Improved error messages. Limit the number of auxiliary columns to 100.
[sqlite.git] / tool / lemon.c
blobc2c9554c89e8fc5090c9b7b729cfed0bfc43b28c
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 static int showPrecedenceConflict = 0;
52 static char *msort(char*,char**,int(*)(const char*,const char*));
55 ** Compilers are getting increasingly pedantic about type conversions
56 ** as C evolves ever closer to Ada.... To work around the latest problems
57 ** we have to define the following variant of strlen().
59 #define lemonStrlen(X) ((int)strlen(X))
62 ** Compilers are starting to complain about the use of sprintf() and strcpy(),
63 ** saying they are unsafe. So we define our own versions of those routines too.
65 ** There are three routines here: lemon_sprintf(), lemon_vsprintf(), and
66 ** lemon_addtext(). The first two are replacements for sprintf() and vsprintf().
67 ** The third is a helper routine for vsnprintf() that adds texts to the end of a
68 ** buffer, making sure the buffer is always zero-terminated.
70 ** The string formatter is a minimal subset of stdlib sprintf() supporting only
71 ** a few simply conversions:
73 ** %d
74 ** %s
75 ** %.*s
78 static void lemon_addtext(
79 char *zBuf, /* The buffer to which text is added */
80 int *pnUsed, /* Slots of the buffer used so far */
81 const char *zIn, /* Text to add */
82 int nIn, /* Bytes of text to add. -1 to use strlen() */
83 int iWidth /* Field width. Negative to left justify */
85 if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){}
86 while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; }
87 if( nIn==0 ) return;
88 memcpy(&zBuf[*pnUsed], zIn, nIn);
89 *pnUsed += nIn;
90 while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; }
91 zBuf[*pnUsed] = 0;
93 static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){
94 int i, j, k, c;
95 int nUsed = 0;
96 const char *z;
97 char zTemp[50];
98 str[0] = 0;
99 for(i=j=0; (c = zFormat[i])!=0; i++){
100 if( c=='%' ){
101 int iWidth = 0;
102 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
103 c = zFormat[++i];
104 if( ISDIGIT(c) || (c=='-' && ISDIGIT(zFormat[i+1])) ){
105 if( c=='-' ) i++;
106 while( ISDIGIT(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0';
107 if( c=='-' ) iWidth = -iWidth;
108 c = zFormat[i];
110 if( c=='d' ){
111 int v = va_arg(ap, int);
112 if( v<0 ){
113 lemon_addtext(str, &nUsed, "-", 1, iWidth);
114 v = -v;
115 }else if( v==0 ){
116 lemon_addtext(str, &nUsed, "0", 1, iWidth);
118 k = 0;
119 while( v>0 ){
120 k++;
121 zTemp[sizeof(zTemp)-k] = (v%10) + '0';
122 v /= 10;
124 lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth);
125 }else if( c=='s' ){
126 z = va_arg(ap, const char*);
127 lemon_addtext(str, &nUsed, z, -1, iWidth);
128 }else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){
129 i += 2;
130 k = va_arg(ap, int);
131 z = va_arg(ap, const char*);
132 lemon_addtext(str, &nUsed, z, k, iWidth);
133 }else if( c=='%' ){
134 lemon_addtext(str, &nUsed, "%", 1, 0);
135 }else{
136 fprintf(stderr, "illegal format\n");
137 exit(1);
139 j = i+1;
142 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
143 return nUsed;
145 static int lemon_sprintf(char *str, const char *format, ...){
146 va_list ap;
147 int rc;
148 va_start(ap, format);
149 rc = lemon_vsprintf(str, format, ap);
150 va_end(ap);
151 return rc;
153 static void lemon_strcpy(char *dest, const char *src){
154 while( (*(dest++) = *(src++))!=0 ){}
156 static void lemon_strcat(char *dest, const char *src){
157 while( *dest ) dest++;
158 lemon_strcpy(dest, src);
162 /* a few forward declarations... */
163 struct rule;
164 struct lemon;
165 struct action;
167 static struct action *Action_new(void);
168 static struct action *Action_sort(struct action *);
170 /********** From the file "build.h" ************************************/
171 void FindRulePrecedences(struct lemon*);
172 void FindFirstSets(struct lemon*);
173 void FindStates(struct lemon*);
174 void FindLinks(struct lemon*);
175 void FindFollowSets(struct lemon*);
176 void FindActions(struct lemon*);
178 /********* From the file "configlist.h" *********************************/
179 void Configlist_init(void);
180 struct config *Configlist_add(struct rule *, int);
181 struct config *Configlist_addbasis(struct rule *, int);
182 void Configlist_closure(struct lemon *);
183 void Configlist_sort(void);
184 void Configlist_sortbasis(void);
185 struct config *Configlist_return(void);
186 struct config *Configlist_basis(void);
187 void Configlist_eat(struct config *);
188 void Configlist_reset(void);
190 /********* From the file "error.h" ***************************************/
191 void ErrorMsg(const char *, int,const char *, ...);
193 /****** From the file "option.h" ******************************************/
194 enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
195 OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR};
196 struct s_options {
197 enum option_type type;
198 const char *label;
199 char *arg;
200 const char *message;
202 int OptInit(char**,struct s_options*,FILE*);
203 int OptNArgs(void);
204 char *OptArg(int);
205 void OptErr(int);
206 void OptPrint(void);
208 /******** From the file "parse.h" *****************************************/
209 void Parse(struct lemon *lemp);
211 /********* From the file "plink.h" ***************************************/
212 struct plink *Plink_new(void);
213 void Plink_add(struct plink **, struct config *);
214 void Plink_copy(struct plink **, struct plink *);
215 void Plink_delete(struct plink *);
217 /********** From the file "report.h" *************************************/
218 void Reprint(struct lemon *);
219 void ReportOutput(struct lemon *);
220 void ReportTable(struct lemon *, int);
221 void ReportHeader(struct lemon *);
222 void CompressTables(struct lemon *);
223 void ResortStates(struct lemon *);
225 /********** From the file "set.h" ****************************************/
226 void SetSize(int); /* All sets will be of size N */
227 char *SetNew(void); /* A new set for element 0..N */
228 void SetFree(char*); /* Deallocate a set */
229 int SetAdd(char*,int); /* Add element to a set */
230 int SetUnion(char *,char *); /* A <- A U B, thru element N */
231 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
233 /********** From the file "struct.h" *************************************/
235 ** Principal data structures for the LEMON parser generator.
238 typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
240 /* Symbols (terminals and nonterminals) of the grammar are stored
241 ** in the following: */
242 enum symbol_type {
243 TERMINAL,
244 NONTERMINAL,
245 MULTITERMINAL
247 enum e_assoc {
248 LEFT,
249 RIGHT,
250 NONE,
253 struct symbol {
254 const char *name; /* Name of the symbol */
255 int index; /* Index number for this symbol */
256 enum symbol_type type; /* Symbols are all either TERMINALS or NTs */
257 struct rule *rule; /* Linked list of rules of this (if an NT) */
258 struct symbol *fallback; /* fallback token in case this token doesn't parse */
259 int prec; /* Precedence if defined (-1 otherwise) */
260 enum e_assoc assoc; /* Associativity if precedence is defined */
261 char *firstset; /* First-set for all rules of this symbol */
262 Boolean lambda; /* True if NT and can generate an empty string */
263 int useCnt; /* Number of times used */
264 char *destructor; /* Code which executes whenever this symbol is
265 ** popped from the stack during error processing */
266 int destLineno; /* Line number for start of destructor. Set to
267 ** -1 for duplicate destructors. */
268 char *datatype; /* The data type of information held by this
269 ** object. Only used if type==NONTERMINAL */
270 int dtnum; /* The data type number. In the parser, the value
271 ** stack is a union. The .yy%d element of this
272 ** union is the correct data type for this object */
273 int bContent; /* True if this symbol ever carries content - if
274 ** it is ever more than just syntax */
275 /* The following fields are used by MULTITERMINALs only */
276 int nsubsym; /* Number of constituent symbols in the MULTI */
277 struct symbol **subsym; /* Array of constituent symbols */
280 /* Each production rule in the grammar is stored in the following
281 ** structure. */
282 struct rule {
283 struct symbol *lhs; /* Left-hand side of the rule */
284 const char *lhsalias; /* Alias for the LHS (NULL if none) */
285 int lhsStart; /* True if left-hand side is the start symbol */
286 int ruleline; /* Line number for the rule */
287 int nrhs; /* Number of RHS symbols */
288 struct symbol **rhs; /* The RHS symbols */
289 const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
290 int line; /* Line number at which code begins */
291 const char *code; /* The code executed when this rule is reduced */
292 const char *codePrefix; /* Setup code before code[] above */
293 const char *codeSuffix; /* Breakdown code after code[] above */
294 int noCode; /* True if this rule has no associated C code */
295 int codeEmitted; /* True if the code has been emitted already */
296 struct symbol *precsym; /* Precedence symbol for this rule */
297 int index; /* An index number for this rule */
298 int iRule; /* Rule number as used in the generated tables */
299 Boolean canReduce; /* True if this rule is ever reduced */
300 Boolean doesReduce; /* Reduce actions occur after optimization */
301 struct rule *nextlhs; /* Next rule with the same LHS */
302 struct rule *next; /* Next rule in the global list */
305 /* A configuration is a production rule of the grammar together with
306 ** a mark (dot) showing how much of that rule has been processed so far.
307 ** Configurations also contain a follow-set which is a list of terminal
308 ** symbols which are allowed to immediately follow the end of the rule.
309 ** Every configuration is recorded as an instance of the following: */
310 enum cfgstatus {
311 COMPLETE,
312 INCOMPLETE
314 struct config {
315 struct rule *rp; /* The rule upon which the configuration is based */
316 int dot; /* The parse point */
317 char *fws; /* Follow-set for this configuration only */
318 struct plink *fplp; /* Follow-set forward propagation links */
319 struct plink *bplp; /* Follow-set backwards propagation links */
320 struct state *stp; /* Pointer to state which contains this */
321 enum cfgstatus status; /* used during followset and shift computations */
322 struct config *next; /* Next configuration in the state */
323 struct config *bp; /* The next basis configuration */
326 enum e_action {
327 SHIFT,
328 ACCEPT,
329 REDUCE,
330 ERROR,
331 SSCONFLICT, /* A shift/shift conflict */
332 SRCONFLICT, /* Was a reduce, but part of a conflict */
333 RRCONFLICT, /* Was a reduce, but part of a conflict */
334 SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
335 RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
336 NOT_USED, /* Deleted by compression */
337 SHIFTREDUCE /* Shift first, then reduce */
340 /* Every shift or reduce operation is stored as one of the following */
341 struct action {
342 struct symbol *sp; /* The look-ahead symbol */
343 enum e_action type;
344 union {
345 struct state *stp; /* The new state, if a shift */
346 struct rule *rp; /* The rule, if a reduce */
347 } x;
348 struct symbol *spOpt; /* SHIFTREDUCE optimization to this symbol */
349 struct action *next; /* Next action for this state */
350 struct action *collide; /* Next action with the same hash */
353 /* Each state of the generated parser's finite state machine
354 ** is encoded as an instance of the following structure. */
355 struct state {
356 struct config *bp; /* The basis configurations for this state */
357 struct config *cfp; /* All configurations in this set */
358 int statenum; /* Sequential number for this state */
359 struct action *ap; /* List of actions for this state */
360 int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
361 int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
362 int iDfltReduce; /* Default action is to REDUCE by this rule */
363 struct rule *pDfltReduce;/* The default REDUCE rule. */
364 int autoReduce; /* True if this is an auto-reduce state */
366 #define NO_OFFSET (-2147483647)
368 /* A followset propagation link indicates that the contents of one
369 ** configuration followset should be propagated to another whenever
370 ** the first changes. */
371 struct plink {
372 struct config *cfp; /* The configuration to which linked */
373 struct plink *next; /* The next propagate link */
376 /* The state vector for the entire parser generator is recorded as
377 ** follows. (LEMON uses no global variables and makes little use of
378 ** static variables. Fields in the following structure can be thought
379 ** of as begin global variables in the program.) */
380 struct lemon {
381 struct state **sorted; /* Table of states sorted by state number */
382 struct rule *rule; /* List of all rules */
383 struct rule *startRule; /* First rule */
384 int nstate; /* Number of states */
385 int nxstate; /* nstate with tail degenerate states removed */
386 int nrule; /* Number of rules */
387 int nsymbol; /* Number of terminal and nonterminal symbols */
388 int nterminal; /* Number of terminal symbols */
389 int minShiftReduce; /* Minimum shift-reduce action value */
390 int errAction; /* Error action value */
391 int accAction; /* Accept action value */
392 int noAction; /* No-op action value */
393 int minReduce; /* Minimum reduce action */
394 int maxAction; /* Maximum action value of any kind */
395 struct symbol **symbols; /* Sorted array of pointers to symbols */
396 int errorcnt; /* Number of errors */
397 struct symbol *errsym; /* The error symbol */
398 struct symbol *wildcard; /* Token that matches anything */
399 char *name; /* Name of the generated parser */
400 char *arg; /* Declaration of the 3th argument to parser */
401 char *ctx; /* Declaration of 2nd argument to constructor */
402 char *tokentype; /* Type of terminal symbols in the parser stack */
403 char *vartype; /* The default type of non-terminal symbols */
404 char *start; /* Name of the start symbol for the grammar */
405 char *stacksize; /* Size of the parser stack */
406 char *include; /* Code to put at the start of the C file */
407 char *error; /* Code to execute when an error is seen */
408 char *overflow; /* Code to execute on a stack overflow */
409 char *failure; /* Code to execute on parser failure */
410 char *accept; /* Code to execute when the parser excepts */
411 char *extracode; /* Code appended to the generated file */
412 char *tokendest; /* Code to execute to destroy token data */
413 char *vardest; /* Code for the default non-terminal destructor */
414 char *filename; /* Name of the input file */
415 char *outname; /* Name of the current output file */
416 char *tokenprefix; /* A prefix added to token names in the .h file */
417 int nconflict; /* Number of parsing conflicts */
418 int nactiontab; /* Number of entries in the yy_action[] table */
419 int nlookaheadtab; /* Number of entries in yy_lookahead[] */
420 int tablesize; /* Total table size of all tables in bytes */
421 int basisflag; /* Print only basis configurations */
422 int has_fallback; /* True if any %fallback is seen in the grammar */
423 int nolinenosflag; /* True if #line statements should not be printed */
424 char *argv0; /* Name of the program */
427 #define MemoryCheck(X) if((X)==0){ \
428 extern void memory_error(); \
429 memory_error(); \
432 /**************** From the file "table.h" *********************************/
434 ** All code in this file has been automatically generated
435 ** from a specification in the file
436 ** "table.q"
437 ** by the associative array code building program "aagen".
438 ** Do not edit this file! Instead, edit the specification
439 ** file, then rerun aagen.
442 ** Code for processing tables in the LEMON parser generator.
444 /* Routines for handling a strings */
446 const char *Strsafe(const char *);
448 void Strsafe_init(void);
449 int Strsafe_insert(const char *);
450 const char *Strsafe_find(const char *);
452 /* Routines for handling symbols of the grammar */
454 struct symbol *Symbol_new(const char *);
455 int Symbolcmpp(const void *, const void *);
456 void Symbol_init(void);
457 int Symbol_insert(struct symbol *, const char *);
458 struct symbol *Symbol_find(const char *);
459 struct symbol *Symbol_Nth(int);
460 int Symbol_count(void);
461 struct symbol **Symbol_arrayof(void);
463 /* Routines to manage the state table */
465 int Configcmp(const char *, const char *);
466 struct state *State_new(void);
467 void State_init(void);
468 int State_insert(struct state *, struct config *);
469 struct state *State_find(struct config *);
470 struct state **State_arrayof(void);
472 /* Routines used for efficiency in Configlist_add */
474 void Configtable_init(void);
475 int Configtable_insert(struct config *);
476 struct config *Configtable_find(struct config *);
477 void Configtable_clear(int(*)(struct config *));
479 /****************** From the file "action.c" *******************************/
481 ** Routines processing parser actions in the LEMON parser generator.
484 /* Allocate a new parser action */
485 static struct action *Action_new(void){
486 static struct action *freelist = 0;
487 struct action *newaction;
489 if( freelist==0 ){
490 int i;
491 int amt = 100;
492 freelist = (struct action *)calloc(amt, sizeof(struct action));
493 if( freelist==0 ){
494 fprintf(stderr,"Unable to allocate memory for a new parser action.");
495 exit(1);
497 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
498 freelist[amt-1].next = 0;
500 newaction = freelist;
501 freelist = freelist->next;
502 return newaction;
505 /* Compare two actions for sorting purposes. Return negative, zero, or
506 ** positive if the first action is less than, equal to, or greater than
507 ** the first
509 static int actioncmp(
510 struct action *ap1,
511 struct action *ap2
513 int rc;
514 rc = ap1->sp->index - ap2->sp->index;
515 if( rc==0 ){
516 rc = (int)ap1->type - (int)ap2->type;
518 if( rc==0 && (ap1->type==REDUCE || ap1->type==SHIFTREDUCE) ){
519 rc = ap1->x.rp->index - ap2->x.rp->index;
521 if( rc==0 ){
522 rc = (int) (ap2 - ap1);
524 return rc;
527 /* Sort parser actions */
528 static struct action *Action_sort(
529 struct action *ap
531 ap = (struct action *)msort((char *)ap,(char **)&ap->next,
532 (int(*)(const char*,const char*))actioncmp);
533 return ap;
536 void Action_add(
537 struct action **app,
538 enum e_action type,
539 struct symbol *sp,
540 char *arg
542 struct action *newaction;
543 newaction = Action_new();
544 newaction->next = *app;
545 *app = newaction;
546 newaction->type = type;
547 newaction->sp = sp;
548 newaction->spOpt = 0;
549 if( type==SHIFT ){
550 newaction->x.stp = (struct state *)arg;
551 }else{
552 newaction->x.rp = (struct rule *)arg;
555 /********************** New code to implement the "acttab" module ***********/
557 ** This module implements routines use to construct the yy_action[] table.
561 ** The state of the yy_action table under construction is an instance of
562 ** the following structure.
564 ** The yy_action table maps the pair (state_number, lookahead) into an
565 ** action_number. The table is an array of integers pairs. The state_number
566 ** determines an initial offset into the yy_action array. The lookahead
567 ** value is then added to this initial offset to get an index X into the
568 ** yy_action array. If the aAction[X].lookahead equals the value of the
569 ** of the lookahead input, then the value of the action_number output is
570 ** aAction[X].action. If the lookaheads do not match then the
571 ** default action for the state_number is returned.
573 ** All actions associated with a single state_number are first entered
574 ** into aLookahead[] using multiple calls to acttab_action(). Then the
575 ** actions for that single state_number are placed into the aAction[]
576 ** array with a single call to acttab_insert(). The acttab_insert() call
577 ** also resets the aLookahead[] array in preparation for the next
578 ** state number.
580 struct lookahead_action {
581 int lookahead; /* Value of the lookahead token */
582 int action; /* Action to take on the given lookahead */
584 typedef struct acttab acttab;
585 struct acttab {
586 int nAction; /* Number of used slots in aAction[] */
587 int nActionAlloc; /* Slots allocated for aAction[] */
588 struct lookahead_action
589 *aAction, /* The yy_action[] table under construction */
590 *aLookahead; /* A single new transaction set */
591 int mnLookahead; /* Minimum aLookahead[].lookahead */
592 int mnAction; /* Action associated with mnLookahead */
593 int mxLookahead; /* Maximum aLookahead[].lookahead */
594 int nLookahead; /* Used slots in aLookahead[] */
595 int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
596 int nterminal; /* Number of terminal symbols */
597 int nsymbol; /* total number of symbols */
600 /* Return the number of entries in the yy_action table */
601 #define acttab_lookahead_size(X) ((X)->nAction)
603 /* The value for the N-th entry in yy_action */
604 #define acttab_yyaction(X,N) ((X)->aAction[N].action)
606 /* The value for the N-th entry in yy_lookahead */
607 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
609 /* Free all memory associated with the given acttab */
610 void acttab_free(acttab *p){
611 free( p->aAction );
612 free( p->aLookahead );
613 free( p );
616 /* Allocate a new acttab structure */
617 acttab *acttab_alloc(int nsymbol, int nterminal){
618 acttab *p = (acttab *) calloc( 1, sizeof(*p) );
619 if( p==0 ){
620 fprintf(stderr,"Unable to allocate memory for a new acttab.");
621 exit(1);
623 memset(p, 0, sizeof(*p));
624 p->nsymbol = nsymbol;
625 p->nterminal = nterminal;
626 return p;
629 /* Add a new action to the current transaction set.
631 ** This routine is called once for each lookahead for a particular
632 ** state.
634 void acttab_action(acttab *p, int lookahead, int action){
635 if( p->nLookahead>=p->nLookaheadAlloc ){
636 p->nLookaheadAlloc += 25;
637 p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead,
638 sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
639 if( p->aLookahead==0 ){
640 fprintf(stderr,"malloc failed\n");
641 exit(1);
644 if( p->nLookahead==0 ){
645 p->mxLookahead = lookahead;
646 p->mnLookahead = lookahead;
647 p->mnAction = action;
648 }else{
649 if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
650 if( p->mnLookahead>lookahead ){
651 p->mnLookahead = lookahead;
652 p->mnAction = action;
655 p->aLookahead[p->nLookahead].lookahead = lookahead;
656 p->aLookahead[p->nLookahead].action = action;
657 p->nLookahead++;
661 ** Add the transaction set built up with prior calls to acttab_action()
662 ** into the current action table. Then reset the transaction set back
663 ** to an empty set in preparation for a new round of acttab_action() calls.
665 ** Return the offset into the action table of the new transaction.
667 ** If the makeItSafe parameter is true, then the offset is chosen so that
668 ** it is impossible to overread the yy_lookaside[] table regardless of
669 ** the lookaside token. This is done for the terminal symbols, as they
670 ** come from external inputs and can contain syntax errors. When makeItSafe
671 ** is false, there is more flexibility in selecting offsets, resulting in
672 ** a smaller table. For non-terminal symbols, which are never syntax errors,
673 ** makeItSafe can be false.
675 int acttab_insert(acttab *p, int makeItSafe){
676 int i, j, k, n, end;
677 assert( p->nLookahead>0 );
679 /* Make sure we have enough space to hold the expanded action table
680 ** in the worst case. The worst case occurs if the transaction set
681 ** must be appended to the current action table
683 n = p->nsymbol + 1;
684 if( p->nAction + n >= p->nActionAlloc ){
685 int oldAlloc = p->nActionAlloc;
686 p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
687 p->aAction = (struct lookahead_action *) realloc( p->aAction,
688 sizeof(p->aAction[0])*p->nActionAlloc);
689 if( p->aAction==0 ){
690 fprintf(stderr,"malloc failed\n");
691 exit(1);
693 for(i=oldAlloc; i<p->nActionAlloc; i++){
694 p->aAction[i].lookahead = -1;
695 p->aAction[i].action = -1;
699 /* Scan the existing action table looking for an offset that is a
700 ** duplicate of the current transaction set. Fall out of the loop
701 ** if and when the duplicate is found.
703 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
705 end = makeItSafe ? p->mnLookahead : 0;
706 for(i=p->nAction-1; i>=end; i--){
707 if( p->aAction[i].lookahead==p->mnLookahead ){
708 /* All lookaheads and actions in the aLookahead[] transaction
709 ** must match against the candidate aAction[i] entry. */
710 if( p->aAction[i].action!=p->mnAction ) continue;
711 for(j=0; j<p->nLookahead; j++){
712 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
713 if( k<0 || k>=p->nAction ) break;
714 if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
715 if( p->aLookahead[j].action!=p->aAction[k].action ) break;
717 if( j<p->nLookahead ) continue;
719 /* No possible lookahead value that is not in the aLookahead[]
720 ** transaction is allowed to match aAction[i] */
721 n = 0;
722 for(j=0; j<p->nAction; j++){
723 if( p->aAction[j].lookahead<0 ) continue;
724 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
726 if( n==p->nLookahead ){
727 break; /* An exact match is found at offset i */
732 /* If no existing offsets exactly match the current transaction, find an
733 ** an empty offset in the aAction[] table in which we can add the
734 ** aLookahead[] transaction.
736 if( i<end ){
737 /* Look for holes in the aAction[] table that fit the current
738 ** aLookahead[] transaction. Leave i set to the offset of the hole.
739 ** If no holes are found, i is left at p->nAction, which means the
740 ** transaction will be appended. */
741 i = makeItSafe ? p->mnLookahead : 0;
742 for(; i<p->nActionAlloc - p->mxLookahead; i++){
743 if( p->aAction[i].lookahead<0 ){
744 for(j=0; j<p->nLookahead; j++){
745 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
746 if( k<0 ) break;
747 if( p->aAction[k].lookahead>=0 ) break;
749 if( j<p->nLookahead ) continue;
750 for(j=0; j<p->nAction; j++){
751 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
753 if( j==p->nAction ){
754 break; /* Fits in empty slots */
759 /* Insert transaction set at index i. */
760 #if 0
761 printf("Acttab:");
762 for(j=0; j<p->nLookahead; j++){
763 printf(" %d", p->aLookahead[j].lookahead);
765 printf(" inserted at %d\n", i);
766 #endif
767 for(j=0; j<p->nLookahead; j++){
768 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
769 p->aAction[k] = p->aLookahead[j];
770 if( k>=p->nAction ) p->nAction = k+1;
772 if( makeItSafe && i+p->nterminal>=p->nAction ) p->nAction = i+p->nterminal+1;
773 p->nLookahead = 0;
775 /* Return the offset that is added to the lookahead in order to get the
776 ** index into yy_action of the action */
777 return i - p->mnLookahead;
781 ** Return the size of the action table without the trailing syntax error
782 ** entries.
784 int acttab_action_size(acttab *p){
785 int n = p->nAction;
786 while( n>0 && p->aAction[n-1].lookahead<0 ){ n--; }
787 return n;
790 /********************** From the file "build.c" *****************************/
792 ** Routines to construction the finite state machine for the LEMON
793 ** parser generator.
796 /* Find a precedence symbol of every rule in the grammar.
798 ** Those rules which have a precedence symbol coded in the input
799 ** grammar using the "[symbol]" construct will already have the
800 ** rp->precsym field filled. Other rules take as their precedence
801 ** symbol the first RHS symbol with a defined precedence. If there
802 ** are not RHS symbols with a defined precedence, the precedence
803 ** symbol field is left blank.
805 void FindRulePrecedences(struct lemon *xp)
807 struct rule *rp;
808 for(rp=xp->rule; rp; rp=rp->next){
809 if( rp->precsym==0 ){
810 int i, j;
811 for(i=0; i<rp->nrhs && rp->precsym==0; i++){
812 struct symbol *sp = rp->rhs[i];
813 if( sp->type==MULTITERMINAL ){
814 for(j=0; j<sp->nsubsym; j++){
815 if( sp->subsym[j]->prec>=0 ){
816 rp->precsym = sp->subsym[j];
817 break;
820 }else if( sp->prec>=0 ){
821 rp->precsym = rp->rhs[i];
826 return;
829 /* Find all nonterminals which will generate the empty string.
830 ** Then go back and compute the first sets of every nonterminal.
831 ** The first set is the set of all terminal symbols which can begin
832 ** a string generated by that nonterminal.
834 void FindFirstSets(struct lemon *lemp)
836 int i, j;
837 struct rule *rp;
838 int progress;
840 for(i=0; i<lemp->nsymbol; i++){
841 lemp->symbols[i]->lambda = LEMON_FALSE;
843 for(i=lemp->nterminal; i<lemp->nsymbol; i++){
844 lemp->symbols[i]->firstset = SetNew();
847 /* First compute all lambdas */
849 progress = 0;
850 for(rp=lemp->rule; rp; rp=rp->next){
851 if( rp->lhs->lambda ) continue;
852 for(i=0; i<rp->nrhs; i++){
853 struct symbol *sp = rp->rhs[i];
854 assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE );
855 if( sp->lambda==LEMON_FALSE ) break;
857 if( i==rp->nrhs ){
858 rp->lhs->lambda = LEMON_TRUE;
859 progress = 1;
862 }while( progress );
864 /* Now compute all first sets */
866 struct symbol *s1, *s2;
867 progress = 0;
868 for(rp=lemp->rule; rp; rp=rp->next){
869 s1 = rp->lhs;
870 for(i=0; i<rp->nrhs; i++){
871 s2 = rp->rhs[i];
872 if( s2->type==TERMINAL ){
873 progress += SetAdd(s1->firstset,s2->index);
874 break;
875 }else if( s2->type==MULTITERMINAL ){
876 for(j=0; j<s2->nsubsym; j++){
877 progress += SetAdd(s1->firstset,s2->subsym[j]->index);
879 break;
880 }else if( s1==s2 ){
881 if( s1->lambda==LEMON_FALSE ) break;
882 }else{
883 progress += SetUnion(s1->firstset,s2->firstset);
884 if( s2->lambda==LEMON_FALSE ) break;
888 }while( progress );
889 return;
892 /* Compute all LR(0) states for the grammar. Links
893 ** are added to between some states so that the LR(1) follow sets
894 ** can be computed later.
896 PRIVATE struct state *getstate(struct lemon *); /* forward reference */
897 void FindStates(struct lemon *lemp)
899 struct symbol *sp;
900 struct rule *rp;
902 Configlist_init();
904 /* Find the start symbol */
905 if( lemp->start ){
906 sp = Symbol_find(lemp->start);
907 if( sp==0 ){
908 ErrorMsg(lemp->filename,0,
909 "The specified start symbol \"%s\" is not \
910 in a nonterminal of the grammar. \"%s\" will be used as the start \
911 symbol instead.",lemp->start,lemp->startRule->lhs->name);
912 lemp->errorcnt++;
913 sp = lemp->startRule->lhs;
915 }else{
916 sp = lemp->startRule->lhs;
919 /* Make sure the start symbol doesn't occur on the right-hand side of
920 ** any rule. Report an error if it does. (YACC would generate a new
921 ** start symbol in this case.) */
922 for(rp=lemp->rule; rp; rp=rp->next){
923 int i;
924 for(i=0; i<rp->nrhs; i++){
925 if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
926 ErrorMsg(lemp->filename,0,
927 "The start symbol \"%s\" occurs on the \
928 right-hand side of a rule. This will result in a parser which \
929 does not work properly.",sp->name);
930 lemp->errorcnt++;
935 /* The basis configuration set for the first state
936 ** is all rules which have the start symbol as their
937 ** left-hand side */
938 for(rp=sp->rule; rp; rp=rp->nextlhs){
939 struct config *newcfp;
940 rp->lhsStart = 1;
941 newcfp = Configlist_addbasis(rp,0);
942 SetAdd(newcfp->fws,0);
945 /* Compute the first state. All other states will be
946 ** computed automatically during the computation of the first one.
947 ** The returned pointer to the first state is not used. */
948 (void)getstate(lemp);
949 return;
952 /* Return a pointer to a state which is described by the configuration
953 ** list which has been built from calls to Configlist_add.
955 PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */
956 PRIVATE struct state *getstate(struct lemon *lemp)
958 struct config *cfp, *bp;
959 struct state *stp;
961 /* Extract the sorted basis of the new state. The basis was constructed
962 ** by prior calls to "Configlist_addbasis()". */
963 Configlist_sortbasis();
964 bp = Configlist_basis();
966 /* Get a state with the same basis */
967 stp = State_find(bp);
968 if( stp ){
969 /* A state with the same basis already exists! Copy all the follow-set
970 ** propagation links from the state under construction into the
971 ** preexisting state, then return a pointer to the preexisting state */
972 struct config *x, *y;
973 for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
974 Plink_copy(&y->bplp,x->bplp);
975 Plink_delete(x->fplp);
976 x->fplp = x->bplp = 0;
978 cfp = Configlist_return();
979 Configlist_eat(cfp);
980 }else{
981 /* This really is a new state. Construct all the details */
982 Configlist_closure(lemp); /* Compute the configuration closure */
983 Configlist_sort(); /* Sort the configuration closure */
984 cfp = Configlist_return(); /* Get a pointer to the config list */
985 stp = State_new(); /* A new state structure */
986 MemoryCheck(stp);
987 stp->bp = bp; /* Remember the configuration basis */
988 stp->cfp = cfp; /* Remember the configuration closure */
989 stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
990 stp->ap = 0; /* No actions, yet. */
991 State_insert(stp,stp->bp); /* Add to the state table */
992 buildshifts(lemp,stp); /* Recursively compute successor states */
994 return stp;
998 ** Return true if two symbols are the same.
1000 int same_symbol(struct symbol *a, struct symbol *b)
1002 int i;
1003 if( a==b ) return 1;
1004 if( a->type!=MULTITERMINAL ) return 0;
1005 if( b->type!=MULTITERMINAL ) return 0;
1006 if( a->nsubsym!=b->nsubsym ) return 0;
1007 for(i=0; i<a->nsubsym; i++){
1008 if( a->subsym[i]!=b->subsym[i] ) return 0;
1010 return 1;
1013 /* Construct all successor states to the given state. A "successor"
1014 ** state is any state which can be reached by a shift action.
1016 PRIVATE void buildshifts(struct lemon *lemp, struct state *stp)
1018 struct config *cfp; /* For looping thru the config closure of "stp" */
1019 struct config *bcfp; /* For the inner loop on config closure of "stp" */
1020 struct config *newcfg; /* */
1021 struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
1022 struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
1023 struct state *newstp; /* A pointer to a successor state */
1025 /* Each configuration becomes complete after it contibutes to a successor
1026 ** state. Initially, all configurations are incomplete */
1027 for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
1029 /* Loop through all configurations of the state "stp" */
1030 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1031 if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
1032 if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
1033 Configlist_reset(); /* Reset the new config set */
1034 sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
1036 /* For every configuration in the state "stp" which has the symbol "sp"
1037 ** following its dot, add the same configuration to the basis set under
1038 ** construction but with the dot shifted one symbol to the right. */
1039 for(bcfp=cfp; bcfp; bcfp=bcfp->next){
1040 if( bcfp->status==COMPLETE ) continue; /* Already used */
1041 if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
1042 bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
1043 if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
1044 bcfp->status = COMPLETE; /* Mark this config as used */
1045 newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
1046 Plink_add(&newcfg->bplp,bcfp);
1049 /* Get a pointer to the state described by the basis configuration set
1050 ** constructed in the preceding loop */
1051 newstp = getstate(lemp);
1053 /* The state "newstp" is reached from the state "stp" by a shift action
1054 ** on the symbol "sp" */
1055 if( sp->type==MULTITERMINAL ){
1056 int i;
1057 for(i=0; i<sp->nsubsym; i++){
1058 Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
1060 }else{
1061 Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
1067 ** Construct the propagation links
1069 void FindLinks(struct lemon *lemp)
1071 int i;
1072 struct config *cfp, *other;
1073 struct state *stp;
1074 struct plink *plp;
1076 /* Housekeeping detail:
1077 ** Add to every propagate link a pointer back to the state to
1078 ** which the link is attached. */
1079 for(i=0; i<lemp->nstate; i++){
1080 stp = lemp->sorted[i];
1081 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1082 cfp->stp = stp;
1086 /* Convert all backlinks into forward links. Only the forward
1087 ** links are used in the follow-set computation. */
1088 for(i=0; i<lemp->nstate; i++){
1089 stp = lemp->sorted[i];
1090 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1091 for(plp=cfp->bplp; plp; plp=plp->next){
1092 other = plp->cfp;
1093 Plink_add(&other->fplp,cfp);
1099 /* Compute all followsets.
1101 ** A followset is the set of all symbols which can come immediately
1102 ** after a configuration.
1104 void FindFollowSets(struct lemon *lemp)
1106 int i;
1107 struct config *cfp;
1108 struct plink *plp;
1109 int progress;
1110 int change;
1112 for(i=0; i<lemp->nstate; i++){
1113 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1114 cfp->status = INCOMPLETE;
1119 progress = 0;
1120 for(i=0; i<lemp->nstate; i++){
1121 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1122 if( cfp->status==COMPLETE ) continue;
1123 for(plp=cfp->fplp; plp; plp=plp->next){
1124 change = SetUnion(plp->cfp->fws,cfp->fws);
1125 if( change ){
1126 plp->cfp->status = INCOMPLETE;
1127 progress = 1;
1130 cfp->status = COMPLETE;
1133 }while( progress );
1136 static int resolve_conflict(struct action *,struct action *);
1138 /* Compute the reduce actions, and resolve conflicts.
1140 void FindActions(struct lemon *lemp)
1142 int i,j;
1143 struct config *cfp;
1144 struct state *stp;
1145 struct symbol *sp;
1146 struct rule *rp;
1148 /* Add all of the reduce actions
1149 ** A reduce action is added for each element of the followset of
1150 ** a configuration which has its dot at the extreme right.
1152 for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
1153 stp = lemp->sorted[i];
1154 for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
1155 if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
1156 for(j=0; j<lemp->nterminal; j++){
1157 if( SetFind(cfp->fws,j) ){
1158 /* Add a reduce action to the state "stp" which will reduce by the
1159 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
1160 Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
1167 /* Add the accepting token */
1168 if( lemp->start ){
1169 sp = Symbol_find(lemp->start);
1170 if( sp==0 ) sp = lemp->startRule->lhs;
1171 }else{
1172 sp = lemp->startRule->lhs;
1174 /* Add to the first state (which is always the starting state of the
1175 ** finite state machine) an action to ACCEPT if the lookahead is the
1176 ** start nonterminal. */
1177 Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
1179 /* Resolve conflicts */
1180 for(i=0; i<lemp->nstate; i++){
1181 struct action *ap, *nap;
1182 stp = lemp->sorted[i];
1183 /* assert( stp->ap ); */
1184 stp->ap = Action_sort(stp->ap);
1185 for(ap=stp->ap; ap && ap->next; ap=ap->next){
1186 for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
1187 /* The two actions "ap" and "nap" have the same lookahead.
1188 ** Figure out which one should be used */
1189 lemp->nconflict += resolve_conflict(ap,nap);
1194 /* Report an error for each rule that can never be reduced. */
1195 for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
1196 for(i=0; i<lemp->nstate; i++){
1197 struct action *ap;
1198 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
1199 if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
1202 for(rp=lemp->rule; rp; rp=rp->next){
1203 if( rp->canReduce ) continue;
1204 ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
1205 lemp->errorcnt++;
1209 /* Resolve a conflict between the two given actions. If the
1210 ** conflict can't be resolved, return non-zero.
1212 ** NO LONGER TRUE:
1213 ** To resolve a conflict, first look to see if either action
1214 ** is on an error rule. In that case, take the action which
1215 ** is not associated with the error rule. If neither or both
1216 ** actions are associated with an error rule, then try to
1217 ** use precedence to resolve the conflict.
1219 ** If either action is a SHIFT, then it must be apx. This
1220 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1222 static int resolve_conflict(
1223 struct action *apx,
1224 struct action *apy
1226 struct symbol *spx, *spy;
1227 int errcnt = 0;
1228 assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
1229 if( apx->type==SHIFT && apy->type==SHIFT ){
1230 apy->type = SSCONFLICT;
1231 errcnt++;
1233 if( apx->type==SHIFT && apy->type==REDUCE ){
1234 spx = apx->sp;
1235 spy = apy->x.rp->precsym;
1236 if( spy==0 || spx->prec<0 || spy->prec<0 ){
1237 /* Not enough precedence information. */
1238 apy->type = SRCONFLICT;
1239 errcnt++;
1240 }else if( spx->prec>spy->prec ){ /* higher precedence wins */
1241 apy->type = RD_RESOLVED;
1242 }else if( spx->prec<spy->prec ){
1243 apx->type = SH_RESOLVED;
1244 }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1245 apy->type = RD_RESOLVED; /* associativity */
1246 }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
1247 apx->type = SH_RESOLVED;
1248 }else{
1249 assert( spx->prec==spy->prec && spx->assoc==NONE );
1250 apx->type = ERROR;
1252 }else if( apx->type==REDUCE && apy->type==REDUCE ){
1253 spx = apx->x.rp->precsym;
1254 spy = apy->x.rp->precsym;
1255 if( spx==0 || spy==0 || spx->prec<0 ||
1256 spy->prec<0 || spx->prec==spy->prec ){
1257 apy->type = RRCONFLICT;
1258 errcnt++;
1259 }else if( spx->prec>spy->prec ){
1260 apy->type = RD_RESOLVED;
1261 }else if( spx->prec<spy->prec ){
1262 apx->type = RD_RESOLVED;
1264 }else{
1265 assert(
1266 apx->type==SH_RESOLVED ||
1267 apx->type==RD_RESOLVED ||
1268 apx->type==SSCONFLICT ||
1269 apx->type==SRCONFLICT ||
1270 apx->type==RRCONFLICT ||
1271 apy->type==SH_RESOLVED ||
1272 apy->type==RD_RESOLVED ||
1273 apy->type==SSCONFLICT ||
1274 apy->type==SRCONFLICT ||
1275 apy->type==RRCONFLICT
1277 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1278 ** REDUCEs on the list. If we reach this point it must be because
1279 ** the parser conflict had already been resolved. */
1281 return errcnt;
1283 /********************* From the file "configlist.c" *************************/
1285 ** Routines to processing a configuration list and building a state
1286 ** in the LEMON parser generator.
1289 static struct config *freelist = 0; /* List of free configurations */
1290 static struct config *current = 0; /* Top of list of configurations */
1291 static struct config **currentend = 0; /* Last on list of configs */
1292 static struct config *basis = 0; /* Top of list of basis configs */
1293 static struct config **basisend = 0; /* End of list of basis configs */
1295 /* Return a pointer to a new configuration */
1296 PRIVATE struct config *newconfig(void){
1297 struct config *newcfg;
1298 if( freelist==0 ){
1299 int i;
1300 int amt = 3;
1301 freelist = (struct config *)calloc( amt, sizeof(struct config) );
1302 if( freelist==0 ){
1303 fprintf(stderr,"Unable to allocate memory for a new configuration.");
1304 exit(1);
1306 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1307 freelist[amt-1].next = 0;
1309 newcfg = freelist;
1310 freelist = freelist->next;
1311 return newcfg;
1314 /* The configuration "old" is no longer used */
1315 PRIVATE void deleteconfig(struct config *old)
1317 old->next = freelist;
1318 freelist = old;
1321 /* Initialized the configuration list builder */
1322 void Configlist_init(void){
1323 current = 0;
1324 currentend = &current;
1325 basis = 0;
1326 basisend = &basis;
1327 Configtable_init();
1328 return;
1331 /* Initialized the configuration list builder */
1332 void Configlist_reset(void){
1333 current = 0;
1334 currentend = &current;
1335 basis = 0;
1336 basisend = &basis;
1337 Configtable_clear(0);
1338 return;
1341 /* Add another configuration to the configuration list */
1342 struct config *Configlist_add(
1343 struct rule *rp, /* The rule */
1344 int dot /* Index into the RHS of the rule where the dot goes */
1346 struct config *cfp, model;
1348 assert( currentend!=0 );
1349 model.rp = rp;
1350 model.dot = dot;
1351 cfp = Configtable_find(&model);
1352 if( cfp==0 ){
1353 cfp = newconfig();
1354 cfp->rp = rp;
1355 cfp->dot = dot;
1356 cfp->fws = SetNew();
1357 cfp->stp = 0;
1358 cfp->fplp = cfp->bplp = 0;
1359 cfp->next = 0;
1360 cfp->bp = 0;
1361 *currentend = cfp;
1362 currentend = &cfp->next;
1363 Configtable_insert(cfp);
1365 return cfp;
1368 /* Add a basis configuration to the configuration list */
1369 struct config *Configlist_addbasis(struct rule *rp, int dot)
1371 struct config *cfp, model;
1373 assert( basisend!=0 );
1374 assert( currentend!=0 );
1375 model.rp = rp;
1376 model.dot = dot;
1377 cfp = Configtable_find(&model);
1378 if( cfp==0 ){
1379 cfp = newconfig();
1380 cfp->rp = rp;
1381 cfp->dot = dot;
1382 cfp->fws = SetNew();
1383 cfp->stp = 0;
1384 cfp->fplp = cfp->bplp = 0;
1385 cfp->next = 0;
1386 cfp->bp = 0;
1387 *currentend = cfp;
1388 currentend = &cfp->next;
1389 *basisend = cfp;
1390 basisend = &cfp->bp;
1391 Configtable_insert(cfp);
1393 return cfp;
1396 /* Compute the closure of the configuration list */
1397 void Configlist_closure(struct lemon *lemp)
1399 struct config *cfp, *newcfp;
1400 struct rule *rp, *newrp;
1401 struct symbol *sp, *xsp;
1402 int i, dot;
1404 assert( currentend!=0 );
1405 for(cfp=current; cfp; cfp=cfp->next){
1406 rp = cfp->rp;
1407 dot = cfp->dot;
1408 if( dot>=rp->nrhs ) continue;
1409 sp = rp->rhs[dot];
1410 if( sp->type==NONTERMINAL ){
1411 if( sp->rule==0 && sp!=lemp->errsym ){
1412 ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1413 sp->name);
1414 lemp->errorcnt++;
1416 for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1417 newcfp = Configlist_add(newrp,0);
1418 for(i=dot+1; i<rp->nrhs; i++){
1419 xsp = rp->rhs[i];
1420 if( xsp->type==TERMINAL ){
1421 SetAdd(newcfp->fws,xsp->index);
1422 break;
1423 }else if( xsp->type==MULTITERMINAL ){
1424 int k;
1425 for(k=0; k<xsp->nsubsym; k++){
1426 SetAdd(newcfp->fws, xsp->subsym[k]->index);
1428 break;
1429 }else{
1430 SetUnion(newcfp->fws,xsp->firstset);
1431 if( xsp->lambda==LEMON_FALSE ) break;
1434 if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1438 return;
1441 /* Sort the configuration list */
1442 void Configlist_sort(void){
1443 current = (struct config*)msort((char*)current,(char**)&(current->next),
1444 Configcmp);
1445 currentend = 0;
1446 return;
1449 /* Sort the basis configuration list */
1450 void Configlist_sortbasis(void){
1451 basis = (struct config*)msort((char*)current,(char**)&(current->bp),
1452 Configcmp);
1453 basisend = 0;
1454 return;
1457 /* Return a pointer to the head of the configuration list and
1458 ** reset the list */
1459 struct config *Configlist_return(void){
1460 struct config *old;
1461 old = current;
1462 current = 0;
1463 currentend = 0;
1464 return old;
1467 /* Return a pointer to the head of the configuration list and
1468 ** reset the list */
1469 struct config *Configlist_basis(void){
1470 struct config *old;
1471 old = basis;
1472 basis = 0;
1473 basisend = 0;
1474 return old;
1477 /* Free all elements of the given configuration list */
1478 void Configlist_eat(struct config *cfp)
1480 struct config *nextcfp;
1481 for(; cfp; cfp=nextcfp){
1482 nextcfp = cfp->next;
1483 assert( cfp->fplp==0 );
1484 assert( cfp->bplp==0 );
1485 if( cfp->fws ) SetFree(cfp->fws);
1486 deleteconfig(cfp);
1488 return;
1490 /***************** From the file "error.c" *********************************/
1492 ** Code for printing error message.
1495 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1496 va_list ap;
1497 fprintf(stderr, "%s:%d: ", filename, lineno);
1498 va_start(ap, format);
1499 vfprintf(stderr,format,ap);
1500 va_end(ap);
1501 fprintf(stderr, "\n");
1503 /**************** From the file "main.c" ************************************/
1505 ** Main program file for the LEMON parser generator.
1508 /* Report an out-of-memory condition and abort. This function
1509 ** is used mostly by the "MemoryCheck" macro in struct.h
1511 void memory_error(void){
1512 fprintf(stderr,"Out of memory. Aborting...\n");
1513 exit(1);
1516 static int nDefine = 0; /* Number of -D options on the command line */
1517 static char **azDefine = 0; /* Name of the -D macros */
1519 /* This routine is called with the argument to each -D command-line option.
1520 ** Add the macro defined to the azDefine array.
1522 static void handle_D_option(char *z){
1523 char **paz;
1524 nDefine++;
1525 azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine);
1526 if( azDefine==0 ){
1527 fprintf(stderr,"out of memory\n");
1528 exit(1);
1530 paz = &azDefine[nDefine-1];
1531 *paz = (char *) malloc( lemonStrlen(z)+1 );
1532 if( *paz==0 ){
1533 fprintf(stderr,"out of memory\n");
1534 exit(1);
1536 lemon_strcpy(*paz, z);
1537 for(z=*paz; *z && *z!='='; z++){}
1538 *z = 0;
1541 /* Rember the name of the output directory
1543 static char *outputDir = NULL;
1544 static void handle_d_option(char *z){
1545 outputDir = (char *) malloc( lemonStrlen(z)+1 );
1546 if( outputDir==0 ){
1547 fprintf(stderr,"out of memory\n");
1548 exit(1);
1550 lemon_strcpy(outputDir, z);
1553 static char *user_templatename = NULL;
1554 static void handle_T_option(char *z){
1555 user_templatename = (char *) malloc( lemonStrlen(z)+1 );
1556 if( user_templatename==0 ){
1557 memory_error();
1559 lemon_strcpy(user_templatename, z);
1562 /* Merge together to lists of rules ordered by rule.iRule */
1563 static struct rule *Rule_merge(struct rule *pA, struct rule *pB){
1564 struct rule *pFirst = 0;
1565 struct rule **ppPrev = &pFirst;
1566 while( pA && pB ){
1567 if( pA->iRule<pB->iRule ){
1568 *ppPrev = pA;
1569 ppPrev = &pA->next;
1570 pA = pA->next;
1571 }else{
1572 *ppPrev = pB;
1573 ppPrev = &pB->next;
1574 pB = pB->next;
1577 if( pA ){
1578 *ppPrev = pA;
1579 }else{
1580 *ppPrev = pB;
1582 return pFirst;
1586 ** Sort a list of rules in order of increasing iRule value
1588 static struct rule *Rule_sort(struct rule *rp){
1589 int i;
1590 struct rule *pNext;
1591 struct rule *x[32];
1592 memset(x, 0, sizeof(x));
1593 while( rp ){
1594 pNext = rp->next;
1595 rp->next = 0;
1596 for(i=0; i<sizeof(x)/sizeof(x[0]) && x[i]; i++){
1597 rp = Rule_merge(x[i], rp);
1598 x[i] = 0;
1600 x[i] = rp;
1601 rp = pNext;
1603 rp = 0;
1604 for(i=0; i<sizeof(x)/sizeof(x[0]); i++){
1605 rp = Rule_merge(x[i], rp);
1607 return rp;
1610 /* forward reference */
1611 static const char *minimum_size_type(int lwr, int upr, int *pnByte);
1613 /* Print a single line of the "Parser Stats" output
1615 static void stats_line(const char *zLabel, int iValue){
1616 int nLabel = lemonStrlen(zLabel);
1617 printf(" %s%.*s %5d\n", zLabel,
1618 35-nLabel, "................................",
1619 iValue);
1622 /* The main program. Parse the command line and do it... */
1623 int main(int argc, char **argv)
1625 static int version = 0;
1626 static int rpflag = 0;
1627 static int basisflag = 0;
1628 static int compress = 0;
1629 static int quiet = 0;
1630 static int statistics = 0;
1631 static int mhflag = 0;
1632 static int nolinenosflag = 0;
1633 static int noResort = 0;
1635 static struct s_options options[] = {
1636 {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1637 {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1638 {OPT_FSTR, "d", (char*)&handle_d_option, "Output directory. Default '.'"},
1639 {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
1640 {OPT_FSTR, "f", 0, "Ignored. (Placeholder for -f compiler options.)"},
1641 {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1642 {OPT_FSTR, "I", 0, "Ignored. (Placeholder for '-I' compiler options.)"},
1643 {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."},
1644 {OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."},
1645 {OPT_FSTR, "O", 0, "Ignored. (Placeholder for '-O' compiler options.)"},
1646 {OPT_FLAG, "p", (char*)&showPrecedenceConflict,
1647 "Show conflicts resolved by precedence rules"},
1648 {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1649 {OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"},
1650 {OPT_FLAG, "s", (char*)&statistics,
1651 "Print parser stats to standard output."},
1652 {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1653 {OPT_FSTR, "T", (char*)handle_T_option, "Specify a template file."},
1654 {OPT_FSTR, "W", 0, "Ignored. (Placeholder for '-W' compiler options.)"},
1655 {OPT_FLAG,0,0,0}
1657 int i;
1658 int exitcode;
1659 struct lemon lem;
1660 struct rule *rp;
1662 OptInit(argv,options,stderr);
1663 if( version ){
1664 printf("Lemon version 1.0\n");
1665 exit(0);
1667 if( OptNArgs()!=1 ){
1668 fprintf(stderr,"Exactly one filename argument is required.\n");
1669 exit(1);
1671 memset(&lem, 0, sizeof(lem));
1672 lem.errorcnt = 0;
1674 /* Initialize the machine */
1675 Strsafe_init();
1676 Symbol_init();
1677 State_init();
1678 lem.argv0 = argv[0];
1679 lem.filename = OptArg(0);
1680 lem.basisflag = basisflag;
1681 lem.nolinenosflag = nolinenosflag;
1682 Symbol_new("$");
1684 /* Parse the input file */
1685 Parse(&lem);
1686 if( lem.errorcnt ) exit(lem.errorcnt);
1687 if( lem.nrule==0 ){
1688 fprintf(stderr,"Empty grammar.\n");
1689 exit(1);
1691 lem.errsym = Symbol_find("error");
1693 /* Count and index the symbols of the grammar */
1694 Symbol_new("{default}");
1695 lem.nsymbol = Symbol_count();
1696 lem.symbols = Symbol_arrayof();
1697 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1698 qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp);
1699 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1700 while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; }
1701 assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 );
1702 lem.nsymbol = i - 1;
1703 for(i=1; ISUPPER(lem.symbols[i]->name[0]); i++);
1704 lem.nterminal = i;
1706 /* Assign sequential rule numbers. Start with 0. Put rules that have no
1707 ** reduce action C-code associated with them last, so that the switch()
1708 ** statement that selects reduction actions will have a smaller jump table.
1710 for(i=0, rp=lem.rule; rp; rp=rp->next){
1711 rp->iRule = rp->code ? i++ : -1;
1713 for(rp=lem.rule; rp; rp=rp->next){
1714 if( rp->iRule<0 ) rp->iRule = i++;
1716 lem.startRule = lem.rule;
1717 lem.rule = Rule_sort(lem.rule);
1719 /* Generate a reprint of the grammar, if requested on the command line */
1720 if( rpflag ){
1721 Reprint(&lem);
1722 }else{
1723 /* Initialize the size for all follow and first sets */
1724 SetSize(lem.nterminal+1);
1726 /* Find the precedence for every production rule (that has one) */
1727 FindRulePrecedences(&lem);
1729 /* Compute the lambda-nonterminals and the first-sets for every
1730 ** nonterminal */
1731 FindFirstSets(&lem);
1733 /* Compute all LR(0) states. Also record follow-set propagation
1734 ** links so that the follow-set can be computed later */
1735 lem.nstate = 0;
1736 FindStates(&lem);
1737 lem.sorted = State_arrayof();
1739 /* Tie up loose ends on the propagation links */
1740 FindLinks(&lem);
1742 /* Compute the follow set of every reducible configuration */
1743 FindFollowSets(&lem);
1745 /* Compute the action tables */
1746 FindActions(&lem);
1748 /* Compress the action tables */
1749 if( compress==0 ) CompressTables(&lem);
1751 /* Reorder and renumber the states so that states with fewer choices
1752 ** occur at the end. This is an optimization that helps make the
1753 ** generated parser tables smaller. */
1754 if( noResort==0 ) ResortStates(&lem);
1756 /* Generate a report of the parser generated. (the "y.output" file) */
1757 if( !quiet ) ReportOutput(&lem);
1759 /* Generate the source code for the parser */
1760 ReportTable(&lem, mhflag);
1762 /* Produce a header file for use by the scanner. (This step is
1763 ** omitted if the "-m" option is used because makeheaders will
1764 ** generate the file for us.) */
1765 if( !mhflag ) ReportHeader(&lem);
1767 if( statistics ){
1768 printf("Parser statistics:\n");
1769 stats_line("terminal symbols", lem.nterminal);
1770 stats_line("non-terminal symbols", lem.nsymbol - lem.nterminal);
1771 stats_line("total symbols", lem.nsymbol);
1772 stats_line("rules", lem.nrule);
1773 stats_line("states", lem.nxstate);
1774 stats_line("conflicts", lem.nconflict);
1775 stats_line("action table entries", lem.nactiontab);
1776 stats_line("lookahead table entries", lem.nlookaheadtab);
1777 stats_line("total table size (bytes)", lem.tablesize);
1779 if( lem.nconflict > 0 ){
1780 fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1783 /* return 0 on success, 1 on failure. */
1784 exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0;
1785 exit(exitcode);
1786 return (exitcode);
1788 /******************** From the file "msort.c" *******************************/
1790 ** A generic merge-sort program.
1792 ** USAGE:
1793 ** Let "ptr" be a pointer to some structure which is at the head of
1794 ** a null-terminated list. Then to sort the list call:
1796 ** ptr = msort(ptr,&(ptr->next),cmpfnc);
1798 ** In the above, "cmpfnc" is a pointer to a function which compares
1799 ** two instances of the structure and returns an integer, as in
1800 ** strcmp. The second argument is a pointer to the pointer to the
1801 ** second element of the linked list. This address is used to compute
1802 ** the offset to the "next" field within the structure. The offset to
1803 ** the "next" field must be constant for all structures in the list.
1805 ** The function returns a new pointer which is the head of the list
1806 ** after sorting.
1808 ** ALGORITHM:
1809 ** Merge-sort.
1813 ** Return a pointer to the next structure in the linked list.
1815 #define NEXT(A) (*(char**)(((char*)A)+offset))
1818 ** Inputs:
1819 ** a: A sorted, null-terminated linked list. (May be null).
1820 ** b: A sorted, null-terminated linked list. (May be null).
1821 ** cmp: A pointer to the comparison function.
1822 ** offset: Offset in the structure to the "next" field.
1824 ** Return Value:
1825 ** A pointer to the head of a sorted list containing the elements
1826 ** of both a and b.
1828 ** Side effects:
1829 ** The "next" pointers for elements in the lists a and b are
1830 ** changed.
1832 static char *merge(
1833 char *a,
1834 char *b,
1835 int (*cmp)(const char*,const char*),
1836 int offset
1838 char *ptr, *head;
1840 if( a==0 ){
1841 head = b;
1842 }else if( b==0 ){
1843 head = a;
1844 }else{
1845 if( (*cmp)(a,b)<=0 ){
1846 ptr = a;
1847 a = NEXT(a);
1848 }else{
1849 ptr = b;
1850 b = NEXT(b);
1852 head = ptr;
1853 while( a && b ){
1854 if( (*cmp)(a,b)<=0 ){
1855 NEXT(ptr) = a;
1856 ptr = a;
1857 a = NEXT(a);
1858 }else{
1859 NEXT(ptr) = b;
1860 ptr = b;
1861 b = NEXT(b);
1864 if( a ) NEXT(ptr) = a;
1865 else NEXT(ptr) = b;
1867 return head;
1871 ** Inputs:
1872 ** list: Pointer to a singly-linked list of structures.
1873 ** next: Pointer to pointer to the second element of the list.
1874 ** cmp: A comparison function.
1876 ** Return Value:
1877 ** A pointer to the head of a sorted list containing the elements
1878 ** orginally in list.
1880 ** Side effects:
1881 ** The "next" pointers for elements in list are changed.
1883 #define LISTSIZE 30
1884 static char *msort(
1885 char *list,
1886 char **next,
1887 int (*cmp)(const char*,const char*)
1889 unsigned long offset;
1890 char *ep;
1891 char *set[LISTSIZE];
1892 int i;
1893 offset = (unsigned long)((char*)next - (char*)list);
1894 for(i=0; i<LISTSIZE; i++) set[i] = 0;
1895 while( list ){
1896 ep = list;
1897 list = NEXT(list);
1898 NEXT(ep) = 0;
1899 for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1900 ep = merge(ep,set[i],cmp,offset);
1901 set[i] = 0;
1903 set[i] = ep;
1905 ep = 0;
1906 for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset);
1907 return ep;
1909 /************************ From the file "option.c" **************************/
1910 static char **argv;
1911 static struct s_options *op;
1912 static FILE *errstream;
1914 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1917 ** Print the command line with a carrot pointing to the k-th character
1918 ** of the n-th field.
1920 static void errline(int n, int k, FILE *err)
1922 int spcnt, i;
1923 if( argv[0] ) fprintf(err,"%s",argv[0]);
1924 spcnt = lemonStrlen(argv[0]) + 1;
1925 for(i=1; i<n && argv[i]; i++){
1926 fprintf(err," %s",argv[i]);
1927 spcnt += lemonStrlen(argv[i])+1;
1929 spcnt += k;
1930 for(; argv[i]; i++) fprintf(err," %s",argv[i]);
1931 if( spcnt<20 ){
1932 fprintf(err,"\n%*s^-- here\n",spcnt,"");
1933 }else{
1934 fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1939 ** Return the index of the N-th non-switch argument. Return -1
1940 ** if N is out of range.
1942 static int argindex(int n)
1944 int i;
1945 int dashdash = 0;
1946 if( argv!=0 && *argv!=0 ){
1947 for(i=1; argv[i]; i++){
1948 if( dashdash || !ISOPT(argv[i]) ){
1949 if( n==0 ) return i;
1950 n--;
1952 if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1955 return -1;
1958 static char emsg[] = "Command line syntax error: ";
1961 ** Process a flag command line argument.
1963 static int handleflags(int i, FILE *err)
1965 int v;
1966 int errcnt = 0;
1967 int j;
1968 for(j=0; op[j].label; j++){
1969 if( strncmp(&argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
1971 v = argv[i][0]=='-' ? 1 : 0;
1972 if( op[j].label==0 ){
1973 if( err ){
1974 fprintf(err,"%sundefined option.\n",emsg);
1975 errline(i,1,err);
1977 errcnt++;
1978 }else if( op[j].arg==0 ){
1979 /* Ignore this option */
1980 }else if( op[j].type==OPT_FLAG ){
1981 *((int*)op[j].arg) = v;
1982 }else if( op[j].type==OPT_FFLAG ){
1983 (*(void(*)(int))(op[j].arg))(v);
1984 }else if( op[j].type==OPT_FSTR ){
1985 (*(void(*)(char *))(op[j].arg))(&argv[i][2]);
1986 }else{
1987 if( err ){
1988 fprintf(err,"%smissing argument on switch.\n",emsg);
1989 errline(i,1,err);
1991 errcnt++;
1993 return errcnt;
1997 ** Process a command line switch which has an argument.
1999 static int handleswitch(int i, FILE *err)
2001 int lv = 0;
2002 double dv = 0.0;
2003 char *sv = 0, *end;
2004 char *cp;
2005 int j;
2006 int errcnt = 0;
2007 cp = strchr(argv[i],'=');
2008 assert( cp!=0 );
2009 *cp = 0;
2010 for(j=0; op[j].label; j++){
2011 if( strcmp(argv[i],op[j].label)==0 ) break;
2013 *cp = '=';
2014 if( op[j].label==0 ){
2015 if( err ){
2016 fprintf(err,"%sundefined option.\n",emsg);
2017 errline(i,0,err);
2019 errcnt++;
2020 }else{
2021 cp++;
2022 switch( op[j].type ){
2023 case OPT_FLAG:
2024 case OPT_FFLAG:
2025 if( err ){
2026 fprintf(err,"%soption requires an argument.\n",emsg);
2027 errline(i,0,err);
2029 errcnt++;
2030 break;
2031 case OPT_DBL:
2032 case OPT_FDBL:
2033 dv = strtod(cp,&end);
2034 if( *end ){
2035 if( err ){
2036 fprintf(err,
2037 "%sillegal character in floating-point argument.\n",emsg);
2038 errline(i,(int)((char*)end-(char*)argv[i]),err);
2040 errcnt++;
2042 break;
2043 case OPT_INT:
2044 case OPT_FINT:
2045 lv = strtol(cp,&end,0);
2046 if( *end ){
2047 if( err ){
2048 fprintf(err,"%sillegal character in integer argument.\n",emsg);
2049 errline(i,(int)((char*)end-(char*)argv[i]),err);
2051 errcnt++;
2053 break;
2054 case OPT_STR:
2055 case OPT_FSTR:
2056 sv = cp;
2057 break;
2059 switch( op[j].type ){
2060 case OPT_FLAG:
2061 case OPT_FFLAG:
2062 break;
2063 case OPT_DBL:
2064 *(double*)(op[j].arg) = dv;
2065 break;
2066 case OPT_FDBL:
2067 (*(void(*)(double))(op[j].arg))(dv);
2068 break;
2069 case OPT_INT:
2070 *(int*)(op[j].arg) = lv;
2071 break;
2072 case OPT_FINT:
2073 (*(void(*)(int))(op[j].arg))((int)lv);
2074 break;
2075 case OPT_STR:
2076 *(char**)(op[j].arg) = sv;
2077 break;
2078 case OPT_FSTR:
2079 (*(void(*)(char *))(op[j].arg))(sv);
2080 break;
2083 return errcnt;
2086 int OptInit(char **a, struct s_options *o, FILE *err)
2088 int errcnt = 0;
2089 argv = a;
2090 op = o;
2091 errstream = err;
2092 if( argv && *argv && op ){
2093 int i;
2094 for(i=1; argv[i]; i++){
2095 if( argv[i][0]=='+' || argv[i][0]=='-' ){
2096 errcnt += handleflags(i,err);
2097 }else if( strchr(argv[i],'=') ){
2098 errcnt += handleswitch(i,err);
2102 if( errcnt>0 ){
2103 fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
2104 OptPrint();
2105 exit(1);
2107 return 0;
2110 int OptNArgs(void){
2111 int cnt = 0;
2112 int dashdash = 0;
2113 int i;
2114 if( argv!=0 && argv[0]!=0 ){
2115 for(i=1; argv[i]; i++){
2116 if( dashdash || !ISOPT(argv[i]) ) cnt++;
2117 if( strcmp(argv[i],"--")==0 ) dashdash = 1;
2120 return cnt;
2123 char *OptArg(int n)
2125 int i;
2126 i = argindex(n);
2127 return i>=0 ? argv[i] : 0;
2130 void OptErr(int n)
2132 int i;
2133 i = argindex(n);
2134 if( i>=0 ) errline(i,0,errstream);
2137 void OptPrint(void){
2138 int i;
2139 int max, len;
2140 max = 0;
2141 for(i=0; op[i].label; i++){
2142 len = lemonStrlen(op[i].label) + 1;
2143 switch( op[i].type ){
2144 case OPT_FLAG:
2145 case OPT_FFLAG:
2146 break;
2147 case OPT_INT:
2148 case OPT_FINT:
2149 len += 9; /* length of "<integer>" */
2150 break;
2151 case OPT_DBL:
2152 case OPT_FDBL:
2153 len += 6; /* length of "<real>" */
2154 break;
2155 case OPT_STR:
2156 case OPT_FSTR:
2157 len += 8; /* length of "<string>" */
2158 break;
2160 if( len>max ) max = len;
2162 for(i=0; op[i].label; i++){
2163 switch( op[i].type ){
2164 case OPT_FLAG:
2165 case OPT_FFLAG:
2166 fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
2167 break;
2168 case OPT_INT:
2169 case OPT_FINT:
2170 fprintf(errstream," -%s<integer>%*s %s\n",op[i].label,
2171 (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
2172 break;
2173 case OPT_DBL:
2174 case OPT_FDBL:
2175 fprintf(errstream," -%s<real>%*s %s\n",op[i].label,
2176 (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
2177 break;
2178 case OPT_STR:
2179 case OPT_FSTR:
2180 fprintf(errstream," -%s<string>%*s %s\n",op[i].label,
2181 (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
2182 break;
2186 /*********************** From the file "parse.c" ****************************/
2188 ** Input file parser for the LEMON parser generator.
2191 /* The state of the parser */
2192 enum e_state {
2193 INITIALIZE,
2194 WAITING_FOR_DECL_OR_RULE,
2195 WAITING_FOR_DECL_KEYWORD,
2196 WAITING_FOR_DECL_ARG,
2197 WAITING_FOR_PRECEDENCE_SYMBOL,
2198 WAITING_FOR_ARROW,
2199 IN_RHS,
2200 LHS_ALIAS_1,
2201 LHS_ALIAS_2,
2202 LHS_ALIAS_3,
2203 RHS_ALIAS_1,
2204 RHS_ALIAS_2,
2205 PRECEDENCE_MARK_1,
2206 PRECEDENCE_MARK_2,
2207 RESYNC_AFTER_RULE_ERROR,
2208 RESYNC_AFTER_DECL_ERROR,
2209 WAITING_FOR_DESTRUCTOR_SYMBOL,
2210 WAITING_FOR_DATATYPE_SYMBOL,
2211 WAITING_FOR_FALLBACK_ID,
2212 WAITING_FOR_WILDCARD_ID,
2213 WAITING_FOR_CLASS_ID,
2214 WAITING_FOR_CLASS_TOKEN,
2215 WAITING_FOR_TOKEN_NAME
2217 struct pstate {
2218 char *filename; /* Name of the input file */
2219 int tokenlineno; /* Linenumber at which current token starts */
2220 int errorcnt; /* Number of errors so far */
2221 char *tokenstart; /* Text of current token */
2222 struct lemon *gp; /* Global state vector */
2223 enum e_state state; /* The state of the parser */
2224 struct symbol *fallback; /* The fallback token */
2225 struct symbol *tkclass; /* Token class symbol */
2226 struct symbol *lhs; /* Left-hand side of current rule */
2227 const char *lhsalias; /* Alias for the LHS */
2228 int nrhs; /* Number of right-hand side symbols seen */
2229 struct symbol *rhs[MAXRHS]; /* RHS symbols */
2230 const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
2231 struct rule *prevrule; /* Previous rule parsed */
2232 const char *declkeyword; /* Keyword of a declaration */
2233 char **declargslot; /* Where the declaration argument should be put */
2234 int insertLineMacro; /* Add #line before declaration insert */
2235 int *decllinenoslot; /* Where to write declaration line number */
2236 enum e_assoc declassoc; /* Assign this association to decl arguments */
2237 int preccounter; /* Assign this precedence to decl arguments */
2238 struct rule *firstrule; /* Pointer to first rule in the grammar */
2239 struct rule *lastrule; /* Pointer to the most recently parsed rule */
2242 /* Parse a single token */
2243 static void parseonetoken(struct pstate *psp)
2245 const char *x;
2246 x = Strsafe(psp->tokenstart); /* Save the token permanently */
2247 #if 0
2248 printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
2249 x,psp->state);
2250 #endif
2251 switch( psp->state ){
2252 case INITIALIZE:
2253 psp->prevrule = 0;
2254 psp->preccounter = 0;
2255 psp->firstrule = psp->lastrule = 0;
2256 psp->gp->nrule = 0;
2257 /* Fall thru to next case */
2258 case WAITING_FOR_DECL_OR_RULE:
2259 if( x[0]=='%' ){
2260 psp->state = WAITING_FOR_DECL_KEYWORD;
2261 }else if( ISLOWER(x[0]) ){
2262 psp->lhs = Symbol_new(x);
2263 psp->nrhs = 0;
2264 psp->lhsalias = 0;
2265 psp->state = WAITING_FOR_ARROW;
2266 }else if( x[0]=='{' ){
2267 if( psp->prevrule==0 ){
2268 ErrorMsg(psp->filename,psp->tokenlineno,
2269 "There is no prior rule upon which to attach the code \
2270 fragment which begins on this line.");
2271 psp->errorcnt++;
2272 }else if( psp->prevrule->code!=0 ){
2273 ErrorMsg(psp->filename,psp->tokenlineno,
2274 "Code fragment beginning on this line is not the first \
2275 to follow the previous rule.");
2276 psp->errorcnt++;
2277 }else{
2278 psp->prevrule->line = psp->tokenlineno;
2279 psp->prevrule->code = &x[1];
2280 psp->prevrule->noCode = 0;
2282 }else if( x[0]=='[' ){
2283 psp->state = PRECEDENCE_MARK_1;
2284 }else{
2285 ErrorMsg(psp->filename,psp->tokenlineno,
2286 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2288 psp->errorcnt++;
2290 break;
2291 case PRECEDENCE_MARK_1:
2292 if( !ISUPPER(x[0]) ){
2293 ErrorMsg(psp->filename,psp->tokenlineno,
2294 "The precedence symbol must be a terminal.");
2295 psp->errorcnt++;
2296 }else if( psp->prevrule==0 ){
2297 ErrorMsg(psp->filename,psp->tokenlineno,
2298 "There is no prior rule to assign precedence \"[%s]\".",x);
2299 psp->errorcnt++;
2300 }else if( psp->prevrule->precsym!=0 ){
2301 ErrorMsg(psp->filename,psp->tokenlineno,
2302 "Precedence mark on this line is not the first \
2303 to follow the previous rule.");
2304 psp->errorcnt++;
2305 }else{
2306 psp->prevrule->precsym = Symbol_new(x);
2308 psp->state = PRECEDENCE_MARK_2;
2309 break;
2310 case PRECEDENCE_MARK_2:
2311 if( x[0]!=']' ){
2312 ErrorMsg(psp->filename,psp->tokenlineno,
2313 "Missing \"]\" on precedence mark.");
2314 psp->errorcnt++;
2316 psp->state = WAITING_FOR_DECL_OR_RULE;
2317 break;
2318 case WAITING_FOR_ARROW:
2319 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2320 psp->state = IN_RHS;
2321 }else if( x[0]=='(' ){
2322 psp->state = LHS_ALIAS_1;
2323 }else{
2324 ErrorMsg(psp->filename,psp->tokenlineno,
2325 "Expected to see a \":\" following the LHS symbol \"%s\".",
2326 psp->lhs->name);
2327 psp->errorcnt++;
2328 psp->state = RESYNC_AFTER_RULE_ERROR;
2330 break;
2331 case LHS_ALIAS_1:
2332 if( ISALPHA(x[0]) ){
2333 psp->lhsalias = x;
2334 psp->state = LHS_ALIAS_2;
2335 }else{
2336 ErrorMsg(psp->filename,psp->tokenlineno,
2337 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2338 x,psp->lhs->name);
2339 psp->errorcnt++;
2340 psp->state = RESYNC_AFTER_RULE_ERROR;
2342 break;
2343 case LHS_ALIAS_2:
2344 if( x[0]==')' ){
2345 psp->state = LHS_ALIAS_3;
2346 }else{
2347 ErrorMsg(psp->filename,psp->tokenlineno,
2348 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2349 psp->errorcnt++;
2350 psp->state = RESYNC_AFTER_RULE_ERROR;
2352 break;
2353 case LHS_ALIAS_3:
2354 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2355 psp->state = IN_RHS;
2356 }else{
2357 ErrorMsg(psp->filename,psp->tokenlineno,
2358 "Missing \"->\" following: \"%s(%s)\".",
2359 psp->lhs->name,psp->lhsalias);
2360 psp->errorcnt++;
2361 psp->state = RESYNC_AFTER_RULE_ERROR;
2363 break;
2364 case IN_RHS:
2365 if( x[0]=='.' ){
2366 struct rule *rp;
2367 rp = (struct rule *)calloc( sizeof(struct rule) +
2368 sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
2369 if( rp==0 ){
2370 ErrorMsg(psp->filename,psp->tokenlineno,
2371 "Can't allocate enough memory for this rule.");
2372 psp->errorcnt++;
2373 psp->prevrule = 0;
2374 }else{
2375 int i;
2376 rp->ruleline = psp->tokenlineno;
2377 rp->rhs = (struct symbol**)&rp[1];
2378 rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]);
2379 for(i=0; i<psp->nrhs; i++){
2380 rp->rhs[i] = psp->rhs[i];
2381 rp->rhsalias[i] = psp->alias[i];
2382 if( rp->rhsalias[i]!=0 ){ rp->rhs[i]->bContent = 1; }
2384 rp->lhs = psp->lhs;
2385 rp->lhsalias = psp->lhsalias;
2386 rp->nrhs = psp->nrhs;
2387 rp->code = 0;
2388 rp->noCode = 1;
2389 rp->precsym = 0;
2390 rp->index = psp->gp->nrule++;
2391 rp->nextlhs = rp->lhs->rule;
2392 rp->lhs->rule = rp;
2393 rp->next = 0;
2394 if( psp->firstrule==0 ){
2395 psp->firstrule = psp->lastrule = rp;
2396 }else{
2397 psp->lastrule->next = rp;
2398 psp->lastrule = rp;
2400 psp->prevrule = rp;
2402 psp->state = WAITING_FOR_DECL_OR_RULE;
2403 }else if( ISALPHA(x[0]) ){
2404 if( psp->nrhs>=MAXRHS ){
2405 ErrorMsg(psp->filename,psp->tokenlineno,
2406 "Too many symbols on RHS of rule beginning at \"%s\".",
2408 psp->errorcnt++;
2409 psp->state = RESYNC_AFTER_RULE_ERROR;
2410 }else{
2411 psp->rhs[psp->nrhs] = Symbol_new(x);
2412 psp->alias[psp->nrhs] = 0;
2413 psp->nrhs++;
2415 }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 ){
2416 struct symbol *msp = psp->rhs[psp->nrhs-1];
2417 if( msp->type!=MULTITERMINAL ){
2418 struct symbol *origsp = msp;
2419 msp = (struct symbol *) calloc(1,sizeof(*msp));
2420 memset(msp, 0, sizeof(*msp));
2421 msp->type = MULTITERMINAL;
2422 msp->nsubsym = 1;
2423 msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*));
2424 msp->subsym[0] = origsp;
2425 msp->name = origsp->name;
2426 psp->rhs[psp->nrhs-1] = msp;
2428 msp->nsubsym++;
2429 msp->subsym = (struct symbol **) realloc(msp->subsym,
2430 sizeof(struct symbol*)*msp->nsubsym);
2431 msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
2432 if( ISLOWER(x[1]) || ISLOWER(msp->subsym[0]->name[0]) ){
2433 ErrorMsg(psp->filename,psp->tokenlineno,
2434 "Cannot form a compound containing a non-terminal");
2435 psp->errorcnt++;
2437 }else if( x[0]=='(' && psp->nrhs>0 ){
2438 psp->state = RHS_ALIAS_1;
2439 }else{
2440 ErrorMsg(psp->filename,psp->tokenlineno,
2441 "Illegal character on RHS of rule: \"%s\".",x);
2442 psp->errorcnt++;
2443 psp->state = RESYNC_AFTER_RULE_ERROR;
2445 break;
2446 case RHS_ALIAS_1:
2447 if( ISALPHA(x[0]) ){
2448 psp->alias[psp->nrhs-1] = x;
2449 psp->state = RHS_ALIAS_2;
2450 }else{
2451 ErrorMsg(psp->filename,psp->tokenlineno,
2452 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2453 x,psp->rhs[psp->nrhs-1]->name);
2454 psp->errorcnt++;
2455 psp->state = RESYNC_AFTER_RULE_ERROR;
2457 break;
2458 case RHS_ALIAS_2:
2459 if( x[0]==')' ){
2460 psp->state = IN_RHS;
2461 }else{
2462 ErrorMsg(psp->filename,psp->tokenlineno,
2463 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2464 psp->errorcnt++;
2465 psp->state = RESYNC_AFTER_RULE_ERROR;
2467 break;
2468 case WAITING_FOR_DECL_KEYWORD:
2469 if( ISALPHA(x[0]) ){
2470 psp->declkeyword = x;
2471 psp->declargslot = 0;
2472 psp->decllinenoslot = 0;
2473 psp->insertLineMacro = 1;
2474 psp->state = WAITING_FOR_DECL_ARG;
2475 if( strcmp(x,"name")==0 ){
2476 psp->declargslot = &(psp->gp->name);
2477 psp->insertLineMacro = 0;
2478 }else if( strcmp(x,"include")==0 ){
2479 psp->declargslot = &(psp->gp->include);
2480 }else if( strcmp(x,"code")==0 ){
2481 psp->declargslot = &(psp->gp->extracode);
2482 }else if( strcmp(x,"token_destructor")==0 ){
2483 psp->declargslot = &psp->gp->tokendest;
2484 }else if( strcmp(x,"default_destructor")==0 ){
2485 psp->declargslot = &psp->gp->vardest;
2486 }else if( strcmp(x,"token_prefix")==0 ){
2487 psp->declargslot = &psp->gp->tokenprefix;
2488 psp->insertLineMacro = 0;
2489 }else if( strcmp(x,"syntax_error")==0 ){
2490 psp->declargslot = &(psp->gp->error);
2491 }else if( strcmp(x,"parse_accept")==0 ){
2492 psp->declargslot = &(psp->gp->accept);
2493 }else if( strcmp(x,"parse_failure")==0 ){
2494 psp->declargslot = &(psp->gp->failure);
2495 }else if( strcmp(x,"stack_overflow")==0 ){
2496 psp->declargslot = &(psp->gp->overflow);
2497 }else if( strcmp(x,"extra_argument")==0 ){
2498 psp->declargslot = &(psp->gp->arg);
2499 psp->insertLineMacro = 0;
2500 }else if( strcmp(x,"extra_context")==0 ){
2501 psp->declargslot = &(psp->gp->ctx);
2502 psp->insertLineMacro = 0;
2503 }else if( strcmp(x,"token_type")==0 ){
2504 psp->declargslot = &(psp->gp->tokentype);
2505 psp->insertLineMacro = 0;
2506 }else if( strcmp(x,"default_type")==0 ){
2507 psp->declargslot = &(psp->gp->vartype);
2508 psp->insertLineMacro = 0;
2509 }else if( strcmp(x,"stack_size")==0 ){
2510 psp->declargslot = &(psp->gp->stacksize);
2511 psp->insertLineMacro = 0;
2512 }else if( strcmp(x,"start_symbol")==0 ){
2513 psp->declargslot = &(psp->gp->start);
2514 psp->insertLineMacro = 0;
2515 }else if( strcmp(x,"left")==0 ){
2516 psp->preccounter++;
2517 psp->declassoc = LEFT;
2518 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2519 }else if( strcmp(x,"right")==0 ){
2520 psp->preccounter++;
2521 psp->declassoc = RIGHT;
2522 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2523 }else if( strcmp(x,"nonassoc")==0 ){
2524 psp->preccounter++;
2525 psp->declassoc = NONE;
2526 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2527 }else if( strcmp(x,"destructor")==0 ){
2528 psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2529 }else if( strcmp(x,"type")==0 ){
2530 psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2531 }else if( strcmp(x,"fallback")==0 ){
2532 psp->fallback = 0;
2533 psp->state = WAITING_FOR_FALLBACK_ID;
2534 }else if( strcmp(x,"token")==0 ){
2535 psp->state = WAITING_FOR_TOKEN_NAME;
2536 }else if( strcmp(x,"wildcard")==0 ){
2537 psp->state = WAITING_FOR_WILDCARD_ID;
2538 }else if( strcmp(x,"token_class")==0 ){
2539 psp->state = WAITING_FOR_CLASS_ID;
2540 }else{
2541 ErrorMsg(psp->filename,psp->tokenlineno,
2542 "Unknown declaration keyword: \"%%%s\".",x);
2543 psp->errorcnt++;
2544 psp->state = RESYNC_AFTER_DECL_ERROR;
2546 }else{
2547 ErrorMsg(psp->filename,psp->tokenlineno,
2548 "Illegal declaration keyword: \"%s\".",x);
2549 psp->errorcnt++;
2550 psp->state = RESYNC_AFTER_DECL_ERROR;
2552 break;
2553 case WAITING_FOR_DESTRUCTOR_SYMBOL:
2554 if( !ISALPHA(x[0]) ){
2555 ErrorMsg(psp->filename,psp->tokenlineno,
2556 "Symbol name missing after %%destructor keyword");
2557 psp->errorcnt++;
2558 psp->state = RESYNC_AFTER_DECL_ERROR;
2559 }else{
2560 struct symbol *sp = Symbol_new(x);
2561 psp->declargslot = &sp->destructor;
2562 psp->decllinenoslot = &sp->destLineno;
2563 psp->insertLineMacro = 1;
2564 psp->state = WAITING_FOR_DECL_ARG;
2566 break;
2567 case WAITING_FOR_DATATYPE_SYMBOL:
2568 if( !ISALPHA(x[0]) ){
2569 ErrorMsg(psp->filename,psp->tokenlineno,
2570 "Symbol name missing after %%type keyword");
2571 psp->errorcnt++;
2572 psp->state = RESYNC_AFTER_DECL_ERROR;
2573 }else{
2574 struct symbol *sp = Symbol_find(x);
2575 if((sp) && (sp->datatype)){
2576 ErrorMsg(psp->filename,psp->tokenlineno,
2577 "Symbol %%type \"%s\" already defined", x);
2578 psp->errorcnt++;
2579 psp->state = RESYNC_AFTER_DECL_ERROR;
2580 }else{
2581 if (!sp){
2582 sp = Symbol_new(x);
2584 psp->declargslot = &sp->datatype;
2585 psp->insertLineMacro = 0;
2586 psp->state = WAITING_FOR_DECL_ARG;
2589 break;
2590 case WAITING_FOR_PRECEDENCE_SYMBOL:
2591 if( x[0]=='.' ){
2592 psp->state = WAITING_FOR_DECL_OR_RULE;
2593 }else if( ISUPPER(x[0]) ){
2594 struct symbol *sp;
2595 sp = Symbol_new(x);
2596 if( sp->prec>=0 ){
2597 ErrorMsg(psp->filename,psp->tokenlineno,
2598 "Symbol \"%s\" has already be given a precedence.",x);
2599 psp->errorcnt++;
2600 }else{
2601 sp->prec = psp->preccounter;
2602 sp->assoc = psp->declassoc;
2604 }else{
2605 ErrorMsg(psp->filename,psp->tokenlineno,
2606 "Can't assign a precedence to \"%s\".",x);
2607 psp->errorcnt++;
2609 break;
2610 case WAITING_FOR_DECL_ARG:
2611 if( x[0]=='{' || x[0]=='\"' || ISALNUM(x[0]) ){
2612 const char *zOld, *zNew;
2613 char *zBuf, *z;
2614 int nOld, n, nLine = 0, nNew, nBack;
2615 int addLineMacro;
2616 char zLine[50];
2617 zNew = x;
2618 if( zNew[0]=='"' || zNew[0]=='{' ) zNew++;
2619 nNew = lemonStrlen(zNew);
2620 if( *psp->declargslot ){
2621 zOld = *psp->declargslot;
2622 }else{
2623 zOld = "";
2625 nOld = lemonStrlen(zOld);
2626 n = nOld + nNew + 20;
2627 addLineMacro = !psp->gp->nolinenosflag && psp->insertLineMacro &&
2628 (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0);
2629 if( addLineMacro ){
2630 for(z=psp->filename, nBack=0; *z; z++){
2631 if( *z=='\\' ) nBack++;
2633 lemon_sprintf(zLine, "#line %d ", psp->tokenlineno);
2634 nLine = lemonStrlen(zLine);
2635 n += nLine + lemonStrlen(psp->filename) + nBack;
2637 *psp->declargslot = (char *) realloc(*psp->declargslot, n);
2638 zBuf = *psp->declargslot + nOld;
2639 if( addLineMacro ){
2640 if( nOld && zBuf[-1]!='\n' ){
2641 *(zBuf++) = '\n';
2643 memcpy(zBuf, zLine, nLine);
2644 zBuf += nLine;
2645 *(zBuf++) = '"';
2646 for(z=psp->filename; *z; z++){
2647 if( *z=='\\' ){
2648 *(zBuf++) = '\\';
2650 *(zBuf++) = *z;
2652 *(zBuf++) = '"';
2653 *(zBuf++) = '\n';
2655 if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){
2656 psp->decllinenoslot[0] = psp->tokenlineno;
2658 memcpy(zBuf, zNew, nNew);
2659 zBuf += nNew;
2660 *zBuf = 0;
2661 psp->state = WAITING_FOR_DECL_OR_RULE;
2662 }else{
2663 ErrorMsg(psp->filename,psp->tokenlineno,
2664 "Illegal argument to %%%s: %s",psp->declkeyword,x);
2665 psp->errorcnt++;
2666 psp->state = RESYNC_AFTER_DECL_ERROR;
2668 break;
2669 case WAITING_FOR_FALLBACK_ID:
2670 if( x[0]=='.' ){
2671 psp->state = WAITING_FOR_DECL_OR_RULE;
2672 }else if( !ISUPPER(x[0]) ){
2673 ErrorMsg(psp->filename, psp->tokenlineno,
2674 "%%fallback argument \"%s\" should be a token", x);
2675 psp->errorcnt++;
2676 }else{
2677 struct symbol *sp = Symbol_new(x);
2678 if( psp->fallback==0 ){
2679 psp->fallback = sp;
2680 }else if( sp->fallback ){
2681 ErrorMsg(psp->filename, psp->tokenlineno,
2682 "More than one fallback assigned to token %s", x);
2683 psp->errorcnt++;
2684 }else{
2685 sp->fallback = psp->fallback;
2686 psp->gp->has_fallback = 1;
2689 break;
2690 case WAITING_FOR_TOKEN_NAME:
2691 /* Tokens do not have to be declared before use. But they can be
2692 ** in order to control their assigned integer number. The number for
2693 ** each token is assigned when it is first seen. So by including
2695 ** %token ONE TWO THREE
2697 ** early in the grammar file, that assigns small consecutive values
2698 ** to each of the tokens ONE TWO and THREE.
2700 if( x[0]=='.' ){
2701 psp->state = WAITING_FOR_DECL_OR_RULE;
2702 }else if( !ISUPPER(x[0]) ){
2703 ErrorMsg(psp->filename, psp->tokenlineno,
2704 "%%token argument \"%s\" should be a token", x);
2705 psp->errorcnt++;
2706 }else{
2707 (void)Symbol_new(x);
2709 break;
2710 case WAITING_FOR_WILDCARD_ID:
2711 if( x[0]=='.' ){
2712 psp->state = WAITING_FOR_DECL_OR_RULE;
2713 }else if( !ISUPPER(x[0]) ){
2714 ErrorMsg(psp->filename, psp->tokenlineno,
2715 "%%wildcard argument \"%s\" should be a token", x);
2716 psp->errorcnt++;
2717 }else{
2718 struct symbol *sp = Symbol_new(x);
2719 if( psp->gp->wildcard==0 ){
2720 psp->gp->wildcard = sp;
2721 }else{
2722 ErrorMsg(psp->filename, psp->tokenlineno,
2723 "Extra wildcard to token: %s", x);
2724 psp->errorcnt++;
2727 break;
2728 case WAITING_FOR_CLASS_ID:
2729 if( !ISLOWER(x[0]) ){
2730 ErrorMsg(psp->filename, psp->tokenlineno,
2731 "%%token_class must be followed by an identifier: ", x);
2732 psp->errorcnt++;
2733 psp->state = RESYNC_AFTER_DECL_ERROR;
2734 }else if( Symbol_find(x) ){
2735 ErrorMsg(psp->filename, psp->tokenlineno,
2736 "Symbol \"%s\" already used", x);
2737 psp->errorcnt++;
2738 psp->state = RESYNC_AFTER_DECL_ERROR;
2739 }else{
2740 psp->tkclass = Symbol_new(x);
2741 psp->tkclass->type = MULTITERMINAL;
2742 psp->state = WAITING_FOR_CLASS_TOKEN;
2744 break;
2745 case WAITING_FOR_CLASS_TOKEN:
2746 if( x[0]=='.' ){
2747 psp->state = WAITING_FOR_DECL_OR_RULE;
2748 }else if( ISUPPER(x[0]) || ((x[0]=='|' || x[0]=='/') && ISUPPER(x[1])) ){
2749 struct symbol *msp = psp->tkclass;
2750 msp->nsubsym++;
2751 msp->subsym = (struct symbol **) realloc(msp->subsym,
2752 sizeof(struct symbol*)*msp->nsubsym);
2753 if( !ISUPPER(x[0]) ) x++;
2754 msp->subsym[msp->nsubsym-1] = Symbol_new(x);
2755 }else{
2756 ErrorMsg(psp->filename, psp->tokenlineno,
2757 "%%token_class argument \"%s\" should be a token", x);
2758 psp->errorcnt++;
2759 psp->state = RESYNC_AFTER_DECL_ERROR;
2761 break;
2762 case RESYNC_AFTER_RULE_ERROR:
2763 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2764 ** break; */
2765 case RESYNC_AFTER_DECL_ERROR:
2766 if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2767 if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2768 break;
2772 /* Run the preprocessor over the input file text. The global variables
2773 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2774 ** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2775 ** comments them out. Text in between is also commented out as appropriate.
2777 static void preprocess_input(char *z){
2778 int i, j, k, n;
2779 int exclude = 0;
2780 int start = 0;
2781 int lineno = 1;
2782 int start_lineno = 1;
2783 for(i=0; z[i]; i++){
2784 if( z[i]=='\n' ) lineno++;
2785 if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue;
2786 if( strncmp(&z[i],"%endif",6)==0 && ISSPACE(z[i+6]) ){
2787 if( exclude ){
2788 exclude--;
2789 if( exclude==0 ){
2790 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2793 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2794 }else if( (strncmp(&z[i],"%ifdef",6)==0 && ISSPACE(z[i+6]))
2795 || (strncmp(&z[i],"%ifndef",7)==0 && ISSPACE(z[i+7])) ){
2796 if( exclude ){
2797 exclude++;
2798 }else{
2799 for(j=i+7; ISSPACE(z[j]); j++){}
2800 for(n=0; z[j+n] && !ISSPACE(z[j+n]); n++){}
2801 exclude = 1;
2802 for(k=0; k<nDefine; k++){
2803 if( strncmp(azDefine[k],&z[j],n)==0 && lemonStrlen(azDefine[k])==n ){
2804 exclude = 0;
2805 break;
2808 if( z[i+3]=='n' ) exclude = !exclude;
2809 if( exclude ){
2810 start = i;
2811 start_lineno = lineno;
2814 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2817 if( exclude ){
2818 fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno);
2819 exit(1);
2823 /* In spite of its name, this function is really a scanner. It read
2824 ** in the entire input file (all at once) then tokenizes it. Each
2825 ** token is passed to the function "parseonetoken" which builds all
2826 ** the appropriate data structures in the global state vector "gp".
2828 void Parse(struct lemon *gp)
2830 struct pstate ps;
2831 FILE *fp;
2832 char *filebuf;
2833 unsigned int filesize;
2834 int lineno;
2835 int c;
2836 char *cp, *nextcp;
2837 int startline = 0;
2839 memset(&ps, '\0', sizeof(ps));
2840 ps.gp = gp;
2841 ps.filename = gp->filename;
2842 ps.errorcnt = 0;
2843 ps.state = INITIALIZE;
2845 /* Begin by reading the input file */
2846 fp = fopen(ps.filename,"rb");
2847 if( fp==0 ){
2848 ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2849 gp->errorcnt++;
2850 return;
2852 fseek(fp,0,2);
2853 filesize = ftell(fp);
2854 rewind(fp);
2855 filebuf = (char *)malloc( filesize+1 );
2856 if( filesize>100000000 || filebuf==0 ){
2857 ErrorMsg(ps.filename,0,"Input file too large.");
2858 gp->errorcnt++;
2859 fclose(fp);
2860 return;
2862 if( fread(filebuf,1,filesize,fp)!=filesize ){
2863 ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2864 filesize);
2865 free(filebuf);
2866 gp->errorcnt++;
2867 fclose(fp);
2868 return;
2870 fclose(fp);
2871 filebuf[filesize] = 0;
2873 /* Make an initial pass through the file to handle %ifdef and %ifndef */
2874 preprocess_input(filebuf);
2876 /* Now scan the text of the input file */
2877 lineno = 1;
2878 for(cp=filebuf; (c= *cp)!=0; ){
2879 if( c=='\n' ) lineno++; /* Keep track of the line number */
2880 if( ISSPACE(c) ){ cp++; continue; } /* Skip all white space */
2881 if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */
2882 cp+=2;
2883 while( (c= *cp)!=0 && c!='\n' ) cp++;
2884 continue;
2886 if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */
2887 cp+=2;
2888 while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
2889 if( c=='\n' ) lineno++;
2890 cp++;
2892 if( c ) cp++;
2893 continue;
2895 ps.tokenstart = cp; /* Mark the beginning of the token */
2896 ps.tokenlineno = lineno; /* Linenumber on which token begins */
2897 if( c=='\"' ){ /* String literals */
2898 cp++;
2899 while( (c= *cp)!=0 && c!='\"' ){
2900 if( c=='\n' ) lineno++;
2901 cp++;
2903 if( c==0 ){
2904 ErrorMsg(ps.filename,startline,
2905 "String starting on this line is not terminated before the end of the file.");
2906 ps.errorcnt++;
2907 nextcp = cp;
2908 }else{
2909 nextcp = cp+1;
2911 }else if( c=='{' ){ /* A block of C code */
2912 int level;
2913 cp++;
2914 for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
2915 if( c=='\n' ) lineno++;
2916 else if( c=='{' ) level++;
2917 else if( c=='}' ) level--;
2918 else if( c=='/' && cp[1]=='*' ){ /* Skip comments */
2919 int prevc;
2920 cp = &cp[2];
2921 prevc = 0;
2922 while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
2923 if( c=='\n' ) lineno++;
2924 prevc = c;
2925 cp++;
2927 }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */
2928 cp = &cp[2];
2929 while( (c= *cp)!=0 && c!='\n' ) cp++;
2930 if( c ) lineno++;
2931 }else if( c=='\'' || c=='\"' ){ /* String a character literals */
2932 int startchar, prevc;
2933 startchar = c;
2934 prevc = 0;
2935 for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
2936 if( c=='\n' ) lineno++;
2937 if( prevc=='\\' ) prevc = 0;
2938 else prevc = c;
2942 if( c==0 ){
2943 ErrorMsg(ps.filename,ps.tokenlineno,
2944 "C code starting on this line is not terminated before the end of the file.");
2945 ps.errorcnt++;
2946 nextcp = cp;
2947 }else{
2948 nextcp = cp+1;
2950 }else if( ISALNUM(c) ){ /* Identifiers */
2951 while( (c= *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
2952 nextcp = cp;
2953 }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
2954 cp += 3;
2955 nextcp = cp;
2956 }else if( (c=='/' || c=='|') && ISALPHA(cp[1]) ){
2957 cp += 2;
2958 while( (c = *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
2959 nextcp = cp;
2960 }else{ /* All other (one character) operators */
2961 cp++;
2962 nextcp = cp;
2964 c = *cp;
2965 *cp = 0; /* Null terminate the token */
2966 parseonetoken(&ps); /* Parse the token */
2967 *cp = (char)c; /* Restore the buffer */
2968 cp = nextcp;
2970 free(filebuf); /* Release the buffer after parsing */
2971 gp->rule = ps.firstrule;
2972 gp->errorcnt = ps.errorcnt;
2974 /*************************** From the file "plink.c" *********************/
2976 ** Routines processing configuration follow-set propagation links
2977 ** in the LEMON parser generator.
2979 static struct plink *plink_freelist = 0;
2981 /* Allocate a new plink */
2982 struct plink *Plink_new(void){
2983 struct plink *newlink;
2985 if( plink_freelist==0 ){
2986 int i;
2987 int amt = 100;
2988 plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) );
2989 if( plink_freelist==0 ){
2990 fprintf(stderr,
2991 "Unable to allocate memory for a new follow-set propagation link.\n");
2992 exit(1);
2994 for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
2995 plink_freelist[amt-1].next = 0;
2997 newlink = plink_freelist;
2998 plink_freelist = plink_freelist->next;
2999 return newlink;
3002 /* Add a plink to a plink list */
3003 void Plink_add(struct plink **plpp, struct config *cfp)
3005 struct plink *newlink;
3006 newlink = Plink_new();
3007 newlink->next = *plpp;
3008 *plpp = newlink;
3009 newlink->cfp = cfp;
3012 /* Transfer every plink on the list "from" to the list "to" */
3013 void Plink_copy(struct plink **to, struct plink *from)
3015 struct plink *nextpl;
3016 while( from ){
3017 nextpl = from->next;
3018 from->next = *to;
3019 *to = from;
3020 from = nextpl;
3024 /* Delete every plink on the list */
3025 void Plink_delete(struct plink *plp)
3027 struct plink *nextpl;
3029 while( plp ){
3030 nextpl = plp->next;
3031 plp->next = plink_freelist;
3032 plink_freelist = plp;
3033 plp = nextpl;
3036 /*********************** From the file "report.c" **************************/
3038 ** Procedures for generating reports and tables in the LEMON parser generator.
3041 /* Generate a filename with the given suffix. Space to hold the
3042 ** name comes from malloc() and must be freed by the calling
3043 ** function.
3045 PRIVATE char *file_makename(struct lemon *lemp, const char *suffix)
3047 char *name;
3048 char *cp;
3049 char *filename = lemp->filename;
3050 int sz;
3052 if( outputDir ){
3053 cp = strrchr(filename, '/');
3054 if( cp ) filename = cp + 1;
3056 sz = lemonStrlen(filename);
3057 sz += lemonStrlen(suffix);
3058 if( outputDir ) sz += lemonStrlen(outputDir) + 1;
3059 sz += 5;
3060 name = (char*)malloc( sz );
3061 if( name==0 ){
3062 fprintf(stderr,"Can't allocate space for a filename.\n");
3063 exit(1);
3065 name[0] = 0;
3066 if( outputDir ){
3067 lemon_strcpy(name, outputDir);
3068 lemon_strcat(name, "/");
3070 lemon_strcat(name,filename);
3071 cp = strrchr(name,'.');
3072 if( cp ) *cp = 0;
3073 lemon_strcat(name,suffix);
3074 return name;
3077 /* Open a file with a name based on the name of the input file,
3078 ** but with a different (specified) suffix, and return a pointer
3079 ** to the stream */
3080 PRIVATE FILE *file_open(
3081 struct lemon *lemp,
3082 const char *suffix,
3083 const char *mode
3085 FILE *fp;
3087 if( lemp->outname ) free(lemp->outname);
3088 lemp->outname = file_makename(lemp, suffix);
3089 fp = fopen(lemp->outname,mode);
3090 if( fp==0 && *mode=='w' ){
3091 fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
3092 lemp->errorcnt++;
3093 return 0;
3095 return fp;
3098 /* Print the text of a rule
3100 void rule_print(FILE *out, struct rule *rp){
3101 int i, j;
3102 fprintf(out, "%s",rp->lhs->name);
3103 /* if( rp->lhsalias ) fprintf(out,"(%s)",rp->lhsalias); */
3104 fprintf(out," ::=");
3105 for(i=0; i<rp->nrhs; i++){
3106 struct symbol *sp = rp->rhs[i];
3107 if( sp->type==MULTITERMINAL ){
3108 fprintf(out," %s", sp->subsym[0]->name);
3109 for(j=1; j<sp->nsubsym; j++){
3110 fprintf(out,"|%s", sp->subsym[j]->name);
3112 }else{
3113 fprintf(out," %s", sp->name);
3115 /* if( rp->rhsalias[i] ) fprintf(out,"(%s)",rp->rhsalias[i]); */
3119 /* Duplicate the input file without comments and without actions
3120 ** on rules */
3121 void Reprint(struct lemon *lemp)
3123 struct rule *rp;
3124 struct symbol *sp;
3125 int i, j, maxlen, len, ncolumns, skip;
3126 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
3127 maxlen = 10;
3128 for(i=0; i<lemp->nsymbol; i++){
3129 sp = lemp->symbols[i];
3130 len = lemonStrlen(sp->name);
3131 if( len>maxlen ) maxlen = len;
3133 ncolumns = 76/(maxlen+5);
3134 if( ncolumns<1 ) ncolumns = 1;
3135 skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
3136 for(i=0; i<skip; i++){
3137 printf("//");
3138 for(j=i; j<lemp->nsymbol; j+=skip){
3139 sp = lemp->symbols[j];
3140 assert( sp->index==j );
3141 printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
3143 printf("\n");
3145 for(rp=lemp->rule; rp; rp=rp->next){
3146 rule_print(stdout, rp);
3147 printf(".");
3148 if( rp->precsym ) printf(" [%s]",rp->precsym->name);
3149 /* if( rp->code ) printf("\n %s",rp->code); */
3150 printf("\n");
3154 /* Print a single rule.
3156 void RulePrint(FILE *fp, struct rule *rp, int iCursor){
3157 struct symbol *sp;
3158 int i, j;
3159 fprintf(fp,"%s ::=",rp->lhs->name);
3160 for(i=0; i<=rp->nrhs; i++){
3161 if( i==iCursor ) fprintf(fp," *");
3162 if( i==rp->nrhs ) break;
3163 sp = rp->rhs[i];
3164 if( sp->type==MULTITERMINAL ){
3165 fprintf(fp," %s", sp->subsym[0]->name);
3166 for(j=1; j<sp->nsubsym; j++){
3167 fprintf(fp,"|%s",sp->subsym[j]->name);
3169 }else{
3170 fprintf(fp," %s", sp->name);
3175 /* Print the rule for a configuration.
3177 void ConfigPrint(FILE *fp, struct config *cfp){
3178 RulePrint(fp, cfp->rp, cfp->dot);
3181 /* #define TEST */
3182 #if 0
3183 /* Print a set */
3184 PRIVATE void SetPrint(out,set,lemp)
3185 FILE *out;
3186 char *set;
3187 struct lemon *lemp;
3189 int i;
3190 char *spacer;
3191 spacer = "";
3192 fprintf(out,"%12s[","");
3193 for(i=0; i<lemp->nterminal; i++){
3194 if( SetFind(set,i) ){
3195 fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
3196 spacer = " ";
3199 fprintf(out,"]\n");
3202 /* Print a plink chain */
3203 PRIVATE void PlinkPrint(out,plp,tag)
3204 FILE *out;
3205 struct plink *plp;
3206 char *tag;
3208 while( plp ){
3209 fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum);
3210 ConfigPrint(out,plp->cfp);
3211 fprintf(out,"\n");
3212 plp = plp->next;
3215 #endif
3217 /* Print an action to the given file descriptor. Return FALSE if
3218 ** nothing was actually printed.
3220 int PrintAction(
3221 struct action *ap, /* The action to print */
3222 FILE *fp, /* Print the action here */
3223 int indent /* Indent by this amount */
3225 int result = 1;
3226 switch( ap->type ){
3227 case SHIFT: {
3228 struct state *stp = ap->x.stp;
3229 fprintf(fp,"%*s shift %-7d",indent,ap->sp->name,stp->statenum);
3230 break;
3232 case REDUCE: {
3233 struct rule *rp = ap->x.rp;
3234 fprintf(fp,"%*s reduce %-7d",indent,ap->sp->name,rp->iRule);
3235 RulePrint(fp, rp, -1);
3236 break;
3238 case SHIFTREDUCE: {
3239 struct rule *rp = ap->x.rp;
3240 fprintf(fp,"%*s shift-reduce %-7d",indent,ap->sp->name,rp->iRule);
3241 RulePrint(fp, rp, -1);
3242 break;
3244 case ACCEPT:
3245 fprintf(fp,"%*s accept",indent,ap->sp->name);
3246 break;
3247 case ERROR:
3248 fprintf(fp,"%*s error",indent,ap->sp->name);
3249 break;
3250 case SRCONFLICT:
3251 case RRCONFLICT:
3252 fprintf(fp,"%*s reduce %-7d ** Parsing conflict **",
3253 indent,ap->sp->name,ap->x.rp->iRule);
3254 break;
3255 case SSCONFLICT:
3256 fprintf(fp,"%*s shift %-7d ** Parsing conflict **",
3257 indent,ap->sp->name,ap->x.stp->statenum);
3258 break;
3259 case SH_RESOLVED:
3260 if( showPrecedenceConflict ){
3261 fprintf(fp,"%*s shift %-7d -- dropped by precedence",
3262 indent,ap->sp->name,ap->x.stp->statenum);
3263 }else{
3264 result = 0;
3266 break;
3267 case RD_RESOLVED:
3268 if( showPrecedenceConflict ){
3269 fprintf(fp,"%*s reduce %-7d -- dropped by precedence",
3270 indent,ap->sp->name,ap->x.rp->iRule);
3271 }else{
3272 result = 0;
3274 break;
3275 case NOT_USED:
3276 result = 0;
3277 break;
3279 if( result && ap->spOpt ){
3280 fprintf(fp," /* because %s==%s */", ap->sp->name, ap->spOpt->name);
3282 return result;
3285 /* Generate the "*.out" log file */
3286 void ReportOutput(struct lemon *lemp)
3288 int i, n;
3289 struct state *stp;
3290 struct config *cfp;
3291 struct action *ap;
3292 struct rule *rp;
3293 FILE *fp;
3295 fp = file_open(lemp,".out","wb");
3296 if( fp==0 ) return;
3297 for(i=0; i<lemp->nxstate; i++){
3298 stp = lemp->sorted[i];
3299 fprintf(fp,"State %d:\n",stp->statenum);
3300 if( lemp->basisflag ) cfp=stp->bp;
3301 else cfp=stp->cfp;
3302 while( cfp ){
3303 char buf[20];
3304 if( cfp->dot==cfp->rp->nrhs ){
3305 lemon_sprintf(buf,"(%d)",cfp->rp->iRule);
3306 fprintf(fp," %5s ",buf);
3307 }else{
3308 fprintf(fp," ");
3310 ConfigPrint(fp,cfp);
3311 fprintf(fp,"\n");
3312 #if 0
3313 SetPrint(fp,cfp->fws,lemp);
3314 PlinkPrint(fp,cfp->fplp,"To ");
3315 PlinkPrint(fp,cfp->bplp,"From");
3316 #endif
3317 if( lemp->basisflag ) cfp=cfp->bp;
3318 else cfp=cfp->next;
3320 fprintf(fp,"\n");
3321 for(ap=stp->ap; ap; ap=ap->next){
3322 if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
3324 fprintf(fp,"\n");
3326 fprintf(fp, "----------------------------------------------------\n");
3327 fprintf(fp, "Symbols:\n");
3328 fprintf(fp, "The first-set of non-terminals is shown after the name.\n\n");
3329 for(i=0; i<lemp->nsymbol; i++){
3330 int j;
3331 struct symbol *sp;
3333 sp = lemp->symbols[i];
3334 fprintf(fp, " %3d: %s", i, sp->name);
3335 if( sp->type==NONTERMINAL ){
3336 fprintf(fp, ":");
3337 if( sp->lambda ){
3338 fprintf(fp, " <lambda>");
3340 for(j=0; j<lemp->nterminal; j++){
3341 if( sp->firstset && SetFind(sp->firstset, j) ){
3342 fprintf(fp, " %s", lemp->symbols[j]->name);
3346 if( sp->prec>=0 ) fprintf(fp," (precedence=%d)", sp->prec);
3347 fprintf(fp, "\n");
3349 fprintf(fp, "----------------------------------------------------\n");
3350 fprintf(fp, "Syntax-only Symbols:\n");
3351 fprintf(fp, "The following symbols never carry semantic content.\n\n");
3352 for(i=n=0; i<lemp->nsymbol; i++){
3353 int w;
3354 struct symbol *sp = lemp->symbols[i];
3355 if( sp->bContent ) continue;
3356 w = (int)strlen(sp->name);
3357 if( n>0 && n+w>75 ){
3358 fprintf(fp,"\n");
3359 n = 0;
3361 if( n>0 ){
3362 fprintf(fp, " ");
3363 n++;
3365 fprintf(fp, "%s", sp->name);
3366 n += w;
3368 if( n>0 ) fprintf(fp, "\n");
3369 fprintf(fp, "----------------------------------------------------\n");
3370 fprintf(fp, "Rules:\n");
3371 for(rp=lemp->rule; rp; rp=rp->next){
3372 fprintf(fp, "%4d: ", rp->iRule);
3373 rule_print(fp, rp);
3374 fprintf(fp,".");
3375 if( rp->precsym ){
3376 fprintf(fp," [%s precedence=%d]",
3377 rp->precsym->name, rp->precsym->prec);
3379 fprintf(fp,"\n");
3381 fclose(fp);
3382 return;
3385 /* Search for the file "name" which is in the same directory as
3386 ** the exacutable */
3387 PRIVATE char *pathsearch(char *argv0, char *name, int modemask)
3389 const char *pathlist;
3390 char *pathbufptr;
3391 char *pathbuf;
3392 char *path,*cp;
3393 char c;
3395 #ifdef __WIN32__
3396 cp = strrchr(argv0,'\\');
3397 #else
3398 cp = strrchr(argv0,'/');
3399 #endif
3400 if( cp ){
3401 c = *cp;
3402 *cp = 0;
3403 path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 );
3404 if( path ) lemon_sprintf(path,"%s/%s",argv0,name);
3405 *cp = c;
3406 }else{
3407 pathlist = getenv("PATH");
3408 if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
3409 pathbuf = (char *) malloc( lemonStrlen(pathlist) + 1 );
3410 path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 );
3411 if( (pathbuf != 0) && (path!=0) ){
3412 pathbufptr = pathbuf;
3413 lemon_strcpy(pathbuf, pathlist);
3414 while( *pathbuf ){
3415 cp = strchr(pathbuf,':');
3416 if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)];
3417 c = *cp;
3418 *cp = 0;
3419 lemon_sprintf(path,"%s/%s",pathbuf,name);
3420 *cp = c;
3421 if( c==0 ) pathbuf[0] = 0;
3422 else pathbuf = &cp[1];
3423 if( access(path,modemask)==0 ) break;
3425 free(pathbufptr);
3428 return path;
3431 /* Given an action, compute the integer value for that action
3432 ** which is to be put in the action table of the generated machine.
3433 ** Return negative if no action should be generated.
3435 PRIVATE int compute_action(struct lemon *lemp, struct action *ap)
3437 int act;
3438 switch( ap->type ){
3439 case SHIFT: act = ap->x.stp->statenum; break;
3440 case SHIFTREDUCE: {
3441 /* Since a SHIFT is inherient after a prior REDUCE, convert any
3442 ** SHIFTREDUCE action with a nonterminal on the LHS into a simple
3443 ** REDUCE action: */
3444 if( ap->sp->index>=lemp->nterminal ){
3445 act = lemp->minReduce + ap->x.rp->iRule;
3446 }else{
3447 act = lemp->minShiftReduce + ap->x.rp->iRule;
3449 break;
3451 case REDUCE: act = lemp->minReduce + ap->x.rp->iRule; break;
3452 case ERROR: act = lemp->errAction; break;
3453 case ACCEPT: act = lemp->accAction; break;
3454 default: act = -1; break;
3456 return act;
3459 #define LINESIZE 1000
3460 /* The next cluster of routines are for reading the template file
3461 ** and writing the results to the generated parser */
3462 /* The first function transfers data from "in" to "out" until
3463 ** a line is seen which begins with "%%". The line number is
3464 ** tracked.
3466 ** if name!=0, then any word that begin with "Parse" is changed to
3467 ** begin with *name instead.
3469 PRIVATE void tplt_xfer(char *name, FILE *in, FILE *out, int *lineno)
3471 int i, iStart;
3472 char line[LINESIZE];
3473 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3474 (*lineno)++;
3475 iStart = 0;
3476 if( name ){
3477 for(i=0; line[i]; i++){
3478 if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
3479 && (i==0 || !ISALPHA(line[i-1]))
3481 if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
3482 fprintf(out,"%s",name);
3483 i += 4;
3484 iStart = i+1;
3488 fprintf(out,"%s",&line[iStart]);
3492 /* The next function finds the template file and opens it, returning
3493 ** a pointer to the opened file. */
3494 PRIVATE FILE *tplt_open(struct lemon *lemp)
3496 static char templatename[] = "lempar.c";
3497 char buf[1000];
3498 FILE *in;
3499 char *tpltname;
3500 char *cp;
3502 /* first, see if user specified a template filename on the command line. */
3503 if (user_templatename != 0) {
3504 if( access(user_templatename,004)==-1 ){
3505 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3506 user_templatename);
3507 lemp->errorcnt++;
3508 return 0;
3510 in = fopen(user_templatename,"rb");
3511 if( in==0 ){
3512 fprintf(stderr,"Can't open the template file \"%s\".\n",
3513 user_templatename);
3514 lemp->errorcnt++;
3515 return 0;
3517 return in;
3520 cp = strrchr(lemp->filename,'.');
3521 if( cp ){
3522 lemon_sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
3523 }else{
3524 lemon_sprintf(buf,"%s.lt",lemp->filename);
3526 if( access(buf,004)==0 ){
3527 tpltname = buf;
3528 }else if( access(templatename,004)==0 ){
3529 tpltname = templatename;
3530 }else{
3531 tpltname = pathsearch(lemp->argv0,templatename,0);
3533 if( tpltname==0 ){
3534 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3535 templatename);
3536 lemp->errorcnt++;
3537 return 0;
3539 in = fopen(tpltname,"rb");
3540 if( in==0 ){
3541 fprintf(stderr,"Can't open the template file \"%s\".\n",templatename);
3542 lemp->errorcnt++;
3543 return 0;
3545 return in;
3548 /* Print a #line directive line to the output file. */
3549 PRIVATE void tplt_linedir(FILE *out, int lineno, char *filename)
3551 fprintf(out,"#line %d \"",lineno);
3552 while( *filename ){
3553 if( *filename == '\\' ) putc('\\',out);
3554 putc(*filename,out);
3555 filename++;
3557 fprintf(out,"\"\n");
3560 /* Print a string to the file and keep the linenumber up to date */
3561 PRIVATE void tplt_print(FILE *out, struct lemon *lemp, char *str, int *lineno)
3563 if( str==0 ) return;
3564 while( *str ){
3565 putc(*str,out);
3566 if( *str=='\n' ) (*lineno)++;
3567 str++;
3569 if( str[-1]!='\n' ){
3570 putc('\n',out);
3571 (*lineno)++;
3573 if (!lemp->nolinenosflag) {
3574 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3576 return;
3580 ** The following routine emits code for the destructor for the
3581 ** symbol sp
3583 void emit_destructor_code(
3584 FILE *out,
3585 struct symbol *sp,
3586 struct lemon *lemp,
3587 int *lineno
3589 char *cp = 0;
3591 if( sp->type==TERMINAL ){
3592 cp = lemp->tokendest;
3593 if( cp==0 ) return;
3594 fprintf(out,"{\n"); (*lineno)++;
3595 }else if( sp->destructor ){
3596 cp = sp->destructor;
3597 fprintf(out,"{\n"); (*lineno)++;
3598 if( !lemp->nolinenosflag ){
3599 (*lineno)++;
3600 tplt_linedir(out,sp->destLineno,lemp->filename);
3602 }else if( lemp->vardest ){
3603 cp = lemp->vardest;
3604 if( cp==0 ) return;
3605 fprintf(out,"{\n"); (*lineno)++;
3606 }else{
3607 assert( 0 ); /* Cannot happen */
3609 for(; *cp; cp++){
3610 if( *cp=='$' && cp[1]=='$' ){
3611 fprintf(out,"(yypminor->yy%d)",sp->dtnum);
3612 cp++;
3613 continue;
3615 if( *cp=='\n' ) (*lineno)++;
3616 fputc(*cp,out);
3618 fprintf(out,"\n"); (*lineno)++;
3619 if (!lemp->nolinenosflag) {
3620 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3622 fprintf(out,"}\n"); (*lineno)++;
3623 return;
3627 ** Return TRUE (non-zero) if the given symbol has a destructor.
3629 int has_destructor(struct symbol *sp, struct lemon *lemp)
3631 int ret;
3632 if( sp->type==TERMINAL ){
3633 ret = lemp->tokendest!=0;
3634 }else{
3635 ret = lemp->vardest!=0 || sp->destructor!=0;
3637 return ret;
3641 ** Append text to a dynamically allocated string. If zText is 0 then
3642 ** reset the string to be empty again. Always return the complete text
3643 ** of the string (which is overwritten with each call).
3645 ** n bytes of zText are stored. If n==0 then all of zText up to the first
3646 ** \000 terminator is stored. zText can contain up to two instances of
3647 ** %d. The values of p1 and p2 are written into the first and second
3648 ** %d.
3650 ** If n==-1, then the previous character is overwritten.
3652 PRIVATE char *append_str(const char *zText, int n, int p1, int p2){
3653 static char empty[1] = { 0 };
3654 static char *z = 0;
3655 static int alloced = 0;
3656 static int used = 0;
3657 int c;
3658 char zInt[40];
3659 if( zText==0 ){
3660 if( used==0 && z!=0 ) z[0] = 0;
3661 used = 0;
3662 return z;
3664 if( n<=0 ){
3665 if( n<0 ){
3666 used += n;
3667 assert( used>=0 );
3669 n = lemonStrlen(zText);
3671 if( (int) (n+sizeof(zInt)*2+used) >= alloced ){
3672 alloced = n + sizeof(zInt)*2 + used + 200;
3673 z = (char *) realloc(z, alloced);
3675 if( z==0 ) return empty;
3676 while( n-- > 0 ){
3677 c = *(zText++);
3678 if( c=='%' && n>0 && zText[0]=='d' ){
3679 lemon_sprintf(zInt, "%d", p1);
3680 p1 = p2;
3681 lemon_strcpy(&z[used], zInt);
3682 used += lemonStrlen(&z[used]);
3683 zText++;
3684 n--;
3685 }else{
3686 z[used++] = (char)c;
3689 z[used] = 0;
3690 return z;
3694 ** Write and transform the rp->code string so that symbols are expanded.
3695 ** Populate the rp->codePrefix and rp->codeSuffix strings, as appropriate.
3697 ** Return 1 if the expanded code requires that "yylhsminor" local variable
3698 ** to be defined.
3700 PRIVATE int translate_code(struct lemon *lemp, struct rule *rp){
3701 char *cp, *xp;
3702 int i;
3703 int rc = 0; /* True if yylhsminor is used */
3704 int dontUseRhs0 = 0; /* If true, use of left-most RHS label is illegal */
3705 const char *zSkip = 0; /* The zOvwrt comment within rp->code, or NULL */
3706 char lhsused = 0; /* True if the LHS element has been used */
3707 char lhsdirect; /* True if LHS writes directly into stack */
3708 char used[MAXRHS]; /* True for each RHS element which is used */
3709 char zLhs[50]; /* Convert the LHS symbol into this string */
3710 char zOvwrt[900]; /* Comment that to allow LHS to overwrite RHS */
3712 for(i=0; i<rp->nrhs; i++) used[i] = 0;
3713 lhsused = 0;
3715 if( rp->code==0 ){
3716 static char newlinestr[2] = { '\n', '\0' };
3717 rp->code = newlinestr;
3718 rp->line = rp->ruleline;
3719 rp->noCode = 1;
3720 }else{
3721 rp->noCode = 0;
3725 if( rp->nrhs==0 ){
3726 /* If there are no RHS symbols, then writing directly to the LHS is ok */
3727 lhsdirect = 1;
3728 }else if( rp->rhsalias[0]==0 ){
3729 /* The left-most RHS symbol has no value. LHS direct is ok. But
3730 ** we have to call the distructor on the RHS symbol first. */
3731 lhsdirect = 1;
3732 if( has_destructor(rp->rhs[0],lemp) ){
3733 append_str(0,0,0,0);
3734 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3735 rp->rhs[0]->index,1-rp->nrhs);
3736 rp->codePrefix = Strsafe(append_str(0,0,0,0));
3737 rp->noCode = 0;
3739 }else if( rp->lhsalias==0 ){
3740 /* There is no LHS value symbol. */
3741 lhsdirect = 1;
3742 }else if( strcmp(rp->lhsalias,rp->rhsalias[0])==0 ){
3743 /* The LHS symbol and the left-most RHS symbol are the same, so
3744 ** direct writing is allowed */
3745 lhsdirect = 1;
3746 lhsused = 1;
3747 used[0] = 1;
3748 if( rp->lhs->dtnum!=rp->rhs[0]->dtnum ){
3749 ErrorMsg(lemp->filename,rp->ruleline,
3750 "%s(%s) and %s(%s) share the same label but have "
3751 "different datatypes.",
3752 rp->lhs->name, rp->lhsalias, rp->rhs[0]->name, rp->rhsalias[0]);
3753 lemp->errorcnt++;
3755 }else{
3756 lemon_sprintf(zOvwrt, "/*%s-overwrites-%s*/",
3757 rp->lhsalias, rp->rhsalias[0]);
3758 zSkip = strstr(rp->code, zOvwrt);
3759 if( zSkip!=0 ){
3760 /* The code contains a special comment that indicates that it is safe
3761 ** for the LHS label to overwrite left-most RHS label. */
3762 lhsdirect = 1;
3763 }else{
3764 lhsdirect = 0;
3767 if( lhsdirect ){
3768 sprintf(zLhs, "yymsp[%d].minor.yy%d",1-rp->nrhs,rp->lhs->dtnum);
3769 }else{
3770 rc = 1;
3771 sprintf(zLhs, "yylhsminor.yy%d",rp->lhs->dtnum);
3774 append_str(0,0,0,0);
3776 /* This const cast is wrong but harmless, if we're careful. */
3777 for(cp=(char *)rp->code; *cp; cp++){
3778 if( cp==zSkip ){
3779 append_str(zOvwrt,0,0,0);
3780 cp += lemonStrlen(zOvwrt)-1;
3781 dontUseRhs0 = 1;
3782 continue;
3784 if( ISALPHA(*cp) && (cp==rp->code || (!ISALNUM(cp[-1]) && cp[-1]!='_')) ){
3785 char saved;
3786 for(xp= &cp[1]; ISALNUM(*xp) || *xp=='_'; xp++);
3787 saved = *xp;
3788 *xp = 0;
3789 if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
3790 append_str(zLhs,0,0,0);
3791 cp = xp;
3792 lhsused = 1;
3793 }else{
3794 for(i=0; i<rp->nrhs; i++){
3795 if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
3796 if( i==0 && dontUseRhs0 ){
3797 ErrorMsg(lemp->filename,rp->ruleline,
3798 "Label %s used after '%s'.",
3799 rp->rhsalias[0], zOvwrt);
3800 lemp->errorcnt++;
3801 }else if( cp!=rp->code && cp[-1]=='@' ){
3802 /* If the argument is of the form @X then substituted
3803 ** the token number of X, not the value of X */
3804 append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
3805 }else{
3806 struct symbol *sp = rp->rhs[i];
3807 int dtnum;
3808 if( sp->type==MULTITERMINAL ){
3809 dtnum = sp->subsym[0]->dtnum;
3810 }else{
3811 dtnum = sp->dtnum;
3813 append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);
3815 cp = xp;
3816 used[i] = 1;
3817 break;
3821 *xp = saved;
3823 append_str(cp, 1, 0, 0);
3824 } /* End loop */
3826 /* Main code generation completed */
3827 cp = append_str(0,0,0,0);
3828 if( cp && cp[0] ) rp->code = Strsafe(cp);
3829 append_str(0,0,0,0);
3831 /* Check to make sure the LHS has been used */
3832 if( rp->lhsalias && !lhsused ){
3833 ErrorMsg(lemp->filename,rp->ruleline,
3834 "Label \"%s\" for \"%s(%s)\" is never used.",
3835 rp->lhsalias,rp->lhs->name,rp->lhsalias);
3836 lemp->errorcnt++;
3839 /* Generate destructor code for RHS minor values which are not referenced.
3840 ** Generate error messages for unused labels and duplicate labels.
3842 for(i=0; i<rp->nrhs; i++){
3843 if( rp->rhsalias[i] ){
3844 if( i>0 ){
3845 int j;
3846 if( rp->lhsalias && strcmp(rp->lhsalias,rp->rhsalias[i])==0 ){
3847 ErrorMsg(lemp->filename,rp->ruleline,
3848 "%s(%s) has the same label as the LHS but is not the left-most "
3849 "symbol on the RHS.",
3850 rp->rhs[i]->name, rp->rhsalias);
3851 lemp->errorcnt++;
3853 for(j=0; j<i; j++){
3854 if( rp->rhsalias[j] && strcmp(rp->rhsalias[j],rp->rhsalias[i])==0 ){
3855 ErrorMsg(lemp->filename,rp->ruleline,
3856 "Label %s used for multiple symbols on the RHS of a rule.",
3857 rp->rhsalias[i]);
3858 lemp->errorcnt++;
3859 break;
3863 if( !used[i] ){
3864 ErrorMsg(lemp->filename,rp->ruleline,
3865 "Label %s for \"%s(%s)\" is never used.",
3866 rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
3867 lemp->errorcnt++;
3869 }else if( i>0 && has_destructor(rp->rhs[i],lemp) ){
3870 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3871 rp->rhs[i]->index,i-rp->nrhs+1);
3875 /* If unable to write LHS values directly into the stack, write the
3876 ** saved LHS value now. */
3877 if( lhsdirect==0 ){
3878 append_str(" yymsp[%d].minor.yy%d = ", 0, 1-rp->nrhs, rp->lhs->dtnum);
3879 append_str(zLhs, 0, 0, 0);
3880 append_str(";\n", 0, 0, 0);
3883 /* Suffix code generation complete */
3884 cp = append_str(0,0,0,0);
3885 if( cp && cp[0] ){
3886 rp->codeSuffix = Strsafe(cp);
3887 rp->noCode = 0;
3890 return rc;
3894 ** Generate code which executes when the rule "rp" is reduced. Write
3895 ** the code to "out". Make sure lineno stays up-to-date.
3897 PRIVATE void emit_code(
3898 FILE *out,
3899 struct rule *rp,
3900 struct lemon *lemp,
3901 int *lineno
3903 const char *cp;
3905 /* Setup code prior to the #line directive */
3906 if( rp->codePrefix && rp->codePrefix[0] ){
3907 fprintf(out, "{%s", rp->codePrefix);
3908 for(cp=rp->codePrefix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
3911 /* Generate code to do the reduce action */
3912 if( rp->code ){
3913 if( !lemp->nolinenosflag ){
3914 (*lineno)++;
3915 tplt_linedir(out,rp->line,lemp->filename);
3917 fprintf(out,"{%s",rp->code);
3918 for(cp=rp->code; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
3919 fprintf(out,"}\n"); (*lineno)++;
3920 if( !lemp->nolinenosflag ){
3921 (*lineno)++;
3922 tplt_linedir(out,*lineno,lemp->outname);
3926 /* Generate breakdown code that occurs after the #line directive */
3927 if( rp->codeSuffix && rp->codeSuffix[0] ){
3928 fprintf(out, "%s", rp->codeSuffix);
3929 for(cp=rp->codeSuffix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
3932 if( rp->codePrefix ){
3933 fprintf(out, "}\n"); (*lineno)++;
3936 return;
3940 ** Print the definition of the union used for the parser's data stack.
3941 ** This union contains fields for every possible data type for tokens
3942 ** and nonterminals. In the process of computing and printing this
3943 ** union, also set the ".dtnum" field of every terminal and nonterminal
3944 ** symbol.
3946 void print_stack_union(
3947 FILE *out, /* The output stream */
3948 struct lemon *lemp, /* The main info structure for this parser */
3949 int *plineno, /* Pointer to the line number */
3950 int mhflag /* True if generating makeheaders output */
3952 int lineno = *plineno; /* The line number of the output */
3953 char **types; /* A hash table of datatypes */
3954 int arraysize; /* Size of the "types" array */
3955 int maxdtlength; /* Maximum length of any ".datatype" field. */
3956 char *stddt; /* Standardized name for a datatype */
3957 int i,j; /* Loop counters */
3958 unsigned hash; /* For hashing the name of a type */
3959 const char *name; /* Name of the parser */
3961 /* Allocate and initialize types[] and allocate stddt[] */
3962 arraysize = lemp->nsymbol * 2;
3963 types = (char**)calloc( arraysize, sizeof(char*) );
3964 if( types==0 ){
3965 fprintf(stderr,"Out of memory.\n");
3966 exit(1);
3968 for(i=0; i<arraysize; i++) types[i] = 0;
3969 maxdtlength = 0;
3970 if( lemp->vartype ){
3971 maxdtlength = lemonStrlen(lemp->vartype);
3973 for(i=0; i<lemp->nsymbol; i++){
3974 int len;
3975 struct symbol *sp = lemp->symbols[i];
3976 if( sp->datatype==0 ) continue;
3977 len = lemonStrlen(sp->datatype);
3978 if( len>maxdtlength ) maxdtlength = len;
3980 stddt = (char*)malloc( maxdtlength*2 + 1 );
3981 if( stddt==0 ){
3982 fprintf(stderr,"Out of memory.\n");
3983 exit(1);
3986 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3987 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
3988 ** used for terminal symbols. If there is no %default_type defined then
3989 ** 0 is also used as the .dtnum value for nonterminals which do not specify
3990 ** a datatype using the %type directive.
3992 for(i=0; i<lemp->nsymbol; i++){
3993 struct symbol *sp = lemp->symbols[i];
3994 char *cp;
3995 if( sp==lemp->errsym ){
3996 sp->dtnum = arraysize+1;
3997 continue;
3999 if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
4000 sp->dtnum = 0;
4001 continue;
4003 cp = sp->datatype;
4004 if( cp==0 ) cp = lemp->vartype;
4005 j = 0;
4006 while( ISSPACE(*cp) ) cp++;
4007 while( *cp ) stddt[j++] = *cp++;
4008 while( j>0 && ISSPACE(stddt[j-1]) ) j--;
4009 stddt[j] = 0;
4010 if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){
4011 sp->dtnum = 0;
4012 continue;
4014 hash = 0;
4015 for(j=0; stddt[j]; j++){
4016 hash = hash*53 + stddt[j];
4018 hash = (hash & 0x7fffffff)%arraysize;
4019 while( types[hash] ){
4020 if( strcmp(types[hash],stddt)==0 ){
4021 sp->dtnum = hash + 1;
4022 break;
4024 hash++;
4025 if( hash>=(unsigned)arraysize ) hash = 0;
4027 if( types[hash]==0 ){
4028 sp->dtnum = hash + 1;
4029 types[hash] = (char*)malloc( lemonStrlen(stddt)+1 );
4030 if( types[hash]==0 ){
4031 fprintf(stderr,"Out of memory.\n");
4032 exit(1);
4034 lemon_strcpy(types[hash],stddt);
4038 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
4039 name = lemp->name ? lemp->name : "Parse";
4040 lineno = *plineno;
4041 if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
4042 fprintf(out,"#define %sTOKENTYPE %s\n",name,
4043 lemp->tokentype?lemp->tokentype:"void*"); lineno++;
4044 if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
4045 fprintf(out,"typedef union {\n"); lineno++;
4046 fprintf(out," int yyinit;\n"); lineno++;
4047 fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++;
4048 for(i=0; i<arraysize; i++){
4049 if( types[i]==0 ) continue;
4050 fprintf(out," %s yy%d;\n",types[i],i+1); lineno++;
4051 free(types[i]);
4053 if( lemp->errsym && lemp->errsym->useCnt ){
4054 fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++;
4056 free(stddt);
4057 free(types);
4058 fprintf(out,"} YYMINORTYPE;\n"); lineno++;
4059 *plineno = lineno;
4063 ** Return the name of a C datatype able to represent values between
4064 ** lwr and upr, inclusive. If pnByte!=NULL then also write the sizeof
4065 ** for that type (1, 2, or 4) into *pnByte.
4067 static const char *minimum_size_type(int lwr, int upr, int *pnByte){
4068 const char *zType = "int";
4069 int nByte = 4;
4070 if( lwr>=0 ){
4071 if( upr<=255 ){
4072 zType = "unsigned char";
4073 nByte = 1;
4074 }else if( upr<65535 ){
4075 zType = "unsigned short int";
4076 nByte = 2;
4077 }else{
4078 zType = "unsigned int";
4079 nByte = 4;
4081 }else if( lwr>=-127 && upr<=127 ){
4082 zType = "signed char";
4083 nByte = 1;
4084 }else if( lwr>=-32767 && upr<32767 ){
4085 zType = "short";
4086 nByte = 2;
4088 if( pnByte ) *pnByte = nByte;
4089 return zType;
4093 ** Each state contains a set of token transaction and a set of
4094 ** nonterminal transactions. Each of these sets makes an instance
4095 ** of the following structure. An array of these structures is used
4096 ** to order the creation of entries in the yy_action[] table.
4098 struct axset {
4099 struct state *stp; /* A pointer to a state */
4100 int isTkn; /* True to use tokens. False for non-terminals */
4101 int nAction; /* Number of actions */
4102 int iOrder; /* Original order of action sets */
4106 ** Compare to axset structures for sorting purposes
4108 static int axset_compare(const void *a, const void *b){
4109 struct axset *p1 = (struct axset*)a;
4110 struct axset *p2 = (struct axset*)b;
4111 int c;
4112 c = p2->nAction - p1->nAction;
4113 if( c==0 ){
4114 c = p1->iOrder - p2->iOrder;
4116 assert( c!=0 || p1==p2 );
4117 return c;
4121 ** Write text on "out" that describes the rule "rp".
4123 static void writeRuleText(FILE *out, struct rule *rp){
4124 int j;
4125 fprintf(out,"%s ::=", rp->lhs->name);
4126 for(j=0; j<rp->nrhs; j++){
4127 struct symbol *sp = rp->rhs[j];
4128 if( sp->type!=MULTITERMINAL ){
4129 fprintf(out," %s", sp->name);
4130 }else{
4131 int k;
4132 fprintf(out," %s", sp->subsym[0]->name);
4133 for(k=1; k<sp->nsubsym; k++){
4134 fprintf(out,"|%s",sp->subsym[k]->name);
4141 /* Generate C source code for the parser */
4142 void ReportTable(
4143 struct lemon *lemp,
4144 int mhflag /* Output in makeheaders format if true */
4146 FILE *out, *in;
4147 char line[LINESIZE];
4148 int lineno;
4149 struct state *stp;
4150 struct action *ap;
4151 struct rule *rp;
4152 struct acttab *pActtab;
4153 int i, j, n, sz;
4154 int szActionType; /* sizeof(YYACTIONTYPE) */
4155 int szCodeType; /* sizeof(YYCODETYPE) */
4156 const char *name;
4157 int mnTknOfst, mxTknOfst;
4158 int mnNtOfst, mxNtOfst;
4159 struct axset *ax;
4161 lemp->minShiftReduce = lemp->nstate;
4162 lemp->errAction = lemp->minShiftReduce + lemp->nrule;
4163 lemp->accAction = lemp->errAction + 1;
4164 lemp->noAction = lemp->accAction + 1;
4165 lemp->minReduce = lemp->noAction + 1;
4166 lemp->maxAction = lemp->minReduce + lemp->nrule;
4168 in = tplt_open(lemp);
4169 if( in==0 ) return;
4170 out = file_open(lemp,".c","wb");
4171 if( out==0 ){
4172 fclose(in);
4173 return;
4175 lineno = 1;
4176 tplt_xfer(lemp->name,in,out,&lineno);
4178 /* Generate the include code, if any */
4179 tplt_print(out,lemp,lemp->include,&lineno);
4180 if( mhflag ){
4181 char *incName = file_makename(lemp, ".h");
4182 fprintf(out,"#include \"%s\"\n", incName); lineno++;
4183 free(incName);
4185 tplt_xfer(lemp->name,in,out,&lineno);
4187 /* Generate #defines for all tokens */
4188 if( mhflag ){
4189 const char *prefix;
4190 fprintf(out,"#if INTERFACE\n"); lineno++;
4191 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4192 else prefix = "";
4193 for(i=1; i<lemp->nterminal; i++){
4194 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
4195 lineno++;
4197 fprintf(out,"#endif\n"); lineno++;
4199 tplt_xfer(lemp->name,in,out,&lineno);
4201 /* Generate the defines */
4202 fprintf(out,"#define YYCODETYPE %s\n",
4203 minimum_size_type(0, lemp->nsymbol, &szCodeType)); lineno++;
4204 fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol); lineno++;
4205 fprintf(out,"#define YYACTIONTYPE %s\n",
4206 minimum_size_type(0,lemp->maxAction,&szActionType)); lineno++;
4207 if( lemp->wildcard ){
4208 fprintf(out,"#define YYWILDCARD %d\n",
4209 lemp->wildcard->index); lineno++;
4211 print_stack_union(out,lemp,&lineno,mhflag);
4212 fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++;
4213 if( lemp->stacksize ){
4214 fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++;
4215 }else{
4216 fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++;
4218 fprintf(out, "#endif\n"); lineno++;
4219 if( mhflag ){
4220 fprintf(out,"#if INTERFACE\n"); lineno++;
4222 name = lemp->name ? lemp->name : "Parse";
4223 if( lemp->arg && lemp->arg[0] ){
4224 i = lemonStrlen(lemp->arg);
4225 while( i>=1 && ISSPACE(lemp->arg[i-1]) ) i--;
4226 while( i>=1 && (ISALNUM(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
4227 fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++;
4228 fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++;
4229 fprintf(out,"#define %sARG_PARAM ,%s\n",name,&lemp->arg[i]); lineno++;
4230 fprintf(out,"#define %sARG_FETCH %s=yypParser->%s;\n",
4231 name,lemp->arg,&lemp->arg[i]); lineno++;
4232 fprintf(out,"#define %sARG_STORE yypParser->%s=%s;\n",
4233 name,&lemp->arg[i],&lemp->arg[i]); lineno++;
4234 }else{
4235 fprintf(out,"#define %sARG_SDECL\n",name); lineno++;
4236 fprintf(out,"#define %sARG_PDECL\n",name); lineno++;
4237 fprintf(out,"#define %sARG_PARAM\n",name); lineno++;
4238 fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
4239 fprintf(out,"#define %sARG_STORE\n",name); lineno++;
4241 if( lemp->ctx && lemp->ctx[0] ){
4242 i = lemonStrlen(lemp->ctx);
4243 while( i>=1 && ISSPACE(lemp->ctx[i-1]) ) i--;
4244 while( i>=1 && (ISALNUM(lemp->ctx[i-1]) || lemp->ctx[i-1]=='_') ) i--;
4245 fprintf(out,"#define %sCTX_SDECL %s;\n",name,lemp->ctx); lineno++;
4246 fprintf(out,"#define %sCTX_PDECL ,%s\n",name,lemp->ctx); lineno++;
4247 fprintf(out,"#define %sCTX_PARAM ,%s\n",name,&lemp->ctx[i]); lineno++;
4248 fprintf(out,"#define %sCTX_FETCH %s=yypParser->%s;\n",
4249 name,lemp->ctx,&lemp->ctx[i]); lineno++;
4250 fprintf(out,"#define %sCTX_STORE yypParser->%s=%s;\n",
4251 name,&lemp->ctx[i],&lemp->ctx[i]); lineno++;
4252 }else{
4253 fprintf(out,"#define %sCTX_SDECL\n",name); lineno++;
4254 fprintf(out,"#define %sCTX_PDECL\n",name); lineno++;
4255 fprintf(out,"#define %sCTX_PARAM\n",name); lineno++;
4256 fprintf(out,"#define %sCTX_FETCH\n",name); lineno++;
4257 fprintf(out,"#define %sCTX_STORE\n",name); lineno++;
4259 if( mhflag ){
4260 fprintf(out,"#endif\n"); lineno++;
4262 if( lemp->errsym && lemp->errsym->useCnt ){
4263 fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++;
4264 fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++;
4266 if( lemp->has_fallback ){
4267 fprintf(out,"#define YYFALLBACK 1\n"); lineno++;
4270 /* Compute the action table, but do not output it yet. The action
4271 ** table must be computed before generating the YYNSTATE macro because
4272 ** we need to know how many states can be eliminated.
4274 ax = (struct axset *) calloc(lemp->nxstate*2, sizeof(ax[0]));
4275 if( ax==0 ){
4276 fprintf(stderr,"malloc failed\n");
4277 exit(1);
4279 for(i=0; i<lemp->nxstate; i++){
4280 stp = lemp->sorted[i];
4281 ax[i*2].stp = stp;
4282 ax[i*2].isTkn = 1;
4283 ax[i*2].nAction = stp->nTknAct;
4284 ax[i*2+1].stp = stp;
4285 ax[i*2+1].isTkn = 0;
4286 ax[i*2+1].nAction = stp->nNtAct;
4288 mxTknOfst = mnTknOfst = 0;
4289 mxNtOfst = mnNtOfst = 0;
4290 /* In an effort to minimize the action table size, use the heuristic
4291 ** of placing the largest action sets first */
4292 for(i=0; i<lemp->nxstate*2; i++) ax[i].iOrder = i;
4293 qsort(ax, lemp->nxstate*2, sizeof(ax[0]), axset_compare);
4294 pActtab = acttab_alloc(lemp->nsymbol, lemp->nterminal);
4295 for(i=0; i<lemp->nxstate*2 && ax[i].nAction>0; i++){
4296 stp = ax[i].stp;
4297 if( ax[i].isTkn ){
4298 for(ap=stp->ap; ap; ap=ap->next){
4299 int action;
4300 if( ap->sp->index>=lemp->nterminal ) continue;
4301 action = compute_action(lemp, ap);
4302 if( action<0 ) continue;
4303 acttab_action(pActtab, ap->sp->index, action);
4305 stp->iTknOfst = acttab_insert(pActtab, 1);
4306 if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
4307 if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
4308 }else{
4309 for(ap=stp->ap; ap; ap=ap->next){
4310 int action;
4311 if( ap->sp->index<lemp->nterminal ) continue;
4312 if( ap->sp->index==lemp->nsymbol ) continue;
4313 action = compute_action(lemp, ap);
4314 if( action<0 ) continue;
4315 acttab_action(pActtab, ap->sp->index, action);
4317 stp->iNtOfst = acttab_insert(pActtab, 0);
4318 if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
4319 if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
4321 #if 0 /* Uncomment for a trace of how the yy_action[] table fills out */
4322 { int jj, nn;
4323 for(jj=nn=0; jj<pActtab->nAction; jj++){
4324 if( pActtab->aAction[jj].action<0 ) nn++;
4326 printf("%4d: State %3d %s n: %2d size: %5d freespace: %d\n",
4327 i, stp->statenum, ax[i].isTkn ? "Token" : "Var ",
4328 ax[i].nAction, pActtab->nAction, nn);
4330 #endif
4332 free(ax);
4334 /* Mark rules that are actually used for reduce actions after all
4335 ** optimizations have been applied
4337 for(rp=lemp->rule; rp; rp=rp->next) rp->doesReduce = LEMON_FALSE;
4338 for(i=0; i<lemp->nxstate; i++){
4339 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
4340 if( ap->type==REDUCE || ap->type==SHIFTREDUCE ){
4341 ap->x.rp->doesReduce = 1;
4346 /* Finish rendering the constants now that the action table has
4347 ** been computed */
4348 fprintf(out,"#define YYNSTATE %d\n",lemp->nxstate); lineno++;
4349 fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++;
4350 fprintf(out,"#define YYNTOKEN %d\n",lemp->nterminal); lineno++;
4351 fprintf(out,"#define YY_MAX_SHIFT %d\n",lemp->nxstate-1); lineno++;
4352 i = lemp->minShiftReduce;
4353 fprintf(out,"#define YY_MIN_SHIFTREDUCE %d\n",i); lineno++;
4354 i += lemp->nrule;
4355 fprintf(out,"#define YY_MAX_SHIFTREDUCE %d\n", i-1); lineno++;
4356 fprintf(out,"#define YY_ERROR_ACTION %d\n", lemp->errAction); lineno++;
4357 fprintf(out,"#define YY_ACCEPT_ACTION %d\n", lemp->accAction); lineno++;
4358 fprintf(out,"#define YY_NO_ACTION %d\n", lemp->noAction); lineno++;
4359 fprintf(out,"#define YY_MIN_REDUCE %d\n", lemp->minReduce); lineno++;
4360 i = lemp->minReduce + lemp->nrule;
4361 fprintf(out,"#define YY_MAX_REDUCE %d\n", i-1); lineno++;
4362 tplt_xfer(lemp->name,in,out,&lineno);
4364 /* Now output the action table and its associates:
4366 ** yy_action[] A single table containing all actions.
4367 ** yy_lookahead[] A table containing the lookahead for each entry in
4368 ** yy_action. Used to detect hash collisions.
4369 ** yy_shift_ofst[] For each state, the offset into yy_action for
4370 ** shifting terminals.
4371 ** yy_reduce_ofst[] For each state, the offset into yy_action for
4372 ** shifting non-terminals after a reduce.
4373 ** yy_default[] Default action for each state.
4376 /* Output the yy_action table */
4377 lemp->nactiontab = n = acttab_action_size(pActtab);
4378 lemp->tablesize += n*szActionType;
4379 fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++;
4380 fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++;
4381 for(i=j=0; i<n; i++){
4382 int action = acttab_yyaction(pActtab, i);
4383 if( action<0 ) action = lemp->noAction;
4384 if( j==0 ) fprintf(out," /* %5d */ ", i);
4385 fprintf(out, " %4d,", action);
4386 if( j==9 || i==n-1 ){
4387 fprintf(out, "\n"); lineno++;
4388 j = 0;
4389 }else{
4390 j++;
4393 fprintf(out, "};\n"); lineno++;
4395 /* Output the yy_lookahead table */
4396 lemp->nlookaheadtab = n = acttab_lookahead_size(pActtab);
4397 lemp->tablesize += n*szCodeType;
4398 fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++;
4399 for(i=j=0; i<n; i++){
4400 int la = acttab_yylookahead(pActtab, i);
4401 if( la<0 ) la = lemp->nsymbol;
4402 if( j==0 ) fprintf(out," /* %5d */ ", i);
4403 fprintf(out, " %4d,", la);
4404 if( j==9 || i==n-1 ){
4405 fprintf(out, "\n"); lineno++;
4406 j = 0;
4407 }else{
4408 j++;
4411 fprintf(out, "};\n"); lineno++;
4413 /* Output the yy_shift_ofst[] table */
4414 n = lemp->nxstate;
4415 while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--;
4416 fprintf(out, "#define YY_SHIFT_COUNT (%d)\n", n-1); lineno++;
4417 fprintf(out, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst); lineno++;
4418 fprintf(out, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst); lineno++;
4419 fprintf(out, "static const %s yy_shift_ofst[] = {\n",
4420 minimum_size_type(mnTknOfst, lemp->nterminal+lemp->nactiontab, &sz));
4421 lineno++;
4422 lemp->tablesize += n*sz;
4423 for(i=j=0; i<n; i++){
4424 int ofst;
4425 stp = lemp->sorted[i];
4426 ofst = stp->iTknOfst;
4427 if( ofst==NO_OFFSET ) ofst = lemp->nactiontab;
4428 if( j==0 ) fprintf(out," /* %5d */ ", i);
4429 fprintf(out, " %4d,", ofst);
4430 if( j==9 || i==n-1 ){
4431 fprintf(out, "\n"); lineno++;
4432 j = 0;
4433 }else{
4434 j++;
4437 fprintf(out, "};\n"); lineno++;
4439 /* Output the yy_reduce_ofst[] table */
4440 n = lemp->nxstate;
4441 while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--;
4442 fprintf(out, "#define YY_REDUCE_COUNT (%d)\n", n-1); lineno++;
4443 fprintf(out, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst); lineno++;
4444 fprintf(out, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst); lineno++;
4445 fprintf(out, "static const %s yy_reduce_ofst[] = {\n",
4446 minimum_size_type(mnNtOfst-1, mxNtOfst, &sz)); lineno++;
4447 lemp->tablesize += n*sz;
4448 for(i=j=0; i<n; i++){
4449 int ofst;
4450 stp = lemp->sorted[i];
4451 ofst = stp->iNtOfst;
4452 if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
4453 if( j==0 ) fprintf(out," /* %5d */ ", i);
4454 fprintf(out, " %4d,", ofst);
4455 if( j==9 || i==n-1 ){
4456 fprintf(out, "\n"); lineno++;
4457 j = 0;
4458 }else{
4459 j++;
4462 fprintf(out, "};\n"); lineno++;
4464 /* Output the default action table */
4465 fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++;
4466 n = lemp->nxstate;
4467 lemp->tablesize += n*szActionType;
4468 for(i=j=0; i<n; i++){
4469 stp = lemp->sorted[i];
4470 if( j==0 ) fprintf(out," /* %5d */ ", i);
4471 if( stp->iDfltReduce<0 ){
4472 fprintf(out, " %4d,", lemp->errAction);
4473 }else{
4474 fprintf(out, " %4d,", stp->iDfltReduce + lemp->minReduce);
4476 if( j==9 || i==n-1 ){
4477 fprintf(out, "\n"); lineno++;
4478 j = 0;
4479 }else{
4480 j++;
4483 fprintf(out, "};\n"); lineno++;
4484 tplt_xfer(lemp->name,in,out,&lineno);
4486 /* Generate the table of fallback tokens.
4488 if( lemp->has_fallback ){
4489 int mx = lemp->nterminal - 1;
4490 while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; }
4491 lemp->tablesize += (mx+1)*szCodeType;
4492 for(i=0; i<=mx; i++){
4493 struct symbol *p = lemp->symbols[i];
4494 if( p->fallback==0 ){
4495 fprintf(out, " 0, /* %10s => nothing */\n", p->name);
4496 }else{
4497 fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index,
4498 p->name, p->fallback->name);
4500 lineno++;
4503 tplt_xfer(lemp->name, in, out, &lineno);
4505 /* Generate a table containing the symbolic name of every symbol
4507 for(i=0; i<lemp->nsymbol; i++){
4508 lemon_sprintf(line,"\"%s\",",lemp->symbols[i]->name);
4509 fprintf(out," /* %4d */ \"%s\",\n",i, lemp->symbols[i]->name); lineno++;
4511 tplt_xfer(lemp->name,in,out,&lineno);
4513 /* Generate a table containing a text string that describes every
4514 ** rule in the rule set of the grammar. This information is used
4515 ** when tracing REDUCE actions.
4517 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4518 assert( rp->iRule==i );
4519 fprintf(out," /* %3d */ \"", i);
4520 writeRuleText(out, rp);
4521 fprintf(out,"\",\n"); lineno++;
4523 tplt_xfer(lemp->name,in,out,&lineno);
4525 /* Generate code which executes every time a symbol is popped from
4526 ** the stack while processing errors or while destroying the parser.
4527 ** (In other words, generate the %destructor actions)
4529 if( lemp->tokendest ){
4530 int once = 1;
4531 for(i=0; i<lemp->nsymbol; i++){
4532 struct symbol *sp = lemp->symbols[i];
4533 if( sp==0 || sp->type!=TERMINAL ) continue;
4534 if( once ){
4535 fprintf(out, " /* TERMINAL Destructor */\n"); lineno++;
4536 once = 0;
4538 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4540 for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
4541 if( i<lemp->nsymbol ){
4542 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4543 fprintf(out," break;\n"); lineno++;
4546 if( lemp->vardest ){
4547 struct symbol *dflt_sp = 0;
4548 int once = 1;
4549 for(i=0; i<lemp->nsymbol; i++){
4550 struct symbol *sp = lemp->symbols[i];
4551 if( sp==0 || sp->type==TERMINAL ||
4552 sp->index<=0 || sp->destructor!=0 ) continue;
4553 if( once ){
4554 fprintf(out, " /* Default NON-TERMINAL Destructor */\n");lineno++;
4555 once = 0;
4557 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4558 dflt_sp = sp;
4560 if( dflt_sp!=0 ){
4561 emit_destructor_code(out,dflt_sp,lemp,&lineno);
4563 fprintf(out," break;\n"); lineno++;
4565 for(i=0; i<lemp->nsymbol; i++){
4566 struct symbol *sp = lemp->symbols[i];
4567 if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
4568 if( sp->destLineno<0 ) continue; /* Already emitted */
4569 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4571 /* Combine duplicate destructors into a single case */
4572 for(j=i+1; j<lemp->nsymbol; j++){
4573 struct symbol *sp2 = lemp->symbols[j];
4574 if( sp2 && sp2->type!=TERMINAL && sp2->destructor
4575 && sp2->dtnum==sp->dtnum
4576 && strcmp(sp->destructor,sp2->destructor)==0 ){
4577 fprintf(out," case %d: /* %s */\n",
4578 sp2->index, sp2->name); lineno++;
4579 sp2->destLineno = -1; /* Avoid emitting this destructor again */
4583 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4584 fprintf(out," break;\n"); lineno++;
4586 tplt_xfer(lemp->name,in,out,&lineno);
4588 /* Generate code which executes whenever the parser stack overflows */
4589 tplt_print(out,lemp,lemp->overflow,&lineno);
4590 tplt_xfer(lemp->name,in,out,&lineno);
4592 /* Generate the table of rule information
4594 ** Note: This code depends on the fact that rules are number
4595 ** sequentually beginning with 0.
4597 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4598 fprintf(out," { %4d, %4d }, /* (%d) ",rp->lhs->index,-rp->nrhs,i);
4599 rule_print(out, rp);
4600 fprintf(out," */\n"); lineno++;
4602 tplt_xfer(lemp->name,in,out,&lineno);
4604 /* Generate code which execution during each REDUCE action */
4605 i = 0;
4606 for(rp=lemp->rule; rp; rp=rp->next){
4607 i += translate_code(lemp, rp);
4609 if( i ){
4610 fprintf(out," YYMINORTYPE yylhsminor;\n"); lineno++;
4612 /* First output rules other than the default: rule */
4613 for(rp=lemp->rule; rp; rp=rp->next){
4614 struct rule *rp2; /* Other rules with the same action */
4615 if( rp->codeEmitted ) continue;
4616 if( rp->noCode ){
4617 /* No C code actions, so this will be part of the "default:" rule */
4618 continue;
4620 fprintf(out," case %d: /* ", rp->iRule);
4621 writeRuleText(out, rp);
4622 fprintf(out, " */\n"); lineno++;
4623 for(rp2=rp->next; rp2; rp2=rp2->next){
4624 if( rp2->code==rp->code && rp2->codePrefix==rp->codePrefix
4625 && rp2->codeSuffix==rp->codeSuffix ){
4626 fprintf(out," case %d: /* ", rp2->iRule);
4627 writeRuleText(out, rp2);
4628 fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->iRule); lineno++;
4629 rp2->codeEmitted = 1;
4632 emit_code(out,rp,lemp,&lineno);
4633 fprintf(out," break;\n"); lineno++;
4634 rp->codeEmitted = 1;
4636 /* Finally, output the default: rule. We choose as the default: all
4637 ** empty actions. */
4638 fprintf(out," default:\n"); lineno++;
4639 for(rp=lemp->rule; rp; rp=rp->next){
4640 if( rp->codeEmitted ) continue;
4641 assert( rp->noCode );
4642 fprintf(out," /* (%d) ", rp->iRule);
4643 writeRuleText(out, rp);
4644 if( rp->doesReduce ){
4645 fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->iRule); lineno++;
4646 }else{
4647 fprintf(out, " (OPTIMIZED OUT) */ assert(yyruleno!=%d);\n",
4648 rp->iRule); lineno++;
4651 fprintf(out," break;\n"); lineno++;
4652 tplt_xfer(lemp->name,in,out,&lineno);
4654 /* Generate code which executes if a parse fails */
4655 tplt_print(out,lemp,lemp->failure,&lineno);
4656 tplt_xfer(lemp->name,in,out,&lineno);
4658 /* Generate code which executes when a syntax error occurs */
4659 tplt_print(out,lemp,lemp->error,&lineno);
4660 tplt_xfer(lemp->name,in,out,&lineno);
4662 /* Generate code which executes when the parser accepts its input */
4663 tplt_print(out,lemp,lemp->accept,&lineno);
4664 tplt_xfer(lemp->name,in,out,&lineno);
4666 /* Append any addition code the user desires */
4667 tplt_print(out,lemp,lemp->extracode,&lineno);
4669 fclose(in);
4670 fclose(out);
4671 return;
4674 /* Generate a header file for the parser */
4675 void ReportHeader(struct lemon *lemp)
4677 FILE *out, *in;
4678 const char *prefix;
4679 char line[LINESIZE];
4680 char pattern[LINESIZE];
4681 int i;
4683 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4684 else prefix = "";
4685 in = file_open(lemp,".h","rb");
4686 if( in ){
4687 int nextChar;
4688 for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
4689 lemon_sprintf(pattern,"#define %s%-30s %3d\n",
4690 prefix,lemp->symbols[i]->name,i);
4691 if( strcmp(line,pattern) ) break;
4693 nextChar = fgetc(in);
4694 fclose(in);
4695 if( i==lemp->nterminal && nextChar==EOF ){
4696 /* No change in the file. Don't rewrite it. */
4697 return;
4700 out = file_open(lemp,".h","wb");
4701 if( out ){
4702 for(i=1; i<lemp->nterminal; i++){
4703 fprintf(out,"#define %s%-30s %3d\n",prefix,lemp->symbols[i]->name,i);
4705 fclose(out);
4707 return;
4710 /* Reduce the size of the action tables, if possible, by making use
4711 ** of defaults.
4713 ** In this version, we take the most frequent REDUCE action and make
4714 ** it the default. Except, there is no default if the wildcard token
4715 ** is a possible look-ahead.
4717 void CompressTables(struct lemon *lemp)
4719 struct state *stp;
4720 struct action *ap, *ap2, *nextap;
4721 struct rule *rp, *rp2, *rbest;
4722 int nbest, n;
4723 int i;
4724 int usesWildcard;
4726 for(i=0; i<lemp->nstate; i++){
4727 stp = lemp->sorted[i];
4728 nbest = 0;
4729 rbest = 0;
4730 usesWildcard = 0;
4732 for(ap=stp->ap; ap; ap=ap->next){
4733 if( ap->type==SHIFT && ap->sp==lemp->wildcard ){
4734 usesWildcard = 1;
4736 if( ap->type!=REDUCE ) continue;
4737 rp = ap->x.rp;
4738 if( rp->lhsStart ) continue;
4739 if( rp==rbest ) continue;
4740 n = 1;
4741 for(ap2=ap->next; ap2; ap2=ap2->next){
4742 if( ap2->type!=REDUCE ) continue;
4743 rp2 = ap2->x.rp;
4744 if( rp2==rbest ) continue;
4745 if( rp2==rp ) n++;
4747 if( n>nbest ){
4748 nbest = n;
4749 rbest = rp;
4753 /* Do not make a default if the number of rules to default
4754 ** is not at least 1 or if the wildcard token is a possible
4755 ** lookahead.
4757 if( nbest<1 || usesWildcard ) continue;
4760 /* Combine matching REDUCE actions into a single default */
4761 for(ap=stp->ap; ap; ap=ap->next){
4762 if( ap->type==REDUCE && ap->x.rp==rbest ) break;
4764 assert( ap );
4765 ap->sp = Symbol_new("{default}");
4766 for(ap=ap->next; ap; ap=ap->next){
4767 if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
4769 stp->ap = Action_sort(stp->ap);
4771 for(ap=stp->ap; ap; ap=ap->next){
4772 if( ap->type==SHIFT ) break;
4773 if( ap->type==REDUCE && ap->x.rp!=rbest ) break;
4775 if( ap==0 ){
4776 stp->autoReduce = 1;
4777 stp->pDfltReduce = rbest;
4781 /* Make a second pass over all states and actions. Convert
4782 ** every action that is a SHIFT to an autoReduce state into
4783 ** a SHIFTREDUCE action.
4785 for(i=0; i<lemp->nstate; i++){
4786 stp = lemp->sorted[i];
4787 for(ap=stp->ap; ap; ap=ap->next){
4788 struct state *pNextState;
4789 if( ap->type!=SHIFT ) continue;
4790 pNextState = ap->x.stp;
4791 if( pNextState->autoReduce && pNextState->pDfltReduce!=0 ){
4792 ap->type = SHIFTREDUCE;
4793 ap->x.rp = pNextState->pDfltReduce;
4798 /* If a SHIFTREDUCE action specifies a rule that has a single RHS term
4799 ** (meaning that the SHIFTREDUCE will land back in the state where it
4800 ** started) and if there is no C-code associated with the reduce action,
4801 ** then we can go ahead and convert the action to be the same as the
4802 ** action for the RHS of the rule.
4804 for(i=0; i<lemp->nstate; i++){
4805 stp = lemp->sorted[i];
4806 for(ap=stp->ap; ap; ap=nextap){
4807 nextap = ap->next;
4808 if( ap->type!=SHIFTREDUCE ) continue;
4809 rp = ap->x.rp;
4810 if( rp->noCode==0 ) continue;
4811 if( rp->nrhs!=1 ) continue;
4812 #if 1
4813 /* Only apply this optimization to non-terminals. It would be OK to
4814 ** apply it to terminal symbols too, but that makes the parser tables
4815 ** larger. */
4816 if( ap->sp->index<lemp->nterminal ) continue;
4817 #endif
4818 /* If we reach this point, it means the optimization can be applied */
4819 nextap = ap;
4820 for(ap2=stp->ap; ap2 && (ap2==ap || ap2->sp!=rp->lhs); ap2=ap2->next){}
4821 assert( ap2!=0 );
4822 ap->spOpt = ap2->sp;
4823 ap->type = ap2->type;
4824 ap->x = ap2->x;
4831 ** Compare two states for sorting purposes. The smaller state is the
4832 ** one with the most non-terminal actions. If they have the same number
4833 ** of non-terminal actions, then the smaller is the one with the most
4834 ** token actions.
4836 static int stateResortCompare(const void *a, const void *b){
4837 const struct state *pA = *(const struct state**)a;
4838 const struct state *pB = *(const struct state**)b;
4839 int n;
4841 n = pB->nNtAct - pA->nNtAct;
4842 if( n==0 ){
4843 n = pB->nTknAct - pA->nTknAct;
4844 if( n==0 ){
4845 n = pB->statenum - pA->statenum;
4848 assert( n!=0 );
4849 return n;
4854 ** Renumber and resort states so that states with fewer choices
4855 ** occur at the end. Except, keep state 0 as the first state.
4857 void ResortStates(struct lemon *lemp)
4859 int i;
4860 struct state *stp;
4861 struct action *ap;
4863 for(i=0; i<lemp->nstate; i++){
4864 stp = lemp->sorted[i];
4865 stp->nTknAct = stp->nNtAct = 0;
4866 stp->iDfltReduce = -1; /* Init dflt action to "syntax error" */
4867 stp->iTknOfst = NO_OFFSET;
4868 stp->iNtOfst = NO_OFFSET;
4869 for(ap=stp->ap; ap; ap=ap->next){
4870 int iAction = compute_action(lemp,ap);
4871 if( iAction>=0 ){
4872 if( ap->sp->index<lemp->nterminal ){
4873 stp->nTknAct++;
4874 }else if( ap->sp->index<lemp->nsymbol ){
4875 stp->nNtAct++;
4876 }else{
4877 assert( stp->autoReduce==0 || stp->pDfltReduce==ap->x.rp );
4878 stp->iDfltReduce = iAction;
4883 qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]),
4884 stateResortCompare);
4885 for(i=0; i<lemp->nstate; i++){
4886 lemp->sorted[i]->statenum = i;
4888 lemp->nxstate = lemp->nstate;
4889 while( lemp->nxstate>1 && lemp->sorted[lemp->nxstate-1]->autoReduce ){
4890 lemp->nxstate--;
4895 /***************** From the file "set.c" ************************************/
4897 ** Set manipulation routines for the LEMON parser generator.
4900 static int size = 0;
4902 /* Set the set size */
4903 void SetSize(int n)
4905 size = n+1;
4908 /* Allocate a new set */
4909 char *SetNew(void){
4910 char *s;
4911 s = (char*)calloc( size, 1);
4912 if( s==0 ){
4913 extern void memory_error();
4914 memory_error();
4916 return s;
4919 /* Deallocate a set */
4920 void SetFree(char *s)
4922 free(s);
4925 /* Add a new element to the set. Return TRUE if the element was added
4926 ** and FALSE if it was already there. */
4927 int SetAdd(char *s, int e)
4929 int rv;
4930 assert( e>=0 && e<size );
4931 rv = s[e];
4932 s[e] = 1;
4933 return !rv;
4936 /* Add every element of s2 to s1. Return TRUE if s1 changes. */
4937 int SetUnion(char *s1, char *s2)
4939 int i, progress;
4940 progress = 0;
4941 for(i=0; i<size; i++){
4942 if( s2[i]==0 ) continue;
4943 if( s1[i]==0 ){
4944 progress = 1;
4945 s1[i] = 1;
4948 return progress;
4950 /********************** From the file "table.c" ****************************/
4952 ** All code in this file has been automatically generated
4953 ** from a specification in the file
4954 ** "table.q"
4955 ** by the associative array code building program "aagen".
4956 ** Do not edit this file! Instead, edit the specification
4957 ** file, then rerun aagen.
4960 ** Code for processing tables in the LEMON parser generator.
4963 PRIVATE unsigned strhash(const char *x)
4965 unsigned h = 0;
4966 while( *x ) h = h*13 + *(x++);
4967 return h;
4970 /* Works like strdup, sort of. Save a string in malloced memory, but
4971 ** keep strings in a table so that the same string is not in more
4972 ** than one place.
4974 const char *Strsafe(const char *y)
4976 const char *z;
4977 char *cpy;
4979 if( y==0 ) return 0;
4980 z = Strsafe_find(y);
4981 if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){
4982 lemon_strcpy(cpy,y);
4983 z = cpy;
4984 Strsafe_insert(z);
4986 MemoryCheck(z);
4987 return z;
4990 /* There is one instance of the following structure for each
4991 ** associative array of type "x1".
4993 struct s_x1 {
4994 int size; /* The number of available slots. */
4995 /* Must be a power of 2 greater than or */
4996 /* equal to 1 */
4997 int count; /* Number of currently slots filled */
4998 struct s_x1node *tbl; /* The data stored here */
4999 struct s_x1node **ht; /* Hash table for lookups */
5002 /* There is one instance of this structure for every data element
5003 ** in an associative array of type "x1".
5005 typedef struct s_x1node {
5006 const char *data; /* The data */
5007 struct s_x1node *next; /* Next entry with the same hash */
5008 struct s_x1node **from; /* Previous link */
5009 } x1node;
5011 /* There is only one instance of the array, which is the following */
5012 static struct s_x1 *x1a;
5014 /* Allocate a new associative array */
5015 void Strsafe_init(void){
5016 if( x1a ) return;
5017 x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
5018 if( x1a ){
5019 x1a->size = 1024;
5020 x1a->count = 0;
5021 x1a->tbl = (x1node*)calloc(1024, sizeof(x1node) + sizeof(x1node*));
5022 if( x1a->tbl==0 ){
5023 free(x1a);
5024 x1a = 0;
5025 }else{
5026 int i;
5027 x1a->ht = (x1node**)&(x1a->tbl[1024]);
5028 for(i=0; i<1024; i++) x1a->ht[i] = 0;
5032 /* Insert a new record into the array. Return TRUE if successful.
5033 ** Prior data with the same key is NOT overwritten */
5034 int Strsafe_insert(const char *data)
5036 x1node *np;
5037 unsigned h;
5038 unsigned ph;
5040 if( x1a==0 ) return 0;
5041 ph = strhash(data);
5042 h = ph & (x1a->size-1);
5043 np = x1a->ht[h];
5044 while( np ){
5045 if( strcmp(np->data,data)==0 ){
5046 /* An existing entry with the same key is found. */
5047 /* Fail because overwrite is not allows. */
5048 return 0;
5050 np = np->next;
5052 if( x1a->count>=x1a->size ){
5053 /* Need to make the hash table bigger */
5054 int i,arrSize;
5055 struct s_x1 array;
5056 array.size = arrSize = x1a->size*2;
5057 array.count = x1a->count;
5058 array.tbl = (x1node*)calloc(arrSize, sizeof(x1node) + sizeof(x1node*));
5059 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5060 array.ht = (x1node**)&(array.tbl[arrSize]);
5061 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5062 for(i=0; i<x1a->count; i++){
5063 x1node *oldnp, *newnp;
5064 oldnp = &(x1a->tbl[i]);
5065 h = strhash(oldnp->data) & (arrSize-1);
5066 newnp = &(array.tbl[i]);
5067 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5068 newnp->next = array.ht[h];
5069 newnp->data = oldnp->data;
5070 newnp->from = &(array.ht[h]);
5071 array.ht[h] = newnp;
5073 free(x1a->tbl);
5074 *x1a = array;
5076 /* Insert the new data */
5077 h = ph & (x1a->size-1);
5078 np = &(x1a->tbl[x1a->count++]);
5079 np->data = data;
5080 if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
5081 np->next = x1a->ht[h];
5082 x1a->ht[h] = np;
5083 np->from = &(x1a->ht[h]);
5084 return 1;
5087 /* Return a pointer to data assigned to the given key. Return NULL
5088 ** if no such key. */
5089 const char *Strsafe_find(const char *key)
5091 unsigned h;
5092 x1node *np;
5094 if( x1a==0 ) return 0;
5095 h = strhash(key) & (x1a->size-1);
5096 np = x1a->ht[h];
5097 while( np ){
5098 if( strcmp(np->data,key)==0 ) break;
5099 np = np->next;
5101 return np ? np->data : 0;
5104 /* Return a pointer to the (terminal or nonterminal) symbol "x".
5105 ** Create a new symbol if this is the first time "x" has been seen.
5107 struct symbol *Symbol_new(const char *x)
5109 struct symbol *sp;
5111 sp = Symbol_find(x);
5112 if( sp==0 ){
5113 sp = (struct symbol *)calloc(1, sizeof(struct symbol) );
5114 MemoryCheck(sp);
5115 sp->name = Strsafe(x);
5116 sp->type = ISUPPER(*x) ? TERMINAL : NONTERMINAL;
5117 sp->rule = 0;
5118 sp->fallback = 0;
5119 sp->prec = -1;
5120 sp->assoc = UNK;
5121 sp->firstset = 0;
5122 sp->lambda = LEMON_FALSE;
5123 sp->destructor = 0;
5124 sp->destLineno = 0;
5125 sp->datatype = 0;
5126 sp->useCnt = 0;
5127 Symbol_insert(sp,sp->name);
5129 sp->useCnt++;
5130 return sp;
5133 /* Compare two symbols for sorting purposes. Return negative,
5134 ** zero, or positive if a is less then, equal to, or greater
5135 ** than b.
5137 ** Symbols that begin with upper case letters (terminals or tokens)
5138 ** must sort before symbols that begin with lower case letters
5139 ** (non-terminals). And MULTITERMINAL symbols (created using the
5140 ** %token_class directive) must sort at the very end. Other than
5141 ** that, the order does not matter.
5143 ** We find experimentally that leaving the symbols in their original
5144 ** order (the order they appeared in the grammar file) gives the
5145 ** smallest parser tables in SQLite.
5147 int Symbolcmpp(const void *_a, const void *_b)
5149 const struct symbol *a = *(const struct symbol **) _a;
5150 const struct symbol *b = *(const struct symbol **) _b;
5151 int i1 = a->type==MULTITERMINAL ? 3 : a->name[0]>'Z' ? 2 : 1;
5152 int i2 = b->type==MULTITERMINAL ? 3 : b->name[0]>'Z' ? 2 : 1;
5153 return i1==i2 ? a->index - b->index : i1 - i2;
5156 /* There is one instance of the following structure for each
5157 ** associative array of type "x2".
5159 struct s_x2 {
5160 int size; /* The number of available slots. */
5161 /* Must be a power of 2 greater than or */
5162 /* equal to 1 */
5163 int count; /* Number of currently slots filled */
5164 struct s_x2node *tbl; /* The data stored here */
5165 struct s_x2node **ht; /* Hash table for lookups */
5168 /* There is one instance of this structure for every data element
5169 ** in an associative array of type "x2".
5171 typedef struct s_x2node {
5172 struct symbol *data; /* The data */
5173 const char *key; /* The key */
5174 struct s_x2node *next; /* Next entry with the same hash */
5175 struct s_x2node **from; /* Previous link */
5176 } x2node;
5178 /* There is only one instance of the array, which is the following */
5179 static struct s_x2 *x2a;
5181 /* Allocate a new associative array */
5182 void Symbol_init(void){
5183 if( x2a ) return;
5184 x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
5185 if( x2a ){
5186 x2a->size = 128;
5187 x2a->count = 0;
5188 x2a->tbl = (x2node*)calloc(128, sizeof(x2node) + sizeof(x2node*));
5189 if( x2a->tbl==0 ){
5190 free(x2a);
5191 x2a = 0;
5192 }else{
5193 int i;
5194 x2a->ht = (x2node**)&(x2a->tbl[128]);
5195 for(i=0; i<128; i++) x2a->ht[i] = 0;
5199 /* Insert a new record into the array. Return TRUE if successful.
5200 ** Prior data with the same key is NOT overwritten */
5201 int Symbol_insert(struct symbol *data, const char *key)
5203 x2node *np;
5204 unsigned h;
5205 unsigned ph;
5207 if( x2a==0 ) return 0;
5208 ph = strhash(key);
5209 h = ph & (x2a->size-1);
5210 np = x2a->ht[h];
5211 while( np ){
5212 if( strcmp(np->key,key)==0 ){
5213 /* An existing entry with the same key is found. */
5214 /* Fail because overwrite is not allows. */
5215 return 0;
5217 np = np->next;
5219 if( x2a->count>=x2a->size ){
5220 /* Need to make the hash table bigger */
5221 int i,arrSize;
5222 struct s_x2 array;
5223 array.size = arrSize = x2a->size*2;
5224 array.count = x2a->count;
5225 array.tbl = (x2node*)calloc(arrSize, sizeof(x2node) + sizeof(x2node*));
5226 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5227 array.ht = (x2node**)&(array.tbl[arrSize]);
5228 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5229 for(i=0; i<x2a->count; i++){
5230 x2node *oldnp, *newnp;
5231 oldnp = &(x2a->tbl[i]);
5232 h = strhash(oldnp->key) & (arrSize-1);
5233 newnp = &(array.tbl[i]);
5234 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5235 newnp->next = array.ht[h];
5236 newnp->key = oldnp->key;
5237 newnp->data = oldnp->data;
5238 newnp->from = &(array.ht[h]);
5239 array.ht[h] = newnp;
5241 free(x2a->tbl);
5242 *x2a = array;
5244 /* Insert the new data */
5245 h = ph & (x2a->size-1);
5246 np = &(x2a->tbl[x2a->count++]);
5247 np->key = key;
5248 np->data = data;
5249 if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
5250 np->next = x2a->ht[h];
5251 x2a->ht[h] = np;
5252 np->from = &(x2a->ht[h]);
5253 return 1;
5256 /* Return a pointer to data assigned to the given key. Return NULL
5257 ** if no such key. */
5258 struct symbol *Symbol_find(const char *key)
5260 unsigned h;
5261 x2node *np;
5263 if( x2a==0 ) return 0;
5264 h = strhash(key) & (x2a->size-1);
5265 np = x2a->ht[h];
5266 while( np ){
5267 if( strcmp(np->key,key)==0 ) break;
5268 np = np->next;
5270 return np ? np->data : 0;
5273 /* Return the n-th data. Return NULL if n is out of range. */
5274 struct symbol *Symbol_Nth(int n)
5276 struct symbol *data;
5277 if( x2a && n>0 && n<=x2a->count ){
5278 data = x2a->tbl[n-1].data;
5279 }else{
5280 data = 0;
5282 return data;
5285 /* Return the size of the array */
5286 int Symbol_count()
5288 return x2a ? x2a->count : 0;
5291 /* Return an array of pointers to all data in the table.
5292 ** The array is obtained from malloc. Return NULL if memory allocation
5293 ** problems, or if the array is empty. */
5294 struct symbol **Symbol_arrayof()
5296 struct symbol **array;
5297 int i,arrSize;
5298 if( x2a==0 ) return 0;
5299 arrSize = x2a->count;
5300 array = (struct symbol **)calloc(arrSize, sizeof(struct symbol *));
5301 if( array ){
5302 for(i=0; i<arrSize; i++) array[i] = x2a->tbl[i].data;
5304 return array;
5307 /* Compare two configurations */
5308 int Configcmp(const char *_a,const char *_b)
5310 const struct config *a = (struct config *) _a;
5311 const struct config *b = (struct config *) _b;
5312 int x;
5313 x = a->rp->index - b->rp->index;
5314 if( x==0 ) x = a->dot - b->dot;
5315 return x;
5318 /* Compare two states */
5319 PRIVATE int statecmp(struct config *a, struct config *b)
5321 int rc;
5322 for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){
5323 rc = a->rp->index - b->rp->index;
5324 if( rc==0 ) rc = a->dot - b->dot;
5326 if( rc==0 ){
5327 if( a ) rc = 1;
5328 if( b ) rc = -1;
5330 return rc;
5333 /* Hash a state */
5334 PRIVATE unsigned statehash(struct config *a)
5336 unsigned h=0;
5337 while( a ){
5338 h = h*571 + a->rp->index*37 + a->dot;
5339 a = a->bp;
5341 return h;
5344 /* Allocate a new state structure */
5345 struct state *State_new()
5347 struct state *newstate;
5348 newstate = (struct state *)calloc(1, sizeof(struct state) );
5349 MemoryCheck(newstate);
5350 return newstate;
5353 /* There is one instance of the following structure for each
5354 ** associative array of type "x3".
5356 struct s_x3 {
5357 int size; /* The number of available slots. */
5358 /* Must be a power of 2 greater than or */
5359 /* equal to 1 */
5360 int count; /* Number of currently slots filled */
5361 struct s_x3node *tbl; /* The data stored here */
5362 struct s_x3node **ht; /* Hash table for lookups */
5365 /* There is one instance of this structure for every data element
5366 ** in an associative array of type "x3".
5368 typedef struct s_x3node {
5369 struct state *data; /* The data */
5370 struct config *key; /* The key */
5371 struct s_x3node *next; /* Next entry with the same hash */
5372 struct s_x3node **from; /* Previous link */
5373 } x3node;
5375 /* There is only one instance of the array, which is the following */
5376 static struct s_x3 *x3a;
5378 /* Allocate a new associative array */
5379 void State_init(void){
5380 if( x3a ) return;
5381 x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
5382 if( x3a ){
5383 x3a->size = 128;
5384 x3a->count = 0;
5385 x3a->tbl = (x3node*)calloc(128, sizeof(x3node) + sizeof(x3node*));
5386 if( x3a->tbl==0 ){
5387 free(x3a);
5388 x3a = 0;
5389 }else{
5390 int i;
5391 x3a->ht = (x3node**)&(x3a->tbl[128]);
5392 for(i=0; i<128; i++) x3a->ht[i] = 0;
5396 /* Insert a new record into the array. Return TRUE if successful.
5397 ** Prior data with the same key is NOT overwritten */
5398 int State_insert(struct state *data, struct config *key)
5400 x3node *np;
5401 unsigned h;
5402 unsigned ph;
5404 if( x3a==0 ) return 0;
5405 ph = statehash(key);
5406 h = ph & (x3a->size-1);
5407 np = x3a->ht[h];
5408 while( np ){
5409 if( statecmp(np->key,key)==0 ){
5410 /* An existing entry with the same key is found. */
5411 /* Fail because overwrite is not allows. */
5412 return 0;
5414 np = np->next;
5416 if( x3a->count>=x3a->size ){
5417 /* Need to make the hash table bigger */
5418 int i,arrSize;
5419 struct s_x3 array;
5420 array.size = arrSize = x3a->size*2;
5421 array.count = x3a->count;
5422 array.tbl = (x3node*)calloc(arrSize, sizeof(x3node) + sizeof(x3node*));
5423 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5424 array.ht = (x3node**)&(array.tbl[arrSize]);
5425 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5426 for(i=0; i<x3a->count; i++){
5427 x3node *oldnp, *newnp;
5428 oldnp = &(x3a->tbl[i]);
5429 h = statehash(oldnp->key) & (arrSize-1);
5430 newnp = &(array.tbl[i]);
5431 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5432 newnp->next = array.ht[h];
5433 newnp->key = oldnp->key;
5434 newnp->data = oldnp->data;
5435 newnp->from = &(array.ht[h]);
5436 array.ht[h] = newnp;
5438 free(x3a->tbl);
5439 *x3a = array;
5441 /* Insert the new data */
5442 h = ph & (x3a->size-1);
5443 np = &(x3a->tbl[x3a->count++]);
5444 np->key = key;
5445 np->data = data;
5446 if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
5447 np->next = x3a->ht[h];
5448 x3a->ht[h] = np;
5449 np->from = &(x3a->ht[h]);
5450 return 1;
5453 /* Return a pointer to data assigned to the given key. Return NULL
5454 ** if no such key. */
5455 struct state *State_find(struct config *key)
5457 unsigned h;
5458 x3node *np;
5460 if( x3a==0 ) return 0;
5461 h = statehash(key) & (x3a->size-1);
5462 np = x3a->ht[h];
5463 while( np ){
5464 if( statecmp(np->key,key)==0 ) break;
5465 np = np->next;
5467 return np ? np->data : 0;
5470 /* Return an array of pointers to all data in the table.
5471 ** The array is obtained from malloc. Return NULL if memory allocation
5472 ** problems, or if the array is empty. */
5473 struct state **State_arrayof(void)
5475 struct state **array;
5476 int i,arrSize;
5477 if( x3a==0 ) return 0;
5478 arrSize = x3a->count;
5479 array = (struct state **)calloc(arrSize, sizeof(struct state *));
5480 if( array ){
5481 for(i=0; i<arrSize; i++) array[i] = x3a->tbl[i].data;
5483 return array;
5486 /* Hash a configuration */
5487 PRIVATE unsigned confighash(struct config *a)
5489 unsigned h=0;
5490 h = h*571 + a->rp->index*37 + a->dot;
5491 return h;
5494 /* There is one instance of the following structure for each
5495 ** associative array of type "x4".
5497 struct s_x4 {
5498 int size; /* The number of available slots. */
5499 /* Must be a power of 2 greater than or */
5500 /* equal to 1 */
5501 int count; /* Number of currently slots filled */
5502 struct s_x4node *tbl; /* The data stored here */
5503 struct s_x4node **ht; /* Hash table for lookups */
5506 /* There is one instance of this structure for every data element
5507 ** in an associative array of type "x4".
5509 typedef struct s_x4node {
5510 struct config *data; /* The data */
5511 struct s_x4node *next; /* Next entry with the same hash */
5512 struct s_x4node **from; /* Previous link */
5513 } x4node;
5515 /* There is only one instance of the array, which is the following */
5516 static struct s_x4 *x4a;
5518 /* Allocate a new associative array */
5519 void Configtable_init(void){
5520 if( x4a ) return;
5521 x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
5522 if( x4a ){
5523 x4a->size = 64;
5524 x4a->count = 0;
5525 x4a->tbl = (x4node*)calloc(64, sizeof(x4node) + sizeof(x4node*));
5526 if( x4a->tbl==0 ){
5527 free(x4a);
5528 x4a = 0;
5529 }else{
5530 int i;
5531 x4a->ht = (x4node**)&(x4a->tbl[64]);
5532 for(i=0; i<64; i++) x4a->ht[i] = 0;
5536 /* Insert a new record into the array. Return TRUE if successful.
5537 ** Prior data with the same key is NOT overwritten */
5538 int Configtable_insert(struct config *data)
5540 x4node *np;
5541 unsigned h;
5542 unsigned ph;
5544 if( x4a==0 ) return 0;
5545 ph = confighash(data);
5546 h = ph & (x4a->size-1);
5547 np = x4a->ht[h];
5548 while( np ){
5549 if( Configcmp((const char *) np->data,(const char *) data)==0 ){
5550 /* An existing entry with the same key is found. */
5551 /* Fail because overwrite is not allows. */
5552 return 0;
5554 np = np->next;
5556 if( x4a->count>=x4a->size ){
5557 /* Need to make the hash table bigger */
5558 int i,arrSize;
5559 struct s_x4 array;
5560 array.size = arrSize = x4a->size*2;
5561 array.count = x4a->count;
5562 array.tbl = (x4node*)calloc(arrSize, sizeof(x4node) + sizeof(x4node*));
5563 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5564 array.ht = (x4node**)&(array.tbl[arrSize]);
5565 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5566 for(i=0; i<x4a->count; i++){
5567 x4node *oldnp, *newnp;
5568 oldnp = &(x4a->tbl[i]);
5569 h = confighash(oldnp->data) & (arrSize-1);
5570 newnp = &(array.tbl[i]);
5571 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5572 newnp->next = array.ht[h];
5573 newnp->data = oldnp->data;
5574 newnp->from = &(array.ht[h]);
5575 array.ht[h] = newnp;
5577 free(x4a->tbl);
5578 *x4a = array;
5580 /* Insert the new data */
5581 h = ph & (x4a->size-1);
5582 np = &(x4a->tbl[x4a->count++]);
5583 np->data = data;
5584 if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
5585 np->next = x4a->ht[h];
5586 x4a->ht[h] = np;
5587 np->from = &(x4a->ht[h]);
5588 return 1;
5591 /* Return a pointer to data assigned to the given key. Return NULL
5592 ** if no such key. */
5593 struct config *Configtable_find(struct config *key)
5595 int h;
5596 x4node *np;
5598 if( x4a==0 ) return 0;
5599 h = confighash(key) & (x4a->size-1);
5600 np = x4a->ht[h];
5601 while( np ){
5602 if( Configcmp((const char *) np->data,(const char *) key)==0 ) break;
5603 np = np->next;
5605 return np ? np->data : 0;
5608 /* Remove all data from the table. Pass each data to the function "f"
5609 ** as it is removed. ("f" may be null to avoid this step.) */
5610 void Configtable_clear(int(*f)(struct config *))
5612 int i;
5613 if( x4a==0 || x4a->count==0 ) return;
5614 if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
5615 for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;
5616 x4a->count = 0;
5617 return;