There is no prior rule upon which to attach the code \ | |
| 2141 fragment which begins on this line."); | |
| 2142 psp->errorcnt++; | |
| 2143 }else if( psp->prevrule->code!=0 ){ | |
| 2144 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2145 "Code fragment beginning on this line is not the first \ | |
| 2146 to follow the previous rule."); | |
| 2147 psp->errorcnt++; | |
| 2148 }else{ | |
| 2149 psp->prevrule->line = psp->tokenlineno; | |
| 2150 psp->prevrule->code = &x[1]; | |
| 2151 } | |
| 2152 }else if( x[0]=='[' ){ | |
| 2153 psp->state = PRECEDENCE_MARK_1; | |
| 2154 }else{ | |
| 2155 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2156 "Token \"%s\" should be either \"%%\" or a nonterminal name.", | |
| 2157 x); | |
| 2158 psp->errorcnt++; | |
| 2159 } | |
| 2160 break; | |
| 2161 case PRECEDENCE_MARK_1: | |
| 2162 if( !ISUPPER(x[0]) ){ | |
| 2163 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2164 "The precedence symbol must be a terminal."); | |
| 2165 psp->errorcnt++; | |
| 2166 }else if( psp->prevrule==0 ){ | |
| 2167 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2168 "There is no prior rule to assign precedence \"[%s]\".",x); | |
| 2169 psp->errorcnt++; | |
| 2170 }else if( psp->prevrule->precsym!=0 ){ | |
| 2171 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2172 "Precedence mark on this line is not the first \ | |
| 2173 to follow the previous rule."); | |
| 2174 psp->errorcnt++; | |
| 2175 }else{ | |
| 2176 psp->prevrule->precsym = Symbol_new(x); | |
| 2177 } | |
| 2178 psp->state = PRECEDENCE_MARK_2; | |
| 2179 break; | |
| 2180 case PRECEDENCE_MARK_2: | |
| 2181 if( x[0]!=']' ){ | |
| 2182 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2183 "Missing \"]\" on precedence mark."); | |
| 2184 psp->errorcnt++; | |
| 2185 } | |
| 2186 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2187 break; | |
| 2188 case WAITING_FOR_ARROW: | |
| 2189 if( x[0]==':' && x[1]==':' && x[2]=='=' ){ | |
| 2190 psp->state = IN_RHS; | |
| 2191 }else if( x[0]=='(' ){ | |
| 2192 psp->state = LHS_ALIAS_1; | |
| 2193 }else{ | |
| 2194 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2195 "Expected to see a \":\" following the LHS symbol \"%s\".", | |
| 2196 psp->lhs->name); | |
| 2197 psp->errorcnt++; | |
| 2198 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2199 } | |
| 2200 break; | |
| 2201 case LHS_ALIAS_1: | |
| 2202 if( ISALPHA(x[0]) ){ | |
| 2203 psp->lhsalias = x; | |
| 2204 psp->state = LHS_ALIAS_2; | |
| 2205 }else{ | |
| 2206 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2207 "\"%s\" is not a valid alias for the LHS \"%s\"\n", | |
| 2208 x,psp->lhs->name); | |
| 2209 psp->errorcnt++; | |
| 2210 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2211 } | |
| 2212 break; | |
| 2213 case LHS_ALIAS_2: | |
| 2214 if( x[0]==')' ){ | |
| 2215 psp->state = LHS_ALIAS_3; | |
| 2216 }else{ | |
| 2217 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2218 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias); | |
| 2219 psp->errorcnt++; | |
| 2220 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2221 } | |
| 2222 break; | |
| 2223 case LHS_ALIAS_3: | |
| 2224 if( x[0]==':' && x[1]==':' && x[2]=='=' ){ | |
| 2225 psp->state = IN_RHS; | |
| 2226 }else{ | |
| 2227 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2228 "Missing \"->\" following: \"%s(%s)\".", | |
| 2229 psp->lhs->name,psp->lhsalias); | |
| 2230 psp->errorcnt++; | |
| 2231 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2232 } | |
| 2233 break; | |
| 2234 case IN_RHS: | |
| 2235 if( x[0]=='.' ){ | |
| 2236 struct rule *rp; | |
| 2237 rp = (struct rule *)calloc( sizeof(struct rule) + | |
| 2238 sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1); | |
| 2239 if( rp==0 ){ | |
| 2240 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2241 "Can't allocate enough memory for this rule."); | |
| 2242 psp->errorcnt++; | |
| 2243 psp->prevrule = 0; | |
| 2244 }else{ | |
| 2245 int i; | |
| 2246 rp->ruleline = psp->tokenlineno; | |
| 2247 rp->rhs = (struct symbol**)&rp[1]; | |
| 2248 rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]); | |
| 2249 for(i=0; i<psp->nrhs; i++){ | |
| 2250 rp->rhs[i] = psp->rhs[i]; | |
| 2251 rp->rhsalias[i] = psp->alias[i]; | |
| 2252 } | |
| 2253 rp->lhs = psp->lhs; | |
| 2254 rp->lhsalias = psp->lhsalias; | |
| 2255 rp->nrhs = psp->nrhs; | |
| 2256 rp->code = 0; | |
| 2257 rp->precsym = 0; | |
| 2258 rp->index = psp->gp->nrule++; | |
| 2259 rp->nextlhs = rp->lhs->rule; | |
| 2260 rp->lhs->rule = rp; | |
| 2261 rp->next = 0; | |
| 2262 if( psp->firstrule==0 ){ | |
| 2263 psp->firstrule = psp->lastrule = rp; | |
| 2264 }else{ | |
| 2265 psp->lastrule->next = rp; | |
| 2266 psp->lastrule = rp; | |
| 2267 } | |
| 2268 psp->prevrule = rp; | |
| 2269 } | |
| 2270 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2271 }else if( ISALPHA(x[0]) ){ | |
| 2272 if( psp->nrhs>=MAXRHS ){ | |
| 2273 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2274 "Too many symbols on RHS of rule beginning at \"%s\".", | |
| 2275 x); | |
| 2276 psp->errorcnt++; | |
| 2277 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2278 }else{ | |
| 2279 psp->rhs[psp->nrhs] = Symbol_new(x); | |
| 2280 psp->alias[psp->nrhs] = 0; | |
| 2281 psp->nrhs++; | |
| 2282 } | |
| 2283 }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 ){ | |
| 2284 struct symbol *msp = psp->rhs[psp->nrhs-1]; | |
| 2285 if( msp->type!=MULTITERMINAL ){ | |
| 2286 struct symbol *origsp = msp; | |
| 2287 msp = (struct symbol *) calloc(1,sizeof(*msp)); | |
| 2288 memset(msp, 0, sizeof(*msp)); | |
| 2289 msp->type = MULTITERMINAL; | |
| 2290 msp->nsubsym = 1; | |
| 2291 msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*)); | |
| 2292 msp->subsym[0] = origsp; | |
| 2293 msp->name = origsp->name; | |
| 2294 psp->rhs[psp->nrhs-1] = msp; | |
| 2295 } | |
| 2296 msp->nsubsym++; | |
| 2297 msp->subsym = (struct symbol **) realloc(msp->subsym, | |
| 2298 sizeof(struct symbol*)*msp->nsubsym); | |
| 2299 msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]); | |
| 2300 if( ISLOWER(x[1]) || ISLOWER(msp->subsym[0]->name[0]) ){ | |
| 2301 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2302 "Cannot form a compound containing a non-terminal"); | |
| 2303 psp->errorcnt++; | |
| 2304 } | |
| 2305 }else if( x[0]=='(' && psp->nrhs>0 ){ | |
| 2306 psp->state = RHS_ALIAS_1; | |
| 2307 }else{ | |
| 2308 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2309 "Illegal character on RHS of rule: \"%s\".",x); | |
| 2310 psp->errorcnt++; | |
| 2311 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2312 } | |
| 2313 break; | |
| 2314 case RHS_ALIAS_1: | |
| 2315 if( ISALPHA(x[0]) ){ | |
| 2316 psp->alias[psp->nrhs-1] = x; | |
| 2317 psp->state = RHS_ALIAS_2; | |
| 2318 }else{ | |
| 2319 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2320 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n", | |
| 2321 x,psp->rhs[psp->nrhs-1]->name); | |
| 2322 psp->errorcnt++; | |
| 2323 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2324 } | |
| 2325 break; | |
| 2326 case RHS_ALIAS_2: | |
| 2327 if( x[0]==')' ){ | |
| 2328 psp->state = IN_RHS; | |
| 2329 }else{ | |
| 2330 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2331 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias); | |
| 2332 psp->errorcnt++; | |
| 2333 psp->state = RESYNC_AFTER_RULE_ERROR; | |
| 2334 } | |
| 2335 break; | |
| 2336 case WAITING_FOR_DECL_KEYWORD: | |
| 2337 if( ISALPHA(x[0]) ){ | |
| 2338 psp->declkeyword = x; | |
| 2339 psp->declargslot = 0; | |
| 2340 psp->decllinenoslot = 0; | |
| 2341 psp->insertLineMacro = 1; | |
| 2342 psp->state = WAITING_FOR_DECL_ARG; | |
| 2343 if( strcmp(x,"name")==0 ){ | |
| 2344 psp->declargslot = &(psp->gp->name); | |
| 2345 psp->insertLineMacro = 0; | |
| 2346 }else if( strcmp(x,"include")==0 ){ | |
| 2347 psp->declargslot = &(psp->gp->include); | |
| 2348 }else if( strcmp(x,"code")==0 ){ | |
| 2349 psp->declargslot = &(psp->gp->extracode); | |
| 2350 }else if( strcmp(x,"token_destructor")==0 ){ | |
| 2351 psp->declargslot = &psp->gp->tokendest; | |
| 2352 }else if( strcmp(x,"default_destructor")==0 ){ | |
| 2353 psp->declargslot = &psp->gp->vardest; | |
| 2354 }else if( strcmp(x,"token_prefix")==0 ){ | |
| 2355 psp->declargslot = &psp->gp->tokenprefix; | |
| 2356 psp->insertLineMacro = 0; | |
| 2357 }else if( strcmp(x,"syntax_error")==0 ){ | |
| 2358 psp->declargslot = &(psp->gp->error); | |
| 2359 }else if( strcmp(x,"parse_accept")==0 ){ | |
| 2360 psp->declargslot = &(psp->gp->accept); | |
| 2361 }else if( strcmp(x,"parse_failure")==0 ){ | |
| 2362 psp->declargslot = &(psp->gp->failure); | |
| 2363 }else if( strcmp(x,"stack_overflow")==0 ){ | |
| 2364 psp->declargslot = &(psp->gp->overflow); | |
| 2365 }else if( strcmp(x,"extra_argument")==0 ){ | |
| 2366 psp->declargslot = &(psp->gp->arg); | |
| 2367 psp->insertLineMacro = 0; | |
| 2368 }else if( strcmp(x,"token_type")==0 ){ | |
| 2369 psp->declargslot = &(psp->gp->tokentype); | |
| 2370 psp->insertLineMacro = 0; | |
| 2371 }else if( strcmp(x,"default_type")==0 ){ | |
| 2372 psp->declargslot = &(psp->gp->vartype); | |
| 2373 psp->insertLineMacro = 0; | |
| 2374 }else if( strcmp(x,"stack_size")==0 ){ | |
| 2375 psp->declargslot = &(psp->gp->stacksize); | |
| 2376 psp->insertLineMacro = 0; | |
| 2377 }else if( strcmp(x,"start_symbol")==0 ){ | |
| 2378 psp->declargslot = &(psp->gp->start); | |
| 2379 psp->insertLineMacro = 0; | |
| 2380 }else if( strcmp(x,"left")==0 ){ | |
| 2381 psp->preccounter++; | |
| 2382 psp->declassoc = LEFT; | |
| 2383 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL; | |
| 2384 }else if( strcmp(x,"right")==0 ){ | |
| 2385 psp->preccounter++; | |
| 2386 psp->declassoc = RIGHT; | |
| 2387 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL; | |
| 2388 }else if( strcmp(x,"nonassoc")==0 ){ | |
| 2389 psp->preccounter++; | |
| 2390 psp->declassoc = NONE; | |
| 2391 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL; | |
| 2392 }else if( strcmp(x,"destructor")==0 ){ | |
| 2393 psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL; | |
| 2394 }else if( strcmp(x,"type")==0 ){ | |
| 2395 psp->state = WAITING_FOR_DATATYPE_SYMBOL; | |
| 2396 }else if( strcmp(x,"fallback")==0 ){ | |
| 2397 psp->fallback = 0; | |
| 2398 psp->state = WAITING_FOR_FALLBACK_ID; | |
| 2399 }else if( strcmp(x,"wildcard")==0 ){ | |
| 2400 psp->state = WAITING_FOR_WILDCARD_ID; | |
| 2401 }else if( strcmp(x,"token_class")==0 ){ | |
| 2402 psp->state = WAITING_FOR_CLASS_ID; | |
| 2403 }else{ | |
| 2404 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2405 "Unknown declaration keyword: \"%%%s\".",x); | |
| 2406 psp->errorcnt++; | |
| 2407 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2408 } | |
| 2409 }else{ | |
| 2410 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2411 "Illegal declaration keyword: \"%s\".",x); | |
| 2412 psp->errorcnt++; | |
| 2413 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2414 } | |
| 2415 break; | |
| 2416 case WAITING_FOR_DESTRUCTOR_SYMBOL: | |
| 2417 if( !ISALPHA(x[0]) ){ | |
| 2418 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2419 "Symbol name missing after %%destructor keyword"); | |
| 2420 psp->errorcnt++; | |
| 2421 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2422 }else{ | |
| 2423 struct symbol *sp = Symbol_new(x); | |
| 2424 psp->declargslot = &sp->destructor; | |
| 2425 psp->decllinenoslot = &sp->destLineno; | |
| 2426 psp->insertLineMacro = 1; | |
| 2427 psp->state = WAITING_FOR_DECL_ARG; | |
| 2428 } | |
| 2429 break; | |
| 2430 case WAITING_FOR_DATATYPE_SYMBOL: | |
| 2431 if( !ISALPHA(x[0]) ){ | |
| 2432 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2433 "Symbol name missing after %%type keyword"); | |
| 2434 psp->errorcnt++; | |
| 2435 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2436 }else{ | |
| 2437 struct symbol *sp = Symbol_find(x); | |
| 2438 if((sp) && (sp->datatype)){ | |
| 2439 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2440 "Symbol %%type \"%s\" already defined", x); | |
| 2441 psp->errorcnt++; | |
| 2442 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2443 }else{ | |
| 2444 if (!sp){ | |
| 2445 sp = Symbol_new(x); | |
| 2446 } | |
| 2447 psp->declargslot = &sp->datatype; | |
| 2448 psp->insertLineMacro = 0; | |
| 2449 psp->state = WAITING_FOR_DECL_ARG; | |
| 2450 } | |
| 2451 } | |
| 2452 break; | |
| 2453 case WAITING_FOR_PRECEDENCE_SYMBOL: | |
| 2454 if( x[0]=='.' ){ | |
| 2455 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2456 }else if( ISUPPER(x[0]) ){ | |
| 2457 struct symbol *sp; | |
| 2458 sp = Symbol_new(x); | |
| 2459 if( sp->prec>=0 ){ | |
| 2460 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2461 "Symbol \"%s\" has already be given a precedence.",x); | |
| 2462 psp->errorcnt++; | |
| 2463 }else{ | |
| 2464 sp->prec = psp->preccounter; | |
| 2465 sp->assoc = psp->declassoc; | |
| 2466 } | |
| 2467 }else{ | |
| 2468 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2469 "Can't assign a precedence to \"%s\".",x); | |
| 2470 psp->errorcnt++; | |
| 2471 } | |
| 2472 break; | |
| 2473 case WAITING_FOR_DECL_ARG: | |
| 2474 if( x[0]=='{' || x[0]=='\"' || ISALNUM(x[0]) ){ | |
| 2475 const char *zOld, *zNew; | |
| 2476 char *zBuf, *z; | |
| 2477 int nOld, n, nLine = 0, nNew, nBack; | |
| 2478 int addLineMacro; | |
| 2479 char zLine[50]; | |
| 2480 zNew = x; | |
| 2481 if( zNew[0]=='"' || zNew[0]=='{' ) zNew++; | |
| 2482 nNew = lemonStrlen(zNew); | |
| 2483 if( *psp->declargslot ){ | |
| 2484 zOld = *psp->declargslot; | |
| 2485 }else{ | |
| 2486 zOld = ""; | |
| 2487 } | |
| 2488 nOld = lemonStrlen(zOld); | |
| 2489 n = nOld + nNew + 20; | |
| 2490 addLineMacro = !psp->gp->nolinenosflag && psp->insertLineMacro && | |
| 2491 (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0); | |
| 2492 if( addLineMacro ){ | |
| 2493 for(z=psp->filename, nBack=0; *z; z++){ | |
| 2494 if( *z=='\\' ) nBack++; | |
| 2495 } | |
| 2496 lemon_sprintf(zLine, "#line %d ", psp->tokenlineno); | |
| 2497 nLine = lemonStrlen(zLine); | |
| 2498 n += nLine + lemonStrlen(psp->filename) + nBack; | |
| 2499 } | |
| 2500 *psp->declargslot = (char *) realloc(*psp->declargslot, n); | |
| 2501 zBuf = *psp->declargslot + nOld; | |
| 2502 if( addLineMacro ){ | |
| 2503 if( nOld && zBuf[-1]!='\n' ){ | |
| 2504 *(zBuf++) = '\n'; | |
| 2505 } | |
| 2506 memcpy(zBuf, zLine, nLine); | |
| 2507 zBuf += nLine; | |
| 2508 *(zBuf++) = '"'; | |
| 2509 for(z=psp->filename; *z; z++){ | |
| 2510 if( *z=='\\' ){ | |
| 2511 *(zBuf++) = '\\'; | |
| 2512 } | |
| 2513 *(zBuf++) = *z; | |
| 2514 } | |
| 2515 *(zBuf++) = '"'; | |
| 2516 *(zBuf++) = '\n'; | |
| 2517 } | |
| 2518 if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){ | |
| 2519 psp->decllinenoslot[0] = psp->tokenlineno; | |
| 2520 } | |
| 2521 memcpy(zBuf, zNew, nNew); | |
| 2522 zBuf += nNew; | |
| 2523 *zBuf = 0; | |
| 2524 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2525 }else{ | |
| 2526 ErrorMsg(psp->filename,psp->tokenlineno, | |
| 2527 "Illegal argument to %%%s: %s",psp->declkeyword,x); | |
| 2528 psp->errorcnt++; | |
| 2529 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2530 } | |
| 2531 break; | |
| 2532 case WAITING_FOR_FALLBACK_ID: | |
| 2533 if( x[0]=='.' ){ | |
| 2534 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2535 }else if( !ISUPPER(x[0]) ){ | |
| 2536 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2537 "%%fallback argument \"%s\" should be a token", x); | |
| 2538 psp->errorcnt++; | |
| 2539 }else{ | |
| 2540 struct symbol *sp = Symbol_new(x); | |
| 2541 if( psp->fallback==0 ){ | |
| 2542 psp->fallback = sp; | |
| 2543 }else if( sp->fallback ){ | |
| 2544 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2545 "More than one fallback assigned to token %s", x); | |
| 2546 psp->errorcnt++; | |
| 2547 }else{ | |
| 2548 sp->fallback = psp->fallback; | |
| 2549 psp->gp->has_fallback = 1; | |
| 2550 } | |
| 2551 } | |
| 2552 break; | |
| 2553 case WAITING_FOR_WILDCARD_ID: | |
| 2554 if( x[0]=='.' ){ | |
| 2555 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2556 }else if( !ISUPPER(x[0]) ){ | |
| 2557 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2558 "%%wildcard argument \"%s\" should be a token", x); | |
| 2559 psp->errorcnt++; | |
| 2560 }else{ | |
| 2561 struct symbol *sp = Symbol_new(x); | |
| 2562 if( psp->gp->wildcard==0 ){ | |
| 2563 psp->gp->wildcard = sp; | |
| 2564 }else{ | |
| 2565 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2566 "Extra wildcard to token: %s", x); | |
| 2567 psp->errorcnt++; | |
| 2568 } | |
| 2569 } | |
| 2570 break; | |
| 2571 case WAITING_FOR_CLASS_ID: | |
| 2572 if( !ISLOWER(x[0]) ){ | |
| 2573 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2574 "%%token_class must be followed by an identifier: ", x); | |
| 2575 psp->errorcnt++; | |
| 2576 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2577 }else if( Symbol_find(x) ){ | |
| 2578 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2579 "Symbol \"%s\" already used", x); | |
| 2580 psp->errorcnt++; | |
| 2581 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2582 }else{ | |
| 2583 psp->tkclass = Symbol_new(x); | |
| 2584 psp->tkclass->type = MULTITERMINAL; | |
| 2585 psp->state = WAITING_FOR_CLASS_TOKEN; | |
| 2586 } | |
| 2587 break; | |
| 2588 case WAITING_FOR_CLASS_TOKEN: | |
| 2589 if( x[0]=='.' ){ | |
| 2590 psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2591 }else if( ISUPPER(x[0]) || ((x[0]=='|' || x[0]=='/') && ISUPPER(x[1])) ){ | |
| 2592 struct symbol *msp = psp->tkclass; | |
| 2593 msp->nsubsym++; | |
| 2594 msp->subsym = (struct symbol **) realloc(msp->subsym, | |
| 2595 sizeof(struct symbol*)*msp->nsubsym); | |
| 2596 if( !ISUPPER(x[0]) ) x++; | |
| 2597 msp->subsym[msp->nsubsym-1] = Symbol_new(x); | |
| 2598 }else{ | |
| 2599 ErrorMsg(psp->filename, psp->tokenlineno, | |
| 2600 "%%token_class argument \"%s\" should be a token", x); | |
| 2601 psp->errorcnt++; | |
| 2602 psp->state = RESYNC_AFTER_DECL_ERROR; | |
| 2603 } | |
| 2604 break; | |
| 2605 case RESYNC_AFTER_RULE_ERROR: | |
| 2606 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2607 ** break; */ | |
| 2608 case RESYNC_AFTER_DECL_ERROR: | |
| 2609 if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE; | |
| 2610 if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD; | |
| 2611 break; | |
| 2612 } | |
| 2613 } | |
| 2614 | |
| 2615 /* Run the preprocessor over the input file text. The global variables | |
| 2616 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined | |
| 2617 ** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and | |
| 2618 ** comments them out. Text in between is also commented out as appropriate. | |
| 2619 */ | |
| 2620 static void preprocess_input(char *z){ | |
| 2621 int i, j, k, n; | |
| 2622 int exclude = 0; | |
| 2623 int start = 0; | |
| 2624 int lineno = 1; | |
| 2625 int start_lineno = 1; | |
| 2626 for(i=0; z[i]; i++){ | |
| 2627 if( z[i]=='\n' ) lineno++; | |
| 2628 if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue; | |
| 2629 if( strncmp(&z[i],"%endif",6)==0 && ISSPACE(z[i+6]) ){ | |
| 2630 if( exclude ){ | |
| 2631 exclude--; | |
| 2632 if( exclude==0 ){ | |
| 2633 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' '; | |
| 2634 } | |
| 2635 } | |
| 2636 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' '; | |
| 2637 }else if( (strncmp(&z[i],"%ifdef",6)==0 && ISSPACE(z[i+6])) | |
| 2638 || (strncmp(&z[i],"%ifndef",7)==0 && ISSPACE(z[i+7])) ){ | |
| 2639 if( exclude ){ | |
| 2640 exclude++; | |
| 2641 }else{ | |
| 2642 for(j=i+7; ISSPACE(z[j]); j++){} | |
| 2643 for(n=0; z[j+n] && !ISSPACE(z[j+n]); n++){} | |
| 2644 exclude = 1; | |
| 2645 for(k=0; k<nDefine; k++){ | |
| 2646 if( strncmp(azDefine[k],&z[j],n)==0 && lemonStrlen(azDefine[k])==n ){ | |
| 2647 exclude = 0; | |
| 2648 break; | |
| 2649 } | |
| 2650 } | |
| 2651 if( z[i+3]=='n' ) exclude = !exclude; | |
| 2652 if( exclude ){ | |
| 2653 start = i; | |
| 2654 start_lineno = lineno; | |
| 2655 } | |
| 2656 } | |
| 2657 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' '; | |
| 2658 } | |
| 2659 } | |
| 2660 if( exclude ){ | |
| 2661 fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno); | |
| 2662 exit(1); | |
| 2663 } | |
| 2664 } | |
| 2665 | |
| 2666 /* In spite of its name, this function is really a scanner. It read | |
| 2667 ** in the entire input file (all at once) then tokenizes it. Each | |
| 2668 ** token is passed to the function "parseonetoken" which builds all | |
| 2669 ** the appropriate data structures in the global state vector "gp". | |
| 2670 */ | |
| 2671 void Parse(struct lemon *gp) | |
| 2672 { | |
| 2673 struct pstate ps; | |
| 2674 FILE *fp; | |
| 2675 char *filebuf; | |
| 2676 unsigned int filesize; | |
| 2677 int lineno; | |
| 2678 int c; | |
| 2679 char *cp, *nextcp; | |
| 2680 int startline = 0; | |
| 2681 | |
| 2682 memset(&ps, '\0', sizeof(ps)); | |
| 2683 ps.gp = gp; | |
| 2684 ps.filename = gp->filename; | |
| 2685 ps.errorcnt = 0; | |
| 2686 ps.state = INITIALIZE; | |
| 2687 | |
| 2688 /* Begin by reading the input file */ | |
| 2689 fp = fopen(ps.filename,"rb"); | |
| 2690 if( fp==0 ){ | |
| 2691 ErrorMsg(ps.filename,0,"Can't open this file for reading."); | |
| 2692 gp->errorcnt++; | |
| 2693 return; | |
| 2694 } | |
| 2695 fseek(fp,0,2); | |
| 2696 filesize = ftell(fp); | |
| 2697 rewind(fp); | |
| 2698 filebuf = (char *)malloc( filesize+1 ); | |
| 2699 if( filesize>100000000 || filebuf==0 ){ | |
| 2700 ErrorMsg(ps.filename,0,"Input file too large."); | |
| 2701 gp->errorcnt++; | |
| 2702 fclose(fp); | |
| 2703 return; | |
| 2704 } | |
| 2705 if( fread(filebuf,1,filesize,fp)!=filesize ){ | |
| 2706 ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.", | |
| 2707 filesize); | |
| 2708 free(filebuf); | |
| 2709 gp->errorcnt++; | |
| 2710 fclose(fp); | |
| 2711 return; | |
| 2712 } | |
| 2713 fclose(fp); | |
| 2714 filebuf[filesize] = 0; | |
| 2715 | |
| 2716 /* Make an initial pass through the file to handle %ifdef and %ifndef */ | |
| 2717 preprocess_input(filebuf); | |
| 2718 | |
| 2719 /* Now scan the text of the input file */ | |
| 2720 lineno = 1; | |
| 2721 for(cp=filebuf; (c= *cp)!=0; ){ | |
| 2722 if( c=='\n' ) lineno++; /* Keep track of the line number */ | |
| 2723 if( ISSPACE(c) ){ cp++; continue; } /* Skip all white space */ | |
| 2724 if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */ | |
| 2725 cp+=2; | |
| 2726 while( (c= *cp)!=0 && c!='\n' ) cp++; | |
| 2727 continue; | |
| 2728 } | |
| 2729 if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */ | |
| 2730 cp+=2; | |
| 2731 while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){ | |
| 2732 if( c=='\n' ) lineno++; | |
| 2733 cp++; | |
| 2734 } | |
| 2735 if( c ) cp++; | |
| 2736 continue; | |
| 2737 } | |
| 2738 ps.tokenstart = cp; /* Mark the beginning of the token */ | |
| 2739 ps.tokenlineno = lineno; /* Linenumber on which token begins */ | |
| 2740 if( c=='\"' ){ /* String literals */ | |
| 2741 cp++; | |
| 2742 while( (c= *cp)!=0 && c!='\"' ){ | |
| 2743 if( c=='\n' ) lineno++; | |
| 2744 cp++; | |
| 2745 } | |
| 2746 if( c==0 ){ | |
| 2747 ErrorMsg(ps.filename,startline, | |
| 2748 "String starting on this line is not terminated before the end of the file."); | |
| 2749 ps.errorcnt++; | |
| 2750 nextcp = cp; | |
| 2751 }else{ | |
| 2752 nextcp = cp+1; | |
| 2753 } | |
| 2754 }else if( c=='{' ){ /* A block of C code */ | |
| 2755 int level; | |
| 2756 cp++; | |
| 2757 for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){ | |
| 2758 if( c=='\n' ) lineno++; | |
| 2759 else if( c=='{' ) level++; | |
| 2760 else if( c=='}' ) level--; | |
| 2761 else if( c=='/' && cp[1]=='*' ){ /* Skip comments */ | |
| 2762 int prevc; | |
| 2763 cp = &cp[2]; | |
| 2764 prevc = 0; | |
| 2765 while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){ | |
| 2766 if( c=='\n' ) lineno++; | |
| 2767 prevc = c; | |
| 2768 cp++; | |
| 2769 } | |
| 2770 }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */ | |
| 2771 cp = &cp[2]; | |
| 2772 while( (c= *cp)!=0 && c!='\n' ) cp++; | |
| 2773 if( c ) lineno++; | |
| 2774 }else if( c=='\'' || c=='\"' ){ /* String a character literals */ | |
| 2775 int startchar, prevc; | |
| 2776 startchar = c; | |
| 2777 prevc = 0; | |
| 2778 for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){ | |
| 2779 if( c=='\n' ) lineno++; | |
| 2780 if( prevc=='\\' ) prevc = 0; | |
| 2781 else prevc = c; | |
| 2782 } | |
| 2783 } | |
| 2784 } | |
| 2785 if( c==0 ){ | |
| 2786 ErrorMsg(ps.filename,ps.tokenlineno, | |
| 2787 "C code starting on this line is not terminated before the end of the file."); | |
| 2788 ps.errorcnt++; | |
| 2789 nextcp = cp; | |
| 2790 }else{ | |
| 2791 nextcp = cp+1; | |
| 2792 } | |
| 2793 }else if( ISALNUM(c) ){ /* Identifiers */ | |
| 2794 while( (c= *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++; | |
| 2795 nextcp = cp; | |
| 2796 }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */ | |
| 2797 cp += 3; | |
| 2798 nextcp = cp; | |
| 2799 }else if( (c=='/' || c=='|') && ISALPHA(cp[1]) ){ | |
| 2800 cp += 2; | |
| 2801 while( (c = *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++; | |
| 2802 nextcp = cp; | |
| 2803 }else{ /* All other (one character) operators */ | |
| 2804 cp++; | |
| 2805 nextcp = cp; | |
| 2806 } | |
| 2807 c = *cp; | |
| 2808 *cp = 0; /* Null terminate the token */ | |
| 2809 parseonetoken(&ps); /* Parse the token */ | |
| 2810 *cp = (char)c; /* Restore the buffer */ | |
| 2811 cp = nextcp; | |
| 2812 } | |
| 2813 free(filebuf); /* Release the buffer after parsing */ | |
| 2814 gp->rule = ps.firstrule; | |
| 2815 gp->errorcnt = ps.errorcnt; | |
| 2816 } | |
| 2817 /*************************** From the file "plink.c" *********************/ | |
| 2818 /* | |
| 2819 ** Routines processing configuration follow-set propagation links | |
| 2820 ** in the LEMON parser generator. | |
| 2821 */ | |
| 2822 static struct plink *plink_freelist = 0; | |
| 2823 | |
| 2824 /* Allocate a new plink */ | |
| 2825 struct plink *Plink_new(){ | |
| 2826 struct plink *newlink; | |
| 2827 | |
| 2828 if( plink_freelist==0 ){ | |
| 2829 int i; | |
| 2830 int amt = 100; | |
| 2831 plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) ); | |
| 2832 if( plink_freelist==0 ){ | |
| 2833 fprintf(stderr, | |
| 2834 "Unable to allocate memory for a new follow-set propagation link.\n"); | |
| 2835 exit(1); | |
| 2836 } | |
| 2837 for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1]; | |
| 2838 plink_freelist[amt-1].next = 0; | |
| 2839 } | |
| 2840 newlink = plink_freelist; | |
| 2841 plink_freelist = plink_freelist->next; | |
| 2842 return newlink; | |
| 2843 } | |
| 2844 | |
| 2845 /* Add a plink to a plink list */ | |
| 2846 void Plink_add(struct plink **plpp, struct config *cfp) | |
| 2847 { | |
| 2848 struct plink *newlink; | |
| 2849 newlink = Plink_new(); | |
| 2850 newlink->next = *plpp; | |
| 2851 *plpp = newlink; | |
| 2852 newlink->cfp = cfp; | |
| 2853 } | |
| 2854 | |
| 2855 /* Transfer every plink on the list "from" to the list "to" */ | |
| 2856 void Plink_copy(struct plink **to, struct plink *from) | |
| 2857 { | |
| 2858 struct plink *nextpl; | |
| 2859 while( from ){ | |
| 2860 nextpl = from->next; | |
| 2861 from->next = *to; | |
| 2862 *to = from; | |
| 2863 from = nextpl; | |
| 2864 } | |
| 2865 } | |
| 2866 | |
| 2867 /* Delete every plink on the list */ | |
| 2868 void Plink_delete(struct plink *plp) | |
| 2869 { | |
| 2870 struct plink *nextpl; | |
| 2871 | |
| 2872 while( plp ){ | |
| 2873 nextpl = plp->next; | |
| 2874 plp->next = plink_freelist; | |
| 2875 plink_freelist = plp; | |
| 2876 plp = nextpl; | |
| 2877 } | |
| 2878 } | |
| 2879 /*********************** From the file "report.c" **************************/ | |
| 2880 /* | |
| 2881 ** Procedures for generating reports and tables in the LEMON parser generator. | |
| 2882 */ | |
| 2883 | |
| 2884 /* Generate a filename with the given suffix. Space to hold the | |
| 2885 ** name comes from malloc() and must be freed by the calling | |
| 2886 ** function. | |
| 2887 */ | |
| 2888 PRIVATE char *file_makename(struct lemon *lemp, const char *suffix) | |
| 2889 { | |
| 2890 char *name; | |
| 2891 char *cp; | |
| 2892 | |
| 2893 name = (char*)malloc( lemonStrlen(lemp->filename) + lemonStrlen(suffix) + 5 ); | |
| 2894 if( name==0 ){ | |
| 2895 fprintf(stderr,"Can't allocate space for a filename.\n"); | |
| 2896 exit(1); | |
| 2897 } | |
| 2898 lemon_strcpy(name,lemp->filename); | |
| 2899 cp = strrchr(name,'.'); | |
| 2900 if( cp ) *cp = 0; | |
| 2901 lemon_strcat(name,suffix); | |
| 2902 return name; | |
| 2903 } | |
| 2904 | |
| 2905 /* Open a file with a name based on the name of the input file, | |
| 2906 ** but with a different (specified) suffix, and return a pointer | |
| 2907 ** to the stream */ | |
| 2908 PRIVATE FILE *file_open( | |
| 2909 struct lemon *lemp, | |
| 2910 const char *suffix, | |
| 2911 const char *mode | |
| 2912 ){ | |
| 2913 FILE *fp; | |
| 2914 | |
| 2915 if( lemp->outname ) free(lemp->outname); | |
| 2916 lemp->outname = file_makename(lemp, suffix); | |
| 2917 fp = fopen(lemp->outname,mode); | |
| 2918 if( fp==0 && *mode=='w' ){ | |
| 2919 fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname); | |
| 2920 lemp->errorcnt++; | |
| 2921 return 0; | |
| 2922 } | |
| 2923 return fp; | |
| 2924 } | |
| 2925 | |
| 2926 /* Duplicate the input file without comments and without actions | |
| 2927 ** on rules */ | |
| 2928 void Reprint(struct lemon *lemp) | |
| 2929 { | |
| 2930 struct rule *rp; | |
| 2931 struct symbol *sp; | |
| 2932 int i, j, maxlen, len, ncolumns, skip; | |
| 2933 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename); | |
| 2934 maxlen = 10; | |
| 2935 for(i=0; i<lemp->nsymbol; i++){ | |
| 2936 sp = lemp->symbols[i]; | |
| 2937 len = lemonStrlen(sp->name); | |
| 2938 if( len>maxlen ) maxlen = len; | |
| 2939 } | |
| 2940 ncolumns = 76/(maxlen+5); | |
| 2941 if( ncolumns<1 ) ncolumns = 1; | |
| 2942 skip = (lemp->nsymbol + ncolumns - 1)/ncolumns; | |
| 2943 for(i=0; i<skip; i++){ | |
| 2944 printf("//"); | |
| 2945 for(j=i; j<lemp->nsymbol; j+=skip){ | |
| 2946 sp = lemp->symbols[j]; | |
| 2947 assert( sp->index==j ); | |
| 2948 printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name); | |
| 2949 } | |
| 2950 printf("\n"); | |
| 2951 } | |
| 2952 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 2953 printf("%s",rp->lhs->name); | |
| 2954 /* if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */ | |
| 2955 printf(" ::="); | |
| 2956 for(i=0; i<rp->nrhs; i++){ | |
| 2957 sp = rp->rhs[i]; | |
| 2958 if( sp->type==MULTITERMINAL ){ | |
| 2959 printf(" %s", sp->subsym[0]->name); | |
| 2960 for(j=1; j<sp->nsubsym; j++){ | |
| 2961 printf("|%s", sp->subsym[j]->name); | |
| 2962 } | |
| 2963 }else{ | |
| 2964 printf(" %s", sp->name); | |
| 2965 } | |
| 2966 /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */ | |
| 2967 } | |
| 2968 printf("."); | |
| 2969 if( rp->precsym ) printf(" [%s]",rp->precsym->name); | |
| 2970 /* if( rp->code ) printf("\n %s",rp->code); */ | |
| 2971 printf("\n"); | |
| 2972 } | |
| 2973 } | |
| 2974 | |
| 2975 /* Print a single rule. | |
| 2976 */ | |
| 2977 void RulePrint(FILE *fp, struct rule *rp, int iCursor){ | |
| 2978 struct symbol *sp; | |
| 2979 int i, j; | |
| 2980 fprintf(fp,"%s ::=",rp->lhs->name); | |
| 2981 for(i=0; i<=rp->nrhs; i++){ | |
| 2982 if( i==iCursor ) fprintf(fp," *"); | |
| 2983 if( i==rp->nrhs ) break; | |
| 2984 sp = rp->rhs[i]; | |
| 2985 if( sp->type==MULTITERMINAL ){ | |
| 2986 fprintf(fp," %s", sp->subsym[0]->name); | |
| 2987 for(j=1; j<sp->nsubsym; j++){ | |
| 2988 fprintf(fp,"|%s",sp->subsym[j]->name); | |
| 2989 } | |
| 2990 }else{ | |
| 2991 fprintf(fp," %s", sp->name); | |
| 2992 } | |
| 2993 } | |
| 2994 } | |
| 2995 | |
| 2996 /* Print the rule for a configuration. | |
| 2997 */ | |
| 2998 void ConfigPrint(FILE *fp, struct config *cfp){ | |
| 2999 RulePrint(fp, cfp->rp, cfp->dot); | |
| 3000 } | |
| 3001 | |
| 3002 /* #define TEST */ | |
| 3003 #if 0 | |
| 3004 /* Print a set */ | |
| 3005 PRIVATE void SetPrint(out,set,lemp) | |
| 3006 FILE *out; | |
| 3007 char *set; | |
| 3008 struct lemon *lemp; | |
| 3009 { | |
| 3010 int i; | |
| 3011 char *spacer; | |
| 3012 spacer = ""; | |
| 3013 fprintf(out,"%12s[",""); | |
| 3014 for(i=0; i<lemp->nterminal; i++){ | |
| 3015 if( SetFind(set,i) ){ | |
| 3016 fprintf(out,"%s%s",spacer,lemp->symbols[i]->name); | |
| 3017 spacer = " "; | |
| 3018 } | |
| 3019 } | |
| 3020 fprintf(out,"]\n"); | |
| 3021 } | |
| 3022 | |
| 3023 /* Print a plink chain */ | |
| 3024 PRIVATE void PlinkPrint(out,plp,tag) | |
| 3025 FILE *out; | |
| 3026 struct plink *plp; | |
| 3027 char *tag; | |
| 3028 { | |
| 3029 while( plp ){ | |
| 3030 fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum); | |
| 3031 ConfigPrint(out,plp->cfp); | |
| 3032 fprintf(out,"\n"); | |
| 3033 plp = plp->next; | |
| 3034 } | |
| 3035 } | |
| 3036 #endif | |
| 3037 | |
| 3038 /* Print an action to the given file descriptor. Return FALSE if | |
| 3039 ** nothing was actually printed. | |
| 3040 */ | |
| 3041 int PrintAction( | |
| 3042 struct action *ap, /* The action to print */ | |
| 3043 FILE *fp, /* Print the action here */ | |
| 3044 int indent /* Indent by this amount */ | |
| 3045 ){ | |
| 3046 int result = 1; | |
| 3047 switch( ap->type ){ | |
| 3048 case SHIFT: { | |
| 3049 struct state *stp = ap->x.stp; | |
| 3050 fprintf(fp,"%*s shift %-7d",indent,ap->sp->name,stp->statenum); | |
| 3051 break; | |
| 3052 } | |
| 3053 case REDUCE: { | |
| 3054 struct rule *rp = ap->x.rp; | |
| 3055 fprintf(fp,"%*s reduce %-7d",indent,ap->sp->name,rp->index); | |
| 3056 RulePrint(fp, rp, -1); | |
| 3057 break; | |
| 3058 } | |
| 3059 case SHIFTREDUCE: { | |
| 3060 struct rule *rp = ap->x.rp; | |
| 3061 fprintf(fp,"%*s shift-reduce %-7d",indent,ap->sp->name,rp->index); | |
| 3062 RulePrint(fp, rp, -1); | |
| 3063 break; | |
| 3064 } | |
| 3065 case ACCEPT: | |
| 3066 fprintf(fp,"%*s accept",indent,ap->sp->name); | |
| 3067 break; | |
| 3068 case ERROR: | |
| 3069 fprintf(fp,"%*s error",indent,ap->sp->name); | |
| 3070 break; | |
| 3071 case SRCONFLICT: | |
| 3072 case RRCONFLICT: | |
| 3073 fprintf(fp,"%*s reduce %-7d ** Parsing conflict **", | |
| 3074 indent,ap->sp->name,ap->x.rp->index); | |
| 3075 break; | |
| 3076 case SSCONFLICT: | |
| 3077 fprintf(fp,"%*s shift %-7d ** Parsing conflict **", | |
| 3078 indent,ap->sp->name,ap->x.stp->statenum); | |
| 3079 break; | |
| 3080 case SH_RESOLVED: | |
| 3081 if( showPrecedenceConflict ){ | |
| 3082 fprintf(fp,"%*s shift %-7d -- dropped by precedence", | |
| 3083 indent,ap->sp->name,ap->x.stp->statenum); | |
| 3084 }else{ | |
| 3085 result = 0; | |
| 3086 } | |
| 3087 break; | |
| 3088 case RD_RESOLVED: | |
| 3089 if( showPrecedenceConflict ){ | |
| 3090 fprintf(fp,"%*s reduce %-7d -- dropped by precedence", | |
| 3091 indent,ap->sp->name,ap->x.rp->index); | |
| 3092 }else{ | |
| 3093 result = 0; | |
| 3094 } | |
| 3095 break; | |
| 3096 case NOT_USED: | |
| 3097 result = 0; | |
| 3098 break; | |
| 3099 } | |
| 3100 return result; | |
| 3101 } | |
| 3102 | |
| 3103 /* Generate the "*.out" log file */ | |
| 3104 void ReportOutput(struct lemon *lemp) | |
| 3105 { | |
| 3106 int i; | |
| 3107 struct state *stp; | |
| 3108 struct config *cfp; | |
| 3109 struct action *ap; | |
| 3110 FILE *fp; | |
| 3111 | |
| 3112 fp = file_open(lemp,".out","wb"); | |
| 3113 if( fp==0 ) return; | |
| 3114 for(i=0; i<lemp->nxstate; i++){ | |
| 3115 stp = lemp->sorted[i]; | |
| 3116 fprintf(fp,"State %d:\n",stp->statenum); | |
| 3117 if( lemp->basisflag ) cfp=stp->bp; | |
| 3118 else cfp=stp->cfp; | |
| 3119 while( cfp ){ | |
| 3120 char buf[20]; | |
| 3121 if( cfp->dot==cfp->rp->nrhs ){ | |
| 3122 lemon_sprintf(buf,"(%d)",cfp->rp->index); | |
| 3123 fprintf(fp," %5s ",buf); | |
| 3124 }else{ | |
| 3125 fprintf(fp," "); | |
| 3126 } | |
| 3127 ConfigPrint(fp,cfp); | |
| 3128 fprintf(fp,"\n"); | |
| 3129 #if 0 | |
| 3130 SetPrint(fp,cfp->fws,lemp); | |
| 3131 PlinkPrint(fp,cfp->fplp,"To "); | |
| 3132 PlinkPrint(fp,cfp->bplp,"From"); | |
| 3133 #endif | |
| 3134 if( lemp->basisflag ) cfp=cfp->bp; | |
| 3135 else cfp=cfp->next; | |
| 3136 } | |
| 3137 fprintf(fp,"\n"); | |
| 3138 for(ap=stp->ap; ap; ap=ap->next){ | |
| 3139 if( PrintAction(ap,fp,30) ) fprintf(fp,"\n"); | |
| 3140 } | |
| 3141 fprintf(fp,"\n"); | |
| 3142 } | |
| 3143 fprintf(fp, "----------------------------------------------------\n"); | |
| 3144 fprintf(fp, "Symbols:\n"); | |
| 3145 for(i=0; i<lemp->nsymbol; i++){ | |
| 3146 int j; | |
| 3147 struct symbol *sp; | |
| 3148 | |
| 3149 sp = lemp->symbols[i]; | |
| 3150 fprintf(fp, " %3d: %s", i, sp->name); | |
| 3151 if( sp->type==NONTERMINAL ){ | |
| 3152 fprintf(fp, ":"); | |
| 3153 if( sp->lambda ){ | |
| 3154 fprintf(fp, " <lambda>"); | |
| 3155 } | |
| 3156 for(j=0; j<lemp->nterminal; j++){ | |
| 3157 if( sp->firstset && SetFind(sp->firstset, j) ){ | |
| 3158 fprintf(fp, " %s", lemp->symbols[j]->name); | |
| 3159 } | |
| 3160 } | |
| 3161 } | |
| 3162 fprintf(fp, "\n"); | |
| 3163 } | |
| 3164 fclose(fp); | |
| 3165 return; | |
| 3166 } | |
| 3167 | |
| 3168 /* Search for the file "name" which is in the same directory as | |
| 3169 ** the exacutable */ | |
| 3170 PRIVATE char *pathsearch(char *argv0, char *name, int modemask) | |
| 3171 { | |
| 3172 const char *pathlist; | |
| 3173 char *pathbufptr; | |
| 3174 char *pathbuf; | |
| 3175 char *path,*cp; | |
| 3176 char c; | |
| 3177 | |
| 3178 #ifdef __WIN32__ | |
| 3179 cp = strrchr(argv0,'\\'); | |
| 3180 #else | |
| 3181 cp = strrchr(argv0,'/'); | |
| 3182 #endif | |
| 3183 if( cp ){ | |
| 3184 c = *cp; | |
| 3185 *cp = 0; | |
| 3186 path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 ); | |
| 3187 if( path ) lemon_sprintf(path,"%s/%s",argv0,name); | |
| 3188 *cp = c; | |
| 3189 }else{ | |
| 3190 pathlist = getenv("PATH"); | |
| 3191 if( pathlist==0 ) pathlist = ".:/bin:/usr/bin"; | |
| 3192 pathbuf = (char *) malloc( lemonStrlen(pathlist) + 1 ); | |
| 3193 path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 ); | |
| 3194 if( (pathbuf != 0) && (path!=0) ){ | |
| 3195 pathbufptr = pathbuf; | |
| 3196 lemon_strcpy(pathbuf, pathlist); | |
| 3197 while( *pathbuf ){ | |
| 3198 cp = strchr(pathbuf,':'); | |
| 3199 if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)]; | |
| 3200 c = *cp; | |
| 3201 *cp = 0; | |
| 3202 lemon_sprintf(path,"%s/%s",pathbuf,name); | |
| 3203 *cp = c; | |
| 3204 if( c==0 ) pathbuf[0] = 0; | |
| 3205 else pathbuf = &cp[1]; | |
| 3206 if( access(path,modemask)==0 ) break; | |
| 3207 } | |
| 3208 free(pathbufptr); | |
| 3209 } | |
| 3210 } | |
| 3211 return path; | |
| 3212 } | |
| 3213 | |
| 3214 /* Given an action, compute the integer value for that action | |
| 3215 ** which is to be put in the action table of the generated machine. | |
| 3216 ** Return negative if no action should be generated. | |
| 3217 */ | |
| 3218 PRIVATE int compute_action(struct lemon *lemp, struct action *ap) | |
| 3219 { | |
| 3220 int act; | |
| 3221 switch( ap->type ){ | |
| 3222 case SHIFT: act = ap->x.stp->statenum; break; | |
| 3223 case SHIFTREDUCE: act = ap->x.rp->index + lemp->nstate; break; | |
| 3224 case REDUCE: act = ap->x.rp->index + lemp->nstate+lemp->nrule; break; | |
| 3225 case ERROR: act = lemp->nstate + lemp->nrule*2; break; | |
| 3226 case ACCEPT: act = lemp->nstate + lemp->nrule*2 + 1; break; | |
| 3227 default: act = -1; break; | |
| 3228 } | |
| 3229 return act; | |
| 3230 } | |
| 3231 | |
| 3232 #define LINESIZE 1000 | |
| 3233 /* The next cluster of routines are for reading the template file | |
| 3234 ** and writing the results to the generated parser */ | |
| 3235 /* The first function transfers data from "in" to "out" until | |
| 3236 ** a line is seen which begins with "%%". The line number is | |
| 3237 ** tracked. | |
| 3238 ** | |
| 3239 ** if name!=0, then any word that begin with "Parse" is changed to | |
| 3240 ** begin with *name instead. | |
| 3241 */ | |
| 3242 PRIVATE void tplt_xfer(char *name, FILE *in, FILE *out, int *lineno) | |
| 3243 { | |
| 3244 int i, iStart; | |
| 3245 char line[LINESIZE]; | |
| 3246 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){ | |
| 3247 (*lineno)++; | |
| 3248 iStart = 0; | |
| 3249 if( name ){ | |
| 3250 for(i=0; line[i]; i++){ | |
| 3251 if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0 | |
| 3252 && (i==0 || !ISALPHA(line[i-1])) | |
| 3253 ){ | |
| 3254 if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]); | |
| 3255 fprintf(out,"%s",name); | |
| 3256 i += 4; | |
| 3257 iStart = i+1; | |
| 3258 } | |
| 3259 } | |
| 3260 } | |
| 3261 fprintf(out,"%s",&line[iStart]); | |
| 3262 } | |
| 3263 } | |
| 3264 | |
| 3265 /* The next function finds the template file and opens it, returning | |
| 3266 ** a pointer to the opened file. */ | |
| 3267 PRIVATE FILE *tplt_open(struct lemon *lemp) | |
| 3268 { | |
| 3269 static char templatename[] = "lempar.c"; | |
| 3270 char buf[1000]; | |
| 3271 FILE *in; | |
| 3272 char *tpltname; | |
| 3273 char *cp; | |
| 3274 | |
| 3275 /* first, see if user specified a template filename on the command line. */ | |
| 3276 if (user_templatename != 0) { | |
| 3277 if( access(user_templatename,004)==-1 ){ | |
| 3278 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n", | |
| 3279 user_templatename); | |
| 3280 lemp->errorcnt++; | |
| 3281 return 0; | |
| 3282 } | |
| 3283 in = fopen(user_templatename,"rb"); | |
| 3284 if( in==0 ){ | |
| 3285 fprintf(stderr,"Can't open the template file \"%s\".\n", | |
| 3286 user_templatename); | |
| 3287 lemp->errorcnt++; | |
| 3288 return 0; | |
| 3289 } | |
| 3290 return in; | |
| 3291 } | |
| 3292 | |
| 3293 cp = strrchr(lemp->filename,'.'); | |
| 3294 if( cp ){ | |
| 3295 lemon_sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename); | |
| 3296 }else{ | |
| 3297 lemon_sprintf(buf,"%s.lt",lemp->filename); | |
| 3298 } | |
| 3299 if( access(buf,004)==0 ){ | |
| 3300 tpltname = buf; | |
| 3301 }else if( access(templatename,004)==0 ){ | |
| 3302 tpltname = templatename; | |
| 3303 }else{ | |
| 3304 tpltname = pathsearch(lemp->argv0,templatename,0); | |
| 3305 } | |
| 3306 if( tpltname==0 ){ | |
| 3307 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n", | |
| 3308 templatename); | |
| 3309 lemp->errorcnt++; | |
| 3310 return 0; | |
| 3311 } | |
| 3312 in = fopen(tpltname,"rb"); | |
| 3313 if( in==0 ){ | |
| 3314 fprintf(stderr,"Can't open the template file \"%s\".\n",templatename); | |
| 3315 lemp->errorcnt++; | |
| 3316 return 0; | |
| 3317 } | |
| 3318 return in; | |
| 3319 } | |
| 3320 | |
| 3321 /* Print a #line directive line to the output file. */ | |
| 3322 PRIVATE void tplt_linedir(FILE *out, int lineno, char *filename) | |
| 3323 { | |
| 3324 fprintf(out,"#line %d \"",lineno); | |
| 3325 while( *filename ){ | |
| 3326 if( *filename == '\\' ) putc('\\',out); | |
| 3327 putc(*filename,out); | |
| 3328 filename++; | |
| 3329 } | |
| 3330 fprintf(out,"\"\n"); | |
| 3331 } | |
| 3332 | |
| 3333 /* Print a string to the file and keep the linenumber up to date */ | |
| 3334 PRIVATE void tplt_print(FILE *out, struct lemon *lemp, char *str, int *lineno) | |
| 3335 { | |
| 3336 if( str==0 ) return; | |
| 3337 while( *str ){ | |
| 3338 putc(*str,out); | |
| 3339 if( *str=='\n' ) (*lineno)++; | |
| 3340 str++; | |
| 3341 } | |
| 3342 if( str[-1]!='\n' ){ | |
| 3343 putc('\n',out); | |
| 3344 (*lineno)++; | |
| 3345 } | |
| 3346 if (!lemp->nolinenosflag) { | |
| 3347 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname); | |
| 3348 } | |
| 3349 return; | |
| 3350 } | |
| 3351 | |
| 3352 /* | |
| 3353 ** The following routine emits code for the destructor for the | |
| 3354 ** symbol sp | |
| 3355 */ | |
| 3356 void emit_destructor_code( | |
| 3357 FILE *out, | |
| 3358 struct symbol *sp, | |
| 3359 struct lemon *lemp, | |
| 3360 int *lineno | |
| 3361 ){ | |
| 3362 char *cp = 0; | |
| 3363 | |
| 3364 if( sp->type==TERMINAL ){ | |
| 3365 cp = lemp->tokendest; | |
| 3366 if( cp==0 ) return; | |
| 3367 fprintf(out,"{\n"); (*lineno)++; | |
| 3368 }else if( sp->destructor ){ | |
| 3369 cp = sp->destructor; | |
| 3370 fprintf(out,"{\n"); (*lineno)++; | |
| 3371 if( !lemp->nolinenosflag ){ | |
| 3372 (*lineno)++; | |
| 3373 tplt_linedir(out,sp->destLineno,lemp->filename); | |
| 3374 } | |
| 3375 }else if( lemp->vardest ){ | |
| 3376 cp = lemp->vardest; | |
| 3377 if( cp==0 ) return; | |
| 3378 fprintf(out,"{\n"); (*lineno)++; | |
| 3379 }else{ | |
| 3380 assert( 0 ); /* Cannot happen */ | |
| 3381 } | |
| 3382 for(; *cp; cp++){ | |
| 3383 if( *cp=='$' && cp[1]=='$' ){ | |
| 3384 fprintf(out,"(yypminor->yy%d)",sp->dtnum); | |
| 3385 cp++; | |
| 3386 continue; | |
| 3387 } | |
| 3388 if( *cp=='\n' ) (*lineno)++; | |
| 3389 fputc(*cp,out); | |
| 3390 } | |
| 3391 fprintf(out,"\n"); (*lineno)++; | |
| 3392 if (!lemp->nolinenosflag) { | |
| 3393 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname); | |
| 3394 } | |
| 3395 fprintf(out,"}\n"); (*lineno)++; | |
| 3396 return; | |
| 3397 } | |
| 3398 | |
| 3399 /* | |
| 3400 ** Return TRUE (non-zero) if the given symbol has a destructor. | |
| 3401 */ | |
| 3402 int has_destructor(struct symbol *sp, struct lemon *lemp) | |
| 3403 { | |
| 3404 int ret; | |
| 3405 if( sp->type==TERMINAL ){ | |
| 3406 ret = lemp->tokendest!=0; | |
| 3407 }else{ | |
| 3408 ret = lemp->vardest!=0 || sp->destructor!=0; | |
| 3409 } | |
| 3410 return ret; | |
| 3411 } | |
| 3412 | |
| 3413 /* | |
| 3414 ** Append text to a dynamically allocated string. If zText is 0 then | |
| 3415 ** reset the string to be empty again. Always return the complete text | |
| 3416 ** of the string (which is overwritten with each call). | |
| 3417 ** | |
| 3418 ** n bytes of zText are stored. If n==0 then all of zText up to the first | |
| 3419 ** \000 terminator is stored. zText can contain up to two instances of | |
| 3420 ** %d. The values of p1 and p2 are written into the first and second | |
| 3421 ** %d. | |
| 3422 ** | |
| 3423 ** If n==-1, then the previous character is overwritten. | |
| 3424 */ | |
| 3425 PRIVATE char *append_str(const char *zText, int n, int p1, int p2){ | |
| 3426 static char empty[1] = { 0 }; | |
| 3427 static char *z = 0; | |
| 3428 static int alloced = 0; | |
| 3429 static int used = 0; | |
| 3430 int c; | |
| 3431 char zInt[40]; | |
| 3432 if( zText==0 ){ | |
| 3433 used = 0; | |
| 3434 return z; | |
| 3435 } | |
| 3436 if( n<=0 ){ | |
| 3437 if( n<0 ){ | |
| 3438 used += n; | |
| 3439 assert( used>=0 ); | |
| 3440 } | |
| 3441 n = lemonStrlen(zText); | |
| 3442 } | |
| 3443 if( (int) (n+sizeof(zInt)*2+used) >= alloced ){ | |
| 3444 alloced = n + sizeof(zInt)*2 + used + 200; | |
| 3445 z = (char *) realloc(z, alloced); | |
| 3446 } | |
| 3447 if( z==0 ) return empty; | |
| 3448 while( n-- > 0 ){ | |
| 3449 c = *(zText++); | |
| 3450 if( c=='%' && n>0 && zText[0]=='d' ){ | |
| 3451 lemon_sprintf(zInt, "%d", p1); | |
| 3452 p1 = p2; | |
| 3453 lemon_strcpy(&z[used], zInt); | |
| 3454 used += lemonStrlen(&z[used]); | |
| 3455 zText++; | |
| 3456 n--; | |
| 3457 }else{ | |
| 3458 z[used++] = (char)c; | |
| 3459 } | |
| 3460 } | |
| 3461 z[used] = 0; | |
| 3462 return z; | |
| 3463 } | |
| 3464 | |
| 3465 /* | |
| 3466 ** zCode is a string that is the action associated with a rule. Expand | |
| 3467 ** the symbols in this string so that the refer to elements of the parser | |
| 3468 ** stack. | |
| 3469 */ | |
| 3470 PRIVATE void translate_code(struct lemon *lemp, struct rule *rp){ | |
| 3471 char *cp, *xp; | |
| 3472 int i; | |
| 3473 char lhsused = 0; /* True if the LHS element has been used */ | |
| 3474 char used[MAXRHS]; /* True for each RHS element which is used */ | |
| 3475 | |
| 3476 for(i=0; i<rp->nrhs; i++) used[i] = 0; | |
| 3477 lhsused = 0; | |
| 3478 | |
| 3479 if( rp->code==0 ){ | |
| 3480 static char newlinestr[2] = { '\n', '\0' }; | |
| 3481 rp->code = newlinestr; | |
| 3482 rp->line = rp->ruleline; | |
| 3483 } | |
| 3484 | |
| 3485 append_str(0,0,0,0); | |
| 3486 | |
| 3487 /* This const cast is wrong but harmless, if we're careful. */ | |
| 3488 for(cp=(char *)rp->code; *cp; cp++){ | |
| 3489 if( ISALPHA(*cp) && (cp==rp->code || (!ISALNUM(cp[-1]) && cp[-1]!='_')) ){ | |
| 3490 char saved; | |
| 3491 for(xp= &cp[1]; ISALNUM(*xp) || *xp=='_'; xp++); | |
| 3492 saved = *xp; | |
| 3493 *xp = 0; | |
| 3494 if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){ | |
| 3495 append_str("yygotominor.yy%d",0,rp->lhs->dtnum,0); | |
| 3496 cp = xp; | |
| 3497 lhsused = 1; | |
| 3498 }else{ | |
| 3499 for(i=0; i<rp->nrhs; i++){ | |
| 3500 if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){ | |
| 3501 if( cp!=rp->code && cp[-1]=='@' ){ | |
| 3502 /* If the argument is of the form @X then substituted | |
| 3503 ** the token number of X, not the value of X */ | |
| 3504 append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0); | |
| 3505 }else{ | |
| 3506 struct symbol *sp = rp->rhs[i]; | |
| 3507 int dtnum; | |
| 3508 if( sp->type==MULTITERMINAL ){ | |
| 3509 dtnum = sp->subsym[0]->dtnum; | |
| 3510 }else{ | |
| 3511 dtnum = sp->dtnum; | |
| 3512 } | |
| 3513 append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum); | |
| 3514 } | |
| 3515 cp = xp; | |
| 3516 used[i] = 1; | |
| 3517 break; | |
| 3518 } | |
| 3519 } | |
| 3520 } | |
| 3521 *xp = saved; | |
| 3522 } | |
| 3523 append_str(cp, 1, 0, 0); | |
| 3524 } /* End loop */ | |
| 3525 | |
| 3526 /* Check to make sure the LHS has been used */ | |
| 3527 if( rp->lhsalias && !lhsused ){ | |
| 3528 ErrorMsg(lemp->filename,rp->ruleline, | |
| 3529 "Label \"%s\" for \"%s(%s)\" is never used.", | |
| 3530 rp->lhsalias,rp->lhs->name,rp->lhsalias); | |
| 3531 lemp->errorcnt++; | |
| 3532 } | |
| 3533 | |
| 3534 /* Generate destructor code for RHS symbols which are not used in the | |
| 3535 ** reduce code */ | |
| 3536 for(i=0; i<rp->nrhs; i++){ | |
| 3537 if( rp->rhsalias[i] && !used[i] ){ | |
| 3538 ErrorMsg(lemp->filename,rp->ruleline, | |
| 3539 "Label %s for \"%s(%s)\" is never used.", | |
| 3540 rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]); | |
| 3541 lemp->errorcnt++; | |
| 3542 }else if( rp->rhsalias[i]==0 ){ | |
| 3543 if( has_destructor(rp->rhs[i],lemp) ){ | |
| 3544 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0, | |
| 3545 rp->rhs[i]->index,i-rp->nrhs+1); | |
| 3546 }else{ | |
| 3547 /* No destructor defined for this term */ | |
| 3548 } | |
| 3549 } | |
| 3550 } | |
| 3551 if( rp->code ){ | |
| 3552 cp = append_str(0,0,0,0); | |
| 3553 rp->code = Strsafe(cp?cp:""); | |
| 3554 } | |
| 3555 } | |
| 3556 | |
| 3557 /* | |
| 3558 ** Generate code which executes when the rule "rp" is reduced. Write | |
| 3559 ** the code to "out". Make sure lineno stays up-to-date. | |
| 3560 */ | |
| 3561 PRIVATE void emit_code( | |
| 3562 FILE *out, | |
| 3563 struct rule *rp, | |
| 3564 struct lemon *lemp, | |
| 3565 int *lineno | |
| 3566 ){ | |
| 3567 const char *cp; | |
| 3568 | |
| 3569 /* Generate code to do the reduce action */ | |
| 3570 if( rp->code ){ | |
| 3571 if( !lemp->nolinenosflag ){ | |
| 3572 (*lineno)++; | |
| 3573 tplt_linedir(out,rp->line,lemp->filename); | |
| 3574 } | |
| 3575 fprintf(out,"{%s",rp->code); | |
| 3576 for(cp=rp->code; *cp; cp++){ | |
| 3577 if( *cp=='\n' ) (*lineno)++; | |
| 3578 } /* End loop */ | |
| 3579 fprintf(out,"}\n"); (*lineno)++; | |
| 3580 if( !lemp->nolinenosflag ){ | |
| 3581 (*lineno)++; | |
| 3582 tplt_linedir(out,*lineno,lemp->outname); | |
| 3583 } | |
| 3584 } /* End if( rp->code ) */ | |
| 3585 | |
| 3586 return; | |
| 3587 } | |
| 3588 | |
| 3589 /* | |
| 3590 ** Print the definition of the union used for the parser's data stack. | |
| 3591 ** This union contains fields for every possible data type for tokens | |
| 3592 ** and nonterminals. In the process of computing and printing this | |
| 3593 ** union, also set the ".dtnum" field of every terminal and nonterminal | |
| 3594 ** symbol. | |
| 3595 */ | |
| 3596 void print_stack_union( | |
| 3597 FILE *out, /* The output stream */ | |
| 3598 struct lemon *lemp, /* The main info structure for this parser */ | |
| 3599 int *plineno, /* Pointer to the line number */ | |
| 3600 int mhflag /* True if generating makeheaders output */ | |
| 3601 ){ | |
| 3602 int lineno = *plineno; /* The line number of the output */ | |
| 3603 char **types; /* A hash table of datatypes */ | |
| 3604 int arraysize; /* Size of the "types" array */ | |
| 3605 int maxdtlength; /* Maximum length of any ".datatype" field. */ | |
| 3606 char *stddt; /* Standardized name for a datatype */ | |
| 3607 int i,j; /* Loop counters */ | |
| 3608 unsigned hash; /* For hashing the name of a type */ | |
| 3609 const char *name; /* Name of the parser */ | |
| 3610 | |
| 3611 /* Allocate and initialize types[] and allocate stddt[] */ | |
| 3612 arraysize = lemp->nsymbol * 2; | |
| 3613 types = (char**)calloc( arraysize, sizeof(char*) ); | |
| 3614 if( types==0 ){ | |
| 3615 fprintf(stderr,"Out of memory.\n"); | |
| 3616 exit(1); | |
| 3617 } | |
| 3618 for(i=0; i<arraysize; i++) types[i] = 0; | |
| 3619 maxdtlength = 0; | |
| 3620 if( lemp->vartype ){ | |
| 3621 maxdtlength = lemonStrlen(lemp->vartype); | |
| 3622 } | |
| 3623 for(i=0; i<lemp->nsymbol; i++){ | |
| 3624 int len; | |
| 3625 struct symbol *sp = lemp->symbols[i]; | |
| 3626 if( sp->datatype==0 ) continue; | |
| 3627 len = lemonStrlen(sp->datatype); | |
| 3628 if( len>maxdtlength ) maxdtlength = len; | |
| 3629 } | |
| 3630 stddt = (char*)malloc( maxdtlength*2 + 1 ); | |
| 3631 if( stddt==0 ){ | |
| 3632 fprintf(stderr,"Out of memory.\n"); | |
| 3633 exit(1); | |
| 3634 } | |
| 3635 | |
| 3636 /* Build a hash table of datatypes. The ".dtnum" field of each symbol | |
| 3637 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is | |
| 3638 ** used for terminal symbols. If there is no %default_type defined then | |
| 3639 ** 0 is also used as the .dtnum value for nonterminals which do not specify | |
| 3640 ** a datatype using the %type directive. | |
| 3641 */ | |
| 3642 for(i=0; i<lemp->nsymbol; i++){ | |
| 3643 struct symbol *sp = lemp->symbols[i]; | |
| 3644 char *cp; | |
| 3645 if( sp==lemp->errsym ){ | |
| 3646 sp->dtnum = arraysize+1; | |
| 3647 continue; | |
| 3648 } | |
| 3649 if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){ | |
| 3650 sp->dtnum = 0; | |
| 3651 continue; | |
| 3652 } | |
| 3653 cp = sp->datatype; | |
| 3654 if( cp==0 ) cp = lemp->vartype; | |
| 3655 j = 0; | |
| 3656 while( ISSPACE(*cp) ) cp++; | |
| 3657 while( *cp ) stddt[j++] = *cp++; | |
| 3658 while( j>0 && ISSPACE(stddt[j-1]) ) j--; | |
| 3659 stddt[j] = 0; | |
| 3660 if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){ | |
| 3661 sp->dtnum = 0; | |
| 3662 continue; | |
| 3663 } | |
| 3664 hash = 0; | |
| 3665 for(j=0; stddt[j]; j++){ | |
| 3666 hash = hash*53 + stddt[j]; | |
| 3667 } | |
| 3668 hash = (hash & 0x7fffffff)%arraysize; | |
| 3669 while( types[hash] ){ | |
| 3670 if( strcmp(types[hash],stddt)==0 ){ | |
| 3671 sp->dtnum = hash + 1; | |
| 3672 break; | |
| 3673 } | |
| 3674 hash++; | |
| 3675 if( hash>=(unsigned)arraysize ) hash = 0; | |
| 3676 } | |
| 3677 if( types[hash]==0 ){ | |
| 3678 sp->dtnum = hash + 1; | |
| 3679 types[hash] = (char*)malloc( lemonStrlen(stddt)+1 ); | |
| 3680 if( types[hash]==0 ){ | |
| 3681 fprintf(stderr,"Out of memory.\n"); | |
| 3682 exit(1); | |
| 3683 } | |
| 3684 lemon_strcpy(types[hash],stddt); | |
| 3685 } | |
| 3686 } | |
| 3687 | |
| 3688 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */ | |
| 3689 name = lemp->name ? lemp->name : "Parse"; | |
| 3690 lineno = *plineno; | |
| 3691 if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; } | |
| 3692 fprintf(out,"#define %sTOKENTYPE %s\n",name, | |
| 3693 lemp->tokentype?lemp->tokentype:"void*"); lineno++; | |
| 3694 if( mhflag ){ fprintf(out,"#endif\n"); lineno++; } | |
| 3695 fprintf(out,"typedef union {\n"); lineno++; | |
| 3696 fprintf(out," int yyinit;\n"); lineno++; | |
| 3697 fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++; | |
| 3698 for(i=0; i<arraysize; i++){ | |
| 3699 if( types[i]==0 ) continue; | |
| 3700 fprintf(out," %s yy%d;\n",types[i],i+1); lineno++; | |
| 3701 free(types[i]); | |
| 3702 } | |
| 3703 if( lemp->errsym->useCnt ){ | |
| 3704 fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++; | |
| 3705 } | |
| 3706 free(stddt); | |
| 3707 free(types); | |
| 3708 fprintf(out,"} YYMINORTYPE;\n"); lineno++; | |
| 3709 *plineno = lineno; | |
| 3710 } | |
| 3711 | |
| 3712 /* | |
| 3713 ** Return the name of a C datatype able to represent values between | |
| 3714 ** lwr and upr, inclusive. If pnByte!=NULL then also write the sizeof | |
| 3715 ** for that type (1, 2, or 4) into *pnByte. | |
| 3716 */ | |
| 3717 static const char *minimum_size_type(int lwr, int upr, int *pnByte){ | |
| 3718 const char *zType = "int"; | |
| 3719 int nByte = 4; | |
| 3720 if( lwr>=0 ){ | |
| 3721 if( upr<=255 ){ | |
| 3722 zType = "unsigned char"; | |
| 3723 nByte = 1; | |
| 3724 }else if( upr<65535 ){ | |
| 3725 zType = "unsigned short int"; | |
| 3726 nByte = 2; | |
| 3727 }else{ | |
| 3728 zType = "unsigned int"; | |
| 3729 nByte = 4; | |
| 3730 } | |
| 3731 }else if( lwr>=-127 && upr<=127 ){ | |
| 3732 zType = "signed char"; | |
| 3733 nByte = 1; | |
| 3734 }else if( lwr>=-32767 && upr<32767 ){ | |
| 3735 zType = "short"; | |
| 3736 nByte = 2; | |
| 3737 } | |
| 3738 if( pnByte ) *pnByte = nByte; | |
| 3739 return zType; | |
| 3740 } | |
| 3741 | |
| 3742 /* | |
| 3743 ** Each state contains a set of token transaction and a set of | |
| 3744 ** nonterminal transactions. Each of these sets makes an instance | |
| 3745 ** of the following structure. An array of these structures is used | |
| 3746 ** to order the creation of entries in the yy_action[] table. | |
| 3747 */ | |
| 3748 struct axset { | |
| 3749 struct state *stp; /* A pointer to a state */ | |
| 3750 int isTkn; /* True to use tokens. False for non-terminals */ | |
| 3751 int nAction; /* Number of actions */ | |
| 3752 int iOrder; /* Original order of action sets */ | |
| 3753 }; | |
| 3754 | |
| 3755 /* | |
| 3756 ** Compare to axset structures for sorting purposes | |
| 3757 */ | |
| 3758 static int axset_compare(const void *a, const void *b){ | |
| 3759 struct axset *p1 = (struct axset*)a; | |
| 3760 struct axset *p2 = (struct axset*)b; | |
| 3761 int c; | |
| 3762 c = p2->nAction - p1->nAction; | |
| 3763 if( c==0 ){ | |
| 3764 c = p1->iOrder - p2->iOrder; | |
| 3765 } | |
| 3766 assert( c!=0 || p1==p2 ); | |
| 3767 return c; | |
| 3768 } | |
| 3769 | |
| 3770 /* | |
| 3771 ** Write text on "out" that describes the rule "rp". | |
| 3772 */ | |
| 3773 static void writeRuleText(FILE *out, struct rule *rp){ | |
| 3774 int j; | |
| 3775 fprintf(out,"%s ::=", rp->lhs->name); | |
| 3776 for(j=0; j<rp->nrhs; j++){ | |
| 3777 struct symbol *sp = rp->rhs[j]; | |
| 3778 if( sp->type!=MULTITERMINAL ){ | |
| 3779 fprintf(out," %s", sp->name); | |
| 3780 }else{ | |
| 3781 int k; | |
| 3782 fprintf(out," %s", sp->subsym[0]->name); | |
| 3783 for(k=1; k<sp->nsubsym; k++){ | |
| 3784 fprintf(out,"|%s",sp->subsym[k]->name); | |
| 3785 } | |
| 3786 } | |
| 3787 } | |
| 3788 } | |
| 3789 | |
| 3790 | |
| 3791 /* Generate C source code for the parser */ | |
| 3792 void ReportTable( | |
| 3793 struct lemon *lemp, | |
| 3794 int mhflag /* Output in makeheaders format if true */ | |
| 3795 ){ | |
| 3796 FILE *out, *in; | |
| 3797 char line[LINESIZE]; | |
| 3798 int lineno; | |
| 3799 struct state *stp; | |
| 3800 struct action *ap; | |
| 3801 struct rule *rp; | |
| 3802 struct acttab *pActtab; | |
| 3803 int i, j, n, sz; | |
| 3804 int szActionType; /* sizeof(YYACTIONTYPE) */ | |
| 3805 int szCodeType; /* sizeof(YYCODETYPE) */ | |
| 3806 const char *name; | |
| 3807 int mnTknOfst, mxTknOfst; | |
| 3808 int mnNtOfst, mxNtOfst; | |
| 3809 struct axset *ax; | |
| 3810 | |
| 3811 in = tplt_open(lemp); | |
| 3812 if( in==0 ) return; | |
| 3813 out = file_open(lemp,".c","wb"); | |
| 3814 if( out==0 ){ | |
| 3815 fclose(in); | |
| 3816 return; | |
| 3817 } | |
| 3818 lineno = 1; | |
| 3819 tplt_xfer(lemp->name,in,out,&lineno); | |
| 3820 | |
| 3821 /* Generate the include code, if any */ | |
| 3822 tplt_print(out,lemp,lemp->include,&lineno); | |
| 3823 if( mhflag ){ | |
| 3824 char *incName = file_makename(lemp, ".h"); | |
| 3825 fprintf(out,"#include \"%s\"\n", incName); lineno++; | |
| 3826 free(incName); | |
| 3827 } | |
| 3828 tplt_xfer(lemp->name,in,out,&lineno); | |
| 3829 | |
| 3830 /* Generate #defines for all tokens */ | |
| 3831 if( mhflag ){ | |
| 3832 const char *prefix; | |
| 3833 fprintf(out,"#if INTERFACE\n"); lineno++; | |
| 3834 if( lemp->tokenprefix ) prefix = lemp->tokenprefix; | |
| 3835 else prefix = ""; | |
| 3836 for(i=1; i<lemp->nterminal; i++){ | |
| 3837 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i); | |
| 3838 lineno++; | |
| 3839 } | |
| 3840 fprintf(out,"#endif\n"); lineno++; | |
| 3841 } | |
| 3842 tplt_xfer(lemp->name,in,out,&lineno); | |
| 3843 | |
| 3844 /* Generate the defines */ | |
| 3845 fprintf(out,"#define YYCODETYPE %s\n", | |
| 3846 minimum_size_type(0, lemp->nsymbol+1, &szCodeType)); lineno++; | |
| 3847 fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol+1); lineno++; | |
| 3848 fprintf(out,"#define YYACTIONTYPE %s\n", | |
| 3849 minimum_size_type(0,lemp->nstate+lemp->nrule*2+5,&szActionType)); lineno++; | |
| 3850 if( lemp->wildcard ){ | |
| 3851 fprintf(out,"#define YYWILDCARD %d\n", | |
| 3852 lemp->wildcard->index); lineno++; | |
| 3853 } | |
| 3854 print_stack_union(out,lemp,&lineno,mhflag); | |
| 3855 fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++; | |
| 3856 if( lemp->stacksize ){ | |
| 3857 fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++; | |
| 3858 }else{ | |
| 3859 fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++; | |
| 3860 } | |
| 3861 fprintf(out, "#endif\n"); lineno++; | |
| 3862 if( mhflag ){ | |
| 3863 fprintf(out,"#if INTERFACE\n"); lineno++; | |
| 3864 } | |
| 3865 name = lemp->name ? lemp->name : "Parse"; | |
| 3866 if( lemp->arg && lemp->arg[0] ){ | |
| 3867 i = lemonStrlen(lemp->arg); | |
| 3868 while( i>=1 && ISSPACE(lemp->arg[i-1]) ) i--; | |
| 3869 while( i>=1 && (ISALNUM(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--; | |
| 3870 fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++; | |
| 3871 fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++; | |
| 3872 fprintf(out,"#define %sARG_FETCH %s = yypParser->%s\n", | |
| 3873 name,lemp->arg,&lemp->arg[i]); lineno++; | |
| 3874 fprintf(out,"#define %sARG_STORE yypParser->%s = %s\n", | |
| 3875 name,&lemp->arg[i],&lemp->arg[i]); lineno++; | |
| 3876 }else{ | |
| 3877 fprintf(out,"#define %sARG_SDECL\n",name); lineno++; | |
| 3878 fprintf(out,"#define %sARG_PDECL\n",name); lineno++; | |
| 3879 fprintf(out,"#define %sARG_FETCH\n",name); lineno++; | |
| 3880 fprintf(out,"#define %sARG_STORE\n",name); lineno++; | |
| 3881 } | |
| 3882 if( mhflag ){ | |
| 3883 fprintf(out,"#endif\n"); lineno++; | |
| 3884 } | |
| 3885 if( lemp->errsym->useCnt ){ | |
| 3886 fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++; | |
| 3887 fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++; | |
| 3888 } | |
| 3889 if( lemp->has_fallback ){ | |
| 3890 fprintf(out,"#define YYFALLBACK 1\n"); lineno++; | |
| 3891 } | |
| 3892 | |
| 3893 /* Compute the action table, but do not output it yet. The action | |
| 3894 ** table must be computed before generating the YYNSTATE macro because | |
| 3895 ** we need to know how many states can be eliminated. | |
| 3896 */ | |
| 3897 ax = (struct axset *) calloc(lemp->nxstate*2, sizeof(ax[0])); | |
| 3898 if( ax==0 ){ | |
| 3899 fprintf(stderr,"malloc failed\n"); | |
| 3900 exit(1); | |
| 3901 } | |
| 3902 for(i=0; i<lemp->nxstate; i++){ | |
| 3903 stp = lemp->sorted[i]; | |
| 3904 ax[i*2].stp = stp; | |
| 3905 ax[i*2].isTkn = 1; | |
| 3906 ax[i*2].nAction = stp->nTknAct; | |
| 3907 ax[i*2+1].stp = stp; | |
| 3908 ax[i*2+1].isTkn = 0; | |
| 3909 ax[i*2+1].nAction = stp->nNtAct; | |
| 3910 } | |
| 3911 mxTknOfst = mnTknOfst = 0; | |
| 3912 mxNtOfst = mnNtOfst = 0; | |
| 3913 /* In an effort to minimize the action table size, use the heuristic | |
| 3914 ** of placing the largest action sets first */ | |
| 3915 for(i=0; i<lemp->nxstate*2; i++) ax[i].iOrder = i; | |
| 3916 qsort(ax, lemp->nxstate*2, sizeof(ax[0]), axset_compare); | |
| 3917 pActtab = acttab_alloc(); | |
| 3918 for(i=0; i<lemp->nxstate*2 && ax[i].nAction>0; i++){ | |
| 3919 stp = ax[i].stp; | |
| 3920 if( ax[i].isTkn ){ | |
| 3921 for(ap=stp->ap; ap; ap=ap->next){ | |
| 3922 int action; | |
| 3923 if( ap->sp->index>=lemp->nterminal ) continue; | |
| 3924 action = compute_action(lemp, ap); | |
| 3925 if( action<0 ) continue; | |
| 3926 acttab_action(pActtab, ap->sp->index, action); | |
| 3927 } | |
| 3928 stp->iTknOfst = acttab_insert(pActtab); | |
| 3929 if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst; | |
| 3930 if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst; | |
| 3931 }else{ | |
| 3932 for(ap=stp->ap; ap; ap=ap->next){ | |
| 3933 int action; | |
| 3934 if( ap->sp->index<lemp->nterminal ) continue; | |
| 3935 if( ap->sp->index==lemp->nsymbol ) continue; | |
| 3936 action = compute_action(lemp, ap); | |
| 3937 if( action<0 ) continue; | |
| 3938 acttab_action(pActtab, ap->sp->index, action); | |
| 3939 } | |
| 3940 stp->iNtOfst = acttab_insert(pActtab); | |
| 3941 if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst; | |
| 3942 if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst; | |
| 3943 } | |
| 3944 #if 0 /* Uncomment for a trace of how the yy_action[] table fills out */ | |
| 3945 { int jj, nn; | |
| 3946 for(jj=nn=0; jj<pActtab->nAction; jj++){ | |
| 3947 if( pActtab->aAction[jj].action<0 ) nn++; | |
| 3948 } | |
| 3949 printf("%4d: State %3d %s n: %2d size: %5d freespace: %d\n", | |
| 3950 i, stp->statenum, ax[i].isTkn ? "Token" : "Var ", | |
| 3951 ax[i].nAction, pActtab->nAction, nn); | |
| 3952 } | |
| 3953 #endif | |
| 3954 } | |
| 3955 free(ax); | |
| 3956 | |
| 3957 /* Finish rendering the constants now that the action table has | |
| 3958 ** been computed */ | |
| 3959 fprintf(out,"#define YYNSTATE %d\n",lemp->nxstate); lineno++; | |
| 3960 fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++; | |
| 3961 fprintf(out,"#define YY_MAX_SHIFT %d\n",lemp->nxstate-1); lineno++; | |
| 3962 fprintf(out,"#define YY_MIN_SHIFTREDUCE %d\n",lemp->nstate); lineno++; | |
| 3963 i = lemp->nstate + lemp->nrule; | |
| 3964 fprintf(out,"#define YY_MAX_SHIFTREDUCE %d\n", i-1); lineno++; | |
| 3965 fprintf(out,"#define YY_MIN_REDUCE %d\n", i); lineno++; | |
| 3966 i = lemp->nstate + lemp->nrule*2; | |
| 3967 fprintf(out,"#define YY_MAX_REDUCE %d\n", i-1); lineno++; | |
| 3968 fprintf(out,"#define YY_ERROR_ACTION %d\n", i); lineno++; | |
| 3969 fprintf(out,"#define YY_ACCEPT_ACTION %d\n", i+1); lineno++; | |
| 3970 fprintf(out,"#define YY_NO_ACTION %d\n", i+2); lineno++; | |
| 3971 tplt_xfer(lemp->name,in,out,&lineno); | |
| 3972 | |
| 3973 /* Now output the action table and its associates: | |
| 3974 ** | |
| 3975 ** yy_action[] A single table containing all actions. | |
| 3976 ** yy_lookahead[] A table containing the lookahead for each entry in | |
| 3977 ** yy_action. Used to detect hash collisions. | |
| 3978 ** yy_shift_ofst[] For each state, the offset into yy_action for | |
| 3979 ** shifting terminals. | |
| 3980 ** yy_reduce_ofst[] For each state, the offset into yy_action for | |
| 3981 ** shifting non-terminals after a reduce. | |
| 3982 ** yy_default[] Default action for each state. | |
| 3983 */ | |
| 3984 | |
| 3985 /* Output the yy_action table */ | |
| 3986 lemp->nactiontab = n = acttab_size(pActtab); | |
| 3987 lemp->tablesize += n*szActionType; | |
| 3988 fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++; | |
| 3989 fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++; | |
| 3990 for(i=j=0; i<n; i++){ | |
| 3991 int action = acttab_yyaction(pActtab, i); | |
| 3992 if( action<0 ) action = lemp->nstate + lemp->nrule + 2; | |
| 3993 if( j==0 ) fprintf(out," /* %5d */ ", i); | |
| 3994 fprintf(out, " %4d,", action); | |
| 3995 if( j==9 || i==n-1 ){ | |
| 3996 fprintf(out, "\n"); lineno++; | |
| 3997 j = 0; | |
| 3998 }else{ | |
| 3999 j++; | |
| 4000 } | |
| 4001 } | |
| 4002 fprintf(out, "};\n"); lineno++; | |
| 4003 | |
| 4004 /* Output the yy_lookahead table */ | |
| 4005 lemp->tablesize += n*szCodeType; | |
| 4006 fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++; | |
| 4007 for(i=j=0; i<n; i++){ | |
| 4008 int la = acttab_yylookahead(pActtab, i); | |
| 4009 if( la<0 ) la = lemp->nsymbol; | |
| 4010 if( j==0 ) fprintf(out," /* %5d */ ", i); | |
| 4011 fprintf(out, " %4d,", la); | |
| 4012 if( j==9 || i==n-1 ){ | |
| 4013 fprintf(out, "\n"); lineno++; | |
| 4014 j = 0; | |
| 4015 }else{ | |
| 4016 j++; | |
| 4017 } | |
| 4018 } | |
| 4019 fprintf(out, "};\n"); lineno++; | |
| 4020 | |
| 4021 /* Output the yy_shift_ofst[] table */ | |
| 4022 fprintf(out, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst-1); lineno++; | |
| 4023 n = lemp->nxstate; | |
| 4024 while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--; | |
| 4025 fprintf(out, "#define YY_SHIFT_COUNT (%d)\n", n-1); lineno++; | |
| 4026 fprintf(out, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst); lineno++; | |
| 4027 fprintf(out, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst); lineno++; | |
| 4028 fprintf(out, "static const %s yy_shift_ofst[] = {\n", | |
| 4029 minimum_size_type(mnTknOfst-1, mxTknOfst, &sz)); lineno++; | |
| 4030 lemp->tablesize += n*sz; | |
| 4031 for(i=j=0; i<n; i++){ | |
| 4032 int ofst; | |
| 4033 stp = lemp->sorted[i]; | |
| 4034 ofst = stp->iTknOfst; | |
| 4035 if( ofst==NO_OFFSET ) ofst = mnTknOfst - 1; | |
| 4036 if( j==0 ) fprintf(out," /* %5d */ ", i); | |
| 4037 fprintf(out, " %4d,", ofst); | |
| 4038 if( j==9 || i==n-1 ){ | |
| 4039 fprintf(out, "\n"); lineno++; | |
| 4040 j = 0; | |
| 4041 }else{ | |
| 4042 j++; | |
| 4043 } | |
| 4044 } | |
| 4045 fprintf(out, "};\n"); lineno++; | |
| 4046 | |
| 4047 /* Output the yy_reduce_ofst[] table */ | |
| 4048 fprintf(out, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst-1); lineno++; | |
| 4049 n = lemp->nxstate; | |
| 4050 while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--; | |
| 4051 fprintf(out, "#define YY_REDUCE_COUNT (%d)\n", n-1); lineno++; | |
| 4052 fprintf(out, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst); lineno++; | |
| 4053 fprintf(out, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst); lineno++; | |
| 4054 fprintf(out, "static const %s yy_reduce_ofst[] = {\n", | |
| 4055 minimum_size_type(mnNtOfst-1, mxNtOfst, &sz)); lineno++; | |
| 4056 lemp->tablesize += n*sz; | |
| 4057 for(i=j=0; i<n; i++){ | |
| 4058 int ofst; | |
| 4059 stp = lemp->sorted[i]; | |
| 4060 ofst = stp->iNtOfst; | |
| 4061 if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1; | |
| 4062 if( j==0 ) fprintf(out," /* %5d */ ", i); | |
| 4063 fprintf(out, " %4d,", ofst); | |
| 4064 if( j==9 || i==n-1 ){ | |
| 4065 fprintf(out, "\n"); lineno++; | |
| 4066 j = 0; | |
| 4067 }else{ | |
| 4068 j++; | |
| 4069 } | |
| 4070 } | |
| 4071 fprintf(out, "};\n"); lineno++; | |
| 4072 | |
| 4073 /* Output the default action table */ | |
| 4074 fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++; | |
| 4075 n = lemp->nxstate; | |
| 4076 lemp->tablesize += n*szActionType; | |
| 4077 for(i=j=0; i<n; i++){ | |
| 4078 stp = lemp->sorted[i]; | |
| 4079 if( j==0 ) fprintf(out," /* %5d */ ", i); | |
| 4080 fprintf(out, " %4d,", stp->iDfltReduce+lemp->nstate+lemp->nrule); | |
| 4081 if( j==9 || i==n-1 ){ | |
| 4082 fprintf(out, "\n"); lineno++; | |
| 4083 j = 0; | |
| 4084 }else{ | |
| 4085 j++; | |
| 4086 } | |
| 4087 } | |
| 4088 fprintf(out, "};\n"); lineno++; | |
| 4089 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4090 | |
| 4091 /* Generate the table of fallback tokens. | |
| 4092 */ | |
| 4093 if( lemp->has_fallback ){ | |
| 4094 int mx = lemp->nterminal - 1; | |
| 4095 while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; } | |
| 4096 lemp->tablesize += (mx+1)*szCodeType; | |
| 4097 for(i=0; i<=mx; i++){ | |
| 4098 struct symbol *p = lemp->symbols[i]; | |
| 4099 if( p->fallback==0 ){ | |
| 4100 fprintf(out, " 0, /* %10s => nothing */\n", p->name); | |
| 4101 }else{ | |
| 4102 fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index, | |
| 4103 p->name, p->fallback->name); | |
| 4104 } | |
| 4105 lineno++; | |
| 4106 } | |
| 4107 } | |
| 4108 tplt_xfer(lemp->name, in, out, &lineno); | |
| 4109 | |
| 4110 /* Generate a table containing the symbolic name of every symbol | |
| 4111 */ | |
| 4112 for(i=0; i<lemp->nsymbol; i++){ | |
| 4113 lemon_sprintf(line,"\"%s\",",lemp->symbols[i]->name); | |
| 4114 fprintf(out," %-15s",line); | |
| 4115 if( (i&3)==3 ){ fprintf(out,"\n"); lineno++; } | |
| 4116 } | |
| 4117 if( (i&3)!=0 ){ fprintf(out,"\n"); lineno++; } | |
| 4118 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4119 | |
| 4120 /* Generate a table containing a text string that describes every | |
| 4121 ** rule in the rule set of the grammar. This information is used | |
| 4122 ** when tracing REDUCE actions. | |
| 4123 */ | |
| 4124 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){ | |
| 4125 assert( rp->index==i ); | |
| 4126 fprintf(out," /* %3d */ \"", i); | |
| 4127 writeRuleText(out, rp); | |
| 4128 fprintf(out,"\",\n"); lineno++; | |
| 4129 } | |
| 4130 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4131 | |
| 4132 /* Generate code which executes every time a symbol is popped from | |
| 4133 ** the stack while processing errors or while destroying the parser. | |
| 4134 ** (In other words, generate the %destructor actions) | |
| 4135 */ | |
| 4136 if( lemp->tokendest ){ | |
| 4137 int once = 1; | |
| 4138 for(i=0; i<lemp->nsymbol; i++){ | |
| 4139 struct symbol *sp = lemp->symbols[i]; | |
| 4140 if( sp==0 || sp->type!=TERMINAL ) continue; | |
| 4141 if( once ){ | |
| 4142 fprintf(out, " /* TERMINAL Destructor */\n"); lineno++; | |
| 4143 once = 0; | |
| 4144 } | |
| 4145 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++; | |
| 4146 } | |
| 4147 for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++); | |
| 4148 if( i<lemp->nsymbol ){ | |
| 4149 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno); | |
| 4150 fprintf(out," break;\n"); lineno++; | |
| 4151 } | |
| 4152 } | |
| 4153 if( lemp->vardest ){ | |
| 4154 struct symbol *dflt_sp = 0; | |
| 4155 int once = 1; | |
| 4156 for(i=0; i<lemp->nsymbol; i++){ | |
| 4157 struct symbol *sp = lemp->symbols[i]; | |
| 4158 if( sp==0 || sp->type==TERMINAL || | |
| 4159 sp->index<=0 || sp->destructor!=0 ) continue; | |
| 4160 if( once ){ | |
| 4161 fprintf(out, " /* Default NON-TERMINAL Destructor */\n"); lineno++; | |
| 4162 once = 0; | |
| 4163 } | |
| 4164 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++; | |
| 4165 dflt_sp = sp; | |
| 4166 } | |
| 4167 if( dflt_sp!=0 ){ | |
| 4168 emit_destructor_code(out,dflt_sp,lemp,&lineno); | |
| 4169 } | |
| 4170 fprintf(out," break;\n"); lineno++; | |
| 4171 } | |
| 4172 for(i=0; i<lemp->nsymbol; i++){ | |
| 4173 struct symbol *sp = lemp->symbols[i]; | |
| 4174 if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue; | |
| 4175 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++; | |
| 4176 | |
| 4177 /* Combine duplicate destructors into a single case */ | |
| 4178 for(j=i+1; j<lemp->nsymbol; j++){ | |
| 4179 struct symbol *sp2 = lemp->symbols[j]; | |
| 4180 if( sp2 && sp2->type!=TERMINAL && sp2->destructor | |
| 4181 && sp2->dtnum==sp->dtnum | |
| 4182 && strcmp(sp->destructor,sp2->destructor)==0 ){ | |
| 4183 fprintf(out," case %d: /* %s */\n", | |
| 4184 sp2->index, sp2->name); lineno++; | |
| 4185 sp2->destructor = 0; | |
| 4186 } | |
| 4187 } | |
| 4188 | |
| 4189 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno); | |
| 4190 fprintf(out," break;\n"); lineno++; | |
| 4191 } | |
| 4192 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4193 | |
| 4194 /* Generate code which executes whenever the parser stack overflows */ | |
| 4195 tplt_print(out,lemp,lemp->overflow,&lineno); | |
| 4196 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4197 | |
| 4198 /* Generate the table of rule information | |
| 4199 ** | |
| 4200 ** Note: This code depends on the fact that rules are number | |
| 4201 ** sequentually beginning with 0. | |
| 4202 */ | |
| 4203 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 4204 fprintf(out," { %d, %d },\n",rp->lhs->index,rp->nrhs); lineno++; | |
| 4205 } | |
| 4206 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4207 | |
| 4208 /* Generate code which execution during each REDUCE action */ | |
| 4209 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 4210 translate_code(lemp, rp); | |
| 4211 } | |
| 4212 /* First output rules other than the default: rule */ | |
| 4213 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 4214 struct rule *rp2; /* Other rules with the same action */ | |
| 4215 if( rp->code==0 ) continue; | |
| 4216 if( rp->code[0]=='\n' && rp->code[1]==0 ) continue; /* Will be default: */ | |
| 4217 fprintf(out," case %d: /* ", rp->index); | |
| 4218 writeRuleText(out, rp); | |
| 4219 fprintf(out, " */\n"); lineno++; | |
| 4220 for(rp2=rp->next; rp2; rp2=rp2->next){ | |
| 4221 if( rp2->code==rp->code ){ | |
| 4222 fprintf(out," case %d: /* ", rp2->index); | |
| 4223 writeRuleText(out, rp2); | |
| 4224 fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->index); lineno++; | |
| 4225 rp2->code = 0; | |
| 4226 } | |
| 4227 } | |
| 4228 emit_code(out,rp,lemp,&lineno); | |
| 4229 fprintf(out," break;\n"); lineno++; | |
| 4230 rp->code = 0; | |
| 4231 } | |
| 4232 /* Finally, output the default: rule. We choose as the default: all | |
| 4233 ** empty actions. */ | |
| 4234 fprintf(out," default:\n"); lineno++; | |
| 4235 for(rp=lemp->rule; rp; rp=rp->next){ | |
| 4236 if( rp->code==0 ) continue; | |
| 4237 assert( rp->code[0]=='\n' && rp->code[1]==0 ); | |
| 4238 fprintf(out," /* (%d) ", rp->index); | |
| 4239 writeRuleText(out, rp); | |
| 4240 fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->index); lineno++; | |
| 4241 } | |
| 4242 fprintf(out," break;\n"); lineno++; | |
| 4243 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4244 | |
| 4245 /* Generate code which executes if a parse fails */ | |
| 4246 tplt_print(out,lemp,lemp->failure,&lineno); | |
| 4247 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4248 | |
| 4249 /* Generate code which executes when a syntax error occurs */ | |
| 4250 tplt_print(out,lemp,lemp->error,&lineno); | |
| 4251 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4252 | |
| 4253 /* Generate code which executes when the parser accepts its input */ | |
| 4254 tplt_print(out,lemp,lemp->accept,&lineno); | |
| 4255 tplt_xfer(lemp->name,in,out,&lineno); | |
| 4256 | |
| 4257 /* Append any addition code the user desires */ | |
| 4258 tplt_print(out,lemp,lemp->extracode,&lineno); | |
| 4259 | |
| 4260 fclose(in); | |
| 4261 fclose(out); | |
| 4262 return; | |
| 4263 } | |
| 4264 | |
| 4265 /* Generate a header file for the parser */ | |
| 4266 void ReportHeader(struct lemon *lemp) | |
| 4267 { | |
| 4268 FILE *out, *in; | |
| 4269 const char *prefix; | |
| 4270 char line[LINESIZE]; | |
| 4271 char pattern[LINESIZE]; | |
| 4272 int i; | |
| 4273 | |
| 4274 if( lemp->tokenprefix ) prefix = lemp->tokenprefix; | |
| 4275 else prefix = ""; | |
| 4276 in = file_open(lemp,".h","rb"); | |
| 4277 if( in ){ | |
| 4278 int nextChar; | |
| 4279 for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){ | |
| 4280 lemon_sprintf(pattern,"#define %s%-30s %3d\n", | |
| 4281 prefix,lemp->symbols[i]->name,i); | |
| 4282 if( strcmp(line,pattern) ) break; | |
| 4283 } | |
| 4284 nextChar = fgetc(in); | |
| 4285 fclose(in); | |
| 4286 if( i==lemp->nterminal && nextChar==EOF ){ | |
| 4287 /* No change in the file. Don't rewrite it. */ | |
| 4288 return; | |
| 4289 } | |
| 4290 } | |
| 4291 out = file_open(lemp,".h","wb"); | |
| 4292 if( out ){ | |
| 4293 for(i=1; i<lemp->nterminal; i++){ | |
| 4294 fprintf(out,"#define %s%-30s %3d\n",prefix,lemp->symbols[i]->name,i); | |
| 4295 } | |
| 4296 fclose(out); | |
| 4297 } | |
| 4298 return; | |
| 4299 } | |
| 4300 | |
| 4301 /* Reduce the size of the action tables, if possible, by making use | |
| 4302 ** of defaults. | |
| 4303 ** | |
| 4304 ** In this version, we take the most frequent REDUCE action and make | |
| 4305 ** it the default. Except, there is no default if the wildcard token | |
| 4306 ** is a possible look-ahead. | |
| 4307 */ | |
| 4308 void CompressTables(struct lemon *lemp) | |
| 4309 { | |
| 4310 struct state *stp; | |
| 4311 struct action *ap, *ap2; | |
| 4312 struct rule *rp, *rp2, *rbest; | |
| 4313 int nbest, n; | |
| 4314 int i; | |
| 4315 int usesWildcard; | |
| 4316 | |
| 4317 for(i=0; i<lemp->nstate; i++){ | |
| 4318 stp = lemp->sorted[i]; | |
| 4319 nbest = 0; | |
| 4320 rbest = 0; | |
| 4321 usesWildcard = 0; | |
| 4322 | |
| 4323 for(ap=stp->ap; ap; ap=ap->next){ | |
| 4324 if( ap->type==SHIFT && ap->sp==lemp->wildcard ){ | |
| 4325 usesWildcard = 1; | |
| 4326 } | |
| 4327 if( ap->type!=REDUCE ) continue; | |
| 4328 rp = ap->x.rp; | |
| 4329 if( rp->lhsStart ) continue; | |
| 4330 if( rp==rbest ) continue; | |
| 4331 n = 1; | |
| 4332 for(ap2=ap->next; ap2; ap2=ap2->next){ | |
| 4333 if( ap2->type!=REDUCE ) continue; | |
| 4334 rp2 = ap2->x.rp; | |
| 4335 if( rp2==rbest ) continue; | |
| 4336 if( rp2==rp ) n++; | |
| 4337 } | |
| 4338 if( n>nbest ){ | |
| 4339 nbest = n; | |
| 4340 rbest = rp; | |
| 4341 } | |
| 4342 } | |
| 4343 | |
| 4344 /* Do not make a default if the number of rules to default | |
| 4345 ** is not at least 1 or if the wildcard token is a possible | |
| 4346 ** lookahead. | |
| 4347 */ | |
| 4348 if( nbest<1 || usesWildcard ) continue; | |
| 4349 | |
| 4350 | |
| 4351 /* Combine matching REDUCE actions into a single default */ | |
| 4352 for(ap=stp->ap; ap; ap=ap->next){ | |
| 4353 if( ap->type==REDUCE && ap->x.rp==rbest ) break; | |
| 4354 } | |
| 4355 assert( ap ); | |
| 4356 ap->sp = Symbol_new("{default}"); | |
| 4357 for(ap=ap->next; ap; ap=ap->next){ | |
| 4358 if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED; | |
| 4359 } | |
| 4360 stp->ap = Action_sort(stp->ap); | |
| 4361 | |
| 4362 for(ap=stp->ap; ap; ap=ap->next){ | |
| 4363 if( ap->type==SHIFT ) break; | |
| 4364 if( ap->type==REDUCE && ap->x.rp!=rbest ) break; | |
| 4365 } | |
| 4366 if( ap==0 ){ | |
| 4367 stp->autoReduce = 1; | |
| 4368 stp->pDfltReduce = rbest; | |
| 4369 } | |
| 4370 } | |
| 4371 | |
| 4372 /* Make a second pass over all states and actions. Convert | |
| 4373 ** every action that is a SHIFT to an autoReduce state into | |
| 4374 ** a SHIFTREDUCE action. | |
| 4375 */ | |
| 4376 for(i=0; i<lemp->nstate; i++){ | |
| 4377 stp = lemp->sorted[i]; | |
| 4378 for(ap=stp->ap; ap; ap=ap->next){ | |
| 4379 struct state *pNextState; | |
| 4380 if( ap->type!=SHIFT ) continue; | |
| 4381 pNextState = ap->x.stp; | |
| 4382 if( pNextState->autoReduce && pNextState->pDfltReduce!=0 ){ | |
| 4383 ap->type = SHIFTREDUCE; | |
| 4384 ap->x.rp = pNextState->pDfltReduce; | |
| 4385 } | |
| 4386 } | |
| 4387 } | |
| 4388 } | |
| 4389 | |
| 4390 | |
| 4391 /* | |
| 4392 ** Compare two states for sorting purposes. The smaller state is the | |
| 4393 ** one with the most non-terminal actions. If they have the same number | |
| 4394 ** of non-terminal actions, then the smaller is the one with the most | |
| 4395 ** token actions. | |
| 4396 */ | |
| 4397 static int stateResortCompare(const void *a, const void *b){ | |
| 4398 const struct state *pA = *(const struct state**)a; | |
| 4399 const struct state *pB = *(const struct state**)b; | |
| 4400 int n; | |
| 4401 | |
| 4402 n = pB->nNtAct - pA->nNtAct; | |
| 4403 if( n==0 ){ | |
| 4404 n = pB->nTknAct - pA->nTknAct; | |
| 4405 if( n==0 ){ | |
| 4406 n = pB->statenum - pA->statenum; | |
| 4407 } | |
| 4408 } | |
| 4409 assert( n!=0 ); | |
| 4410 return n; | |
| 4411 } | |
| 4412 | |
| 4413 | |
| 4414 /* | |
| 4415 ** Renumber and resort states so that states with fewer choices | |
| 4416 ** occur at the end. Except, keep state 0 as the first state. | |
| 4417 */ | |
| 4418 void ResortStates(struct lemon *lemp) | |
| 4419 { | |
| 4420 int i; | |
| 4421 struct state *stp; | |
| 4422 struct action *ap; | |
| 4423 | |
| 4424 for(i=0; i<lemp->nstate; i++){ | |
| 4425 stp = lemp->sorted[i]; | |
| 4426 stp->nTknAct = stp->nNtAct = 0; | |
| 4427 stp->iDfltReduce = lemp->nrule; /* Init dflt action to "syntax error" */ | |
| 4428 stp->iTknOfst = NO_OFFSET; | |
| 4429 stp->iNtOfst = NO_OFFSET; | |
| 4430 for(ap=stp->ap; ap; ap=ap->next){ | |
| 4431 int iAction = compute_action(lemp,ap); | |
| 4432 if( iAction>=0 ){ | |
| 4433 if( ap->sp->index<lemp->nterminal ){ | |
| 4434 stp->nTknAct++; | |
| 4435 }else if( ap->sp->index<lemp->nsymbol ){ | |
| 4436 stp->nNtAct++; | |
| 4437 }else{ | |
| 4438 assert( stp->autoReduce==0 || stp->pDfltReduce==ap->x.rp ); | |
| 4439 stp->iDfltReduce = iAction - lemp->nstate - lemp->nrule; | |
| 4440 } | |
| 4441 } | |
| 4442 } | |
| 4443 } | |
| 4444 qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]), | |
| 4445 stateResortCompare); | |
| 4446 for(i=0; i<lemp->nstate; i++){ | |
| 4447 lemp->sorted[i]->statenum = i; | |
| 4448 } | |
| 4449 lemp->nxstate = lemp->nstate; | |
| 4450 while( lemp->nxstate>1 && lemp->sorted[lemp->nxstate-1]->autoReduce ){ | |
| 4451 lemp->nxstate--; | |
| 4452 } | |
| 4453 } | |
| 4454 | |
| 4455 | |
| 4456 /***************** From the file "set.c" ************************************/ | |
| 4457 /* | |
| 4458 ** Set manipulation routines for the LEMON parser generator. | |
| 4459 */ | |
| 4460 | |
| 4461 static int size = 0; | |
| 4462 | |
| 4463 /* Set the set size */ | |
| 4464 void SetSize(int n) | |
| 4465 { | |
| 4466 size = n+1; | |
| 4467 } | |
| 4468 | |
| 4469 /* Allocate a new set */ | |
| 4470 char *SetNew(){ | |
| 4471 char *s; | |
| 4472 s = (char*)calloc( size, 1); | |
| 4473 if( s==0 ){ | |
| 4474 extern void memory_error(); | |
| 4475 memory_error(); | |
| 4476 } | |
| 4477 return s; | |
| 4478 } | |
| 4479 | |
| 4480 /* Deallocate a set */ | |
| 4481 void SetFree(char *s) | |
| 4482 { | |
| 4483 free(s); | |
| 4484 } | |
| 4485 | |
| 4486 /* Add a new element to the set. Return TRUE if the element was added | |
| 4487 ** and FALSE if it was already there. */ | |
| 4488 int SetAdd(char *s, int e) | |
| 4489 { | |
| 4490 int rv; | |
| 4491 assert( e>=0 && e<size ); | |
| 4492 rv = s[e]; | |
| 4493 s[e] = 1; | |
| 4494 return !rv; | |
| 4495 } | |
| 4496 | |
| 4497 /* Add every element of s2 to s1. Return TRUE if s1 changes. */ | |
| 4498 int SetUnion(char *s1, char *s2) | |
| 4499 { | |
| 4500 int i, progress; | |
| 4501 progress = 0; | |
| 4502 for(i=0; i<size; i++){ | |
| 4503 if( s2[i]==0 ) continue; | |
| 4504 if( s1[i]==0 ){ | |
| 4505 progress = 1; | |
| 4506 s1[i] = 1; | |
| 4507 } | |
| 4508 } | |
| 4509 return progress; | |
| 4510 } | |
| 4511 /********************** From the file "table.c" ****************************/ | |
| 4512 /* | |
| 4513 ** All code in this file has been automatically generated | |
| 4514 ** from a specification in the file | |
| 4515 ** "table.q" | |
| 4516 ** by the associative array code building program "aagen". | |
| 4517 ** Do not edit this file! Instead, edit the specification | |
| 4518 ** file, then rerun aagen. | |
| 4519 */ | |
| 4520 /* | |
| 4521 ** Code for processing tables in the LEMON parser generator. | |
| 4522 */ | |
| 4523 | |
| 4524 PRIVATE unsigned strhash(const char *x) | |
| 4525 { | |
| 4526 unsigned h = 0; | |
| 4527 while( *x ) h = h*13 + *(x++); | |
| 4528 return h; | |
| 4529 } | |
| 4530 | |
| 4531 /* Works like strdup, sort of. Save a string in malloced memory, but | |
| 4532 ** keep strings in a table so that the same string is not in more | |
| 4533 ** than one place. | |
| 4534 */ | |
| 4535 const char *Strsafe(const char *y) | |
| 4536 { | |
| 4537 const char *z; | |
| 4538 char *cpy; | |
| 4539 | |
| 4540 if( y==0 ) return 0; | |
| 4541 z = Strsafe_find(y); | |
| 4542 if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){ | |
| 4543 lemon_strcpy(cpy,y); | |
| 4544 z = cpy; | |
| 4545 Strsafe_insert(z); | |
| 4546 } | |
| 4547 MemoryCheck(z); | |
| 4548 return z; | |
| 4549 } | |
| 4550 | |
| 4551 /* There is one instance of the following structure for each | |
| 4552 ** associative array of type "x1". | |
| 4553 */ | |
| 4554 struct s_x1 { | |
| 4555 int size; /* The number of available slots. */ | |
| 4556 /* Must be a power of 2 greater than or */ | |
| 4557 /* equal to 1 */ | |
| 4558 int count; /* Number of currently slots filled */ | |
| 4559 struct s_x1node *tbl; /* The data stored here */ | |
| 4560 struct s_x1node **ht; /* Hash table for lookups */ | |
| 4561 }; | |
| 4562 | |
| 4563 /* There is one instance of this structure for every data element | |
| 4564 ** in an associative array of type "x1". | |
| 4565 */ | |
| 4566 typedef struct s_x1node { | |
| 4567 const char *data; /* The data */ | |
| 4568 struct s_x1node *next; /* Next entry with the same hash */ | |
| 4569 struct s_x1node **from; /* Previous link */ | |
| 4570 } x1node; | |
| 4571 | |
| 4572 /* There is only one instance of the array, which is the following */ | |
| 4573 static struct s_x1 *x1a; | |
| 4574 | |
| 4575 /* Allocate a new associative array */ | |
| 4576 void Strsafe_init(){ | |
| 4577 if( x1a ) return; | |
| 4578 x1a = (struct s_x1*)malloc( sizeof(struct s_x1) ); | |
| 4579 if( x1a ){ | |
| 4580 x1a->size = 1024; | |
| 4581 x1a->count = 0; | |
| 4582 x1a->tbl = (x1node*)calloc(1024, sizeof(x1node) + sizeof(x1node*)); | |
| 4583 if( x1a->tbl==0 ){ | |
| 4584 free(x1a); | |
| 4585 x1a = 0; | |
| 4586 }else{ | |
| 4587 int i; | |
| 4588 x1a->ht = (x1node**)&(x1a->tbl[1024]); | |
| 4589 for(i=0; i<1024; i++) x1a->ht[i] = 0; | |
| 4590 } | |
| 4591 } | |
| 4592 } | |
| 4593 /* Insert a new record into the array. Return TRUE if successful. | |
| 4594 ** Prior data with the same key is NOT overwritten */ | |
| 4595 int Strsafe_insert(const char *data) | |
| 4596 { | |
| 4597 x1node *np; | |
| 4598 unsigned h; | |
| 4599 unsigned ph; | |
| 4600 | |
| 4601 if( x1a==0 ) return 0; | |
| 4602 ph = strhash(data); | |
| 4603 h = ph & (x1a->size-1); | |
| 4604 np = x1a->ht[h]; | |
| 4605 while( np ){ | |
| 4606 if( strcmp(np->data,data)==0 ){ | |
| 4607 /* An existing entry with the same key is found. */ | |
| 4608 /* Fail because overwrite is not allows. */ | |
| 4609 return 0; | |
| 4610 } | |
| 4611 np = np->next; | |
| 4612 } | |
| 4613 if( x1a->count>=x1a->size ){ | |
| 4614 /* Need to make the hash table bigger */ | |
| 4615 int i,arrSize; | |
| 4616 struct s_x1 array; | |
| 4617 array.size = arrSize = x1a->size*2; | |
| 4618 array.count = x1a->count; | |
| 4619 array.tbl = (x1node*)calloc(arrSize, sizeof(x1node) + sizeof(x1node*)); | |
| 4620 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ | |
| 4621 array.ht = (x1node**)&(array.tbl[arrSize]); | |
| 4622 for(i=0; i<arrSize; i++) array.ht[i] = 0; | |
| 4623 for(i=0; i<x1a->count; i++){ | |
| 4624 x1node *oldnp, *newnp; | |
| 4625 oldnp = &(x1a->tbl[i]); | |
| 4626 h = strhash(oldnp->data) & (arrSize-1); | |
| 4627 newnp = &(array.tbl[i]); | |
| 4628 if( array.ht[h] ) array.ht[h]->from = &(newnp->next); | |
| 4629 newnp->next = array.ht[h]; | |
| 4630 newnp->data = oldnp->data; | |
| 4631 newnp->from = &(array.ht[h]); | |
| 4632 array.ht[h] = newnp; | |
| 4633 } | |
| 4634 free(x1a->tbl); | |
| 4635 *x1a = array; | |
| 4636 } | |
| 4637 /* Insert the new data */ | |
| 4638 h = ph & (x1a->size-1); | |
| 4639 np = &(x1a->tbl[x1a->count++]); | |
| 4640 np->data = data; | |
| 4641 if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next); | |
| 4642 np->next = x1a->ht[h]; | |
| 4643 x1a->ht[h] = np; | |
| 4644 np->from = &(x1a->ht[h]); | |
| 4645 return 1; | |
| 4646 } | |
| 4647 | |
| 4648 /* Return a pointer to data assigned to the given key. Return NULL | |
| 4649 ** if no such key. */ | |
| 4650 const char *Strsafe_find(const char *key) | |
| 4651 { | |
| 4652 unsigned h; | |
| 4653 x1node *np; | |
| 4654 | |
| 4655 if( x1a==0 ) return 0; | |
| 4656 h = strhash(key) & (x1a->size-1); | |
| 4657 np = x1a->ht[h]; | |
| 4658 while( np ){ | |
| 4659 if( strcmp(np->data,key)==0 ) break; | |
| 4660 np = np->next; | |
| 4661 } | |
| 4662 return np ? np->data : 0; | |
| 4663 } | |
| 4664 | |
| 4665 /* Return a pointer to the (terminal or nonterminal) symbol "x". | |
| 4666 ** Create a new symbol if this is the first time "x" has been seen. | |
| 4667 */ | |
| 4668 struct symbol *Symbol_new(const char *x) | |
| 4669 { | |
| 4670 struct symbol *sp; | |
| 4671 | |
| 4672 sp = Symbol_find(x); | |
| 4673 if( sp==0 ){ | |
| 4674 sp = (struct symbol *)calloc(1, sizeof(struct symbol) ); | |
| 4675 MemoryCheck(sp); | |
| 4676 sp->name = Strsafe(x); | |
| 4677 sp->type = ISUPPER(*x) ? TERMINAL : NONTERMINAL; | |
| 4678 sp->rule = 0; | |
| 4679 sp->fallback = 0; | |
| 4680 sp->prec = -1; | |
| 4681 sp->assoc = UNK; | |
| 4682 sp->firstset = 0; | |
| 4683 sp->lambda = LEMON_FALSE; | |
| 4684 sp->destructor = 0; | |
| 4685 sp->destLineno = 0; | |
| 4686 sp->datatype = 0; | |
| 4687 sp->useCnt = 0; | |
| 4688 Symbol_insert(sp,sp->name); | |
| 4689 } | |
| 4690 sp->useCnt++; | |
| 4691 return sp; | |
| 4692 } | |
| 4693 | |
| 4694 /* Compare two symbols for sorting purposes. Return negative, | |
| 4695 ** zero, or positive if a is less then, equal to, or greater | |
| 4696 ** than b. | |
| 4697 ** | |
| 4698 ** Symbols that begin with upper case letters (terminals or tokens) | |
| 4699 ** must sort before symbols that begin with lower case letters | |
| 4700 ** (non-terminals). And MULTITERMINAL symbols (created using the | |
| 4701 ** %token_class directive) must sort at the very end. Other than | |
| 4702 ** that, the order does not matter. | |
| 4703 ** | |
| 4704 ** We find experimentally that leaving the symbols in their original | |
| 4705 ** order (the order they appeared in the grammar file) gives the | |
| 4706 ** smallest parser tables in SQLite. | |
| 4707 */ | |
| 4708 int Symbolcmpp(const void *_a, const void *_b) | |
| 4709 { | |
| 4710 const struct symbol *a = *(const struct symbol **) _a; | |
| 4711 const struct symbol *b = *(const struct symbol **) _b; | |
| 4712 int i1 = a->type==MULTITERMINAL ? 3 : a->name[0]>'Z' ? 2 : 1; | |
| 4713 int i2 = b->type==MULTITERMINAL ? 3 : b->name[0]>'Z' ? 2 : 1; | |
| 4714 return i1==i2 ? a->index - b->index : i1 - i2; | |
| 4715 } | |
| 4716 | |
| 4717 /* There is one instance of the following structure for each | |
| 4718 ** associative array of type "x2". | |
| 4719 */ | |
| 4720 struct s_x2 { | |
| 4721 int size; /* The number of available slots. */ | |
| 4722 /* Must be a power of 2 greater than or */ | |
| 4723 /* equal to 1 */ | |
| 4724 int count; /* Number of currently slots filled */ | |
| 4725 struct s_x2node *tbl; /* The data stored here */ | |
| 4726 struct s_x2node **ht; /* Hash table for lookups */ | |
| 4727 }; | |
| 4728 | |
| 4729 /* There is one instance of this structure for every data element | |
| 4730 ** in an associative array of type "x2". | |
| 4731 */ | |
| 4732 typedef struct s_x2node { | |
| 4733 struct symbol *data; /* The data */ | |
| 4734 const char *key; /* The key */ | |
| 4735 struct s_x2node *next; /* Next entry with the same hash */ | |
| 4736 struct s_x2node **from; /* Previous link */ | |
| 4737 } x2node; | |
| 4738 | |
| 4739 /* There is only one instance of the array, which is the following */ | |
| 4740 static struct s_x2 *x2a; | |
| 4741 | |
| 4742 /* Allocate a new associative array */ | |
| 4743 void Symbol_init(){ | |
| 4744 if( x2a ) return; | |
| 4745 x2a = (struct s_x2*)malloc( sizeof(struct s_x2) ); | |
| 4746 if( x2a ){ | |
| 4747 x2a->size = 128; | |
| 4748 x2a->count = 0; | |
| 4749 x2a->tbl = (x2node*)calloc(128, sizeof(x2node) + sizeof(x2node*)); | |
| 4750 if( x2a->tbl==0 ){ | |
| 4751 free(x2a); | |
| 4752 x2a = 0; | |
| 4753 }else{ | |
| 4754 int i; | |
| 4755 x2a->ht = (x2node**)&(x2a->tbl[128]); | |
| 4756 for(i=0; i<128; i++) x2a->ht[i] = 0; | |
| 4757 } | |
| 4758 } | |
| 4759 } | |
| 4760 /* Insert a new record into the array. Return TRUE if successful. | |
| 4761 ** Prior data with the same key is NOT overwritten */ | |
| 4762 int Symbol_insert(struct symbol *data, const char *key) | |
| 4763 { | |
| 4764 x2node *np; | |
| 4765 unsigned h; | |
| 4766 unsigned ph; | |
| 4767 | |
| 4768 if( x2a==0 ) return 0; | |
| 4769 ph = strhash(key); | |
| 4770 h = ph & (x2a->size-1); | |
| 4771 np = x2a->ht[h]; | |
| 4772 while( np ){ | |
| 4773 if( strcmp(np->key,key)==0 ){ | |
| 4774 /* An existing entry with the same key is found. */ | |
| 4775 /* Fail because overwrite is not allows. */ | |
| 4776 return 0; | |
| 4777 } | |
| 4778 np = np->next; | |
| 4779 } | |
| 4780 if( x2a->count>=x2a->size ){ | |
| 4781 /* Need to make the hash table bigger */ | |
| 4782 int i,arrSize; | |
| 4783 struct s_x2 array; | |
| 4784 array.size = arrSize = x2a->size*2; | |
| 4785 array.count = x2a->count; | |
| 4786 array.tbl = (x2node*)calloc(arrSize, sizeof(x2node) + sizeof(x2node*)); | |
| 4787 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ | |
| 4788 array.ht = (x2node**)&(array.tbl[arrSize]); | |
| 4789 for(i=0; i<arrSize; i++) array.ht[i] = 0; | |
| 4790 for(i=0; i<x2a->count; i++){ | |
| 4791 x2node *oldnp, *newnp; | |
| 4792 oldnp = &(x2a->tbl[i]); | |
| 4793 h = strhash(oldnp->key) & (arrSize-1); | |
| 4794 newnp = &(array.tbl[i]); | |
| 4795 if( array.ht[h] ) array.ht[h]->from = &(newnp->next); | |
| 4796 newnp->next = array.ht[h]; | |
| 4797 newnp->key = oldnp->key; | |
| 4798 newnp->data = oldnp->data; | |
| 4799 newnp->from = &(array.ht[h]); | |
| 4800 array.ht[h] = newnp; | |
| 4801 } | |
| 4802 free(x2a->tbl); | |
| 4803 *x2a = array; | |
| 4804 } | |
| 4805 /* Insert the new data */ | |
| 4806 h = ph & (x2a->size-1); | |
| 4807 np = &(x2a->tbl[x2a->count++]); | |
| 4808 np->key = key; | |
| 4809 np->data = data; | |
| 4810 if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next); | |
| 4811 np->next = x2a->ht[h]; | |
| 4812 x2a->ht[h] = np; | |
| 4813 np->from = &(x2a->ht[h]); | |
| 4814 return 1; | |
| 4815 } | |
| 4816 | |
| 4817 /* Return a pointer to data assigned to the given key. Return NULL | |
| 4818 ** if no such key. */ | |
| 4819 struct symbol *Symbol_find(const char *key) | |
| 4820 { | |
| 4821 unsigned h; | |
| 4822 x2node *np; | |
| 4823 | |
| 4824 if( x2a==0 ) return 0; | |
| 4825 h = strhash(key) & (x2a->size-1); | |
| 4826 np = x2a->ht[h]; | |
| 4827 while( np ){ | |
| 4828 if( strcmp(np->key,key)==0 ) break; | |
| 4829 np = np->next; | |
| 4830 } | |
| 4831 return np ? np->data : 0; | |
| 4832 } | |
| 4833 | |
| 4834 /* Return the n-th data. Return NULL if n is out of range. */ | |
| 4835 struct symbol *Symbol_Nth(int n) | |
| 4836 { | |
| 4837 struct symbol *data; | |
| 4838 if( x2a && n>0 && n<=x2a->count ){ | |
| 4839 data = x2a->tbl[n-1].data; | |
| 4840 }else{ | |
| 4841 data = 0; | |
| 4842 } | |
| 4843 return data; | |
| 4844 } | |
| 4845 | |
| 4846 /* Return the size of the array */ | |
| 4847 int Symbol_count() | |
| 4848 { | |
| 4849 return x2a ? x2a->count : 0; | |
| 4850 } | |
| 4851 | |
| 4852 /* Return an array of pointers to all data in the table. | |
| 4853 ** The array is obtained from malloc. Return NULL if memory allocation | |
| 4854 ** problems, or if the array is empty. */ | |
| 4855 struct symbol **Symbol_arrayof() | |
| 4856 { | |
| 4857 struct symbol **array; | |
| 4858 int i,arrSize; | |
| 4859 if( x2a==0 ) return 0; | |
| 4860 arrSize = x2a->count; | |
| 4861 array = (struct symbol **)calloc(arrSize, sizeof(struct symbol *)); | |
| 4862 if( array ){ | |
| 4863 for(i=0; i<arrSize; i++) array[i] = x2a->tbl[i].data; | |
| 4864 } | |
| 4865 return array; | |
| 4866 } | |
| 4867 | |
| 4868 /* Compare two configurations */ | |
| 4869 int Configcmp(const char *_a,const char *_b) | |
| 4870 { | |
| 4871 const struct config *a = (struct config *) _a; | |
| 4872 const struct config *b = (struct config *) _b; | |
| 4873 int x; | |
| 4874 x = a->rp->index - b->rp->index; | |
| 4875 if( x==0 ) x = a->dot - b->dot; | |
| 4876 return x; | |
| 4877 } | |
| 4878 | |
| 4879 /* Compare two states */ | |
| 4880 PRIVATE int statecmp(struct config *a, struct config *b) | |
| 4881 { | |
| 4882 int rc; | |
| 4883 for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){ | |
| 4884 rc = a->rp->index - b->rp->index; | |
| 4885 if( rc==0 ) rc = a->dot - b->dot; | |
| 4886 } | |
| 4887 if( rc==0 ){ | |
| 4888 if( a ) rc = 1; | |
| 4889 if( b ) rc = -1; | |
| 4890 } | |
| 4891 return rc; | |
| 4892 } | |
| 4893 | |
| 4894 /* Hash a state */ | |
| 4895 PRIVATE unsigned statehash(struct config *a) | |
| 4896 { | |
| 4897 unsigned h=0; | |
| 4898 while( a ){ | |
| 4899 h = h*571 + a->rp->index*37 + a->dot; | |
| 4900 a = a->bp; | |
| 4901 } | |
| 4902 return h; | |
| 4903 } | |
| 4904 | |
| 4905 /* Allocate a new state structure */ | |
| 4906 struct state *State_new() | |
| 4907 { | |
| 4908 struct state *newstate; | |
| 4909 newstate = (struct state *)calloc(1, sizeof(struct state) ); | |
| 4910 MemoryCheck(newstate); | |
| 4911 return newstate; | |
| 4912 } | |
| 4913 | |
| 4914 /* There is one instance of the following structure for each | |
| 4915 ** associative array of type "x3". | |
| 4916 */ | |
| 4917 struct s_x3 { | |
| 4918 int size; /* The number of available slots. */ | |
| 4919 /* Must be a power of 2 greater than or */ | |
| 4920 /* equal to 1 */ | |
| 4921 int count; /* Number of currently slots filled */ | |
| 4922 struct s_x3node *tbl; /* The data stored here */ | |
| 4923 struct s_x3node **ht; /* Hash table for lookups */ | |
| 4924 }; | |
| 4925 | |
| 4926 /* There is one instance of this structure for every data element | |
| 4927 ** in an associative array of type "x3". | |
| 4928 */ | |
| 4929 typedef struct s_x3node { | |
| 4930 struct state *data; /* The data */ | |
| 4931 struct config *key; /* The key */ | |
| 4932 struct s_x3node *next; /* Next entry with the same hash */ | |
| 4933 struct s_x3node **from; /* Previous link */ | |
| 4934 } x3node; | |
| 4935 | |
| 4936 /* There is only one instance of the array, which is the following */ | |
| 4937 static struct s_x3 *x3a; | |
| 4938 | |
| 4939 /* Allocate a new associative array */ | |
| 4940 void State_init(){ | |
| 4941 if( x3a ) return; | |
| 4942 x3a = (struct s_x3*)malloc( sizeof(struct s_x3) ); | |
| 4943 if( x3a ){ | |
| 4944 x3a->size = 128; | |
| 4945 x3a->count = 0; | |
| 4946 x3a->tbl = (x3node*)calloc(128, sizeof(x3node) + sizeof(x3node*)); | |
| 4947 if( x3a->tbl==0 ){ | |
| 4948 free(x3a); | |
| 4949 x3a = 0; | |
| 4950 }else{ | |
| 4951 int i; | |
| 4952 x3a->ht = (x3node**)&(x3a->tbl[128]); | |
| 4953 for(i=0; i<128; i++) x3a->ht[i] = 0; | |
| 4954 } | |
| 4955 } | |
| 4956 } | |
| 4957 /* Insert a new record into the array. Return TRUE if successful. | |
| 4958 ** Prior data with the same key is NOT overwritten */ | |
| 4959 int State_insert(struct state *data, struct config *key) | |
| 4960 { | |
| 4961 x3node *np; | |
| 4962 unsigned h; | |
| 4963 unsigned ph; | |
| 4964 | |
| 4965 if( x3a==0 ) return 0; | |
| 4966 ph = statehash(key); | |
| 4967 h = ph & (x3a->size-1); | |
| 4968 np = x3a->ht[h]; | |
| 4969 while( np ){ | |
| 4970 if( statecmp(np->key,key)==0 ){ | |
| 4971 /* An existing entry with the same key is found. */ | |
| 4972 /* Fail because overwrite is not allows. */ | |
| 4973 return 0; | |
| 4974 } | |
| 4975 np = np->next; | |
| 4976 } | |
| 4977 if( x3a->count>=x3a->size ){ | |
| 4978 /* Need to make the hash table bigger */ | |
| 4979 int i,arrSize; | |
| 4980 struct s_x3 array; | |
| 4981 array.size = arrSize = x3a->size*2; | |
| 4982 array.count = x3a->count; | |
| 4983 array.tbl = (x3node*)calloc(arrSize, sizeof(x3node) + sizeof(x3node*)); | |
| 4984 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ | |
| 4985 array.ht = (x3node**)&(array.tbl[arrSize]); | |
| 4986 for(i=0; i<arrSize; i++) array.ht[i] = 0; | |
| 4987 for(i=0; i<x3a->count; i++){ | |
| 4988 x3node *oldnp, *newnp; | |
| 4989 oldnp = &(x3a->tbl[i]); | |
| 4990 h = statehash(oldnp->key) & (arrSize-1); | |
| 4991 newnp = &(array.tbl[i]); | |
| 4992 if( array.ht[h] ) array.ht[h]->from = &(newnp->next); | |
| 4993 newnp->next = array.ht[h]; | |
| 4994 newnp->key = oldnp->key; | |
| 4995 newnp->data = oldnp->data; | |
| 4996 newnp->from = &(array.ht[h]); | |
| 4997 array.ht[h] = newnp; | |
| 4998 } | |
| 4999 free(x3a->tbl); | |
| 5000 *x3a = array; | |
| 5001 } | |
| 5002 /* Insert the new data */ | |
| 5003 h = ph & (x3a->size-1); | |
| 5004 np = &(x3a->tbl[x3a->count++]); | |
| 5005 np->key = key; | |
| 5006 np->data = data; | |
| 5007 if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next); | |
| 5008 np->next = x3a->ht[h]; | |
| 5009 x3a->ht[h] = np; | |
| 5010 np->from = &(x3a->ht[h]); | |
| 5011 return 1; | |
| 5012 } | |
| 5013 | |
| 5014 /* Return a pointer to data assigned to the given key. Return NULL | |
| 5015 ** if no such key. */ | |
| 5016 struct state *State_find(struct config *key) | |
| 5017 { | |
| 5018 unsigned h; | |
| 5019 x3node *np; | |
| 5020 | |
| 5021 if( x3a==0 ) return 0; | |
| 5022 h = statehash(key) & (x3a->size-1); | |
| 5023 np = x3a->ht[h]; | |
| 5024 while( np ){ | |
| 5025 if( statecmp(np->key,key)==0 ) break; | |
| 5026 np = np->next; | |
| 5027 } | |
| 5028 return np ? np->data : 0; | |
| 5029 } | |
| 5030 | |
| 5031 /* Return an array of pointers to all data in the table. | |
| 5032 ** The array is obtained from malloc. Return NULL if memory allocation | |
| 5033 ** problems, or if the array is empty. */ | |
| 5034 struct state **State_arrayof() | |
| 5035 { | |
| 5036 struct state **array; | |
| 5037 int i,arrSize; | |
| 5038 if( x3a==0 ) return 0; | |
| 5039 arrSize = x3a->count; | |
| 5040 array = (struct state **)calloc(arrSize, sizeof(struct state *)); | |
| 5041 if( array ){ | |
| 5042 for(i=0; i<arrSize; i++) array[i] = x3a->tbl[i].data; | |
| 5043 } | |
| 5044 return array; | |
| 5045 } | |
| 5046 | |
| 5047 /* Hash a configuration */ | |
| 5048 PRIVATE unsigned confighash(struct config *a) | |
| 5049 { | |
| 5050 unsigned h=0; | |
| 5051 h = h*571 + a->rp->index*37 + a->dot; | |
| 5052 return h; | |
| 5053 } | |
| 5054 | |
| 5055 /* There is one instance of the following structure for each | |
| 5056 ** associative array of type "x4". | |
| 5057 */ | |
| 5058 struct s_x4 { | |
| 5059 int size; /* The number of available slots. */ | |
| 5060 /* Must be a power of 2 greater than or */ | |
| 5061 /* equal to 1 */ | |
| 5062 int count; /* Number of currently slots filled */ | |
| 5063 struct s_x4node *tbl; /* The data stored here */ | |
| 5064 struct s_x4node **ht; /* Hash table for lookups */ | |
| 5065 }; | |
| 5066 | |
| 5067 /* There is one instance of this structure for every data element | |
| 5068 ** in an associative array of type "x4". | |
| 5069 */ | |
| 5070 typedef struct s_x4node { | |
| 5071 struct config *data; /* The data */ | |
| 5072 struct s_x4node *next; /* Next entry with the same hash */ | |
| 5073 struct s_x4node **from; /* Previous link */ | |
| 5074 } x4node; | |
| 5075 | |
| 5076 /* There is only one instance of the array, which is the following */ | |
| 5077 static struct s_x4 *x4a; | |
| 5078 | |
| 5079 /* Allocate a new associative array */ | |
| 5080 void Configtable_init(){ | |
| 5081 if( x4a ) return; | |
| 5082 x4a = (struct s_x4*)malloc( sizeof(struct s_x4) ); | |
| 5083 if( x4a ){ | |
| 5084 x4a->size = 64; | |
| 5085 x4a->count = 0; | |
| 5086 x4a->tbl = (x4node*)calloc(64, sizeof(x4node) + sizeof(x4node*)); | |
| 5087 if( x4a->tbl==0 ){ | |
| 5088 free(x4a); | |
| 5089 x4a = 0; | |
| 5090 }else{ | |
| 5091 int i; | |
| 5092 x4a->ht = (x4node**)&(x4a->tbl[64]); | |
| 5093 for(i=0; i<64; i++) x4a->ht[i] = 0; | |
| 5094 } | |
| 5095 } | |
| 5096 } | |
| 5097 /* Insert a new record into the array. Return TRUE if successful. | |
| 5098 ** Prior data with the same key is NOT overwritten */ | |
| 5099 int Configtable_insert(struct config *data) | |
| 5100 { | |
| 5101 x4node *np; | |
| 5102 unsigned h; | |
| 5103 unsigned ph; | |
| 5104 | |
| 5105 if( x4a==0 ) return 0; | |
| 5106 ph = confighash(data); | |
| 5107 h = ph & (x4a->size-1); | |
| 5108 np = x4a->ht[h]; | |
| 5109 while( np ){ | |
| 5110 if( Configcmp((const char *) np->data,(const char *) data)==0 ){ | |
| 5111 /* An existing entry with the same key is found. */ | |
| 5112 /* Fail because overwrite is not allows. */ | |
| 5113 return 0; | |
| 5114 } | |
| 5115 np = np->next; | |
| 5116 } | |
| 5117 if( x4a->count>=x4a->size ){ | |
| 5118 /* Need to make the hash table bigger */ | |
| 5119 int i,arrSize; | |
| 5120 struct s_x4 array; | |
| 5121 array.size = arrSize = x4a->size*2; | |
| 5122 array.count = x4a->count; | |
| 5123 array.tbl = (x4node*)calloc(arrSize, sizeof(x4node) + sizeof(x4node*)); | |
| 5124 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ | |
| 5125 array.ht = (x4node**)&(array.tbl[arrSize]); | |
| 5126 for(i=0; i<arrSize; i++) array.ht[i] = 0; | |
| 5127 for(i=0; i<x4a->count; i++){ | |
| 5128 x4node *oldnp, *newnp; | |
| 5129 oldnp = &(x4a->tbl[i]); | |
| 5130 h = confighash(oldnp->data) & (arrSize-1); | |
| 5131 newnp = &(array.tbl[i]); | |
| 5132 if( array.ht[h] ) array.ht[h]->from = &(newnp->next); | |
| 5133 newnp->next = array.ht[h]; | |
| 5134 newnp->data = oldnp->data; | |
| 5135 newnp->from = &(array.ht[h]); | |
| 5136 array.ht[h] = newnp; | |
| 5137 } | |
| 5138 free(x4a->tbl); | |
| 5139 *x4a = array; | |
| 5140 } | |
| 5141 /* Insert the new data */ | |
| 5142 h = ph & (x4a->size-1); | |
| 5143 np = &(x4a->tbl[x4a->count++]); | |
| 5144 np->data = data; | |
| 5145 if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next); | |
| 5146 np->next = x4a->ht[h]; | |
| 5147 x4a->ht[h] = np; | |
| 5148 np->from = &(x4a->ht[h]); | |
| 5149 return 1; | |
| 5150 } | |
| 5151 | |
| 5152 /* Return a pointer to data assigned to the given key. Return NULL | |
| 5153 ** if no such key. */ | |
| 5154 struct config *Configtable_find(struct config *key) | |
| 5155 { | |
| 5156 int h; | |
| 5157 x4node *np; | |
| 5158 | |
| 5159 if( x4a==0 ) return 0; | |
| 5160 h = confighash(key) & (x4a->size-1); | |
| 5161 np = x4a->ht[h]; | |
| 5162 while( np ){ | |
| 5163 if( Configcmp((const char *) np->data,(const char *) key)==0 ) break; | |
| 5164 np = np->next; | |
| 5165 } | |
| 5166 return np ? np->data : 0; | |
| 5167 } | |
| 5168 | |
| 5169 /* Remove all data from the table. Pass each data to the function "f" | |
| 5170 ** as it is removed. ("f" may be null to avoid this step.) */ | |
| 5171 void Configtable_clear(int(*f)(struct config *)) | |
| 5172 { | |
| 5173 int i; | |
| 5174 if( x4a==0 || x4a->count==0 ) return; | |
| 5175 if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data); | |
| 5176 for(i=0; i<x4a->size; i++) x4a->ht[i] = 0; | |
| 5177 x4a->count = 0; | |
| 5178 return; | |
| 5179 } | |
| OLD | NEW |