replace use of sqlite3StrICmp with public sqlite3_stricmp API
[sqlcipher.git] / tool / lemon.c
blob75fc7aa2fb9988e49ba4754792f05d14b908dcba
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>
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))
24 #ifndef __WIN32__
25 # if defined(_WIN32) || defined(WIN32)
26 # define __WIN32__
27 # endif
28 #endif
30 #ifdef __WIN32__
31 #ifdef __cplusplus
32 extern "C" {
33 #endif
34 extern int access(const char *path, int mode);
35 #ifdef __cplusplus
37 #endif
38 #else
39 #include <unistd.h>
40 #endif
42 /* #define PRIVATE static */
43 #define PRIVATE
45 #ifdef TEST
46 #define MAXRHS 5 /* Set low to exercise exception code */
47 #else
48 #define MAXRHS 1000
49 #endif
51 extern void memory_error();
52 static int showPrecedenceConflict = 0;
53 static char *msort(char*,char**,int(*)(const char*,const char*));
56 ** Compilers are getting increasingly pedantic about type conversions
57 ** as C evolves ever closer to Ada.... To work around the latest problems
58 ** we have to define the following variant of strlen().
60 #define lemonStrlen(X) ((int)strlen(X))
63 ** Compilers are starting to complain about the use of sprintf() and strcpy(),
64 ** saying they are unsafe. So we define our own versions of those routines too.
66 ** There are three routines here: lemon_sprintf(), lemon_vsprintf(), and
67 ** lemon_addtext(). The first two are replacements for sprintf() and vsprintf().
68 ** The third is a helper routine for vsnprintf() that adds texts to the end of a
69 ** buffer, making sure the buffer is always zero-terminated.
71 ** The string formatter is a minimal subset of stdlib sprintf() supporting only
72 ** a few simply conversions:
74 ** %d
75 ** %s
76 ** %.*s
79 static void lemon_addtext(
80 char *zBuf, /* The buffer to which text is added */
81 int *pnUsed, /* Slots of the buffer used so far */
82 const char *zIn, /* Text to add */
83 int nIn, /* Bytes of text to add. -1 to use strlen() */
84 int iWidth /* Field width. Negative to left justify */
86 if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){}
87 while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; }
88 if( nIn==0 ) return;
89 memcpy(&zBuf[*pnUsed], zIn, nIn);
90 *pnUsed += nIn;
91 while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; }
92 zBuf[*pnUsed] = 0;
94 static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){
95 int i, j, k, c;
96 int nUsed = 0;
97 const char *z;
98 char zTemp[50];
99 str[0] = 0;
100 for(i=j=0; (c = zFormat[i])!=0; i++){
101 if( c=='%' ){
102 int iWidth = 0;
103 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
104 c = zFormat[++i];
105 if( ISDIGIT(c) || (c=='-' && ISDIGIT(zFormat[i+1])) ){
106 if( c=='-' ) i++;
107 while( ISDIGIT(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0';
108 if( c=='-' ) iWidth = -iWidth;
109 c = zFormat[i];
111 if( c=='d' ){
112 int v = va_arg(ap, int);
113 if( v<0 ){
114 lemon_addtext(str, &nUsed, "-", 1, iWidth);
115 v = -v;
116 }else if( v==0 ){
117 lemon_addtext(str, &nUsed, "0", 1, iWidth);
119 k = 0;
120 while( v>0 ){
121 k++;
122 zTemp[sizeof(zTemp)-k] = (v%10) + '0';
123 v /= 10;
125 lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth);
126 }else if( c=='s' ){
127 z = va_arg(ap, const char*);
128 lemon_addtext(str, &nUsed, z, -1, iWidth);
129 }else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){
130 i += 2;
131 k = va_arg(ap, int);
132 z = va_arg(ap, const char*);
133 lemon_addtext(str, &nUsed, z, k, iWidth);
134 }else if( c=='%' ){
135 lemon_addtext(str, &nUsed, "%", 1, 0);
136 }else{
137 fprintf(stderr, "illegal format\n");
138 exit(1);
140 j = i+1;
143 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
144 return nUsed;
146 static int lemon_sprintf(char *str, const char *format, ...){
147 va_list ap;
148 int rc;
149 va_start(ap, format);
150 rc = lemon_vsprintf(str, format, ap);
151 va_end(ap);
152 return rc;
154 static void lemon_strcpy(char *dest, const char *src){
155 while( (*(dest++) = *(src++))!=0 ){}
157 static void lemon_strcat(char *dest, const char *src){
158 while( *dest ) dest++;
159 lemon_strcpy(dest, src);
163 /* a few forward declarations... */
164 struct rule;
165 struct lemon;
166 struct action;
168 static struct action *Action_new(void);
169 static struct action *Action_sort(struct action *);
171 /********** From the file "build.h" ************************************/
172 void FindRulePrecedences(struct lemon*);
173 void FindFirstSets(struct lemon*);
174 void FindStates(struct lemon*);
175 void FindLinks(struct lemon*);
176 void FindFollowSets(struct lemon*);
177 void FindActions(struct lemon*);
179 /********* From the file "configlist.h" *********************************/
180 void Configlist_init(void);
181 struct config *Configlist_add(struct rule *, int);
182 struct config *Configlist_addbasis(struct rule *, int);
183 void Configlist_closure(struct lemon *);
184 void Configlist_sort(void);
185 void Configlist_sortbasis(void);
186 struct config *Configlist_return(void);
187 struct config *Configlist_basis(void);
188 void Configlist_eat(struct config *);
189 void Configlist_reset(void);
191 /********* From the file "error.h" ***************************************/
192 void ErrorMsg(const char *, int,const char *, ...);
194 /****** From the file "option.h" ******************************************/
195 enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
196 OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR};
197 struct s_options {
198 enum option_type type;
199 const char *label;
200 char *arg;
201 const char *message;
203 int OptInit(char**,struct s_options*,FILE*);
204 int OptNArgs(void);
205 char *OptArg(int);
206 void OptErr(int);
207 void OptPrint(void);
209 /******** From the file "parse.h" *****************************************/
210 void Parse(struct lemon *lemp);
212 /********* From the file "plink.h" ***************************************/
213 struct plink *Plink_new(void);
214 void Plink_add(struct plink **, struct config *);
215 void Plink_copy(struct plink **, struct plink *);
216 void Plink_delete(struct plink *);
218 /********** From the file "report.h" *************************************/
219 void Reprint(struct lemon *);
220 void ReportOutput(struct lemon *);
221 void ReportTable(struct lemon *, int, int);
222 void ReportHeader(struct lemon *);
223 void CompressTables(struct lemon *);
224 void ResortStates(struct lemon *);
226 /********** From the file "set.h" ****************************************/
227 void SetSize(int); /* All sets will be of size N */
228 char *SetNew(void); /* A new set for element 0..N */
229 void SetFree(char*); /* Deallocate a set */
230 int SetAdd(char*,int); /* Add element to a set */
231 int SetUnion(char *,char *); /* A <- A U B, thru element N */
232 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
234 /********** From the file "struct.h" *************************************/
236 ** Principal data structures for the LEMON parser generator.
239 typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
241 /* Symbols (terminals and nonterminals) of the grammar are stored
242 ** in the following: */
243 enum symbol_type {
244 TERMINAL,
245 NONTERMINAL,
246 MULTITERMINAL
248 enum e_assoc {
249 LEFT,
250 RIGHT,
251 NONE,
254 struct symbol {
255 const char *name; /* Name of the symbol */
256 int index; /* Index number for this symbol */
257 enum symbol_type type; /* Symbols are all either TERMINALS or NTs */
258 struct rule *rule; /* Linked list of rules of this (if an NT) */
259 struct symbol *fallback; /* fallback token in case this token doesn't parse */
260 int prec; /* Precedence if defined (-1 otherwise) */
261 enum e_assoc assoc; /* Associativity if precedence is defined */
262 char *firstset; /* First-set for all rules of this symbol */
263 Boolean lambda; /* True if NT and can generate an empty string */
264 int useCnt; /* Number of times used */
265 char *destructor; /* Code which executes whenever this symbol is
266 ** popped from the stack during error processing */
267 int destLineno; /* Line number for start of destructor. Set to
268 ** -1 for duplicate destructors. */
269 char *datatype; /* The data type of information held by this
270 ** object. Only used if type==NONTERMINAL */
271 int dtnum; /* The data type number. In the parser, the value
272 ** stack is a union. The .yy%d element of this
273 ** union is the correct data type for this object */
274 int bContent; /* True if this symbol ever carries content - if
275 ** it is ever more than just syntax */
276 /* The following fields are used by MULTITERMINALs only */
277 int nsubsym; /* Number of constituent symbols in the MULTI */
278 struct symbol **subsym; /* Array of constituent symbols */
281 /* Each production rule in the grammar is stored in the following
282 ** structure. */
283 struct rule {
284 struct symbol *lhs; /* Left-hand side of the rule */
285 const char *lhsalias; /* Alias for the LHS (NULL if none) */
286 int lhsStart; /* True if left-hand side is the start symbol */
287 int ruleline; /* Line number for the rule */
288 int nrhs; /* Number of RHS symbols */
289 struct symbol **rhs; /* The RHS symbols */
290 const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
291 int line; /* Line number at which code begins */
292 const char *code; /* The code executed when this rule is reduced */
293 const char *codePrefix; /* Setup code before code[] above */
294 const char *codeSuffix; /* Breakdown code after code[] above */
295 struct symbol *precsym; /* Precedence symbol for this rule */
296 int index; /* An index number for this rule */
297 int iRule; /* Rule number as used in the generated tables */
298 Boolean noCode; /* True if this rule has no associated C code */
299 Boolean codeEmitted; /* True if the code has been emitted already */
300 Boolean canReduce; /* True if this rule is ever reduced */
301 Boolean doesReduce; /* Reduce actions occur after optimization */
302 Boolean neverReduce; /* Reduce is theoretically possible, but prevented
303 ** by actions or other outside implementation */
304 struct rule *nextlhs; /* Next rule with the same LHS */
305 struct rule *next; /* Next rule in the global list */
308 /* A configuration is a production rule of the grammar together with
309 ** a mark (dot) showing how much of that rule has been processed so far.
310 ** Configurations also contain a follow-set which is a list of terminal
311 ** symbols which are allowed to immediately follow the end of the rule.
312 ** Every configuration is recorded as an instance of the following: */
313 enum cfgstatus {
314 COMPLETE,
315 INCOMPLETE
317 struct config {
318 struct rule *rp; /* The rule upon which the configuration is based */
319 int dot; /* The parse point */
320 char *fws; /* Follow-set for this configuration only */
321 struct plink *fplp; /* Follow-set forward propagation links */
322 struct plink *bplp; /* Follow-set backwards propagation links */
323 struct state *stp; /* Pointer to state which contains this */
324 enum cfgstatus status; /* used during followset and shift computations */
325 struct config *next; /* Next configuration in the state */
326 struct config *bp; /* The next basis configuration */
329 enum e_action {
330 SHIFT,
331 ACCEPT,
332 REDUCE,
333 ERROR,
334 SSCONFLICT, /* A shift/shift conflict */
335 SRCONFLICT, /* Was a reduce, but part of a conflict */
336 RRCONFLICT, /* Was a reduce, but part of a conflict */
337 SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
338 RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
339 NOT_USED, /* Deleted by compression */
340 SHIFTREDUCE /* Shift first, then reduce */
343 /* Every shift or reduce operation is stored as one of the following */
344 struct action {
345 struct symbol *sp; /* The look-ahead symbol */
346 enum e_action type;
347 union {
348 struct state *stp; /* The new state, if a shift */
349 struct rule *rp; /* The rule, if a reduce */
350 } x;
351 struct symbol *spOpt; /* SHIFTREDUCE optimization to this symbol */
352 struct action *next; /* Next action for this state */
353 struct action *collide; /* Next action with the same hash */
356 /* Each state of the generated parser's finite state machine
357 ** is encoded as an instance of the following structure. */
358 struct state {
359 struct config *bp; /* The basis configurations for this state */
360 struct config *cfp; /* All configurations in this set */
361 int statenum; /* Sequential number for this state */
362 struct action *ap; /* List of actions for this state */
363 int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
364 int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
365 int iDfltReduce; /* Default action is to REDUCE by this rule */
366 struct rule *pDfltReduce;/* The default REDUCE rule. */
367 int autoReduce; /* True if this is an auto-reduce state */
369 #define NO_OFFSET (-2147483647)
371 /* A followset propagation link indicates that the contents of one
372 ** configuration followset should be propagated to another whenever
373 ** the first changes. */
374 struct plink {
375 struct config *cfp; /* The configuration to which linked */
376 struct plink *next; /* The next propagate link */
379 /* The state vector for the entire parser generator is recorded as
380 ** follows. (LEMON uses no global variables and makes little use of
381 ** static variables. Fields in the following structure can be thought
382 ** of as begin global variables in the program.) */
383 struct lemon {
384 struct state **sorted; /* Table of states sorted by state number */
385 struct rule *rule; /* List of all rules */
386 struct rule *startRule; /* First rule */
387 int nstate; /* Number of states */
388 int nxstate; /* nstate with tail degenerate states removed */
389 int nrule; /* Number of rules */
390 int nruleWithAction; /* Number of rules with actions */
391 int nsymbol; /* Number of terminal and nonterminal symbols */
392 int nterminal; /* Number of terminal symbols */
393 int minShiftReduce; /* Minimum shift-reduce action value */
394 int errAction; /* Error action value */
395 int accAction; /* Accept action value */
396 int noAction; /* No-op action value */
397 int minReduce; /* Minimum reduce action */
398 int maxAction; /* Maximum action value of any kind */
399 struct symbol **symbols; /* Sorted array of pointers to symbols */
400 int errorcnt; /* Number of errors */
401 struct symbol *errsym; /* The error symbol */
402 struct symbol *wildcard; /* Token that matches anything */
403 char *name; /* Name of the generated parser */
404 char *arg; /* Declaration of the 3rd argument to parser */
405 char *ctx; /* Declaration of 2nd argument to constructor */
406 char *tokentype; /* Type of terminal symbols in the parser stack */
407 char *vartype; /* The default type of non-terminal symbols */
408 char *start; /* Name of the start symbol for the grammar */
409 char *stacksize; /* Size of the parser stack */
410 char *include; /* Code to put at the start of the C file */
411 char *error; /* Code to execute when an error is seen */
412 char *overflow; /* Code to execute on a stack overflow */
413 char *failure; /* Code to execute on parser failure */
414 char *accept; /* Code to execute when the parser excepts */
415 char *extracode; /* Code appended to the generated file */
416 char *tokendest; /* Code to execute to destroy token data */
417 char *vardest; /* Code for the default non-terminal destructor */
418 char *filename; /* Name of the input file */
419 char *outname; /* Name of the current output file */
420 char *tokenprefix; /* A prefix added to token names in the .h file */
421 int nconflict; /* Number of parsing conflicts */
422 int nactiontab; /* Number of entries in the yy_action[] table */
423 int nlookaheadtab; /* Number of entries in yy_lookahead[] */
424 int tablesize; /* Total table size of all tables in bytes */
425 int basisflag; /* Print only basis configurations */
426 int printPreprocessed; /* Show preprocessor output on stdout */
427 int has_fallback; /* True if any %fallback is seen in the grammar */
428 int nolinenosflag; /* True if #line statements should not be printed */
429 char *argv0; /* Name of the program */
432 #define MemoryCheck(X) if((X)==0){ \
433 extern void memory_error(); \
434 memory_error(); \
437 /**************** From the file "table.h" *********************************/
439 ** All code in this file has been automatically generated
440 ** from a specification in the file
441 ** "table.q"
442 ** by the associative array code building program "aagen".
443 ** Do not edit this file! Instead, edit the specification
444 ** file, then rerun aagen.
447 ** Code for processing tables in the LEMON parser generator.
449 /* Routines for handling a strings */
451 const char *Strsafe(const char *);
453 void Strsafe_init(void);
454 int Strsafe_insert(const char *);
455 const char *Strsafe_find(const char *);
457 /* Routines for handling symbols of the grammar */
459 struct symbol *Symbol_new(const char *);
460 int Symbolcmpp(const void *, const void *);
461 void Symbol_init(void);
462 int Symbol_insert(struct symbol *, const char *);
463 struct symbol *Symbol_find(const char *);
464 struct symbol *Symbol_Nth(int);
465 int Symbol_count(void);
466 struct symbol **Symbol_arrayof(void);
468 /* Routines to manage the state table */
470 int Configcmp(const char *, const char *);
471 struct state *State_new(void);
472 void State_init(void);
473 int State_insert(struct state *, struct config *);
474 struct state *State_find(struct config *);
475 struct state **State_arrayof(void);
477 /* Routines used for efficiency in Configlist_add */
479 void Configtable_init(void);
480 int Configtable_insert(struct config *);
481 struct config *Configtable_find(struct config *);
482 void Configtable_clear(int(*)(struct config *));
484 /****************** From the file "action.c" *******************************/
486 ** Routines processing parser actions in the LEMON parser generator.
489 /* Allocate a new parser action */
490 static struct action *Action_new(void){
491 static struct action *actionfreelist = 0;
492 struct action *newaction;
494 if( actionfreelist==0 ){
495 int i;
496 int amt = 100;
497 actionfreelist = (struct action *)calloc(amt, sizeof(struct action));
498 if( actionfreelist==0 ){
499 fprintf(stderr,"Unable to allocate memory for a new parser action.");
500 exit(1);
502 for(i=0; i<amt-1; i++) actionfreelist[i].next = &actionfreelist[i+1];
503 actionfreelist[amt-1].next = 0;
505 newaction = actionfreelist;
506 actionfreelist = actionfreelist->next;
507 return newaction;
510 /* Compare two actions for sorting purposes. Return negative, zero, or
511 ** positive if the first action is less than, equal to, or greater than
512 ** the first
514 static int actioncmp(
515 struct action *ap1,
516 struct action *ap2
518 int rc;
519 rc = ap1->sp->index - ap2->sp->index;
520 if( rc==0 ){
521 rc = (int)ap1->type - (int)ap2->type;
523 if( rc==0 && (ap1->type==REDUCE || ap1->type==SHIFTREDUCE) ){
524 rc = ap1->x.rp->index - ap2->x.rp->index;
526 if( rc==0 ){
527 rc = (int) (ap2 - ap1);
529 return rc;
532 /* Sort parser actions */
533 static struct action *Action_sort(
534 struct action *ap
536 ap = (struct action *)msort((char *)ap,(char **)&ap->next,
537 (int(*)(const char*,const char*))actioncmp);
538 return ap;
541 void Action_add(
542 struct action **app,
543 enum e_action type,
544 struct symbol *sp,
545 char *arg
547 struct action *newaction;
548 newaction = Action_new();
549 newaction->next = *app;
550 *app = newaction;
551 newaction->type = type;
552 newaction->sp = sp;
553 newaction->spOpt = 0;
554 if( type==SHIFT ){
555 newaction->x.stp = (struct state *)arg;
556 }else{
557 newaction->x.rp = (struct rule *)arg;
560 /********************** New code to implement the "acttab" module ***********/
562 ** This module implements routines use to construct the yy_action[] table.
566 ** The state of the yy_action table under construction is an instance of
567 ** the following structure.
569 ** The yy_action table maps the pair (state_number, lookahead) into an
570 ** action_number. The table is an array of integers pairs. The state_number
571 ** determines an initial offset into the yy_action array. The lookahead
572 ** value is then added to this initial offset to get an index X into the
573 ** yy_action array. If the aAction[X].lookahead equals the value of the
574 ** of the lookahead input, then the value of the action_number output is
575 ** aAction[X].action. If the lookaheads do not match then the
576 ** default action for the state_number is returned.
578 ** All actions associated with a single state_number are first entered
579 ** into aLookahead[] using multiple calls to acttab_action(). Then the
580 ** actions for that single state_number are placed into the aAction[]
581 ** array with a single call to acttab_insert(). The acttab_insert() call
582 ** also resets the aLookahead[] array in preparation for the next
583 ** state number.
585 struct lookahead_action {
586 int lookahead; /* Value of the lookahead token */
587 int action; /* Action to take on the given lookahead */
589 typedef struct acttab acttab;
590 struct acttab {
591 int nAction; /* Number of used slots in aAction[] */
592 int nActionAlloc; /* Slots allocated for aAction[] */
593 struct lookahead_action
594 *aAction, /* The yy_action[] table under construction */
595 *aLookahead; /* A single new transaction set */
596 int mnLookahead; /* Minimum aLookahead[].lookahead */
597 int mnAction; /* Action associated with mnLookahead */
598 int mxLookahead; /* Maximum aLookahead[].lookahead */
599 int nLookahead; /* Used slots in aLookahead[] */
600 int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
601 int nterminal; /* Number of terminal symbols */
602 int nsymbol; /* total number of symbols */
605 /* Return the number of entries in the yy_action table */
606 #define acttab_lookahead_size(X) ((X)->nAction)
608 /* The value for the N-th entry in yy_action */
609 #define acttab_yyaction(X,N) ((X)->aAction[N].action)
611 /* The value for the N-th entry in yy_lookahead */
612 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
614 /* Free all memory associated with the given acttab */
615 void acttab_free(acttab *p){
616 free( p->aAction );
617 free( p->aLookahead );
618 free( p );
621 /* Allocate a new acttab structure */
622 acttab *acttab_alloc(int nsymbol, int nterminal){
623 acttab *p = (acttab *) calloc( 1, sizeof(*p) );
624 if( p==0 ){
625 fprintf(stderr,"Unable to allocate memory for a new acttab.");
626 exit(1);
628 memset(p, 0, sizeof(*p));
629 p->nsymbol = nsymbol;
630 p->nterminal = nterminal;
631 return p;
634 /* Add a new action to the current transaction set.
636 ** This routine is called once for each lookahead for a particular
637 ** state.
639 void acttab_action(acttab *p, int lookahead, int action){
640 if( p->nLookahead>=p->nLookaheadAlloc ){
641 p->nLookaheadAlloc += 25;
642 p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead,
643 sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
644 if( p->aLookahead==0 ){
645 fprintf(stderr,"malloc failed\n");
646 exit(1);
649 if( p->nLookahead==0 ){
650 p->mxLookahead = lookahead;
651 p->mnLookahead = lookahead;
652 p->mnAction = action;
653 }else{
654 if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
655 if( p->mnLookahead>lookahead ){
656 p->mnLookahead = lookahead;
657 p->mnAction = action;
660 p->aLookahead[p->nLookahead].lookahead = lookahead;
661 p->aLookahead[p->nLookahead].action = action;
662 p->nLookahead++;
666 ** Add the transaction set built up with prior calls to acttab_action()
667 ** into the current action table. Then reset the transaction set back
668 ** to an empty set in preparation for a new round of acttab_action() calls.
670 ** Return the offset into the action table of the new transaction.
672 ** If the makeItSafe parameter is true, then the offset is chosen so that
673 ** it is impossible to overread the yy_lookaside[] table regardless of
674 ** the lookaside token. This is done for the terminal symbols, as they
675 ** come from external inputs and can contain syntax errors. When makeItSafe
676 ** is false, there is more flexibility in selecting offsets, resulting in
677 ** a smaller table. For non-terminal symbols, which are never syntax errors,
678 ** makeItSafe can be false.
680 int acttab_insert(acttab *p, int makeItSafe){
681 int i, j, k, n, end;
682 assert( p->nLookahead>0 );
684 /* Make sure we have enough space to hold the expanded action table
685 ** in the worst case. The worst case occurs if the transaction set
686 ** must be appended to the current action table
688 n = p->nsymbol + 1;
689 if( p->nAction + n >= p->nActionAlloc ){
690 int oldAlloc = p->nActionAlloc;
691 p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
692 p->aAction = (struct lookahead_action *) realloc( p->aAction,
693 sizeof(p->aAction[0])*p->nActionAlloc);
694 if( p->aAction==0 ){
695 fprintf(stderr,"malloc failed\n");
696 exit(1);
698 for(i=oldAlloc; i<p->nActionAlloc; i++){
699 p->aAction[i].lookahead = -1;
700 p->aAction[i].action = -1;
704 /* Scan the existing action table looking for an offset that is a
705 ** duplicate of the current transaction set. Fall out of the loop
706 ** if and when the duplicate is found.
708 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
710 end = makeItSafe ? p->mnLookahead : 0;
711 for(i=p->nAction-1; i>=end; i--){
712 if( p->aAction[i].lookahead==p->mnLookahead ){
713 /* All lookaheads and actions in the aLookahead[] transaction
714 ** must match against the candidate aAction[i] entry. */
715 if( p->aAction[i].action!=p->mnAction ) continue;
716 for(j=0; j<p->nLookahead; j++){
717 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
718 if( k<0 || k>=p->nAction ) break;
719 if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
720 if( p->aLookahead[j].action!=p->aAction[k].action ) break;
722 if( j<p->nLookahead ) continue;
724 /* No possible lookahead value that is not in the aLookahead[]
725 ** transaction is allowed to match aAction[i] */
726 n = 0;
727 for(j=0; j<p->nAction; j++){
728 if( p->aAction[j].lookahead<0 ) continue;
729 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
731 if( n==p->nLookahead ){
732 break; /* An exact match is found at offset i */
737 /* If no existing offsets exactly match the current transaction, find an
738 ** an empty offset in the aAction[] table in which we can add the
739 ** aLookahead[] transaction.
741 if( i<end ){
742 /* Look for holes in the aAction[] table that fit the current
743 ** aLookahead[] transaction. Leave i set to the offset of the hole.
744 ** If no holes are found, i is left at p->nAction, which means the
745 ** transaction will be appended. */
746 i = makeItSafe ? p->mnLookahead : 0;
747 for(; i<p->nActionAlloc - p->mxLookahead; i++){
748 if( p->aAction[i].lookahead<0 ){
749 for(j=0; j<p->nLookahead; j++){
750 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
751 if( k<0 ) break;
752 if( p->aAction[k].lookahead>=0 ) break;
754 if( j<p->nLookahead ) continue;
755 for(j=0; j<p->nAction; j++){
756 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
758 if( j==p->nAction ){
759 break; /* Fits in empty slots */
764 /* Insert transaction set at index i. */
765 #if 0
766 printf("Acttab:");
767 for(j=0; j<p->nLookahead; j++){
768 printf(" %d", p->aLookahead[j].lookahead);
770 printf(" inserted at %d\n", i);
771 #endif
772 for(j=0; j<p->nLookahead; j++){
773 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
774 p->aAction[k] = p->aLookahead[j];
775 if( k>=p->nAction ) p->nAction = k+1;
777 if( makeItSafe && i+p->nterminal>=p->nAction ) p->nAction = i+p->nterminal+1;
778 p->nLookahead = 0;
780 /* Return the offset that is added to the lookahead in order to get the
781 ** index into yy_action of the action */
782 return i - p->mnLookahead;
786 ** Return the size of the action table without the trailing syntax error
787 ** entries.
789 int acttab_action_size(acttab *p){
790 int n = p->nAction;
791 while( n>0 && p->aAction[n-1].lookahead<0 ){ n--; }
792 return n;
795 /********************** From the file "build.c" *****************************/
797 ** Routines to construction the finite state machine for the LEMON
798 ** parser generator.
801 /* Find a precedence symbol of every rule in the grammar.
803 ** Those rules which have a precedence symbol coded in the input
804 ** grammar using the "[symbol]" construct will already have the
805 ** rp->precsym field filled. Other rules take as their precedence
806 ** symbol the first RHS symbol with a defined precedence. If there
807 ** are not RHS symbols with a defined precedence, the precedence
808 ** symbol field is left blank.
810 void FindRulePrecedences(struct lemon *xp)
812 struct rule *rp;
813 for(rp=xp->rule; rp; rp=rp->next){
814 if( rp->precsym==0 ){
815 int i, j;
816 for(i=0; i<rp->nrhs && rp->precsym==0; i++){
817 struct symbol *sp = rp->rhs[i];
818 if( sp->type==MULTITERMINAL ){
819 for(j=0; j<sp->nsubsym; j++){
820 if( sp->subsym[j]->prec>=0 ){
821 rp->precsym = sp->subsym[j];
822 break;
825 }else if( sp->prec>=0 ){
826 rp->precsym = rp->rhs[i];
831 return;
834 /* Find all nonterminals which will generate the empty string.
835 ** Then go back and compute the first sets of every nonterminal.
836 ** The first set is the set of all terminal symbols which can begin
837 ** a string generated by that nonterminal.
839 void FindFirstSets(struct lemon *lemp)
841 int i, j;
842 struct rule *rp;
843 int progress;
845 for(i=0; i<lemp->nsymbol; i++){
846 lemp->symbols[i]->lambda = LEMON_FALSE;
848 for(i=lemp->nterminal; i<lemp->nsymbol; i++){
849 lemp->symbols[i]->firstset = SetNew();
852 /* First compute all lambdas */
854 progress = 0;
855 for(rp=lemp->rule; rp; rp=rp->next){
856 if( rp->lhs->lambda ) continue;
857 for(i=0; i<rp->nrhs; i++){
858 struct symbol *sp = rp->rhs[i];
859 assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE );
860 if( sp->lambda==LEMON_FALSE ) break;
862 if( i==rp->nrhs ){
863 rp->lhs->lambda = LEMON_TRUE;
864 progress = 1;
867 }while( progress );
869 /* Now compute all first sets */
871 struct symbol *s1, *s2;
872 progress = 0;
873 for(rp=lemp->rule; rp; rp=rp->next){
874 s1 = rp->lhs;
875 for(i=0; i<rp->nrhs; i++){
876 s2 = rp->rhs[i];
877 if( s2->type==TERMINAL ){
878 progress += SetAdd(s1->firstset,s2->index);
879 break;
880 }else if( s2->type==MULTITERMINAL ){
881 for(j=0; j<s2->nsubsym; j++){
882 progress += SetAdd(s1->firstset,s2->subsym[j]->index);
884 break;
885 }else if( s1==s2 ){
886 if( s1->lambda==LEMON_FALSE ) break;
887 }else{
888 progress += SetUnion(s1->firstset,s2->firstset);
889 if( s2->lambda==LEMON_FALSE ) break;
893 }while( progress );
894 return;
897 /* Compute all LR(0) states for the grammar. Links
898 ** are added to between some states so that the LR(1) follow sets
899 ** can be computed later.
901 PRIVATE struct state *getstate(struct lemon *); /* forward reference */
902 void FindStates(struct lemon *lemp)
904 struct symbol *sp;
905 struct rule *rp;
907 Configlist_init();
909 /* Find the start symbol */
910 if( lemp->start ){
911 sp = Symbol_find(lemp->start);
912 if( sp==0 ){
913 ErrorMsg(lemp->filename,0,
914 "The specified start symbol \"%s\" is not "
915 "in a nonterminal of the grammar. \"%s\" will be used as the start "
916 "symbol instead.",lemp->start,lemp->startRule->lhs->name);
917 lemp->errorcnt++;
918 sp = lemp->startRule->lhs;
920 }else{
921 sp = lemp->startRule->lhs;
924 /* Make sure the start symbol doesn't occur on the right-hand side of
925 ** any rule. Report an error if it does. (YACC would generate a new
926 ** start symbol in this case.) */
927 for(rp=lemp->rule; rp; rp=rp->next){
928 int i;
929 for(i=0; i<rp->nrhs; i++){
930 if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
931 ErrorMsg(lemp->filename,0,
932 "The start symbol \"%s\" occurs on the "
933 "right-hand side of a rule. This will result in a parser which "
934 "does not work properly.",sp->name);
935 lemp->errorcnt++;
940 /* The basis configuration set for the first state
941 ** is all rules which have the start symbol as their
942 ** left-hand side */
943 for(rp=sp->rule; rp; rp=rp->nextlhs){
944 struct config *newcfp;
945 rp->lhsStart = 1;
946 newcfp = Configlist_addbasis(rp,0);
947 SetAdd(newcfp->fws,0);
950 /* Compute the first state. All other states will be
951 ** computed automatically during the computation of the first one.
952 ** The returned pointer to the first state is not used. */
953 (void)getstate(lemp);
954 return;
957 /* Return a pointer to a state which is described by the configuration
958 ** list which has been built from calls to Configlist_add.
960 PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */
961 PRIVATE struct state *getstate(struct lemon *lemp)
963 struct config *cfp, *bp;
964 struct state *stp;
966 /* Extract the sorted basis of the new state. The basis was constructed
967 ** by prior calls to "Configlist_addbasis()". */
968 Configlist_sortbasis();
969 bp = Configlist_basis();
971 /* Get a state with the same basis */
972 stp = State_find(bp);
973 if( stp ){
974 /* A state with the same basis already exists! Copy all the follow-set
975 ** propagation links from the state under construction into the
976 ** preexisting state, then return a pointer to the preexisting state */
977 struct config *x, *y;
978 for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
979 Plink_copy(&y->bplp,x->bplp);
980 Plink_delete(x->fplp);
981 x->fplp = x->bplp = 0;
983 cfp = Configlist_return();
984 Configlist_eat(cfp);
985 }else{
986 /* This really is a new state. Construct all the details */
987 Configlist_closure(lemp); /* Compute the configuration closure */
988 Configlist_sort(); /* Sort the configuration closure */
989 cfp = Configlist_return(); /* Get a pointer to the config list */
990 stp = State_new(); /* A new state structure */
991 MemoryCheck(stp);
992 stp->bp = bp; /* Remember the configuration basis */
993 stp->cfp = cfp; /* Remember the configuration closure */
994 stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
995 stp->ap = 0; /* No actions, yet. */
996 State_insert(stp,stp->bp); /* Add to the state table */
997 buildshifts(lemp,stp); /* Recursively compute successor states */
999 return stp;
1003 ** Return true if two symbols are the same.
1005 int same_symbol(struct symbol *a, struct symbol *b)
1007 int i;
1008 if( a==b ) return 1;
1009 if( a->type!=MULTITERMINAL ) return 0;
1010 if( b->type!=MULTITERMINAL ) return 0;
1011 if( a->nsubsym!=b->nsubsym ) return 0;
1012 for(i=0; i<a->nsubsym; i++){
1013 if( a->subsym[i]!=b->subsym[i] ) return 0;
1015 return 1;
1018 /* Construct all successor states to the given state. A "successor"
1019 ** state is any state which can be reached by a shift action.
1021 PRIVATE void buildshifts(struct lemon *lemp, struct state *stp)
1023 struct config *cfp; /* For looping thru the config closure of "stp" */
1024 struct config *bcfp; /* For the inner loop on config closure of "stp" */
1025 struct config *newcfg; /* */
1026 struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
1027 struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
1028 struct state *newstp; /* A pointer to a successor state */
1030 /* Each configuration becomes complete after it contributes to a successor
1031 ** state. Initially, all configurations are incomplete */
1032 for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
1034 /* Loop through all configurations of the state "stp" */
1035 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1036 if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
1037 if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
1038 Configlist_reset(); /* Reset the new config set */
1039 sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
1041 /* For every configuration in the state "stp" which has the symbol "sp"
1042 ** following its dot, add the same configuration to the basis set under
1043 ** construction but with the dot shifted one symbol to the right. */
1044 for(bcfp=cfp; bcfp; bcfp=bcfp->next){
1045 if( bcfp->status==COMPLETE ) continue; /* Already used */
1046 if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
1047 bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
1048 if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
1049 bcfp->status = COMPLETE; /* Mark this config as used */
1050 newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
1051 Plink_add(&newcfg->bplp,bcfp);
1054 /* Get a pointer to the state described by the basis configuration set
1055 ** constructed in the preceding loop */
1056 newstp = getstate(lemp);
1058 /* The state "newstp" is reached from the state "stp" by a shift action
1059 ** on the symbol "sp" */
1060 if( sp->type==MULTITERMINAL ){
1061 int i;
1062 for(i=0; i<sp->nsubsym; i++){
1063 Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
1065 }else{
1066 Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
1072 ** Construct the propagation links
1074 void FindLinks(struct lemon *lemp)
1076 int i;
1077 struct config *cfp, *other;
1078 struct state *stp;
1079 struct plink *plp;
1081 /* Housekeeping detail:
1082 ** Add to every propagate link a pointer back to the state to
1083 ** which the link is attached. */
1084 for(i=0; i<lemp->nstate; i++){
1085 stp = lemp->sorted[i];
1086 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1087 cfp->stp = stp;
1091 /* Convert all backlinks into forward links. Only the forward
1092 ** links are used in the follow-set computation. */
1093 for(i=0; i<lemp->nstate; i++){
1094 stp = lemp->sorted[i];
1095 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1096 for(plp=cfp->bplp; plp; plp=plp->next){
1097 other = plp->cfp;
1098 Plink_add(&other->fplp,cfp);
1104 /* Compute all followsets.
1106 ** A followset is the set of all symbols which can come immediately
1107 ** after a configuration.
1109 void FindFollowSets(struct lemon *lemp)
1111 int i;
1112 struct config *cfp;
1113 struct plink *plp;
1114 int progress;
1115 int change;
1117 for(i=0; i<lemp->nstate; i++){
1118 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1119 cfp->status = INCOMPLETE;
1124 progress = 0;
1125 for(i=0; i<lemp->nstate; i++){
1126 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1127 if( cfp->status==COMPLETE ) continue;
1128 for(plp=cfp->fplp; plp; plp=plp->next){
1129 change = SetUnion(plp->cfp->fws,cfp->fws);
1130 if( change ){
1131 plp->cfp->status = INCOMPLETE;
1132 progress = 1;
1135 cfp->status = COMPLETE;
1138 }while( progress );
1141 static int resolve_conflict(struct action *,struct action *);
1143 /* Compute the reduce actions, and resolve conflicts.
1145 void FindActions(struct lemon *lemp)
1147 int i,j;
1148 struct config *cfp;
1149 struct state *stp;
1150 struct symbol *sp;
1151 struct rule *rp;
1153 /* Add all of the reduce actions
1154 ** A reduce action is added for each element of the followset of
1155 ** a configuration which has its dot at the extreme right.
1157 for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
1158 stp = lemp->sorted[i];
1159 for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
1160 if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
1161 for(j=0; j<lemp->nterminal; j++){
1162 if( SetFind(cfp->fws,j) ){
1163 /* Add a reduce action to the state "stp" which will reduce by the
1164 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
1165 Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
1172 /* Add the accepting token */
1173 if( lemp->start ){
1174 sp = Symbol_find(lemp->start);
1175 if( sp==0 ) sp = lemp->startRule->lhs;
1176 }else{
1177 sp = lemp->startRule->lhs;
1179 /* Add to the first state (which is always the starting state of the
1180 ** finite state machine) an action to ACCEPT if the lookahead is the
1181 ** start nonterminal. */
1182 Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
1184 /* Resolve conflicts */
1185 for(i=0; i<lemp->nstate; i++){
1186 struct action *ap, *nap;
1187 stp = lemp->sorted[i];
1188 /* assert( stp->ap ); */
1189 stp->ap = Action_sort(stp->ap);
1190 for(ap=stp->ap; ap && ap->next; ap=ap->next){
1191 for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
1192 /* The two actions "ap" and "nap" have the same lookahead.
1193 ** Figure out which one should be used */
1194 lemp->nconflict += resolve_conflict(ap,nap);
1199 /* Report an error for each rule that can never be reduced. */
1200 for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
1201 for(i=0; i<lemp->nstate; i++){
1202 struct action *ap;
1203 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
1204 if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
1207 for(rp=lemp->rule; rp; rp=rp->next){
1208 if( rp->canReduce ) continue;
1209 ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
1210 lemp->errorcnt++;
1214 /* Resolve a conflict between the two given actions. If the
1215 ** conflict can't be resolved, return non-zero.
1217 ** NO LONGER TRUE:
1218 ** To resolve a conflict, first look to see if either action
1219 ** is on an error rule. In that case, take the action which
1220 ** is not associated with the error rule. If neither or both
1221 ** actions are associated with an error rule, then try to
1222 ** use precedence to resolve the conflict.
1224 ** If either action is a SHIFT, then it must be apx. This
1225 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1227 static int resolve_conflict(
1228 struct action *apx,
1229 struct action *apy
1231 struct symbol *spx, *spy;
1232 int errcnt = 0;
1233 assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
1234 if( apx->type==SHIFT && apy->type==SHIFT ){
1235 apy->type = SSCONFLICT;
1236 errcnt++;
1238 if( apx->type==SHIFT && apy->type==REDUCE ){
1239 spx = apx->sp;
1240 spy = apy->x.rp->precsym;
1241 if( spy==0 || spx->prec<0 || spy->prec<0 ){
1242 /* Not enough precedence information. */
1243 apy->type = SRCONFLICT;
1244 errcnt++;
1245 }else if( spx->prec>spy->prec ){ /* higher precedence wins */
1246 apy->type = RD_RESOLVED;
1247 }else if( spx->prec<spy->prec ){
1248 apx->type = SH_RESOLVED;
1249 }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1250 apy->type = RD_RESOLVED; /* associativity */
1251 }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
1252 apx->type = SH_RESOLVED;
1253 }else{
1254 assert( spx->prec==spy->prec && spx->assoc==NONE );
1255 apx->type = ERROR;
1257 }else if( apx->type==REDUCE && apy->type==REDUCE ){
1258 spx = apx->x.rp->precsym;
1259 spy = apy->x.rp->precsym;
1260 if( spx==0 || spy==0 || spx->prec<0 ||
1261 spy->prec<0 || spx->prec==spy->prec ){
1262 apy->type = RRCONFLICT;
1263 errcnt++;
1264 }else if( spx->prec>spy->prec ){
1265 apy->type = RD_RESOLVED;
1266 }else if( spx->prec<spy->prec ){
1267 apx->type = RD_RESOLVED;
1269 }else{
1270 assert(
1271 apx->type==SH_RESOLVED ||
1272 apx->type==RD_RESOLVED ||
1273 apx->type==SSCONFLICT ||
1274 apx->type==SRCONFLICT ||
1275 apx->type==RRCONFLICT ||
1276 apy->type==SH_RESOLVED ||
1277 apy->type==RD_RESOLVED ||
1278 apy->type==SSCONFLICT ||
1279 apy->type==SRCONFLICT ||
1280 apy->type==RRCONFLICT
1282 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1283 ** REDUCEs on the list. If we reach this point it must be because
1284 ** the parser conflict had already been resolved. */
1286 return errcnt;
1288 /********************* From the file "configlist.c" *************************/
1290 ** Routines to processing a configuration list and building a state
1291 ** in the LEMON parser generator.
1294 static struct config *freelist = 0; /* List of free configurations */
1295 static struct config *current = 0; /* Top of list of configurations */
1296 static struct config **currentend = 0; /* Last on list of configs */
1297 static struct config *basis = 0; /* Top of list of basis configs */
1298 static struct config **basisend = 0; /* End of list of basis configs */
1300 /* Return a pointer to a new configuration */
1301 PRIVATE struct config *newconfig(void){
1302 struct config *newcfg;
1303 if( freelist==0 ){
1304 int i;
1305 int amt = 3;
1306 freelist = (struct config *)calloc( amt, sizeof(struct config) );
1307 if( freelist==0 ){
1308 fprintf(stderr,"Unable to allocate memory for a new configuration.");
1309 exit(1);
1311 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1312 freelist[amt-1].next = 0;
1314 newcfg = freelist;
1315 freelist = freelist->next;
1316 return newcfg;
1319 /* The configuration "old" is no longer used */
1320 PRIVATE void deleteconfig(struct config *old)
1322 old->next = freelist;
1323 freelist = old;
1326 /* Initialized the configuration list builder */
1327 void Configlist_init(void){
1328 current = 0;
1329 currentend = &current;
1330 basis = 0;
1331 basisend = &basis;
1332 Configtable_init();
1333 return;
1336 /* Initialized the configuration list builder */
1337 void Configlist_reset(void){
1338 current = 0;
1339 currentend = &current;
1340 basis = 0;
1341 basisend = &basis;
1342 Configtable_clear(0);
1343 return;
1346 /* Add another configuration to the configuration list */
1347 struct config *Configlist_add(
1348 struct rule *rp, /* The rule */
1349 int dot /* Index into the RHS of the rule where the dot goes */
1351 struct config *cfp, model;
1353 assert( currentend!=0 );
1354 model.rp = rp;
1355 model.dot = dot;
1356 cfp = Configtable_find(&model);
1357 if( cfp==0 ){
1358 cfp = newconfig();
1359 cfp->rp = rp;
1360 cfp->dot = dot;
1361 cfp->fws = SetNew();
1362 cfp->stp = 0;
1363 cfp->fplp = cfp->bplp = 0;
1364 cfp->next = 0;
1365 cfp->bp = 0;
1366 *currentend = cfp;
1367 currentend = &cfp->next;
1368 Configtable_insert(cfp);
1370 return cfp;
1373 /* Add a basis configuration to the configuration list */
1374 struct config *Configlist_addbasis(struct rule *rp, int dot)
1376 struct config *cfp, model;
1378 assert( basisend!=0 );
1379 assert( currentend!=0 );
1380 model.rp = rp;
1381 model.dot = dot;
1382 cfp = Configtable_find(&model);
1383 if( cfp==0 ){
1384 cfp = newconfig();
1385 cfp->rp = rp;
1386 cfp->dot = dot;
1387 cfp->fws = SetNew();
1388 cfp->stp = 0;
1389 cfp->fplp = cfp->bplp = 0;
1390 cfp->next = 0;
1391 cfp->bp = 0;
1392 *currentend = cfp;
1393 currentend = &cfp->next;
1394 *basisend = cfp;
1395 basisend = &cfp->bp;
1396 Configtable_insert(cfp);
1398 return cfp;
1401 /* Compute the closure of the configuration list */
1402 void Configlist_closure(struct lemon *lemp)
1404 struct config *cfp, *newcfp;
1405 struct rule *rp, *newrp;
1406 struct symbol *sp, *xsp;
1407 int i, dot;
1409 assert( currentend!=0 );
1410 for(cfp=current; cfp; cfp=cfp->next){
1411 rp = cfp->rp;
1412 dot = cfp->dot;
1413 if( dot>=rp->nrhs ) continue;
1414 sp = rp->rhs[dot];
1415 if( sp->type==NONTERMINAL ){
1416 if( sp->rule==0 && sp!=lemp->errsym ){
1417 ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1418 sp->name);
1419 lemp->errorcnt++;
1421 for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1422 newcfp = Configlist_add(newrp,0);
1423 for(i=dot+1; i<rp->nrhs; i++){
1424 xsp = rp->rhs[i];
1425 if( xsp->type==TERMINAL ){
1426 SetAdd(newcfp->fws,xsp->index);
1427 break;
1428 }else if( xsp->type==MULTITERMINAL ){
1429 int k;
1430 for(k=0; k<xsp->nsubsym; k++){
1431 SetAdd(newcfp->fws, xsp->subsym[k]->index);
1433 break;
1434 }else{
1435 SetUnion(newcfp->fws,xsp->firstset);
1436 if( xsp->lambda==LEMON_FALSE ) break;
1439 if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1443 return;
1446 /* Sort the configuration list */
1447 void Configlist_sort(void){
1448 current = (struct config*)msort((char*)current,(char**)&(current->next),
1449 Configcmp);
1450 currentend = 0;
1451 return;
1454 /* Sort the basis configuration list */
1455 void Configlist_sortbasis(void){
1456 basis = (struct config*)msort((char*)current,(char**)&(current->bp),
1457 Configcmp);
1458 basisend = 0;
1459 return;
1462 /* Return a pointer to the head of the configuration list and
1463 ** reset the list */
1464 struct config *Configlist_return(void){
1465 struct config *old;
1466 old = current;
1467 current = 0;
1468 currentend = 0;
1469 return old;
1472 /* Return a pointer to the head of the configuration list and
1473 ** reset the list */
1474 struct config *Configlist_basis(void){
1475 struct config *old;
1476 old = basis;
1477 basis = 0;
1478 basisend = 0;
1479 return old;
1482 /* Free all elements of the given configuration list */
1483 void Configlist_eat(struct config *cfp)
1485 struct config *nextcfp;
1486 for(; cfp; cfp=nextcfp){
1487 nextcfp = cfp->next;
1488 assert( cfp->fplp==0 );
1489 assert( cfp->bplp==0 );
1490 if( cfp->fws ) SetFree(cfp->fws);
1491 deleteconfig(cfp);
1493 return;
1495 /***************** From the file "error.c" *********************************/
1497 ** Code for printing error message.
1500 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1501 va_list ap;
1502 fprintf(stderr, "%s:%d: ", filename, lineno);
1503 va_start(ap, format);
1504 vfprintf(stderr,format,ap);
1505 va_end(ap);
1506 fprintf(stderr, "\n");
1508 /**************** From the file "main.c" ************************************/
1510 ** Main program file for the LEMON parser generator.
1513 /* Report an out-of-memory condition and abort. This function
1514 ** is used mostly by the "MemoryCheck" macro in struct.h
1516 void memory_error(void){
1517 fprintf(stderr,"Out of memory. Aborting...\n");
1518 exit(1);
1521 static int nDefine = 0; /* Number of -D options on the command line */
1522 static char **azDefine = 0; /* Name of the -D macros */
1524 /* This routine is called with the argument to each -D command-line option.
1525 ** Add the macro defined to the azDefine array.
1527 static void handle_D_option(char *z){
1528 char **paz;
1529 nDefine++;
1530 azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine);
1531 if( azDefine==0 ){
1532 fprintf(stderr,"out of memory\n");
1533 exit(1);
1535 paz = &azDefine[nDefine-1];
1536 *paz = (char *) malloc( lemonStrlen(z)+1 );
1537 if( *paz==0 ){
1538 fprintf(stderr,"out of memory\n");
1539 exit(1);
1541 lemon_strcpy(*paz, z);
1542 for(z=*paz; *z && *z!='='; z++){}
1543 *z = 0;
1546 /* Rember the name of the output directory
1548 static char *outputDir = NULL;
1549 static void handle_d_option(char *z){
1550 outputDir = (char *) malloc( lemonStrlen(z)+1 );
1551 if( outputDir==0 ){
1552 fprintf(stderr,"out of memory\n");
1553 exit(1);
1555 lemon_strcpy(outputDir, z);
1558 static char *user_templatename = NULL;
1559 static void handle_T_option(char *z){
1560 user_templatename = (char *) malloc( lemonStrlen(z)+1 );
1561 if( user_templatename==0 ){
1562 memory_error();
1564 lemon_strcpy(user_templatename, z);
1567 /* Merge together to lists of rules ordered by rule.iRule */
1568 static struct rule *Rule_merge(struct rule *pA, struct rule *pB){
1569 struct rule *pFirst = 0;
1570 struct rule **ppPrev = &pFirst;
1571 while( pA && pB ){
1572 if( pA->iRule<pB->iRule ){
1573 *ppPrev = pA;
1574 ppPrev = &pA->next;
1575 pA = pA->next;
1576 }else{
1577 *ppPrev = pB;
1578 ppPrev = &pB->next;
1579 pB = pB->next;
1582 if( pA ){
1583 *ppPrev = pA;
1584 }else{
1585 *ppPrev = pB;
1587 return pFirst;
1591 ** Sort a list of rules in order of increasing iRule value
1593 static struct rule *Rule_sort(struct rule *rp){
1594 unsigned int i;
1595 struct rule *pNext;
1596 struct rule *x[32];
1597 memset(x, 0, sizeof(x));
1598 while( rp ){
1599 pNext = rp->next;
1600 rp->next = 0;
1601 for(i=0; i<sizeof(x)/sizeof(x[0])-1 && x[i]; i++){
1602 rp = Rule_merge(x[i], rp);
1603 x[i] = 0;
1605 x[i] = rp;
1606 rp = pNext;
1608 rp = 0;
1609 for(i=0; i<sizeof(x)/sizeof(x[0]); i++){
1610 rp = Rule_merge(x[i], rp);
1612 return rp;
1615 /* forward reference */
1616 static const char *minimum_size_type(int lwr, int upr, int *pnByte);
1618 /* Print a single line of the "Parser Stats" output
1620 static void stats_line(const char *zLabel, int iValue){
1621 int nLabel = lemonStrlen(zLabel);
1622 printf(" %s%.*s %5d\n", zLabel,
1623 35-nLabel, "................................",
1624 iValue);
1627 /* The main program. Parse the command line and do it... */
1628 int main(int argc, char **argv){
1629 static int version = 0;
1630 static int rpflag = 0;
1631 static int basisflag = 0;
1632 static int compress = 0;
1633 static int quiet = 0;
1634 static int statistics = 0;
1635 static int mhflag = 0;
1636 static int nolinenosflag = 0;
1637 static int noResort = 0;
1638 static int sqlFlag = 0;
1639 static int printPP = 0;
1641 static struct s_options options[] = {
1642 {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1643 {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1644 {OPT_FSTR, "d", (char*)&handle_d_option, "Output directory. Default '.'"},
1645 {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
1646 {OPT_FLAG, "E", (char*)&printPP, "Print input file after preprocessing."},
1647 {OPT_FSTR, "f", 0, "Ignored. (Placeholder for -f compiler options.)"},
1648 {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1649 {OPT_FSTR, "I", 0, "Ignored. (Placeholder for '-I' compiler options.)"},
1650 {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."},
1651 {OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."},
1652 {OPT_FSTR, "O", 0, "Ignored. (Placeholder for '-O' compiler options.)"},
1653 {OPT_FLAG, "p", (char*)&showPrecedenceConflict,
1654 "Show conflicts resolved by precedence rules"},
1655 {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1656 {OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"},
1657 {OPT_FLAG, "s", (char*)&statistics,
1658 "Print parser stats to standard output."},
1659 {OPT_FLAG, "S", (char*)&sqlFlag,
1660 "Generate the *.sql file describing the parser tables."},
1661 {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1662 {OPT_FSTR, "T", (char*)handle_T_option, "Specify a template file."},
1663 {OPT_FSTR, "W", 0, "Ignored. (Placeholder for '-W' compiler options.)"},
1664 {OPT_FLAG,0,0,0}
1666 int i;
1667 int exitcode;
1668 struct lemon lem;
1669 struct rule *rp;
1671 (void)argc;
1672 OptInit(argv,options,stderr);
1673 if( version ){
1674 printf("Lemon version 1.0\n");
1675 exit(0);
1677 if( OptNArgs()!=1 ){
1678 fprintf(stderr,"Exactly one filename argument is required.\n");
1679 exit(1);
1681 memset(&lem, 0, sizeof(lem));
1682 lem.errorcnt = 0;
1684 /* Initialize the machine */
1685 Strsafe_init();
1686 Symbol_init();
1687 State_init();
1688 lem.argv0 = argv[0];
1689 lem.filename = OptArg(0);
1690 lem.basisflag = basisflag;
1691 lem.nolinenosflag = nolinenosflag;
1692 lem.printPreprocessed = printPP;
1693 Symbol_new("$");
1695 /* Parse the input file */
1696 Parse(&lem);
1697 if( lem.printPreprocessed || lem.errorcnt ) exit(lem.errorcnt);
1698 if( lem.nrule==0 ){
1699 fprintf(stderr,"Empty grammar.\n");
1700 exit(1);
1702 lem.errsym = Symbol_find("error");
1704 /* Count and index the symbols of the grammar */
1705 Symbol_new("{default}");
1706 lem.nsymbol = Symbol_count();
1707 lem.symbols = Symbol_arrayof();
1708 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1709 qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp);
1710 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1711 while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; }
1712 assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 );
1713 lem.nsymbol = i - 1;
1714 for(i=1; ISUPPER(lem.symbols[i]->name[0]); i++);
1715 lem.nterminal = i;
1717 /* Assign sequential rule numbers. Start with 0. Put rules that have no
1718 ** reduce action C-code associated with them last, so that the switch()
1719 ** statement that selects reduction actions will have a smaller jump table.
1721 for(i=0, rp=lem.rule; rp; rp=rp->next){
1722 rp->iRule = rp->code ? i++ : -1;
1724 lem.nruleWithAction = i;
1725 for(rp=lem.rule; rp; rp=rp->next){
1726 if( rp->iRule<0 ) rp->iRule = i++;
1728 lem.startRule = lem.rule;
1729 lem.rule = Rule_sort(lem.rule);
1731 /* Generate a reprint of the grammar, if requested on the command line */
1732 if( rpflag ){
1733 Reprint(&lem);
1734 }else{
1735 /* Initialize the size for all follow and first sets */
1736 SetSize(lem.nterminal+1);
1738 /* Find the precedence for every production rule (that has one) */
1739 FindRulePrecedences(&lem);
1741 /* Compute the lambda-nonterminals and the first-sets for every
1742 ** nonterminal */
1743 FindFirstSets(&lem);
1745 /* Compute all LR(0) states. Also record follow-set propagation
1746 ** links so that the follow-set can be computed later */
1747 lem.nstate = 0;
1748 FindStates(&lem);
1749 lem.sorted = State_arrayof();
1751 /* Tie up loose ends on the propagation links */
1752 FindLinks(&lem);
1754 /* Compute the follow set of every reducible configuration */
1755 FindFollowSets(&lem);
1757 /* Compute the action tables */
1758 FindActions(&lem);
1760 /* Compress the action tables */
1761 if( compress==0 ) CompressTables(&lem);
1763 /* Reorder and renumber the states so that states with fewer choices
1764 ** occur at the end. This is an optimization that helps make the
1765 ** generated parser tables smaller. */
1766 if( noResort==0 ) ResortStates(&lem);
1768 /* Generate a report of the parser generated. (the "y.output" file) */
1769 if( !quiet ) ReportOutput(&lem);
1771 /* Generate the source code for the parser */
1772 ReportTable(&lem, mhflag, sqlFlag);
1774 /* Produce a header file for use by the scanner. (This step is
1775 ** omitted if the "-m" option is used because makeheaders will
1776 ** generate the file for us.) */
1777 if( !mhflag ) ReportHeader(&lem);
1779 if( statistics ){
1780 printf("Parser statistics:\n");
1781 stats_line("terminal symbols", lem.nterminal);
1782 stats_line("non-terminal symbols", lem.nsymbol - lem.nterminal);
1783 stats_line("total symbols", lem.nsymbol);
1784 stats_line("rules", lem.nrule);
1785 stats_line("states", lem.nxstate);
1786 stats_line("conflicts", lem.nconflict);
1787 stats_line("action table entries", lem.nactiontab);
1788 stats_line("lookahead table entries", lem.nlookaheadtab);
1789 stats_line("total table size (bytes)", lem.tablesize);
1791 if( lem.nconflict > 0 ){
1792 fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1795 /* return 0 on success, 1 on failure. */
1796 exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0;
1797 exit(exitcode);
1798 return (exitcode);
1800 /******************** From the file "msort.c" *******************************/
1802 ** A generic merge-sort program.
1804 ** USAGE:
1805 ** Let "ptr" be a pointer to some structure which is at the head of
1806 ** a null-terminated list. Then to sort the list call:
1808 ** ptr = msort(ptr,&(ptr->next),cmpfnc);
1810 ** In the above, "cmpfnc" is a pointer to a function which compares
1811 ** two instances of the structure and returns an integer, as in
1812 ** strcmp. The second argument is a pointer to the pointer to the
1813 ** second element of the linked list. This address is used to compute
1814 ** the offset to the "next" field within the structure. The offset to
1815 ** the "next" field must be constant for all structures in the list.
1817 ** The function returns a new pointer which is the head of the list
1818 ** after sorting.
1820 ** ALGORITHM:
1821 ** Merge-sort.
1825 ** Return a pointer to the next structure in the linked list.
1827 #define NEXT(A) (*(char**)(((char*)A)+offset))
1830 ** Inputs:
1831 ** a: A sorted, null-terminated linked list. (May be null).
1832 ** b: A sorted, null-terminated linked list. (May be null).
1833 ** cmp: A pointer to the comparison function.
1834 ** offset: Offset in the structure to the "next" field.
1836 ** Return Value:
1837 ** A pointer to the head of a sorted list containing the elements
1838 ** of both a and b.
1840 ** Side effects:
1841 ** The "next" pointers for elements in the lists a and b are
1842 ** changed.
1844 static char *merge(
1845 char *a,
1846 char *b,
1847 int (*cmp)(const char*,const char*),
1848 int offset
1850 char *ptr, *head;
1852 if( a==0 ){
1853 head = b;
1854 }else if( b==0 ){
1855 head = a;
1856 }else{
1857 if( (*cmp)(a,b)<=0 ){
1858 ptr = a;
1859 a = NEXT(a);
1860 }else{
1861 ptr = b;
1862 b = NEXT(b);
1864 head = ptr;
1865 while( a && b ){
1866 if( (*cmp)(a,b)<=0 ){
1867 NEXT(ptr) = a;
1868 ptr = a;
1869 a = NEXT(a);
1870 }else{
1871 NEXT(ptr) = b;
1872 ptr = b;
1873 b = NEXT(b);
1876 if( a ) NEXT(ptr) = a;
1877 else NEXT(ptr) = b;
1879 return head;
1883 ** Inputs:
1884 ** list: Pointer to a singly-linked list of structures.
1885 ** next: Pointer to pointer to the second element of the list.
1886 ** cmp: A comparison function.
1888 ** Return Value:
1889 ** A pointer to the head of a sorted list containing the elements
1890 ** originally in list.
1892 ** Side effects:
1893 ** The "next" pointers for elements in list are changed.
1895 #define LISTSIZE 30
1896 static char *msort(
1897 char *list,
1898 char **next,
1899 int (*cmp)(const char*,const char*)
1901 unsigned long offset;
1902 char *ep;
1903 char *set[LISTSIZE];
1904 int i;
1905 offset = (unsigned long)((char*)next - (char*)list);
1906 for(i=0; i<LISTSIZE; i++) set[i] = 0;
1907 while( list ){
1908 ep = list;
1909 list = NEXT(list);
1910 NEXT(ep) = 0;
1911 for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1912 ep = merge(ep,set[i],cmp,offset);
1913 set[i] = 0;
1915 set[i] = ep;
1917 ep = 0;
1918 for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset);
1919 return ep;
1921 /************************ From the file "option.c" **************************/
1922 static char **g_argv;
1923 static struct s_options *op;
1924 static FILE *errstream;
1926 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1929 ** Print the command line with a carrot pointing to the k-th character
1930 ** of the n-th field.
1932 static void errline(int n, int k, FILE *err)
1934 int spcnt, i;
1935 if( g_argv[0] ) fprintf(err,"%s",g_argv[0]);
1936 spcnt = lemonStrlen(g_argv[0]) + 1;
1937 for(i=1; i<n && g_argv[i]; i++){
1938 fprintf(err," %s",g_argv[i]);
1939 spcnt += lemonStrlen(g_argv[i])+1;
1941 spcnt += k;
1942 for(; g_argv[i]; i++) fprintf(err," %s",g_argv[i]);
1943 if( spcnt<20 ){
1944 fprintf(err,"\n%*s^-- here\n",spcnt,"");
1945 }else{
1946 fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1951 ** Return the index of the N-th non-switch argument. Return -1
1952 ** if N is out of range.
1954 static int argindex(int n)
1956 int i;
1957 int dashdash = 0;
1958 if( g_argv!=0 && *g_argv!=0 ){
1959 for(i=1; g_argv[i]; i++){
1960 if( dashdash || !ISOPT(g_argv[i]) ){
1961 if( n==0 ) return i;
1962 n--;
1964 if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
1967 return -1;
1970 static char emsg[] = "Command line syntax error: ";
1973 ** Process a flag command line argument.
1975 static int handleflags(int i, FILE *err)
1977 int v;
1978 int errcnt = 0;
1979 int j;
1980 for(j=0; op[j].label; j++){
1981 if( strncmp(&g_argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
1983 v = g_argv[i][0]=='-' ? 1 : 0;
1984 if( op[j].label==0 ){
1985 if( err ){
1986 fprintf(err,"%sundefined option.\n",emsg);
1987 errline(i,1,err);
1989 errcnt++;
1990 }else if( op[j].arg==0 ){
1991 /* Ignore this option */
1992 }else if( op[j].type==OPT_FLAG ){
1993 *((int*)op[j].arg) = v;
1994 }else if( op[j].type==OPT_FFLAG ){
1995 (*(void(*)(int))(op[j].arg))(v);
1996 }else if( op[j].type==OPT_FSTR ){
1997 (*(void(*)(char *))(op[j].arg))(&g_argv[i][2]);
1998 }else{
1999 if( err ){
2000 fprintf(err,"%smissing argument on switch.\n",emsg);
2001 errline(i,1,err);
2003 errcnt++;
2005 return errcnt;
2009 ** Process a command line switch which has an argument.
2011 static int handleswitch(int i, FILE *err)
2013 int lv = 0;
2014 double dv = 0.0;
2015 char *sv = 0, *end;
2016 char *cp;
2017 int j;
2018 int errcnt = 0;
2019 cp = strchr(g_argv[i],'=');
2020 assert( cp!=0 );
2021 *cp = 0;
2022 for(j=0; op[j].label; j++){
2023 if( strcmp(g_argv[i],op[j].label)==0 ) break;
2025 *cp = '=';
2026 if( op[j].label==0 ){
2027 if( err ){
2028 fprintf(err,"%sundefined option.\n",emsg);
2029 errline(i,0,err);
2031 errcnt++;
2032 }else{
2033 cp++;
2034 switch( op[j].type ){
2035 case OPT_FLAG:
2036 case OPT_FFLAG:
2037 if( err ){
2038 fprintf(err,"%soption requires an argument.\n",emsg);
2039 errline(i,0,err);
2041 errcnt++;
2042 break;
2043 case OPT_DBL:
2044 case OPT_FDBL:
2045 dv = strtod(cp,&end);
2046 if( *end ){
2047 if( err ){
2048 fprintf(err,
2049 "%sillegal character in floating-point argument.\n",emsg);
2050 errline(i,(int)((char*)end-(char*)g_argv[i]),err);
2052 errcnt++;
2054 break;
2055 case OPT_INT:
2056 case OPT_FINT:
2057 lv = strtol(cp,&end,0);
2058 if( *end ){
2059 if( err ){
2060 fprintf(err,"%sillegal character in integer argument.\n",emsg);
2061 errline(i,(int)((char*)end-(char*)g_argv[i]),err);
2063 errcnt++;
2065 break;
2066 case OPT_STR:
2067 case OPT_FSTR:
2068 sv = cp;
2069 break;
2071 switch( op[j].type ){
2072 case OPT_FLAG:
2073 case OPT_FFLAG:
2074 break;
2075 case OPT_DBL:
2076 *(double*)(op[j].arg) = dv;
2077 break;
2078 case OPT_FDBL:
2079 (*(void(*)(double))(op[j].arg))(dv);
2080 break;
2081 case OPT_INT:
2082 *(int*)(op[j].arg) = lv;
2083 break;
2084 case OPT_FINT:
2085 (*(void(*)(int))(op[j].arg))((int)lv);
2086 break;
2087 case OPT_STR:
2088 *(char**)(op[j].arg) = sv;
2089 break;
2090 case OPT_FSTR:
2091 (*(void(*)(char *))(op[j].arg))(sv);
2092 break;
2095 return errcnt;
2098 int OptInit(char **a, struct s_options *o, FILE *err)
2100 int errcnt = 0;
2101 g_argv = a;
2102 op = o;
2103 errstream = err;
2104 if( g_argv && *g_argv && op ){
2105 int i;
2106 for(i=1; g_argv[i]; i++){
2107 if( g_argv[i][0]=='+' || g_argv[i][0]=='-' ){
2108 errcnt += handleflags(i,err);
2109 }else if( strchr(g_argv[i],'=') ){
2110 errcnt += handleswitch(i,err);
2114 if( errcnt>0 ){
2115 fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
2116 OptPrint();
2117 exit(1);
2119 return 0;
2122 int OptNArgs(void){
2123 int cnt = 0;
2124 int dashdash = 0;
2125 int i;
2126 if( g_argv!=0 && g_argv[0]!=0 ){
2127 for(i=1; g_argv[i]; i++){
2128 if( dashdash || !ISOPT(g_argv[i]) ) cnt++;
2129 if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
2132 return cnt;
2135 char *OptArg(int n)
2137 int i;
2138 i = argindex(n);
2139 return i>=0 ? g_argv[i] : 0;
2142 void OptErr(int n)
2144 int i;
2145 i = argindex(n);
2146 if( i>=0 ) errline(i,0,errstream);
2149 void OptPrint(void){
2150 int i;
2151 int max, len;
2152 max = 0;
2153 for(i=0; op[i].label; i++){
2154 len = lemonStrlen(op[i].label) + 1;
2155 switch( op[i].type ){
2156 case OPT_FLAG:
2157 case OPT_FFLAG:
2158 break;
2159 case OPT_INT:
2160 case OPT_FINT:
2161 len += 9; /* length of "<integer>" */
2162 break;
2163 case OPT_DBL:
2164 case OPT_FDBL:
2165 len += 6; /* length of "<real>" */
2166 break;
2167 case OPT_STR:
2168 case OPT_FSTR:
2169 len += 8; /* length of "<string>" */
2170 break;
2172 if( len>max ) max = len;
2174 for(i=0; op[i].label; i++){
2175 switch( op[i].type ){
2176 case OPT_FLAG:
2177 case OPT_FFLAG:
2178 fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
2179 break;
2180 case OPT_INT:
2181 case OPT_FINT:
2182 fprintf(errstream," -%s<integer>%*s %s\n",op[i].label,
2183 (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
2184 break;
2185 case OPT_DBL:
2186 case OPT_FDBL:
2187 fprintf(errstream," -%s<real>%*s %s\n",op[i].label,
2188 (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
2189 break;
2190 case OPT_STR:
2191 case OPT_FSTR:
2192 fprintf(errstream," -%s<string>%*s %s\n",op[i].label,
2193 (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
2194 break;
2198 /*********************** From the file "parse.c" ****************************/
2200 ** Input file parser for the LEMON parser generator.
2203 /* The state of the parser */
2204 enum e_state {
2205 INITIALIZE,
2206 WAITING_FOR_DECL_OR_RULE,
2207 WAITING_FOR_DECL_KEYWORD,
2208 WAITING_FOR_DECL_ARG,
2209 WAITING_FOR_PRECEDENCE_SYMBOL,
2210 WAITING_FOR_ARROW,
2211 IN_RHS,
2212 LHS_ALIAS_1,
2213 LHS_ALIAS_2,
2214 LHS_ALIAS_3,
2215 RHS_ALIAS_1,
2216 RHS_ALIAS_2,
2217 PRECEDENCE_MARK_1,
2218 PRECEDENCE_MARK_2,
2219 RESYNC_AFTER_RULE_ERROR,
2220 RESYNC_AFTER_DECL_ERROR,
2221 WAITING_FOR_DESTRUCTOR_SYMBOL,
2222 WAITING_FOR_DATATYPE_SYMBOL,
2223 WAITING_FOR_FALLBACK_ID,
2224 WAITING_FOR_WILDCARD_ID,
2225 WAITING_FOR_CLASS_ID,
2226 WAITING_FOR_CLASS_TOKEN,
2227 WAITING_FOR_TOKEN_NAME
2229 struct pstate {
2230 char *filename; /* Name of the input file */
2231 int tokenlineno; /* Linenumber at which current token starts */
2232 int errorcnt; /* Number of errors so far */
2233 char *tokenstart; /* Text of current token */
2234 struct lemon *gp; /* Global state vector */
2235 enum e_state state; /* The state of the parser */
2236 struct symbol *fallback; /* The fallback token */
2237 struct symbol *tkclass; /* Token class symbol */
2238 struct symbol *lhs; /* Left-hand side of current rule */
2239 const char *lhsalias; /* Alias for the LHS */
2240 int nrhs; /* Number of right-hand side symbols seen */
2241 struct symbol *rhs[MAXRHS]; /* RHS symbols */
2242 const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
2243 struct rule *prevrule; /* Previous rule parsed */
2244 const char *declkeyword; /* Keyword of a declaration */
2245 char **declargslot; /* Where the declaration argument should be put */
2246 int insertLineMacro; /* Add #line before declaration insert */
2247 int *decllinenoslot; /* Where to write declaration line number */
2248 enum e_assoc declassoc; /* Assign this association to decl arguments */
2249 int preccounter; /* Assign this precedence to decl arguments */
2250 struct rule *firstrule; /* Pointer to first rule in the grammar */
2251 struct rule *lastrule; /* Pointer to the most recently parsed rule */
2254 /* Parse a single token */
2255 static void parseonetoken(struct pstate *psp)
2257 const char *x;
2258 x = Strsafe(psp->tokenstart); /* Save the token permanently */
2259 #if 0
2260 printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
2261 x,psp->state);
2262 #endif
2263 switch( psp->state ){
2264 case INITIALIZE:
2265 psp->prevrule = 0;
2266 psp->preccounter = 0;
2267 psp->firstrule = psp->lastrule = 0;
2268 psp->gp->nrule = 0;
2269 /* fall through */
2270 case WAITING_FOR_DECL_OR_RULE:
2271 if( x[0]=='%' ){
2272 psp->state = WAITING_FOR_DECL_KEYWORD;
2273 }else if( ISLOWER(x[0]) ){
2274 psp->lhs = Symbol_new(x);
2275 psp->nrhs = 0;
2276 psp->lhsalias = 0;
2277 psp->state = WAITING_FOR_ARROW;
2278 }else if( x[0]=='{' ){
2279 if( psp->prevrule==0 ){
2280 ErrorMsg(psp->filename,psp->tokenlineno,
2281 "There is no prior rule upon which to attach the code "
2282 "fragment which begins on this line.");
2283 psp->errorcnt++;
2284 }else if( psp->prevrule->code!=0 ){
2285 ErrorMsg(psp->filename,psp->tokenlineno,
2286 "Code fragment beginning on this line is not the first "
2287 "to follow the previous rule.");
2288 psp->errorcnt++;
2289 }else if( strcmp(x, "{NEVER-REDUCE")==0 ){
2290 psp->prevrule->neverReduce = 1;
2291 }else{
2292 psp->prevrule->line = psp->tokenlineno;
2293 psp->prevrule->code = &x[1];
2294 psp->prevrule->noCode = 0;
2296 }else if( x[0]=='[' ){
2297 psp->state = PRECEDENCE_MARK_1;
2298 }else{
2299 ErrorMsg(psp->filename,psp->tokenlineno,
2300 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2302 psp->errorcnt++;
2304 break;
2305 case PRECEDENCE_MARK_1:
2306 if( !ISUPPER(x[0]) ){
2307 ErrorMsg(psp->filename,psp->tokenlineno,
2308 "The precedence symbol must be a terminal.");
2309 psp->errorcnt++;
2310 }else if( psp->prevrule==0 ){
2311 ErrorMsg(psp->filename,psp->tokenlineno,
2312 "There is no prior rule to assign precedence \"[%s]\".",x);
2313 psp->errorcnt++;
2314 }else if( psp->prevrule->precsym!=0 ){
2315 ErrorMsg(psp->filename,psp->tokenlineno,
2316 "Precedence mark on this line is not the first "
2317 "to follow the previous rule.");
2318 psp->errorcnt++;
2319 }else{
2320 psp->prevrule->precsym = Symbol_new(x);
2322 psp->state = PRECEDENCE_MARK_2;
2323 break;
2324 case PRECEDENCE_MARK_2:
2325 if( x[0]!=']' ){
2326 ErrorMsg(psp->filename,psp->tokenlineno,
2327 "Missing \"]\" on precedence mark.");
2328 psp->errorcnt++;
2330 psp->state = WAITING_FOR_DECL_OR_RULE;
2331 break;
2332 case WAITING_FOR_ARROW:
2333 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2334 psp->state = IN_RHS;
2335 }else if( x[0]=='(' ){
2336 psp->state = LHS_ALIAS_1;
2337 }else{
2338 ErrorMsg(psp->filename,psp->tokenlineno,
2339 "Expected to see a \":\" following the LHS symbol \"%s\".",
2340 psp->lhs->name);
2341 psp->errorcnt++;
2342 psp->state = RESYNC_AFTER_RULE_ERROR;
2344 break;
2345 case LHS_ALIAS_1:
2346 if( ISALPHA(x[0]) ){
2347 psp->lhsalias = x;
2348 psp->state = LHS_ALIAS_2;
2349 }else{
2350 ErrorMsg(psp->filename,psp->tokenlineno,
2351 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2352 x,psp->lhs->name);
2353 psp->errorcnt++;
2354 psp->state = RESYNC_AFTER_RULE_ERROR;
2356 break;
2357 case LHS_ALIAS_2:
2358 if( x[0]==')' ){
2359 psp->state = LHS_ALIAS_3;
2360 }else{
2361 ErrorMsg(psp->filename,psp->tokenlineno,
2362 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2363 psp->errorcnt++;
2364 psp->state = RESYNC_AFTER_RULE_ERROR;
2366 break;
2367 case LHS_ALIAS_3:
2368 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2369 psp->state = IN_RHS;
2370 }else{
2371 ErrorMsg(psp->filename,psp->tokenlineno,
2372 "Missing \"->\" following: \"%s(%s)\".",
2373 psp->lhs->name,psp->lhsalias);
2374 psp->errorcnt++;
2375 psp->state = RESYNC_AFTER_RULE_ERROR;
2377 break;
2378 case IN_RHS:
2379 if( x[0]=='.' ){
2380 struct rule *rp;
2381 rp = (struct rule *)calloc( sizeof(struct rule) +
2382 sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
2383 if( rp==0 ){
2384 ErrorMsg(psp->filename,psp->tokenlineno,
2385 "Can't allocate enough memory for this rule.");
2386 psp->errorcnt++;
2387 psp->prevrule = 0;
2388 }else{
2389 int i;
2390 rp->ruleline = psp->tokenlineno;
2391 rp->rhs = (struct symbol**)&rp[1];
2392 rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]);
2393 for(i=0; i<psp->nrhs; i++){
2394 rp->rhs[i] = psp->rhs[i];
2395 rp->rhsalias[i] = psp->alias[i];
2396 if( rp->rhsalias[i]!=0 ){ rp->rhs[i]->bContent = 1; }
2398 rp->lhs = psp->lhs;
2399 rp->lhsalias = psp->lhsalias;
2400 rp->nrhs = psp->nrhs;
2401 rp->code = 0;
2402 rp->noCode = 1;
2403 rp->precsym = 0;
2404 rp->index = psp->gp->nrule++;
2405 rp->nextlhs = rp->lhs->rule;
2406 rp->lhs->rule = rp;
2407 rp->next = 0;
2408 if( psp->firstrule==0 ){
2409 psp->firstrule = psp->lastrule = rp;
2410 }else{
2411 psp->lastrule->next = rp;
2412 psp->lastrule = rp;
2414 psp->prevrule = rp;
2416 psp->state = WAITING_FOR_DECL_OR_RULE;
2417 }else if( ISALPHA(x[0]) ){
2418 if( psp->nrhs>=MAXRHS ){
2419 ErrorMsg(psp->filename,psp->tokenlineno,
2420 "Too many symbols on RHS of rule beginning at \"%s\".",
2422 psp->errorcnt++;
2423 psp->state = RESYNC_AFTER_RULE_ERROR;
2424 }else{
2425 psp->rhs[psp->nrhs] = Symbol_new(x);
2426 psp->alias[psp->nrhs] = 0;
2427 psp->nrhs++;
2429 }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 && ISUPPER(x[1]) ){
2430 struct symbol *msp = psp->rhs[psp->nrhs-1];
2431 if( msp->type!=MULTITERMINAL ){
2432 struct symbol *origsp = msp;
2433 msp = (struct symbol *) calloc(1,sizeof(*msp));
2434 memset(msp, 0, sizeof(*msp));
2435 msp->type = MULTITERMINAL;
2436 msp->nsubsym = 1;
2437 msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*));
2438 msp->subsym[0] = origsp;
2439 msp->name = origsp->name;
2440 psp->rhs[psp->nrhs-1] = msp;
2442 msp->nsubsym++;
2443 msp->subsym = (struct symbol **) realloc(msp->subsym,
2444 sizeof(struct symbol*)*msp->nsubsym);
2445 msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
2446 if( ISLOWER(x[1]) || ISLOWER(msp->subsym[0]->name[0]) ){
2447 ErrorMsg(psp->filename,psp->tokenlineno,
2448 "Cannot form a compound containing a non-terminal");
2449 psp->errorcnt++;
2451 }else if( x[0]=='(' && psp->nrhs>0 ){
2452 psp->state = RHS_ALIAS_1;
2453 }else{
2454 ErrorMsg(psp->filename,psp->tokenlineno,
2455 "Illegal character on RHS of rule: \"%s\".",x);
2456 psp->errorcnt++;
2457 psp->state = RESYNC_AFTER_RULE_ERROR;
2459 break;
2460 case RHS_ALIAS_1:
2461 if( ISALPHA(x[0]) ){
2462 psp->alias[psp->nrhs-1] = x;
2463 psp->state = RHS_ALIAS_2;
2464 }else{
2465 ErrorMsg(psp->filename,psp->tokenlineno,
2466 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2467 x,psp->rhs[psp->nrhs-1]->name);
2468 psp->errorcnt++;
2469 psp->state = RESYNC_AFTER_RULE_ERROR;
2471 break;
2472 case RHS_ALIAS_2:
2473 if( x[0]==')' ){
2474 psp->state = IN_RHS;
2475 }else{
2476 ErrorMsg(psp->filename,psp->tokenlineno,
2477 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2478 psp->errorcnt++;
2479 psp->state = RESYNC_AFTER_RULE_ERROR;
2481 break;
2482 case WAITING_FOR_DECL_KEYWORD:
2483 if( ISALPHA(x[0]) ){
2484 psp->declkeyword = x;
2485 psp->declargslot = 0;
2486 psp->decllinenoslot = 0;
2487 psp->insertLineMacro = 1;
2488 psp->state = WAITING_FOR_DECL_ARG;
2489 if( strcmp(x,"name")==0 ){
2490 psp->declargslot = &(psp->gp->name);
2491 psp->insertLineMacro = 0;
2492 }else if( strcmp(x,"include")==0 ){
2493 psp->declargslot = &(psp->gp->include);
2494 }else if( strcmp(x,"code")==0 ){
2495 psp->declargslot = &(psp->gp->extracode);
2496 }else if( strcmp(x,"token_destructor")==0 ){
2497 psp->declargslot = &psp->gp->tokendest;
2498 }else if( strcmp(x,"default_destructor")==0 ){
2499 psp->declargslot = &psp->gp->vardest;
2500 }else if( strcmp(x,"token_prefix")==0 ){
2501 psp->declargslot = &psp->gp->tokenprefix;
2502 psp->insertLineMacro = 0;
2503 }else if( strcmp(x,"syntax_error")==0 ){
2504 psp->declargslot = &(psp->gp->error);
2505 }else if( strcmp(x,"parse_accept")==0 ){
2506 psp->declargslot = &(psp->gp->accept);
2507 }else if( strcmp(x,"parse_failure")==0 ){
2508 psp->declargslot = &(psp->gp->failure);
2509 }else if( strcmp(x,"stack_overflow")==0 ){
2510 psp->declargslot = &(psp->gp->overflow);
2511 }else if( strcmp(x,"extra_argument")==0 ){
2512 psp->declargslot = &(psp->gp->arg);
2513 psp->insertLineMacro = 0;
2514 }else if( strcmp(x,"extra_context")==0 ){
2515 psp->declargslot = &(psp->gp->ctx);
2516 psp->insertLineMacro = 0;
2517 }else if( strcmp(x,"token_type")==0 ){
2518 psp->declargslot = &(psp->gp->tokentype);
2519 psp->insertLineMacro = 0;
2520 }else if( strcmp(x,"default_type")==0 ){
2521 psp->declargslot = &(psp->gp->vartype);
2522 psp->insertLineMacro = 0;
2523 }else if( strcmp(x,"stack_size")==0 ){
2524 psp->declargslot = &(psp->gp->stacksize);
2525 psp->insertLineMacro = 0;
2526 }else if( strcmp(x,"start_symbol")==0 ){
2527 psp->declargslot = &(psp->gp->start);
2528 psp->insertLineMacro = 0;
2529 }else if( strcmp(x,"left")==0 ){
2530 psp->preccounter++;
2531 psp->declassoc = LEFT;
2532 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2533 }else if( strcmp(x,"right")==0 ){
2534 psp->preccounter++;
2535 psp->declassoc = RIGHT;
2536 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2537 }else if( strcmp(x,"nonassoc")==0 ){
2538 psp->preccounter++;
2539 psp->declassoc = NONE;
2540 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2541 }else if( strcmp(x,"destructor")==0 ){
2542 psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2543 }else if( strcmp(x,"type")==0 ){
2544 psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2545 }else if( strcmp(x,"fallback")==0 ){
2546 psp->fallback = 0;
2547 psp->state = WAITING_FOR_FALLBACK_ID;
2548 }else if( strcmp(x,"token")==0 ){
2549 psp->state = WAITING_FOR_TOKEN_NAME;
2550 }else if( strcmp(x,"wildcard")==0 ){
2551 psp->state = WAITING_FOR_WILDCARD_ID;
2552 }else if( strcmp(x,"token_class")==0 ){
2553 psp->state = WAITING_FOR_CLASS_ID;
2554 }else{
2555 ErrorMsg(psp->filename,psp->tokenlineno,
2556 "Unknown declaration keyword: \"%%%s\".",x);
2557 psp->errorcnt++;
2558 psp->state = RESYNC_AFTER_DECL_ERROR;
2560 }else{
2561 ErrorMsg(psp->filename,psp->tokenlineno,
2562 "Illegal declaration keyword: \"%s\".",x);
2563 psp->errorcnt++;
2564 psp->state = RESYNC_AFTER_DECL_ERROR;
2566 break;
2567 case WAITING_FOR_DESTRUCTOR_SYMBOL:
2568 if( !ISALPHA(x[0]) ){
2569 ErrorMsg(psp->filename,psp->tokenlineno,
2570 "Symbol name missing after %%destructor keyword");
2571 psp->errorcnt++;
2572 psp->state = RESYNC_AFTER_DECL_ERROR;
2573 }else{
2574 struct symbol *sp = Symbol_new(x);
2575 psp->declargslot = &sp->destructor;
2576 psp->decllinenoslot = &sp->destLineno;
2577 psp->insertLineMacro = 1;
2578 psp->state = WAITING_FOR_DECL_ARG;
2580 break;
2581 case WAITING_FOR_DATATYPE_SYMBOL:
2582 if( !ISALPHA(x[0]) ){
2583 ErrorMsg(psp->filename,psp->tokenlineno,
2584 "Symbol name missing after %%type keyword");
2585 psp->errorcnt++;
2586 psp->state = RESYNC_AFTER_DECL_ERROR;
2587 }else{
2588 struct symbol *sp = Symbol_find(x);
2589 if((sp) && (sp->datatype)){
2590 ErrorMsg(psp->filename,psp->tokenlineno,
2591 "Symbol %%type \"%s\" already defined", x);
2592 psp->errorcnt++;
2593 psp->state = RESYNC_AFTER_DECL_ERROR;
2594 }else{
2595 if (!sp){
2596 sp = Symbol_new(x);
2598 psp->declargslot = &sp->datatype;
2599 psp->insertLineMacro = 0;
2600 psp->state = WAITING_FOR_DECL_ARG;
2603 break;
2604 case WAITING_FOR_PRECEDENCE_SYMBOL:
2605 if( x[0]=='.' ){
2606 psp->state = WAITING_FOR_DECL_OR_RULE;
2607 }else if( ISUPPER(x[0]) ){
2608 struct symbol *sp;
2609 sp = Symbol_new(x);
2610 if( sp->prec>=0 ){
2611 ErrorMsg(psp->filename,psp->tokenlineno,
2612 "Symbol \"%s\" has already be given a precedence.",x);
2613 psp->errorcnt++;
2614 }else{
2615 sp->prec = psp->preccounter;
2616 sp->assoc = psp->declassoc;
2618 }else{
2619 ErrorMsg(psp->filename,psp->tokenlineno,
2620 "Can't assign a precedence to \"%s\".",x);
2621 psp->errorcnt++;
2623 break;
2624 case WAITING_FOR_DECL_ARG:
2625 if( x[0]=='{' || x[0]=='\"' || ISALNUM(x[0]) ){
2626 const char *zOld, *zNew;
2627 char *zBuf, *z;
2628 int nOld, n, nLine = 0, nNew, nBack;
2629 int addLineMacro;
2630 char zLine[50];
2631 zNew = x;
2632 if( zNew[0]=='"' || zNew[0]=='{' ) zNew++;
2633 nNew = lemonStrlen(zNew);
2634 if( *psp->declargslot ){
2635 zOld = *psp->declargslot;
2636 }else{
2637 zOld = "";
2639 nOld = lemonStrlen(zOld);
2640 n = nOld + nNew + 20;
2641 addLineMacro = !psp->gp->nolinenosflag
2642 && psp->insertLineMacro
2643 && psp->tokenlineno>1
2644 && (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0);
2645 if( addLineMacro ){
2646 for(z=psp->filename, nBack=0; *z; z++){
2647 if( *z=='\\' ) nBack++;
2649 lemon_sprintf(zLine, "#line %d ", psp->tokenlineno);
2650 nLine = lemonStrlen(zLine);
2651 n += nLine + lemonStrlen(psp->filename) + nBack;
2653 *psp->declargslot = (char *) realloc(*psp->declargslot, n);
2654 zBuf = *psp->declargslot + nOld;
2655 if( addLineMacro ){
2656 if( nOld && zBuf[-1]!='\n' ){
2657 *(zBuf++) = '\n';
2659 memcpy(zBuf, zLine, nLine);
2660 zBuf += nLine;
2661 *(zBuf++) = '"';
2662 for(z=psp->filename; *z; z++){
2663 if( *z=='\\' ){
2664 *(zBuf++) = '\\';
2666 *(zBuf++) = *z;
2668 *(zBuf++) = '"';
2669 *(zBuf++) = '\n';
2671 if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){
2672 psp->decllinenoslot[0] = psp->tokenlineno;
2674 memcpy(zBuf, zNew, nNew);
2675 zBuf += nNew;
2676 *zBuf = 0;
2677 psp->state = WAITING_FOR_DECL_OR_RULE;
2678 }else{
2679 ErrorMsg(psp->filename,psp->tokenlineno,
2680 "Illegal argument to %%%s: %s",psp->declkeyword,x);
2681 psp->errorcnt++;
2682 psp->state = RESYNC_AFTER_DECL_ERROR;
2684 break;
2685 case WAITING_FOR_FALLBACK_ID:
2686 if( x[0]=='.' ){
2687 psp->state = WAITING_FOR_DECL_OR_RULE;
2688 }else if( !ISUPPER(x[0]) ){
2689 ErrorMsg(psp->filename, psp->tokenlineno,
2690 "%%fallback argument \"%s\" should be a token", x);
2691 psp->errorcnt++;
2692 }else{
2693 struct symbol *sp = Symbol_new(x);
2694 if( psp->fallback==0 ){
2695 psp->fallback = sp;
2696 }else if( sp->fallback ){
2697 ErrorMsg(psp->filename, psp->tokenlineno,
2698 "More than one fallback assigned to token %s", x);
2699 psp->errorcnt++;
2700 }else{
2701 sp->fallback = psp->fallback;
2702 psp->gp->has_fallback = 1;
2705 break;
2706 case WAITING_FOR_TOKEN_NAME:
2707 /* Tokens do not have to be declared before use. But they can be
2708 ** in order to control their assigned integer number. The number for
2709 ** each token is assigned when it is first seen. So by including
2711 ** %token ONE TWO THREE.
2713 ** early in the grammar file, that assigns small consecutive values
2714 ** to each of the tokens ONE TWO and THREE.
2716 if( x[0]=='.' ){
2717 psp->state = WAITING_FOR_DECL_OR_RULE;
2718 }else if( !ISUPPER(x[0]) ){
2719 ErrorMsg(psp->filename, psp->tokenlineno,
2720 "%%token argument \"%s\" should be a token", x);
2721 psp->errorcnt++;
2722 }else{
2723 (void)Symbol_new(x);
2725 break;
2726 case WAITING_FOR_WILDCARD_ID:
2727 if( x[0]=='.' ){
2728 psp->state = WAITING_FOR_DECL_OR_RULE;
2729 }else if( !ISUPPER(x[0]) ){
2730 ErrorMsg(psp->filename, psp->tokenlineno,
2731 "%%wildcard argument \"%s\" should be a token", x);
2732 psp->errorcnt++;
2733 }else{
2734 struct symbol *sp = Symbol_new(x);
2735 if( psp->gp->wildcard==0 ){
2736 psp->gp->wildcard = sp;
2737 }else{
2738 ErrorMsg(psp->filename, psp->tokenlineno,
2739 "Extra wildcard to token: %s", x);
2740 psp->errorcnt++;
2743 break;
2744 case WAITING_FOR_CLASS_ID:
2745 if( !ISLOWER(x[0]) ){
2746 ErrorMsg(psp->filename, psp->tokenlineno,
2747 "%%token_class must be followed by an identifier: %s", x);
2748 psp->errorcnt++;
2749 psp->state = RESYNC_AFTER_DECL_ERROR;
2750 }else if( Symbol_find(x) ){
2751 ErrorMsg(psp->filename, psp->tokenlineno,
2752 "Symbol \"%s\" already used", x);
2753 psp->errorcnt++;
2754 psp->state = RESYNC_AFTER_DECL_ERROR;
2755 }else{
2756 psp->tkclass = Symbol_new(x);
2757 psp->tkclass->type = MULTITERMINAL;
2758 psp->state = WAITING_FOR_CLASS_TOKEN;
2760 break;
2761 case WAITING_FOR_CLASS_TOKEN:
2762 if( x[0]=='.' ){
2763 psp->state = WAITING_FOR_DECL_OR_RULE;
2764 }else if( ISUPPER(x[0]) || ((x[0]=='|' || x[0]=='/') && ISUPPER(x[1])) ){
2765 struct symbol *msp = psp->tkclass;
2766 msp->nsubsym++;
2767 msp->subsym = (struct symbol **) realloc(msp->subsym,
2768 sizeof(struct symbol*)*msp->nsubsym);
2769 if( !ISUPPER(x[0]) ) x++;
2770 msp->subsym[msp->nsubsym-1] = Symbol_new(x);
2771 }else{
2772 ErrorMsg(psp->filename, psp->tokenlineno,
2773 "%%token_class argument \"%s\" should be a token", x);
2774 psp->errorcnt++;
2775 psp->state = RESYNC_AFTER_DECL_ERROR;
2777 break;
2778 case RESYNC_AFTER_RULE_ERROR:
2779 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2780 ** break; */
2781 case RESYNC_AFTER_DECL_ERROR:
2782 if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2783 if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2784 break;
2788 /* The text in the input is part of the argument to an %ifdef or %ifndef.
2789 ** Evaluate the text as a boolean expression. Return true or false.
2791 static int eval_preprocessor_boolean(char *z, int lineno){
2792 int neg = 0;
2793 int res = 0;
2794 int okTerm = 1;
2795 int i;
2796 for(i=0; z[i]!=0; i++){
2797 if( ISSPACE(z[i]) ) continue;
2798 if( z[i]=='!' ){
2799 if( !okTerm ) goto pp_syntax_error;
2800 neg = !neg;
2801 continue;
2803 if( z[i]=='|' && z[i+1]=='|' ){
2804 if( okTerm ) goto pp_syntax_error;
2805 if( res ) return 1;
2806 i++;
2807 okTerm = 1;
2808 continue;
2810 if( z[i]=='&' && z[i+1]=='&' ){
2811 if( okTerm ) goto pp_syntax_error;
2812 if( !res ) return 0;
2813 i++;
2814 okTerm = 1;
2815 continue;
2817 if( z[i]=='(' ){
2818 int k;
2819 int n = 1;
2820 if( !okTerm ) goto pp_syntax_error;
2821 for(k=i+1; z[k]; k++){
2822 if( z[k]==')' ){
2823 n--;
2824 if( n==0 ){
2825 z[k] = 0;
2826 res = eval_preprocessor_boolean(&z[i+1], -1);
2827 z[k] = ')';
2828 if( res<0 ){
2829 i = i-res;
2830 goto pp_syntax_error;
2832 i = k;
2833 break;
2835 }else if( z[k]=='(' ){
2836 n++;
2837 }else if( z[k]==0 ){
2838 i = k;
2839 goto pp_syntax_error;
2842 if( neg ){
2843 res = !res;
2844 neg = 0;
2846 okTerm = 0;
2847 continue;
2849 if( ISALPHA(z[i]) ){
2850 int j, k, n;
2851 if( !okTerm ) goto pp_syntax_error;
2852 for(k=i+1; ISALNUM(z[k]) || z[k]=='_'; k++){}
2853 n = k - i;
2854 res = 0;
2855 for(j=0; j<nDefine; j++){
2856 if( strncmp(azDefine[j],&z[i],n)==0 && azDefine[j][n]==0 ){
2857 res = 1;
2858 break;
2861 i = k-1;
2862 if( neg ){
2863 res = !res;
2864 neg = 0;
2866 okTerm = 0;
2867 continue;
2869 goto pp_syntax_error;
2871 return res;
2873 pp_syntax_error:
2874 if( lineno>0 ){
2875 fprintf(stderr, "%%if syntax error on line %d.\n", lineno);
2876 fprintf(stderr, " %.*s <-- syntax error here\n", i+1, z);
2877 exit(1);
2878 }else{
2879 return -(i+1);
2883 /* Run the preprocessor over the input file text. The global variables
2884 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2885 ** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2886 ** comments them out. Text in between is also commented out as appropriate.
2888 static void preprocess_input(char *z){
2889 int i, j, k;
2890 int exclude = 0;
2891 int start = 0;
2892 int lineno = 1;
2893 int start_lineno = 1;
2894 for(i=0; z[i]; i++){
2895 if( z[i]=='\n' ) lineno++;
2896 if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue;
2897 if( strncmp(&z[i],"%endif",6)==0 && ISSPACE(z[i+6]) ){
2898 if( exclude ){
2899 exclude--;
2900 if( exclude==0 ){
2901 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2904 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2905 }else if( strncmp(&z[i],"%else",5)==0 && ISSPACE(z[i+5]) ){
2906 if( exclude==1){
2907 exclude = 0;
2908 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2909 }else if( exclude==0 ){
2910 exclude = 1;
2911 start = i;
2912 start_lineno = lineno;
2914 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2915 }else if( strncmp(&z[i],"%ifdef ",7)==0
2916 || strncmp(&z[i],"%if ",4)==0
2917 || strncmp(&z[i],"%ifndef ",8)==0 ){
2918 if( exclude ){
2919 exclude++;
2920 }else{
2921 int isNot;
2922 int iBool;
2923 for(j=i; z[j] && !ISSPACE(z[j]); j++){}
2924 iBool = j;
2925 isNot = (j==i+7);
2926 while( z[j] && z[j]!='\n' ){ j++; }
2927 k = z[j];
2928 z[j] = 0;
2929 exclude = eval_preprocessor_boolean(&z[iBool], lineno);
2930 z[j] = k;
2931 if( !isNot ) exclude = !exclude;
2932 if( exclude ){
2933 start = i;
2934 start_lineno = lineno;
2937 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2940 if( exclude ){
2941 fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno);
2942 exit(1);
2946 /* In spite of its name, this function is really a scanner. It read
2947 ** in the entire input file (all at once) then tokenizes it. Each
2948 ** token is passed to the function "parseonetoken" which builds all
2949 ** the appropriate data structures in the global state vector "gp".
2951 void Parse(struct lemon *gp)
2953 struct pstate ps;
2954 FILE *fp;
2955 char *filebuf;
2956 unsigned int filesize;
2957 int lineno;
2958 int c;
2959 char *cp, *nextcp;
2960 int startline = 0;
2962 memset(&ps, '\0', sizeof(ps));
2963 ps.gp = gp;
2964 ps.filename = gp->filename;
2965 ps.errorcnt = 0;
2966 ps.state = INITIALIZE;
2968 /* Begin by reading the input file */
2969 fp = fopen(ps.filename,"rb");
2970 if( fp==0 ){
2971 ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2972 gp->errorcnt++;
2973 return;
2975 fseek(fp,0,2);
2976 filesize = ftell(fp);
2977 rewind(fp);
2978 filebuf = (char *)malloc( filesize+1 );
2979 if( filesize>100000000 || filebuf==0 ){
2980 ErrorMsg(ps.filename,0,"Input file too large.");
2981 free(filebuf);
2982 gp->errorcnt++;
2983 fclose(fp);
2984 return;
2986 if( fread(filebuf,1,filesize,fp)!=filesize ){
2987 ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2988 filesize);
2989 free(filebuf);
2990 gp->errorcnt++;
2991 fclose(fp);
2992 return;
2994 fclose(fp);
2995 filebuf[filesize] = 0;
2997 /* Make an initial pass through the file to handle %ifdef and %ifndef */
2998 preprocess_input(filebuf);
2999 if( gp->printPreprocessed ){
3000 printf("%s\n", filebuf);
3001 return;
3004 /* Now scan the text of the input file */
3005 lineno = 1;
3006 for(cp=filebuf; (c= *cp)!=0; ){
3007 if( c=='\n' ) lineno++; /* Keep track of the line number */
3008 if( ISSPACE(c) ){ cp++; continue; } /* Skip all white space */
3009 if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */
3010 cp+=2;
3011 while( (c= *cp)!=0 && c!='\n' ) cp++;
3012 continue;
3014 if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */
3015 cp+=2;
3016 while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
3017 if( c=='\n' ) lineno++;
3018 cp++;
3020 if( c ) cp++;
3021 continue;
3023 ps.tokenstart = cp; /* Mark the beginning of the token */
3024 ps.tokenlineno = lineno; /* Linenumber on which token begins */
3025 if( c=='\"' ){ /* String literals */
3026 cp++;
3027 while( (c= *cp)!=0 && c!='\"' ){
3028 if( c=='\n' ) lineno++;
3029 cp++;
3031 if( c==0 ){
3032 ErrorMsg(ps.filename,startline,
3033 "String starting on this line is not terminated before "
3034 "the end of the file.");
3035 ps.errorcnt++;
3036 nextcp = cp;
3037 }else{
3038 nextcp = cp+1;
3040 }else if( c=='{' ){ /* A block of C code */
3041 int level;
3042 cp++;
3043 for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
3044 if( c=='\n' ) lineno++;
3045 else if( c=='{' ) level++;
3046 else if( c=='}' ) level--;
3047 else if( c=='/' && cp[1]=='*' ){ /* Skip comments */
3048 int prevc;
3049 cp = &cp[2];
3050 prevc = 0;
3051 while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
3052 if( c=='\n' ) lineno++;
3053 prevc = c;
3054 cp++;
3056 }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */
3057 cp = &cp[2];
3058 while( (c= *cp)!=0 && c!='\n' ) cp++;
3059 if( c ) lineno++;
3060 }else if( c=='\'' || c=='\"' ){ /* String a character literals */
3061 int startchar, prevc;
3062 startchar = c;
3063 prevc = 0;
3064 for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
3065 if( c=='\n' ) lineno++;
3066 if( prevc=='\\' ) prevc = 0;
3067 else prevc = c;
3071 if( c==0 ){
3072 ErrorMsg(ps.filename,ps.tokenlineno,
3073 "C code starting on this line is not terminated before "
3074 "the end of the file.");
3075 ps.errorcnt++;
3076 nextcp = cp;
3077 }else{
3078 nextcp = cp+1;
3080 }else if( ISALNUM(c) ){ /* Identifiers */
3081 while( (c= *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
3082 nextcp = cp;
3083 }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
3084 cp += 3;
3085 nextcp = cp;
3086 }else if( (c=='/' || c=='|') && ISALPHA(cp[1]) ){
3087 cp += 2;
3088 while( (c = *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
3089 nextcp = cp;
3090 }else{ /* All other (one character) operators */
3091 cp++;
3092 nextcp = cp;
3094 c = *cp;
3095 *cp = 0; /* Null terminate the token */
3096 parseonetoken(&ps); /* Parse the token */
3097 *cp = (char)c; /* Restore the buffer */
3098 cp = nextcp;
3100 free(filebuf); /* Release the buffer after parsing */
3101 gp->rule = ps.firstrule;
3102 gp->errorcnt = ps.errorcnt;
3104 /*************************** From the file "plink.c" *********************/
3106 ** Routines processing configuration follow-set propagation links
3107 ** in the LEMON parser generator.
3109 static struct plink *plink_freelist = 0;
3111 /* Allocate a new plink */
3112 struct plink *Plink_new(void){
3113 struct plink *newlink;
3115 if( plink_freelist==0 ){
3116 int i;
3117 int amt = 100;
3118 plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) );
3119 if( plink_freelist==0 ){
3120 fprintf(stderr,
3121 "Unable to allocate memory for a new follow-set propagation link.\n");
3122 exit(1);
3124 for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
3125 plink_freelist[amt-1].next = 0;
3127 newlink = plink_freelist;
3128 plink_freelist = plink_freelist->next;
3129 return newlink;
3132 /* Add a plink to a plink list */
3133 void Plink_add(struct plink **plpp, struct config *cfp)
3135 struct plink *newlink;
3136 newlink = Plink_new();
3137 newlink->next = *plpp;
3138 *plpp = newlink;
3139 newlink->cfp = cfp;
3142 /* Transfer every plink on the list "from" to the list "to" */
3143 void Plink_copy(struct plink **to, struct plink *from)
3145 struct plink *nextpl;
3146 while( from ){
3147 nextpl = from->next;
3148 from->next = *to;
3149 *to = from;
3150 from = nextpl;
3154 /* Delete every plink on the list */
3155 void Plink_delete(struct plink *plp)
3157 struct plink *nextpl;
3159 while( plp ){
3160 nextpl = plp->next;
3161 plp->next = plink_freelist;
3162 plink_freelist = plp;
3163 plp = nextpl;
3166 /*********************** From the file "report.c" **************************/
3168 ** Procedures for generating reports and tables in the LEMON parser generator.
3171 /* Generate a filename with the given suffix. Space to hold the
3172 ** name comes from malloc() and must be freed by the calling
3173 ** function.
3175 PRIVATE char *file_makename(struct lemon *lemp, const char *suffix)
3177 char *name;
3178 char *cp;
3179 char *filename = lemp->filename;
3180 int sz;
3182 if( outputDir ){
3183 cp = strrchr(filename, '/');
3184 if( cp ) filename = cp + 1;
3186 sz = lemonStrlen(filename);
3187 sz += lemonStrlen(suffix);
3188 if( outputDir ) sz += lemonStrlen(outputDir) + 1;
3189 sz += 5;
3190 name = (char*)malloc( sz );
3191 if( name==0 ){
3192 fprintf(stderr,"Can't allocate space for a filename.\n");
3193 exit(1);
3195 name[0] = 0;
3196 if( outputDir ){
3197 lemon_strcpy(name, outputDir);
3198 lemon_strcat(name, "/");
3200 lemon_strcat(name,filename);
3201 cp = strrchr(name,'.');
3202 if( cp ) *cp = 0;
3203 lemon_strcat(name,suffix);
3204 return name;
3207 /* Open a file with a name based on the name of the input file,
3208 ** but with a different (specified) suffix, and return a pointer
3209 ** to the stream */
3210 PRIVATE FILE *file_open(
3211 struct lemon *lemp,
3212 const char *suffix,
3213 const char *mode
3215 FILE *fp;
3217 if( lemp->outname ) free(lemp->outname);
3218 lemp->outname = file_makename(lemp, suffix);
3219 fp = fopen(lemp->outname,mode);
3220 if( fp==0 && *mode=='w' ){
3221 fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
3222 lemp->errorcnt++;
3223 return 0;
3225 return fp;
3228 /* Print the text of a rule
3230 void rule_print(FILE *out, struct rule *rp){
3231 int i, j;
3232 fprintf(out, "%s",rp->lhs->name);
3233 /* if( rp->lhsalias ) fprintf(out,"(%s)",rp->lhsalias); */
3234 fprintf(out," ::=");
3235 for(i=0; i<rp->nrhs; i++){
3236 struct symbol *sp = rp->rhs[i];
3237 if( sp->type==MULTITERMINAL ){
3238 fprintf(out," %s", sp->subsym[0]->name);
3239 for(j=1; j<sp->nsubsym; j++){
3240 fprintf(out,"|%s", sp->subsym[j]->name);
3242 }else{
3243 fprintf(out," %s", sp->name);
3245 /* if( rp->rhsalias[i] ) fprintf(out,"(%s)",rp->rhsalias[i]); */
3249 /* Duplicate the input file without comments and without actions
3250 ** on rules */
3251 void Reprint(struct lemon *lemp)
3253 struct rule *rp;
3254 struct symbol *sp;
3255 int i, j, maxlen, len, ncolumns, skip;
3256 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
3257 maxlen = 10;
3258 for(i=0; i<lemp->nsymbol; i++){
3259 sp = lemp->symbols[i];
3260 len = lemonStrlen(sp->name);
3261 if( len>maxlen ) maxlen = len;
3263 ncolumns = 76/(maxlen+5);
3264 if( ncolumns<1 ) ncolumns = 1;
3265 skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
3266 for(i=0; i<skip; i++){
3267 printf("//");
3268 for(j=i; j<lemp->nsymbol; j+=skip){
3269 sp = lemp->symbols[j];
3270 assert( sp->index==j );
3271 printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
3273 printf("\n");
3275 for(rp=lemp->rule; rp; rp=rp->next){
3276 rule_print(stdout, rp);
3277 printf(".");
3278 if( rp->precsym ) printf(" [%s]",rp->precsym->name);
3279 /* if( rp->code ) printf("\n %s",rp->code); */
3280 printf("\n");
3284 /* Print a single rule.
3286 void RulePrint(FILE *fp, struct rule *rp, int iCursor){
3287 struct symbol *sp;
3288 int i, j;
3289 fprintf(fp,"%s ::=",rp->lhs->name);
3290 for(i=0; i<=rp->nrhs; i++){
3291 if( i==iCursor ) fprintf(fp," *");
3292 if( i==rp->nrhs ) break;
3293 sp = rp->rhs[i];
3294 if( sp->type==MULTITERMINAL ){
3295 fprintf(fp," %s", sp->subsym[0]->name);
3296 for(j=1; j<sp->nsubsym; j++){
3297 fprintf(fp,"|%s",sp->subsym[j]->name);
3299 }else{
3300 fprintf(fp," %s", sp->name);
3305 /* Print the rule for a configuration.
3307 void ConfigPrint(FILE *fp, struct config *cfp){
3308 RulePrint(fp, cfp->rp, cfp->dot);
3311 /* #define TEST */
3312 #if 0
3313 /* Print a set */
3314 PRIVATE void SetPrint(out,set,lemp)
3315 FILE *out;
3316 char *set;
3317 struct lemon *lemp;
3319 int i;
3320 char *spacer;
3321 spacer = "";
3322 fprintf(out,"%12s[","");
3323 for(i=0; i<lemp->nterminal; i++){
3324 if( SetFind(set,i) ){
3325 fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
3326 spacer = " ";
3329 fprintf(out,"]\n");
3332 /* Print a plink chain */
3333 PRIVATE void PlinkPrint(out,plp,tag)
3334 FILE *out;
3335 struct plink *plp;
3336 char *tag;
3338 while( plp ){
3339 fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum);
3340 ConfigPrint(out,plp->cfp);
3341 fprintf(out,"\n");
3342 plp = plp->next;
3345 #endif
3347 /* Print an action to the given file descriptor. Return FALSE if
3348 ** nothing was actually printed.
3350 int PrintAction(
3351 struct action *ap, /* The action to print */
3352 FILE *fp, /* Print the action here */
3353 int indent /* Indent by this amount */
3355 int result = 1;
3356 switch( ap->type ){
3357 case SHIFT: {
3358 struct state *stp = ap->x.stp;
3359 fprintf(fp,"%*s shift %-7d",indent,ap->sp->name,stp->statenum);
3360 break;
3362 case REDUCE: {
3363 struct rule *rp = ap->x.rp;
3364 fprintf(fp,"%*s reduce %-7d",indent,ap->sp->name,rp->iRule);
3365 RulePrint(fp, rp, -1);
3366 break;
3368 case SHIFTREDUCE: {
3369 struct rule *rp = ap->x.rp;
3370 fprintf(fp,"%*s shift-reduce %-7d",indent,ap->sp->name,rp->iRule);
3371 RulePrint(fp, rp, -1);
3372 break;
3374 case ACCEPT:
3375 fprintf(fp,"%*s accept",indent,ap->sp->name);
3376 break;
3377 case ERROR:
3378 fprintf(fp,"%*s error",indent,ap->sp->name);
3379 break;
3380 case SRCONFLICT:
3381 case RRCONFLICT:
3382 fprintf(fp,"%*s reduce %-7d ** Parsing conflict **",
3383 indent,ap->sp->name,ap->x.rp->iRule);
3384 break;
3385 case SSCONFLICT:
3386 fprintf(fp,"%*s shift %-7d ** Parsing conflict **",
3387 indent,ap->sp->name,ap->x.stp->statenum);
3388 break;
3389 case SH_RESOLVED:
3390 if( showPrecedenceConflict ){
3391 fprintf(fp,"%*s shift %-7d -- dropped by precedence",
3392 indent,ap->sp->name,ap->x.stp->statenum);
3393 }else{
3394 result = 0;
3396 break;
3397 case RD_RESOLVED:
3398 if( showPrecedenceConflict ){
3399 fprintf(fp,"%*s reduce %-7d -- dropped by precedence",
3400 indent,ap->sp->name,ap->x.rp->iRule);
3401 }else{
3402 result = 0;
3404 break;
3405 case NOT_USED:
3406 result = 0;
3407 break;
3409 if( result && ap->spOpt ){
3410 fprintf(fp," /* because %s==%s */", ap->sp->name, ap->spOpt->name);
3412 return result;
3415 /* Generate the "*.out" log file */
3416 void ReportOutput(struct lemon *lemp)
3418 int i, n;
3419 struct state *stp;
3420 struct config *cfp;
3421 struct action *ap;
3422 struct rule *rp;
3423 FILE *fp;
3425 fp = file_open(lemp,".out","wb");
3426 if( fp==0 ) return;
3427 for(i=0; i<lemp->nxstate; i++){
3428 stp = lemp->sorted[i];
3429 fprintf(fp,"State %d:\n",stp->statenum);
3430 if( lemp->basisflag ) cfp=stp->bp;
3431 else cfp=stp->cfp;
3432 while( cfp ){
3433 char buf[20];
3434 if( cfp->dot==cfp->rp->nrhs ){
3435 lemon_sprintf(buf,"(%d)",cfp->rp->iRule);
3436 fprintf(fp," %5s ",buf);
3437 }else{
3438 fprintf(fp," ");
3440 ConfigPrint(fp,cfp);
3441 fprintf(fp,"\n");
3442 #if 0
3443 SetPrint(fp,cfp->fws,lemp);
3444 PlinkPrint(fp,cfp->fplp,"To ");
3445 PlinkPrint(fp,cfp->bplp,"From");
3446 #endif
3447 if( lemp->basisflag ) cfp=cfp->bp;
3448 else cfp=cfp->next;
3450 fprintf(fp,"\n");
3451 for(ap=stp->ap; ap; ap=ap->next){
3452 if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
3454 fprintf(fp,"\n");
3456 fprintf(fp, "----------------------------------------------------\n");
3457 fprintf(fp, "Symbols:\n");
3458 fprintf(fp, "The first-set of non-terminals is shown after the name.\n\n");
3459 for(i=0; i<lemp->nsymbol; i++){
3460 int j;
3461 struct symbol *sp;
3463 sp = lemp->symbols[i];
3464 fprintf(fp, " %3d: %s", i, sp->name);
3465 if( sp->type==NONTERMINAL ){
3466 fprintf(fp, ":");
3467 if( sp->lambda ){
3468 fprintf(fp, " <lambda>");
3470 for(j=0; j<lemp->nterminal; j++){
3471 if( sp->firstset && SetFind(sp->firstset, j) ){
3472 fprintf(fp, " %s", lemp->symbols[j]->name);
3476 if( sp->prec>=0 ) fprintf(fp," (precedence=%d)", sp->prec);
3477 fprintf(fp, "\n");
3479 fprintf(fp, "----------------------------------------------------\n");
3480 fprintf(fp, "Syntax-only Symbols:\n");
3481 fprintf(fp, "The following symbols never carry semantic content.\n\n");
3482 for(i=n=0; i<lemp->nsymbol; i++){
3483 int w;
3484 struct symbol *sp = lemp->symbols[i];
3485 if( sp->bContent ) continue;
3486 w = (int)strlen(sp->name);
3487 if( n>0 && n+w>75 ){
3488 fprintf(fp,"\n");
3489 n = 0;
3491 if( n>0 ){
3492 fprintf(fp, " ");
3493 n++;
3495 fprintf(fp, "%s", sp->name);
3496 n += w;
3498 if( n>0 ) fprintf(fp, "\n");
3499 fprintf(fp, "----------------------------------------------------\n");
3500 fprintf(fp, "Rules:\n");
3501 for(rp=lemp->rule; rp; rp=rp->next){
3502 fprintf(fp, "%4d: ", rp->iRule);
3503 rule_print(fp, rp);
3504 fprintf(fp,".");
3505 if( rp->precsym ){
3506 fprintf(fp," [%s precedence=%d]",
3507 rp->precsym->name, rp->precsym->prec);
3509 fprintf(fp,"\n");
3511 fclose(fp);
3512 return;
3515 /* Search for the file "name" which is in the same directory as
3516 ** the executable */
3517 PRIVATE char *pathsearch(char *argv0, char *name, int modemask)
3519 const char *pathlist;
3520 char *pathbufptr = 0;
3521 char *pathbuf = 0;
3522 char *path,*cp;
3523 char c;
3525 #ifdef __WIN32__
3526 cp = strrchr(argv0,'\\');
3527 #else
3528 cp = strrchr(argv0,'/');
3529 #endif
3530 if( cp ){
3531 c = *cp;
3532 *cp = 0;
3533 path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 );
3534 if( path ) lemon_sprintf(path,"%s/%s",argv0,name);
3535 *cp = c;
3536 }else{
3537 pathlist = getenv("PATH");
3538 if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
3539 pathbuf = (char *) malloc( lemonStrlen(pathlist) + 1 );
3540 path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 );
3541 if( (pathbuf != 0) && (path!=0) ){
3542 pathbufptr = pathbuf;
3543 lemon_strcpy(pathbuf, pathlist);
3544 while( *pathbuf ){
3545 cp = strchr(pathbuf,':');
3546 if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)];
3547 c = *cp;
3548 *cp = 0;
3549 lemon_sprintf(path,"%s/%s",pathbuf,name);
3550 *cp = c;
3551 if( c==0 ) pathbuf[0] = 0;
3552 else pathbuf = &cp[1];
3553 if( access(path,modemask)==0 ) break;
3556 free(pathbufptr);
3558 return path;
3561 /* Given an action, compute the integer value for that action
3562 ** which is to be put in the action table of the generated machine.
3563 ** Return negative if no action should be generated.
3565 PRIVATE int compute_action(struct lemon *lemp, struct action *ap)
3567 int act;
3568 switch( ap->type ){
3569 case SHIFT: act = ap->x.stp->statenum; break;
3570 case SHIFTREDUCE: {
3571 /* Since a SHIFT is inherient after a prior REDUCE, convert any
3572 ** SHIFTREDUCE action with a nonterminal on the LHS into a simple
3573 ** REDUCE action: */
3574 if( ap->sp->index>=lemp->nterminal ){
3575 act = lemp->minReduce + ap->x.rp->iRule;
3576 }else{
3577 act = lemp->minShiftReduce + ap->x.rp->iRule;
3579 break;
3581 case REDUCE: act = lemp->minReduce + ap->x.rp->iRule; break;
3582 case ERROR: act = lemp->errAction; break;
3583 case ACCEPT: act = lemp->accAction; break;
3584 default: act = -1; break;
3586 return act;
3589 #define LINESIZE 1000
3590 /* The next cluster of routines are for reading the template file
3591 ** and writing the results to the generated parser */
3592 /* The first function transfers data from "in" to "out" until
3593 ** a line is seen which begins with "%%". The line number is
3594 ** tracked.
3596 ** if name!=0, then any word that begin with "Parse" is changed to
3597 ** begin with *name instead.
3599 PRIVATE void tplt_xfer(char *name, FILE *in, FILE *out, int *lineno)
3601 int i, iStart;
3602 char line[LINESIZE];
3603 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3604 (*lineno)++;
3605 iStart = 0;
3606 if( name ){
3607 for(i=0; line[i]; i++){
3608 if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
3609 && (i==0 || !ISALPHA(line[i-1]))
3611 if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
3612 fprintf(out,"%s",name);
3613 i += 4;
3614 iStart = i+1;
3618 fprintf(out,"%s",&line[iStart]);
3622 /* Skip forward past the header of the template file to the first "%%"
3624 PRIVATE void tplt_skip_header(FILE *in, int *lineno)
3626 char line[LINESIZE];
3627 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3628 (*lineno)++;
3632 /* The next function finds the template file and opens it, returning
3633 ** a pointer to the opened file. */
3634 PRIVATE FILE *tplt_open(struct lemon *lemp)
3636 static char templatename[] = "lempar.c";
3637 char buf[1000];
3638 FILE *in;
3639 char *tpltname;
3640 char *toFree = 0;
3641 char *cp;
3643 /* first, see if user specified a template filename on the command line. */
3644 if (user_templatename != 0) {
3645 if( access(user_templatename,004)==-1 ){
3646 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3647 user_templatename);
3648 lemp->errorcnt++;
3649 return 0;
3651 in = fopen(user_templatename,"rb");
3652 if( in==0 ){
3653 fprintf(stderr,"Can't open the template file \"%s\".\n",
3654 user_templatename);
3655 lemp->errorcnt++;
3656 return 0;
3658 return in;
3661 cp = strrchr(lemp->filename,'.');
3662 if( cp ){
3663 lemon_sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
3664 }else{
3665 lemon_sprintf(buf,"%s.lt",lemp->filename);
3667 if( access(buf,004)==0 ){
3668 tpltname = buf;
3669 }else if( access(templatename,004)==0 ){
3670 tpltname = templatename;
3671 }else{
3672 toFree = tpltname = pathsearch(lemp->argv0,templatename,0);
3674 if( tpltname==0 ){
3675 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3676 templatename);
3677 lemp->errorcnt++;
3678 return 0;
3680 in = fopen(tpltname,"rb");
3681 if( in==0 ){
3682 fprintf(stderr,"Can't open the template file \"%s\".\n",tpltname);
3683 lemp->errorcnt++;
3685 free(toFree);
3686 return in;
3689 /* Print a #line directive line to the output file. */
3690 PRIVATE void tplt_linedir(FILE *out, int lineno, char *filename)
3692 fprintf(out,"#line %d \"",lineno);
3693 while( *filename ){
3694 if( *filename == '\\' ) putc('\\',out);
3695 putc(*filename,out);
3696 filename++;
3698 fprintf(out,"\"\n");
3701 /* Print a string to the file and keep the linenumber up to date */
3702 PRIVATE void tplt_print(FILE *out, struct lemon *lemp, char *str, int *lineno)
3704 if( str==0 ) return;
3705 while( *str ){
3706 putc(*str,out);
3707 if( *str=='\n' ) (*lineno)++;
3708 str++;
3710 if( str[-1]!='\n' ){
3711 putc('\n',out);
3712 (*lineno)++;
3714 if (!lemp->nolinenosflag) {
3715 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3717 return;
3721 ** The following routine emits code for the destructor for the
3722 ** symbol sp
3724 void emit_destructor_code(
3725 FILE *out,
3726 struct symbol *sp,
3727 struct lemon *lemp,
3728 int *lineno
3730 char *cp = 0;
3732 if( sp->type==TERMINAL ){
3733 cp = lemp->tokendest;
3734 if( cp==0 ) return;
3735 fprintf(out,"{\n"); (*lineno)++;
3736 }else if( sp->destructor ){
3737 cp = sp->destructor;
3738 fprintf(out,"{\n"); (*lineno)++;
3739 if( !lemp->nolinenosflag ){
3740 (*lineno)++;
3741 tplt_linedir(out,sp->destLineno,lemp->filename);
3743 }else if( lemp->vardest ){
3744 cp = lemp->vardest;
3745 if( cp==0 ) return;
3746 fprintf(out,"{\n"); (*lineno)++;
3747 }else{
3748 assert( 0 ); /* Cannot happen */
3750 for(; *cp; cp++){
3751 if( *cp=='$' && cp[1]=='$' ){
3752 fprintf(out,"(yypminor->yy%d)",sp->dtnum);
3753 cp++;
3754 continue;
3756 if( *cp=='\n' ) (*lineno)++;
3757 fputc(*cp,out);
3759 fprintf(out,"\n"); (*lineno)++;
3760 if (!lemp->nolinenosflag) {
3761 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3763 fprintf(out,"}\n"); (*lineno)++;
3764 return;
3768 ** Return TRUE (non-zero) if the given symbol has a destructor.
3770 int has_destructor(struct symbol *sp, struct lemon *lemp)
3772 int ret;
3773 if( sp->type==TERMINAL ){
3774 ret = lemp->tokendest!=0;
3775 }else{
3776 ret = lemp->vardest!=0 || sp->destructor!=0;
3778 return ret;
3782 ** Append text to a dynamically allocated string. If zText is 0 then
3783 ** reset the string to be empty again. Always return the complete text
3784 ** of the string (which is overwritten with each call).
3786 ** n bytes of zText are stored. If n==0 then all of zText up to the first
3787 ** \000 terminator is stored. zText can contain up to two instances of
3788 ** %d. The values of p1 and p2 are written into the first and second
3789 ** %d.
3791 ** If n==-1, then the previous character is overwritten.
3793 PRIVATE char *append_str(const char *zText, int n, int p1, int p2){
3794 static char empty[1] = { 0 };
3795 static char *z = 0;
3796 static int alloced = 0;
3797 static int used = 0;
3798 int c;
3799 char zInt[40];
3800 if( zText==0 ){
3801 if( used==0 && z!=0 ) z[0] = 0;
3802 used = 0;
3803 return z;
3805 if( n<=0 ){
3806 if( n<0 ){
3807 used += n;
3808 assert( used>=0 );
3810 n = lemonStrlen(zText);
3812 if( (int) (n+sizeof(zInt)*2+used) >= alloced ){
3813 alloced = n + sizeof(zInt)*2 + used + 200;
3814 z = (char *) realloc(z, alloced);
3816 if( z==0 ) return empty;
3817 while( n-- > 0 ){
3818 c = *(zText++);
3819 if( c=='%' && n>0 && zText[0]=='d' ){
3820 lemon_sprintf(zInt, "%d", p1);
3821 p1 = p2;
3822 lemon_strcpy(&z[used], zInt);
3823 used += lemonStrlen(&z[used]);
3824 zText++;
3825 n--;
3826 }else{
3827 z[used++] = (char)c;
3830 z[used] = 0;
3831 return z;
3835 ** Write and transform the rp->code string so that symbols are expanded.
3836 ** Populate the rp->codePrefix and rp->codeSuffix strings, as appropriate.
3838 ** Return 1 if the expanded code requires that "yylhsminor" local variable
3839 ** to be defined.
3841 PRIVATE int translate_code(struct lemon *lemp, struct rule *rp){
3842 char *cp, *xp;
3843 int i;
3844 int rc = 0; /* True if yylhsminor is used */
3845 int dontUseRhs0 = 0; /* If true, use of left-most RHS label is illegal */
3846 const char *zSkip = 0; /* The zOvwrt comment within rp->code, or NULL */
3847 char lhsused = 0; /* True if the LHS element has been used */
3848 char lhsdirect; /* True if LHS writes directly into stack */
3849 char used[MAXRHS]; /* True for each RHS element which is used */
3850 char zLhs[50]; /* Convert the LHS symbol into this string */
3851 char zOvwrt[900]; /* Comment that to allow LHS to overwrite RHS */
3853 for(i=0; i<rp->nrhs; i++) used[i] = 0;
3854 lhsused = 0;
3856 if( rp->code==0 ){
3857 static char newlinestr[2] = { '\n', '\0' };
3858 rp->code = newlinestr;
3859 rp->line = rp->ruleline;
3860 rp->noCode = 1;
3861 }else{
3862 rp->noCode = 0;
3866 if( rp->nrhs==0 ){
3867 /* If there are no RHS symbols, then writing directly to the LHS is ok */
3868 lhsdirect = 1;
3869 }else if( rp->rhsalias[0]==0 ){
3870 /* The left-most RHS symbol has no value. LHS direct is ok. But
3871 ** we have to call the destructor on the RHS symbol first. */
3872 lhsdirect = 1;
3873 if( has_destructor(rp->rhs[0],lemp) ){
3874 append_str(0,0,0,0);
3875 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3876 rp->rhs[0]->index,1-rp->nrhs);
3877 rp->codePrefix = Strsafe(append_str(0,0,0,0));
3878 rp->noCode = 0;
3880 }else if( rp->lhsalias==0 ){
3881 /* There is no LHS value symbol. */
3882 lhsdirect = 1;
3883 }else if( strcmp(rp->lhsalias,rp->rhsalias[0])==0 ){
3884 /* The LHS symbol and the left-most RHS symbol are the same, so
3885 ** direct writing is allowed */
3886 lhsdirect = 1;
3887 lhsused = 1;
3888 used[0] = 1;
3889 if( rp->lhs->dtnum!=rp->rhs[0]->dtnum ){
3890 ErrorMsg(lemp->filename,rp->ruleline,
3891 "%s(%s) and %s(%s) share the same label but have "
3892 "different datatypes.",
3893 rp->lhs->name, rp->lhsalias, rp->rhs[0]->name, rp->rhsalias[0]);
3894 lemp->errorcnt++;
3896 }else{
3897 lemon_sprintf(zOvwrt, "/*%s-overwrites-%s*/",
3898 rp->lhsalias, rp->rhsalias[0]);
3899 zSkip = strstr(rp->code, zOvwrt);
3900 if( zSkip!=0 ){
3901 /* The code contains a special comment that indicates that it is safe
3902 ** for the LHS label to overwrite left-most RHS label. */
3903 lhsdirect = 1;
3904 }else{
3905 lhsdirect = 0;
3908 if( lhsdirect ){
3909 sprintf(zLhs, "yymsp[%d].minor.yy%d",1-rp->nrhs,rp->lhs->dtnum);
3910 }else{
3911 rc = 1;
3912 sprintf(zLhs, "yylhsminor.yy%d",rp->lhs->dtnum);
3915 append_str(0,0,0,0);
3917 /* This const cast is wrong but harmless, if we're careful. */
3918 for(cp=(char *)rp->code; *cp; cp++){
3919 if( cp==zSkip ){
3920 append_str(zOvwrt,0,0,0);
3921 cp += lemonStrlen(zOvwrt)-1;
3922 dontUseRhs0 = 1;
3923 continue;
3925 if( ISALPHA(*cp) && (cp==rp->code || (!ISALNUM(cp[-1]) && cp[-1]!='_')) ){
3926 char saved;
3927 for(xp= &cp[1]; ISALNUM(*xp) || *xp=='_'; xp++);
3928 saved = *xp;
3929 *xp = 0;
3930 if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
3931 append_str(zLhs,0,0,0);
3932 cp = xp;
3933 lhsused = 1;
3934 }else{
3935 for(i=0; i<rp->nrhs; i++){
3936 if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
3937 if( i==0 && dontUseRhs0 ){
3938 ErrorMsg(lemp->filename,rp->ruleline,
3939 "Label %s used after '%s'.",
3940 rp->rhsalias[0], zOvwrt);
3941 lemp->errorcnt++;
3942 }else if( cp!=rp->code && cp[-1]=='@' ){
3943 /* If the argument is of the form @X then substituted
3944 ** the token number of X, not the value of X */
3945 append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
3946 }else{
3947 struct symbol *sp = rp->rhs[i];
3948 int dtnum;
3949 if( sp->type==MULTITERMINAL ){
3950 dtnum = sp->subsym[0]->dtnum;
3951 }else{
3952 dtnum = sp->dtnum;
3954 append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);
3956 cp = xp;
3957 used[i] = 1;
3958 break;
3962 *xp = saved;
3964 append_str(cp, 1, 0, 0);
3965 } /* End loop */
3967 /* Main code generation completed */
3968 cp = append_str(0,0,0,0);
3969 if( cp && cp[0] ) rp->code = Strsafe(cp);
3970 append_str(0,0,0,0);
3972 /* Check to make sure the LHS has been used */
3973 if( rp->lhsalias && !lhsused ){
3974 ErrorMsg(lemp->filename,rp->ruleline,
3975 "Label \"%s\" for \"%s(%s)\" is never used.",
3976 rp->lhsalias,rp->lhs->name,rp->lhsalias);
3977 lemp->errorcnt++;
3980 /* Generate destructor code for RHS minor values which are not referenced.
3981 ** Generate error messages for unused labels and duplicate labels.
3983 for(i=0; i<rp->nrhs; i++){
3984 if( rp->rhsalias[i] ){
3985 if( i>0 ){
3986 int j;
3987 if( rp->lhsalias && strcmp(rp->lhsalias,rp->rhsalias[i])==0 ){
3988 ErrorMsg(lemp->filename,rp->ruleline,
3989 "%s(%s) has the same label as the LHS but is not the left-most "
3990 "symbol on the RHS.",
3991 rp->rhs[i]->name, rp->rhsalias[i]);
3992 lemp->errorcnt++;
3994 for(j=0; j<i; j++){
3995 if( rp->rhsalias[j] && strcmp(rp->rhsalias[j],rp->rhsalias[i])==0 ){
3996 ErrorMsg(lemp->filename,rp->ruleline,
3997 "Label %s used for multiple symbols on the RHS of a rule.",
3998 rp->rhsalias[i]);
3999 lemp->errorcnt++;
4000 break;
4004 if( !used[i] ){
4005 ErrorMsg(lemp->filename,rp->ruleline,
4006 "Label %s for \"%s(%s)\" is never used.",
4007 rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
4008 lemp->errorcnt++;
4010 }else if( i>0 && has_destructor(rp->rhs[i],lemp) ){
4011 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
4012 rp->rhs[i]->index,i-rp->nrhs+1);
4016 /* If unable to write LHS values directly into the stack, write the
4017 ** saved LHS value now. */
4018 if( lhsdirect==0 ){
4019 append_str(" yymsp[%d].minor.yy%d = ", 0, 1-rp->nrhs, rp->lhs->dtnum);
4020 append_str(zLhs, 0, 0, 0);
4021 append_str(";\n", 0, 0, 0);
4024 /* Suffix code generation complete */
4025 cp = append_str(0,0,0,0);
4026 if( cp && cp[0] ){
4027 rp->codeSuffix = Strsafe(cp);
4028 rp->noCode = 0;
4031 return rc;
4035 ** Generate code which executes when the rule "rp" is reduced. Write
4036 ** the code to "out". Make sure lineno stays up-to-date.
4038 PRIVATE void emit_code(
4039 FILE *out,
4040 struct rule *rp,
4041 struct lemon *lemp,
4042 int *lineno
4044 const char *cp;
4046 /* Setup code prior to the #line directive */
4047 if( rp->codePrefix && rp->codePrefix[0] ){
4048 fprintf(out, "{%s", rp->codePrefix);
4049 for(cp=rp->codePrefix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4052 /* Generate code to do the reduce action */
4053 if( rp->code ){
4054 if( !lemp->nolinenosflag ){
4055 (*lineno)++;
4056 tplt_linedir(out,rp->line,lemp->filename);
4058 fprintf(out,"{%s",rp->code);
4059 for(cp=rp->code; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4060 fprintf(out,"}\n"); (*lineno)++;
4061 if( !lemp->nolinenosflag ){
4062 (*lineno)++;
4063 tplt_linedir(out,*lineno,lemp->outname);
4067 /* Generate breakdown code that occurs after the #line directive */
4068 if( rp->codeSuffix && rp->codeSuffix[0] ){
4069 fprintf(out, "%s", rp->codeSuffix);
4070 for(cp=rp->codeSuffix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4073 if( rp->codePrefix ){
4074 fprintf(out, "}\n"); (*lineno)++;
4077 return;
4081 ** Print the definition of the union used for the parser's data stack.
4082 ** This union contains fields for every possible data type for tokens
4083 ** and nonterminals. In the process of computing and printing this
4084 ** union, also set the ".dtnum" field of every terminal and nonterminal
4085 ** symbol.
4087 void print_stack_union(
4088 FILE *out, /* The output stream */
4089 struct lemon *lemp, /* The main info structure for this parser */
4090 int *plineno, /* Pointer to the line number */
4091 int mhflag /* True if generating makeheaders output */
4093 int lineno = *plineno; /* The line number of the output */
4094 char **types; /* A hash table of datatypes */
4095 int arraysize; /* Size of the "types" array */
4096 int maxdtlength; /* Maximum length of any ".datatype" field. */
4097 char *stddt; /* Standardized name for a datatype */
4098 int i,j; /* Loop counters */
4099 unsigned hash; /* For hashing the name of a type */
4100 const char *name; /* Name of the parser */
4102 /* Allocate and initialize types[] and allocate stddt[] */
4103 arraysize = lemp->nsymbol * 2;
4104 types = (char**)calloc( arraysize, sizeof(char*) );
4105 if( types==0 ){
4106 fprintf(stderr,"Out of memory.\n");
4107 exit(1);
4109 for(i=0; i<arraysize; i++) types[i] = 0;
4110 maxdtlength = 0;
4111 if( lemp->vartype ){
4112 maxdtlength = lemonStrlen(lemp->vartype);
4114 for(i=0; i<lemp->nsymbol; i++){
4115 int len;
4116 struct symbol *sp = lemp->symbols[i];
4117 if( sp->datatype==0 ) continue;
4118 len = lemonStrlen(sp->datatype);
4119 if( len>maxdtlength ) maxdtlength = len;
4121 stddt = (char*)malloc( maxdtlength*2 + 1 );
4122 if( stddt==0 ){
4123 fprintf(stderr,"Out of memory.\n");
4124 exit(1);
4127 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
4128 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
4129 ** used for terminal symbols. If there is no %default_type defined then
4130 ** 0 is also used as the .dtnum value for nonterminals which do not specify
4131 ** a datatype using the %type directive.
4133 for(i=0; i<lemp->nsymbol; i++){
4134 struct symbol *sp = lemp->symbols[i];
4135 char *cp;
4136 if( sp==lemp->errsym ){
4137 sp->dtnum = arraysize+1;
4138 continue;
4140 if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
4141 sp->dtnum = 0;
4142 continue;
4144 cp = sp->datatype;
4145 if( cp==0 ) cp = lemp->vartype;
4146 j = 0;
4147 while( ISSPACE(*cp) ) cp++;
4148 while( *cp ) stddt[j++] = *cp++;
4149 while( j>0 && ISSPACE(stddt[j-1]) ) j--;
4150 stddt[j] = 0;
4151 if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){
4152 sp->dtnum = 0;
4153 continue;
4155 hash = 0;
4156 for(j=0; stddt[j]; j++){
4157 hash = hash*53 + stddt[j];
4159 hash = (hash & 0x7fffffff)%arraysize;
4160 while( types[hash] ){
4161 if( strcmp(types[hash],stddt)==0 ){
4162 sp->dtnum = hash + 1;
4163 break;
4165 hash++;
4166 if( hash>=(unsigned)arraysize ) hash = 0;
4168 if( types[hash]==0 ){
4169 sp->dtnum = hash + 1;
4170 types[hash] = (char*)malloc( lemonStrlen(stddt)+1 );
4171 if( types[hash]==0 ){
4172 fprintf(stderr,"Out of memory.\n");
4173 exit(1);
4175 lemon_strcpy(types[hash],stddt);
4179 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
4180 name = lemp->name ? lemp->name : "Parse";
4181 lineno = *plineno;
4182 if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
4183 fprintf(out,"#define %sTOKENTYPE %s\n",name,
4184 lemp->tokentype?lemp->tokentype:"void*"); lineno++;
4185 if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
4186 fprintf(out,"typedef union {\n"); lineno++;
4187 fprintf(out," int yyinit;\n"); lineno++;
4188 fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++;
4189 for(i=0; i<arraysize; i++){
4190 if( types[i]==0 ) continue;
4191 fprintf(out," %s yy%d;\n",types[i],i+1); lineno++;
4192 free(types[i]);
4194 if( lemp->errsym && lemp->errsym->useCnt ){
4195 fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++;
4197 free(stddt);
4198 free(types);
4199 fprintf(out,"} YYMINORTYPE;\n"); lineno++;
4200 *plineno = lineno;
4204 ** Return the name of a C datatype able to represent values between
4205 ** lwr and upr, inclusive. If pnByte!=NULL then also write the sizeof
4206 ** for that type (1, 2, or 4) into *pnByte.
4208 static const char *minimum_size_type(int lwr, int upr, int *pnByte){
4209 const char *zType = "int";
4210 int nByte = 4;
4211 if( lwr>=0 ){
4212 if( upr<=255 ){
4213 zType = "unsigned char";
4214 nByte = 1;
4215 }else if( upr<65535 ){
4216 zType = "unsigned short int";
4217 nByte = 2;
4218 }else{
4219 zType = "unsigned int";
4220 nByte = 4;
4222 }else if( lwr>=-127 && upr<=127 ){
4223 zType = "signed char";
4224 nByte = 1;
4225 }else if( lwr>=-32767 && upr<32767 ){
4226 zType = "short";
4227 nByte = 2;
4229 if( pnByte ) *pnByte = nByte;
4230 return zType;
4234 ** Each state contains a set of token transaction and a set of
4235 ** nonterminal transactions. Each of these sets makes an instance
4236 ** of the following structure. An array of these structures is used
4237 ** to order the creation of entries in the yy_action[] table.
4239 struct axset {
4240 struct state *stp; /* A pointer to a state */
4241 int isTkn; /* True to use tokens. False for non-terminals */
4242 int nAction; /* Number of actions */
4243 int iOrder; /* Original order of action sets */
4247 ** Compare to axset structures for sorting purposes
4249 static int axset_compare(const void *a, const void *b){
4250 struct axset *p1 = (struct axset*)a;
4251 struct axset *p2 = (struct axset*)b;
4252 int c;
4253 c = p2->nAction - p1->nAction;
4254 if( c==0 ){
4255 c = p1->iOrder - p2->iOrder;
4257 assert( c!=0 || p1==p2 );
4258 return c;
4262 ** Write text on "out" that describes the rule "rp".
4264 static void writeRuleText(FILE *out, struct rule *rp){
4265 int j;
4266 fprintf(out,"%s ::=", rp->lhs->name);
4267 for(j=0; j<rp->nrhs; j++){
4268 struct symbol *sp = rp->rhs[j];
4269 if( sp->type!=MULTITERMINAL ){
4270 fprintf(out," %s", sp->name);
4271 }else{
4272 int k;
4273 fprintf(out," %s", sp->subsym[0]->name);
4274 for(k=1; k<sp->nsubsym; k++){
4275 fprintf(out,"|%s",sp->subsym[k]->name);
4282 /* Generate C source code for the parser */
4283 void ReportTable(
4284 struct lemon *lemp,
4285 int mhflag, /* Output in makeheaders format if true */
4286 int sqlFlag /* Generate the *.sql file too */
4288 FILE *out, *in, *sql;
4289 char line[LINESIZE];
4290 int lineno;
4291 struct state *stp;
4292 struct action *ap;
4293 struct rule *rp;
4294 struct acttab *pActtab;
4295 int i, j, n, sz;
4296 int nLookAhead;
4297 int szActionType; /* sizeof(YYACTIONTYPE) */
4298 int szCodeType; /* sizeof(YYCODETYPE) */
4299 const char *name;
4300 int mnTknOfst, mxTknOfst;
4301 int mnNtOfst, mxNtOfst;
4302 struct axset *ax;
4303 char *prefix;
4305 lemp->minShiftReduce = lemp->nstate;
4306 lemp->errAction = lemp->minShiftReduce + lemp->nrule;
4307 lemp->accAction = lemp->errAction + 1;
4308 lemp->noAction = lemp->accAction + 1;
4309 lemp->minReduce = lemp->noAction + 1;
4310 lemp->maxAction = lemp->minReduce + lemp->nrule;
4312 in = tplt_open(lemp);
4313 if( in==0 ) return;
4314 out = file_open(lemp,".c","wb");
4315 if( out==0 ){
4316 fclose(in);
4317 return;
4319 if( sqlFlag==0 ){
4320 sql = 0;
4321 }else{
4322 sql = file_open(lemp, ".sql", "wb");
4323 if( sql==0 ){
4324 fclose(in);
4325 fclose(out);
4326 return;
4328 fprintf(sql,
4329 "BEGIN;\n"
4330 "CREATE TABLE symbol(\n"
4331 " id INTEGER PRIMARY KEY,\n"
4332 " name TEXT NOT NULL,\n"
4333 " isTerminal BOOLEAN NOT NULL,\n"
4334 " fallback INTEGER REFERENCES symbol"
4335 " DEFERRABLE INITIALLY DEFERRED\n"
4336 ");\n"
4338 for(i=0; i<lemp->nsymbol; i++){
4339 fprintf(sql,
4340 "INSERT INTO symbol(id,name,isTerminal,fallback)"
4341 "VALUES(%d,'%s',%s",
4342 i, lemp->symbols[i]->name,
4343 i<lemp->nterminal ? "TRUE" : "FALSE"
4345 if( lemp->symbols[i]->fallback ){
4346 fprintf(sql, ",%d);\n", lemp->symbols[i]->fallback->index);
4347 }else{
4348 fprintf(sql, ",NULL);\n");
4351 fprintf(sql,
4352 "CREATE TABLE rule(\n"
4353 " ruleid INTEGER PRIMARY KEY,\n"
4354 " lhs INTEGER REFERENCES symbol(id),\n"
4355 " txt TEXT\n"
4356 ");\n"
4357 "CREATE TABLE rulerhs(\n"
4358 " ruleid INTEGER REFERENCES rule(ruleid),\n"
4359 " pos INTEGER,\n"
4360 " sym INTEGER REFERENCES symbol(id)\n"
4361 ");\n"
4363 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4364 assert( i==rp->iRule );
4365 fprintf(sql,
4366 "INSERT INTO rule(ruleid,lhs,txt)VALUES(%d,%d,'",
4367 rp->iRule, rp->lhs->index
4369 writeRuleText(sql, rp);
4370 fprintf(sql,"');\n");
4371 for(j=0; j<rp->nrhs; j++){
4372 struct symbol *sp = rp->rhs[j];
4373 if( sp->type!=MULTITERMINAL ){
4374 fprintf(sql,
4375 "INSERT INTO rulerhs(ruleid,pos,sym)VALUES(%d,%d,%d);\n",
4376 i,j,sp->index
4378 }else{
4379 int k;
4380 for(k=0; k<sp->nsubsym; k++){
4381 fprintf(sql,
4382 "INSERT INTO rulerhs(ruleid,pos,sym)VALUES(%d,%d,%d);\n",
4383 i,j,sp->subsym[k]->index
4389 fprintf(sql, "COMMIT;\n");
4391 lineno = 1;
4393 fprintf(out,
4394 "/* This file is automatically generated by Lemon from input grammar\n"
4395 "** source file \"%s\". */\n", lemp->filename); lineno += 2;
4397 /* The first %include directive begins with a C-language comment,
4398 ** then skip over the header comment of the template file
4400 if( lemp->include==0 ) lemp->include = "";
4401 for(i=0; ISSPACE(lemp->include[i]); i++){
4402 if( lemp->include[i]=='\n' ){
4403 lemp->include += i+1;
4404 i = -1;
4407 if( lemp->include[0]=='/' ){
4408 tplt_skip_header(in,&lineno);
4409 }else{
4410 tplt_xfer(lemp->name,in,out,&lineno);
4413 /* Generate the include code, if any */
4414 tplt_print(out,lemp,lemp->include,&lineno);
4415 if( mhflag ){
4416 char *incName = file_makename(lemp, ".h");
4417 fprintf(out,"#include \"%s\"\n", incName); lineno++;
4418 free(incName);
4420 tplt_xfer(lemp->name,in,out,&lineno);
4422 /* Generate #defines for all tokens */
4423 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4424 else prefix = "";
4425 if( mhflag ){
4426 fprintf(out,"#if INTERFACE\n"); lineno++;
4427 }else{
4428 fprintf(out,"#ifndef %s%s\n", prefix, lemp->symbols[1]->name);
4430 for(i=1; i<lemp->nterminal; i++){
4431 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
4432 lineno++;
4434 fprintf(out,"#endif\n"); lineno++;
4435 tplt_xfer(lemp->name,in,out,&lineno);
4437 /* Generate the defines */
4438 fprintf(out,"#define YYCODETYPE %s\n",
4439 minimum_size_type(0, lemp->nsymbol, &szCodeType)); lineno++;
4440 fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol); lineno++;
4441 fprintf(out,"#define YYACTIONTYPE %s\n",
4442 minimum_size_type(0,lemp->maxAction,&szActionType)); lineno++;
4443 if( lemp->wildcard ){
4444 fprintf(out,"#define YYWILDCARD %d\n",
4445 lemp->wildcard->index); lineno++;
4447 print_stack_union(out,lemp,&lineno,mhflag);
4448 fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++;
4449 if( lemp->stacksize ){
4450 fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++;
4451 }else{
4452 fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++;
4454 fprintf(out, "#endif\n"); lineno++;
4455 if( mhflag ){
4456 fprintf(out,"#if INTERFACE\n"); lineno++;
4458 name = lemp->name ? lemp->name : "Parse";
4459 if( lemp->arg && lemp->arg[0] ){
4460 i = lemonStrlen(lemp->arg);
4461 while( i>=1 && ISSPACE(lemp->arg[i-1]) ) i--;
4462 while( i>=1 && (ISALNUM(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
4463 fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++;
4464 fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++;
4465 fprintf(out,"#define %sARG_PARAM ,%s\n",name,&lemp->arg[i]); lineno++;
4466 fprintf(out,"#define %sARG_FETCH %s=yypParser->%s;\n",
4467 name,lemp->arg,&lemp->arg[i]); lineno++;
4468 fprintf(out,"#define %sARG_STORE yypParser->%s=%s;\n",
4469 name,&lemp->arg[i],&lemp->arg[i]); lineno++;
4470 }else{
4471 fprintf(out,"#define %sARG_SDECL\n",name); lineno++;
4472 fprintf(out,"#define %sARG_PDECL\n",name); lineno++;
4473 fprintf(out,"#define %sARG_PARAM\n",name); lineno++;
4474 fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
4475 fprintf(out,"#define %sARG_STORE\n",name); lineno++;
4477 if( lemp->ctx && lemp->ctx[0] ){
4478 i = lemonStrlen(lemp->ctx);
4479 while( i>=1 && ISSPACE(lemp->ctx[i-1]) ) i--;
4480 while( i>=1 && (ISALNUM(lemp->ctx[i-1]) || lemp->ctx[i-1]=='_') ) i--;
4481 fprintf(out,"#define %sCTX_SDECL %s;\n",name,lemp->ctx); lineno++;
4482 fprintf(out,"#define %sCTX_PDECL ,%s\n",name,lemp->ctx); lineno++;
4483 fprintf(out,"#define %sCTX_PARAM ,%s\n",name,&lemp->ctx[i]); lineno++;
4484 fprintf(out,"#define %sCTX_FETCH %s=yypParser->%s;\n",
4485 name,lemp->ctx,&lemp->ctx[i]); lineno++;
4486 fprintf(out,"#define %sCTX_STORE yypParser->%s=%s;\n",
4487 name,&lemp->ctx[i],&lemp->ctx[i]); lineno++;
4488 }else{
4489 fprintf(out,"#define %sCTX_SDECL\n",name); lineno++;
4490 fprintf(out,"#define %sCTX_PDECL\n",name); lineno++;
4491 fprintf(out,"#define %sCTX_PARAM\n",name); lineno++;
4492 fprintf(out,"#define %sCTX_FETCH\n",name); lineno++;
4493 fprintf(out,"#define %sCTX_STORE\n",name); lineno++;
4495 if( mhflag ){
4496 fprintf(out,"#endif\n"); lineno++;
4498 if( lemp->errsym && lemp->errsym->useCnt ){
4499 fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++;
4500 fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++;
4502 if( lemp->has_fallback ){
4503 fprintf(out,"#define YYFALLBACK 1\n"); lineno++;
4506 /* Compute the action table, but do not output it yet. The action
4507 ** table must be computed before generating the YYNSTATE macro because
4508 ** we need to know how many states can be eliminated.
4510 ax = (struct axset *) calloc(lemp->nxstate*2, sizeof(ax[0]));
4511 if( ax==0 ){
4512 fprintf(stderr,"malloc failed\n");
4513 exit(1);
4515 for(i=0; i<lemp->nxstate; i++){
4516 stp = lemp->sorted[i];
4517 ax[i*2].stp = stp;
4518 ax[i*2].isTkn = 1;
4519 ax[i*2].nAction = stp->nTknAct;
4520 ax[i*2+1].stp = stp;
4521 ax[i*2+1].isTkn = 0;
4522 ax[i*2+1].nAction = stp->nNtAct;
4524 mxTknOfst = mnTknOfst = 0;
4525 mxNtOfst = mnNtOfst = 0;
4526 /* In an effort to minimize the action table size, use the heuristic
4527 ** of placing the largest action sets first */
4528 for(i=0; i<lemp->nxstate*2; i++) ax[i].iOrder = i;
4529 qsort(ax, lemp->nxstate*2, sizeof(ax[0]), axset_compare);
4530 pActtab = acttab_alloc(lemp->nsymbol, lemp->nterminal);
4531 for(i=0; i<lemp->nxstate*2 && ax[i].nAction>0; i++){
4532 stp = ax[i].stp;
4533 if( ax[i].isTkn ){
4534 for(ap=stp->ap; ap; ap=ap->next){
4535 int action;
4536 if( ap->sp->index>=lemp->nterminal ) continue;
4537 action = compute_action(lemp, ap);
4538 if( action<0 ) continue;
4539 acttab_action(pActtab, ap->sp->index, action);
4541 stp->iTknOfst = acttab_insert(pActtab, 1);
4542 if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
4543 if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
4544 }else{
4545 for(ap=stp->ap; ap; ap=ap->next){
4546 int action;
4547 if( ap->sp->index<lemp->nterminal ) continue;
4548 if( ap->sp->index==lemp->nsymbol ) continue;
4549 action = compute_action(lemp, ap);
4550 if( action<0 ) continue;
4551 acttab_action(pActtab, ap->sp->index, action);
4553 stp->iNtOfst = acttab_insert(pActtab, 0);
4554 if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
4555 if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
4557 #if 0 /* Uncomment for a trace of how the yy_action[] table fills out */
4558 { int jj, nn;
4559 for(jj=nn=0; jj<pActtab->nAction; jj++){
4560 if( pActtab->aAction[jj].action<0 ) nn++;
4562 printf("%4d: State %3d %s n: %2d size: %5d freespace: %d\n",
4563 i, stp->statenum, ax[i].isTkn ? "Token" : "Var ",
4564 ax[i].nAction, pActtab->nAction, nn);
4566 #endif
4568 free(ax);
4570 /* Mark rules that are actually used for reduce actions after all
4571 ** optimizations have been applied
4573 for(rp=lemp->rule; rp; rp=rp->next) rp->doesReduce = LEMON_FALSE;
4574 for(i=0; i<lemp->nxstate; i++){
4575 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
4576 if( ap->type==REDUCE || ap->type==SHIFTREDUCE ){
4577 ap->x.rp->doesReduce = 1;
4582 /* Finish rendering the constants now that the action table has
4583 ** been computed */
4584 fprintf(out,"#define YYNSTATE %d\n",lemp->nxstate); lineno++;
4585 fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++;
4586 fprintf(out,"#define YYNRULE_WITH_ACTION %d\n",lemp->nruleWithAction);
4587 lineno++;
4588 fprintf(out,"#define YYNTOKEN %d\n",lemp->nterminal); lineno++;
4589 fprintf(out,"#define YY_MAX_SHIFT %d\n",lemp->nxstate-1); lineno++;
4590 i = lemp->minShiftReduce;
4591 fprintf(out,"#define YY_MIN_SHIFTREDUCE %d\n",i); lineno++;
4592 i += lemp->nrule;
4593 fprintf(out,"#define YY_MAX_SHIFTREDUCE %d\n", i-1); lineno++;
4594 fprintf(out,"#define YY_ERROR_ACTION %d\n", lemp->errAction); lineno++;
4595 fprintf(out,"#define YY_ACCEPT_ACTION %d\n", lemp->accAction); lineno++;
4596 fprintf(out,"#define YY_NO_ACTION %d\n", lemp->noAction); lineno++;
4597 fprintf(out,"#define YY_MIN_REDUCE %d\n", lemp->minReduce); lineno++;
4598 i = lemp->minReduce + lemp->nrule;
4599 fprintf(out,"#define YY_MAX_REDUCE %d\n", i-1); lineno++;
4600 tplt_xfer(lemp->name,in,out,&lineno);
4602 /* Now output the action table and its associates:
4604 ** yy_action[] A single table containing all actions.
4605 ** yy_lookahead[] A table containing the lookahead for each entry in
4606 ** yy_action. Used to detect hash collisions.
4607 ** yy_shift_ofst[] For each state, the offset into yy_action for
4608 ** shifting terminals.
4609 ** yy_reduce_ofst[] For each state, the offset into yy_action for
4610 ** shifting non-terminals after a reduce.
4611 ** yy_default[] Default action for each state.
4614 /* Output the yy_action table */
4615 lemp->nactiontab = n = acttab_action_size(pActtab);
4616 lemp->tablesize += n*szActionType;
4617 fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++;
4618 fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++;
4619 for(i=j=0; i<n; i++){
4620 int action = acttab_yyaction(pActtab, i);
4621 if( action<0 ) action = lemp->noAction;
4622 if( j==0 ) fprintf(out," /* %5d */ ", i);
4623 fprintf(out, " %4d,", action);
4624 if( j==9 || i==n-1 ){
4625 fprintf(out, "\n"); lineno++;
4626 j = 0;
4627 }else{
4628 j++;
4631 fprintf(out, "};\n"); lineno++;
4633 /* Output the yy_lookahead table */
4634 lemp->nlookaheadtab = n = acttab_lookahead_size(pActtab);
4635 lemp->tablesize += n*szCodeType;
4636 fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++;
4637 for(i=j=0; i<n; i++){
4638 int la = acttab_yylookahead(pActtab, i);
4639 if( la<0 ) la = lemp->nsymbol;
4640 if( j==0 ) fprintf(out," /* %5d */ ", i);
4641 fprintf(out, " %4d,", la);
4642 if( j==9 ){
4643 fprintf(out, "\n"); lineno++;
4644 j = 0;
4645 }else{
4646 j++;
4649 /* Add extra entries to the end of the yy_lookahead[] table so that
4650 ** yy_shift_ofst[]+iToken will always be a valid index into the array,
4651 ** even for the largest possible value of yy_shift_ofst[] and iToken. */
4652 nLookAhead = lemp->nterminal + lemp->nactiontab;
4653 while( i<nLookAhead ){
4654 if( j==0 ) fprintf(out," /* %5d */ ", i);
4655 fprintf(out, " %4d,", lemp->nterminal);
4656 if( j==9 ){
4657 fprintf(out, "\n"); lineno++;
4658 j = 0;
4659 }else{
4660 j++;
4662 i++;
4664 if( j>0 ){ fprintf(out, "\n"); lineno++; }
4665 fprintf(out, "};\n"); lineno++;
4667 /* Output the yy_shift_ofst[] table */
4668 n = lemp->nxstate;
4669 while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--;
4670 fprintf(out, "#define YY_SHIFT_COUNT (%d)\n", n-1); lineno++;
4671 fprintf(out, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst); lineno++;
4672 fprintf(out, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst); lineno++;
4673 fprintf(out, "static const %s yy_shift_ofst[] = {\n",
4674 minimum_size_type(mnTknOfst, lemp->nterminal+lemp->nactiontab, &sz));
4675 lineno++;
4676 lemp->tablesize += n*sz;
4677 for(i=j=0; i<n; i++){
4678 int ofst;
4679 stp = lemp->sorted[i];
4680 ofst = stp->iTknOfst;
4681 if( ofst==NO_OFFSET ) ofst = lemp->nactiontab;
4682 if( j==0 ) fprintf(out," /* %5d */ ", i);
4683 fprintf(out, " %4d,", ofst);
4684 if( j==9 || i==n-1 ){
4685 fprintf(out, "\n"); lineno++;
4686 j = 0;
4687 }else{
4688 j++;
4691 fprintf(out, "};\n"); lineno++;
4693 /* Output the yy_reduce_ofst[] table */
4694 n = lemp->nxstate;
4695 while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--;
4696 fprintf(out, "#define YY_REDUCE_COUNT (%d)\n", n-1); lineno++;
4697 fprintf(out, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst); lineno++;
4698 fprintf(out, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst); lineno++;
4699 fprintf(out, "static const %s yy_reduce_ofst[] = {\n",
4700 minimum_size_type(mnNtOfst-1, mxNtOfst, &sz)); lineno++;
4701 lemp->tablesize += n*sz;
4702 for(i=j=0; i<n; i++){
4703 int ofst;
4704 stp = lemp->sorted[i];
4705 ofst = stp->iNtOfst;
4706 if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
4707 if( j==0 ) fprintf(out," /* %5d */ ", i);
4708 fprintf(out, " %4d,", ofst);
4709 if( j==9 || i==n-1 ){
4710 fprintf(out, "\n"); lineno++;
4711 j = 0;
4712 }else{
4713 j++;
4716 fprintf(out, "};\n"); lineno++;
4718 /* Output the default action table */
4719 fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++;
4720 n = lemp->nxstate;
4721 lemp->tablesize += n*szActionType;
4722 for(i=j=0; i<n; i++){
4723 stp = lemp->sorted[i];
4724 if( j==0 ) fprintf(out," /* %5d */ ", i);
4725 if( stp->iDfltReduce<0 ){
4726 fprintf(out, " %4d,", lemp->errAction);
4727 }else{
4728 fprintf(out, " %4d,", stp->iDfltReduce + lemp->minReduce);
4730 if( j==9 || i==n-1 ){
4731 fprintf(out, "\n"); lineno++;
4732 j = 0;
4733 }else{
4734 j++;
4737 fprintf(out, "};\n"); lineno++;
4738 tplt_xfer(lemp->name,in,out,&lineno);
4740 /* Generate the table of fallback tokens.
4742 if( lemp->has_fallback ){
4743 int mx = lemp->nterminal - 1;
4744 /* 2019-08-28: Generate fallback entries for every token to avoid
4745 ** having to do a range check on the index */
4746 /* while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; } */
4747 lemp->tablesize += (mx+1)*szCodeType;
4748 for(i=0; i<=mx; i++){
4749 struct symbol *p = lemp->symbols[i];
4750 if( p->fallback==0 ){
4751 fprintf(out, " 0, /* %10s => nothing */\n", p->name);
4752 }else{
4753 fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index,
4754 p->name, p->fallback->name);
4756 lineno++;
4759 tplt_xfer(lemp->name, in, out, &lineno);
4761 /* Generate a table containing the symbolic name of every symbol
4763 for(i=0; i<lemp->nsymbol; i++){
4764 lemon_sprintf(line,"\"%s\",",lemp->symbols[i]->name);
4765 fprintf(out," /* %4d */ \"%s\",\n",i, lemp->symbols[i]->name); lineno++;
4767 tplt_xfer(lemp->name,in,out,&lineno);
4769 /* Generate a table containing a text string that describes every
4770 ** rule in the rule set of the grammar. This information is used
4771 ** when tracing REDUCE actions.
4773 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4774 assert( rp->iRule==i );
4775 fprintf(out," /* %3d */ \"", i);
4776 writeRuleText(out, rp);
4777 fprintf(out,"\",\n"); lineno++;
4779 tplt_xfer(lemp->name,in,out,&lineno);
4781 /* Generate code which executes every time a symbol is popped from
4782 ** the stack while processing errors or while destroying the parser.
4783 ** (In other words, generate the %destructor actions)
4785 if( lemp->tokendest ){
4786 int once = 1;
4787 for(i=0; i<lemp->nsymbol; i++){
4788 struct symbol *sp = lemp->symbols[i];
4789 if( sp==0 || sp->type!=TERMINAL ) continue;
4790 if( once ){
4791 fprintf(out, " /* TERMINAL Destructor */\n"); lineno++;
4792 once = 0;
4794 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4796 for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
4797 if( i<lemp->nsymbol ){
4798 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4799 fprintf(out," break;\n"); lineno++;
4802 if( lemp->vardest ){
4803 struct symbol *dflt_sp = 0;
4804 int once = 1;
4805 for(i=0; i<lemp->nsymbol; i++){
4806 struct symbol *sp = lemp->symbols[i];
4807 if( sp==0 || sp->type==TERMINAL ||
4808 sp->index<=0 || sp->destructor!=0 ) continue;
4809 if( once ){
4810 fprintf(out, " /* Default NON-TERMINAL Destructor */\n");lineno++;
4811 once = 0;
4813 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4814 dflt_sp = sp;
4816 if( dflt_sp!=0 ){
4817 emit_destructor_code(out,dflt_sp,lemp,&lineno);
4819 fprintf(out," break;\n"); lineno++;
4821 for(i=0; i<lemp->nsymbol; i++){
4822 struct symbol *sp = lemp->symbols[i];
4823 if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
4824 if( sp->destLineno<0 ) continue; /* Already emitted */
4825 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4827 /* Combine duplicate destructors into a single case */
4828 for(j=i+1; j<lemp->nsymbol; j++){
4829 struct symbol *sp2 = lemp->symbols[j];
4830 if( sp2 && sp2->type!=TERMINAL && sp2->destructor
4831 && sp2->dtnum==sp->dtnum
4832 && strcmp(sp->destructor,sp2->destructor)==0 ){
4833 fprintf(out," case %d: /* %s */\n",
4834 sp2->index, sp2->name); lineno++;
4835 sp2->destLineno = -1; /* Avoid emitting this destructor again */
4839 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4840 fprintf(out," break;\n"); lineno++;
4842 tplt_xfer(lemp->name,in,out,&lineno);
4844 /* Generate code which executes whenever the parser stack overflows */
4845 tplt_print(out,lemp,lemp->overflow,&lineno);
4846 tplt_xfer(lemp->name,in,out,&lineno);
4848 /* Generate the tables of rule information. yyRuleInfoLhs[] and
4849 ** yyRuleInfoNRhs[].
4851 ** Note: This code depends on the fact that rules are number
4852 ** sequentially beginning with 0.
4854 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4855 fprintf(out," %4d, /* (%d) ", rp->lhs->index, i);
4856 rule_print(out, rp);
4857 fprintf(out," */\n"); lineno++;
4859 tplt_xfer(lemp->name,in,out,&lineno);
4860 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4861 fprintf(out," %3d, /* (%d) ", -rp->nrhs, i);
4862 rule_print(out, rp);
4863 fprintf(out," */\n"); lineno++;
4865 tplt_xfer(lemp->name,in,out,&lineno);
4867 /* Generate code which execution during each REDUCE action */
4868 i = 0;
4869 for(rp=lemp->rule; rp; rp=rp->next){
4870 i += translate_code(lemp, rp);
4872 if( i ){
4873 fprintf(out," YYMINORTYPE yylhsminor;\n"); lineno++;
4875 /* First output rules other than the default: rule */
4876 for(rp=lemp->rule; rp; rp=rp->next){
4877 struct rule *rp2; /* Other rules with the same action */
4878 if( rp->codeEmitted ) continue;
4879 if( rp->noCode ){
4880 /* No C code actions, so this will be part of the "default:" rule */
4881 continue;
4883 fprintf(out," case %d: /* ", rp->iRule);
4884 writeRuleText(out, rp);
4885 fprintf(out, " */\n"); lineno++;
4886 for(rp2=rp->next; rp2; rp2=rp2->next){
4887 if( rp2->code==rp->code && rp2->codePrefix==rp->codePrefix
4888 && rp2->codeSuffix==rp->codeSuffix ){
4889 fprintf(out," case %d: /* ", rp2->iRule);
4890 writeRuleText(out, rp2);
4891 fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->iRule); lineno++;
4892 rp2->codeEmitted = 1;
4895 emit_code(out,rp,lemp,&lineno);
4896 fprintf(out," break;\n"); lineno++;
4897 rp->codeEmitted = 1;
4899 /* Finally, output the default: rule. We choose as the default: all
4900 ** empty actions. */
4901 fprintf(out," default:\n"); lineno++;
4902 for(rp=lemp->rule; rp; rp=rp->next){
4903 if( rp->codeEmitted ) continue;
4904 assert( rp->noCode );
4905 fprintf(out," /* (%d) ", rp->iRule);
4906 writeRuleText(out, rp);
4907 if( rp->neverReduce ){
4908 fprintf(out, " (NEVER REDUCES) */ assert(yyruleno!=%d);\n",
4909 rp->iRule); lineno++;
4910 }else if( rp->doesReduce ){
4911 fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->iRule); lineno++;
4912 }else{
4913 fprintf(out, " (OPTIMIZED OUT) */ assert(yyruleno!=%d);\n",
4914 rp->iRule); lineno++;
4917 fprintf(out," break;\n"); lineno++;
4918 tplt_xfer(lemp->name,in,out,&lineno);
4920 /* Generate code which executes if a parse fails */
4921 tplt_print(out,lemp,lemp->failure,&lineno);
4922 tplt_xfer(lemp->name,in,out,&lineno);
4924 /* Generate code which executes when a syntax error occurs */
4925 tplt_print(out,lemp,lemp->error,&lineno);
4926 tplt_xfer(lemp->name,in,out,&lineno);
4928 /* Generate code which executes when the parser accepts its input */
4929 tplt_print(out,lemp,lemp->accept,&lineno);
4930 tplt_xfer(lemp->name,in,out,&lineno);
4932 /* Append any addition code the user desires */
4933 tplt_print(out,lemp,lemp->extracode,&lineno);
4935 acttab_free(pActtab);
4936 fclose(in);
4937 fclose(out);
4938 if( sql ) fclose(sql);
4939 return;
4942 /* Generate a header file for the parser */
4943 void ReportHeader(struct lemon *lemp)
4945 FILE *out, *in;
4946 const char *prefix;
4947 char line[LINESIZE];
4948 char pattern[LINESIZE];
4949 int i;
4951 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4952 else prefix = "";
4953 in = file_open(lemp,".h","rb");
4954 if( in ){
4955 int nextChar;
4956 for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
4957 lemon_sprintf(pattern,"#define %s%-30s %3d\n",
4958 prefix,lemp->symbols[i]->name,i);
4959 if( strcmp(line,pattern) ) break;
4961 nextChar = fgetc(in);
4962 fclose(in);
4963 if( i==lemp->nterminal && nextChar==EOF ){
4964 /* No change in the file. Don't rewrite it. */
4965 return;
4968 out = file_open(lemp,".h","wb");
4969 if( out ){
4970 for(i=1; i<lemp->nterminal; i++){
4971 fprintf(out,"#define %s%-30s %3d\n",prefix,lemp->symbols[i]->name,i);
4973 fclose(out);
4975 return;
4978 /* Reduce the size of the action tables, if possible, by making use
4979 ** of defaults.
4981 ** In this version, we take the most frequent REDUCE action and make
4982 ** it the default. Except, there is no default if the wildcard token
4983 ** is a possible look-ahead.
4985 void CompressTables(struct lemon *lemp)
4987 struct state *stp;
4988 struct action *ap, *ap2, *nextap;
4989 struct rule *rp, *rp2, *rbest;
4990 int nbest, n;
4991 int i;
4992 int usesWildcard;
4994 for(i=0; i<lemp->nstate; i++){
4995 stp = lemp->sorted[i];
4996 nbest = 0;
4997 rbest = 0;
4998 usesWildcard = 0;
5000 for(ap=stp->ap; ap; ap=ap->next){
5001 if( ap->type==SHIFT && ap->sp==lemp->wildcard ){
5002 usesWildcard = 1;
5004 if( ap->type!=REDUCE ) continue;
5005 rp = ap->x.rp;
5006 if( rp->lhsStart ) continue;
5007 if( rp==rbest ) continue;
5008 n = 1;
5009 for(ap2=ap->next; ap2; ap2=ap2->next){
5010 if( ap2->type!=REDUCE ) continue;
5011 rp2 = ap2->x.rp;
5012 if( rp2==rbest ) continue;
5013 if( rp2==rp ) n++;
5015 if( n>nbest ){
5016 nbest = n;
5017 rbest = rp;
5021 /* Do not make a default if the number of rules to default
5022 ** is not at least 1 or if the wildcard token is a possible
5023 ** lookahead.
5025 if( nbest<1 || usesWildcard ) continue;
5028 /* Combine matching REDUCE actions into a single default */
5029 for(ap=stp->ap; ap; ap=ap->next){
5030 if( ap->type==REDUCE && ap->x.rp==rbest ) break;
5032 assert( ap );
5033 ap->sp = Symbol_new("{default}");
5034 for(ap=ap->next; ap; ap=ap->next){
5035 if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
5037 stp->ap = Action_sort(stp->ap);
5039 for(ap=stp->ap; ap; ap=ap->next){
5040 if( ap->type==SHIFT ) break;
5041 if( ap->type==REDUCE && ap->x.rp!=rbest ) break;
5043 if( ap==0 ){
5044 stp->autoReduce = 1;
5045 stp->pDfltReduce = rbest;
5049 /* Make a second pass over all states and actions. Convert
5050 ** every action that is a SHIFT to an autoReduce state into
5051 ** a SHIFTREDUCE action.
5053 for(i=0; i<lemp->nstate; i++){
5054 stp = lemp->sorted[i];
5055 for(ap=stp->ap; ap; ap=ap->next){
5056 struct state *pNextState;
5057 if( ap->type!=SHIFT ) continue;
5058 pNextState = ap->x.stp;
5059 if( pNextState->autoReduce && pNextState->pDfltReduce!=0 ){
5060 ap->type = SHIFTREDUCE;
5061 ap->x.rp = pNextState->pDfltReduce;
5066 /* If a SHIFTREDUCE action specifies a rule that has a single RHS term
5067 ** (meaning that the SHIFTREDUCE will land back in the state where it
5068 ** started) and if there is no C-code associated with the reduce action,
5069 ** then we can go ahead and convert the action to be the same as the
5070 ** action for the RHS of the rule.
5072 for(i=0; i<lemp->nstate; i++){
5073 stp = lemp->sorted[i];
5074 for(ap=stp->ap; ap; ap=nextap){
5075 nextap = ap->next;
5076 if( ap->type!=SHIFTREDUCE ) continue;
5077 rp = ap->x.rp;
5078 if( rp->noCode==0 ) continue;
5079 if( rp->nrhs!=1 ) continue;
5080 #if 1
5081 /* Only apply this optimization to non-terminals. It would be OK to
5082 ** apply it to terminal symbols too, but that makes the parser tables
5083 ** larger. */
5084 if( ap->sp->index<lemp->nterminal ) continue;
5085 #endif
5086 /* If we reach this point, it means the optimization can be applied */
5087 nextap = ap;
5088 for(ap2=stp->ap; ap2 && (ap2==ap || ap2->sp!=rp->lhs); ap2=ap2->next){}
5089 assert( ap2!=0 );
5090 ap->spOpt = ap2->sp;
5091 ap->type = ap2->type;
5092 ap->x = ap2->x;
5099 ** Compare two states for sorting purposes. The smaller state is the
5100 ** one with the most non-terminal actions. If they have the same number
5101 ** of non-terminal actions, then the smaller is the one with the most
5102 ** token actions.
5104 static int stateResortCompare(const void *a, const void *b){
5105 const struct state *pA = *(const struct state**)a;
5106 const struct state *pB = *(const struct state**)b;
5107 int n;
5109 n = pB->nNtAct - pA->nNtAct;
5110 if( n==0 ){
5111 n = pB->nTknAct - pA->nTknAct;
5112 if( n==0 ){
5113 n = pB->statenum - pA->statenum;
5116 assert( n!=0 );
5117 return n;
5122 ** Renumber and resort states so that states with fewer choices
5123 ** occur at the end. Except, keep state 0 as the first state.
5125 void ResortStates(struct lemon *lemp)
5127 int i;
5128 struct state *stp;
5129 struct action *ap;
5131 for(i=0; i<lemp->nstate; i++){
5132 stp = lemp->sorted[i];
5133 stp->nTknAct = stp->nNtAct = 0;
5134 stp->iDfltReduce = -1; /* Init dflt action to "syntax error" */
5135 stp->iTknOfst = NO_OFFSET;
5136 stp->iNtOfst = NO_OFFSET;
5137 for(ap=stp->ap; ap; ap=ap->next){
5138 int iAction = compute_action(lemp,ap);
5139 if( iAction>=0 ){
5140 if( ap->sp->index<lemp->nterminal ){
5141 stp->nTknAct++;
5142 }else if( ap->sp->index<lemp->nsymbol ){
5143 stp->nNtAct++;
5144 }else{
5145 assert( stp->autoReduce==0 || stp->pDfltReduce==ap->x.rp );
5146 stp->iDfltReduce = iAction;
5151 qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]),
5152 stateResortCompare);
5153 for(i=0; i<lemp->nstate; i++){
5154 lemp->sorted[i]->statenum = i;
5156 lemp->nxstate = lemp->nstate;
5157 while( lemp->nxstate>1 && lemp->sorted[lemp->nxstate-1]->autoReduce ){
5158 lemp->nxstate--;
5163 /***************** From the file "set.c" ************************************/
5165 ** Set manipulation routines for the LEMON parser generator.
5168 static int size = 0;
5170 /* Set the set size */
5171 void SetSize(int n)
5173 size = n+1;
5176 /* Allocate a new set */
5177 char *SetNew(void){
5178 char *s;
5179 s = (char*)calloc( size, 1);
5180 if( s==0 ){
5181 memory_error();
5183 return s;
5186 /* Deallocate a set */
5187 void SetFree(char *s)
5189 free(s);
5192 /* Add a new element to the set. Return TRUE if the element was added
5193 ** and FALSE if it was already there. */
5194 int SetAdd(char *s, int e)
5196 int rv;
5197 assert( e>=0 && e<size );
5198 rv = s[e];
5199 s[e] = 1;
5200 return !rv;
5203 /* Add every element of s2 to s1. Return TRUE if s1 changes. */
5204 int SetUnion(char *s1, char *s2)
5206 int i, progress;
5207 progress = 0;
5208 for(i=0; i<size; i++){
5209 if( s2[i]==0 ) continue;
5210 if( s1[i]==0 ){
5211 progress = 1;
5212 s1[i] = 1;
5215 return progress;
5217 /********************** From the file "table.c" ****************************/
5219 ** All code in this file has been automatically generated
5220 ** from a specification in the file
5221 ** "table.q"
5222 ** by the associative array code building program "aagen".
5223 ** Do not edit this file! Instead, edit the specification
5224 ** file, then rerun aagen.
5227 ** Code for processing tables in the LEMON parser generator.
5230 PRIVATE unsigned strhash(const char *x)
5232 unsigned h = 0;
5233 while( *x ) h = h*13 + *(x++);
5234 return h;
5237 /* Works like strdup, sort of. Save a string in malloced memory, but
5238 ** keep strings in a table so that the same string is not in more
5239 ** than one place.
5241 const char *Strsafe(const char *y)
5243 const char *z;
5244 char *cpy;
5246 if( y==0 ) return 0;
5247 z = Strsafe_find(y);
5248 if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){
5249 lemon_strcpy(cpy,y);
5250 z = cpy;
5251 Strsafe_insert(z);
5253 MemoryCheck(z);
5254 return z;
5257 /* There is one instance of the following structure for each
5258 ** associative array of type "x1".
5260 struct s_x1 {
5261 int size; /* The number of available slots. */
5262 /* Must be a power of 2 greater than or */
5263 /* equal to 1 */
5264 int count; /* Number of currently slots filled */
5265 struct s_x1node *tbl; /* The data stored here */
5266 struct s_x1node **ht; /* Hash table for lookups */
5269 /* There is one instance of this structure for every data element
5270 ** in an associative array of type "x1".
5272 typedef struct s_x1node {
5273 const char *data; /* The data */
5274 struct s_x1node *next; /* Next entry with the same hash */
5275 struct s_x1node **from; /* Previous link */
5276 } x1node;
5278 /* There is only one instance of the array, which is the following */
5279 static struct s_x1 *x1a;
5281 /* Allocate a new associative array */
5282 void Strsafe_init(void){
5283 if( x1a ) return;
5284 x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
5285 if( x1a ){
5286 x1a->size = 1024;
5287 x1a->count = 0;
5288 x1a->tbl = (x1node*)calloc(1024, sizeof(x1node) + sizeof(x1node*));
5289 if( x1a->tbl==0 ){
5290 free(x1a);
5291 x1a = 0;
5292 }else{
5293 int i;
5294 x1a->ht = (x1node**)&(x1a->tbl[1024]);
5295 for(i=0; i<1024; i++) x1a->ht[i] = 0;
5299 /* Insert a new record into the array. Return TRUE if successful.
5300 ** Prior data with the same key is NOT overwritten */
5301 int Strsafe_insert(const char *data)
5303 x1node *np;
5304 unsigned h;
5305 unsigned ph;
5307 if( x1a==0 ) return 0;
5308 ph = strhash(data);
5309 h = ph & (x1a->size-1);
5310 np = x1a->ht[h];
5311 while( np ){
5312 if( strcmp(np->data,data)==0 ){
5313 /* An existing entry with the same key is found. */
5314 /* Fail because overwrite is not allows. */
5315 return 0;
5317 np = np->next;
5319 if( x1a->count>=x1a->size ){
5320 /* Need to make the hash table bigger */
5321 int i,arrSize;
5322 struct s_x1 array;
5323 array.size = arrSize = x1a->size*2;
5324 array.count = x1a->count;
5325 array.tbl = (x1node*)calloc(arrSize, sizeof(x1node) + sizeof(x1node*));
5326 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5327 array.ht = (x1node**)&(array.tbl[arrSize]);
5328 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5329 for(i=0; i<x1a->count; i++){
5330 x1node *oldnp, *newnp;
5331 oldnp = &(x1a->tbl[i]);
5332 h = strhash(oldnp->data) & (arrSize-1);
5333 newnp = &(array.tbl[i]);
5334 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5335 newnp->next = array.ht[h];
5336 newnp->data = oldnp->data;
5337 newnp->from = &(array.ht[h]);
5338 array.ht[h] = newnp;
5340 free(x1a->tbl);
5341 *x1a = array;
5343 /* Insert the new data */
5344 h = ph & (x1a->size-1);
5345 np = &(x1a->tbl[x1a->count++]);
5346 np->data = data;
5347 if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
5348 np->next = x1a->ht[h];
5349 x1a->ht[h] = np;
5350 np->from = &(x1a->ht[h]);
5351 return 1;
5354 /* Return a pointer to data assigned to the given key. Return NULL
5355 ** if no such key. */
5356 const char *Strsafe_find(const char *key)
5358 unsigned h;
5359 x1node *np;
5361 if( x1a==0 ) return 0;
5362 h = strhash(key) & (x1a->size-1);
5363 np = x1a->ht[h];
5364 while( np ){
5365 if( strcmp(np->data,key)==0 ) break;
5366 np = np->next;
5368 return np ? np->data : 0;
5371 /* Return a pointer to the (terminal or nonterminal) symbol "x".
5372 ** Create a new symbol if this is the first time "x" has been seen.
5374 struct symbol *Symbol_new(const char *x)
5376 struct symbol *sp;
5378 sp = Symbol_find(x);
5379 if( sp==0 ){
5380 sp = (struct symbol *)calloc(1, sizeof(struct symbol) );
5381 MemoryCheck(sp);
5382 sp->name = Strsafe(x);
5383 sp->type = ISUPPER(*x) ? TERMINAL : NONTERMINAL;
5384 sp->rule = 0;
5385 sp->fallback = 0;
5386 sp->prec = -1;
5387 sp->assoc = UNK;
5388 sp->firstset = 0;
5389 sp->lambda = LEMON_FALSE;
5390 sp->destructor = 0;
5391 sp->destLineno = 0;
5392 sp->datatype = 0;
5393 sp->useCnt = 0;
5394 Symbol_insert(sp,sp->name);
5396 sp->useCnt++;
5397 return sp;
5400 /* Compare two symbols for sorting purposes. Return negative,
5401 ** zero, or positive if a is less then, equal to, or greater
5402 ** than b.
5404 ** Symbols that begin with upper case letters (terminals or tokens)
5405 ** must sort before symbols that begin with lower case letters
5406 ** (non-terminals). And MULTITERMINAL symbols (created using the
5407 ** %token_class directive) must sort at the very end. Other than
5408 ** that, the order does not matter.
5410 ** We find experimentally that leaving the symbols in their original
5411 ** order (the order they appeared in the grammar file) gives the
5412 ** smallest parser tables in SQLite.
5414 int Symbolcmpp(const void *_a, const void *_b)
5416 const struct symbol *a = *(const struct symbol **) _a;
5417 const struct symbol *b = *(const struct symbol **) _b;
5418 int i1 = a->type==MULTITERMINAL ? 3 : a->name[0]>'Z' ? 2 : 1;
5419 int i2 = b->type==MULTITERMINAL ? 3 : b->name[0]>'Z' ? 2 : 1;
5420 return i1==i2 ? a->index - b->index : i1 - i2;
5423 /* There is one instance of the following structure for each
5424 ** associative array of type "x2".
5426 struct s_x2 {
5427 int size; /* The number of available slots. */
5428 /* Must be a power of 2 greater than or */
5429 /* equal to 1 */
5430 int count; /* Number of currently slots filled */
5431 struct s_x2node *tbl; /* The data stored here */
5432 struct s_x2node **ht; /* Hash table for lookups */
5435 /* There is one instance of this structure for every data element
5436 ** in an associative array of type "x2".
5438 typedef struct s_x2node {
5439 struct symbol *data; /* The data */
5440 const char *key; /* The key */
5441 struct s_x2node *next; /* Next entry with the same hash */
5442 struct s_x2node **from; /* Previous link */
5443 } x2node;
5445 /* There is only one instance of the array, which is the following */
5446 static struct s_x2 *x2a;
5448 /* Allocate a new associative array */
5449 void Symbol_init(void){
5450 if( x2a ) return;
5451 x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
5452 if( x2a ){
5453 x2a->size = 128;
5454 x2a->count = 0;
5455 x2a->tbl = (x2node*)calloc(128, sizeof(x2node) + sizeof(x2node*));
5456 if( x2a->tbl==0 ){
5457 free(x2a);
5458 x2a = 0;
5459 }else{
5460 int i;
5461 x2a->ht = (x2node**)&(x2a->tbl[128]);
5462 for(i=0; i<128; i++) x2a->ht[i] = 0;
5466 /* Insert a new record into the array. Return TRUE if successful.
5467 ** Prior data with the same key is NOT overwritten */
5468 int Symbol_insert(struct symbol *data, const char *key)
5470 x2node *np;
5471 unsigned h;
5472 unsigned ph;
5474 if( x2a==0 ) return 0;
5475 ph = strhash(key);
5476 h = ph & (x2a->size-1);
5477 np = x2a->ht[h];
5478 while( np ){
5479 if( strcmp(np->key,key)==0 ){
5480 /* An existing entry with the same key is found. */
5481 /* Fail because overwrite is not allows. */
5482 return 0;
5484 np = np->next;
5486 if( x2a->count>=x2a->size ){
5487 /* Need to make the hash table bigger */
5488 int i,arrSize;
5489 struct s_x2 array;
5490 array.size = arrSize = x2a->size*2;
5491 array.count = x2a->count;
5492 array.tbl = (x2node*)calloc(arrSize, sizeof(x2node) + sizeof(x2node*));
5493 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5494 array.ht = (x2node**)&(array.tbl[arrSize]);
5495 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5496 for(i=0; i<x2a->count; i++){
5497 x2node *oldnp, *newnp;
5498 oldnp = &(x2a->tbl[i]);
5499 h = strhash(oldnp->key) & (arrSize-1);
5500 newnp = &(array.tbl[i]);
5501 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5502 newnp->next = array.ht[h];
5503 newnp->key = oldnp->key;
5504 newnp->data = oldnp->data;
5505 newnp->from = &(array.ht[h]);
5506 array.ht[h] = newnp;
5508 free(x2a->tbl);
5509 *x2a = array;
5511 /* Insert the new data */
5512 h = ph & (x2a->size-1);
5513 np = &(x2a->tbl[x2a->count++]);
5514 np->key = key;
5515 np->data = data;
5516 if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
5517 np->next = x2a->ht[h];
5518 x2a->ht[h] = np;
5519 np->from = &(x2a->ht[h]);
5520 return 1;
5523 /* Return a pointer to data assigned to the given key. Return NULL
5524 ** if no such key. */
5525 struct symbol *Symbol_find(const char *key)
5527 unsigned h;
5528 x2node *np;
5530 if( x2a==0 ) return 0;
5531 h = strhash(key) & (x2a->size-1);
5532 np = x2a->ht[h];
5533 while( np ){
5534 if( strcmp(np->key,key)==0 ) break;
5535 np = np->next;
5537 return np ? np->data : 0;
5540 /* Return the n-th data. Return NULL if n is out of range. */
5541 struct symbol *Symbol_Nth(int n)
5543 struct symbol *data;
5544 if( x2a && n>0 && n<=x2a->count ){
5545 data = x2a->tbl[n-1].data;
5546 }else{
5547 data = 0;
5549 return data;
5552 /* Return the size of the array */
5553 int Symbol_count()
5555 return x2a ? x2a->count : 0;
5558 /* Return an array of pointers to all data in the table.
5559 ** The array is obtained from malloc. Return NULL if memory allocation
5560 ** problems, or if the array is empty. */
5561 struct symbol **Symbol_arrayof()
5563 struct symbol **array;
5564 int i,arrSize;
5565 if( x2a==0 ) return 0;
5566 arrSize = x2a->count;
5567 array = (struct symbol **)calloc(arrSize, sizeof(struct symbol *));
5568 if( array ){
5569 for(i=0; i<arrSize; i++) array[i] = x2a->tbl[i].data;
5571 return array;
5574 /* Compare two configurations */
5575 int Configcmp(const char *_a,const char *_b)
5577 const struct config *a = (struct config *) _a;
5578 const struct config *b = (struct config *) _b;
5579 int x;
5580 x = a->rp->index - b->rp->index;
5581 if( x==0 ) x = a->dot - b->dot;
5582 return x;
5585 /* Compare two states */
5586 PRIVATE int statecmp(struct config *a, struct config *b)
5588 int rc;
5589 for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){
5590 rc = a->rp->index - b->rp->index;
5591 if( rc==0 ) rc = a->dot - b->dot;
5593 if( rc==0 ){
5594 if( a ) rc = 1;
5595 if( b ) rc = -1;
5597 return rc;
5600 /* Hash a state */
5601 PRIVATE unsigned statehash(struct config *a)
5603 unsigned h=0;
5604 while( a ){
5605 h = h*571 + a->rp->index*37 + a->dot;
5606 a = a->bp;
5608 return h;
5611 /* Allocate a new state structure */
5612 struct state *State_new()
5614 struct state *newstate;
5615 newstate = (struct state *)calloc(1, sizeof(struct state) );
5616 MemoryCheck(newstate);
5617 return newstate;
5620 /* There is one instance of the following structure for each
5621 ** associative array of type "x3".
5623 struct s_x3 {
5624 int size; /* The number of available slots. */
5625 /* Must be a power of 2 greater than or */
5626 /* equal to 1 */
5627 int count; /* Number of currently slots filled */
5628 struct s_x3node *tbl; /* The data stored here */
5629 struct s_x3node **ht; /* Hash table for lookups */
5632 /* There is one instance of this structure for every data element
5633 ** in an associative array of type "x3".
5635 typedef struct s_x3node {
5636 struct state *data; /* The data */
5637 struct config *key; /* The key */
5638 struct s_x3node *next; /* Next entry with the same hash */
5639 struct s_x3node **from; /* Previous link */
5640 } x3node;
5642 /* There is only one instance of the array, which is the following */
5643 static struct s_x3 *x3a;
5645 /* Allocate a new associative array */
5646 void State_init(void){
5647 if( x3a ) return;
5648 x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
5649 if( x3a ){
5650 x3a->size = 128;
5651 x3a->count = 0;
5652 x3a->tbl = (x3node*)calloc(128, sizeof(x3node) + sizeof(x3node*));
5653 if( x3a->tbl==0 ){
5654 free(x3a);
5655 x3a = 0;
5656 }else{
5657 int i;
5658 x3a->ht = (x3node**)&(x3a->tbl[128]);
5659 for(i=0; i<128; i++) x3a->ht[i] = 0;
5663 /* Insert a new record into the array. Return TRUE if successful.
5664 ** Prior data with the same key is NOT overwritten */
5665 int State_insert(struct state *data, struct config *key)
5667 x3node *np;
5668 unsigned h;
5669 unsigned ph;
5671 if( x3a==0 ) return 0;
5672 ph = statehash(key);
5673 h = ph & (x3a->size-1);
5674 np = x3a->ht[h];
5675 while( np ){
5676 if( statecmp(np->key,key)==0 ){
5677 /* An existing entry with the same key is found. */
5678 /* Fail because overwrite is not allows. */
5679 return 0;
5681 np = np->next;
5683 if( x3a->count>=x3a->size ){
5684 /* Need to make the hash table bigger */
5685 int i,arrSize;
5686 struct s_x3 array;
5687 array.size = arrSize = x3a->size*2;
5688 array.count = x3a->count;
5689 array.tbl = (x3node*)calloc(arrSize, sizeof(x3node) + sizeof(x3node*));
5690 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5691 array.ht = (x3node**)&(array.tbl[arrSize]);
5692 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5693 for(i=0; i<x3a->count; i++){
5694 x3node *oldnp, *newnp;
5695 oldnp = &(x3a->tbl[i]);
5696 h = statehash(oldnp->key) & (arrSize-1);
5697 newnp = &(array.tbl[i]);
5698 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5699 newnp->next = array.ht[h];
5700 newnp->key = oldnp->key;
5701 newnp->data = oldnp->data;
5702 newnp->from = &(array.ht[h]);
5703 array.ht[h] = newnp;
5705 free(x3a->tbl);
5706 *x3a = array;
5708 /* Insert the new data */
5709 h = ph & (x3a->size-1);
5710 np = &(x3a->tbl[x3a->count++]);
5711 np->key = key;
5712 np->data = data;
5713 if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
5714 np->next = x3a->ht[h];
5715 x3a->ht[h] = np;
5716 np->from = &(x3a->ht[h]);
5717 return 1;
5720 /* Return a pointer to data assigned to the given key. Return NULL
5721 ** if no such key. */
5722 struct state *State_find(struct config *key)
5724 unsigned h;
5725 x3node *np;
5727 if( x3a==0 ) return 0;
5728 h = statehash(key) & (x3a->size-1);
5729 np = x3a->ht[h];
5730 while( np ){
5731 if( statecmp(np->key,key)==0 ) break;
5732 np = np->next;
5734 return np ? np->data : 0;
5737 /* Return an array of pointers to all data in the table.
5738 ** The array is obtained from malloc. Return NULL if memory allocation
5739 ** problems, or if the array is empty. */
5740 struct state **State_arrayof(void)
5742 struct state **array;
5743 int i,arrSize;
5744 if( x3a==0 ) return 0;
5745 arrSize = x3a->count;
5746 array = (struct state **)calloc(arrSize, sizeof(struct state *));
5747 if( array ){
5748 for(i=0; i<arrSize; i++) array[i] = x3a->tbl[i].data;
5750 return array;
5753 /* Hash a configuration */
5754 PRIVATE unsigned confighash(struct config *a)
5756 unsigned h=0;
5757 h = h*571 + a->rp->index*37 + a->dot;
5758 return h;
5761 /* There is one instance of the following structure for each
5762 ** associative array of type "x4".
5764 struct s_x4 {
5765 int size; /* The number of available slots. */
5766 /* Must be a power of 2 greater than or */
5767 /* equal to 1 */
5768 int count; /* Number of currently slots filled */
5769 struct s_x4node *tbl; /* The data stored here */
5770 struct s_x4node **ht; /* Hash table for lookups */
5773 /* There is one instance of this structure for every data element
5774 ** in an associative array of type "x4".
5776 typedef struct s_x4node {
5777 struct config *data; /* The data */
5778 struct s_x4node *next; /* Next entry with the same hash */
5779 struct s_x4node **from; /* Previous link */
5780 } x4node;
5782 /* There is only one instance of the array, which is the following */
5783 static struct s_x4 *x4a;
5785 /* Allocate a new associative array */
5786 void Configtable_init(void){
5787 if( x4a ) return;
5788 x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
5789 if( x4a ){
5790 x4a->size = 64;
5791 x4a->count = 0;
5792 x4a->tbl = (x4node*)calloc(64, sizeof(x4node) + sizeof(x4node*));
5793 if( x4a->tbl==0 ){
5794 free(x4a);
5795 x4a = 0;
5796 }else{
5797 int i;
5798 x4a->ht = (x4node**)&(x4a->tbl[64]);
5799 for(i=0; i<64; i++) x4a->ht[i] = 0;
5803 /* Insert a new record into the array. Return TRUE if successful.
5804 ** Prior data with the same key is NOT overwritten */
5805 int Configtable_insert(struct config *data)
5807 x4node *np;
5808 unsigned h;
5809 unsigned ph;
5811 if( x4a==0 ) return 0;
5812 ph = confighash(data);
5813 h = ph & (x4a->size-1);
5814 np = x4a->ht[h];
5815 while( np ){
5816 if( Configcmp((const char *) np->data,(const char *) data)==0 ){
5817 /* An existing entry with the same key is found. */
5818 /* Fail because overwrite is not allows. */
5819 return 0;
5821 np = np->next;
5823 if( x4a->count>=x4a->size ){
5824 /* Need to make the hash table bigger */
5825 int i,arrSize;
5826 struct s_x4 array;
5827 array.size = arrSize = x4a->size*2;
5828 array.count = x4a->count;
5829 array.tbl = (x4node*)calloc(arrSize, sizeof(x4node) + sizeof(x4node*));
5830 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5831 array.ht = (x4node**)&(array.tbl[arrSize]);
5832 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5833 for(i=0; i<x4a->count; i++){
5834 x4node *oldnp, *newnp;
5835 oldnp = &(x4a->tbl[i]);
5836 h = confighash(oldnp->data) & (arrSize-1);
5837 newnp = &(array.tbl[i]);
5838 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5839 newnp->next = array.ht[h];
5840 newnp->data = oldnp->data;
5841 newnp->from = &(array.ht[h]);
5842 array.ht[h] = newnp;
5844 free(x4a->tbl);
5845 *x4a = array;
5847 /* Insert the new data */
5848 h = ph & (x4a->size-1);
5849 np = &(x4a->tbl[x4a->count++]);
5850 np->data = data;
5851 if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
5852 np->next = x4a->ht[h];
5853 x4a->ht[h] = np;
5854 np->from = &(x4a->ht[h]);
5855 return 1;
5858 /* Return a pointer to data assigned to the given key. Return NULL
5859 ** if no such key. */
5860 struct config *Configtable_find(struct config *key)
5862 int h;
5863 x4node *np;
5865 if( x4a==0 ) return 0;
5866 h = confighash(key) & (x4a->size-1);
5867 np = x4a->ht[h];
5868 while( np ){
5869 if( Configcmp((const char *) np->data,(const char *) key)==0 ) break;
5870 np = np->next;
5872 return np ? np->data : 0;
5875 /* Remove all data from the table. Pass each data to the function "f"
5876 ** as it is removed. ("f" may be null to avoid this step.) */
5877 void Configtable_clear(int(*f)(struct config *))
5879 int i;
5880 if( x4a==0 || x4a->count==0 ) return;
5881 if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
5882 for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;
5883 x4a->count = 0;
5884 return;