4 ** This file contains all sources (including headers) to the LEMON
5 ** LALR(1) parser generator. The sources have been combined into a
6 ** single file to make it easy to include LEMON in the source tree
7 ** and Makefile of another program.
9 ** The author of this program disclaims copyright.
20 #ifdef HAVE_INTTYPES_H
21 # include <inttypes.h>
24 #define UNUSED(x) ( (void)(x) )
27 extern double strtod();
32 extern char *getenv();
35 # if defined(_WIN32) || defined(WIN32)
41 #define NORETURN __attribute__ ((__noreturn__))
46 /* #define PRIVATE static */
47 #define PRIVATE static
50 #define MAXRHS 5 /* Set low to exercise exception code */
55 void *msort(void *list
, void **next
, int(*cmp
)(void *, void *));
57 extern void memory_error() NORETURN
;
59 /******** From the file "action.h" *************************************/
60 struct action
*Action_new();
61 struct action
*Action_sort();
64 /********* From the file "assert.h" ************************************/
65 void myassert() NORETURN
;
67 # define assert(X) if(!(X))myassert(__FILE__,__LINE__)
72 /********** From the file "build.h" ************************************/
73 void FindRulePrecedences();
77 void FindFollowSets();
80 /********* From the file "configlist.h" *********************************/
81 void Configlist_init(/* void */);
82 struct config
*Configlist_add(/* struct rule *, int */);
83 struct config
*Configlist_addbasis(/* struct rule *, int */);
84 void Configlist_closure(/* void */);
85 void Configlist_sort(/* void */);
86 void Configlist_sortbasis(/* void */);
87 struct config
*Configlist_return(/* void */);
88 struct config
*Configlist_basis(/* void */);
89 void Configlist_eat(/* struct config * */);
90 void Configlist_reset(/* void */);
92 /********* From the file "error.h" ***************************************/
93 void ErrorMsg(const char *, int,const char *, ...);
95 /****** From the file "option.h" ******************************************/
97 enum { OPT_FLAG
=1, OPT_INT
, OPT_DBL
, OPT_STR
,
98 OPT_FFLAG
, OPT_FINT
, OPT_FDBL
, OPT_FSTR
} type
;
103 int OptInit(/* char**,struct s_options*,FILE* */);
104 int OptNArgs(/* void */);
105 char *OptArg(/* int */);
106 void OptErr(/* int */);
107 void OptPrint(/* void */);
109 /******** From the file "parse.h" *****************************************/
110 void Parse(/* struct lemon *lemp */);
112 /********* From the file "plink.h" ***************************************/
113 struct plink
*Plink_new(/* void */);
114 void Plink_add(/* struct plink **, struct config * */);
115 void Plink_copy(/* struct plink **, struct plink * */);
116 void Plink_delete(/* struct plink * */);
118 /********** From the file "report.h" *************************************/
119 void Reprint(/* struct lemon * */);
120 void ReportOutput(/* struct lemon * */);
121 void ReportTable(/* struct lemon * */);
122 void ReportHeader(/* struct lemon * */);
123 void CompressTables(/* struct lemon * */);
125 /********** From the file "set.h" ****************************************/
126 void SetSize(/* int N */); /* All sets will be of size N */
127 char *SetNew(/* void */); /* A new set for element 0..N */
128 void SetFree(/* char* */); /* Deallocate a set */
130 int SetAdd(/* char*,int */); /* Add element to a set */
131 int SetUnion(/* char *A,char *B */); /* A <- A U B, thru element N */
133 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
135 /********** From the file "struct.h" *************************************/
137 ** Principal data structures for the LEMON parser generator.
140 typedef enum {Bo_FALSE
=0, Bo_TRUE
} Boolean
;
142 /* Symbols (terminals and nonterminals) of the grammar are stored
143 ** in the following: */
145 char *name
; /* Name of the symbol */
146 int index
; /* Index number for this symbol */
150 } type
; /* Symbols are all either TERMINALS or NTs */
151 struct rule
*rule
; /* Linked list of rules of this (if an NT) */
152 struct symbol
*fallback
; /* fallback token in case this token doesn't parse */
153 int prec
; /* Precedence if defined (-1 otherwise) */
159 } assoc
; /* Associativity if predecence is defined */
160 char *firstset
; /* First-set for all rules of this symbol */
161 Boolean lambda
; /* True if NT and can generate an empty string */
162 char *destructor
; /* Code which executes whenever this symbol is
163 ** popped from the stack during error processing */
164 int destructorln
; /* Line number of destructor code */
165 char *datatype
; /* The data type of information held by this
166 ** object. Only used if type==NONTERMINAL */
167 int dtnum
; /* The data type number. In the parser, the value
168 ** stack is a union. The .yy%d element of this
169 ** union is the correct data type for this object */
172 /* Each production rule in the grammar is stored in the following
175 struct symbol
*lhs
; /* Left-hand side of the rule */
176 char *lhsalias
; /* Alias for the LHS (NULL if none) */
177 int ruleline
; /* Line number for the rule */
178 int nrhs
; /* Number of RHS symbols */
179 struct symbol
**rhs
; /* The RHS symbols */
180 char **rhsalias
; /* An alias for each RHS symbol (NULL if none) */
181 int line
; /* Line number at which code begins */
182 char *code
; /* The code executed when this rule is reduced */
183 struct symbol
*precsym
; /* Precedence symbol for this rule */
184 int index
; /* An index number for this rule */
185 Boolean canReduce
; /* True if this rule is ever reduced */
186 struct rule
*nextlhs
; /* Next rule with the same LHS */
187 struct rule
*next
; /* Next rule in the global list */
190 /* A configuration is a production rule of the grammar together with
191 ** a mark (dot) showing how much of that rule has been processed so far.
192 ** Configurations also contain a follow-set which is a list of terminal
193 ** symbols which are allowed to immediately follow the end of the rule.
194 ** Every configuration is recorded as an instance of the following: */
196 struct rule
*rp
; /* The rule upon which the configuration is based */
197 int dot
; /* The parse point */
198 char *fws
; /* Follow-set for this configuration only */
199 struct plink
*fplp
; /* Follow-set forward propagation links */
200 struct plink
*bplp
; /* Follow-set backwards propagation links */
201 struct state
*stp
; /* Pointer to state which contains this */
203 COMPLETE
, /* The status is used during followset and */
204 INCOMPLETE
/* shift computations */
206 struct config
*next
; /* Next configuration in the state */
207 struct config
*bp
; /* The next basis configuration */
210 /* Every shift or reduce operation is stored as one of the following */
212 struct symbol
*sp
; /* The look-ahead symbol */
218 CONFLICT
, /* Was a reduce, but part of a conflict */
219 SH_RESOLVED
, /* Was a shift. Precedence resolved conflict */
220 RD_RESOLVED
, /* Was reduce. Precedence resolved conflict */
221 NOT_USED
/* Deleted by compression */
224 struct state
*stp
; /* The new state, if a shift */
225 struct rule
*rp
; /* The rule, if a reduce */
227 struct action
*next
; /* Next action for this state */
228 struct action
*collide
; /* Next action with the same hash */
231 /* Each state of the generated parser's finite state machine
232 ** is encoded as an instance of the following structure. */
234 struct config
*bp
; /* The basis configurations for this state */
235 struct config
*cfp
; /* All configurations in this set */
236 int index
; /* Sequencial number for this state */
237 struct action
*ap
; /* Array of actions for this state */
238 int nTknAct
, nNtAct
; /* Number of actions on terminals and nonterminals */
239 int iTknOfst
, iNtOfst
; /* yy_action[] offset for terminals and nonterms */
240 int iDflt
; /* Default action */
242 #define NO_OFFSET (-2147483647)
244 /* A followset propagation link indicates that the contents of one
245 ** configuration followset should be propagated to another whenever
246 ** the first changes. */
248 struct config
*cfp
; /* The configuration to which linked */
249 struct plink
*next
; /* The next propagate link */
252 /* The state vector for the entire parser generator is recorded as
253 ** follows. (LEMON uses no global variables and makes little use of
254 ** static variables. Fields in the following structure can be thought
255 ** of as begin global variables in the program.) */
257 struct state
**sorted
; /* Table of states sorted by state number */
258 struct rule
*rule
; /* List of all rules */
259 int nstate
; /* Number of states */
260 int nrule
; /* Number of rules */
261 int nsymbol
; /* Number of terminal and nonterminal symbols */
262 int nterminal
; /* Number of terminal symbols */
263 struct symbol
**symbols
; /* Sorted array of pointers to symbols */
264 int errorcnt
; /* Number of errors */
265 struct symbol
*errsym
; /* The error symbol */
266 char *name
; /* Name of the generated parser */
267 char *arg
; /* Declaration of the 3th argument to parser */
268 char *tokentype
; /* Type of terminal symbols in the parser stack */
269 char *vartype
; /* The default type of non-terminal symbols */
270 char *start
; /* Name of the start symbol for the grammar */
271 char *stacksize
; /* Size of the parser stack */
272 char *include
; /* Code to put at the start of the C file */
273 int includeln
; /* Line number for start of include code */
274 char *error
; /* Code to execute when an error is seen */
275 int errorln
; /* Line number for start of error code */
276 char *overflow
; /* Code to execute on a stack overflow */
277 int overflowln
; /* Line number for start of overflow code */
278 char *failure
; /* Code to execute on parser failure */
279 int failureln
; /* Line number for start of failure code */
280 char *accept
; /* Code to execute when the parser excepts */
281 int acceptln
; /* Line number for the start of accept code */
282 char *extracode
; /* Code appended to the generated file */
283 int extracodeln
; /* Line number for the start of the extra code */
284 char *tokendest
; /* Code to execute to destroy token data */
285 int tokendestln
; /* Line number for token destroyer code */
286 char *vardest
; /* Code for the default non-terminal destructor */
287 int vardestln
; /* Line number for default non-term destructor code*/
288 char *filename
; /* Name of the input file */
289 char *tmplname
; /* Name of the template file */
290 char *outname
; /* Name of the current output file */
291 char *tokenprefix
; /* A prefix added to token names in the .h file */
292 int nconflict
; /* Number of parsing conflicts */
293 int tablesize
; /* Size of the parse tables */
294 int basisflag
; /* Print only basis configurations */
295 int has_fallback
; /* True if any %fallback is seen in the grammer */
296 char *argv0
; /* Name of the program */
299 #define MemoryCheck(X) if((X)==0){ \
303 /**************** From the file "table.h" *********************************/
305 ** All code in this file has been automatically generated
306 ** from a specification in the file
308 ** by the associative array code building program "aagen".
309 ** Do not edit this file! Instead, edit the specification
310 ** file, then rerun aagen.
313 ** Code for processing tables in the LEMON parser generator.
316 /* Routines for handling a strings */
320 void Strsafe_init(/* void */);
321 int Strsafe_insert(/* char * */);
322 char *Strsafe_find(/* char * */);
324 /* Routines for handling symbols of the grammar */
326 struct symbol
*Symbol_new();
327 int Symbolcmpp(/* struct symbol **, struct symbol ** */);
328 void Symbol_init(/* void */);
329 int Symbol_insert(/* struct symbol *, char * */);
330 struct symbol
*Symbol_find(/* char * */);
331 struct symbol
*Symbol_Nth(/* int */);
332 int Symbol_count(/* */);
333 int State_count(void);
334 struct symbol
**Symbol_arrayof(/* */);
336 /* Routines to manage the state table */
338 int Configcmp(/* struct config *, struct config * */);
339 struct state
*State_new();
340 void State_init(/* void */);
341 int State_insert(/* struct state *, struct config * */);
342 struct state
*State_find(/* struct config * */);
343 struct state
**State_arrayof(/* */);
345 /* Routines used for efficiency in Configlist_add */
347 void Configtable_init(/* void */);
348 int Configtable_insert(/* struct config * */);
349 struct config
*Configtable_find(/* struct config * */);
350 void Configtable_clear(/* int(*)(struct config *) */);
351 /****************** From the file "action.c" *******************************/
353 ** Routines processing parser actions in the LEMON parser generator.
356 /* Allocate a new parser action */
357 struct action
*Action_new(){
358 static struct action
*freelist
= NULL
;
361 if( freelist
==NULL
){
364 freelist
= (struct action
*)malloc( sizeof(struct action
)*amt
);
366 fprintf(stderr
,"Unable to allocate memory for a new parser action.");
369 for(i
=0; i
<amt
-1; i
++) freelist
[i
].next
= &freelist
[i
+1];
370 freelist
[amt
-1].next
= 0;
373 freelist
= freelist
->next
;
377 /* Compare two actions */
378 static int actioncmp(ap1
,ap2
)
383 rc
= ap1
->sp
->index
- ap2
->sp
->index
;
384 if( rc
==0 ) rc
= (int)ap1
->type
- (int)ap2
->type
;
386 assert( ap1
->type
==REDUCE
|| ap1
->type
==RD_RESOLVED
|| ap1
->type
==CONFLICT
);
387 assert( ap2
->type
==REDUCE
|| ap2
->type
==RD_RESOLVED
|| ap2
->type
==CONFLICT
);
388 rc
= ap1
->x
.rp
->index
- ap2
->x
.rp
->index
;
393 /* Sort parser actions */
394 struct action
*Action_sort(ap
)
397 ap
= (struct action
*)msort(ap
,(void **)&ap
->next
,actioncmp
);
401 void Action_add(app
,type
,sp
,arg
)
414 new->x
.stp
= (struct state
*)arg
;
416 new->x
.rp
= (struct rule
*)arg
;
419 /********************** New code to implement the "acttab" module ***********/
421 ** This module implements routines use to construct the yy_action[] table.
425 ** The state of the yy_action table under construction is an instance of
426 ** the following structure
428 typedef struct acttab acttab
;
430 int nAction
; /* Number of used slots in aAction[] */
431 int nActionAlloc
; /* Slots allocated for aAction[] */
433 int lookahead
; /* Value of the lookahead token */
434 int action
; /* Action to take on the given lookahead */
435 } *aAction
, /* The yy_action[] table under construction */
436 *aLookahead
; /* A single new transaction set */
437 int mnLookahead
; /* Minimum aLookahead[].lookahead */
438 int mnAction
; /* Action associated with mnLookahead */
439 int mxLookahead
; /* Maximum aLookahead[].lookahead */
440 int nLookahead
; /* Used slots in aLookahead[] */
441 int nLookaheadAlloc
; /* Slots allocated in aLookahead[] */
444 /* Return the number of entries in the yy_action table */
445 #define acttab_size(X) ((X)->nAction)
447 /* The value for the N-th entry in yy_action */
448 #define acttab_yyaction(X,N) ((X)->aAction[N].action)
450 /* The value for the N-th entry in yy_lookahead */
451 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
453 /* Free all memory associated with the given acttab */
455 PRIVATE void acttab_free(acttab *p){
457 free( p->aLookahead );
462 /* Allocate a new acttab structure */
463 PRIVATE acttab
*acttab_alloc(void){
464 acttab
*p
= malloc( sizeof(*p
) );
466 fprintf(stderr
,"Unable to allocate memory for a new acttab.");
469 memset(p
, 0, sizeof(*p
));
473 /* Add a new action to the current transaction set
475 PRIVATE
void acttab_action(acttab
*p
, int lookahead
, int action
){
476 if( p
->nLookahead
>=p
->nLookaheadAlloc
){
477 p
->nLookaheadAlloc
+= 25;
478 p
->aLookahead
= realloc( p
->aLookahead
,
479 sizeof(p
->aLookahead
[0])*p
->nLookaheadAlloc
);
480 if( p
->aLookahead
==0 ){
481 fprintf(stderr
,"malloc failed\n");
485 if( p
->nLookahead
==0 ){
486 p
->mxLookahead
= lookahead
;
487 p
->mnLookahead
= lookahead
;
488 p
->mnAction
= action
;
490 if( p
->mxLookahead
<lookahead
) p
->mxLookahead
= lookahead
;
491 if( p
->mnLookahead
>lookahead
){
492 p
->mnLookahead
= lookahead
;
493 p
->mnAction
= action
;
496 p
->aLookahead
[p
->nLookahead
].lookahead
= lookahead
;
497 p
->aLookahead
[p
->nLookahead
].action
= action
;
502 ** Add the transaction set built up with prior calls to acttab_action()
503 ** into the current action table. Then reset the transaction set back
504 ** to an empty set in preparation for a new round of acttab_action() calls.
506 ** Return the offset into the action table of the new transaction.
508 PRIVATE
int acttab_insert(acttab
*p
){
510 assert( p
->nLookahead
>0 );
512 /* Make sure we have enough space to hold the expanded action table
513 ** in the worst case. The worst case occurs if the transaction set
514 ** must be appended to the current action table
516 n
= p
->mxLookahead
+ 1;
517 if( p
->nAction
+ n
>= p
->nActionAlloc
){
518 int oldAlloc
= p
->nActionAlloc
;
519 p
->nActionAlloc
= p
->nAction
+ n
+ p
->nActionAlloc
+ 20;
520 p
->aAction
= realloc( p
->aAction
,
521 sizeof(p
->aAction
[0])*p
->nActionAlloc
);
523 fprintf(stderr
,"malloc failed\n");
526 for(i
=oldAlloc
; i
<p
->nActionAlloc
; i
++){
527 p
->aAction
[i
].lookahead
= -1;
528 p
->aAction
[i
].action
= -1;
532 /* Scan the existing action table looking for an offset where we can
533 ** insert the current transaction set. Fall out of the loop when that
534 ** offset is found. In the worst case, we fall out of the loop when
535 ** i reaches p->nAction, which means we append the new transaction set.
537 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
539 for(i
=0; i
<p
->nAction
+p
->mnLookahead
; i
++){
540 if( p
->aAction
[i
].lookahead
<0 ){
541 for(j
=0; j
<p
->nLookahead
; j
++){
542 k
= p
->aLookahead
[j
].lookahead
- p
->mnLookahead
+ i
;
544 if( p
->aAction
[k
].lookahead
>=0 ) break;
546 if( j
<p
->nLookahead
) continue;
547 for(j
=0; j
<p
->nAction
; j
++){
548 if( p
->aAction
[j
].lookahead
==j
+p
->mnLookahead
-i
) break;
551 break; /* Fits in empty slots */
553 }else if( p
->aAction
[i
].lookahead
==p
->mnLookahead
){
554 if( p
->aAction
[i
].action
!=p
->mnAction
) continue;
555 for(j
=0; j
<p
->nLookahead
; j
++){
556 k
= p
->aLookahead
[j
].lookahead
- p
->mnLookahead
+ i
;
557 if( k
<0 || k
>=p
->nAction
) break;
558 if( p
->aLookahead
[j
].lookahead
!=p
->aAction
[k
].lookahead
) break;
559 if( p
->aLookahead
[j
].action
!=p
->aAction
[k
].action
) break;
561 if( j
<p
->nLookahead
) continue;
563 for(j
=0; j
<p
->nAction
; j
++){
564 if( p
->aAction
[j
].lookahead
<0 ) continue;
565 if( p
->aAction
[j
].lookahead
==j
+p
->mnLookahead
-i
) n
++;
567 if( n
==p
->nLookahead
){
568 break; /* Same as a prior transaction set */
572 /* Insert transaction set at index i. */
573 for(j
=0; j
<p
->nLookahead
; j
++){
574 k
= p
->aLookahead
[j
].lookahead
- p
->mnLookahead
+ i
;
575 p
->aAction
[k
] = p
->aLookahead
[j
];
576 if( k
>=p
->nAction
) p
->nAction
= k
+1;
580 /* Return the offset that is added to the lookahead in order to get the
581 ** index into yy_action of the action */
582 return i
- p
->mnLookahead
;
585 /********************** From the file "assert.c" ****************************/
587 ** A more efficient way of handling assertions.
589 void myassert(file
,line
)
593 fprintf(stderr
,"Assertion failed on line %d of file \"%s\"\n",line
,file
);
596 /********************** From the file "build.c" *****************************/
598 ** Routines to construction the finite state machine for the LEMON
602 /* Find a precedence symbol of every rule in the grammar.
604 ** Those rules which have a precedence symbol coded in the input
605 ** grammar using the "[symbol]" construct will already have the
606 ** rp->precsym field filled. Other rules take as their precedence
607 ** symbol the first RHS symbol with a defined precedence. If there
608 ** are not RHS symbols with a defined precedence, the precedence
609 ** symbol field is left blank.
611 void FindRulePrecedences(xp
)
615 for(rp
=xp
->rule
; rp
; rp
=rp
->next
){
616 if( rp
->precsym
==0 ){
618 for(i
=0; i
<rp
->nrhs
; i
++){
619 if( rp
->rhs
[i
]->prec
>=0 ){
620 rp
->precsym
= rp
->rhs
[i
];
629 /* Find all nonterminals which will generate the empty string.
630 ** Then go back and compute the first sets of every nonterminal.
631 ** The first set is the set of all terminal symbols which can begin
632 ** a string generated by that nonterminal.
634 void FindFirstSets(lemp
)
641 for(i
=0; i
<lemp
->nsymbol
; i
++){
642 lemp
->symbols
[i
]->lambda
= Bo_FALSE
;
644 for(i
=lemp
->nterminal
; i
<lemp
->nsymbol
; i
++){
645 lemp
->symbols
[i
]->firstset
= SetNew();
648 /* First compute all lambdas */
651 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
652 if( rp
->lhs
->lambda
) continue;
653 for(i
=0; i
<rp
->nrhs
; i
++){
654 if( rp
->rhs
[i
]->lambda
==Bo_FALSE
) break;
657 rp
->lhs
->lambda
= Bo_TRUE
;
663 /* Now compute all first sets */
665 struct symbol
*s1
, *s2
;
667 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
669 for(i
=0; i
<rp
->nrhs
; i
++){
671 if( s2
->type
==TERMINAL
){
672 progress
+= SetAdd(s1
->firstset
,s2
->index
);
675 if( s1
->lambda
==Bo_FALSE
) break;
677 progress
+= SetUnion(s1
->firstset
,s2
->firstset
);
678 if( s2
->lambda
==Bo_FALSE
) break;
686 /* Compute all LR(0) states for the grammar. Links
687 ** are added to between some states so that the LR(1) follow sets
688 ** can be computed later.
690 PRIVATE
struct state
*getstate(/* struct lemon * */); /* forward reference */
691 void FindStates(lemp
)
699 /* Find the start symbol */
701 sp
= Symbol_find(lemp
->start
);
703 ErrorMsg(lemp
->filename
,0,
704 "The specified start symbol \"%s\" is not \
705 in a nonterminal of the grammar. \"%s\" will be used as the start \
706 symbol instead.",lemp
->start
,lemp
->rule
->lhs
->name
);
708 sp
= lemp
->rule
->lhs
;
711 sp
= lemp
->rule
->lhs
;
714 /* Make sure the start symbol doesn't occur on the right-hand side of
715 ** any rule. Report an error if it does. (YACC would generate a new
716 ** start symbol in this case.) */
717 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
719 for(i
=0; i
<rp
->nrhs
; i
++){
720 if( rp
->rhs
[i
]==sp
){
721 ErrorMsg(lemp
->filename
,0,
722 "The start symbol \"%s\" occurs on the \
723 right-hand side of a rule. This will result in a parser which \
724 does not work properly.",sp
->name
);
730 /* The basis configuration set for the first state
731 ** is all rules which have the start symbol as their
733 for(rp
=sp
->rule
; rp
; rp
=rp
->nextlhs
){
734 struct config
*newcfp
;
735 newcfp
= Configlist_addbasis(rp
,0);
736 SetAdd(newcfp
->fws
,0);
739 /* Compute the first state. All other states will be
740 ** computed automatically during the computation of the first one.
741 ** The returned pointer to the first state is not used. */
742 (void)getstate(lemp
);
746 /* Return a pointer to a state which is described by the configuration
747 ** list which has been built from calls to Configlist_add.
749 PRIVATE
void buildshifts(/* struct lemon *, struct state * */); /* Forwd ref */
750 PRIVATE
struct state
*getstate(lemp
)
753 struct config
*cfp
, *bp
;
756 /* Extract the sorted basis of the new state. The basis was constructed
757 ** by prior calls to "Configlist_addbasis()". */
758 Configlist_sortbasis();
759 bp
= Configlist_basis();
761 /* Get a state with the same basis */
762 stp
= State_find(bp
);
764 /* A state with the same basis already exists! Copy all the follow-set
765 ** propagation links from the state under construction into the
766 ** preexisting state, then return a pointer to the preexisting state */
767 struct config
*x
, *y
;
768 for(x
=bp
, y
=stp
->bp
; x
&& y
; x
=x
->bp
, y
=y
->bp
){
769 Plink_copy(&y
->bplp
,x
->bplp
);
770 Plink_delete(x
->fplp
);
771 x
->fplp
= x
->bplp
= 0;
773 cfp
= Configlist_return();
776 /* This really is a new state. Construct all the details */
777 Configlist_closure(lemp
); /* Compute the configuration closure */
778 Configlist_sort(); /* Sort the configuration closure */
779 cfp
= Configlist_return(); /* Get a pointer to the config list */
780 stp
= State_new(); /* A new state structure */
782 stp
->bp
= bp
; /* Remember the configuration basis */
783 stp
->cfp
= cfp
; /* Remember the configuration closure */
784 stp
->index
= lemp
->nstate
++; /* Every state gets a sequence number */
785 stp
->ap
= 0; /* No actions, yet. */
786 State_insert(stp
,stp
->bp
); /* Add to the state table */
787 buildshifts(lemp
,stp
); /* Recursively compute successor states */
792 /* Construct all successor states to the given state. A "successor"
793 ** state is any state which can be reached by a shift action.
795 PRIVATE
void buildshifts(lemp
,stp
)
797 struct state
*stp
; /* The state from which successors are computed */
799 struct config
*cfp
; /* For looping thru the config closure of "stp" */
800 struct config
*bcfp
; /* For the inner loop on config closure of "stp" */
801 struct config
*new; /* */
802 struct symbol
*sp
; /* Symbol following the dot in configuration "cfp" */
803 struct symbol
*bsp
; /* Symbol following the dot in configuration "bcfp" */
804 struct state
*newstp
; /* A pointer to a successor state */
806 /* Each configuration becomes complete after it contibutes to a successor
807 ** state. Initially, all configurations are incomplete */
808 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
) cfp
->status
= INCOMPLETE
;
810 /* Loop through all configurations of the state "stp" */
811 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
812 if( cfp
->status
==COMPLETE
) continue; /* Already used by inner loop */
813 if( cfp
->dot
>=cfp
->rp
->nrhs
) continue; /* Can't shift this config */
814 Configlist_reset(); /* Reset the new config set */
815 sp
= cfp
->rp
->rhs
[cfp
->dot
]; /* Symbol after the dot */
817 /* For every configuration in the state "stp" which has the symbol "sp"
818 ** following its dot, add the same configuration to the basis set under
819 ** construction but with the dot shifted one symbol to the right. */
820 for(bcfp
=cfp
; bcfp
; bcfp
=bcfp
->next
){
821 if( bcfp
->status
==COMPLETE
) continue; /* Already used */
822 if( bcfp
->dot
>=bcfp
->rp
->nrhs
) continue; /* Can't shift this one */
823 bsp
= bcfp
->rp
->rhs
[bcfp
->dot
]; /* Get symbol after dot */
824 if( bsp
!=sp
) continue; /* Must be same as for "cfp" */
825 bcfp
->status
= COMPLETE
; /* Mark this config as used */
826 new = Configlist_addbasis(bcfp
->rp
,bcfp
->dot
+1);
827 Plink_add(&new->bplp
,bcfp
);
830 /* Get a pointer to the state described by the basis configuration set
831 ** constructed in the preceding loop */
832 newstp
= getstate(lemp
);
834 /* The state "newstp" is reached from the state "stp" by a shift action
835 ** on the symbol "sp" */
836 Action_add(&stp
->ap
,SHIFT
,sp
,newstp
);
841 ** Construct the propagation links
847 struct config
*cfp
, *other
;
851 /* Housekeeping detail:
852 ** Add to every propagate link a pointer back to the state to
853 ** which the link is attached. */
854 for(i
=0; i
<lemp
->nstate
; i
++){
855 stp
= lemp
->sorted
[i
];
856 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
861 /* Convert all backlinks into forward links. Only the forward
862 ** links are used in the follow-set computation. */
863 for(i
=0; i
<lemp
->nstate
; i
++){
864 stp
= lemp
->sorted
[i
];
865 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
866 for(plp
=cfp
->bplp
; plp
; plp
=plp
->next
){
868 Plink_add(&other
->fplp
,cfp
);
874 /* Compute all followsets.
876 ** A followset is the set of all symbols which can come immediately
877 ** after a configuration.
879 void FindFollowSets(lemp
)
889 for(i
=0; i
<lemp
->nstate
; i
++){
890 stp
= lemp
->sorted
[i
];
891 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
892 cfp
->status
= INCOMPLETE
;
898 for(i
=0; i
<lemp
->nstate
; i
++){
899 stp
= lemp
->sorted
[i
];
900 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
901 if( cfp
->status
==COMPLETE
) continue;
902 for(plp
=cfp
->fplp
; plp
; plp
=plp
->next
){
903 change
= SetUnion(plp
->cfp
->fws
,cfp
->fws
);
905 plp
->cfp
->status
= INCOMPLETE
;
909 cfp
->status
= COMPLETE
;
915 static int resolve_conflict();
917 /* Compute the reduce actions, and resolve conflicts.
919 void FindActions(lemp
)
927 /* Add all of the reduce actions
928 ** A reduce action is added for each element of the followset of
929 ** a configuration which has its dot at the extreme right.
931 for(i
=0; i
<lemp
->nstate
; i
++){ /* Loop over all states */
933 stp
= lemp
->sorted
[i
];
934 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){ /* Loop over all configurations */
935 if( cfp
->rp
->nrhs
==cfp
->dot
){ /* Is dot at extreme right? */
936 for(j
=0; j
<lemp
->nterminal
; j
++){
937 if( SetFind(cfp
->fws
,j
) ){
938 /* Add a reduce action to the state "stp" which will reduce by the
939 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
940 Action_add(&stp
->ap
,REDUCE
,lemp
->symbols
[j
],cfp
->rp
);
947 /* Add the accepting token */
949 sp
= Symbol_find(lemp
->start
);
950 if( sp
==0 ) sp
= lemp
->rule
->lhs
;
952 sp
= lemp
->rule
->lhs
;
954 /* Add to the first state (which is always the starting state of the
955 ** finite state machine) an action to ACCEPT if the lookahead is the
956 ** start nonterminal. */
957 if (lemp
->nstate
) { /*(should always be true)*/
959 stp
= lemp
->sorted
[0];
960 Action_add(&stp
->ap
,ACCEPT
,sp
,0);
963 /* Resolve conflicts */
964 for(i
=0; i
<lemp
->nstate
; i
++){
965 struct action
*ap
, *nap
;
967 stp
= lemp
->sorted
[i
];
969 stp
->ap
= Action_sort(stp
->ap
);
970 for(ap
=stp
->ap
; ap
&& ap
->next
; ap
=ap
->next
){
971 for(nap
=ap
->next
; nap
&& nap
->sp
==ap
->sp
; nap
=nap
->next
){
972 /* The two actions "ap" and "nap" have the same lookahead.
973 ** Figure out which one should be used */
974 lemp
->nconflict
+= resolve_conflict(ap
,nap
,lemp
->errsym
);
979 /* Report an error for each rule that can never be reduced. */
980 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
) rp
->canReduce
= Bo_FALSE
;
981 for(i
=0; i
<lemp
->nstate
; i
++){
983 for(ap
=lemp
->sorted
[i
]->ap
; ap
; ap
=ap
->next
){
984 if( ap
->type
==REDUCE
) ap
->x
.rp
->canReduce
= Bo_TRUE
;
987 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
988 if( rp
->canReduce
) continue;
989 ErrorMsg(lemp
->filename
,rp
->ruleline
,"This rule can not be reduced.\n");
994 /* Resolve a conflict between the two given actions. If the
995 ** conflict can't be resolve, return non-zero.
998 ** To resolve a conflict, first look to see if either action
999 ** is on an error rule. In that case, take the action which
1000 ** is not associated with the error rule. If neither or both
1001 ** actions are associated with an error rule, then try to
1002 ** use precedence to resolve the conflict.
1004 ** If either action is a SHIFT, then it must be apx. This
1005 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1007 static int resolve_conflict(apx
,apy
,errsym
)
1010 struct symbol
*errsym
; /* The error symbol (if defined. NULL otherwise) */
1012 struct symbol
*spx
, *spy
;
1015 assert( apx
->sp
==apy
->sp
); /* Otherwise there would be no conflict */
1016 if( apx
->type
==SHIFT
&& apy
->type
==REDUCE
){
1018 spy
= apy
->x
.rp
->precsym
;
1019 if( spy
==0 || spx
->prec
<0 || spy
->prec
<0 ){
1020 /* Not enough precedence information. */
1021 apy
->type
= CONFLICT
;
1023 }else if( spx
->prec
>spy
->prec
){ /* Lower precedence wins */
1024 apy
->type
= RD_RESOLVED
;
1025 }else if( spx
->prec
<spy
->prec
){
1026 apx
->type
= SH_RESOLVED
;
1027 }else if( spx
->prec
==spy
->prec
&& spx
->assoc
==RIGHT
){ /* Use operator */
1028 apy
->type
= RD_RESOLVED
; /* associativity */
1029 }else if( spx
->prec
==spy
->prec
&& spx
->assoc
==LEFT
){ /* to break tie */
1030 apx
->type
= SH_RESOLVED
;
1032 assert( spx
->prec
==spy
->prec
&& spx
->assoc
==NONE
);
1033 apy
->type
= CONFLICT
;
1036 }else if( apx
->type
==REDUCE
&& apy
->type
==REDUCE
){
1037 spx
= apx
->x
.rp
->precsym
;
1038 spy
= apy
->x
.rp
->precsym
;
1039 if( spx
==0 || spy
==0 || spx
->prec
<0 ||
1040 spy
->prec
<0 || spx
->prec
==spy
->prec
){
1041 apy
->type
= CONFLICT
;
1043 }else if( spx
->prec
>spy
->prec
){
1044 apy
->type
= RD_RESOLVED
;
1045 }else if( spx
->prec
<spy
->prec
){
1046 apx
->type
= RD_RESOLVED
;
1050 apx
->type
==SH_RESOLVED
||
1051 apx
->type
==RD_RESOLVED
||
1052 apx
->type
==CONFLICT
||
1053 apy
->type
==SH_RESOLVED
||
1054 apy
->type
==RD_RESOLVED
||
1057 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1058 ** REDUCEs on the list. If we reach this point it must be because
1059 ** the parser conflict had already been resolved. */
1063 /********************* From the file "configlist.c" *************************/
1065 ** Routines to processing a configuration list and building a state
1066 ** in the LEMON parser generator.
1069 static struct config
*freelist
= 0; /* List of free configurations */
1070 static struct config
*current
= 0; /* Top of list of configurations */
1071 static struct config
**currentend
= 0; /* Last on list of configs */
1072 static struct config
*basis
= 0; /* Top of list of basis configs */
1073 static struct config
**basisend
= 0; /* End of list of basis configs */
1075 /* Return a pointer to a new configuration */
1076 PRIVATE
struct config
*newconfig(){
1081 freelist
= (struct config
*)malloc( sizeof(struct config
)*amt
);
1083 fprintf(stderr
,"Unable to allocate memory for a new configuration.");
1086 for(i
=0; i
<amt
-1; i
++) freelist
[i
].next
= &freelist
[i
+1];
1087 freelist
[amt
-1].next
= 0;
1090 freelist
= freelist
->next
;
1094 /* The configuration "old" is no longer used */
1095 PRIVATE
void deleteconfig(old
)
1098 old
->next
= freelist
;
1102 /* Initialized the configuration list builder */
1103 void Configlist_init(){
1105 currentend
= ¤t
;
1112 /* Initialized the configuration list builder */
1113 void Configlist_reset(){
1115 currentend
= ¤t
;
1118 Configtable_clear(0);
1122 /* Add another configuration to the configuration list */
1123 struct config
*Configlist_add(rp
,dot
)
1124 struct rule
*rp
; /* The rule */
1125 int dot
; /* Index into the RHS of the rule where the dot goes */
1127 struct config
*cfp
, model
;
1129 assert( currentend
!=0 );
1132 cfp
= Configtable_find(&model
);
1137 cfp
->fws
= SetNew();
1139 cfp
->fplp
= cfp
->bplp
= 0;
1143 currentend
= &cfp
->next
;
1144 Configtable_insert(cfp
);
1149 /* Add a basis configuration to the configuration list */
1150 struct config
*Configlist_addbasis(rp
,dot
)
1154 struct config
*cfp
, model
;
1156 assert( basisend
!=0 );
1157 assert( currentend
!=0 );
1160 cfp
= Configtable_find(&model
);
1165 cfp
->fws
= SetNew();
1167 cfp
->fplp
= cfp
->bplp
= 0;
1171 currentend
= &cfp
->next
;
1173 basisend
= &cfp
->bp
;
1174 Configtable_insert(cfp
);
1179 /* Compute the closure of the configuration list */
1180 void Configlist_closure(lemp
)
1183 struct config
*cfp
, *newcfp
;
1184 struct rule
*rp
, *newrp
;
1185 struct symbol
*sp
, *xsp
;
1188 assert( currentend
!=0 );
1189 for(cfp
=current
; cfp
; cfp
=cfp
->next
){
1192 if( dot
>=rp
->nrhs
) continue;
1194 if( sp
->type
==NONTERMINAL
){
1195 if( sp
->rule
==0 && sp
!=lemp
->errsym
){
1196 ErrorMsg(lemp
->filename
,rp
->line
,"Nonterminal \"%s\" has no rules.",
1200 for(newrp
=sp
->rule
; newrp
; newrp
=newrp
->nextlhs
){
1201 newcfp
= Configlist_add(newrp
,0);
1202 for(i
=dot
+1; i
<rp
->nrhs
; i
++){
1204 if( xsp
->type
==TERMINAL
){
1205 SetAdd(newcfp
->fws
,xsp
->index
);
1208 SetUnion(newcfp
->fws
,xsp
->firstset
);
1209 if( xsp
->lambda
==Bo_FALSE
) break;
1212 if( i
==rp
->nrhs
) Plink_add(&cfp
->fplp
,newcfp
);
1219 /* Sort the configuration list */
1220 void Configlist_sort(){
1221 current
= (struct config
*)msort(current
,(void **)&(current
->next
),Configcmp
);
1226 /* Sort the basis configuration list */
1227 void Configlist_sortbasis(){
1228 basis
= (struct config
*)msort(current
,(void **)&(current
->bp
),Configcmp
);
1233 /* Return a pointer to the head of the configuration list and
1234 ** reset the list */
1235 struct config
*Configlist_return(){
1243 /* Return a pointer to the head of the configuration list and
1244 ** reset the list */
1245 struct config
*Configlist_basis(){
1253 /* Free all elements of the given configuration list */
1254 void Configlist_eat(cfp
)
1257 struct config
*nextcfp
;
1258 for(; cfp
; cfp
=nextcfp
){
1259 nextcfp
= cfp
->next
;
1260 assert( cfp
->fplp
==0 );
1261 assert( cfp
->bplp
==0 );
1262 if( cfp
->fws
) SetFree(cfp
->fws
);
1267 /***************** From the file "error.c" *********************************/
1269 ** Code for printing error message.
1272 /* Find a good place to break "msg" so that its length is at least "min"
1273 ** but no more than "max". Make the point as close to max as possible.
1275 static int findbreak(msg
,min
,max
)
1282 for(i
=spot
=min
; i
<=max
; i
++){
1284 if( c
=='\t' ) msg
[i
] = ' ';
1285 if( c
=='\n' ){ msg
[i
] = ' '; spot
= i
; break; }
1286 if( c
==0 ){ spot
= i
; break; }
1287 if( c
=='-' && i
<max
-1 ) spot
= i
+1;
1288 if( c
==' ' ) spot
= i
;
1294 ** The error message is split across multiple lines if necessary. The
1295 ** splits occur at a space, if there is a space available near the end
1298 #define ERRMSGSIZE 10000 /* Hope this is big enough. No way to error check */
1299 #define LINEWIDTH 79 /* Max width of any output line */
1300 #define PREFIXLIMIT 30 /* Max width of the prefix on each line */
1301 void ErrorMsg(const char *filename
, int lineno
, const char *format
, ...){
1302 char errmsg
[ERRMSGSIZE
];
1303 char prefix
[PREFIXLIMIT
+10];
1308 int end
, restart
, base
;
1310 va_start(ap
, format
);
1311 /* Prepare a prefix to be prepended to every output line */
1313 sprintf(prefix
,"%.*s:%d: ",PREFIXLIMIT
-10,filename
,lineno
);
1315 sprintf(prefix
,"%.*s: ",PREFIXLIMIT
-10,filename
);
1317 prefixsize
= strlen(prefix
);
1318 availablewidth
= LINEWIDTH
- prefixsize
;
1320 /* Generate the error message */
1321 vsprintf(errmsg
,format
,ap
);
1323 errmsgsize
= strlen(errmsg
);
1324 /* Remove trailing '\n's from the error message. */
1325 while( errmsgsize
>0 && errmsg
[errmsgsize
-1]=='\n' ){
1326 errmsg
[--errmsgsize
] = 0;
1329 /* Print the error message */
1331 while( errmsg
[base
]!=0 ){
1332 end
= restart
= findbreak(&errmsg
[base
],0,availablewidth
);
1334 while( errmsg
[restart
]==' ' ) restart
++;
1335 fprintf(stdout
,"%s%.*s\n",prefix
,end
,&errmsg
[base
]);
1339 /**************** From the file "main.c" ************************************/
1341 ** Main program file for the LEMON parser generator.
1344 /* Report an out-of-memory condition and abort. This function
1345 ** is used mostly by the "MemoryCheck" macro in struct.h
1347 void memory_error() {
1348 fprintf(stderr
,"Out of memory. Aborting...\n");
1353 /* The main program. Parse the command line and do it... */
1358 static int version
= 0;
1359 static int rpflag
= 0;
1360 static int basisflag
= 0;
1361 static int compress
= 0;
1362 static int quiet
= 0;
1363 static int statistics
= 0;
1364 static int mhflag
= 0;
1365 static struct s_options options
[] = {
1366 {OPT_FLAG
, "b", (char*)&basisflag
, "Print only the basis in report."},
1367 {OPT_FLAG
, "c", (char*)&compress
, "Don't compress the action table."},
1368 {OPT_FLAG
, "g", (char*)&rpflag
, "Print grammar without actions."},
1369 {OPT_FLAG
, "m", (char*)&mhflag
, "Output a makeheaders compatible file"},
1370 {OPT_FLAG
, "q", (char*)&quiet
, "(Quiet) Don't print the report file."},
1371 {OPT_FLAG
, "s", (char*)&statistics
, "Print parser stats to standard output."},
1372 {OPT_FLAG
, "x", (char*)&version
, "Print the version number."},
1377 char *def_tmpl_name
= "lempar.c";
1380 OptInit(argv
,options
,stderr
);
1382 printf("Lemon version 1.0\n");
1385 if( OptNArgs() < 1 ){
1386 fprintf(stderr
,"Exactly one filename argument is required.\n");
1391 /* Initialize the machine */
1395 lem
.argv0
= argv
[0];
1396 lem
.filename
= OptArg(0);
1397 lem
.tmplname
= (OptNArgs() == 2) ? OptArg(1) : def_tmpl_name
;
1398 lem
.basisflag
= basisflag
;
1399 lem
.has_fallback
= 0;
1401 lem
.name
= lem
.include
= lem
.arg
= lem
.tokentype
= lem
.start
= 0;
1404 lem
.error
= lem
.overflow
= lem
.failure
= lem
.accept
= lem
.tokendest
=
1405 lem
.tokenprefix
= lem
.outname
= lem
.extracode
= 0;
1409 lem
.errsym
= Symbol_new("error");
1411 /* Parse the input file */
1413 if( lem
.errorcnt
) exit(lem
.errorcnt
);
1415 fprintf(stderr
,"Empty grammar.\n");
1419 /* Count and index the symbols of the grammar */
1420 Symbol_new("{default}");
1421 lem
.nsymbol
= Symbol_count();
1422 lem
.symbols
= Symbol_arrayof();
1423 for(i
=0; i
<lem
.nsymbol
; i
++) lem
.symbols
[i
]->index
= i
;
1424 qsort(lem
.symbols
,lem
.nsymbol
,sizeof(struct symbol
*),
1425 (int(*)())Symbolcmpp
);
1426 for(i
=0; i
<lem
.nsymbol
; i
++) lem
.symbols
[i
]->index
= i
;
1427 for(i
=1; i
<lem
.nsymbol
&& isupper(lem
.symbols
[i
]->name
[0]); i
++);
1428 lem
.nsymbol
--; /*(do not count "{default}")*/
1431 /* Generate a reprint of the grammar, if requested on the command line */
1435 /* Initialize the size for all follow and first sets */
1436 SetSize(lem
.nterminal
);
1438 /* Find the precedence for every production rule (that has one) */
1439 FindRulePrecedences(&lem
);
1441 /* Compute the lambda-nonterminals and the first-sets for every
1443 FindFirstSets(&lem
);
1445 /* Compute all LR(0) states. Also record follow-set propagation
1446 ** links so that the follow-set can be computed later */
1449 lem
.nstate
= State_count();
1450 lem
.sorted
= State_arrayof();
1452 /* Tie up loose ends on the propagation links */
1455 /* Compute the follow set of every reducible configuration */
1456 FindFollowSets(&lem
);
1458 /* Compute the action tables */
1461 /* Compress the action tables */
1462 if( compress
==0 ) CompressTables(&lem
);
1464 /* Generate a report of the parser generated. (the "y.output" file) */
1465 if( !quiet
) ReportOutput(&lem
);
1467 /* Generate the source code for the parser */
1468 ReportTable(&lem
, mhflag
);
1470 /* Produce a header file for use by the scanner. (This step is
1471 ** omitted if the "-m" option is used because makeheaders will
1472 ** generate the file for us.) */
1473 if( !mhflag
) ReportHeader(&lem
);
1476 printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
1477 lem
.nterminal
, lem
.nsymbol
- lem
.nterminal
, lem
.nrule
);
1478 printf(" %d states, %d parser table entries, %d conflicts\n",
1479 lem
.nstate
, lem
.tablesize
, lem
.nconflict
);
1481 if( lem
.nconflict
){
1482 fprintf(stderr
,"%d parsing conflicts.\n",lem
.nconflict
);
1484 exit(lem
.errorcnt
+ lem
.nconflict
);
1486 /******************** From the file "msort.c" *******************************/
1488 ** A generic merge-sort program.
1491 ** Let "ptr" be a pointer to some structure which is at the head of
1492 ** a null-terminated list. Then to sort the list call:
1494 ** ptr = msort(ptr,&(ptr->next),cmpfnc);
1496 ** In the above, "cmpfnc" is a pointer to a function which compares
1497 ** two instances of the structure and returns an integer, as in
1498 ** strcmp. The second argument is a pointer to the pointer to the
1499 ** second element of the linked list. This address is used to compute
1500 ** the offset to the "next" field within the structure. The offset to
1501 ** the "next" field must be constant for all structures in the list.
1503 ** The function returns a new pointer which is the head of the list
1511 ** Return a pointer to the next structure in the linked list.
1513 #define NEXT(A) (*(char**)(((unsigned long)A)+offset))
1517 ** a: A sorted, null-terminated linked list. (May be null).
1518 ** b: A sorted, null-terminated linked list. (May be null).
1519 ** cmp: A pointer to the comparison function.
1520 ** offset: Offset in the structure to the "next" field.
1523 ** A pointer to the head of a sorted list containing the elements
1527 ** The "next" pointers for elements in the lists a and b are
1530 static char *merge(a
,b
,cmp
,offset
)
1543 if( (*cmp
)(a
,b
)<0 ){
1552 if( (*cmp
)(a
,b
)<0 ){
1562 if( a
) NEXT(ptr
) = a
;
1570 ** list: Pointer to a singly-linked list of structures.
1571 ** next: Pointer to pointer to the second element of the list.
1572 ** cmp: A comparison function.
1575 ** A pointer to the head of a sorted list containing the elements
1576 ** orginally in list.
1579 ** The "next" pointers for elements in list are changed.
1582 void *msort(void *list
, void **next
, int(*cmp
)(void *, void *))
1584 unsigned long offset
;
1586 char *set
[LISTSIZE
];
1588 offset
= (unsigned long)next
- (unsigned long)list
;
1589 for(i
=0; i
<LISTSIZE
; i
++) set
[i
] = 0;
1594 for(i
=0; i
<LISTSIZE
-1 && set
[i
]!=0; i
++){
1595 ep
= merge(ep
,set
[i
],cmp
,offset
);
1601 for(i
=0; i
<LISTSIZE
; i
++) if( set
[i
] ) ep
= merge(ep
,set
[i
],cmp
,offset
);
1604 /************************ From the file "option.c" **************************/
1606 static struct s_options
*op
;
1607 static FILE *errstream
;
1609 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1612 ** Print the command line with a carrot pointing to the k-th character
1613 ** of the n-th field.
1615 static void errline(n
,k
,err
)
1622 fprintf(err
,"%s",argv
[0]);
1623 spcnt
+= strlen(argv
[0]) + 1;
1625 for(i
=1; i
<n
&& argv
[i
]; i
++){
1626 fprintf(err
," %s",argv
[i
]);
1627 spcnt
+= strlen(argv
[i
]) + 1;
1630 for(; argv
[i
]; i
++) fprintf(err
," %s",argv
[i
]);
1632 fprintf(err
,"\n%*s^-- here\n",spcnt
,"");
1634 fprintf(err
,"\n%*shere --^\n",spcnt
-7,"");
1639 ** Return the index of the N-th non-switch argument. Return -1
1640 ** if N is out of range.
1642 static int argindex(n
)
1647 if( argv
!=0 && *argv
!=0 ){
1648 for(i
=1; argv
[i
]; i
++){
1649 if( dashdash
|| !ISOPT(argv
[i
]) ){
1650 if( n
==0 ) return i
;
1653 if( strcmp(argv
[i
],"--")==0 ) dashdash
= 1;
1659 static char emsg
[] = "Command line syntax error: ";
1662 ** Process a flag command line argument.
1664 static int handleflags(i
,err
)
1671 for(j
=0; op
[j
].label
; j
++){
1672 if( strcmp(&argv
[i
][1],op
[j
].label
)==0 ) break;
1674 v
= argv
[i
][0]=='-' ? 1 : 0;
1675 if( op
[j
].label
==0 ){
1677 fprintf(err
,"%sundefined option.\n",emsg
);
1681 }else if( op
[j
].type
==OPT_FLAG
){
1682 *((int*)op
[j
].arg
) = v
;
1683 }else if( op
[j
].type
==OPT_FFLAG
){
1684 (*(void(*)())(intptr_t)(op
[j
].arg
))(v
);
1687 fprintf(err
,"%smissing argument on switch.\n",emsg
);
1696 ** Process a command line switch which has an argument.
1698 static int handleswitch(i
,err
)
1708 cp
= strchr(argv
[i
],'=');
1710 for(j
=0; op
[j
].label
; j
++){
1711 if( strcmp(argv
[i
],op
[j
].label
)==0 ) break;
1714 if( op
[j
].label
==0 ){
1716 fprintf(err
,"%sundefined option.\n",emsg
);
1722 switch( op
[j
].type
){
1726 fprintf(err
,"%soption requires an argument.\n",emsg
);
1733 dv
= strtod(cp
,&end
);
1736 fprintf(err
,"%sillegal character in floating-point argument.\n",emsg
);
1737 errline(i
,((unsigned long)end
)-(unsigned long)argv
[i
],err
);
1744 lv
= strtol(cp
,&end
,0);
1747 fprintf(err
,"%sillegal character in integer argument.\n",emsg
);
1748 errline(i
,((unsigned long)end
)-(unsigned long)argv
[i
],err
);
1758 switch( op
[j
].type
){
1763 *(double*)(op
[j
].arg
) = dv
;
1766 (*(void(*)())(intptr_t)(op
[j
].arg
))(dv
);
1769 *(int*)(op
[j
].arg
) = lv
;
1772 (*(void(*)())(intptr_t)(op
[j
].arg
))((int)lv
);
1775 *(char**)(op
[j
].arg
) = sv
;
1778 (*(void(*)())(intptr_t)(op
[j
].arg
))(sv
);
1785 int OptInit(a
,o
,err
)
1787 struct s_options
*o
;
1794 if( argv
&& *argv
&& op
){
1796 for(i
=1; argv
[i
]; i
++){
1797 if( argv
[i
][0]=='+' || argv
[i
][0]=='-' ){
1798 errcnt
+= handleflags(i
,err
);
1799 }else if( strchr(argv
[i
],'=') ){
1800 errcnt
+= handleswitch(i
,err
);
1805 fprintf(err
,"Valid command line options for \"%s\" are:\n",*a
);
1816 if( argv
!=0 && argv
[0]!=0 ){
1817 for(i
=1; argv
[i
]; i
++){
1818 if( dashdash
|| !ISOPT(argv
[i
]) ) cnt
++;
1819 if( strcmp(argv
[i
],"--")==0 ) dashdash
= 1;
1830 return i
>=0 ? argv
[i
] : 0;
1838 if( i
>=0 ) errline(i
,0,errstream
);
1845 for(i
=0; op
[i
].label
; i
++){
1846 len
= strlen(op
[i
].label
) + 1;
1847 switch( op
[i
].type
){
1853 len
+= 9; /* length of "<integer>" */
1857 len
+= 6; /* length of "<real>" */
1861 len
+= 8; /* length of "<string>" */
1864 if( len
>max
) max
= len
;
1866 for(i
=0; op
[i
].label
; i
++){
1867 switch( op
[i
].type
){
1870 fprintf(errstream
," -%-*s %s\n",max
,op
[i
].label
,op
[i
].message
);
1874 fprintf(errstream
," %s=<integer>%*s %s\n",op
[i
].label
,
1875 (int)(max
-strlen(op
[i
].label
)-9),"",op
[i
].message
);
1879 fprintf(errstream
," %s=<real>%*s %s\n",op
[i
].label
,
1880 (int)(max
-strlen(op
[i
].label
)-6),"",op
[i
].message
);
1884 fprintf(errstream
," %s=<string>%*s %s\n",op
[i
].label
,
1885 (int)(max
-strlen(op
[i
].label
)-8),"",op
[i
].message
);
1890 /*********************** From the file "parse.c" ****************************/
1892 ** Input file parser for the LEMON parser generator.
1895 /* The state of the parser */
1897 char *filename
; /* Name of the input file */
1898 int tokenlineno
; /* Linenumber at which current token starts */
1899 int errorcnt
; /* Number of errors so far */
1900 char *tokenstart
; /* Text of current token */
1901 struct lemon
*gp
; /* Global state vector */
1904 WAITING_FOR_DECL_OR_RULE
,
1905 WAITING_FOR_DECL_KEYWORD
,
1906 WAITING_FOR_DECL_ARG
,
1907 WAITING_FOR_PRECEDENCE_SYMBOL
,
1917 RESYNC_AFTER_RULE_ERROR
,
1918 RESYNC_AFTER_DECL_ERROR
,
1919 WAITING_FOR_DESTRUCTOR_SYMBOL
,
1920 WAITING_FOR_DATATYPE_SYMBOL
,
1921 WAITING_FOR_FALLBACK_ID
1922 } state
; /* The state of the parser */
1923 struct symbol
*fallback
; /* The fallback token */
1924 struct symbol
*lhs
; /* Left-hand side of current rule */
1925 char *lhsalias
; /* Alias for the LHS */
1926 int nrhs
; /* Number of right-hand side symbols seen */
1927 struct symbol
*rhs
[MAXRHS
]; /* RHS symbols */
1928 char *alias
[MAXRHS
]; /* Aliases for each RHS symbol (or NULL) */
1929 struct rule
*prevrule
; /* Previous rule parsed */
1930 char *declkeyword
; /* Keyword of a declaration */
1931 char **declargslot
; /* Where the declaration argument should be put */
1932 int *decllnslot
; /* Where the declaration linenumber is put */
1933 enum e_assoc declassoc
; /* Assign this association to decl arguments */
1934 int preccounter
; /* Assign this precedence to decl arguments */
1935 struct rule
*firstrule
; /* Pointer to first rule in the grammar */
1936 struct rule
*lastrule
; /* Pointer to the most recently parsed rule */
1939 /* Parse a single token */
1940 static void parseonetoken(psp
)
1944 x
= Strsafe(psp
->tokenstart
); /* Save the token permanently */
1946 printf("%s:%d: Token=[%s] state=%d\n",psp
->filename
,psp
->tokenlineno
,
1949 switch( psp
->state
){
1952 psp
->preccounter
= 0;
1953 psp
->firstrule
= psp
->lastrule
= 0;
1955 /* Fall thru to next case */
1956 case WAITING_FOR_DECL_OR_RULE
:
1958 psp
->state
= WAITING_FOR_DECL_KEYWORD
;
1959 }else if( islower(x
[0]) ){
1960 psp
->lhs
= Symbol_new(x
);
1963 psp
->state
= WAITING_FOR_ARROW
;
1964 }else if( x
[0]=='{' ){
1965 if( psp
->prevrule
==0 ){
1966 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1967 "There is not prior rule opon which to attach the code \
1968 fragment which begins on this line.");
1970 }else if( psp
->prevrule
->code
!=0 ){
1971 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1972 "Code fragment beginning on this line is not the first \
1973 to follow the previous rule.");
1976 psp
->prevrule
->line
= psp
->tokenlineno
;
1977 psp
->prevrule
->code
= &x
[1];
1979 }else if( x
[0]=='[' ){
1980 psp
->state
= PRECEDENCE_MARK_1
;
1982 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1983 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
1988 case PRECEDENCE_MARK_1
:
1989 if( !isupper(x
[0]) ){
1990 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1991 "The precedence symbol must be a terminal.");
1993 }else if( psp
->prevrule
==0 ){
1994 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1995 "There is no prior rule to assign precedence \"[%s]\".",x
);
1997 }else if( psp
->prevrule
->precsym
!=0 ){
1998 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1999 "Precedence mark on this line is not the first \
2000 to follow the previous rule.");
2003 psp
->prevrule
->precsym
= Symbol_new(x
);
2005 psp
->state
= PRECEDENCE_MARK_2
;
2007 case PRECEDENCE_MARK_2
:
2009 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2010 "Missing \"]\" on precedence mark.");
2013 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2015 case WAITING_FOR_ARROW
:
2016 if( x
[0]==':' && x
[1]==':' && x
[2]=='=' ){
2017 psp
->state
= IN_RHS
;
2018 }else if( x
[0]=='(' ){
2019 psp
->state
= LHS_ALIAS_1
;
2021 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2022 "Expected to see a \":\" following the LHS symbol \"%s\".",
2025 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2029 if( isalpha(x
[0]) ){
2031 psp
->state
= LHS_ALIAS_2
;
2033 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2034 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2037 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2042 psp
->state
= LHS_ALIAS_3
;
2044 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2045 "Missing \")\" following LHS alias name \"%s\".",psp
->lhsalias
);
2047 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2051 if( x
[0]==':' && x
[1]==':' && x
[2]=='=' ){
2052 psp
->state
= IN_RHS
;
2054 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2055 "Missing \"->\" following: \"%s(%s)\".",
2056 psp
->lhs
->name
,psp
->lhsalias
);
2058 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2064 rp
= (struct rule
*)malloc( sizeof(struct rule
) +
2065 sizeof(struct symbol
*)*psp
->nrhs
+ sizeof(char*)*psp
->nrhs
);
2067 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2068 "Can't allocate enough memory for this rule.");
2073 rp
->ruleline
= psp
->tokenlineno
;
2074 rp
->rhs
= (struct symbol
**)&rp
[1];
2075 rp
->rhsalias
= (char**)&(rp
->rhs
[psp
->nrhs
]);
2076 for(i
=0; i
<psp
->nrhs
; i
++){
2077 rp
->rhs
[i
] = psp
->rhs
[i
];
2078 rp
->rhsalias
[i
] = psp
->alias
[i
];
2081 rp
->lhsalias
= psp
->lhsalias
;
2082 rp
->nrhs
= psp
->nrhs
;
2085 rp
->index
= psp
->gp
->nrule
++;
2086 rp
->nextlhs
= rp
->lhs
->rule
;
2089 if( psp
->firstrule
==0 ){
2090 psp
->firstrule
= psp
->lastrule
= rp
;
2092 psp
->lastrule
->next
= rp
;
2097 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2098 }else if( isalpha(x
[0]) ){
2099 if( psp
->nrhs
>=MAXRHS
){
2100 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2101 "Too many symbol on RHS or rule beginning at \"%s\".",
2104 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2106 psp
->rhs
[psp
->nrhs
] = Symbol_new(x
);
2107 psp
->alias
[psp
->nrhs
] = 0;
2110 }else if( x
[0]=='(' && psp
->nrhs
>0 ){
2111 psp
->state
= RHS_ALIAS_1
;
2113 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2114 "Illegal character on RHS of rule: \"%s\".",x
);
2116 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2120 if( isalpha(x
[0]) ){
2121 psp
->alias
[psp
->nrhs
-1] = x
;
2122 psp
->state
= RHS_ALIAS_2
;
2124 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2125 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2126 x
,psp
->rhs
[psp
->nrhs
-1]->name
);
2128 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2133 psp
->state
= IN_RHS
;
2135 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2136 "Missing \")\" following LHS alias name \"%s\".",psp
->lhsalias
);
2138 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2141 case WAITING_FOR_DECL_KEYWORD
:
2142 if( isalpha(x
[0]) ){
2143 psp
->declkeyword
= x
;
2144 psp
->declargslot
= 0;
2145 psp
->decllnslot
= 0;
2146 psp
->state
= WAITING_FOR_DECL_ARG
;
2147 if( strcmp(x
,"name")==0 ){
2148 psp
->declargslot
= &(psp
->gp
->name
);
2149 }else if( strcmp(x
,"include")==0 ){
2150 psp
->declargslot
= &(psp
->gp
->include
);
2151 psp
->decllnslot
= &psp
->gp
->includeln
;
2152 }else if( strcmp(x
,"code")==0 ){
2153 psp
->declargslot
= &(psp
->gp
->extracode
);
2154 psp
->decllnslot
= &psp
->gp
->extracodeln
;
2155 }else if( strcmp(x
,"token_destructor")==0 ){
2156 psp
->declargslot
= &psp
->gp
->tokendest
;
2157 psp
->decllnslot
= &psp
->gp
->tokendestln
;
2158 }else if( strcmp(x
,"default_destructor")==0 ){
2159 psp
->declargslot
= &psp
->gp
->vardest
;
2160 psp
->decllnslot
= &psp
->gp
->vardestln
;
2161 }else if( strcmp(x
,"token_prefix")==0 ){
2162 psp
->declargslot
= &psp
->gp
->tokenprefix
;
2163 }else if( strcmp(x
,"syntax_error")==0 ){
2164 psp
->declargslot
= &(psp
->gp
->error
);
2165 psp
->decllnslot
= &psp
->gp
->errorln
;
2166 }else if( strcmp(x
,"parse_accept")==0 ){
2167 psp
->declargslot
= &(psp
->gp
->accept
);
2168 psp
->decllnslot
= &psp
->gp
->acceptln
;
2169 }else if( strcmp(x
,"parse_failure")==0 ){
2170 psp
->declargslot
= &(psp
->gp
->failure
);
2171 psp
->decllnslot
= &psp
->gp
->failureln
;
2172 }else if( strcmp(x
,"stack_overflow")==0 ){
2173 psp
->declargslot
= &(psp
->gp
->overflow
);
2174 psp
->decllnslot
= &psp
->gp
->overflowln
;
2175 }else if( strcmp(x
,"extra_argument")==0 ){
2176 psp
->declargslot
= &(psp
->gp
->arg
);
2177 }else if( strcmp(x
,"token_type")==0 ){
2178 psp
->declargslot
= &(psp
->gp
->tokentype
);
2179 }else if( strcmp(x
,"default_type")==0 ){
2180 psp
->declargslot
= &(psp
->gp
->vartype
);
2181 }else if( strcmp(x
,"stack_size")==0 ){
2182 psp
->declargslot
= &(psp
->gp
->stacksize
);
2183 }else if( strcmp(x
,"start_symbol")==0 ){
2184 psp
->declargslot
= &(psp
->gp
->start
);
2185 }else if( strcmp(x
,"left")==0 ){
2187 psp
->declassoc
= LEFT
;
2188 psp
->state
= WAITING_FOR_PRECEDENCE_SYMBOL
;
2189 }else if( strcmp(x
,"right")==0 ){
2191 psp
->declassoc
= RIGHT
;
2192 psp
->state
= WAITING_FOR_PRECEDENCE_SYMBOL
;
2193 }else if( strcmp(x
,"nonassoc")==0 ){
2195 psp
->declassoc
= NONE
;
2196 psp
->state
= WAITING_FOR_PRECEDENCE_SYMBOL
;
2197 }else if( strcmp(x
,"destructor")==0 ){
2198 psp
->state
= WAITING_FOR_DESTRUCTOR_SYMBOL
;
2199 }else if( strcmp(x
,"type")==0 ){
2200 psp
->state
= WAITING_FOR_DATATYPE_SYMBOL
;
2201 }else if( strcmp(x
,"fallback")==0 ){
2203 psp
->state
= WAITING_FOR_FALLBACK_ID
;
2205 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2206 "Unknown declaration keyword: \"%%%s\".",x
);
2208 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2211 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2212 "Illegal declaration keyword: \"%s\".",x
);
2214 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2217 case WAITING_FOR_DESTRUCTOR_SYMBOL
:
2218 if( !isalpha(x
[0]) ){
2219 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2220 "Symbol name missing after %destructor keyword");
2222 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2224 struct symbol
*sp
= Symbol_new(x
);
2225 psp
->declargslot
= &sp
->destructor
;
2226 psp
->decllnslot
= &sp
->destructorln
;
2227 psp
->state
= WAITING_FOR_DECL_ARG
;
2230 case WAITING_FOR_DATATYPE_SYMBOL
:
2231 if( !isalpha(x
[0]) ){
2232 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2233 "Symbol name missing after %destructor keyword");
2235 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2237 struct symbol
*sp
= Symbol_new(x
);
2238 psp
->declargslot
= &sp
->datatype
;
2239 psp
->decllnslot
= 0;
2240 psp
->state
= WAITING_FOR_DECL_ARG
;
2243 case WAITING_FOR_PRECEDENCE_SYMBOL
:
2245 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2246 }else if( isupper(x
[0]) ){
2250 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2251 "Symbol \"%s\" has already be given a precedence.",x
);
2254 sp
->prec
= psp
->preccounter
;
2255 sp
->assoc
= psp
->declassoc
;
2258 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2259 "Can't assign a precedence to \"%s\".",x
);
2263 case WAITING_FOR_DECL_ARG
:
2264 if( (x
[0]=='{' || x
[0]=='\"' || isalnum(x
[0])) ){
2265 if( *(psp
->declargslot
)!=0 ){
2266 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2267 "The argument \"%s\" to declaration \"%%%s\" is not the first.",
2268 x
[0]=='\"' ? &x
[1] : x
,psp
->declkeyword
);
2270 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2272 *(psp
->declargslot
) = (x
[0]=='\"' || x
[0]=='{') ? &x
[1] : x
;
2273 if( psp
->decllnslot
) *psp
->decllnslot
= psp
->tokenlineno
;
2274 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2277 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2278 "Illegal argument to %%%s: %s",psp
->declkeyword
,x
);
2280 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2283 case WAITING_FOR_FALLBACK_ID
:
2285 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2286 }else if( !isupper(x
[0]) ){
2287 ErrorMsg(psp
->filename
, psp
->tokenlineno
,
2288 "%%fallback argument \"%s\" should be a token", x
);
2291 struct symbol
*sp
= Symbol_new(x
);
2292 if( psp
->fallback
==0 ){
2294 }else if( sp
->fallback
){
2295 ErrorMsg(psp
->filename
, psp
->tokenlineno
,
2296 "More than one fallback assigned to token %s", x
);
2299 sp
->fallback
= psp
->fallback
;
2300 psp
->gp
->has_fallback
= 1;
2304 case RESYNC_AFTER_RULE_ERROR
:
2305 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2307 case RESYNC_AFTER_DECL_ERROR
:
2308 if( x
[0]=='.' ) psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2309 if( x
[0]=='%' ) psp
->state
= WAITING_FOR_DECL_KEYWORD
;
2314 /* In spite of its name, this function is really a scanner. It read
2315 ** in the entire input file (all at once) then tokenizes it. Each
2316 ** token is passed to the function "parseonetoken" which builds all
2317 ** the appropriate data structures in the global state vector "gp".
2332 ps
.filename
= gp
->filename
;
2334 ps
.state
= INITIALIZE
;
2336 /* Begin by reading the input file */
2337 fp
= fopen(ps
.filename
,"rb");
2339 ErrorMsg(ps
.filename
,0,"Can't open this file for reading.");
2344 filesize
= ftell(fp
);
2346 filebuf
= (char *)malloc( filesize
+1 );
2348 ErrorMsg(ps
.filename
,0,"Can't allocate %d of memory to hold this file.",
2354 if( fread(filebuf
,1,filesize
,fp
)!=filesize
){
2355 ErrorMsg(ps
.filename
,0,"Can't read in all %d bytes of this file.",
2363 filebuf
[filesize
] = 0;
2365 /* Now scan the text of the input file */
2367 for(cp
=filebuf
; (c
= *cp
)!=0; ){
2368 if( c
=='\n' ) lineno
++; /* Keep track of the line number */
2369 if( isspace(c
) ){ cp
++; continue; } /* Skip all white space */
2370 if( c
=='/' && cp
[1]=='/' ){ /* Skip C++ style comments */
2372 while( (c
= *cp
)!=0 && c
!='\n' ) cp
++;
2375 if( c
=='/' && cp
[1]=='*' ){ /* Skip C style comments */
2377 while( (c
= *cp
)!=0 && (c
!='/' || cp
[-1]!='*') ){
2378 if( c
=='\n' ) lineno
++;
2384 ps
.tokenstart
= cp
; /* Mark the beginning of the token */
2385 ps
.tokenlineno
= lineno
; /* Linenumber on which token begins */
2386 if( c
=='\"' ){ /* String literals */
2388 while( (c
= *cp
)!=0 && c
!='\"' ){
2389 if( c
=='\n' ) lineno
++;
2393 ErrorMsg(ps
.filename
,startline
,
2394 "String starting on this line is not terminated before the end of the file.");
2400 }else if( c
=='{' ){ /* A block of C code */
2403 for(level
=1; (c
= *cp
)!=0 && (level
>1 || c
!='}'); cp
++){
2404 if( c
=='\n' ) lineno
++;
2405 else if( c
=='{' ) level
++;
2406 else if( c
=='}' ) level
--;
2407 else if( c
=='/' && cp
[1]=='*' ){ /* Skip comments */
2411 while( (c
= *cp
)!=0 && (c
!='/' || prevc
!='*') ){
2412 if( c
=='\n' ) lineno
++;
2416 }else if( c
=='/' && cp
[1]=='/' ){ /* Skip C++ style comments too */
2418 while( (c
= *cp
)!=0 && c
!='\n' ) cp
++;
2420 }else if( c
=='\'' || c
=='\"' ){ /* String a character literals */
2421 int startchar
, prevc
;
2424 for(cp
++; (c
= *cp
)!=0 && (c
!=startchar
|| prevc
=='\\'); cp
++){
2425 if( c
=='\n' ) lineno
++;
2426 if( prevc
=='\\' ) prevc
= 0;
2432 ErrorMsg(ps
.filename
,ps
.tokenlineno
,
2433 "C code starting on this line is not terminated before the end of the file.");
2439 }else if( isalnum(c
) ){ /* Identifiers */
2440 while( (c
= *cp
)!=0 && (isalnum(c
) || c
=='_') ) cp
++;
2442 }else if( c
==':' && cp
[1]==':' && cp
[2]=='=' ){ /* The operator "::=" */
2445 }else{ /* All other (one character) operators */
2450 *cp
= 0; /* Null terminate the token */
2451 parseonetoken(&ps
); /* Parse the token */
2452 *cp
= c
; /* Restore the buffer */
2455 free(filebuf
); /* Release the buffer after parsing */
2456 gp
->rule
= ps
.firstrule
;
2457 gp
->errorcnt
= ps
.errorcnt
;
2459 /*************************** From the file "plink.c" *********************/
2461 ** Routines processing configuration follow-set propagation links
2462 ** in the LEMON parser generator.
2464 static struct plink
*plink_freelist
= 0;
2466 /* Allocate a new plink */
2467 struct plink
*Plink_new(){
2470 if( plink_freelist
==0 ){
2473 plink_freelist
= (struct plink
*)malloc( sizeof(struct plink
)*amt
);
2474 if( plink_freelist
==0 ){
2476 "Unable to allocate memory for a new follow-set propagation link.\n");
2479 for(i
=0; i
<amt
-1; i
++) plink_freelist
[i
].next
= &plink_freelist
[i
+1];
2480 plink_freelist
[amt
-1].next
= 0;
2482 new = plink_freelist
;
2483 plink_freelist
= plink_freelist
->next
;
2487 /* Add a plink to a plink list */
2488 void Plink_add(plpp
,cfp
)
2489 struct plink
**plpp
;
2499 /* Transfer every plink on the list "from" to the list "to" */
2500 void Plink_copy(to
,from
)
2504 struct plink
*nextpl
;
2506 nextpl
= from
->next
;
2513 /* Delete every plink on the list */
2514 void Plink_delete(plp
)
2517 struct plink
*nextpl
;
2521 plp
->next
= plink_freelist
;
2522 plink_freelist
= plp
;
2526 /*********************** From the file "report.c" **************************/
2528 ** Procedures for generating reports and tables in the LEMON parser generator.
2531 /* Generate a filename with the given suffix. Space to hold the
2532 ** name comes from malloc() and must be freed by the calling
2535 PRIVATE
char *file_makename(lemp
,suffix
)
2542 name
= malloc( strlen(lemp
->filename
) + strlen(suffix
) + 5 );
2544 fprintf(stderr
,"Can't allocate space for a filename.\n");
2547 /* skip directory, JK */
2548 if (NULL
== (cp
= strrchr(lemp
->filename
, '/'))) {
2549 cp
= lemp
->filename
;
2554 cp
= strrchr(name
,'.');
2556 strcat(name
,suffix
);
2560 /* Open a file with a name based on the name of the input file,
2561 ** but with a different (specified) suffix, and return a pointer
2563 PRIVATE
FILE *file_open(lemp
,suffix
,mode
)
2570 if( lemp
->outname
) free(lemp
->outname
);
2571 lemp
->outname
= file_makename(lemp
, suffix
);
2572 fp
= fopen(lemp
->outname
,mode
);
2573 if( fp
==0 && *mode
=='w' ){
2574 fprintf(stderr
,"Can't open file \"%s\".\n",lemp
->outname
);
2581 /* Duplicate the input file without comments and without actions
2588 int i
, j
, maxlen
, len
, ncolumns
, skip
;
2589 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp
->filename
);
2591 for(i
=0; i
<lemp
->nsymbol
; i
++){
2592 sp
= lemp
->symbols
[i
];
2593 len
= strlen(sp
->name
);
2594 if( len
>maxlen
) maxlen
= len
;
2596 ncolumns
= 76/(maxlen
+5);
2597 if( ncolumns
<1 ) ncolumns
= 1;
2598 skip
= (lemp
->nsymbol
+ ncolumns
- 1)/ncolumns
;
2599 for(i
=0; i
<skip
; i
++){
2601 for(j
=i
; j
<lemp
->nsymbol
; j
+=skip
){
2602 sp
= lemp
->symbols
[j
];
2603 assert( sp
->index
==j
);
2604 printf(" %3d %-*.*s",j
,maxlen
,maxlen
,sp
->name
);
2608 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
2609 printf("%s",rp
->lhs
->name
);
2610 /* if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
2612 for(i
=0; i
<rp
->nrhs
; i
++){
2613 printf(" %s",rp
->rhs
[i
]->name
);
2614 /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
2617 if( rp
->precsym
) printf(" [%s]",rp
->precsym
->name
);
2618 /* if( rp->code ) printf("\n %s",rp->code); */
2623 PRIVATE
void ConfigPrint(fp
,cfp
)
2630 fprintf(fp
,"%s ::=",rp
->lhs
->name
);
2631 for(i
=0; i
<=rp
->nrhs
; i
++){
2632 if( i
==cfp
->dot
) fprintf(fp
," *");
2633 if( i
==rp
->nrhs
) break;
2634 fprintf(fp
," %s",rp
->rhs
[i
]->name
);
2641 PRIVATE
void SetPrint(out
,set
,lemp
)
2649 fprintf(out
,"%12s[","");
2650 for(i
=0; i
<lemp
->nterminal
; i
++){
2651 if( SetFind(set
,i
) ){
2652 fprintf(out
,"%s%s",spacer
,lemp
->symbols
[i
]->name
);
2659 /* Print a plink chain */
2660 void PlinkPrint(out
,plp
,tag
)
2666 fprintf(out
,"%12s%s (state %2d) ","",tag
,plp
->cfp
->stp
->index
);
2667 ConfigPrint(out
,plp
->cfp
);
2674 /* Print an action to the given file descriptor. Return FALSE if
2675 ** nothing was actually printed.
2677 PRIVATE
int PrintAction(struct action
*ap
, FILE *fp
, int indent
){
2681 fprintf(fp
,"%*s shift %d",indent
,ap
->sp
->name
,ap
->x
.stp
->index
);
2684 fprintf(fp
,"%*s reduce %d",indent
,ap
->sp
->name
,ap
->x
.rp
->index
);
2687 fprintf(fp
,"%*s accept",indent
,ap
->sp
->name
);
2690 fprintf(fp
,"%*s error",indent
,ap
->sp
->name
);
2693 fprintf(fp
,"%*s reduce %-3d ** Parsing conflict **",
2694 indent
,ap
->sp
->name
,ap
->x
.rp
->index
);
2705 /* Generate the "y.output" log file */
2706 void ReportOutput(lemp
)
2715 fp
= file_open(lemp
,".out","w");
2718 for(i
=0; i
<lemp
->nstate
; i
++){
2719 stp
= lemp
->sorted
[i
];
2720 fprintf(fp
,"State %d:\n",stp
->index
);
2721 if( lemp
->basisflag
) cfp
=stp
->bp
;
2725 if( cfp
->dot
==cfp
->rp
->nrhs
){
2726 sprintf(buf
,"(%d)",cfp
->rp
->index
);
2727 fprintf(fp
," %5s ",buf
);
2731 ConfigPrint(fp
,cfp
);
2734 SetPrint(fp
,cfp
->fws
,lemp
);
2735 PlinkPrint(fp
,cfp
->fplp
,"To ");
2736 PlinkPrint(fp
,cfp
->bplp
,"From");
2738 if( lemp
->basisflag
) cfp
=cfp
->bp
;
2742 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
2743 if( PrintAction(ap
,fp
,30) ) fprintf(fp
,"\n");
2751 extern int access();
2752 /* Search for the file "name" which is in the same directory as
2753 ** the exacutable */
2754 PRIVATE
char *pathsearch(argv0
,name
,modemask
)
2764 cp
= strrchr(argv0
,'\\');
2766 cp
= strrchr(argv0
,'/');
2771 path
= (char *)malloc( strlen(argv0
) + strlen(name
) + 2 );
2772 if( path
) sprintf(path
,"%s/%s",argv0
,name
);
2775 pathlist
= getenv("PATH");
2776 if( pathlist
==0 ) pathlist
= ".:/bin:/usr/bin";
2777 path
= (char *)malloc( strlen(pathlist
)+strlen(name
)+2 );
2780 cp
= strchr(pathlist
,':');
2781 if( cp
==0 ) cp
= &pathlist
[strlen(pathlist
)];
2784 sprintf(path
,"%s/%s",pathlist
,name
);
2786 if( c
==0 ) pathlist
= "";
2787 else pathlist
= &cp
[1];
2788 if( access(path
,modemask
)==0 ) break;
2795 /* Given an action, compute the integer value for that action
2796 ** which is to be put in the action table of the generated machine.
2797 ** Return negative if no action should be generated.
2799 PRIVATE
int compute_action(lemp
,ap
)
2805 case SHIFT
: act
= ap
->x
.stp
->index
; break;
2806 case REDUCE
: act
= ap
->x
.rp
->index
+ lemp
->nstate
; break;
2807 case ERROR
: act
= lemp
->nstate
+ lemp
->nrule
; break;
2808 case ACCEPT
: act
= lemp
->nstate
+ lemp
->nrule
+ 1; break;
2809 default: act
= -1; break;
2814 #define LINESIZE 1000
2815 /* The next cluster of routines are for reading the template file
2816 ** and writing the results to the generated parser */
2817 /* The first function transfers data from "in" to "out" until
2818 ** a line is seen which begins with "%%". The line number is
2821 ** if name!=0, then any word that begin with "Parse" is changed to
2822 ** begin with *name instead.
2824 PRIVATE
void tplt_xfer(name
,in
,out
,lineno
)
2831 char line
[LINESIZE
];
2832 while( fgets(line
,LINESIZE
,in
) && (line
[0]!='%' || line
[1]!='%') ){
2836 for(i
=0; line
[i
]; i
++){
2837 if( line
[i
]=='P' && strncmp(&line
[i
],"Parse",5)==0
2838 && (i
==0 || !isalpha(line
[i
-1]))
2840 if( i
>iStart
) fprintf(out
,"%.*s",i
-iStart
,&line
[iStart
]);
2841 fprintf(out
,"%s",name
);
2847 fprintf(out
,"%s",&line
[iStart
]);
2851 /* The next function finds the template file and opens it, returning
2852 ** a pointer to the opened file. */
2853 PRIVATE
FILE *tplt_open(lemp
)
2860 char *tpltname_alloc
= NULL
;
2863 cp
= strrchr(lemp
->filename
,'.');
2865 sprintf(buf
,"%.*s.lt",(int)(cp
-lemp
->filename
),lemp
->filename
);
2867 sprintf(buf
,"%s.lt",lemp
->filename
);
2869 if( access(buf
,004)==0 ){
2871 }else if( access(lemp
->tmplname
,004)==0 ){
2872 tpltname
= lemp
->tmplname
;
2874 tpltname
= tpltname_alloc
= pathsearch(lemp
->argv0
,lemp
->tmplname
,0);
2877 fprintf(stderr
,"Can't find the parser driver template file \"%s\".\n",
2882 in
= fopen(tpltname
,"r");
2884 fprintf(stderr
,"Can't open the template file \"%s\".\n",tpltname
);
2887 if (tpltname_alloc
) free(tpltname_alloc
);
2891 /* Print a string to the file and keep the linenumber up to date */
2892 PRIVATE
void tplt_print(out
,lemp
,str
,strln
,lineno
)
2899 if( str
==0 ) return;
2900 fprintf(out
,"#line %d \"%s\"\n",strln
,lemp
->filename
); (*lineno
)++;
2902 if( *str
=='\n' ) (*lineno
)++;
2906 fprintf(out
,"\n#line %d \"%s\"\n",*lineno
+2,lemp
->outname
); (*lineno
)+=2;
2911 ** The following routine emits code for the destructor for the
2914 PRIVATE
void emit_destructor_code(out
,sp
,lemp
,lineno
)
2923 if( sp
->type
==TERMINAL
){
2924 cp
= lemp
->tokendest
;
2926 fprintf(out
,"#line %d \"%s\"\n{",lemp
->tokendestln
,lemp
->filename
);
2927 }else if( sp
->destructor
){
2928 cp
= sp
->destructor
;
2929 fprintf(out
,"#line %d \"%s\"\n{",sp
->destructorln
,lemp
->filename
);
2933 fprintf(out
,"#line %d \"%s\"\n{",lemp
->vardestln
,lemp
->filename
);
2936 if( *cp
=='$' && cp
[1]=='$' ){
2937 fprintf(out
,"(yypminor->yy%d)",sp
->dtnum
);
2941 if( *cp
=='\n' ) linecnt
++;
2944 (*lineno
) += 3 + linecnt
;
2945 fprintf(out
,"}\n#line %d \"%s\"\n",*lineno
,lemp
->outname
);
2950 ** Return TRUE (non-zero) if the given symbol has a destructor.
2952 PRIVATE
int has_destructor(sp
, lemp
)
2957 if( sp
->type
==TERMINAL
){
2958 ret
= lemp
->tokendest
!=0;
2960 ret
= lemp
->vardest
!=0 || sp
->destructor
!=0;
2966 ** Generate code which executes when the rule "rp" is reduced. Write
2967 ** the code to "out". Make sure lineno stays up-to-date.
2969 PRIVATE
void emit_code(out
,rp
,lemp
,lineno
)
2978 char lhsused
= 0; /* True if the LHS element has been used */
2979 char used
[MAXRHS
]; /* True for each RHS element which is used */
2981 for(i
=0; i
<rp
->nrhs
; i
++) used
[i
] = 0;
2984 /* Generate code to do the reduce action */
2986 fprintf(out
,"#line %d \"%s\"\n{",rp
->line
,lemp
->filename
);
2987 for(cp
=rp
->code
; *cp
; cp
++){
2988 if( isalpha(*cp
) && (cp
==rp
->code
|| (!isalnum(cp
[-1]) && cp
[-1]!='_')) ){
2990 for(xp
= &cp
[1]; isalnum(*xp
) || *xp
=='_'; xp
++);
2993 if( rp
->lhsalias
&& strcmp(cp
,rp
->lhsalias
)==0 ){
2994 fprintf(out
,"yygotominor.yy%d",rp
->lhs
->dtnum
);
2998 for(i
=0; i
<rp
->nrhs
; i
++){
2999 if( rp
->rhsalias
[i
] && strcmp(cp
,rp
->rhsalias
[i
])==0 ){
3000 fprintf(out
,"yymsp[%d].minor.yy%d",i
-rp
->nrhs
+1,rp
->rhs
[i
]->dtnum
);
3009 if( *cp
=='\n' ) linecnt
++;
3012 (*lineno
) += 3 + linecnt
;
3013 fprintf(out
,"}\n#line %d \"%s\"\n",*lineno
,lemp
->outname
);
3014 } /* End if( rp->code ) */
3016 /* Check to make sure the LHS has been used */
3017 if( rp
->lhsalias
&& !lhsused
){
3018 ErrorMsg(lemp
->filename
,rp
->ruleline
,
3019 "Label \"%s\" for \"%s(%s)\" is never used.",
3020 rp
->lhsalias
,rp
->lhs
->name
,rp
->lhsalias
);
3024 /* Generate destructor code for RHS symbols which are not used in the
3026 for(i
=0; i
<rp
->nrhs
; i
++){
3027 if( rp
->rhsalias
[i
] && !used
[i
] ){
3028 ErrorMsg(lemp
->filename
,rp
->ruleline
,
3029 "Label %s for \"%s(%s)\" is never used.",
3030 rp
->rhsalias
[i
],rp
->rhs
[i
]->name
,rp
->rhsalias
[i
]);
3032 }else if( rp
->rhsalias
[i
]==0 ){
3033 if( has_destructor(rp
->rhs
[i
],lemp
) ){
3034 fprintf(out
," yy_destructor(%d,&yymsp[%d].minor);\n",
3035 rp
->rhs
[i
]->index
,i
-rp
->nrhs
+1); (*lineno
)++;
3037 fprintf(out
," /* No destructor defined for %s */\n",
3047 ** Print the definition of the union used for the parser's data stack.
3048 ** This union contains fields for every possible data type for tokens
3049 ** and nonterminals. In the process of computing and printing this
3050 ** union, also set the ".dtnum" field of every terminal and nonterminal
3053 PRIVATE
void print_stack_union(out
,lemp
,plineno
,mhflag
)
3054 FILE *out
; /* The output stream */
3055 struct lemon
*lemp
; /* The main info structure for this parser */
3056 int *plineno
; /* Pointer to the line number */
3057 int mhflag
; /* True if generating makeheaders output */
3059 int lineno
; /* The line number of the output */
3060 char **types
; /* A hash table of datatypes */
3061 int arraysize
; /* Size of the "types" array */
3062 int maxdtlength
; /* Maximum length of any ".datatype" field. */
3063 char *stddt
; /* Standardized name for a datatype */
3064 int i
,j
; /* Loop counters */
3065 int hash
; /* For hashing the name of a type */
3066 char *name
; /* Name of the parser */
3068 /* Allocate and initialize types[] and allocate stddt[] */
3069 arraysize
= lemp
->nsymbol
* 2;
3070 types
= (char**)malloc( arraysize
* sizeof(char*) );
3071 for(i
=0; i
<arraysize
; i
++) types
[i
] = 0;
3073 if( lemp
->vartype
){
3074 maxdtlength
= strlen(lemp
->vartype
);
3076 for(i
=0; i
<lemp
->nsymbol
; i
++){
3078 struct symbol
*sp
= lemp
->symbols
[i
];
3079 if( sp
->datatype
==0 ) continue;
3080 len
= strlen(sp
->datatype
);
3081 if( len
>maxdtlength
) maxdtlength
= len
;
3083 stddt
= (char*)malloc( maxdtlength
*2 + 1 );
3084 if( types
==0 || stddt
==0 ){
3085 fprintf(stderr
,"Out of memory.\n");
3089 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3090 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
3091 ** used for terminal symbols. If there is no %default_type defined then
3092 ** 0 is also used as the .dtnum value for nonterminals which do not specify
3093 ** a datatype using the %type directive.
3095 for(i
=0; i
<lemp
->nsymbol
; i
++){
3096 struct symbol
*sp
= lemp
->symbols
[i
];
3098 if( sp
==lemp
->errsym
){
3099 sp
->dtnum
= arraysize
+1;
3102 if( sp
->type
!=NONTERMINAL
|| (sp
->datatype
==0 && lemp
->vartype
==0) ){
3107 if( cp
==0 ) cp
= lemp
->vartype
;
3109 while( isspace(*cp
) ) cp
++;
3110 while( *cp
) stddt
[j
++] = *cp
++;
3111 while( j
>0 && isspace(stddt
[j
-1]) ) j
--;
3114 for(j
=0; stddt
[j
]; j
++){
3115 hash
= (unsigned int)hash
*53u + (unsigned int) stddt
[j
];
3117 hash
= (hash
& 0x7fffffff)%arraysize
;
3118 while( types
[hash
] ){
3119 if( strcmp(types
[hash
],stddt
)==0 ){
3120 sp
->dtnum
= hash
+ 1;
3124 if( hash
>=arraysize
) hash
= 0;
3126 if( types
[hash
]==0 ){
3127 sp
->dtnum
= hash
+ 1;
3128 types
[hash
] = (char*)malloc( strlen(stddt
)+1 );
3129 if( types
[hash
]==0 ){
3130 fprintf(stderr
,"Out of memory.\n");
3133 strcpy(types
[hash
],stddt
);
3137 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
3138 name
= lemp
->name
? lemp
->name
: "Parse";
3140 if( mhflag
){ fprintf(out
,"#if INTERFACE\n"); lineno
++; }
3141 fprintf(out
,"#define %sTOKENTYPE %s\n",name
,
3142 lemp
->tokentype
?lemp
->tokentype
:"void*"); lineno
++;
3143 if( mhflag
){ fprintf(out
,"#endif\n"); lineno
++; }
3144 fprintf(out
,"typedef union {\n"); lineno
++;
3145 fprintf(out
," %sTOKENTYPE yy0;\n",name
); lineno
++;
3146 for(i
=0; i
<arraysize
; i
++){
3147 if( types
[i
]==0 ) continue;
3148 fprintf(out
," %s yy%d;\n",types
[i
],i
+1); lineno
++;
3151 fprintf(out
," int yy%d;\n",lemp
->errsym
->dtnum
); lineno
++;
3154 fprintf(out
,"} YYMINORTYPE;\n"); lineno
++;
3159 ** Return the name of a C datatype able to represent values between
3160 ** lwr and upr, inclusive.
3162 static const char *minimum_size_type(int lwr
, int upr
){
3165 return "unsigned char";
3166 }else if( upr
<65535 ){
3167 return "unsigned short int";
3169 return "unsigned int";
3171 }else if( lwr
>=-127 && upr
<=127 ){
3172 return "signed char";
3173 }else if( lwr
>=-32767 && upr
<32767 ){
3181 ** Each state contains a set of token transaction and a set of
3182 ** nonterminal transactions. Each of these sets makes an instance
3183 ** of the following structure. An array of these structures is used
3184 ** to order the creation of entries in the yy_action[] table.
3187 struct state
*stp
; /* A pointer to a state */
3188 int isTkn
; /* True to use tokens. False for non-terminals */
3189 int nAction
; /* Number of actions */
3193 ** Compare to axset structures for sorting purposes
3195 static int axset_compare(const void *a
, const void *b
){
3196 struct axset
*p1
= (struct axset
*)a
;
3197 struct axset
*p2
= (struct axset
*)b
;
3198 return p2
->nAction
- p1
->nAction
;
3201 /* Generate C source code for the parser */
3202 void ReportTable(lemp
, mhflag
)
3204 int mhflag
; /* Output in makeheaders format if true */
3207 char line
[LINESIZE
];
3212 struct acttab
*pActtab
;
3214 int mnTknOfst
, mxTknOfst
;
3215 int mnNtOfst
, mxNtOfst
;
3219 in
= tplt_open(lemp
);
3221 out
= file_open(lemp
,".c","w");
3227 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3229 /* Generate the include code, if any */
3230 tplt_print(out
,lemp
,lemp
->include
,lemp
->includeln
,&lineno
);
3232 name
= file_makename(lemp
, ".h");
3233 fprintf(out
,"#include \"%s\"\n", name
); lineno
++;
3236 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3238 /* Generate #defines for all tokens */
3241 fprintf(out
,"#if INTERFACE\n"); lineno
++;
3242 if( lemp
->tokenprefix
) prefix
= lemp
->tokenprefix
;
3244 for(i
=1; i
<lemp
->nterminal
; i
++){
3245 fprintf(out
,"#define %s%-30s %2d\n",prefix
,lemp
->symbols
[i
]->name
,i
);
3248 fprintf(out
,"#endif\n"); lineno
++;
3250 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3252 /* Generate the defines */
3253 fprintf(out
,"/* \001 */\n");
3254 fprintf(out
,"#define YYCODETYPE %s\n",
3255 minimum_size_type(0, lemp
->nsymbol
+5)); lineno
++;
3256 fprintf(out
,"#define YYNOCODE %d\n",lemp
->nsymbol
+1); lineno
++;
3257 fprintf(out
,"#define YYACTIONTYPE %s\n",
3258 minimum_size_type(0, lemp
->nstate
+lemp
->nrule
+5)); lineno
++;
3259 print_stack_union(out
,lemp
,&lineno
,mhflag
);
3260 if( lemp
->stacksize
){
3261 if( atoi(lemp
->stacksize
)<=0 ){
3262 ErrorMsg(lemp
->filename
,0,
3263 "Illegal stack size: [%s]. The stack size should be an integer constant.",
3266 lemp
->stacksize
= "100";
3268 fprintf(out
,"#define YYSTACKDEPTH %s\n",lemp
->stacksize
); lineno
++;
3270 fprintf(out
,"#define YYSTACKDEPTH 100\n"); lineno
++;
3273 fprintf(out
,"#if INTERFACE\n"); lineno
++;
3275 name
= lemp
->name
? lemp
->name
: "Parse";
3276 if( lemp
->arg
&& lemp
->arg
[0] ){
3277 i
= strlen(lemp
->arg
);
3278 while( i
>=1 && isspace(lemp
->arg
[i
-1]) ) i
--;
3279 while( i
>=1 && (isalnum(lemp
->arg
[i
-1]) || lemp
->arg
[i
-1]=='_') ) i
--;
3280 fprintf(out
,"#define %sARG_SDECL %s;\n",name
,lemp
->arg
); lineno
++;
3281 fprintf(out
,"#define %sARG_PDECL ,%s\n",name
,lemp
->arg
); lineno
++;
3282 fprintf(out
,"#define %sARG_FETCH %s = yypParser->%s\n",
3283 name
,lemp
->arg
,&lemp
->arg
[i
]); lineno
++;
3284 fprintf(out
,"#define %sARG_STORE yypParser->%s = %s\n",
3285 name
,&lemp
->arg
[i
],&lemp
->arg
[i
]); lineno
++;
3287 fprintf(out
,"#define %sARG_SDECL\n",name
); lineno
++;
3288 fprintf(out
,"#define %sARG_PDECL\n",name
); lineno
++;
3289 fprintf(out
,"#define %sARG_FETCH\n",name
); lineno
++;
3290 fprintf(out
,"#define %sARG_STORE\n",name
); lineno
++;
3293 fprintf(out
,"#endif\n"); lineno
++;
3295 fprintf(out
,"#define YYNSTATE %d\n",lemp
->nstate
); lineno
++;
3296 fprintf(out
,"#define YYNRULE %d\n",lemp
->nrule
); lineno
++;
3297 fprintf(out
,"#define YYERRORSYMBOL %d\n",lemp
->errsym
->index
); lineno
++;
3298 fprintf(out
,"#define YYERRSYMDT yy%d\n",lemp
->errsym
->dtnum
); lineno
++;
3299 if( lemp
->has_fallback
){
3300 fprintf(out
,"#define YYFALLBACK 1\n"); lineno
++;
3302 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3304 /* Generate the action table and its associates:
3306 ** yy_action[] A single table containing all actions.
3307 ** yy_lookahead[] A table containing the lookahead for each entry in
3308 ** yy_action. Used to detect hash collisions.
3309 ** yy_shift_ofst[] For each state, the offset into yy_action for
3310 ** shifting terminals.
3311 ** yy_reduce_ofst[] For each state, the offset into yy_action for
3312 ** shifting non-terminals after a reduce.
3313 ** yy_default[] Default action for each state.
3316 /* Compute the actions on all states and count them up */
3317 ax
= malloc( sizeof(ax
[0])*lemp
->nstate
*2 );
3319 fprintf(stderr
,"malloc failed\n");
3322 for(i
=0; i
<lemp
->nstate
; i
++){
3323 stp
= lemp
->sorted
[i
];
3324 stp
->nTknAct
= stp
->nNtAct
= 0;
3325 stp
->iDflt
= lemp
->nstate
+ lemp
->nrule
;
3326 stp
->iTknOfst
= NO_OFFSET
;
3327 stp
->iNtOfst
= NO_OFFSET
;
3328 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3329 if( compute_action(lemp
,ap
)>=0 ){
3330 if( ap
->sp
->index
<lemp
->nterminal
){
3332 }else if( ap
->sp
->index
<lemp
->nsymbol
){
3335 stp
->iDflt
= compute_action(lemp
, ap
);
3341 ax
[i
*2].nAction
= stp
->nTknAct
;
3342 ax
[i
*2+1].stp
= stp
;
3343 ax
[i
*2+1].isTkn
= 0;
3344 ax
[i
*2+1].nAction
= stp
->nNtAct
;
3346 mxTknOfst
= mnTknOfst
= 0;
3347 mxNtOfst
= mnNtOfst
= 0;
3349 /* Compute the action table. In order to try to keep the size of the
3350 ** action table to a minimum, the heuristic of placing the largest action
3351 ** sets first is used.
3353 qsort(ax
, lemp
->nstate
*2, sizeof(ax
[0]), axset_compare
);
3354 pActtab
= acttab_alloc();
3355 for(i
=0; i
<lemp
->nstate
*2 && ax
[i
].nAction
>0; i
++){
3358 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3360 if( ap
->sp
->index
>=lemp
->nterminal
) continue;
3361 action
= compute_action(lemp
, ap
);
3362 if( action
<0 ) continue;
3363 acttab_action(pActtab
, ap
->sp
->index
, action
);
3365 stp
->iTknOfst
= acttab_insert(pActtab
);
3366 if( stp
->iTknOfst
<mnTknOfst
) mnTknOfst
= stp
->iTknOfst
;
3367 if( stp
->iTknOfst
>mxTknOfst
) mxTknOfst
= stp
->iTknOfst
;
3369 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3371 if( ap
->sp
->index
<lemp
->nterminal
) continue;
3372 if( ap
->sp
->index
==lemp
->nsymbol
) continue;
3373 action
= compute_action(lemp
, ap
);
3374 if( action
<0 ) continue;
3375 acttab_action(pActtab
, ap
->sp
->index
, action
);
3377 stp
->iNtOfst
= acttab_insert(pActtab
);
3378 if( stp
->iNtOfst
<mnNtOfst
) mnNtOfst
= stp
->iNtOfst
;
3379 if( stp
->iNtOfst
>mxNtOfst
) mxNtOfst
= stp
->iNtOfst
;
3384 /* Output the yy_action table */
3385 fprintf(out
,"static YYACTIONTYPE yy_action[] = {\n"); lineno
++;
3386 n
= acttab_size(pActtab
);
3387 for(i
=j
=0; i
<n
; i
++){
3388 int action
= acttab_yyaction(pActtab
, i
);
3389 if( action
<0 ) action
= lemp
->nsymbol
+ lemp
->nrule
+ 2;
3390 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3391 fprintf(out
, " %4d,", action
);
3392 if( j
==9 || i
==n
-1 ){
3393 fprintf(out
, "\n"); lineno
++;
3399 fprintf(out
, "};\n"); lineno
++;
3401 /* Output the yy_lookahead table */
3402 fprintf(out
,"static YYCODETYPE yy_lookahead[] = {\n"); lineno
++;
3403 for(i
=j
=0; i
<n
; i
++){
3404 int la
= acttab_yylookahead(pActtab
, i
);
3405 if( la
<0 ) la
= lemp
->nsymbol
;
3406 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3407 fprintf(out
, " %4d,", la
);
3408 if( j
==9 || i
==n
-1 ){
3409 fprintf(out
, "\n"); lineno
++;
3415 fprintf(out
, "};\n"); lineno
++;
3417 /* Output the yy_shift_ofst[] table */
3418 fprintf(out
, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst
-1); lineno
++;
3419 fprintf(out
, "static %s yy_shift_ofst[] = {\n",
3420 minimum_size_type(mnTknOfst
-1, mxTknOfst
)); lineno
++;
3422 for(i
=j
=0; i
<n
; i
++){
3424 stp
= lemp
->sorted
[i
];
3425 ofst
= stp
->iTknOfst
;
3426 if( ofst
==NO_OFFSET
) ofst
= mnTknOfst
- 1;
3427 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3428 fprintf(out
, " %4d,", ofst
);
3429 if( j
==9 || i
==n
-1 ){
3430 fprintf(out
, "\n"); lineno
++;
3436 fprintf(out
, "};\n"); lineno
++;
3438 /* Output the yy_reduce_ofst[] table */
3439 fprintf(out
, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst
-1); lineno
++;
3440 fprintf(out
, "static %s yy_reduce_ofst[] = {\n",
3441 minimum_size_type(mnNtOfst
-1, mxNtOfst
)); lineno
++;
3443 for(i
=j
=0; i
<n
; i
++){
3445 stp
= lemp
->sorted
[i
];
3446 ofst
= stp
->iNtOfst
;
3447 if( ofst
==NO_OFFSET
) ofst
= mnNtOfst
- 1;
3448 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3449 fprintf(out
, " %4d,", ofst
);
3450 if( j
==9 || i
==n
-1 ){
3451 fprintf(out
, "\n"); lineno
++;
3457 fprintf(out
, "};\n"); lineno
++;
3459 /* Output the default action table */
3460 fprintf(out
, "static YYACTIONTYPE yy_default[] = {\n"); lineno
++;
3462 for(i
=j
=0; i
<n
; i
++){
3463 stp
= lemp
->sorted
[i
];
3464 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3465 fprintf(out
, " %4d,", stp
->iDflt
);
3466 if( j
==9 || i
==n
-1 ){
3467 fprintf(out
, "\n"); lineno
++;
3473 fprintf(out
, "};\n"); lineno
++;
3474 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3476 /* Generate the table of fallback tokens.
3478 if( lemp
->has_fallback
){
3479 for(i
=0; i
<lemp
->nterminal
; i
++){
3480 struct symbol
*p
= lemp
->symbols
[i
];
3481 if( p
->fallback
==0 ){
3482 fprintf(out
, " 0, /* %10s => nothing */\n", p
->name
);
3484 fprintf(out
, " %3d, /* %10s => %s */\n", p
->fallback
->index
,
3485 p
->name
, p
->fallback
->name
);
3490 tplt_xfer(lemp
->name
, in
, out
, &lineno
);
3492 /* Generate a table containing the symbolic name of every symbol
3494 for(i
=0; i
<lemp
->nsymbol
; i
++){
3495 sprintf(line
,"\"%s\",",lemp
->symbols
[i
]->name
);
3496 fprintf(out
," %-15s",line
);
3497 if( (i
&3)==3 ){ fprintf(out
,"\n"); lineno
++; }
3499 if( (i
&3)!=0 ){ fprintf(out
,"\n"); lineno
++; }
3500 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3502 /* Generate a table containing a text string that describes every
3503 ** rule in the rule set of the grammer. This information is used
3504 ** when tracing REDUCE actions.
3506 for(i
=0, rp
=lemp
->rule
; rp
; rp
=rp
->next
, i
++){
3507 assert( rp
->index
==i
);
3508 fprintf(out
," /* %3d */ \"%s ::=", i
, rp
->lhs
->name
);
3509 for(j
=0; j
<rp
->nrhs
; j
++) fprintf(out
," %s",rp
->rhs
[j
]->name
);
3510 fprintf(out
,"\",\n"); lineno
++;
3512 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3514 /* Generate code which executes every time a symbol is popped from
3515 ** the stack while processing errors or while destroying the parser.
3516 ** (In other words, generate the %destructor actions)
3518 if( lemp
->tokendest
){
3519 for(i
=0; i
<lemp
->nsymbol
; i
++){
3520 struct symbol
*sp
= lemp
->symbols
[i
];
3521 if( sp
==0 || sp
->type
!=TERMINAL
) continue;
3522 fprintf(out
," case %d:\n",sp
->index
); lineno
++;
3524 for(i
=0; i
<lemp
->nsymbol
&& lemp
->symbols
[i
]->type
!=TERMINAL
; i
++);
3525 if( i
<lemp
->nsymbol
){
3526 emit_destructor_code(out
,lemp
->symbols
[i
],lemp
,&lineno
);
3527 fprintf(out
," break;\n"); lineno
++;
3530 for(i
=0; i
<lemp
->nsymbol
; i
++){
3531 struct symbol
*sp
= lemp
->symbols
[i
];
3532 if( sp
==0 || sp
->type
==TERMINAL
|| sp
->destructor
==0 ) continue;
3533 fprintf(out
," case %d:\n",sp
->index
); lineno
++;
3534 emit_destructor_code(out
,lemp
->symbols
[i
],lemp
,&lineno
);
3535 fprintf(out
," break;\n"); lineno
++;
3537 if( lemp
->vardest
){
3538 struct symbol
*dflt_sp
= 0;
3539 for(i
=0; i
<lemp
->nsymbol
; i
++){
3540 struct symbol
*sp
= lemp
->symbols
[i
];
3541 if( sp
==0 || sp
->type
==TERMINAL
||
3542 sp
->index
<=0 || sp
->destructor
!=0 ) continue;
3543 fprintf(out
," case %d:\n",sp
->index
); lineno
++;
3547 emit_destructor_code(out
,dflt_sp
,lemp
,&lineno
);
3548 fprintf(out
," break;\n"); lineno
++;
3551 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3553 /* Generate code which executes whenever the parser stack overflows */
3554 tplt_print(out
,lemp
,lemp
->overflow
,lemp
->overflowln
,&lineno
);
3555 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3557 /* Generate the table of rule information
3559 ** Note: This code depends on the fact that rules are number
3560 ** sequentually beginning with 0.
3562 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
3563 fprintf(out
," { %d, %d },\n",rp
->lhs
->index
,rp
->nrhs
); lineno
++;
3565 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3567 /* Generate code which execution during each REDUCE action */
3568 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
3569 fprintf(out
," case %d:\n",rp
->index
); lineno
++;
3570 emit_code(out
,rp
,lemp
,&lineno
);
3571 fprintf(out
," break;\n"); lineno
++;
3573 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3575 /* Generate code which executes if a parse fails */
3576 tplt_print(out
,lemp
,lemp
->failure
,lemp
->failureln
,&lineno
);
3577 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3579 /* Generate code which executes when a syntax error occurs */
3580 tplt_print(out
,lemp
,lemp
->error
,lemp
->errorln
,&lineno
);
3581 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3583 /* Generate code which executes when the parser accepts its input */
3584 tplt_print(out
,lemp
,lemp
->accept
,lemp
->acceptln
,&lineno
);
3585 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3587 /* Append any addition code the user desires */
3588 tplt_print(out
,lemp
,lemp
->extracode
,lemp
->extracodeln
,&lineno
);
3595 /* Generate a header file for the parser */
3596 void ReportHeader(lemp
)
3601 char line
[LINESIZE
];
3602 char pattern
[LINESIZE
];
3605 if( lemp
->tokenprefix
) prefix
= lemp
->tokenprefix
;
3607 in
= file_open(lemp
,".h","r");
3609 for(i
=1; i
<lemp
->nterminal
&& fgets(line
,LINESIZE
,in
); i
++){
3610 sprintf(pattern
,"#define %s%-30s %2d\n",prefix
,lemp
->symbols
[i
]->name
,i
);
3611 if( strcmp(line
,pattern
) ) break;
3614 if( i
==lemp
->nterminal
){
3615 /* No change in the file. Don't rewrite it. */
3619 out
= file_open(lemp
,".h","w");
3621 for(i
=1; i
<lemp
->nterminal
; i
++){
3622 fprintf(out
,"#define %s%-30s %2d\n",prefix
,lemp
->symbols
[i
]->name
,i
);
3629 /* Reduce the size of the action tables, if possible, by making use
3632 ** In this version, we take the most frequent REDUCE action and make
3633 ** it the default. Only default a reduce if there are more than one.
3635 void CompressTables(lemp
)
3639 struct action
*ap
, *ap2
;
3640 struct rule
*rp
, *rp2
, *rbest
;
3644 for(i
=0; i
<lemp
->nstate
; i
++){
3645 stp
= lemp
->sorted
[i
];
3649 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3650 if( ap
->type
!=REDUCE
) continue;
3652 if( rp
==rbest
) continue;
3654 for(ap2
=ap
->next
; ap2
; ap2
=ap2
->next
){
3655 if( ap2
->type
!=REDUCE
) continue;
3657 if( rp2
==rbest
) continue;
3666 /* Do not make a default if the number of rules to default
3667 ** is not at least 2 */
3668 if( nbest
<2 ) continue;
3671 /* Combine matching REDUCE actions into a single default */
3672 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3673 if( ap
->type
==REDUCE
&& ap
->x
.rp
==rbest
) break;
3676 ap
->sp
= Symbol_new("{default}");
3677 for(ap
=ap
->next
; ap
; ap
=ap
->next
){
3678 if( ap
->type
==REDUCE
&& ap
->x
.rp
==rbest
) ap
->type
= NOT_USED
;
3680 stp
->ap
= Action_sort(stp
->ap
);
3684 /***************** From the file "set.c" ************************************/
3686 ** Set manipulation routines for the LEMON parser generator.
3689 static int global_size
= 0;
3691 /* Set the set size */
3698 /* Allocate a new set */
3702 s
= (char*)malloc( global_size
);
3706 for(i
=0; i
<global_size
; i
++) s
[i
] = 0;
3710 /* Deallocate a set */
3717 /* Add a new element to the set. Return TRUE if the element was added
3718 ** and FALSE if it was already there. */
3729 /* Add every element of s2 to s1. Return TRUE if s1 changes. */
3736 for(i
=0; i
<global_size
; i
++){
3737 if( s2
[i
]==0 ) continue;
3745 /********************** From the file "table.c" ****************************/
3747 ** All code in this file has been automatically generated
3748 ** from a specification in the file
3750 ** by the associative array code building program "aagen".
3751 ** Do not edit this file! Instead, edit the specification
3752 ** file, then rerun aagen.
3755 ** Code for processing tables in the LEMON parser generator.
3758 PRIVATE
int strhash(x
)
3762 while( *x
) h
= h
*13u + (unsigned int) *(x
++);
3766 /* Works like strdup, sort of. Save a string in malloced memory, but
3767 ** keep strings in a table so that the same string is not in more
3775 z
= Strsafe_find(y
);
3776 if( z
==0 && (z
=malloc( strlen(y
)+1 ))!=0 ){
3784 /* There is one instance of the following structure for each
3785 ** associative array of type "x1".
3788 int size
; /* The number of available slots. */
3789 /* Must be a power of 2 greater than or */
3791 int count
; /* Number of currently slots filled */
3792 struct s_x1node
*tbl
; /* The data stored here */
3793 struct s_x1node
**ht
; /* Hash table for lookups */
3796 /* There is one instance of this structure for every data element
3797 ** in an associative array of type "x1".
3799 typedef struct s_x1node
{
3800 char *data
; /* The data */
3801 struct s_x1node
*next
; /* Next entry with the same hash */
3802 struct s_x1node
**from
; /* Previous link */
3805 /* There is only one instance of the array, which is the following */
3806 static struct s_x1
*x1a
;
3808 /* Allocate a new associative array */
3809 void Strsafe_init(){
3811 x1a
= (struct s_x1
*)malloc( sizeof(struct s_x1
) );
3815 x1a
->tbl
= (x1node
*)malloc(
3816 (sizeof(x1node
) + sizeof(x1node
*))*1024 );
3822 x1a
->ht
= (x1node
**)&(x1a
->tbl
[1024]);
3823 for(i
=0; i
<1024; i
++) x1a
->ht
[i
] = 0;
3827 /* Insert a new record into the array. Return TRUE if successful.
3828 ** Prior data with the same key is NOT overwritten */
3829 int Strsafe_insert(data
)
3836 if( x1a
==0 ) return 0;
3838 h
= ph
& (x1a
->size
-1);
3841 if( strcmp(np
->data
,data
)==0 ){
3842 /* An existing entry with the same key is found. */
3843 /* Fail because overwrite is not allows. */
3848 if( x1a
->count
>=x1a
->size
){
3849 /* Need to make the hash table bigger */
3852 array
.size
= size
= x1a
->size
*2;
3853 array
.count
= x1a
->count
;
3854 array
.tbl
= (x1node
*)malloc(
3855 (sizeof(x1node
) + sizeof(x1node
*))*size
);
3856 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
3857 array
.ht
= (x1node
**)&(array
.tbl
[size
]);
3858 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
3859 for(i
=0; i
<x1a
->count
; i
++){
3860 x1node
*oldnp
, *newnp
;
3861 oldnp
= &(x1a
->tbl
[i
]);
3862 h
= strhash(oldnp
->data
) & (size
-1);
3863 newnp
= &(array
.tbl
[i
]);
3864 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
3865 newnp
->next
= array
.ht
[h
];
3866 newnp
->data
= oldnp
->data
;
3867 newnp
->from
= &(array
.ht
[h
]);
3868 array
.ht
[h
] = newnp
;
3871 /* *x1a = array; *//* copy 'array' */
3872 memcpy(x1a
, &array
, sizeof(array
));
3874 /* Insert the new data */
3875 h
= ph
& (x1a
->size
-1);
3876 np
= &(x1a
->tbl
[x1a
->count
++]);
3878 if( x1a
->ht
[h
] ) x1a
->ht
[h
]->from
= &(np
->next
);
3879 np
->next
= x1a
->ht
[h
];
3881 np
->from
= &(x1a
->ht
[h
]);
3885 /* Return a pointer to data assigned to the given key. Return NULL
3886 ** if no such key. */
3887 char *Strsafe_find(key
)
3893 if( x1a
==0 ) return 0;
3894 h
= strhash(key
) & (x1a
->size
-1);
3897 if( strcmp(np
->data
,key
)==0 ) break;
3900 return np
? np
->data
: 0;
3903 /* Return a pointer to the (terminal or nonterminal) symbol "x".
3904 ** Create a new symbol if this is the first time "x" has been seen.
3906 struct symbol
*Symbol_new(x
)
3911 sp
= Symbol_find(x
);
3913 sp
= (struct symbol
*)malloc( sizeof(struct symbol
) );
3915 sp
->name
= Strsafe(x
);
3916 sp
->type
= isupper(*x
) ? TERMINAL
: NONTERMINAL
;
3922 sp
->lambda
= Bo_FALSE
;
3925 Symbol_insert(sp
,sp
->name
);
3930 /* Compare two symbols for working purposes
3932 ** Symbols that begin with upper case letters (terminals or tokens)
3933 ** must sort before symbols that begin with lower case letters
3934 ** (non-terminals). Other than that, the order does not matter.
3936 ** We find experimentally that leaving the symbols in their original
3937 ** order (the order they appeared in the grammar file) gives the
3938 ** smallest parser tables in SQLite.
3940 int Symbolcmpp(struct symbol
**a
, struct symbol
**b
){
3941 int i1
= (**a
).index
+ 10000000*((**a
).name
[0]>'Z');
3942 int i2
= (**b
).index
+ 10000000*((**b
).name
[0]>'Z');
3946 /* There is one instance of the following structure for each
3947 ** associative array of type "x2".
3950 int size
; /* The number of available slots. */
3951 /* Must be a power of 2 greater than or */
3953 int count
; /* Number of currently slots filled */
3954 struct s_x2node
*tbl
; /* The data stored here */
3955 struct s_x2node
**ht
; /* Hash table for lookups */
3958 /* There is one instance of this structure for every data element
3959 ** in an associative array of type "x2".
3961 typedef struct s_x2node
{
3962 struct symbol
*data
; /* The data */
3963 char *key
; /* The key */
3964 struct s_x2node
*next
; /* Next entry with the same hash */
3965 struct s_x2node
**from
; /* Previous link */
3968 /* There is only one instance of the array, which is the following */
3969 static struct s_x2
*x2a
;
3971 /* Allocate a new associative array */
3974 x2a
= (struct s_x2
*)malloc( sizeof(struct s_x2
) );
3978 x2a
->tbl
= (x2node
*)malloc(
3979 (sizeof(x2node
) + sizeof(x2node
*))*128 );
3985 x2a
->ht
= (x2node
**)&(x2a
->tbl
[128]);
3986 for(i
=0; i
<128; i
++) x2a
->ht
[i
] = 0;
3990 /* Insert a new record into the array. Return TRUE if successful.
3991 ** Prior data with the same key is NOT overwritten */
3992 int Symbol_insert(data
,key
)
3993 struct symbol
*data
;
4000 if( x2a
==0 ) return 0;
4002 h
= ph
& (x2a
->size
-1);
4005 if( strcmp(np
->key
,key
)==0 ){
4006 /* An existing entry with the same key is found. */
4007 /* Fail because overwrite is not allows. */
4012 if( x2a
->count
>=x2a
->size
){
4013 /* Need to make the hash table bigger */
4016 array
.size
= size
= x2a
->size
*2;
4017 array
.count
= x2a
->count
;
4018 array
.tbl
= (x2node
*)malloc(
4019 (sizeof(x2node
) + sizeof(x2node
*))*size
);
4020 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4021 array
.ht
= (x2node
**)&(array
.tbl
[size
]);
4022 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4023 for(i
=0; i
<x2a
->count
; i
++){
4024 x2node
*oldnp
, *newnp
;
4025 oldnp
= &(x2a
->tbl
[i
]);
4026 h
= strhash(oldnp
->key
) & (size
-1);
4027 newnp
= &(array
.tbl
[i
]);
4028 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4029 newnp
->next
= array
.ht
[h
];
4030 newnp
->key
= oldnp
->key
;
4031 newnp
->data
= oldnp
->data
;
4032 newnp
->from
= &(array
.ht
[h
]);
4033 array
.ht
[h
] = newnp
;
4036 /* *x2a = array; *//* copy 'array' */
4037 memcpy(x2a
, &array
, sizeof(array
));
4039 /* Insert the new data */
4040 h
= ph
& (x2a
->size
-1);
4041 np
= &(x2a
->tbl
[x2a
->count
++]);
4044 if( x2a
->ht
[h
] ) x2a
->ht
[h
]->from
= &(np
->next
);
4045 np
->next
= x2a
->ht
[h
];
4047 np
->from
= &(x2a
->ht
[h
]);
4051 /* Return a pointer to data assigned to the given key. Return NULL
4052 ** if no such key. */
4053 struct symbol
*Symbol_find(key
)
4059 if( x2a
==0 ) return 0;
4060 h
= strhash(key
) & (x2a
->size
-1);
4063 if( strcmp(np
->key
,key
)==0 ) break;
4066 return np
? np
->data
: 0;
4069 /* Return the n-th data. Return NULL if n is out of range. */
4070 struct symbol
*Symbol_Nth(n
)
4073 struct symbol
*data
;
4074 if( x2a
&& n
>0 && n
<=x2a
->count
){
4075 data
= x2a
->tbl
[n
-1].data
;
4082 /* Return the size of the array */
4085 return x2a
? x2a
->count
: 0;
4088 /* Return an array of pointers to all data in the table.
4089 ** The array is obtained from malloc. Return NULL if memory allocation
4090 ** problems, or if the array is empty. */
4091 struct symbol
**Symbol_arrayof()
4093 struct symbol
**array
;
4095 if( x2a
==0 ) return 0;
4097 array
= (struct symbol
**)malloc( sizeof(struct symbol
*)*size
);
4099 for(i
=0; i
<size
; i
++) array
[i
] = x2a
->tbl
[i
].data
;
4104 /* Compare two configurations */
4110 x
= a
->rp
->index
- b
->rp
->index
;
4111 if( x
==0 ) x
= a
->dot
- b
->dot
;
4115 /* Compare two states */
4116 PRIVATE
int statecmp(a
,b
)
4121 for(rc
=0; rc
==0 && a
&& b
; a
=a
->bp
, b
=b
->bp
){
4122 rc
= a
->rp
->index
- b
->rp
->index
;
4123 if( rc
==0 ) rc
= a
->dot
- b
->dot
;
4133 PRIVATE
int statehash(a
)
4138 h
= h
*571u + (unsigned int)a
->rp
->index
*37u + (unsigned int)a
->dot
;
4144 /* Allocate a new state structure */
4145 struct state
*State_new()
4148 new = (struct state
*)malloc( sizeof(struct state
) );
4153 /* There is one instance of the following structure for each
4154 ** associative array of type "x3".
4157 int size
; /* The number of available slots. */
4158 /* Must be a power of 2 greater than or */
4160 int count
; /* Number of currently slots filled */
4161 struct s_x3node
*tbl
; /* The data stored here */
4162 struct s_x3node
**ht
; /* Hash table for lookups */
4165 /* There is one instance of this structure for every data element
4166 ** in an associative array of type "x3".
4168 typedef struct s_x3node
{
4169 struct state
*data
; /* The data */
4170 struct config
*key
; /* The key */
4171 struct s_x3node
*next
; /* Next entry with the same hash */
4172 struct s_x3node
**from
; /* Previous link */
4175 /* There is only one instance of the array, which is the following */
4176 static struct s_x3
*x3a
;
4178 /* Allocate a new associative array */
4181 x3a
= (struct s_x3
*)malloc( sizeof(struct s_x3
) );
4185 x3a
->tbl
= (x3node
*)malloc(
4186 (sizeof(x3node
) + sizeof(x3node
*))*128 );
4192 x3a
->ht
= (x3node
**)&(x3a
->tbl
[128]);
4193 for(i
=0; i
<128; i
++) x3a
->ht
[i
] = 0;
4197 /* Insert a new record into the array. Return TRUE if successful.
4198 ** Prior data with the same key is NOT overwritten */
4199 int State_insert(data
,key
)
4207 if( x3a
==0 ) return 0;
4208 ph
= statehash(key
);
4209 h
= ph
& (x3a
->size
-1);
4212 if( statecmp(np
->key
,key
)==0 ){
4213 /* An existing entry with the same key is found. */
4214 /* Fail because overwrite is not allows. */
4219 if( x3a
->count
>=x3a
->size
){
4220 /* Need to make the hash table bigger */
4223 array
.size
= size
= x3a
->size
*2;
4224 array
.count
= x3a
->count
;
4225 array
.tbl
= (x3node
*)malloc(
4226 (sizeof(x3node
) + sizeof(x3node
*))*size
);
4227 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4228 array
.ht
= (x3node
**)&(array
.tbl
[size
]);
4229 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4230 for(i
=0; i
<x3a
->count
; i
++){
4231 x3node
*oldnp
, *newnp
;
4232 oldnp
= &(x3a
->tbl
[i
]);
4233 h
= statehash(oldnp
->key
) & (size
-1);
4234 newnp
= &(array
.tbl
[i
]);
4235 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4236 newnp
->next
= array
.ht
[h
];
4237 newnp
->key
= oldnp
->key
;
4238 newnp
->data
= oldnp
->data
;
4239 newnp
->from
= &(array
.ht
[h
]);
4240 array
.ht
[h
] = newnp
;
4243 /* *x3a = array; *//* copy 'array' */
4244 memcpy(x3a
, &array
, sizeof(array
));
4246 /* Insert the new data */
4247 h
= ph
& (x3a
->size
-1);
4248 np
= &(x3a
->tbl
[x3a
->count
++]);
4251 if( x3a
->ht
[h
] ) x3a
->ht
[h
]->from
= &(np
->next
);
4252 np
->next
= x3a
->ht
[h
];
4254 np
->from
= &(x3a
->ht
[h
]);
4258 /* Return a pointer to data assigned to the given key. Return NULL
4259 ** if no such key. */
4260 struct state
*State_find(key
)
4266 if( x3a
==0 ) return 0;
4267 h
= statehash(key
) & (x3a
->size
-1);
4270 if( statecmp(np
->key
,key
)==0 ) break;
4273 return np
? np
->data
: 0;
4276 /* Return the size of the array */
4277 int State_count(void)
4279 return x3a
? x3a
->count
: 0;
4282 /* Return an array of pointers to all data in the table.
4283 ** The array is obtained from malloc. Return NULL if memory allocation
4284 ** problems, or if the array is empty. */
4285 struct state
**State_arrayof()
4287 struct state
**array
;
4289 if( x3a
==0 ) return 0;
4291 array
= (struct state
**)malloc( sizeof(struct state
*)*size
);
4293 for(i
=0; i
<size
; i
++) array
[i
] = x3a
->tbl
[i
].data
;
4298 /* Hash a configuration */
4299 PRIVATE
int confighash(a
)
4303 h
= h
*571 + a
->rp
->index
*37 + a
->dot
;
4307 /* There is one instance of the following structure for each
4308 ** associative array of type "x4".
4311 int size
; /* The number of available slots. */
4312 /* Must be a power of 2 greater than or */
4314 int count
; /* Number of currently slots filled */
4315 struct s_x4node
*tbl
; /* The data stored here */
4316 struct s_x4node
**ht
; /* Hash table for lookups */
4319 /* There is one instance of this structure for every data element
4320 ** in an associative array of type "x4".
4322 typedef struct s_x4node
{
4323 struct config
*data
; /* The data */
4324 struct s_x4node
*next
; /* Next entry with the same hash */
4325 struct s_x4node
**from
; /* Previous link */
4328 /* There is only one instance of the array, which is the following */
4329 static struct s_x4
*x4a
;
4331 /* Allocate a new associative array */
4332 void Configtable_init(){
4334 x4a
= (struct s_x4
*)malloc( sizeof(struct s_x4
) );
4338 x4a
->tbl
= (x4node
*)malloc(
4339 (sizeof(x4node
) + sizeof(x4node
*))*64 );
4345 x4a
->ht
= (x4node
**)&(x4a
->tbl
[64]);
4346 for(i
=0; i
<64; i
++) x4a
->ht
[i
] = 0;
4350 /* Insert a new record into the array. Return TRUE if successful.
4351 ** Prior data with the same key is NOT overwritten */
4352 int Configtable_insert(data
)
4353 struct config
*data
;
4359 if( x4a
==0 ) return 0;
4360 ph
= confighash(data
);
4361 h
= ph
& (x4a
->size
-1);
4364 if( Configcmp(np
->data
,data
)==0 ){
4365 /* An existing entry with the same key is found. */
4366 /* Fail because overwrite is not allows. */
4371 if( x4a
->count
>=x4a
->size
){
4372 /* Need to make the hash table bigger */
4375 array
.size
= size
= x4a
->size
*2;
4376 array
.count
= x4a
->count
;
4377 array
.tbl
= (x4node
*)malloc(
4378 (sizeof(x4node
) + sizeof(x4node
*))*size
);
4379 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4380 array
.ht
= (x4node
**)&(array
.tbl
[size
]);
4381 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4382 for(i
=0; i
<x4a
->count
; i
++){
4383 x4node
*oldnp
, *newnp
;
4384 oldnp
= &(x4a
->tbl
[i
]);
4385 h
= confighash(oldnp
->data
) & (size
-1);
4386 newnp
= &(array
.tbl
[i
]);
4387 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4388 newnp
->next
= array
.ht
[h
];
4389 newnp
->data
= oldnp
->data
;
4390 newnp
->from
= &(array
.ht
[h
]);
4391 array
.ht
[h
] = newnp
;
4394 /* *x4a = array; *//* copy 'array' */
4395 memcpy(x4a
, &array
, sizeof(array
));
4397 /* Insert the new data */
4398 h
= ph
& (x4a
->size
-1);
4399 np
= &(x4a
->tbl
[x4a
->count
++]);
4401 if( x4a
->ht
[h
] ) x4a
->ht
[h
]->from
= &(np
->next
);
4402 np
->next
= x4a
->ht
[h
];
4404 np
->from
= &(x4a
->ht
[h
]);
4408 /* Return a pointer to data assigned to the given key. Return NULL
4409 ** if no such key. */
4410 struct config
*Configtable_find(key
)
4416 if( x4a
==0 ) return 0;
4417 h
= confighash(key
) & (x4a
->size
-1);
4420 if( Configcmp(np
->data
,key
)==0 ) break;
4423 return np
? np
->data
: 0;
4426 /* Remove all data from the table. Pass each data to the function "f"
4427 ** as it is removed. ("f" may be null to avoid this step.) */
4428 void Configtable_clear(f
)
4429 int(*f
)(/* struct config * */);
4432 if( x4a
==0 || x4a
->count
==0 ) return;
4433 if( f
) for(i
=0; i
<x4a
->count
; i
++) (*f
)(x4a
->tbl
[i
].data
);
4434 for(i
=0; i
<x4a
->size
; i
++) x4a
->ht
[i
] = 0;