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.
7 ** The author of this program disclaims copyright.
16 # if defined(_WIN32) || defined(WIN32)
21 /* #define PRIVATE static */
25 #define MAXRHS 5 /* Set low to exercise exception code */
31 extern void *malloc();
33 /******** From the file "action.h" *************************************/
34 struct action *Action_new();
35 struct action *Action_sort();
37 /********* From the file "assert.h" ************************************/
40 # define assert(X) if(!(X))myassert(__FILE__,__LINE__)
45 /********** From the file "build.h" ************************************/
46 void FindRulePrecedences();
50 void FindFollowSets();
53 /********* From the file "configlist.h" *********************************/
54 void Configlist_init(/* void */);
55 struct config *Configlist_add(/* struct rule *, int */);
56 struct config *Configlist_addbasis(/* struct rule *, int */);
57 void Configlist_closure(/* void */);
58 void Configlist_sort(/* void */);
59 void Configlist_sortbasis(/* void */);
60 struct config *Configlist_return(/* void */);
61 struct config *Configlist_basis(/* void */);
62 void Configlist_eat(/* struct config * */);
63 void Configlist_reset(/* void */);
65 /********* From the file "error.h" ***************************************/
66 void ErrorMsg(const char *, int,const char *, ...);
68 /****** From the file "option.h" ******************************************/
70 enum { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
71 OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR} type;
76 int OptInit(/* char**,struct s_options*,FILE* */);
77 int OptNArgs(/* void */);
78 char *OptArg(/* int */);
79 void OptErr(/* int */);
80 void OptPrint(/* void */);
82 /******** From the file "parse.h" *****************************************/
83 void Parse(/* struct lemon *lemp */);
85 /********* From the file "plink.h" ***************************************/
86 struct plink *Plink_new(/* void */);
87 void Plink_add(/* struct plink **, struct config * */);
88 void Plink_copy(/* struct plink **, struct plink * */);
89 void Plink_delete(/* struct plink * */);
91 /********** From the file "report.h" *************************************/
92 void Reprint(/* struct lemon * */);
93 void ReportOutput(/* struct lemon * */);
94 void ReportTable(/* struct lemon * */);
95 void ReportHeader(/* struct lemon * */);
96 void CompressTables(/* struct lemon * */);
97 void ResortStates(/* struct lemon * */);
99 /********** From the file "set.h" ****************************************/
100 void SetSize(/* int N */); /* All sets will be of size N */
101 char *SetNew(/* void */); /* A new set for element 0..N */
102 void SetFree(/* char* */); /* Deallocate a set */
104 int SetAdd(/* char*,int */); /* Add element to a set */
105 int SetUnion(/* char *A,char *B */); /* A <- A U B, thru element N */
107 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
109 /********** From the file "struct.h" *************************************/
111 ** Principal data structures for the LEMON parser generator.
114 typedef enum {B_FALSE=0, B_TRUE} Boolean;
116 /* Symbols (terminals and nonterminals) of the grammar are stored
117 ** in the following: */
119 char *name; /* Name of the symbol */
120 int index; /* Index number for this symbol */
125 } type; /* Symbols are all either TERMINALS or NTs */
126 struct rule *rule; /* Linked list of rules of this (if an NT) */
127 struct symbol *fallback; /* fallback token in case this token doesn't parse */
128 int prec; /* Precedence if defined (-1 otherwise) */
134 } assoc; /* Associativity if predecence is defined */
135 char *firstset; /* First-set for all rules of this symbol */
136 Boolean lambda; /* True if NT and can generate an empty string */
137 char *destructor; /* Code which executes whenever this symbol is
138 ** popped from the stack during error processing */
139 int destructorln; /* Line number of destructor code */
140 char *datatype; /* The data type of information held by this
141 ** object. Only used if type==NONTERMINAL */
142 int dtnum; /* The data type number. In the parser, the value
143 ** stack is a union. The .yy%d element of this
144 ** union is the correct data type for this object */
145 /* The following fields are used by MULTITERMINALs only */
146 int nsubsym; /* Number of constituent symbols in the MULTI */
147 struct symbol **subsym; /* Array of constituent symbols */
150 /* Each production rule in the grammar is stored in the following
153 struct symbol *lhs; /* Left-hand side of the rule */
154 char *lhsalias; /* Alias for the LHS (NULL if none) */
155 int ruleline; /* Line number for the rule */
156 int nrhs; /* Number of RHS symbols */
157 struct symbol **rhs; /* The RHS symbols */
158 char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
159 int line; /* Line number at which code begins */
160 char *code; /* The code executed when this rule is reduced */
161 struct symbol *precsym; /* Precedence symbol for this rule */
162 int index; /* An index number for this rule */
163 Boolean canReduce; /* True if this rule is ever reduced */
164 struct rule *nextlhs; /* Next rule with the same LHS */
165 struct rule *next; /* Next rule in the global list */
168 /* A configuration is a production rule of the grammar together with
169 ** a mark (dot) showing how much of that rule has been processed so far.
170 ** Configurations also contain a follow-set which is a list of terminal
171 ** symbols which are allowed to immediately follow the end of the rule.
172 ** Every configuration is recorded as an instance of the following: */
174 struct rule *rp; /* The rule upon which the configuration is based */
175 int dot; /* The parse point */
176 char *fws; /* Follow-set for this configuration only */
177 struct plink *fplp; /* Follow-set forward propagation links */
178 struct plink *bplp; /* Follow-set backwards propagation links */
179 struct state *stp; /* Pointer to state which contains this */
181 COMPLETE, /* The status is used during followset and */
182 INCOMPLETE /* shift computations */
184 struct config *next; /* Next configuration in the state */
185 struct config *bp; /* The next basis configuration */
188 /* Every shift or reduce operation is stored as one of the following */
190 struct symbol *sp; /* The look-ahead symbol */
196 CONFLICT, /* Was a reduce, but part of a conflict */
197 SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
198 RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
199 NOT_USED /* Deleted by compression */
202 struct state *stp; /* The new state, if a shift */
203 struct rule *rp; /* The rule, if a reduce */
205 struct action *next; /* Next action for this state */
206 struct action *collide; /* Next action with the same hash */
209 /* Each state of the generated parser's finite state machine
210 ** is encoded as an instance of the following structure. */
212 struct config *bp; /* The basis configurations for this state */
213 struct config *cfp; /* All configurations in this set */
214 int statenum; /* Sequencial number for this state */
215 struct action *ap; /* Array of actions for this state */
216 int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
217 int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
218 int iDflt; /* Default action */
220 #define NO_OFFSET (-2147483647)
222 /* A followset propagation link indicates that the contents of one
223 ** configuration followset should be propagated to another whenever
224 ** the first changes. */
226 struct config *cfp; /* The configuration to which linked */
227 struct plink *next; /* The next propagate link */
230 /* The state vector for the entire parser generator is recorded as
231 ** follows. (LEMON uses no global variables and makes little use of
232 ** static variables. Fields in the following structure can be thought
233 ** of as begin global variables in the program.) */
235 struct state **sorted; /* Table of states sorted by state number */
236 struct rule *rule; /* List of all rules */
237 int nstate; /* Number of states */
238 int nrule; /* Number of rules */
239 int nsymbol; /* Number of terminal and nonterminal symbols */
240 int nterminal; /* Number of terminal symbols */
241 struct symbol **symbols; /* Sorted array of pointers to symbols */
242 int errorcnt; /* Number of errors */
243 struct symbol *errsym; /* The error symbol */
244 struct symbol *wildcard; /* Token that matches anything */
245 char *name; /* Name of the generated parser */
246 char *arg; /* Declaration of the 3th argument to parser */
247 char *tokentype; /* Type of terminal symbols in the parser stack */
248 char *vartype; /* The default type of non-terminal symbols */
249 char *start; /* Name of the start symbol for the grammar */
250 char *stacksize; /* Size of the parser stack */
251 char *include; /* Code to put at the start of the C file */
252 int includeln; /* Line number for start of include code */
253 char *error; /* Code to execute when an error is seen */
254 int errorln; /* Line number for start of error code */
255 char *overflow; /* Code to execute on a stack overflow */
256 int overflowln; /* Line number for start of overflow code */
257 char *failure; /* Code to execute on parser failure */
258 int failureln; /* Line number for start of failure code */
259 char *accept; /* Code to execute when the parser excepts */
260 int acceptln; /* Line number for the start of accept code */
261 char *extracode; /* Code appended to the generated file */
262 int extracodeln; /* Line number for the start of the extra code */
263 char *tokendest; /* Code to execute to destroy token data */
264 int tokendestln; /* Line number for token destroyer code */
265 char *vardest; /* Code for the default non-terminal destructor */
266 int vardestln; /* Line number for default non-term destructor code*/
267 char *filename; /* Name of the input file */
268 char *outname; /* Name of the current output file */
269 char *tokenprefix; /* A prefix added to token names in the .h file */
270 int nconflict; /* Number of parsing conflicts */
271 int tablesize; /* Size of the parse tables */
272 int basisflag; /* Print only basis configurations */
273 int has_fallback; /* True if any %fallback is seen in the grammer */
274 char *argv0; /* Name of the program */
277 #define MemoryCheck(X) if((X)==0){ \
278 extern void memory_error(); \
282 /**************** From the file "table.h" *********************************/
284 ** All code in this file has been automatically generated
285 ** from a specification in the file
287 ** by the associative array code building program "aagen".
288 ** Do not edit this file! Instead, edit the specification
289 ** file, then rerun aagen.
292 ** Code for processing tables in the LEMON parser generator.
295 /* Routines for handling a strings */
299 void Strsafe_init(/* void */);
300 int Strsafe_insert(/* char * */);
301 char *Strsafe_find(/* char * */);
303 /* Routines for handling symbols of the grammar */
305 struct symbol *Symbol_new();
306 int Symbolcmpp(/* struct symbol **, struct symbol ** */);
307 void Symbol_init(/* void */);
308 int Symbol_insert(/* struct symbol *, char * */);
309 struct symbol *Symbol_find(/* char * */);
310 struct symbol *Symbol_Nth(/* int */);
311 int Symbol_count(/* */);
312 struct symbol **Symbol_arrayof(/* */);
314 /* Routines to manage the state table */
316 int Configcmp(/* struct config *, struct config * */);
317 struct state *State_new();
318 void State_init(/* void */);
319 int State_insert(/* struct state *, struct config * */);
320 struct state *State_find(/* struct config * */);
321 struct state **State_arrayof(/* */);
323 /* Routines used for efficiency in Configlist_add */
325 void Configtable_init(/* void */);
326 int Configtable_insert(/* struct config * */);
327 struct config *Configtable_find(/* struct config * */);
328 void Configtable_clear(/* int(*)(struct config *) */);
329 /****************** From the file "action.c" *******************************/
331 ** Routines processing parser actions in the LEMON parser generator.
334 /* Allocate a new parser action */
335 struct action *Action_new(){
336 static struct action *freelist = 0;
342 freelist = (struct action *)malloc( sizeof(struct action)*amt );
344 fprintf(stderr,"Unable to allocate memory for a new parser action.");
347 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
348 freelist[amt-1].next = 0;
351 freelist = freelist->next;
355 /* Compare two actions */
356 static int actioncmp(ap1,ap2)
361 rc = ap1->sp->index - ap2->sp->index;
362 if( rc==0 ) rc = (int)ap1->type - (int)ap2->type;
364 assert( ap1->type==REDUCE || ap1->type==RD_RESOLVED || ap1->type==CONFLICT);
365 assert( ap2->type==REDUCE || ap2->type==RD_RESOLVED || ap2->type==CONFLICT);
366 rc = ap1->x.rp->index - ap2->x.rp->index;
371 /* Sort parser actions */
372 struct action *Action_sort(ap)
375 ap = (struct action *)msort((char *)ap,(char **)&ap->next,actioncmp);
379 void Action_add(app,type,sp,arg)
392 new->x.stp = (struct state *)arg;
394 new->x.rp = (struct rule *)arg;
397 /********************** New code to implement the "acttab" module ***********/
399 ** This module implements routines use to construct the yy_action[] table.
403 ** The state of the yy_action table under construction is an instance of
404 ** the following structure
406 typedef struct acttab acttab;
408 int nAction; /* Number of used slots in aAction[] */
409 int nActionAlloc; /* Slots allocated for aAction[] */
411 int lookahead; /* Value of the lookahead token */
412 int action; /* Action to take on the given lookahead */
413 } *aAction, /* The yy_action[] table under construction */
414 *aLookahead; /* A single new transaction set */
415 int mnLookahead; /* Minimum aLookahead[].lookahead */
416 int mnAction; /* Action associated with mnLookahead */
417 int mxLookahead; /* Maximum aLookahead[].lookahead */
418 int nLookahead; /* Used slots in aLookahead[] */
419 int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
422 /* Return the number of entries in the yy_action table */
423 #define acttab_size(X) ((X)->nAction)
425 /* The value for the N-th entry in yy_action */
426 #define acttab_yyaction(X,N) ((X)->aAction[N].action)
428 /* The value for the N-th entry in yy_lookahead */
429 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
431 /* Free all memory associated with the given acttab */
432 void acttab_free(acttab *p){
434 free( p->aLookahead );
438 /* Allocate a new acttab structure */
439 acttab *acttab_alloc(void){
440 acttab *p = malloc( sizeof(*p) );
442 fprintf(stderr,"Unable to allocate memory for a new acttab.");
445 memset(p, 0, sizeof(*p));
449 /* Add a new action to the current transaction set
451 void acttab_action(acttab *p, int lookahead, int action){
452 if( p->nLookahead>=p->nLookaheadAlloc ){
453 p->nLookaheadAlloc += 25;
454 p->aLookahead = realloc( p->aLookahead,
455 sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
456 if( p->aLookahead==0 ){
457 fprintf(stderr,"malloc failed\n");
461 if( p->nLookahead==0 ){
462 p->mxLookahead = lookahead;
463 p->mnLookahead = lookahead;
464 p->mnAction = action;
466 if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
467 if( p->mnLookahead>lookahead ){
468 p->mnLookahead = lookahead;
469 p->mnAction = action;
472 p->aLookahead[p->nLookahead].lookahead = lookahead;
473 p->aLookahead[p->nLookahead].action = action;
478 ** Add the transaction set built up with prior calls to acttab_action()
479 ** into the current action table. Then reset the transaction set back
480 ** to an empty set in preparation for a new round of acttab_action() calls.
482 ** Return the offset into the action table of the new transaction.
484 int acttab_insert(acttab *p){
486 assert( p->nLookahead>0 );
488 /* Make sure we have enough space to hold the expanded action table
489 ** in the worst case. The worst case occurs if the transaction set
490 ** must be appended to the current action table
492 n = p->mxLookahead + 1;
493 if( p->nAction + n >= p->nActionAlloc ){
494 int oldAlloc = p->nActionAlloc;
495 p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
496 p->aAction = realloc( p->aAction,
497 sizeof(p->aAction[0])*p->nActionAlloc);
499 fprintf(stderr,"malloc failed\n");
502 for(i=oldAlloc; i<p->nActionAlloc; i++){
503 p->aAction[i].lookahead = -1;
504 p->aAction[i].action = -1;
508 /* Scan the existing action table looking for an offset where we can
509 ** insert the current transaction set. Fall out of the loop when that
510 ** offset is found. In the worst case, we fall out of the loop when
511 ** i reaches p->nAction, which means we append the new transaction set.
513 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
515 for(i=0; i<p->nAction+p->mnLookahead; i++){
516 if( p->aAction[i].lookahead<0 ){
517 for(j=0; j<p->nLookahead; j++){
518 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
520 if( p->aAction[k].lookahead>=0 ) break;
522 if( j<p->nLookahead ) continue;
523 for(j=0; j<p->nAction; j++){
524 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
527 break; /* Fits in empty slots */
529 }else if( p->aAction[i].lookahead==p->mnLookahead ){
530 if( p->aAction[i].action!=p->mnAction ) continue;
531 for(j=0; j<p->nLookahead; j++){
532 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
533 if( k<0 || k>=p->nAction ) break;
534 if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
535 if( p->aLookahead[j].action!=p->aAction[k].action ) break;
537 if( j<p->nLookahead ) continue;
539 for(j=0; j<p->nAction; j++){
540 if( p->aAction[j].lookahead<0 ) continue;
541 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
543 if( n==p->nLookahead ){
544 break; /* Same as a prior transaction set */
548 /* Insert transaction set at index i. */
549 for(j=0; j<p->nLookahead; j++){
550 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
551 p->aAction[k] = p->aLookahead[j];
552 if( k>=p->nAction ) p->nAction = k+1;
556 /* Return the offset that is added to the lookahead in order to get the
557 ** index into yy_action of the action */
558 return i - p->mnLookahead;
561 /********************** From the file "assert.c" ****************************/
563 ** A more efficient way of handling assertions.
565 void myassert(file,line)
569 fprintf(stderr,"Assertion failed on line %d of file \"%s\"\n",line,file);
572 /********************** From the file "build.c" *****************************/
574 ** Routines to construction the finite state machine for the LEMON
578 /* Find a precedence symbol of every rule in the grammar.
580 ** Those rules which have a precedence symbol coded in the input
581 ** grammar using the "[symbol]" construct will already have the
582 ** rp->precsym field filled. Other rules take as their precedence
583 ** symbol the first RHS symbol with a defined precedence. If there
584 ** are not RHS symbols with a defined precedence, the precedence
585 ** symbol field is left blank.
587 void FindRulePrecedences(xp)
591 for(rp=xp->rule; rp; rp=rp->next){
592 if( rp->precsym==0 ){
594 for(i=0; i<rp->nrhs && rp->precsym==0; i++){
595 struct symbol *sp = rp->rhs[i];
596 if( sp->type==MULTITERMINAL ){
597 for(j=0; j<sp->nsubsym; j++){
598 if( sp->subsym[j]->prec>=0 ){
599 rp->precsym = sp->subsym[j];
603 }else if( sp->prec>=0 ){
604 rp->precsym = rp->rhs[i];
612 /* Find all nonterminals which will generate the empty string.
613 ** Then go back and compute the first sets of every nonterminal.
614 ** The first set is the set of all terminal symbols which can begin
615 ** a string generated by that nonterminal.
617 void FindFirstSets(lemp)
624 for(i=0; i<lemp->nsymbol; i++){
625 lemp->symbols[i]->lambda = B_FALSE;
627 for(i=lemp->nterminal; i<lemp->nsymbol; i++){
628 lemp->symbols[i]->firstset = SetNew();
631 /* First compute all lambdas */
634 for(rp=lemp->rule; rp; rp=rp->next){
635 if( rp->lhs->lambda ) continue;
636 for(i=0; i<rp->nrhs; i++){
637 struct symbol *sp = rp->rhs[i];
638 if( sp->type!=TERMINAL || sp->lambda==B_FALSE ) break;
641 rp->lhs->lambda = B_TRUE;
647 /* Now compute all first sets */
649 struct symbol *s1, *s2;
651 for(rp=lemp->rule; rp; rp=rp->next){
653 for(i=0; i<rp->nrhs; i++){
655 if( s2->type==TERMINAL ){
656 progress += SetAdd(s1->firstset,s2->index);
658 }else if( s2->type==MULTITERMINAL ){
659 for(j=0; j<s2->nsubsym; j++){
660 progress += SetAdd(s1->firstset,s2->subsym[j]->index);
664 if( s1->lambda==B_FALSE ) break;
666 progress += SetUnion(s1->firstset,s2->firstset);
667 if( s2->lambda==B_FALSE ) break;
675 /* Compute all LR(0) states for the grammar. Links
676 ** are added to between some states so that the LR(1) follow sets
677 ** can be computed later.
679 PRIVATE struct state *getstate(/* struct lemon * */); /* forward reference */
680 void FindStates(lemp)
688 /* Find the start symbol */
690 sp = Symbol_find(lemp->start);
692 ErrorMsg(lemp->filename,0,
693 "The specified start symbol \"%s\" is not \
694 in a nonterminal of the grammar. \"%s\" will be used as the start \
695 symbol instead.",lemp->start,lemp->rule->lhs->name);
697 sp = lemp->rule->lhs;
700 sp = lemp->rule->lhs;
703 /* Make sure the start symbol doesn't occur on the right-hand side of
704 ** any rule. Report an error if it does. (YACC would generate a new
705 ** start symbol in this case.) */
706 for(rp=lemp->rule; rp; rp=rp->next){
708 for(i=0; i<rp->nrhs; i++){
709 if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
710 ErrorMsg(lemp->filename,0,
711 "The start symbol \"%s\" occurs on the \
712 right-hand side of a rule. This will result in a parser which \
713 does not work properly.",sp->name);
719 /* The basis configuration set for the first state
720 ** is all rules which have the start symbol as their
722 for(rp=sp->rule; rp; rp=rp->nextlhs){
723 struct config *newcfp;
724 newcfp = Configlist_addbasis(rp,0);
725 SetAdd(newcfp->fws,0);
728 /* Compute the first state. All other states will be
729 ** computed automatically during the computation of the first one.
730 ** The returned pointer to the first state is not used. */
731 (void)getstate(lemp);
735 /* Return a pointer to a state which is described by the configuration
736 ** list which has been built from calls to Configlist_add.
738 PRIVATE void buildshifts(/* struct lemon *, struct state * */); /* Forwd ref */
739 PRIVATE struct state *getstate(lemp)
742 struct config *cfp, *bp;
745 /* Extract the sorted basis of the new state. The basis was constructed
746 ** by prior calls to "Configlist_addbasis()". */
747 Configlist_sortbasis();
748 bp = Configlist_basis();
750 /* Get a state with the same basis */
751 stp = State_find(bp);
753 /* A state with the same basis already exists! Copy all the follow-set
754 ** propagation links from the state under construction into the
755 ** preexisting state, then return a pointer to the preexisting state */
756 struct config *x, *y;
757 for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
758 Plink_copy(&y->bplp,x->bplp);
759 Plink_delete(x->fplp);
760 x->fplp = x->bplp = 0;
762 cfp = Configlist_return();
765 /* This really is a new state. Construct all the details */
766 Configlist_closure(lemp); /* Compute the configuration closure */
767 Configlist_sort(); /* Sort the configuration closure */
768 cfp = Configlist_return(); /* Get a pointer to the config list */
769 stp = State_new(); /* A new state structure */
771 stp->bp = bp; /* Remember the configuration basis */
772 stp->cfp = cfp; /* Remember the configuration closure */
773 stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
774 stp->ap = 0; /* No actions, yet. */
775 State_insert(stp,stp->bp); /* Add to the state table */
776 buildshifts(lemp,stp); /* Recursively compute successor states */
782 ** Return true if two symbols are the same.
790 if( a->type!=MULTITERMINAL ) return 0;
791 if( b->type!=MULTITERMINAL ) return 0;
792 if( a->nsubsym!=b->nsubsym ) return 0;
793 for(i=0; i<a->nsubsym; i++){
794 if( a->subsym[i]!=b->subsym[i] ) return 0;
799 /* Construct all successor states to the given state. A "successor"
800 ** state is any state which can be reached by a shift action.
802 PRIVATE void buildshifts(lemp,stp)
804 struct state *stp; /* The state from which successors are computed */
806 struct config *cfp; /* For looping thru the config closure of "stp" */
807 struct config *bcfp; /* For the inner loop on config closure of "stp" */
808 struct config *new; /* */
809 struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
810 struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
811 struct state *newstp; /* A pointer to a successor state */
813 /* Each configuration becomes complete after it contibutes to a successor
814 ** state. Initially, all configurations are incomplete */
815 for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
817 /* Loop through all configurations of the state "stp" */
818 for(cfp=stp->cfp; cfp; cfp=cfp->next){
819 if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
820 if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
821 Configlist_reset(); /* Reset the new config set */
822 sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
824 /* For every configuration in the state "stp" which has the symbol "sp"
825 ** following its dot, add the same configuration to the basis set under
826 ** construction but with the dot shifted one symbol to the right. */
827 for(bcfp=cfp; bcfp; bcfp=bcfp->next){
828 if( bcfp->status==COMPLETE ) continue; /* Already used */
829 if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
830 bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
831 if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
832 bcfp->status = COMPLETE; /* Mark this config as used */
833 new = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
834 Plink_add(&new->bplp,bcfp);
837 /* Get a pointer to the state described by the basis configuration set
838 ** constructed in the preceding loop */
839 newstp = getstate(lemp);
841 /* The state "newstp" is reached from the state "stp" by a shift action
842 ** on the symbol "sp" */
843 if( sp->type==MULTITERMINAL ){
845 for(i=0; i<sp->nsubsym; i++){
846 Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
849 Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
855 ** Construct the propagation links
861 struct config *cfp, *other;
865 /* Housekeeping detail:
866 ** Add to every propagate link a pointer back to the state to
867 ** which the link is attached. */
868 for(i=0; i<lemp->nstate; i++){
869 stp = lemp->sorted[i];
870 for(cfp=stp->cfp; cfp; cfp=cfp->next){
875 /* Convert all backlinks into forward links. Only the forward
876 ** links are used in the follow-set computation. */
877 for(i=0; i<lemp->nstate; i++){
878 stp = lemp->sorted[i];
879 for(cfp=stp->cfp; cfp; cfp=cfp->next){
880 for(plp=cfp->bplp; plp; plp=plp->next){
882 Plink_add(&other->fplp,cfp);
888 /* Compute all followsets.
890 ** A followset is the set of all symbols which can come immediately
891 ** after a configuration.
893 void FindFollowSets(lemp)
902 for(i=0; i<lemp->nstate; i++){
903 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
904 cfp->status = INCOMPLETE;
910 for(i=0; i<lemp->nstate; i++){
911 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
912 if( cfp->status==COMPLETE ) continue;
913 for(plp=cfp->fplp; plp; plp=plp->next){
914 change = SetUnion(plp->cfp->fws,cfp->fws);
916 plp->cfp->status = INCOMPLETE;
920 cfp->status = COMPLETE;
926 static int resolve_conflict();
928 /* Compute the reduce actions, and resolve conflicts.
930 void FindActions(lemp)
939 /* Add all of the reduce actions
940 ** A reduce action is added for each element of the followset of
941 ** a configuration which has its dot at the extreme right.
943 for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
944 stp = lemp->sorted[i];
945 for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
946 if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
947 for(j=0; j<lemp->nterminal; j++){
948 if( SetFind(cfp->fws,j) ){
949 /* Add a reduce action to the state "stp" which will reduce by the
950 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
951 Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
958 /* Add the accepting token */
960 sp = Symbol_find(lemp->start);
961 if( sp==0 ) sp = lemp->rule->lhs;
963 sp = lemp->rule->lhs;
965 /* Add to the first state (which is always the starting state of the
966 ** finite state machine) an action to ACCEPT if the lookahead is the
967 ** start nonterminal. */
968 Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
970 /* Resolve conflicts */
971 for(i=0; i<lemp->nstate; i++){
972 struct action *ap, *nap;
974 stp = lemp->sorted[i];
976 stp->ap = Action_sort(stp->ap);
977 for(ap=stp->ap; ap && ap->next; ap=ap->next){
978 for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
979 /* The two actions "ap" and "nap" have the same lookahead.
980 ** Figure out which one should be used */
981 lemp->nconflict += resolve_conflict(ap,nap,lemp->errsym);
986 /* Report an error for each rule that can never be reduced. */
987 for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = B_FALSE;
988 for(i=0; i<lemp->nstate; i++){
990 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
991 if( ap->type==REDUCE ) ap->x.rp->canReduce = B_TRUE;
994 for(rp=lemp->rule; rp; rp=rp->next){
995 if( rp->canReduce ) continue;
996 ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
1001 /* Resolve a conflict between the two given actions. If the
1002 ** conflict can't be resolve, return non-zero.
1005 ** To resolve a conflict, first look to see if either action
1006 ** is on an error rule. In that case, take the action which
1007 ** is not associated with the error rule. If neither or both
1008 ** actions are associated with an error rule, then try to
1009 ** use precedence to resolve the conflict.
1011 ** If either action is a SHIFT, then it must be apx. This
1012 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1014 static int resolve_conflict(apx,apy,errsym)
1017 struct symbol *errsym; /* The error symbol (if defined. NULL otherwise) */
1019 struct symbol *spx, *spy;
1021 assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
1022 if( apx->type==SHIFT && apy->type==REDUCE ){
1024 spy = apy->x.rp->precsym;
1025 if( spy==0 || spx->prec<0 || spy->prec<0 ){
1026 /* Not enough precedence information. */
1027 apy->type = CONFLICT;
1029 }else if( spx->prec>spy->prec ){ /* Lower precedence wins */
1030 apy->type = RD_RESOLVED;
1031 }else if( spx->prec<spy->prec ){
1032 apx->type = SH_RESOLVED;
1033 }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1034 apy->type = RD_RESOLVED; /* associativity */
1035 }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
1036 apx->type = SH_RESOLVED;
1038 assert( spx->prec==spy->prec && spx->assoc==NONE );
1039 apy->type = CONFLICT;
1042 }else if( apx->type==REDUCE && apy->type==REDUCE ){
1043 spx = apx->x.rp->precsym;
1044 spy = apy->x.rp->precsym;
1045 if( spx==0 || spy==0 || spx->prec<0 ||
1046 spy->prec<0 || spx->prec==spy->prec ){
1047 apy->type = CONFLICT;
1049 }else if( spx->prec>spy->prec ){
1050 apy->type = RD_RESOLVED;
1051 }else if( spx->prec<spy->prec ){
1052 apx->type = RD_RESOLVED;
1056 apx->type==SH_RESOLVED ||
1057 apx->type==RD_RESOLVED ||
1058 apx->type==CONFLICT ||
1059 apy->type==SH_RESOLVED ||
1060 apy->type==RD_RESOLVED ||
1063 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1064 ** REDUCEs on the list. If we reach this point it must be because
1065 ** the parser conflict had already been resolved. */
1069 /********************* From the file "configlist.c" *************************/
1071 ** Routines to processing a configuration list and building a state
1072 ** in the LEMON parser generator.
1075 static struct config *freelist = 0; /* List of free configurations */
1076 static struct config *current = 0; /* Top of list of configurations */
1077 static struct config **currentend = 0; /* Last on list of configs */
1078 static struct config *basis = 0; /* Top of list of basis configs */
1079 static struct config **basisend = 0; /* End of list of basis configs */
1081 /* Return a pointer to a new configuration */
1082 PRIVATE struct config *newconfig(){
1087 freelist = (struct config *)malloc( sizeof(struct config)*amt );
1089 fprintf(stderr,"Unable to allocate memory for a new configuration.");
1092 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1093 freelist[amt-1].next = 0;
1096 freelist = freelist->next;
1100 /* The configuration "old" is no longer used */
1101 PRIVATE void deleteconfig(old)
1104 old->next = freelist;
1108 /* Initialized the configuration list builder */
1109 void Configlist_init(){
1111 currentend = ¤t;
1118 /* Initialized the configuration list builder */
1119 void Configlist_reset(){
1121 currentend = ¤t;
1124 Configtable_clear(0);
1128 /* Add another configuration to the configuration list */
1129 struct config *Configlist_add(rp,dot)
1130 struct rule *rp; /* The rule */
1131 int dot; /* Index into the RHS of the rule where the dot goes */
1133 struct config *cfp, model;
1135 assert( currentend!=0 );
1138 cfp = Configtable_find(&model);
1143 cfp->fws = SetNew();
1145 cfp->fplp = cfp->bplp = 0;
1149 currentend = &cfp->next;
1150 Configtable_insert(cfp);
1155 /* Add a basis configuration to the configuration list */
1156 struct config *Configlist_addbasis(rp,dot)
1160 struct config *cfp, model;
1162 assert( basisend!=0 );
1163 assert( currentend!=0 );
1166 cfp = Configtable_find(&model);
1171 cfp->fws = SetNew();
1173 cfp->fplp = cfp->bplp = 0;
1177 currentend = &cfp->next;
1179 basisend = &cfp->bp;
1180 Configtable_insert(cfp);
1185 /* Compute the closure of the configuration list */
1186 void Configlist_closure(lemp)
1189 struct config *cfp, *newcfp;
1190 struct rule *rp, *newrp;
1191 struct symbol *sp, *xsp;
1194 assert( currentend!=0 );
1195 for(cfp=current; cfp; cfp=cfp->next){
1198 if( dot>=rp->nrhs ) continue;
1200 if( sp->type==NONTERMINAL ){
1201 if( sp->rule==0 && sp!=lemp->errsym ){
1202 ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1206 for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1207 newcfp = Configlist_add(newrp,0);
1208 for(i=dot+1; i<rp->nrhs; i++){
1210 if( xsp->type==TERMINAL ){
1211 SetAdd(newcfp->fws,xsp->index);
1213 }else if( xsp->type==MULTITERMINAL ){
1215 for(k=0; k<xsp->nsubsym; k++){
1216 SetAdd(newcfp->fws, xsp->subsym[k]->index);
1220 SetUnion(newcfp->fws,xsp->firstset);
1221 if( xsp->lambda==B_FALSE ) break;
1224 if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1231 /* Sort the configuration list */
1232 void Configlist_sort(){
1233 current = (struct config *)msort((char *)current,(char **)&(current->next),Configcmp);
1238 /* Sort the basis configuration list */
1239 void Configlist_sortbasis(){
1240 basis = (struct config *)msort((char *)current,(char **)&(current->bp),Configcmp);
1245 /* Return a pointer to the head of the configuration list and
1246 ** reset the list */
1247 struct config *Configlist_return(){
1255 /* Return a pointer to the head of the configuration list and
1256 ** reset the list */
1257 struct config *Configlist_basis(){
1265 /* Free all elements of the given configuration list */
1266 void Configlist_eat(cfp)
1269 struct config *nextcfp;
1270 for(; cfp; cfp=nextcfp){
1271 nextcfp = cfp->next;
1272 assert( cfp->fplp==0 );
1273 assert( cfp->bplp==0 );
1274 if( cfp->fws ) SetFree(cfp->fws);
1279 /***************** From the file "error.c" *********************************/
1281 ** Code for printing error message.
1284 /* Find a good place to break "msg" so that its length is at least "min"
1285 ** but no more than "max". Make the point as close to max as possible.
1287 static int findbreak(msg,min,max)
1294 for(i=spot=min; i<=max; i++){
1296 if( c=='\t' ) msg[i] = ' ';
1297 if( c=='\n' ){ msg[i] = ' '; spot = i; break; }
1298 if( c==0 ){ spot = i; break; }
1299 if( c=='-' && i<max-1 ) spot = i+1;
1300 if( c==' ' ) spot = i;
1306 ** The error message is split across multiple lines if necessary. The
1307 ** splits occur at a space, if there is a space available near the end
1310 #define ERRMSGSIZE 10000 /* Hope this is big enough. No way to error check */
1311 #define LINEWIDTH 79 /* Max width of any output line */
1312 #define PREFIXLIMIT 30 /* Max width of the prefix on each line */
1313 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1314 char errmsg[ERRMSGSIZE];
1315 char prefix[PREFIXLIMIT+10];
1320 int end, restart, base;
1322 va_start(ap, format);
1323 /* Prepare a prefix to be prepended to every output line */
1325 sprintf(prefix,"%.*s:%d: ",PREFIXLIMIT-10,filename,lineno);
1327 sprintf(prefix,"%.*s: ",PREFIXLIMIT-10,filename);
1329 prefixsize = strlen(prefix);
1330 availablewidth = LINEWIDTH - prefixsize;
1332 /* Generate the error message */
1333 vsprintf(errmsg,format,ap);
1335 errmsgsize = strlen(errmsg);
1336 /* Remove trailing '\n's from the error message. */
1337 while( errmsgsize>0 && errmsg[errmsgsize-1]=='\n' ){
1338 errmsg[--errmsgsize] = 0;
1341 /* Print the error message */
1343 while( errmsg[base]!=0 ){
1344 end = restart = findbreak(&errmsg[base],0,availablewidth);
1346 while( errmsg[restart]==' ' ) restart++;
1347 fprintf(stdout,"%s%.*s\n",prefix,end,&errmsg[base]);
1351 /**************** From the file "main.c" ************************************/
1353 ** Main program file for the LEMON parser generator.
1356 /* Report an out-of-memory condition and abort. This function
1357 ** is used mostly by the "MemoryCheck" macro in struct.h
1359 void memory_error(){
1360 fprintf(stderr,"Out of memory. Aborting...\n");
1364 static int nDefine = 0; /* Number of -D options on the command line */
1365 static char **azDefine = 0; /* Name of the -D macros */
1367 /* This routine is called with the argument to each -D command-line option.
1368 ** Add the macro defined to the azDefine array.
1370 static void handle_D_option(char *z){
1373 azDefine = realloc(azDefine, sizeof(azDefine[0])*nDefine);
1375 fprintf(stderr,"out of memory\n");
1378 paz = &azDefine[nDefine-1];
1379 *paz = malloc( strlen(z)+1 );
1381 fprintf(stderr,"out of memory\n");
1385 for(z=*paz; *z && *z!='='; z++){}
1390 /* The main program. Parse the command line and do it... */
1395 static int version = 0;
1396 static int rpflag = 0;
1397 static int basisflag = 0;
1398 static int compress = 0;
1399 static int quiet = 0;
1400 static int statistics = 0;
1401 static int mhflag = 0;
1402 static struct s_options options[] = {
1403 {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1404 {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1405 {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
1406 {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1407 {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file"},
1408 {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1409 {OPT_FLAG, "s", (char*)&statistics,
1410 "Print parser stats to standard output."},
1411 {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1417 OptInit(argv,options,stderr);
1419 printf("Lemon version 1.0\n");
1422 if( OptNArgs()!=1 ){
1423 fprintf(stderr,"Exactly one filename argument is required.\n");
1426 memset(&lem, 0, sizeof(lem));
1429 /* Initialize the machine */
1433 lem.argv0 = argv[0];
1434 lem.filename = OptArg(0);
1435 lem.basisflag = basisflag;
1437 lem.errsym = Symbol_new("error");
1439 /* Parse the input file */
1441 if( lem.errorcnt ) exit(lem.errorcnt);
1443 fprintf(stderr,"Empty grammar.\n");
1447 /* Count and index the symbols of the grammar */
1448 lem.nsymbol = Symbol_count();
1449 Symbol_new("{default}");
1450 lem.symbols = Symbol_arrayof();
1451 for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
1452 qsort(lem.symbols,lem.nsymbol+1,sizeof(struct symbol*),
1453 (int(*)())Symbolcmpp);
1454 for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
1455 for(i=1; isupper(lem.symbols[i]->name[0]); i++);
1458 /* Generate a reprint of the grammar, if requested on the command line */
1462 /* Initialize the size for all follow and first sets */
1463 SetSize(lem.nterminal);
1465 /* Find the precedence for every production rule (that has one) */
1466 FindRulePrecedences(&lem);
1468 /* Compute the lambda-nonterminals and the first-sets for every
1470 FindFirstSets(&lem);
1472 /* Compute all LR(0) states. Also record follow-set propagation
1473 ** links so that the follow-set can be computed later */
1476 lem.sorted = State_arrayof();
1478 /* Tie up loose ends on the propagation links */
1481 /* Compute the follow set of every reducible configuration */
1482 FindFollowSets(&lem);
1484 /* Compute the action tables */
1487 /* Compress the action tables */
1488 if( compress==0 ) CompressTables(&lem);
1490 /* Reorder and renumber the states so that states with fewer choices
1491 ** occur at the end. */
1494 /* Generate a report of the parser generated. (the "y.output" file) */
1495 if( !quiet ) ReportOutput(&lem);
1497 /* Generate the source code for the parser */
1498 ReportTable(&lem, mhflag);
1500 /* Produce a header file for use by the scanner. (This step is
1501 ** omitted if the "-m" option is used because makeheaders will
1502 ** generate the file for us.) */
1503 if( !mhflag ) ReportHeader(&lem);
1506 printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
1507 lem.nterminal, lem.nsymbol - lem.nterminal, lem.nrule);
1508 printf(" %d states, %d parser table entries, %d conflicts\n",
1509 lem.nstate, lem.tablesize, lem.nconflict);
1511 if( lem.nconflict ){
1512 fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1514 exit(lem.errorcnt + lem.nconflict);
1515 return (lem.errorcnt + lem.nconflict);
1517 /******************** From the file "msort.c" *******************************/
1519 ** A generic merge-sort program.
1522 ** Let "ptr" be a pointer to some structure which is at the head of
1523 ** a null-terminated list. Then to sort the list call:
1525 ** ptr = msort(ptr,&(ptr->next),cmpfnc);
1527 ** In the above, "cmpfnc" is a pointer to a function which compares
1528 ** two instances of the structure and returns an integer, as in
1529 ** strcmp. The second argument is a pointer to the pointer to the
1530 ** second element of the linked list. This address is used to compute
1531 ** the offset to the "next" field within the structure. The offset to
1532 ** the "next" field must be constant for all structures in the list.
1534 ** The function returns a new pointer which is the head of the list
1542 ** Return a pointer to the next structure in the linked list.
1544 #define NEXT(A) (*(char**)(((unsigned long)A)+offset))
1548 ** a: A sorted, null-terminated linked list. (May be null).
1549 ** b: A sorted, null-terminated linked list. (May be null).
1550 ** cmp: A pointer to the comparison function.
1551 ** offset: Offset in the structure to the "next" field.
1554 ** A pointer to the head of a sorted list containing the elements
1558 ** The "next" pointers for elements in the lists a and b are
1561 static char *merge(a,b,cmp,offset)
1574 if( (*cmp)(a,b)<0 ){
1583 if( (*cmp)(a,b)<0 ){
1593 if( a ) NEXT(ptr) = a;
1601 ** list: Pointer to a singly-linked list of structures.
1602 ** next: Pointer to pointer to the second element of the list.
1603 ** cmp: A comparison function.
1606 ** A pointer to the head of a sorted list containing the elements
1607 ** orginally in list.
1610 ** The "next" pointers for elements in list are changed.
1613 char *msort(list,next,cmp)
1618 unsigned long offset;
1620 char *set[LISTSIZE];
1622 offset = (unsigned long)next - (unsigned long)list;
1623 for(i=0; i<LISTSIZE; i++) set[i] = 0;
1628 for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1629 ep = merge(ep,set[i],cmp,offset);
1635 for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(ep,set[i],cmp,offset);
1638 /************************ From the file "option.c" **************************/
1640 static struct s_options *op;
1641 static FILE *errstream;
1643 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1646 ** Print the command line with a carrot pointing to the k-th character
1647 ** of the n-th field.
1649 static void errline(n,k,err)
1655 if( argv[0] ) fprintf(err,"%s",argv[0]);
1656 spcnt = strlen(argv[0]) + 1;
1657 for(i=1; i<n && argv[i]; i++){
1658 fprintf(err," %s",argv[i]);
1659 spcnt += strlen(argv[i])+1;
1662 for(; argv[i]; i++) fprintf(err," %s",argv[i]);
1664 fprintf(err,"\n%*s^-- here\n",spcnt,"");
1666 fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1671 ** Return the index of the N-th non-switch argument. Return -1
1672 ** if N is out of range.
1674 static int argindex(n)
1679 if( argv!=0 && *argv!=0 ){
1680 for(i=1; argv[i]; i++){
1681 if( dashdash || !ISOPT(argv[i]) ){
1682 if( n==0 ) return i;
1685 if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1691 static char emsg[] = "Command line syntax error: ";
1694 ** Process a flag command line argument.
1696 static int handleflags(i,err)
1703 for(j=0; op[j].label; j++){
1704 if( strncmp(&argv[i][1],op[j].label,strlen(op[j].label))==0 ) break;
1706 v = argv[i][0]=='-' ? 1 : 0;
1707 if( op[j].label==0 ){
1709 fprintf(err,"%sundefined option.\n",emsg);
1713 }else if( op[j].type==OPT_FLAG ){
1714 *((int*)op[j].arg) = v;
1715 }else if( op[j].type==OPT_FFLAG ){
1716 (*(void(*)())(op[j].arg))(v);
1717 }else if( op[j].type==OPT_FSTR ){
1718 (*(void(*)())(op[j].arg))(&argv[i][2]);
1721 fprintf(err,"%smissing argument on switch.\n",emsg);
1730 ** Process a command line switch which has an argument.
1732 static int handleswitch(i,err)
1742 cp = strchr(argv[i],'=');
1745 for(j=0; op[j].label; j++){
1746 if( strcmp(argv[i],op[j].label)==0 ) break;
1749 if( op[j].label==0 ){
1751 fprintf(err,"%sundefined option.\n",emsg);
1757 switch( op[j].type ){
1761 fprintf(err,"%soption requires an argument.\n",emsg);
1768 dv = strtod(cp,&end);
1771 fprintf(err,"%sillegal character in floating-point argument.\n",emsg);
1772 errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1779 lv = strtol(cp,&end,0);
1782 fprintf(err,"%sillegal character in integer argument.\n",emsg);
1783 errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1793 switch( op[j].type ){
1798 *(double*)(op[j].arg) = dv;
1801 (*(void(*)())(op[j].arg))(dv);
1804 *(int*)(op[j].arg) = lv;
1807 (*(void(*)())(op[j].arg))((int)lv);
1810 *(char**)(op[j].arg) = sv;
1813 (*(void(*)())(op[j].arg))(sv);
1820 int OptInit(a,o,err)
1822 struct s_options *o;
1829 if( argv && *argv && op ){
1831 for(i=1; argv[i]; i++){
1832 if( argv[i][0]=='+' || argv[i][0]=='-' ){
1833 errcnt += handleflags(i,err);
1834 }else if( strchr(argv[i],'=') ){
1835 errcnt += handleswitch(i,err);
1840 fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
1851 if( argv!=0 && argv[0]!=0 ){
1852 for(i=1; argv[i]; i++){
1853 if( dashdash || !ISOPT(argv[i]) ) cnt++;
1854 if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1865 return i>=0 ? argv[i] : 0;
1873 if( i>=0 ) errline(i,0,errstream);
1880 for(i=0; op[i].label; i++){
1881 len = strlen(op[i].label) + 1;
1882 switch( op[i].type ){
1888 len += 9; /* length of "<integer>" */
1892 len += 6; /* length of "<real>" */
1896 len += 8; /* length of "<string>" */
1899 if( len>max ) max = len;
1901 for(i=0; op[i].label; i++){
1902 switch( op[i].type ){
1905 fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
1909 fprintf(errstream," %s=<integer>%*s %s\n",op[i].label,
1910 (int)(max-strlen(op[i].label)-9),"",op[i].message);
1914 fprintf(errstream," %s=<real>%*s %s\n",op[i].label,
1915 (int)(max-strlen(op[i].label)-6),"",op[i].message);
1919 fprintf(errstream," %s=<string>%*s %s\n",op[i].label,
1920 (int)(max-strlen(op[i].label)-8),"",op[i].message);
1925 /*********************** From the file "parse.c" ****************************/
1927 ** Input file parser for the LEMON parser generator.
1930 /* The state of the parser */
1932 char *filename; /* Name of the input file */
1933 int tokenlineno; /* Linenumber at which current token starts */
1934 int errorcnt; /* Number of errors so far */
1935 char *tokenstart; /* Text of current token */
1936 struct lemon *gp; /* Global state vector */
1939 WAITING_FOR_DECL_OR_RULE,
1940 WAITING_FOR_DECL_KEYWORD,
1941 WAITING_FOR_DECL_ARG,
1942 WAITING_FOR_PRECEDENCE_SYMBOL,
1952 RESYNC_AFTER_RULE_ERROR,
1953 RESYNC_AFTER_DECL_ERROR,
1954 WAITING_FOR_DESTRUCTOR_SYMBOL,
1955 WAITING_FOR_DATATYPE_SYMBOL,
1956 WAITING_FOR_FALLBACK_ID,
1957 WAITING_FOR_WILDCARD_ID
1958 } state; /* The state of the parser */
1959 struct symbol *fallback; /* The fallback token */
1960 struct symbol *lhs; /* Left-hand side of current rule */
1961 char *lhsalias; /* Alias for the LHS */
1962 int nrhs; /* Number of right-hand side symbols seen */
1963 struct symbol *rhs[MAXRHS]; /* RHS symbols */
1964 char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
1965 struct rule *prevrule; /* Previous rule parsed */
1966 char *declkeyword; /* Keyword of a declaration */
1967 char **declargslot; /* Where the declaration argument should be put */
1968 int *decllnslot; /* Where the declaration linenumber is put */
1969 enum e_assoc declassoc; /* Assign this association to decl arguments */
1970 int preccounter; /* Assign this precedence to decl arguments */
1971 struct rule *firstrule; /* Pointer to first rule in the grammar */
1972 struct rule *lastrule; /* Pointer to the most recently parsed rule */
1975 /* Parse a single token */
1976 static void parseonetoken(psp)
1980 x = Strsafe(psp->tokenstart); /* Save the token permanently */
1982 printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
1985 switch( psp->state ){
1988 psp->preccounter = 0;
1989 psp->firstrule = psp->lastrule = 0;
1991 /* Fall thru to next case */
1992 case WAITING_FOR_DECL_OR_RULE:
1994 psp->state = WAITING_FOR_DECL_KEYWORD;
1995 }else if( islower(x[0]) ){
1996 psp->lhs = Symbol_new(x);
1999 psp->state = WAITING_FOR_ARROW;
2000 }else if( x[0]=='{' ){
2001 if( psp->prevrule==0 ){
2002 ErrorMsg(psp->filename,psp->tokenlineno,
2003 "There is not prior rule opon which to attach the code \
2004 fragment which begins on this line.");
2006 }else if( psp->prevrule->code!=0 ){
2007 ErrorMsg(psp->filename,psp->tokenlineno,
2008 "Code fragment beginning on this line is not the first \
2009 to follow the previous rule.");
2012 psp->prevrule->line = psp->tokenlineno;
2013 psp->prevrule->code = &x[1];
2015 }else if( x[0]=='[' ){
2016 psp->state = PRECEDENCE_MARK_1;
2018 ErrorMsg(psp->filename,psp->tokenlineno,
2019 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2024 case PRECEDENCE_MARK_1:
2025 if( !isupper(x[0]) ){
2026 ErrorMsg(psp->filename,psp->tokenlineno,
2027 "The precedence symbol must be a terminal.");
2029 }else if( psp->prevrule==0 ){
2030 ErrorMsg(psp->filename,psp->tokenlineno,
2031 "There is no prior rule to assign precedence \"[%s]\".",x);
2033 }else if( psp->prevrule->precsym!=0 ){
2034 ErrorMsg(psp->filename,psp->tokenlineno,
2035 "Precedence mark on this line is not the first \
2036 to follow the previous rule.");
2039 psp->prevrule->precsym = Symbol_new(x);
2041 psp->state = PRECEDENCE_MARK_2;
2043 case PRECEDENCE_MARK_2:
2045 ErrorMsg(psp->filename,psp->tokenlineno,
2046 "Missing \"]\" on precedence mark.");
2049 psp->state = WAITING_FOR_DECL_OR_RULE;
2051 case WAITING_FOR_ARROW:
2052 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2053 psp->state = IN_RHS;
2054 }else if( x[0]=='(' ){
2055 psp->state = LHS_ALIAS_1;
2057 ErrorMsg(psp->filename,psp->tokenlineno,
2058 "Expected to see a \":\" following the LHS symbol \"%s\".",
2061 psp->state = RESYNC_AFTER_RULE_ERROR;
2065 if( isalpha(x[0]) ){
2067 psp->state = LHS_ALIAS_2;
2069 ErrorMsg(psp->filename,psp->tokenlineno,
2070 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2073 psp->state = RESYNC_AFTER_RULE_ERROR;
2078 psp->state = LHS_ALIAS_3;
2080 ErrorMsg(psp->filename,psp->tokenlineno,
2081 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2083 psp->state = RESYNC_AFTER_RULE_ERROR;
2087 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2088 psp->state = IN_RHS;
2090 ErrorMsg(psp->filename,psp->tokenlineno,
2091 "Missing \"->\" following: \"%s(%s)\".",
2092 psp->lhs->name,psp->lhsalias);
2094 psp->state = RESYNC_AFTER_RULE_ERROR;
2100 rp = (struct rule *)malloc( sizeof(struct rule) +
2101 sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs );
2103 ErrorMsg(psp->filename,psp->tokenlineno,
2104 "Can't allocate enough memory for this rule.");
2109 rp->ruleline = psp->tokenlineno;
2110 rp->rhs = (struct symbol**)&rp[1];
2111 rp->rhsalias = (char**)&(rp->rhs[psp->nrhs]);
2112 for(i=0; i<psp->nrhs; i++){
2113 rp->rhs[i] = psp->rhs[i];
2114 rp->rhsalias[i] = psp->alias[i];
2117 rp->lhsalias = psp->lhsalias;
2118 rp->nrhs = psp->nrhs;
2121 rp->index = psp->gp->nrule++;
2122 rp->nextlhs = rp->lhs->rule;
2125 if( psp->firstrule==0 ){
2126 psp->firstrule = psp->lastrule = rp;
2128 psp->lastrule->next = rp;
2133 psp->state = WAITING_FOR_DECL_OR_RULE;
2134 }else if( isalpha(x[0]) ){
2135 if( psp->nrhs>=MAXRHS ){
2136 ErrorMsg(psp->filename,psp->tokenlineno,
2137 "Too many symbols on RHS or rule beginning at \"%s\".",
2140 psp->state = RESYNC_AFTER_RULE_ERROR;
2142 psp->rhs[psp->nrhs] = Symbol_new(x);
2143 psp->alias[psp->nrhs] = 0;
2146 }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 ){
2147 struct symbol *msp = psp->rhs[psp->nrhs-1];
2148 if( msp->type!=MULTITERMINAL ){
2149 struct symbol *origsp = msp;
2150 msp = malloc(sizeof(*msp));
2151 memset(msp, 0, sizeof(*msp));
2152 msp->type = MULTITERMINAL;
2154 msp->subsym = malloc(sizeof(struct symbol*));
2155 msp->subsym[0] = origsp;
2156 msp->name = origsp->name;
2157 psp->rhs[psp->nrhs-1] = msp;
2160 msp->subsym = realloc(msp->subsym, sizeof(struct symbol*)*msp->nsubsym);
2161 msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
2162 if( islower(x[1]) || islower(msp->subsym[0]->name[0]) ){
2163 ErrorMsg(psp->filename,psp->tokenlineno,
2164 "Cannot form a compound containing a non-terminal");
2167 }else if( x[0]=='(' && psp->nrhs>0 ){
2168 psp->state = RHS_ALIAS_1;
2170 ErrorMsg(psp->filename,psp->tokenlineno,
2171 "Illegal character on RHS of rule: \"%s\".",x);
2173 psp->state = RESYNC_AFTER_RULE_ERROR;
2177 if( isalpha(x[0]) ){
2178 psp->alias[psp->nrhs-1] = x;
2179 psp->state = RHS_ALIAS_2;
2181 ErrorMsg(psp->filename,psp->tokenlineno,
2182 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2183 x,psp->rhs[psp->nrhs-1]->name);
2185 psp->state = RESYNC_AFTER_RULE_ERROR;
2190 psp->state = IN_RHS;
2192 ErrorMsg(psp->filename,psp->tokenlineno,
2193 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2195 psp->state = RESYNC_AFTER_RULE_ERROR;
2198 case WAITING_FOR_DECL_KEYWORD:
2199 if( isalpha(x[0]) ){
2200 psp->declkeyword = x;
2201 psp->declargslot = 0;
2202 psp->decllnslot = 0;
2203 psp->state = WAITING_FOR_DECL_ARG;
2204 if( strcmp(x,"name")==0 ){
2205 psp->declargslot = &(psp->gp->name);
2206 }else if( strcmp(x,"include")==0 ){
2207 psp->declargslot = &(psp->gp->include);
2208 psp->decllnslot = &psp->gp->includeln;
2209 }else if( strcmp(x,"code")==0 ){
2210 psp->declargslot = &(psp->gp->extracode);
2211 psp->decllnslot = &psp->gp->extracodeln;
2212 }else if( strcmp(x,"token_destructor")==0 ){
2213 psp->declargslot = &psp->gp->tokendest;
2214 psp->decllnslot = &psp->gp->tokendestln;
2215 }else if( strcmp(x,"default_destructor")==0 ){
2216 psp->declargslot = &psp->gp->vardest;
2217 psp->decllnslot = &psp->gp->vardestln;
2218 }else if( strcmp(x,"token_prefix")==0 ){
2219 psp->declargslot = &psp->gp->tokenprefix;
2220 }else if( strcmp(x,"syntax_error")==0 ){
2221 psp->declargslot = &(psp->gp->error);
2222 psp->decllnslot = &psp->gp->errorln;
2223 }else if( strcmp(x,"parse_accept")==0 ){
2224 psp->declargslot = &(psp->gp->accept);
2225 psp->decllnslot = &psp->gp->acceptln;
2226 }else if( strcmp(x,"parse_failure")==0 ){
2227 psp->declargslot = &(psp->gp->failure);
2228 psp->decllnslot = &psp->gp->failureln;
2229 }else if( strcmp(x,"stack_overflow")==0 ){
2230 psp->declargslot = &(psp->gp->overflow);
2231 psp->decllnslot = &psp->gp->overflowln;
2232 }else if( strcmp(x,"extra_argument")==0 ){
2233 psp->declargslot = &(psp->gp->arg);
2234 }else if( strcmp(x,"token_type")==0 ){
2235 psp->declargslot = &(psp->gp->tokentype);
2236 }else if( strcmp(x,"default_type")==0 ){
2237 psp->declargslot = &(psp->gp->vartype);
2238 }else if( strcmp(x,"stack_size")==0 ){
2239 psp->declargslot = &(psp->gp->stacksize);
2240 }else if( strcmp(x,"start_symbol")==0 ){
2241 psp->declargslot = &(psp->gp->start);
2242 }else if( strcmp(x,"left")==0 ){
2244 psp->declassoc = LEFT;
2245 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2246 }else if( strcmp(x,"right")==0 ){
2248 psp->declassoc = RIGHT;
2249 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2250 }else if( strcmp(x,"nonassoc")==0 ){
2252 psp->declassoc = NONE;
2253 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2254 }else if( strcmp(x,"destructor")==0 ){
2255 psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2256 }else if( strcmp(x,"type")==0 ){
2257 psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2258 }else if( strcmp(x,"fallback")==0 ){
2260 psp->state = WAITING_FOR_FALLBACK_ID;
2261 }else if( strcmp(x,"wildcard")==0 ){
2262 psp->state = WAITING_FOR_WILDCARD_ID;
2264 ErrorMsg(psp->filename,psp->tokenlineno,
2265 "Unknown declaration keyword: \"%%%s\".",x);
2267 psp->state = RESYNC_AFTER_DECL_ERROR;
2270 ErrorMsg(psp->filename,psp->tokenlineno,
2271 "Illegal declaration keyword: \"%s\".",x);
2273 psp->state = RESYNC_AFTER_DECL_ERROR;
2276 case WAITING_FOR_DESTRUCTOR_SYMBOL:
2277 if( !isalpha(x[0]) ){
2278 ErrorMsg(psp->filename,psp->tokenlineno,
2279 "Symbol name missing after %destructor keyword");
2281 psp->state = RESYNC_AFTER_DECL_ERROR;
2283 struct symbol *sp = Symbol_new(x);
2284 psp->declargslot = &sp->destructor;
2285 psp->decllnslot = &sp->destructorln;
2286 psp->state = WAITING_FOR_DECL_ARG;
2289 case WAITING_FOR_DATATYPE_SYMBOL:
2290 if( !isalpha(x[0]) ){
2291 ErrorMsg(psp->filename,psp->tokenlineno,
2292 "Symbol name missing after %destructor keyword");
2294 psp->state = RESYNC_AFTER_DECL_ERROR;
2296 struct symbol *sp = Symbol_new(x);
2297 psp->declargslot = &sp->datatype;
2298 psp->decllnslot = 0;
2299 psp->state = WAITING_FOR_DECL_ARG;
2302 case WAITING_FOR_PRECEDENCE_SYMBOL:
2304 psp->state = WAITING_FOR_DECL_OR_RULE;
2305 }else if( isupper(x[0]) ){
2309 ErrorMsg(psp->filename,psp->tokenlineno,
2310 "Symbol \"%s\" has already be given a precedence.",x);
2313 sp->prec = psp->preccounter;
2314 sp->assoc = psp->declassoc;
2317 ErrorMsg(psp->filename,psp->tokenlineno,
2318 "Can't assign a precedence to \"%s\".",x);
2322 case WAITING_FOR_DECL_ARG:
2323 if( (x[0]=='{' || x[0]=='\"' || isalnum(x[0])) ){
2324 if( *(psp->declargslot)!=0 ){
2325 ErrorMsg(psp->filename,psp->tokenlineno,
2326 "The argument \"%s\" to declaration \"%%%s\" is not the first.",
2327 x[0]=='\"' ? &x[1] : x,psp->declkeyword);
2329 psp->state = RESYNC_AFTER_DECL_ERROR;
2331 *(psp->declargslot) = (x[0]=='\"' || x[0]=='{') ? &x[1] : x;
2332 if( psp->decllnslot ) *psp->decllnslot = psp->tokenlineno;
2333 psp->state = WAITING_FOR_DECL_OR_RULE;
2336 ErrorMsg(psp->filename,psp->tokenlineno,
2337 "Illegal argument to %%%s: %s",psp->declkeyword,x);
2339 psp->state = RESYNC_AFTER_DECL_ERROR;
2342 case WAITING_FOR_FALLBACK_ID:
2344 psp->state = WAITING_FOR_DECL_OR_RULE;
2345 }else if( !isupper(x[0]) ){
2346 ErrorMsg(psp->filename, psp->tokenlineno,
2347 "%%fallback argument \"%s\" should be a token", x);
2350 struct symbol *sp = Symbol_new(x);
2351 if( psp->fallback==0 ){
2353 }else if( sp->fallback ){
2354 ErrorMsg(psp->filename, psp->tokenlineno,
2355 "More than one fallback assigned to token %s", x);
2358 sp->fallback = psp->fallback;
2359 psp->gp->has_fallback = 1;
2363 case WAITING_FOR_WILDCARD_ID:
2365 psp->state = WAITING_FOR_DECL_OR_RULE;
2366 }else if( !isupper(x[0]) ){
2367 ErrorMsg(psp->filename, psp->tokenlineno,
2368 "%%wildcard argument \"%s\" should be a token", x);
2371 struct symbol *sp = Symbol_new(x);
2372 if( psp->gp->wildcard==0 ){
2373 psp->gp->wildcard = sp;
2375 ErrorMsg(psp->filename, psp->tokenlineno,
2376 "Extra wildcard to token: %s", x);
2381 case RESYNC_AFTER_RULE_ERROR:
2382 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2384 case RESYNC_AFTER_DECL_ERROR:
2385 if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2386 if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2391 /* Run the proprocessor over the input file text. The global variables
2392 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2393 ** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2394 ** comments them out. Text in between is also commented out as appropriate.
2396 static void preprocess_input(char *z){
2402 for(i=0; z[i]; i++){
2403 if( z[i]=='\n' ) lineno++;
2404 if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue;
2405 if( strncmp(&z[i],"%endif",6)==0 && isspace(z[i+6]) ){
2409 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2412 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2413 }else if( (strncmp(&z[i],"%ifdef",6)==0 && isspace(z[i+6]))
2414 || (strncmp(&z[i],"%ifndef",7)==0 && isspace(z[i+7])) ){
2418 for(j=i+7; isspace(z[j]); j++){}
2419 for(n=0; z[j+n] && !isspace(z[j+n]); n++){}
2421 for(k=0; k<nDefine; k++){
2422 if( strncmp(azDefine[k],&z[j],n)==0 && strlen(azDefine[k])==n ){
2427 if( z[i+3]=='n' ) exclude = !exclude;
2430 start_lineno = lineno;
2433 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2437 fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno);
2442 /* In spite of its name, this function is really a scanner. It read
2443 ** in the entire input file (all at once) then tokenizes it. Each
2444 ** token is passed to the function "parseonetoken" which builds all
2445 ** the appropriate data structures in the global state vector "gp".
2460 ps.filename = gp->filename;
2462 ps.state = INITIALIZE;
2464 /* Begin by reading the input file */
2465 fp = fopen(ps.filename,"rb");
2467 ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2472 filesize = ftell(fp);
2474 filebuf = (char *)malloc( filesize+1 );
2476 ErrorMsg(ps.filename,0,"Can't allocate %d of memory to hold this file.",
2481 if( fread(filebuf,1,filesize,fp)!=filesize ){
2482 ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2489 filebuf[filesize] = 0;
2491 /* Make an initial pass through the file to handle %ifdef and %ifndef */
2492 preprocess_input(filebuf);
2494 /* Now scan the text of the input file */
2496 for(cp=filebuf; (c= *cp)!=0; ){
2497 if( c=='\n' ) lineno++; /* Keep track of the line number */
2498 if( isspace(c) ){ cp++; continue; } /* Skip all white space */
2499 if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */
2501 while( (c= *cp)!=0 && c!='\n' ) cp++;
2504 if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */
2506 while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
2507 if( c=='\n' ) lineno++;
2513 ps.tokenstart = cp; /* Mark the beginning of the token */
2514 ps.tokenlineno = lineno; /* Linenumber on which token begins */
2515 if( c=='\"' ){ /* String literals */
2517 while( (c= *cp)!=0 && c!='\"' ){
2518 if( c=='\n' ) lineno++;
2522 ErrorMsg(ps.filename,startline,
2523 "String starting on this line is not terminated before the end of the file.");
2529 }else if( c=='{' ){ /* A block of C code */
2532 for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
2533 if( c=='\n' ) lineno++;
2534 else if( c=='{' ) level++;
2535 else if( c=='}' ) level--;
2536 else if( c=='/' && cp[1]=='*' ){ /* Skip comments */
2540 while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
2541 if( c=='\n' ) lineno++;
2545 }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */
2547 while( (c= *cp)!=0 && c!='\n' ) cp++;
2549 }else if( c=='\'' || c=='\"' ){ /* String a character literals */
2550 int startchar, prevc;
2553 for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
2554 if( c=='\n' ) lineno++;
2555 if( prevc=='\\' ) prevc = 0;
2561 ErrorMsg(ps.filename,ps.tokenlineno,
2562 "C code starting on this line is not terminated before the end of the file.");
2568 }else if( isalnum(c) ){ /* Identifiers */
2569 while( (c= *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
2571 }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
2574 }else if( (c=='/' || c=='|') && isalpha(cp[1]) ){
2576 while( (c = *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
2578 }else{ /* All other (one character) operators */
2583 *cp = 0; /* Null terminate the token */
2584 parseonetoken(&ps); /* Parse the token */
2585 *cp = c; /* Restore the buffer */
2588 free(filebuf); /* Release the buffer after parsing */
2589 gp->rule = ps.firstrule;
2590 gp->errorcnt = ps.errorcnt;
2592 /*************************** From the file "plink.c" *********************/
2594 ** Routines processing configuration follow-set propagation links
2595 ** in the LEMON parser generator.
2597 static struct plink *plink_freelist = 0;
2599 /* Allocate a new plink */
2600 struct plink *Plink_new(){
2603 if( plink_freelist==0 ){
2606 plink_freelist = (struct plink *)malloc( sizeof(struct plink)*amt );
2607 if( plink_freelist==0 ){
2609 "Unable to allocate memory for a new follow-set propagation link.\n");
2612 for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
2613 plink_freelist[amt-1].next = 0;
2615 new = plink_freelist;
2616 plink_freelist = plink_freelist->next;
2620 /* Add a plink to a plink list */
2621 void Plink_add(plpp,cfp)
2622 struct plink **plpp;
2632 /* Transfer every plink on the list "from" to the list "to" */
2633 void Plink_copy(to,from)
2637 struct plink *nextpl;
2639 nextpl = from->next;
2646 /* Delete every plink on the list */
2647 void Plink_delete(plp)
2650 struct plink *nextpl;
2654 plp->next = plink_freelist;
2655 plink_freelist = plp;
2659 /*********************** From the file "report.c" **************************/
2661 ** Procedures for generating reports and tables in the LEMON parser generator.
2664 /* Generate a filename with the given suffix. Space to hold the
2665 ** name comes from malloc() and must be freed by the calling
2668 PRIVATE char *file_makename(lemp,suffix)
2675 name = malloc( strlen(lemp->filename) + strlen(suffix) + 5 );
2677 fprintf(stderr,"Can't allocate space for a filename.\n");
2680 strcpy(name,lemp->filename);
2681 cp = strrchr(name,'.');
2683 strcat(name,suffix);
2687 /* Open a file with a name based on the name of the input file,
2688 ** but with a different (specified) suffix, and return a pointer
2690 PRIVATE FILE *file_open(lemp,suffix,mode)
2697 if( lemp->outname ) free(lemp->outname);
2698 lemp->outname = file_makename(lemp, suffix);
2699 fp = fopen(lemp->outname,mode);
2700 if( fp==0 && *mode=='w' ){
2701 fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
2708 /* Duplicate the input file without comments and without actions
2715 int i, j, maxlen, len, ncolumns, skip;
2716 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
2718 for(i=0; i<lemp->nsymbol; i++){
2719 sp = lemp->symbols[i];
2720 len = strlen(sp->name);
2721 if( len>maxlen ) maxlen = len;
2723 ncolumns = 76/(maxlen+5);
2724 if( ncolumns<1 ) ncolumns = 1;
2725 skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
2726 for(i=0; i<skip; i++){
2728 for(j=i; j<lemp->nsymbol; j+=skip){
2729 sp = lemp->symbols[j];
2730 assert( sp->index==j );
2731 printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
2735 for(rp=lemp->rule; rp; rp=rp->next){
2736 printf("%s",rp->lhs->name);
2737 /* if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
2739 for(i=0; i<rp->nrhs; i++){
2741 printf(" %s", sp->name);
2742 if( sp->type==MULTITERMINAL ){
2743 for(j=1; j<sp->nsubsym; j++){
2744 printf("|%s", sp->subsym[j]->name);
2747 /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
2750 if( rp->precsym ) printf(" [%s]",rp->precsym->name);
2751 /* if( rp->code ) printf("\n %s",rp->code); */
2756 void ConfigPrint(fp,cfp)
2764 fprintf(fp,"%s ::=",rp->lhs->name);
2765 for(i=0; i<=rp->nrhs; i++){
2766 if( i==cfp->dot ) fprintf(fp," *");
2767 if( i==rp->nrhs ) break;
2769 fprintf(fp," %s", sp->name);
2770 if( sp->type==MULTITERMINAL ){
2771 for(j=1; j<sp->nsubsym; j++){
2772 fprintf(fp,"|%s",sp->subsym[j]->name);
2781 PRIVATE void SetPrint(out,set,lemp)
2789 fprintf(out,"%12s[","");
2790 for(i=0; i<lemp->nterminal; i++){
2791 if( SetFind(set,i) ){
2792 fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
2799 /* Print a plink chain */
2800 PRIVATE void PlinkPrint(out,plp,tag)
2806 fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum);
2807 ConfigPrint(out,plp->cfp);
2814 /* Print an action to the given file descriptor. Return FALSE if
2815 ** nothing was actually printed.
2817 int PrintAction(struct action *ap, FILE *fp, int indent){
2821 fprintf(fp,"%*s shift %d",indent,ap->sp->name,ap->x.stp->statenum);
2824 fprintf(fp,"%*s reduce %d",indent,ap->sp->name,ap->x.rp->index);
2827 fprintf(fp,"%*s accept",indent,ap->sp->name);
2830 fprintf(fp,"%*s error",indent,ap->sp->name);
2833 fprintf(fp,"%*s reduce %-3d ** Parsing conflict **",
2834 indent,ap->sp->name,ap->x.rp->index);
2845 /* Generate the "y.output" log file */
2846 void ReportOutput(lemp)
2855 fp = file_open(lemp,".out","wb");
2858 for(i=0; i<lemp->nstate; i++){
2859 stp = lemp->sorted[i];
2860 fprintf(fp,"State %d:\n",stp->statenum);
2861 if( lemp->basisflag ) cfp=stp->bp;
2865 if( cfp->dot==cfp->rp->nrhs ){
2866 sprintf(buf,"(%d)",cfp->rp->index);
2867 fprintf(fp," %5s ",buf);
2871 ConfigPrint(fp,cfp);
2874 SetPrint(fp,cfp->fws,lemp);
2875 PlinkPrint(fp,cfp->fplp,"To ");
2876 PlinkPrint(fp,cfp->bplp,"From");
2878 if( lemp->basisflag ) cfp=cfp->bp;
2882 for(ap=stp->ap; ap; ap=ap->next){
2883 if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
2891 /* Search for the file "name" which is in the same directory as
2892 ** the exacutable */
2893 PRIVATE char *pathsearch(argv0,name,modemask)
2901 extern int access();
2904 cp = strrchr(argv0,'\\');
2906 cp = strrchr(argv0,'/');
2911 path = (char *)malloc( strlen(argv0) + strlen(name) + 2 );
2912 if( path ) sprintf(path,"%s/%s",argv0,name);
2915 extern char *getenv();
2916 pathlist = getenv("PATH");
2917 if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
2918 path = (char *)malloc( strlen(pathlist)+strlen(name)+2 );
2921 cp = strchr(pathlist,':');
2922 if( cp==0 ) cp = &pathlist[strlen(pathlist)];
2925 sprintf(path,"%s/%s",pathlist,name);
2927 if( c==0 ) pathlist = "";
2928 else pathlist = &cp[1];
2929 if( access(path,modemask)==0 ) break;
2936 /* Given an action, compute the integer value for that action
2937 ** which is to be put in the action table of the generated machine.
2938 ** Return negative if no action should be generated.
2940 PRIVATE int compute_action(lemp,ap)
2946 case SHIFT: act = ap->x.stp->statenum; break;
2947 case REDUCE: act = ap->x.rp->index + lemp->nstate; break;
2948 case ERROR: act = lemp->nstate + lemp->nrule; break;
2949 case ACCEPT: act = lemp->nstate + lemp->nrule + 1; break;
2950 default: act = -1; break;
2955 #define LINESIZE 1000
2956 /* The next cluster of routines are for reading the template file
2957 ** and writing the results to the generated parser */
2958 /* The first function transfers data from "in" to "out" until
2959 ** a line is seen which begins with "%%". The line number is
2962 ** if name!=0, then any word that begin with "Parse" is changed to
2963 ** begin with *name instead.
2965 PRIVATE void tplt_xfer(name,in,out,lineno)
2972 char line[LINESIZE];
2973 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
2977 for(i=0; line[i]; i++){
2978 if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
2979 && (i==0 || !isalpha(line[i-1]))
2981 if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
2982 fprintf(out,"%s",name);
2988 fprintf(out,"%s",&line[iStart]);
2992 /* The next function finds the template file and opens it, returning
2993 ** a pointer to the opened file. */
2994 PRIVATE FILE *tplt_open(lemp)
2997 static char templatename[] = "lempar.c";
3003 cp = strrchr(lemp->filename,'.');
3005 sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
3007 sprintf(buf,"%s.lt",lemp->filename);
3009 if( access(buf,004)==0 ){
3011 }else if( access(templatename,004)==0 ){
3012 tpltname = templatename;
3014 tpltname = pathsearch(lemp->argv0,templatename,0);
3017 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3022 in = fopen(tpltname,"rb");
3024 fprintf(stderr,"Can't open the template file \"%s\".\n",templatename);
3031 /* Print a #line directive line to the output file. */
3032 PRIVATE void tplt_linedir(out,lineno,filename)
3037 fprintf(out,"#line %d \"",lineno);
3039 if( *filename == '\\' ) putc('\\',out);
3040 putc(*filename,out);
3043 fprintf(out,"\"\n");
3046 /* Print a string to the file and keep the linenumber up to date */
3047 PRIVATE void tplt_print(out,lemp,str,strln,lineno)
3054 if( str==0 ) return;
3055 tplt_linedir(out,strln,lemp->filename);
3058 if( *str=='\n' ) (*lineno)++;
3062 if( str[-1]!='\n' ){
3066 tplt_linedir(out,*lineno+2,lemp->outname);
3072 ** The following routine emits code for the destructor for the
3075 void emit_destructor_code(out,sp,lemp,lineno)
3084 if( sp->type==TERMINAL ){
3085 cp = lemp->tokendest;
3087 tplt_linedir(out,lemp->tokendestln,lemp->filename);
3089 }else if( sp->destructor ){
3090 cp = sp->destructor;
3091 tplt_linedir(out,sp->destructorln,lemp->filename);
3093 }else if( lemp->vardest ){
3096 tplt_linedir(out,lemp->vardestln,lemp->filename);
3099 assert( 0 ); /* Cannot happen */
3102 if( *cp=='$' && cp[1]=='$' ){
3103 fprintf(out,"(yypminor->yy%d)",sp->dtnum);
3107 if( *cp=='\n' ) linecnt++;
3110 (*lineno) += 3 + linecnt;
3112 tplt_linedir(out,*lineno,lemp->outname);
3117 ** Return TRUE (non-zero) if the given symbol has a destructor.
3119 int has_destructor(sp, lemp)
3124 if( sp->type==TERMINAL ){
3125 ret = lemp->tokendest!=0;
3127 ret = lemp->vardest!=0 || sp->destructor!=0;
3133 ** Append text to a dynamically allocated string. If zText is 0 then
3134 ** reset the string to be empty again. Always return the complete text
3135 ** of the string (which is overwritten with each call).
3137 ** n bytes of zText are stored. If n==0 then all of zText up to the first
3138 ** \000 terminator is stored. zText can contain up to two instances of
3139 ** %d. The values of p1 and p2 are written into the first and second
3142 ** If n==-1, then the previous character is overwritten.
3144 PRIVATE char *append_str(char *zText, int n, int p1, int p2){
3146 static int alloced = 0;
3147 static int used = 0;
3162 if( n+sizeof(zInt)*2+used >= alloced ){
3163 alloced = n + sizeof(zInt)*2 + used + 200;
3164 z = realloc(z, alloced);
3166 if( z==0 ) return "";
3169 if( c=='%' && zText[0]=='d' ){
3170 sprintf(zInt, "%d", p1);
3172 strcpy(&z[used], zInt);
3173 used += strlen(&z[used]);
3185 ** zCode is a string that is the action associated with a rule. Expand
3186 ** the symbols in this string so that the refer to elements of the parser
3189 PRIVATE void translate_code(struct lemon *lemp, struct rule *rp){
3192 char lhsused = 0; /* True if the LHS element has been used */
3193 char used[MAXRHS]; /* True for each RHS element which is used */
3195 for(i=0; i<rp->nrhs; i++) used[i] = 0;
3198 append_str(0,0,0,0);
3199 for(cp=rp->code; *cp; cp++){
3200 if( isalpha(*cp) && (cp==rp->code || (!isalnum(cp[-1]) && cp[-1]!='_')) ){
3202 for(xp= &cp[1]; isalnum(*xp) || *xp=='_'; xp++);
3205 if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
3206 append_str("yygotominor.yy%d",0,rp->lhs->dtnum,0);
3210 for(i=0; i<rp->nrhs; i++){
3211 if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
3212 if( cp!=rp->code && cp[-1]=='@' ){
3213 /* If the argument is of the form @X then substituted
3214 ** the token number of X, not the value of X */
3215 append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
3217 struct symbol *sp = rp->rhs[i];
3219 if( sp->type==MULTITERMINAL ){
3220 dtnum = sp->subsym[0]->dtnum;
3224 append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);
3234 append_str(cp, 1, 0, 0);
3237 /* Check to make sure the LHS has been used */
3238 if( rp->lhsalias && !lhsused ){
3239 ErrorMsg(lemp->filename,rp->ruleline,
3240 "Label \"%s\" for \"%s(%s)\" is never used.",
3241 rp->lhsalias,rp->lhs->name,rp->lhsalias);
3245 /* Generate destructor code for RHS symbols which are not used in the
3247 for(i=0; i<rp->nrhs; i++){
3248 if( rp->rhsalias[i] && !used[i] ){
3249 ErrorMsg(lemp->filename,rp->ruleline,
3250 "Label %s for \"%s(%s)\" is never used.",
3251 rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
3253 }else if( rp->rhsalias[i]==0 ){
3254 if( has_destructor(rp->rhs[i],lemp) ){
3255 append_str(" yy_destructor(%d,&yymsp[%d].minor);\n", 0,
3256 rp->rhs[i]->index,i-rp->nrhs+1);
3258 /* No destructor defined for this term */
3262 cp = append_str(0,0,0,0);
3263 rp->code = Strsafe(cp);
3267 ** Generate code which executes when the rule "rp" is reduced. Write
3268 ** the code to "out". Make sure lineno stays up-to-date.
3270 PRIVATE void emit_code(out,rp,lemp,lineno)
3279 /* Generate code to do the reduce action */
3281 tplt_linedir(out,rp->line,lemp->filename);
3282 fprintf(out,"{%s",rp->code);
3283 for(cp=rp->code; *cp; cp++){
3284 if( *cp=='\n' ) linecnt++;
3286 (*lineno) += 3 + linecnt;
3288 tplt_linedir(out,*lineno,lemp->outname);
3289 } /* End if( rp->code ) */
3295 ** Print the definition of the union used for the parser's data stack.
3296 ** This union contains fields for every possible data type for tokens
3297 ** and nonterminals. In the process of computing and printing this
3298 ** union, also set the ".dtnum" field of every terminal and nonterminal
3301 void print_stack_union(out,lemp,plineno,mhflag)
3302 FILE *out; /* The output stream */
3303 struct lemon *lemp; /* The main info structure for this parser */
3304 int *plineno; /* Pointer to the line number */
3305 int mhflag; /* True if generating makeheaders output */
3307 int lineno = *plineno; /* The line number of the output */
3308 char **types; /* A hash table of datatypes */
3309 int arraysize; /* Size of the "types" array */
3310 int maxdtlength; /* Maximum length of any ".datatype" field. */
3311 char *stddt; /* Standardized name for a datatype */
3312 int i,j; /* Loop counters */
3313 int hash; /* For hashing the name of a type */
3314 char *name; /* Name of the parser */
3316 /* Allocate and initialize types[] and allocate stddt[] */
3317 arraysize = lemp->nsymbol * 2;
3318 types = (char**)malloc( arraysize * sizeof(char*) );
3319 for(i=0; i<arraysize; i++) types[i] = 0;
3321 if( lemp->vartype ){
3322 maxdtlength = strlen(lemp->vartype);
3324 for(i=0; i<lemp->nsymbol; i++){
3326 struct symbol *sp = lemp->symbols[i];
3327 if( sp->datatype==0 ) continue;
3328 len = strlen(sp->datatype);
3329 if( len>maxdtlength ) maxdtlength = len;
3331 stddt = (char*)malloc( maxdtlength*2 + 1 );
3332 if( types==0 || stddt==0 ){
3333 fprintf(stderr,"Out of memory.\n");
3337 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3338 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
3339 ** used for terminal symbols. If there is no %default_type defined then
3340 ** 0 is also used as the .dtnum value for nonterminals which do not specify
3341 ** a datatype using the %type directive.
3343 for(i=0; i<lemp->nsymbol; i++){
3344 struct symbol *sp = lemp->symbols[i];
3346 if( sp==lemp->errsym ){
3347 sp->dtnum = arraysize+1;
3350 if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
3355 if( cp==0 ) cp = lemp->vartype;
3357 while( isspace(*cp) ) cp++;
3358 while( *cp ) stddt[j++] = *cp++;
3359 while( j>0 && isspace(stddt[j-1]) ) j--;
3362 for(j=0; stddt[j]; j++){
3363 hash = hash*53 + stddt[j];
3365 hash = (hash & 0x7fffffff)%arraysize;
3366 while( types[hash] ){
3367 if( strcmp(types[hash],stddt)==0 ){
3368 sp->dtnum = hash + 1;
3372 if( hash>=arraysize ) hash = 0;
3374 if( types[hash]==0 ){
3375 sp->dtnum = hash + 1;
3376 types[hash] = (char*)malloc( strlen(stddt)+1 );
3377 if( types[hash]==0 ){
3378 fprintf(stderr,"Out of memory.\n");
3381 strcpy(types[hash],stddt);
3385 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
3386 name = lemp->name ? lemp->name : "Parse";
3388 if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
3389 fprintf(out,"#define %sTOKENTYPE %s\n",name,
3390 lemp->tokentype?lemp->tokentype:"void*"); lineno++;
3391 if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
3392 fprintf(out,"typedef union {\n"); lineno++;
3393 fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++;
3394 for(i=0; i<arraysize; i++){
3395 if( types[i]==0 ) continue;
3396 fprintf(out," %s yy%d;\n",types[i],i+1); lineno++;
3399 fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++;
3402 fprintf(out,"} YYMINORTYPE;\n"); lineno++;
3407 ** Return the name of a C datatype able to represent values between
3408 ** lwr and upr, inclusive.
3410 static const char *minimum_size_type(int lwr, int upr){
3413 return "unsigned char";
3414 }else if( upr<65535 ){
3415 return "unsigned short int";
3417 return "unsigned int";
3419 }else if( lwr>=-127 && upr<=127 ){
3420 return "signed char";
3421 }else if( lwr>=-32767 && upr<32767 ){
3429 ** Each state contains a set of token transaction and a set of
3430 ** nonterminal transactions. Each of these sets makes an instance
3431 ** of the following structure. An array of these structures is used
3432 ** to order the creation of entries in the yy_action[] table.
3435 struct state *stp; /* A pointer to a state */
3436 int isTkn; /* True to use tokens. False for non-terminals */
3437 int nAction; /* Number of actions */
3441 ** Compare to axset structures for sorting purposes
3443 static int axset_compare(const void *a, const void *b){
3444 struct axset *p1 = (struct axset*)a;
3445 struct axset *p2 = (struct axset*)b;
3446 return p2->nAction - p1->nAction;
3449 /* Generate C source code for the parser */
3450 void ReportTable(lemp, mhflag)
3452 int mhflag; /* Output in makeheaders format if true */
3455 char line[LINESIZE];
3460 struct acttab *pActtab;
3463 int mnTknOfst, mxTknOfst;
3464 int mnNtOfst, mxNtOfst;
3467 in = tplt_open(lemp);
3469 out = file_open(lemp,".c","wb");
3475 tplt_xfer(lemp->name,in,out,&lineno);
3477 /* Generate the include code, if any */
3478 tplt_print(out,lemp,lemp->include,lemp->includeln,&lineno);
3480 char *name = file_makename(lemp, ".h");
3481 fprintf(out,"#include \"%s\"\n", name); lineno++;
3484 tplt_xfer(lemp->name,in,out,&lineno);
3486 /* Generate #defines for all tokens */
3489 fprintf(out,"#if INTERFACE\n"); lineno++;
3490 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
3492 for(i=1; i<lemp->nterminal; i++){
3493 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3496 fprintf(out,"#endif\n"); lineno++;
3498 tplt_xfer(lemp->name,in,out,&lineno);
3500 /* Generate the defines */
3501 fprintf(out,"#define YYCODETYPE %s\n",
3502 minimum_size_type(0, lemp->nsymbol+5)); lineno++;
3503 fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol+1); lineno++;
3504 fprintf(out,"#define YYACTIONTYPE %s\n",
3505 minimum_size_type(0, lemp->nstate+lemp->nrule+5)); lineno++;
3506 if( lemp->wildcard ){
3507 fprintf(out,"#define YYWILDCARD %d\n",
3508 lemp->wildcard->index); lineno++;
3510 print_stack_union(out,lemp,&lineno,mhflag);
3511 if( lemp->stacksize ){
3512 if( atoi(lemp->stacksize)<=0 ){
3513 ErrorMsg(lemp->filename,0,
3514 "Illegal stack size: [%s]. The stack size should be an integer constant.",
3517 lemp->stacksize = "100";
3519 fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++;
3521 fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++;
3524 fprintf(out,"#if INTERFACE\n"); lineno++;
3526 name = lemp->name ? lemp->name : "Parse";
3527 if( lemp->arg && lemp->arg[0] ){
3529 i = strlen(lemp->arg);
3530 while( i>=1 && isspace(lemp->arg[i-1]) ) i--;
3531 while( i>=1 && (isalnum(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
3532 fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++;
3533 fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++;
3534 fprintf(out,"#define %sARG_FETCH %s = yypParser->%s\n",
3535 name,lemp->arg,&lemp->arg[i]); lineno++;
3536 fprintf(out,"#define %sARG_STORE yypParser->%s = %s\n",
3537 name,&lemp->arg[i],&lemp->arg[i]); lineno++;
3539 fprintf(out,"#define %sARG_SDECL\n",name); lineno++;
3540 fprintf(out,"#define %sARG_PDECL\n",name); lineno++;
3541 fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
3542 fprintf(out,"#define %sARG_STORE\n",name); lineno++;
3545 fprintf(out,"#endif\n"); lineno++;
3547 fprintf(out,"#define YYNSTATE %d\n",lemp->nstate); lineno++;
3548 fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++;
3549 fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++;
3550 fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++;
3551 if( lemp->has_fallback ){
3552 fprintf(out,"#define YYFALLBACK 1\n"); lineno++;
3554 tplt_xfer(lemp->name,in,out,&lineno);
3556 /* Generate the action table and its associates:
3558 ** yy_action[] A single table containing all actions.
3559 ** yy_lookahead[] A table containing the lookahead for each entry in
3560 ** yy_action. Used to detect hash collisions.
3561 ** yy_shift_ofst[] For each state, the offset into yy_action for
3562 ** shifting terminals.
3563 ** yy_reduce_ofst[] For each state, the offset into yy_action for
3564 ** shifting non-terminals after a reduce.
3565 ** yy_default[] Default action for each state.
3568 /* Compute the actions on all states and count them up */
3569 ax = malloc( sizeof(ax[0])*lemp->nstate*2 );
3571 fprintf(stderr,"malloc failed\n");
3574 for(i=0; i<lemp->nstate; i++){
3575 stp = lemp->sorted[i];
3578 ax[i*2].nAction = stp->nTknAct;
3579 ax[i*2+1].stp = stp;
3580 ax[i*2+1].isTkn = 0;
3581 ax[i*2+1].nAction = stp->nNtAct;
3583 mxTknOfst = mnTknOfst = 0;
3584 mxNtOfst = mnNtOfst = 0;
3586 /* Compute the action table. In order to try to keep the size of the
3587 ** action table to a minimum, the heuristic of placing the largest action
3588 ** sets first is used.
3590 qsort(ax, lemp->nstate*2, sizeof(ax[0]), axset_compare);
3591 pActtab = acttab_alloc();
3592 for(i=0; i<lemp->nstate*2 && ax[i].nAction>0; i++){
3595 for(ap=stp->ap; ap; ap=ap->next){
3597 if( ap->sp->index>=lemp->nterminal ) continue;
3598 action = compute_action(lemp, ap);
3599 if( action<0 ) continue;
3600 acttab_action(pActtab, ap->sp->index, action);
3602 stp->iTknOfst = acttab_insert(pActtab);
3603 if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
3604 if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
3606 for(ap=stp->ap; ap; ap=ap->next){
3608 if( ap->sp->index<lemp->nterminal ) continue;
3609 if( ap->sp->index==lemp->nsymbol ) continue;
3610 action = compute_action(lemp, ap);
3611 if( action<0 ) continue;
3612 acttab_action(pActtab, ap->sp->index, action);
3614 stp->iNtOfst = acttab_insert(pActtab);
3615 if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
3616 if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
3621 /* Output the yy_action table */
3622 fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++;
3623 n = acttab_size(pActtab);
3624 for(i=j=0; i<n; i++){
3625 int action = acttab_yyaction(pActtab, i);
3626 if( action<0 ) action = lemp->nsymbol + lemp->nrule + 2;
3627 if( j==0 ) fprintf(out," /* %5d */ ", i);
3628 fprintf(out, " %4d,", action);
3629 if( j==9 || i==n-1 ){
3630 fprintf(out, "\n"); lineno++;
3636 fprintf(out, "};\n"); lineno++;
3638 /* Output the yy_lookahead table */
3639 fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++;
3640 for(i=j=0; i<n; i++){
3641 int la = acttab_yylookahead(pActtab, i);
3642 if( la<0 ) la = lemp->nsymbol;
3643 if( j==0 ) fprintf(out," /* %5d */ ", i);
3644 fprintf(out, " %4d,", la);
3645 if( j==9 || i==n-1 ){
3646 fprintf(out, "\n"); lineno++;
3652 fprintf(out, "};\n"); lineno++;
3654 /* Output the yy_shift_ofst[] table */
3655 fprintf(out, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst-1); lineno++;
3657 while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--;
3658 fprintf(out, "#define YY_SHIFT_MAX %d\n", n-1); lineno++;
3659 fprintf(out, "static const %s yy_shift_ofst[] = {\n",
3660 minimum_size_type(mnTknOfst-1, mxTknOfst)); lineno++;
3661 for(i=j=0; i<n; i++){
3663 stp = lemp->sorted[i];
3664 ofst = stp->iTknOfst;
3665 if( ofst==NO_OFFSET ) ofst = mnTknOfst - 1;
3666 if( j==0 ) fprintf(out," /* %5d */ ", i);
3667 fprintf(out, " %4d,", ofst);
3668 if( j==9 || i==n-1 ){
3669 fprintf(out, "\n"); lineno++;
3675 fprintf(out, "};\n"); lineno++;
3677 /* Output the yy_reduce_ofst[] table */
3678 fprintf(out, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst-1); lineno++;
3680 while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--;
3681 fprintf(out, "#define YY_REDUCE_MAX %d\n", n-1); lineno++;
3682 fprintf(out, "static const %s yy_reduce_ofst[] = {\n",
3683 minimum_size_type(mnNtOfst-1, mxNtOfst)); lineno++;
3684 for(i=j=0; i<n; i++){
3686 stp = lemp->sorted[i];
3687 ofst = stp->iNtOfst;
3688 if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
3689 if( j==0 ) fprintf(out," /* %5d */ ", i);
3690 fprintf(out, " %4d,", ofst);
3691 if( j==9 || i==n-1 ){
3692 fprintf(out, "\n"); lineno++;
3698 fprintf(out, "};\n"); lineno++;
3700 /* Output the default action table */
3701 fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++;
3703 for(i=j=0; i<n; i++){
3704 stp = lemp->sorted[i];
3705 if( j==0 ) fprintf(out," /* %5d */ ", i);
3706 fprintf(out, " %4d,", stp->iDflt);
3707 if( j==9 || i==n-1 ){
3708 fprintf(out, "\n"); lineno++;
3714 fprintf(out, "};\n"); lineno++;
3715 tplt_xfer(lemp->name,in,out,&lineno);
3717 /* Generate the table of fallback tokens.
3719 if( lemp->has_fallback ){
3720 for(i=0; i<lemp->nterminal; i++){
3721 struct symbol *p = lemp->symbols[i];
3722 if( p->fallback==0 ){
3723 fprintf(out, " 0, /* %10s => nothing */\n", p->name);
3725 fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index,
3726 p->name, p->fallback->name);
3731 tplt_xfer(lemp->name, in, out, &lineno);
3733 /* Generate a table containing the symbolic name of every symbol
3735 for(i=0; i<lemp->nsymbol; i++){
3736 sprintf(line,"\"%s\",",lemp->symbols[i]->name);
3737 fprintf(out," %-15s",line);
3738 if( (i&3)==3 ){ fprintf(out,"\n"); lineno++; }
3740 if( (i&3)!=0 ){ fprintf(out,"\n"); lineno++; }
3741 tplt_xfer(lemp->name,in,out,&lineno);
3743 /* Generate a table containing a text string that describes every
3744 ** rule in the rule set of the grammer. This information is used
3745 ** when tracing REDUCE actions.
3747 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
3748 assert( rp->index==i );
3749 fprintf(out," /* %3d */ \"%s ::=", i, rp->lhs->name);
3750 for(j=0; j<rp->nrhs; j++){
3751 struct symbol *sp = rp->rhs[j];
3752 fprintf(out," %s", sp->name);
3753 if( sp->type==MULTITERMINAL ){
3755 for(k=1; k<sp->nsubsym; k++){
3756 fprintf(out,"|%s",sp->subsym[k]->name);
3760 fprintf(out,"\",\n"); lineno++;
3762 tplt_xfer(lemp->name,in,out,&lineno);
3764 /* Generate code which executes every time a symbol is popped from
3765 ** the stack while processing errors or while destroying the parser.
3766 ** (In other words, generate the %destructor actions)
3768 if( lemp->tokendest ){
3769 for(i=0; i<lemp->nsymbol; i++){
3770 struct symbol *sp = lemp->symbols[i];
3771 if( sp==0 || sp->type!=TERMINAL ) continue;
3772 fprintf(out," case %d:\n",sp->index); lineno++;
3774 for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
3775 if( i<lemp->nsymbol ){
3776 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
3777 fprintf(out," break;\n"); lineno++;
3780 if( lemp->vardest ){
3781 struct symbol *dflt_sp = 0;
3782 for(i=0; i<lemp->nsymbol; i++){
3783 struct symbol *sp = lemp->symbols[i];
3784 if( sp==0 || sp->type==TERMINAL ||
3785 sp->index<=0 || sp->destructor!=0 ) continue;
3786 fprintf(out," case %d:\n",sp->index); lineno++;
3790 emit_destructor_code(out,dflt_sp,lemp,&lineno);
3791 fprintf(out," break;\n"); lineno++;
3794 for(i=0; i<lemp->nsymbol; i++){
3795 struct symbol *sp = lemp->symbols[i];
3796 if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
3797 fprintf(out," case %d:\n",sp->index); lineno++;
3799 /* Combine duplicate destructors into a single case */
3800 for(j=i+1; j<lemp->nsymbol; j++){
3801 struct symbol *sp2 = lemp->symbols[j];
3802 if( sp2 && sp2->type!=TERMINAL && sp2->destructor
3803 && sp2->dtnum==sp->dtnum
3804 && strcmp(sp->destructor,sp2->destructor)==0 ){
3805 fprintf(out," case %d:\n",sp2->index); lineno++;
3806 sp2->destructor = 0;
3810 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
3811 fprintf(out," break;\n"); lineno++;
3813 tplt_xfer(lemp->name,in,out,&lineno);
3815 /* Generate code which executes whenever the parser stack overflows */
3816 tplt_print(out,lemp,lemp->overflow,lemp->overflowln,&lineno);
3817 tplt_xfer(lemp->name,in,out,&lineno);
3819 /* Generate the table of rule information
3821 ** Note: This code depends on the fact that rules are number
3822 ** sequentually beginning with 0.
3824 for(rp=lemp->rule; rp; rp=rp->next){
3825 fprintf(out," { %d, %d },\n",rp->lhs->index,rp->nrhs); lineno++;
3827 tplt_xfer(lemp->name,in,out,&lineno);
3829 /* Generate code which execution during each REDUCE action */
3830 for(rp=lemp->rule; rp; rp=rp->next){
3831 if( rp->code ) translate_code(lemp, rp);
3833 for(rp=lemp->rule; rp; rp=rp->next){
3835 if( rp->code==0 ) continue;
3836 fprintf(out," case %d:\n",rp->index); lineno++;
3837 for(rp2=rp->next; rp2; rp2=rp2->next){
3838 if( rp2->code==rp->code ){
3839 fprintf(out," case %d:\n",rp2->index); lineno++;
3843 emit_code(out,rp,lemp,&lineno);
3844 fprintf(out," break;\n"); lineno++;
3846 tplt_xfer(lemp->name,in,out,&lineno);
3848 /* Generate code which executes if a parse fails */
3849 tplt_print(out,lemp,lemp->failure,lemp->failureln,&lineno);
3850 tplt_xfer(lemp->name,in,out,&lineno);
3852 /* Generate code which executes when a syntax error occurs */
3853 tplt_print(out,lemp,lemp->error,lemp->errorln,&lineno);
3854 tplt_xfer(lemp->name,in,out,&lineno);
3856 /* Generate code which executes when the parser accepts its input */
3857 tplt_print(out,lemp,lemp->accept,lemp->acceptln,&lineno);
3858 tplt_xfer(lemp->name,in,out,&lineno);
3860 /* Append any addition code the user desires */
3861 tplt_print(out,lemp,lemp->extracode,lemp->extracodeln,&lineno);
3868 /* Generate a header file for the parser */
3869 void ReportHeader(lemp)
3874 char line[LINESIZE];
3875 char pattern[LINESIZE];
3878 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
3880 in = file_open(lemp,".h","rb");
3882 for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
3883 sprintf(pattern,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3884 if( strcmp(line,pattern) ) break;
3887 if( i==lemp->nterminal ){
3888 /* No change in the file. Don't rewrite it. */
3892 out = file_open(lemp,".h","wb");
3894 for(i=1; i<lemp->nterminal; i++){
3895 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3902 /* Reduce the size of the action tables, if possible, by making use
3905 ** In this version, we take the most frequent REDUCE action and make
3906 ** it the default. Except, there is no default if the wildcard token
3907 ** is a possible look-ahead.
3909 void CompressTables(lemp)
3913 struct action *ap, *ap2;
3914 struct rule *rp, *rp2, *rbest;
3919 for(i=0; i<lemp->nstate; i++){
3920 stp = lemp->sorted[i];
3925 for(ap=stp->ap; ap; ap=ap->next){
3926 if( ap->type==SHIFT && ap->sp==lemp->wildcard ){
3929 if( ap->type!=REDUCE ) continue;
3931 if( rp==rbest ) continue;
3933 for(ap2=ap->next; ap2; ap2=ap2->next){
3934 if( ap2->type!=REDUCE ) continue;
3936 if( rp2==rbest ) continue;
3945 /* Do not make a default if the number of rules to default
3946 ** is not at least 1 or if the wildcard token is a possible
3949 if( nbest<1 || usesWildcard ) continue;
3952 /* Combine matching REDUCE actions into a single default */
3953 for(ap=stp->ap; ap; ap=ap->next){
3954 if( ap->type==REDUCE && ap->x.rp==rbest ) break;
3957 ap->sp = Symbol_new("{default}");
3958 for(ap=ap->next; ap; ap=ap->next){
3959 if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
3961 stp->ap = Action_sort(stp->ap);
3967 ** Compare two states for sorting purposes. The smaller state is the
3968 ** one with the most non-terminal actions. If they have the same number
3969 ** of non-terminal actions, then the smaller is the one with the most
3972 static int stateResortCompare(const void *a, const void *b){
3973 const struct state *pA = *(const struct state**)a;
3974 const struct state *pB = *(const struct state**)b;
3977 n = pB->nNtAct - pA->nNtAct;
3979 n = pB->nTknAct - pA->nTknAct;
3986 ** Renumber and resort states so that states with fewer choices
3987 ** occur at the end. Except, keep state 0 as the first state.
3989 void ResortStates(lemp)
3996 for(i=0; i<lemp->nstate; i++){
3997 stp = lemp->sorted[i];
3998 stp->nTknAct = stp->nNtAct = 0;
3999 stp->iDflt = lemp->nstate + lemp->nrule;
4000 stp->iTknOfst = NO_OFFSET;
4001 stp->iNtOfst = NO_OFFSET;
4002 for(ap=stp->ap; ap; ap=ap->next){
4003 if( compute_action(lemp,ap)>=0 ){
4004 if( ap->sp->index<lemp->nterminal ){
4006 }else if( ap->sp->index<lemp->nsymbol ){
4009 stp->iDflt = compute_action(lemp, ap);
4014 qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]),
4015 stateResortCompare);
4016 for(i=0; i<lemp->nstate; i++){
4017 lemp->sorted[i]->statenum = i;
4022 /***************** From the file "set.c" ************************************/
4024 ** Set manipulation routines for the LEMON parser generator.
4027 static int size = 0;
4029 /* Set the set size */
4036 /* Allocate a new set */
4040 s = (char*)malloc( size );
4042 extern void memory_error();
4045 for(i=0; i<size; i++) s[i] = 0;
4049 /* Deallocate a set */
4056 /* Add a new element to the set. Return TRUE if the element was added
4057 ** and FALSE if it was already there. */
4068 /* Add every element of s2 to s1. Return TRUE if s1 changes. */
4075 for(i=0; i<size; i++){
4076 if( s2[i]==0 ) continue;
4084 /********************** From the file "table.c" ****************************/
4086 ** All code in this file has been automatically generated
4087 ** from a specification in the file
4089 ** by the associative array code building program "aagen".
4090 ** Do not edit this file! Instead, edit the specification
4091 ** file, then rerun aagen.
4094 ** Code for processing tables in the LEMON parser generator.
4097 PRIVATE int strhash(x)
4101 while( *x) h = h*13 + *(x++);
4105 /* Works like strdup, sort of. Save a string in malloced memory, but
4106 ** keep strings in a table so that the same string is not in more
4114 if( y==0 ) return 0;
4115 z = Strsafe_find(y);
4116 if( z==0 && (z=malloc( strlen(y)+1 ))!=0 ){
4124 /* There is one instance of the following structure for each
4125 ** associative array of type "x1".
4128 int size; /* The number of available slots. */
4129 /* Must be a power of 2 greater than or */
4131 int count; /* Number of currently slots filled */
4132 struct s_x1node *tbl; /* The data stored here */
4133 struct s_x1node **ht; /* Hash table for lookups */
4136 /* There is one instance of this structure for every data element
4137 ** in an associative array of type "x1".
4139 typedef struct s_x1node {
4140 char *data; /* The data */
4141 struct s_x1node *next; /* Next entry with the same hash */
4142 struct s_x1node **from; /* Previous link */
4145 /* There is only one instance of the array, which is the following */
4146 static struct s_x1 *x1a;
4148 /* Allocate a new associative array */
4149 void Strsafe_init(){
4151 x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
4155 x1a->tbl = (x1node*)malloc(
4156 (sizeof(x1node) + sizeof(x1node*))*1024 );
4162 x1a->ht = (x1node**)&(x1a->tbl[1024]);
4163 for(i=0; i<1024; i++) x1a->ht[i] = 0;
4167 /* Insert a new record into the array. Return TRUE if successful.
4168 ** Prior data with the same key is NOT overwritten */
4169 int Strsafe_insert(data)
4176 if( x1a==0 ) return 0;
4178 h = ph & (x1a->size-1);
4181 if( strcmp(np->data,data)==0 ){
4182 /* An existing entry with the same key is found. */
4183 /* Fail because overwrite is not allows. */
4188 if( x1a->count>=x1a->size ){
4189 /* Need to make the hash table bigger */
4192 array.size = size = x1a->size*2;
4193 array.count = x1a->count;
4194 array.tbl = (x1node*)malloc(
4195 (sizeof(x1node) + sizeof(x1node*))*size );
4196 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4197 array.ht = (x1node**)&(array.tbl[size]);
4198 for(i=0; i<size; i++) array.ht[i] = 0;
4199 for(i=0; i<x1a->count; i++){
4200 x1node *oldnp, *newnp;
4201 oldnp = &(x1a->tbl[i]);
4202 h = strhash(oldnp->data) & (size-1);
4203 newnp = &(array.tbl[i]);
4204 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4205 newnp->next = array.ht[h];
4206 newnp->data = oldnp->data;
4207 newnp->from = &(array.ht[h]);
4208 array.ht[h] = newnp;
4213 /* Insert the new data */
4214 h = ph & (x1a->size-1);
4215 np = &(x1a->tbl[x1a->count++]);
4217 if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
4218 np->next = x1a->ht[h];
4220 np->from = &(x1a->ht[h]);
4224 /* Return a pointer to data assigned to the given key. Return NULL
4225 ** if no such key. */
4226 char *Strsafe_find(key)
4232 if( x1a==0 ) return 0;
4233 h = strhash(key) & (x1a->size-1);
4236 if( strcmp(np->data,key)==0 ) break;
4239 return np ? np->data : 0;
4242 /* Return a pointer to the (terminal or nonterminal) symbol "x".
4243 ** Create a new symbol if this is the first time "x" has been seen.
4245 struct symbol *Symbol_new(x)
4250 sp = Symbol_find(x);
4252 sp = (struct symbol *)malloc( sizeof(struct symbol) );
4254 sp->name = Strsafe(x);
4255 sp->type = isupper(*x) ? TERMINAL : NONTERMINAL;
4261 sp->lambda = B_FALSE;
4264 Symbol_insert(sp,sp->name);
4269 /* Compare two symbols for working purposes
4271 ** Symbols that begin with upper case letters (terminals or tokens)
4272 ** must sort before symbols that begin with lower case letters
4273 ** (non-terminals). Other than that, the order does not matter.
4275 ** We find experimentally that leaving the symbols in their original
4276 ** order (the order they appeared in the grammar file) gives the
4277 ** smallest parser tables in SQLite.
4279 int Symbolcmpp(struct symbol **a, struct symbol **b){
4280 int i1 = (**a).index + 10000000*((**a).name[0]>'Z');
4281 int i2 = (**b).index + 10000000*((**b).name[0]>'Z');
4285 /* There is one instance of the following structure for each
4286 ** associative array of type "x2".
4289 int size; /* The number of available slots. */
4290 /* Must be a power of 2 greater than or */
4292 int count; /* Number of currently slots filled */
4293 struct s_x2node *tbl; /* The data stored here */
4294 struct s_x2node **ht; /* Hash table for lookups */
4297 /* There is one instance of this structure for every data element
4298 ** in an associative array of type "x2".
4300 typedef struct s_x2node {
4301 struct symbol *data; /* The data */
4302 char *key; /* The key */
4303 struct s_x2node *next; /* Next entry with the same hash */
4304 struct s_x2node **from; /* Previous link */
4307 /* There is only one instance of the array, which is the following */
4308 static struct s_x2 *x2a;
4310 /* Allocate a new associative array */
4313 x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
4317 x2a->tbl = (x2node*)malloc(
4318 (sizeof(x2node) + sizeof(x2node*))*128 );
4324 x2a->ht = (x2node**)&(x2a->tbl[128]);
4325 for(i=0; i<128; i++) x2a->ht[i] = 0;
4329 /* Insert a new record into the array. Return TRUE if successful.
4330 ** Prior data with the same key is NOT overwritten */
4331 int Symbol_insert(data,key)
4332 struct symbol *data;
4339 if( x2a==0 ) return 0;
4341 h = ph & (x2a->size-1);
4344 if( strcmp(np->key,key)==0 ){
4345 /* An existing entry with the same key is found. */
4346 /* Fail because overwrite is not allows. */
4351 if( x2a->count>=x2a->size ){
4352 /* Need to make the hash table bigger */
4355 array.size = size = x2a->size*2;
4356 array.count = x2a->count;
4357 array.tbl = (x2node*)malloc(
4358 (sizeof(x2node) + sizeof(x2node*))*size );
4359 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4360 array.ht = (x2node**)&(array.tbl[size]);
4361 for(i=0; i<size; i++) array.ht[i] = 0;
4362 for(i=0; i<x2a->count; i++){
4363 x2node *oldnp, *newnp;
4364 oldnp = &(x2a->tbl[i]);
4365 h = strhash(oldnp->key) & (size-1);
4366 newnp = &(array.tbl[i]);
4367 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4368 newnp->next = array.ht[h];
4369 newnp->key = oldnp->key;
4370 newnp->data = oldnp->data;
4371 newnp->from = &(array.ht[h]);
4372 array.ht[h] = newnp;
4377 /* Insert the new data */
4378 h = ph & (x2a->size-1);
4379 np = &(x2a->tbl[x2a->count++]);
4382 if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
4383 np->next = x2a->ht[h];
4385 np->from = &(x2a->ht[h]);
4389 /* Return a pointer to data assigned to the given key. Return NULL
4390 ** if no such key. */
4391 struct symbol *Symbol_find(key)
4397 if( x2a==0 ) return 0;
4398 h = strhash(key) & (x2a->size-1);
4401 if( strcmp(np->key,key)==0 ) break;
4404 return np ? np->data : 0;
4407 /* Return the n-th data. Return NULL if n is out of range. */
4408 struct symbol *Symbol_Nth(n)
4411 struct symbol *data;
4412 if( x2a && n>0 && n<=x2a->count ){
4413 data = x2a->tbl[n-1].data;
4420 /* Return the size of the array */
4423 return x2a ? x2a->count : 0;
4426 /* Return an array of pointers to all data in the table.
4427 ** The array is obtained from malloc. Return NULL if memory allocation
4428 ** problems, or if the array is empty. */
4429 struct symbol **Symbol_arrayof()
4431 struct symbol **array;
4433 if( x2a==0 ) return 0;
4435 array = (struct symbol **)malloc( sizeof(struct symbol *)*size );
4437 for(i=0; i<size; i++) array[i] = x2a->tbl[i].data;
4442 /* Compare two configurations */
4448 x = a->rp->index - b->rp->index;
4449 if( x==0 ) x = a->dot - b->dot;
4453 /* Compare two states */
4454 PRIVATE int statecmp(a,b)
4459 for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){
4460 rc = a->rp->index - b->rp->index;
4461 if( rc==0 ) rc = a->dot - b->dot;
4471 PRIVATE int statehash(a)
4476 h = h*571 + a->rp->index*37 + a->dot;
4482 /* Allocate a new state structure */
4483 struct state *State_new()
4486 new = (struct state *)malloc( sizeof(struct state) );
4491 /* There is one instance of the following structure for each
4492 ** associative array of type "x3".
4495 int size; /* The number of available slots. */
4496 /* Must be a power of 2 greater than or */
4498 int count; /* Number of currently slots filled */
4499 struct s_x3node *tbl; /* The data stored here */
4500 struct s_x3node **ht; /* Hash table for lookups */
4503 /* There is one instance of this structure for every data element
4504 ** in an associative array of type "x3".
4506 typedef struct s_x3node {
4507 struct state *data; /* The data */
4508 struct config *key; /* The key */
4509 struct s_x3node *next; /* Next entry with the same hash */
4510 struct s_x3node **from; /* Previous link */
4513 /* There is only one instance of the array, which is the following */
4514 static struct s_x3 *x3a;
4516 /* Allocate a new associative array */
4519 x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
4523 x3a->tbl = (x3node*)malloc(
4524 (sizeof(x3node) + sizeof(x3node*))*128 );
4530 x3a->ht = (x3node**)&(x3a->tbl[128]);
4531 for(i=0; i<128; i++) x3a->ht[i] = 0;
4535 /* Insert a new record into the array. Return TRUE if successful.
4536 ** Prior data with the same key is NOT overwritten */
4537 int State_insert(data,key)
4545 if( x3a==0 ) return 0;
4546 ph = statehash(key);
4547 h = ph & (x3a->size-1);
4550 if( statecmp(np->key,key)==0 ){
4551 /* An existing entry with the same key is found. */
4552 /* Fail because overwrite is not allows. */
4557 if( x3a->count>=x3a->size ){
4558 /* Need to make the hash table bigger */
4561 array.size = size = x3a->size*2;
4562 array.count = x3a->count;
4563 array.tbl = (x3node*)malloc(
4564 (sizeof(x3node) + sizeof(x3node*))*size );
4565 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4566 array.ht = (x3node**)&(array.tbl[size]);
4567 for(i=0; i<size; i++) array.ht[i] = 0;
4568 for(i=0; i<x3a->count; i++){
4569 x3node *oldnp, *newnp;
4570 oldnp = &(x3a->tbl[i]);
4571 h = statehash(oldnp->key) & (size-1);
4572 newnp = &(array.tbl[i]);
4573 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4574 newnp->next = array.ht[h];
4575 newnp->key = oldnp->key;
4576 newnp->data = oldnp->data;
4577 newnp->from = &(array.ht[h]);
4578 array.ht[h] = newnp;
4583 /* Insert the new data */
4584 h = ph & (x3a->size-1);
4585 np = &(x3a->tbl[x3a->count++]);
4588 if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
4589 np->next = x3a->ht[h];
4591 np->from = &(x3a->ht[h]);
4595 /* Return a pointer to data assigned to the given key. Return NULL
4596 ** if no such key. */
4597 struct state *State_find(key)
4603 if( x3a==0 ) return 0;
4604 h = statehash(key) & (x3a->size-1);
4607 if( statecmp(np->key,key)==0 ) break;
4610 return np ? np->data : 0;
4613 /* Return an array of pointers to all data in the table.
4614 ** The array is obtained from malloc. Return NULL if memory allocation
4615 ** problems, or if the array is empty. */
4616 struct state **State_arrayof()
4618 struct state **array;
4620 if( x3a==0 ) return 0;
4622 array = (struct state **)malloc( sizeof(struct state *)*size );
4624 for(i=0; i<size; i++) array[i] = x3a->tbl[i].data;
4629 /* Hash a configuration */
4630 PRIVATE int confighash(a)
4634 h = h*571 + a->rp->index*37 + a->dot;
4638 /* There is one instance of the following structure for each
4639 ** associative array of type "x4".
4642 int size; /* The number of available slots. */
4643 /* Must be a power of 2 greater than or */
4645 int count; /* Number of currently slots filled */
4646 struct s_x4node *tbl; /* The data stored here */
4647 struct s_x4node **ht; /* Hash table for lookups */
4650 /* There is one instance of this structure for every data element
4651 ** in an associative array of type "x4".
4653 typedef struct s_x4node {
4654 struct config *data; /* The data */
4655 struct s_x4node *next; /* Next entry with the same hash */
4656 struct s_x4node **from; /* Previous link */
4659 /* There is only one instance of the array, which is the following */
4660 static struct s_x4 *x4a;
4662 /* Allocate a new associative array */
4663 void Configtable_init(){
4665 x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
4669 x4a->tbl = (x4node*)malloc(
4670 (sizeof(x4node) + sizeof(x4node*))*64 );
4676 x4a->ht = (x4node**)&(x4a->tbl[64]);
4677 for(i=0; i<64; i++) x4a->ht[i] = 0;
4681 /* Insert a new record into the array. Return TRUE if successful.
4682 ** Prior data with the same key is NOT overwritten */
4683 int Configtable_insert(data)
4684 struct config *data;
4690 if( x4a==0 ) return 0;
4691 ph = confighash(data);
4692 h = ph & (x4a->size-1);
4695 if( Configcmp(np->data,data)==0 ){
4696 /* An existing entry with the same key is found. */
4697 /* Fail because overwrite is not allows. */
4702 if( x4a->count>=x4a->size ){
4703 /* Need to make the hash table bigger */
4706 array.size = size = x4a->size*2;
4707 array.count = x4a->count;
4708 array.tbl = (x4node*)malloc(
4709 (sizeof(x4node) + sizeof(x4node*))*size );
4710 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4711 array.ht = (x4node**)&(array.tbl[size]);
4712 for(i=0; i<size; i++) array.ht[i] = 0;
4713 for(i=0; i<x4a->count; i++){
4714 x4node *oldnp, *newnp;
4715 oldnp = &(x4a->tbl[i]);
4716 h = confighash(oldnp->data) & (size-1);
4717 newnp = &(array.tbl[i]);
4718 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4719 newnp->next = array.ht[h];
4720 newnp->data = oldnp->data;
4721 newnp->from = &(array.ht[h]);
4722 array.ht[h] = newnp;
4727 /* Insert the new data */
4728 h = ph & (x4a->size-1);
4729 np = &(x4a->tbl[x4a->count++]);
4731 if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
4732 np->next = x4a->ht[h];
4734 np->from = &(x4a->ht[h]);
4738 /* Return a pointer to data assigned to the given key. Return NULL
4739 ** if no such key. */
4740 struct config *Configtable_find(key)
4746 if( x4a==0 ) return 0;
4747 h = confighash(key) & (x4a->size-1);
4750 if( Configcmp(np->data,key)==0 ) break;
4753 return np ? np->data : 0;
4756 /* Remove all data from the table. Pass each data to the function "f"
4757 ** as it is removed. ("f" may be null to avoid this step.) */
4758 void Configtable_clear(f)
4759 int(*f)(/* struct config * */);
4762 if( x4a==0 || x4a->count==0 ) return;
4763 if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
4764 for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;