2 ** This file contains all sources (including headers) to the LEMON
3 ** LALR(1) parser generator. The sources have been combined into a
4 ** single file to make it easy to include LEMON in the source tree
5 ** and Makefile of another program.
7 ** The author of this program disclaims copyright.
17 # if defined(_WIN32) || defined(WIN32)
26 extern int access(const char *path
, int mode
);
34 /* #define PRIVATE static */
38 #define MAXRHS 5 /* Set low to exercise exception code */
43 static int showPrecedenceConflict
= 0;
44 static char *msort(char*,char**,int(*)(const char*,const char*));
47 ** Compilers are getting increasingly pedantic about type conversions
48 ** as C evolves ever closer to Ada.... To work around the latest problems
49 ** we have to define the following variant of strlen().
51 #define lemonStrlen(X) ((int)strlen(X))
53 /* a few forward declarations... */
58 static struct action
*Action_new(void);
59 static struct action
*Action_sort(struct action
*);
61 /********** From the file "build.h" ************************************/
62 void FindRulePrecedences();
66 void FindFollowSets();
69 /********* From the file "configlist.h" *********************************/
70 void Configlist_init(void);
71 struct config
*Configlist_add(struct rule
*, int);
72 struct config
*Configlist_addbasis(struct rule
*, int);
73 void Configlist_closure(struct lemon
*);
74 void Configlist_sort(void);
75 void Configlist_sortbasis(void);
76 struct config
*Configlist_return(void);
77 struct config
*Configlist_basis(void);
78 void Configlist_eat(struct config
*);
79 void Configlist_reset(void);
81 /********* From the file "error.h" ***************************************/
82 void ErrorMsg(const char *, int,const char *, ...);
84 /****** From the file "option.h" ******************************************/
85 enum option_type
{ OPT_FLAG
=1, OPT_INT
, OPT_DBL
, OPT_STR
,
86 OPT_FFLAG
, OPT_FINT
, OPT_FDBL
, OPT_FSTR
};
88 enum option_type type
;
93 int OptInit(char**,struct s_options
*,FILE*);
99 /******** From the file "parse.h" *****************************************/
100 void Parse(struct lemon
*lemp
);
102 /********* From the file "plink.h" ***************************************/
103 struct plink
*Plink_new(void);
104 void Plink_add(struct plink
**, struct config
*);
105 void Plink_copy(struct plink
**, struct plink
*);
106 void Plink_delete(struct plink
*);
108 /********** From the file "report.h" *************************************/
109 void Reprint(struct lemon
*);
110 void ReportOutput(struct lemon
*);
111 void ReportTable(struct lemon
*, int);
112 void ReportHeader(struct lemon
*);
113 void CompressTables(struct lemon
*);
114 void ResortStates(struct lemon
*);
116 /********** From the file "set.h" ****************************************/
117 void SetSize(int); /* All sets will be of size N */
118 char *SetNew(void); /* A new set for element 0..N */
119 void SetFree(char*); /* Deallocate a set */
120 int SetAdd(char*,int); /* Add element to a set */
121 int SetUnion(char *,char *); /* A <- A U B, thru element N */
122 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
124 /********** From the file "struct.h" *************************************/
126 ** Principal data structures for the LEMON parser generator.
129 typedef enum {LEMON_FALSE
=0, LEMON_TRUE
} Boolean
;
131 /* Symbols (terminals and nonterminals) of the grammar are stored
132 ** in the following: */
145 const char *name
; /* Name of the symbol */
146 int index
; /* Index number for this symbol */
147 enum symbol_type type
; /* Symbols are all either TERMINALS or NTs */
148 struct rule
*rule
; /* Linked list of rules of this (if an NT) */
149 struct symbol
*fallback
; /* fallback token in case this token doesn't parse */
150 int prec
; /* Precedence if defined (-1 otherwise) */
151 enum e_assoc assoc
; /* Associativity if precedence is defined */
152 char *firstset
; /* First-set for all rules of this symbol */
153 Boolean lambda
; /* True if NT and can generate an empty string */
154 int useCnt
; /* Number of times used */
155 char *destructor
; /* Code which executes whenever this symbol is
156 ** popped from the stack during error processing */
157 int destLineno
; /* Line number for start of destructor */
158 char *datatype
; /* The data type of information held by this
159 ** object. Only used if type==NONTERMINAL */
160 int dtnum
; /* The data type number. In the parser, the value
161 ** stack is a union. The .yy%d element of this
162 ** union is the correct data type for this object */
163 /* The following fields are used by MULTITERMINALs only */
164 int nsubsym
; /* Number of constituent symbols in the MULTI */
165 struct symbol
**subsym
; /* Array of constituent symbols */
168 /* Each production rule in the grammar is stored in the following
171 struct symbol
*lhs
; /* Left-hand side of the rule */
172 const char *lhsalias
; /* Alias for the LHS (NULL if none) */
173 int lhsStart
; /* True if left-hand side is the start symbol */
174 int ruleline
; /* Line number for the rule */
175 int nrhs
; /* Number of RHS symbols */
176 struct symbol
**rhs
; /* The RHS symbols */
177 const char **rhsalias
; /* An alias for each RHS symbol (NULL if none) */
178 int line
; /* Line number at which code begins */
179 const char *code
; /* The code executed when this rule is reduced */
180 struct symbol
*precsym
; /* Precedence symbol for this rule */
181 int index
; /* An index number for this rule */
182 Boolean canReduce
; /* True if this rule is ever reduced */
183 struct rule
*nextlhs
; /* Next rule with the same LHS */
184 struct rule
*next
; /* Next rule in the global list */
187 /* A configuration is a production rule of the grammar together with
188 ** a mark (dot) showing how much of that rule has been processed so far.
189 ** Configurations also contain a follow-set which is a list of terminal
190 ** symbols which are allowed to immediately follow the end of the rule.
191 ** Every configuration is recorded as an instance of the following: */
197 struct rule
*rp
; /* The rule upon which the configuration is based */
198 int dot
; /* The parse point */
199 char *fws
; /* Follow-set for this configuration only */
200 struct plink
*fplp
; /* Follow-set forward propagation links */
201 struct plink
*bplp
; /* Follow-set backwards propagation links */
202 struct state
*stp
; /* Pointer to state which contains this */
203 enum cfgstatus status
; /* used during followset and shift computations */
204 struct config
*next
; /* Next configuration in the state */
205 struct config
*bp
; /* The next basis configuration */
213 SSCONFLICT
, /* A shift/shift conflict */
214 SRCONFLICT
, /* Was a reduce, but part of a conflict */
215 RRCONFLICT
, /* Was a reduce, but part of a conflict */
216 SH_RESOLVED
, /* Was a shift. Precedence resolved conflict */
217 RD_RESOLVED
, /* Was reduce. Precedence resolved conflict */
218 NOT_USED
/* Deleted by compression */
221 /* Every shift or reduce operation is stored as one of the following */
223 struct symbol
*sp
; /* The look-ahead symbol */
226 struct state
*stp
; /* The new state, if a shift */
227 struct rule
*rp
; /* The rule, if a reduce */
229 struct action
*next
; /* Next action for this state */
230 struct action
*collide
; /* Next action with the same hash */
233 /* Each state of the generated parser's finite state machine
234 ** is encoded as an instance of the following structure. */
236 struct config
*bp
; /* The basis configurations for this state */
237 struct config
*cfp
; /* All configurations in this set */
238 int statenum
; /* Sequential number for this state */
239 struct action
*ap
; /* Array of actions for this state */
240 int nTknAct
, nNtAct
; /* Number of actions on terminals and nonterminals */
241 int iTknOfst
, iNtOfst
; /* yy_action[] offset for terminals and nonterms */
242 int iDflt
; /* Default action */
244 #define NO_OFFSET (-2147483647)
246 /* A followset propagation link indicates that the contents of one
247 ** configuration followset should be propagated to another whenever
248 ** the first changes. */
250 struct config
*cfp
; /* The configuration to which linked */
251 struct plink
*next
; /* The next propagate link */
254 /* The state vector for the entire parser generator is recorded as
255 ** follows. (LEMON uses no global variables and makes little use of
256 ** static variables. Fields in the following structure can be thought
257 ** of as begin global variables in the program.) */
259 struct state
**sorted
; /* Table of states sorted by state number */
260 struct rule
*rule
; /* List of all rules */
261 int nstate
; /* Number of states */
262 int nrule
; /* Number of rules */
263 int nsymbol
; /* Number of terminal and nonterminal symbols */
264 int nterminal
; /* Number of terminal symbols */
265 struct symbol
**symbols
; /* Sorted array of pointers to symbols */
266 int errorcnt
; /* Number of errors */
267 struct symbol
*errsym
; /* The error symbol */
268 struct symbol
*wildcard
; /* Token that matches anything */
269 char *name
; /* Name of the generated parser */
270 char *arg
; /* Declaration of the 3th argument to parser */
271 char *tokentype
; /* Type of terminal symbols in the parser stack */
272 char *vartype
; /* The default type of non-terminal symbols */
273 char *start
; /* Name of the start symbol for the grammar */
274 char *stacksize
; /* Size of the parser stack */
275 char *include
; /* Code to put at the start of the C file */
276 char *error
; /* Code to execute when an error is seen */
277 char *overflow
; /* Code to execute on a stack overflow */
278 char *failure
; /* Code to execute on parser failure */
279 char *accept
; /* Code to execute when the parser excepts */
280 char *extracode
; /* Code appended to the generated file */
281 char *tokendest
; /* Code to execute to destroy token data */
282 char *vardest
; /* Code for the default non-terminal destructor */
283 char *filename
; /* Name of the input file */
284 char *outname
; /* Name of the current output file */
285 char *tokenprefix
; /* A prefix added to token names in the .h file */
286 int nconflict
; /* Number of parsing conflicts */
287 int tablesize
; /* Size of the parse tables */
288 int basisflag
; /* Print only basis configurations */
289 int has_fallback
; /* True if any %fallback is seen in the grammar */
290 int nolinenosflag
; /* True if #line statements should not be printed */
291 char *argv0
; /* Name of the program */
294 #define MemoryCheck(X) if((X)==0){ \
295 extern void memory_error(); \
299 /**************** From the file "table.h" *********************************/
301 ** All code in this file has been automatically generated
302 ** from a specification in the file
304 ** by the associative array code building program "aagen".
305 ** Do not edit this file! Instead, edit the specification
306 ** file, then rerun aagen.
309 ** Code for processing tables in the LEMON parser generator.
311 /* Routines for handling a strings */
313 const char *Strsafe(const char *);
315 void Strsafe_init(void);
316 int Strsafe_insert(const char *);
317 const char *Strsafe_find(const char *);
319 /* Routines for handling symbols of the grammar */
321 struct symbol
*Symbol_new(const char *);
322 int Symbolcmpp(const void *, const void *);
323 void Symbol_init(void);
324 int Symbol_insert(struct symbol
*, const char *);
325 struct symbol
*Symbol_find(const char *);
326 struct symbol
*Symbol_Nth(int);
327 int Symbol_count(void);
328 struct symbol
**Symbol_arrayof(void);
330 /* Routines to manage the state table */
332 int Configcmp(const char *, const char *);
333 struct state
*State_new(void);
334 void State_init(void);
335 int State_insert(struct state
*, struct config
*);
336 struct state
*State_find(struct config
*);
337 struct state
**State_arrayof(/* */);
339 /* Routines used for efficiency in Configlist_add */
341 void Configtable_init(void);
342 int Configtable_insert(struct config
*);
343 struct config
*Configtable_find(struct config
*);
344 void Configtable_clear(int(*)(struct config
*));
346 /****************** From the file "action.c" *******************************/
348 ** Routines processing parser actions in the LEMON parser generator.
351 /* Allocate a new parser action */
352 static struct action
*Action_new(void){
353 static struct action
*freelist
= 0;
354 struct action
*newaction
;
359 freelist
= (struct action
*)calloc(amt
, sizeof(struct action
));
361 fprintf(stderr
,"Unable to allocate memory for a new parser action.");
364 for(i
=0; i
<amt
-1; i
++) freelist
[i
].next
= &freelist
[i
+1];
365 freelist
[amt
-1].next
= 0;
367 newaction
= freelist
;
368 freelist
= freelist
->next
;
372 /* Compare two actions for sorting purposes. Return negative, zero, or
373 ** positive if the first action is less than, equal to, or greater than
376 static int actioncmp(
381 rc
= ap1
->sp
->index
- ap2
->sp
->index
;
383 rc
= (int)ap1
->type
- (int)ap2
->type
;
385 if( rc
==0 && ap1
->type
==REDUCE
){
386 rc
= ap1
->x
.rp
->index
- ap2
->x
.rp
->index
;
389 rc
= (int) (ap2
- ap1
);
394 /* Sort parser actions */
395 static struct action
*Action_sort(
398 ap
= (struct action
*)msort((char *)ap
,(char **)&ap
->next
,
399 (int(*)(const char*,const char*))actioncmp
);
409 struct action
*newaction
;
410 newaction
= Action_new();
411 newaction
->next
= *app
;
413 newaction
->type
= type
;
416 newaction
->x
.stp
= (struct state
*)arg
;
418 newaction
->x
.rp
= (struct rule
*)arg
;
421 /********************** New code to implement the "acttab" module ***********/
423 ** This module implements routines use to construct the yy_action[] table.
427 ** The state of the yy_action table under construction is an instance of
428 ** the following structure.
430 ** The yy_action table maps the pair (state_number, lookahead) into an
431 ** action_number. The table is an array of integers pairs. The state_number
432 ** determines an initial offset into the yy_action array. The lookahead
433 ** value is then added to this initial offset to get an index X into the
434 ** yy_action array. If the aAction[X].lookahead equals the value of the
435 ** of the lookahead input, then the value of the action_number output is
436 ** aAction[X].action. If the lookaheads do not match then the
437 ** default action for the state_number is returned.
439 ** All actions associated with a single state_number are first entered
440 ** into aLookahead[] using multiple calls to acttab_action(). Then the
441 ** actions for that single state_number are placed into the aAction[]
442 ** array with a single call to acttab_insert(). The acttab_insert() call
443 ** also resets the aLookahead[] array in preparation for the next
446 struct lookahead_action
{
447 int lookahead
; /* Value of the lookahead token */
448 int action
; /* Action to take on the given lookahead */
450 typedef struct acttab acttab
;
452 int nAction
; /* Number of used slots in aAction[] */
453 int nActionAlloc
; /* Slots allocated for aAction[] */
454 struct lookahead_action
455 *aAction
, /* The yy_action[] table under construction */
456 *aLookahead
; /* A single new transaction set */
457 int mnLookahead
; /* Minimum aLookahead[].lookahead */
458 int mnAction
; /* Action associated with mnLookahead */
459 int mxLookahead
; /* Maximum aLookahead[].lookahead */
460 int nLookahead
; /* Used slots in aLookahead[] */
461 int nLookaheadAlloc
; /* Slots allocated in aLookahead[] */
464 /* Return the number of entries in the yy_action table */
465 #define acttab_size(X) ((X)->nAction)
467 /* The value for the N-th entry in yy_action */
468 #define acttab_yyaction(X,N) ((X)->aAction[N].action)
470 /* The value for the N-th entry in yy_lookahead */
471 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
473 /* Free all memory associated with the given acttab */
474 void acttab_free(acttab
*p
){
476 free( p
->aLookahead
);
480 /* Allocate a new acttab structure */
481 acttab
*acttab_alloc(void){
482 acttab
*p
= (acttab
*) calloc( 1, sizeof(*p
) );
484 fprintf(stderr
,"Unable to allocate memory for a new acttab.");
487 memset(p
, 0, sizeof(*p
));
491 /* Add a new action to the current transaction set.
493 ** This routine is called once for each lookahead for a particular
496 void acttab_action(acttab
*p
, int lookahead
, int action
){
497 if( p
->nLookahead
>=p
->nLookaheadAlloc
){
498 p
->nLookaheadAlloc
+= 25;
499 p
->aLookahead
= (struct lookahead_action
*) realloc( p
->aLookahead
,
500 sizeof(p
->aLookahead
[0])*p
->nLookaheadAlloc
);
501 if( p
->aLookahead
==0 ){
502 fprintf(stderr
,"malloc failed\n");
506 if( p
->nLookahead
==0 ){
507 p
->mxLookahead
= lookahead
;
508 p
->mnLookahead
= lookahead
;
509 p
->mnAction
= action
;
511 if( p
->mxLookahead
<lookahead
) p
->mxLookahead
= lookahead
;
512 if( p
->mnLookahead
>lookahead
){
513 p
->mnLookahead
= lookahead
;
514 p
->mnAction
= action
;
517 p
->aLookahead
[p
->nLookahead
].lookahead
= lookahead
;
518 p
->aLookahead
[p
->nLookahead
].action
= action
;
523 ** Add the transaction set built up with prior calls to acttab_action()
524 ** into the current action table. Then reset the transaction set back
525 ** to an empty set in preparation for a new round of acttab_action() calls.
527 ** Return the offset into the action table of the new transaction.
529 int acttab_insert(acttab
*p
){
531 assert( p
->nLookahead
>0 );
533 /* Make sure we have enough space to hold the expanded action table
534 ** in the worst case. The worst case occurs if the transaction set
535 ** must be appended to the current action table
537 n
= p
->mxLookahead
+ 1;
538 if( p
->nAction
+ n
>= p
->nActionAlloc
){
539 int oldAlloc
= p
->nActionAlloc
;
540 p
->nActionAlloc
= p
->nAction
+ n
+ p
->nActionAlloc
+ 20;
541 p
->aAction
= (struct lookahead_action
*) realloc( p
->aAction
,
542 sizeof(p
->aAction
[0])*p
->nActionAlloc
);
544 fprintf(stderr
,"malloc failed\n");
547 for(i
=oldAlloc
; i
<p
->nActionAlloc
; i
++){
548 p
->aAction
[i
].lookahead
= -1;
549 p
->aAction
[i
].action
= -1;
553 /* Scan the existing action table looking for an offset that is a
554 ** duplicate of the current transaction set. Fall out of the loop
555 ** if and when the duplicate is found.
557 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
559 for(i
=p
->nAction
-1; i
>=0; i
--){
560 if( p
->aAction
[i
].lookahead
==p
->mnLookahead
){
561 /* All lookaheads and actions in the aLookahead[] transaction
562 ** must match against the candidate aAction[i] entry. */
563 if( p
->aAction
[i
].action
!=p
->mnAction
) continue;
564 for(j
=0; j
<p
->nLookahead
; j
++){
565 k
= p
->aLookahead
[j
].lookahead
- p
->mnLookahead
+ i
;
566 if( k
<0 || k
>=p
->nAction
) break;
567 if( p
->aLookahead
[j
].lookahead
!=p
->aAction
[k
].lookahead
) break;
568 if( p
->aLookahead
[j
].action
!=p
->aAction
[k
].action
) break;
570 if( j
<p
->nLookahead
) continue;
572 /* No possible lookahead value that is not in the aLookahead[]
573 ** transaction is allowed to match aAction[i] */
575 for(j
=0; j
<p
->nAction
; j
++){
576 if( p
->aAction
[j
].lookahead
<0 ) continue;
577 if( p
->aAction
[j
].lookahead
==j
+p
->mnLookahead
-i
) n
++;
579 if( n
==p
->nLookahead
){
580 break; /* An exact match is found at offset i */
585 /* If no existing offsets exactly match the current transaction, find an
586 ** an empty offset in the aAction[] table in which we can add the
587 ** aLookahead[] transaction.
590 /* Look for holes in the aAction[] table that fit the current
591 ** aLookahead[] transaction. Leave i set to the offset of the hole.
592 ** If no holes are found, i is left at p->nAction, which means the
593 ** transaction will be appended. */
594 for(i
=0; i
<p
->nActionAlloc
- p
->mxLookahead
; i
++){
595 if( p
->aAction
[i
].lookahead
<0 ){
596 for(j
=0; j
<p
->nLookahead
; j
++){
597 k
= p
->aLookahead
[j
].lookahead
- p
->mnLookahead
+ i
;
599 if( p
->aAction
[k
].lookahead
>=0 ) break;
601 if( j
<p
->nLookahead
) continue;
602 for(j
=0; j
<p
->nAction
; j
++){
603 if( p
->aAction
[j
].lookahead
==j
+p
->mnLookahead
-i
) break;
606 break; /* Fits in empty slots */
611 /* Insert transaction set at index i. */
612 for(j
=0; j
<p
->nLookahead
; j
++){
613 k
= p
->aLookahead
[j
].lookahead
- p
->mnLookahead
+ i
;
614 p
->aAction
[k
] = p
->aLookahead
[j
];
615 if( k
>=p
->nAction
) p
->nAction
= k
+1;
619 /* Return the offset that is added to the lookahead in order to get the
620 ** index into yy_action of the action */
621 return i
- p
->mnLookahead
;
624 /********************** From the file "build.c" *****************************/
626 ** Routines to construction the finite state machine for the LEMON
630 /* Find a precedence symbol of every rule in the grammar.
632 ** Those rules which have a precedence symbol coded in the input
633 ** grammar using the "[symbol]" construct will already have the
634 ** rp->precsym field filled. Other rules take as their precedence
635 ** symbol the first RHS symbol with a defined precedence. If there
636 ** are not RHS symbols with a defined precedence, the precedence
637 ** symbol field is left blank.
639 void FindRulePrecedences(struct lemon
*xp
)
642 for(rp
=xp
->rule
; rp
; rp
=rp
->next
){
643 if( rp
->precsym
==0 ){
645 for(i
=0; i
<rp
->nrhs
&& rp
->precsym
==0; i
++){
646 struct symbol
*sp
= rp
->rhs
[i
];
647 if( sp
->type
==MULTITERMINAL
){
648 for(j
=0; j
<sp
->nsubsym
; j
++){
649 if( sp
->subsym
[j
]->prec
>=0 ){
650 rp
->precsym
= sp
->subsym
[j
];
654 }else if( sp
->prec
>=0 ){
655 rp
->precsym
= rp
->rhs
[i
];
663 /* Find all nonterminals which will generate the empty string.
664 ** Then go back and compute the first sets of every nonterminal.
665 ** The first set is the set of all terminal symbols which can begin
666 ** a string generated by that nonterminal.
668 void FindFirstSets(struct lemon
*lemp
)
674 for(i
=0; i
<lemp
->nsymbol
; i
++){
675 lemp
->symbols
[i
]->lambda
= LEMON_FALSE
;
677 for(i
=lemp
->nterminal
; i
<lemp
->nsymbol
; i
++){
678 lemp
->symbols
[i
]->firstset
= SetNew();
681 /* First compute all lambdas */
684 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
685 if( rp
->lhs
->lambda
) continue;
686 for(i
=0; i
<rp
->nrhs
; i
++){
687 struct symbol
*sp
= rp
->rhs
[i
];
688 assert( sp
->type
==NONTERMINAL
|| sp
->lambda
==LEMON_FALSE
);
689 if( sp
->lambda
==LEMON_FALSE
) break;
692 rp
->lhs
->lambda
= LEMON_TRUE
;
698 /* Now compute all first sets */
700 struct symbol
*s1
, *s2
;
702 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
704 for(i
=0; i
<rp
->nrhs
; i
++){
706 if( s2
->type
==TERMINAL
){
707 progress
+= SetAdd(s1
->firstset
,s2
->index
);
709 }else if( s2
->type
==MULTITERMINAL
){
710 for(j
=0; j
<s2
->nsubsym
; j
++){
711 progress
+= SetAdd(s1
->firstset
,s2
->subsym
[j
]->index
);
715 if( s1
->lambda
==LEMON_FALSE
) break;
717 progress
+= SetUnion(s1
->firstset
,s2
->firstset
);
718 if( s2
->lambda
==LEMON_FALSE
) break;
726 /* Compute all LR(0) states for the grammar. Links
727 ** are added to between some states so that the LR(1) follow sets
728 ** can be computed later.
730 PRIVATE
struct state
*getstate(struct lemon
*); /* forward reference */
731 void FindStates(struct lemon
*lemp
)
738 /* Find the start symbol */
740 sp
= Symbol_find(lemp
->start
);
742 ErrorMsg(lemp
->filename
,0,
743 "The specified start symbol \"%s\" is not \
744 in a nonterminal of the grammar. \"%s\" will be used as the start \
745 symbol instead.",lemp
->start
,lemp
->rule
->lhs
->name
);
747 sp
= lemp
->rule
->lhs
;
750 sp
= lemp
->rule
->lhs
;
753 /* Make sure the start symbol doesn't occur on the right-hand side of
754 ** any rule. Report an error if it does. (YACC would generate a new
755 ** start symbol in this case.) */
756 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
758 for(i
=0; i
<rp
->nrhs
; i
++){
759 if( rp
->rhs
[i
]==sp
){ /* FIX ME: Deal with multiterminals */
760 ErrorMsg(lemp
->filename
,0,
761 "The start symbol \"%s\" occurs on the \
762 right-hand side of a rule. This will result in a parser which \
763 does not work properly.",sp
->name
);
769 /* The basis configuration set for the first state
770 ** is all rules which have the start symbol as their
772 for(rp
=sp
->rule
; rp
; rp
=rp
->nextlhs
){
773 struct config
*newcfp
;
775 newcfp
= Configlist_addbasis(rp
,0);
776 SetAdd(newcfp
->fws
,0);
779 /* Compute the first state. All other states will be
780 ** computed automatically during the computation of the first one.
781 ** The returned pointer to the first state is not used. */
782 (void)getstate(lemp
);
786 /* Return a pointer to a state which is described by the configuration
787 ** list which has been built from calls to Configlist_add.
789 PRIVATE
void buildshifts(struct lemon
*, struct state
*); /* Forwd ref */
790 PRIVATE
struct state
*getstate(struct lemon
*lemp
)
792 struct config
*cfp
, *bp
;
795 /* Extract the sorted basis of the new state. The basis was constructed
796 ** by prior calls to "Configlist_addbasis()". */
797 Configlist_sortbasis();
798 bp
= Configlist_basis();
800 /* Get a state with the same basis */
801 stp
= State_find(bp
);
803 /* A state with the same basis already exists! Copy all the follow-set
804 ** propagation links from the state under construction into the
805 ** preexisting state, then return a pointer to the preexisting state */
806 struct config
*x
, *y
;
807 for(x
=bp
, y
=stp
->bp
; x
&& y
; x
=x
->bp
, y
=y
->bp
){
808 Plink_copy(&y
->bplp
,x
->bplp
);
809 Plink_delete(x
->fplp
);
810 x
->fplp
= x
->bplp
= 0;
812 cfp
= Configlist_return();
815 /* This really is a new state. Construct all the details */
816 Configlist_closure(lemp
); /* Compute the configuration closure */
817 Configlist_sort(); /* Sort the configuration closure */
818 cfp
= Configlist_return(); /* Get a pointer to the config list */
819 stp
= State_new(); /* A new state structure */
821 stp
->bp
= bp
; /* Remember the configuration basis */
822 stp
->cfp
= cfp
; /* Remember the configuration closure */
823 stp
->statenum
= lemp
->nstate
++; /* Every state gets a sequence number */
824 stp
->ap
= 0; /* No actions, yet. */
825 State_insert(stp
,stp
->bp
); /* Add to the state table */
826 buildshifts(lemp
,stp
); /* Recursively compute successor states */
832 ** Return true if two symbols are the same.
834 int same_symbol(struct symbol
*a
, struct symbol
*b
)
838 if( a
->type
!=MULTITERMINAL
) return 0;
839 if( b
->type
!=MULTITERMINAL
) return 0;
840 if( a
->nsubsym
!=b
->nsubsym
) return 0;
841 for(i
=0; i
<a
->nsubsym
; i
++){
842 if( a
->subsym
[i
]!=b
->subsym
[i
] ) return 0;
847 /* Construct all successor states to the given state. A "successor"
848 ** state is any state which can be reached by a shift action.
850 PRIVATE
void buildshifts(struct lemon
*lemp
, struct state
*stp
)
852 struct config
*cfp
; /* For looping thru the config closure of "stp" */
853 struct config
*bcfp
; /* For the inner loop on config closure of "stp" */
854 struct config
*newcfg
; /* */
855 struct symbol
*sp
; /* Symbol following the dot in configuration "cfp" */
856 struct symbol
*bsp
; /* Symbol following the dot in configuration "bcfp" */
857 struct state
*newstp
; /* A pointer to a successor state */
859 /* Each configuration becomes complete after it contibutes to a successor
860 ** state. Initially, all configurations are incomplete */
861 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
) cfp
->status
= INCOMPLETE
;
863 /* Loop through all configurations of the state "stp" */
864 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
865 if( cfp
->status
==COMPLETE
) continue; /* Already used by inner loop */
866 if( cfp
->dot
>=cfp
->rp
->nrhs
) continue; /* Can't shift this config */
867 Configlist_reset(); /* Reset the new config set */
868 sp
= cfp
->rp
->rhs
[cfp
->dot
]; /* Symbol after the dot */
870 /* For every configuration in the state "stp" which has the symbol "sp"
871 ** following its dot, add the same configuration to the basis set under
872 ** construction but with the dot shifted one symbol to the right. */
873 for(bcfp
=cfp
; bcfp
; bcfp
=bcfp
->next
){
874 if( bcfp
->status
==COMPLETE
) continue; /* Already used */
875 if( bcfp
->dot
>=bcfp
->rp
->nrhs
) continue; /* Can't shift this one */
876 bsp
= bcfp
->rp
->rhs
[bcfp
->dot
]; /* Get symbol after dot */
877 if( !same_symbol(bsp
,sp
) ) continue; /* Must be same as for "cfp" */
878 bcfp
->status
= COMPLETE
; /* Mark this config as used */
879 newcfg
= Configlist_addbasis(bcfp
->rp
,bcfp
->dot
+1);
880 Plink_add(&newcfg
->bplp
,bcfp
);
883 /* Get a pointer to the state described by the basis configuration set
884 ** constructed in the preceding loop */
885 newstp
= getstate(lemp
);
887 /* The state "newstp" is reached from the state "stp" by a shift action
888 ** on the symbol "sp" */
889 if( sp
->type
==MULTITERMINAL
){
891 for(i
=0; i
<sp
->nsubsym
; i
++){
892 Action_add(&stp
->ap
,SHIFT
,sp
->subsym
[i
],(char*)newstp
);
895 Action_add(&stp
->ap
,SHIFT
,sp
,(char *)newstp
);
901 ** Construct the propagation links
903 void FindLinks(struct lemon
*lemp
)
906 struct config
*cfp
, *other
;
910 /* Housekeeping detail:
911 ** Add to every propagate link a pointer back to the state to
912 ** which the link is attached. */
913 for(i
=0; i
<lemp
->nstate
; i
++){
914 stp
= lemp
->sorted
[i
];
915 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
920 /* Convert all backlinks into forward links. Only the forward
921 ** links are used in the follow-set computation. */
922 for(i
=0; i
<lemp
->nstate
; i
++){
923 stp
= lemp
->sorted
[i
];
924 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){
925 for(plp
=cfp
->bplp
; plp
; plp
=plp
->next
){
927 Plink_add(&other
->fplp
,cfp
);
933 /* Compute all followsets.
935 ** A followset is the set of all symbols which can come immediately
936 ** after a configuration.
938 void FindFollowSets(struct lemon
*lemp
)
946 for(i
=0; i
<lemp
->nstate
; i
++){
947 for(cfp
=lemp
->sorted
[i
]->cfp
; cfp
; cfp
=cfp
->next
){
948 cfp
->status
= INCOMPLETE
;
954 for(i
=0; i
<lemp
->nstate
; i
++){
955 for(cfp
=lemp
->sorted
[i
]->cfp
; cfp
; cfp
=cfp
->next
){
956 if( cfp
->status
==COMPLETE
) continue;
957 for(plp
=cfp
->fplp
; plp
; plp
=plp
->next
){
958 change
= SetUnion(plp
->cfp
->fws
,cfp
->fws
);
960 plp
->cfp
->status
= INCOMPLETE
;
964 cfp
->status
= COMPLETE
;
970 static int resolve_conflict(struct action
*,struct action
*);
972 /* Compute the reduce actions, and resolve conflicts.
974 void FindActions(struct lemon
*lemp
)
982 /* Add all of the reduce actions
983 ** A reduce action is added for each element of the followset of
984 ** a configuration which has its dot at the extreme right.
986 for(i
=0; i
<lemp
->nstate
; i
++){ /* Loop over all states */
987 stp
= lemp
->sorted
[i
];
988 for(cfp
=stp
->cfp
; cfp
; cfp
=cfp
->next
){ /* Loop over all configurations */
989 if( cfp
->rp
->nrhs
==cfp
->dot
){ /* Is dot at extreme right? */
990 for(j
=0; j
<lemp
->nterminal
; j
++){
991 if( SetFind(cfp
->fws
,j
) ){
992 /* Add a reduce action to the state "stp" which will reduce by the
993 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
994 Action_add(&stp
->ap
,REDUCE
,lemp
->symbols
[j
],(char *)cfp
->rp
);
1001 /* Add the accepting token */
1003 sp
= Symbol_find(lemp
->start
);
1004 if( sp
==0 ) sp
= lemp
->rule
->lhs
;
1006 sp
= lemp
->rule
->lhs
;
1008 /* Add to the first state (which is always the starting state of the
1009 ** finite state machine) an action to ACCEPT if the lookahead is the
1010 ** start nonterminal. */
1011 Action_add(&lemp
->sorted
[0]->ap
,ACCEPT
,sp
,0);
1013 /* Resolve conflicts */
1014 for(i
=0; i
<lemp
->nstate
; i
++){
1015 struct action
*ap
, *nap
;
1017 stp
= lemp
->sorted
[i
];
1018 /* assert( stp->ap ); */
1019 stp
->ap
= Action_sort(stp
->ap
);
1020 for(ap
=stp
->ap
; ap
&& ap
->next
; ap
=ap
->next
){
1021 for(nap
=ap
->next
; nap
&& nap
->sp
==ap
->sp
; nap
=nap
->next
){
1022 /* The two actions "ap" and "nap" have the same lookahead.
1023 ** Figure out which one should be used */
1024 lemp
->nconflict
+= resolve_conflict(ap
,nap
);
1029 /* Report an error for each rule that can never be reduced. */
1030 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
) rp
->canReduce
= LEMON_FALSE
;
1031 for(i
=0; i
<lemp
->nstate
; i
++){
1033 for(ap
=lemp
->sorted
[i
]->ap
; ap
; ap
=ap
->next
){
1034 if( ap
->type
==REDUCE
) ap
->x
.rp
->canReduce
= LEMON_TRUE
;
1037 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
1038 if( rp
->canReduce
) continue;
1039 ErrorMsg(lemp
->filename
,rp
->ruleline
,"This rule can not be reduced.\n");
1044 /* Resolve a conflict between the two given actions. If the
1045 ** conflict can't be resolved, return non-zero.
1048 ** To resolve a conflict, first look to see if either action
1049 ** is on an error rule. In that case, take the action which
1050 ** is not associated with the error rule. If neither or both
1051 ** actions are associated with an error rule, then try to
1052 ** use precedence to resolve the conflict.
1054 ** If either action is a SHIFT, then it must be apx. This
1055 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1057 static int resolve_conflict(
1061 struct symbol
*spx
, *spy
;
1063 assert( apx
->sp
==apy
->sp
); /* Otherwise there would be no conflict */
1064 if( apx
->type
==SHIFT
&& apy
->type
==SHIFT
){
1065 apy
->type
= SSCONFLICT
;
1068 if( apx
->type
==SHIFT
&& apy
->type
==REDUCE
){
1070 spy
= apy
->x
.rp
->precsym
;
1071 if( spy
==0 || spx
->prec
<0 || spy
->prec
<0 ){
1072 /* Not enough precedence information. */
1073 apy
->type
= SRCONFLICT
;
1075 }else if( spx
->prec
>spy
->prec
){ /* higher precedence wins */
1076 apy
->type
= RD_RESOLVED
;
1077 }else if( spx
->prec
<spy
->prec
){
1078 apx
->type
= SH_RESOLVED
;
1079 }else if( spx
->prec
==spy
->prec
&& spx
->assoc
==RIGHT
){ /* Use operator */
1080 apy
->type
= RD_RESOLVED
; /* associativity */
1081 }else if( spx
->prec
==spy
->prec
&& spx
->assoc
==LEFT
){ /* to break tie */
1082 apx
->type
= SH_RESOLVED
;
1084 assert( spx
->prec
==spy
->prec
&& spx
->assoc
==NONE
);
1085 apy
->type
= SRCONFLICT
;
1088 }else if( apx
->type
==REDUCE
&& apy
->type
==REDUCE
){
1089 spx
= apx
->x
.rp
->precsym
;
1090 spy
= apy
->x
.rp
->precsym
;
1091 if( spx
==0 || spy
==0 || spx
->prec
<0 ||
1092 spy
->prec
<0 || spx
->prec
==spy
->prec
){
1093 apy
->type
= RRCONFLICT
;
1095 }else if( spx
->prec
>spy
->prec
){
1096 apy
->type
= RD_RESOLVED
;
1097 }else if( spx
->prec
<spy
->prec
){
1098 apx
->type
= RD_RESOLVED
;
1102 apx
->type
==SH_RESOLVED
||
1103 apx
->type
==RD_RESOLVED
||
1104 apx
->type
==SSCONFLICT
||
1105 apx
->type
==SRCONFLICT
||
1106 apx
->type
==RRCONFLICT
||
1107 apy
->type
==SH_RESOLVED
||
1108 apy
->type
==RD_RESOLVED
||
1109 apy
->type
==SSCONFLICT
||
1110 apy
->type
==SRCONFLICT
||
1111 apy
->type
==RRCONFLICT
1113 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1114 ** REDUCEs on the list. If we reach this point it must be because
1115 ** the parser conflict had already been resolved. */
1119 /********************* From the file "configlist.c" *************************/
1121 ** Routines to processing a configuration list and building a state
1122 ** in the LEMON parser generator.
1125 static struct config
*freelist
= 0; /* List of free configurations */
1126 static struct config
*current
= 0; /* Top of list of configurations */
1127 static struct config
**currentend
= 0; /* Last on list of configs */
1128 static struct config
*basis
= 0; /* Top of list of basis configs */
1129 static struct config
**basisend
= 0; /* End of list of basis configs */
1131 /* Return a pointer to a new configuration */
1132 PRIVATE
struct config
*newconfig(){
1133 struct config
*newcfg
;
1137 freelist
= (struct config
*)calloc( amt
, sizeof(struct config
) );
1139 fprintf(stderr
,"Unable to allocate memory for a new configuration.");
1142 for(i
=0; i
<amt
-1; i
++) freelist
[i
].next
= &freelist
[i
+1];
1143 freelist
[amt
-1].next
= 0;
1146 freelist
= freelist
->next
;
1150 /* The configuration "old" is no longer used */
1151 PRIVATE
void deleteconfig(struct config
*old
)
1153 old
->next
= freelist
;
1157 /* Initialized the configuration list builder */
1158 void Configlist_init(){
1160 currentend
= ¤t
;
1167 /* Initialized the configuration list builder */
1168 void Configlist_reset(){
1170 currentend
= ¤t
;
1173 Configtable_clear(0);
1177 /* Add another configuration to the configuration list */
1178 struct config
*Configlist_add(
1179 struct rule
*rp
, /* The rule */
1180 int dot
/* Index into the RHS of the rule where the dot goes */
1182 struct config
*cfp
, model
;
1184 assert( currentend
!=0 );
1187 cfp
= Configtable_find(&model
);
1192 cfp
->fws
= SetNew();
1194 cfp
->fplp
= cfp
->bplp
= 0;
1198 currentend
= &cfp
->next
;
1199 Configtable_insert(cfp
);
1204 /* Add a basis configuration to the configuration list */
1205 struct config
*Configlist_addbasis(struct rule
*rp
, int dot
)
1207 struct config
*cfp
, model
;
1209 assert( basisend
!=0 );
1210 assert( currentend
!=0 );
1213 cfp
= Configtable_find(&model
);
1218 cfp
->fws
= SetNew();
1220 cfp
->fplp
= cfp
->bplp
= 0;
1224 currentend
= &cfp
->next
;
1226 basisend
= &cfp
->bp
;
1227 Configtable_insert(cfp
);
1232 /* Compute the closure of the configuration list */
1233 void Configlist_closure(struct lemon
*lemp
)
1235 struct config
*cfp
, *newcfp
;
1236 struct rule
*rp
, *newrp
;
1237 struct symbol
*sp
, *xsp
;
1240 assert( currentend
!=0 );
1241 for(cfp
=current
; cfp
; cfp
=cfp
->next
){
1244 if( dot
>=rp
->nrhs
) continue;
1246 if( sp
->type
==NONTERMINAL
){
1247 if( sp
->rule
==0 && sp
!=lemp
->errsym
){
1248 ErrorMsg(lemp
->filename
,rp
->line
,"Nonterminal \"%s\" has no rules.",
1252 for(newrp
=sp
->rule
; newrp
; newrp
=newrp
->nextlhs
){
1253 newcfp
= Configlist_add(newrp
,0);
1254 for(i
=dot
+1; i
<rp
->nrhs
; i
++){
1256 if( xsp
->type
==TERMINAL
){
1257 SetAdd(newcfp
->fws
,xsp
->index
);
1259 }else if( xsp
->type
==MULTITERMINAL
){
1261 for(k
=0; k
<xsp
->nsubsym
; k
++){
1262 SetAdd(newcfp
->fws
, xsp
->subsym
[k
]->index
);
1266 SetUnion(newcfp
->fws
,xsp
->firstset
);
1267 if( xsp
->lambda
==LEMON_FALSE
) break;
1270 if( i
==rp
->nrhs
) Plink_add(&cfp
->fplp
,newcfp
);
1277 /* Sort the configuration list */
1278 void Configlist_sort(){
1279 current
= (struct config
*)msort((char *)current
,(char **)&(current
->next
),Configcmp
);
1284 /* Sort the basis configuration list */
1285 void Configlist_sortbasis(){
1286 basis
= (struct config
*)msort((char *)current
,(char **)&(current
->bp
),Configcmp
);
1291 /* Return a pointer to the head of the configuration list and
1292 ** reset the list */
1293 struct config
*Configlist_return(){
1301 /* Return a pointer to the head of the configuration list and
1302 ** reset the list */
1303 struct config
*Configlist_basis(){
1311 /* Free all elements of the given configuration list */
1312 void Configlist_eat(struct config
*cfp
)
1314 struct config
*nextcfp
;
1315 for(; cfp
; cfp
=nextcfp
){
1316 nextcfp
= cfp
->next
;
1317 assert( cfp
->fplp
==0 );
1318 assert( cfp
->bplp
==0 );
1319 if( cfp
->fws
) SetFree(cfp
->fws
);
1324 /***************** From the file "error.c" *********************************/
1326 ** Code for printing error message.
1329 void ErrorMsg(const char *filename
, int lineno
, const char *format
, ...){
1331 fprintf(stderr
, "%s:%d: ", filename
, lineno
);
1332 va_start(ap
, format
);
1333 vfprintf(stderr
,format
,ap
);
1335 fprintf(stderr
, "\n");
1337 /**************** From the file "main.c" ************************************/
1339 ** Main program file for the LEMON parser generator.
1342 /* Report an out-of-memory condition and abort. This function
1343 ** is used mostly by the "MemoryCheck" macro in struct.h
1345 void memory_error(){
1346 fprintf(stderr
,"Out of memory. Aborting...\n");
1350 static int nDefine
= 0; /* Number of -D options on the command line */
1351 static char **azDefine
= 0; /* Name of the -D macros */
1353 /* This routine is called with the argument to each -D command-line option.
1354 ** Add the macro defined to the azDefine array.
1356 static void handle_D_option(char *z
){
1359 azDefine
= (char **) realloc(azDefine
, sizeof(azDefine
[0])*nDefine
);
1361 fprintf(stderr
,"out of memory\n");
1364 paz
= &azDefine
[nDefine
-1];
1365 *paz
= (char *) malloc( lemonStrlen(z
)+1 );
1367 fprintf(stderr
,"out of memory\n");
1371 for(z
=*paz
; *z
&& *z
!='='; z
++){}
1375 static char *user_templatename
= NULL
;
1376 static void handle_T_option(char *z
){
1377 user_templatename
= (char *) malloc( lemonStrlen(z
)+1 );
1378 if( user_templatename
==0 ){
1381 strcpy(user_templatename
, z
);
1384 /* The main program. Parse the command line and do it... */
1385 int main(int argc
, char **argv
)
1387 static int version
= 0;
1388 static int rpflag
= 0;
1389 static int basisflag
= 0;
1390 static int compress
= 0;
1391 static int quiet
= 0;
1392 static int statistics
= 0;
1393 static int mhflag
= 0;
1394 static int nolinenosflag
= 0;
1395 static int noResort
= 0;
1396 static struct s_options options
[] = {
1397 {OPT_FLAG
, "b", (char*)&basisflag
, "Print only the basis in report."},
1398 {OPT_FLAG
, "c", (char*)&compress
, "Don't compress the action table."},
1399 {OPT_FSTR
, "D", (char*)handle_D_option
, "Define an %ifdef macro."},
1400 {OPT_FSTR
, "T", (char*)handle_T_option
, "Specify a template file."},
1401 {OPT_FLAG
, "g", (char*)&rpflag
, "Print grammar without actions."},
1402 {OPT_FLAG
, "m", (char*)&mhflag
, "Output a makeheaders compatible file."},
1403 {OPT_FLAG
, "l", (char*)&nolinenosflag
, "Do not print #line statements."},
1404 {OPT_FLAG
, "p", (char*)&showPrecedenceConflict
,
1405 "Show conflicts resolved by precedence rules"},
1406 {OPT_FLAG
, "q", (char*)&quiet
, "(Quiet) Don't print the report file."},
1407 {OPT_FLAG
, "r", (char*)&noResort
, "Do not sort or renumber states"},
1408 {OPT_FLAG
, "s", (char*)&statistics
,
1409 "Print parser stats to standard output."},
1410 {OPT_FLAG
, "x", (char*)&version
, "Print the version number."},
1417 OptInit(argv
,options
,stderr
);
1419 printf("Lemon version 1.0\n");
1422 if( OptNArgs()!=1 ){
1423 fprintf(stderr
,"Exactly one filename argument is required.\n");
1426 memset(&lem
, 0, sizeof(lem
));
1429 /* Initialize the machine */
1433 lem
.argv0
= argv
[0];
1434 lem
.filename
= OptArg(0);
1435 lem
.basisflag
= basisflag
;
1436 lem
.nolinenosflag
= nolinenosflag
;
1438 lem
.errsym
= Symbol_new("error");
1439 lem
.errsym
->useCnt
= 0;
1441 /* Parse the input file */
1443 if( lem
.errorcnt
) exit(lem
.errorcnt
);
1445 fprintf(stderr
,"Empty grammar.\n");
1449 /* Count and index the symbols of the grammar */
1450 lem
.nsymbol
= Symbol_count();
1451 Symbol_new("{default}");
1452 lem
.symbols
= Symbol_arrayof();
1453 for(i
=0; i
<=lem
.nsymbol
; i
++) lem
.symbols
[i
]->index
= i
;
1454 qsort(lem
.symbols
,lem
.nsymbol
+1,sizeof(struct symbol
*), Symbolcmpp
);
1455 for(i
=0; i
<=lem
.nsymbol
; i
++) lem
.symbols
[i
]->index
= i
;
1456 for(i
=1; isupper(lem
.symbols
[i
]->name
[0]); i
++);
1459 /* Generate a reprint of the grammar, if requested on the command line */
1463 /* Initialize the size for all follow and first sets */
1464 SetSize(lem
.nterminal
+1);
1466 /* Find the precedence for every production rule (that has one) */
1467 FindRulePrecedences(&lem
);
1469 /* Compute the lambda-nonterminals and the first-sets for every
1471 FindFirstSets(&lem
);
1473 /* Compute all LR(0) states. Also record follow-set propagation
1474 ** links so that the follow-set can be computed later */
1477 lem
.sorted
= State_arrayof();
1479 /* Tie up loose ends on the propagation links */
1482 /* Compute the follow set of every reducible configuration */
1483 FindFollowSets(&lem
);
1485 /* Compute the action tables */
1488 /* Compress the action tables */
1489 if( compress
==0 ) CompressTables(&lem
);
1491 /* Reorder and renumber the states so that states with fewer choices
1492 ** occur at the end. This is an optimization that helps make the
1493 ** generated parser tables smaller. */
1494 if( noResort
==0 ) ResortStates(&lem
);
1496 /* Generate a report of the parser generated. (the "y.output" file) */
1497 if( !quiet
) ReportOutput(&lem
);
1499 /* Generate the source code for the parser */
1500 ReportTable(&lem
, mhflag
);
1502 /* Produce a header file for use by the scanner. (This step is
1503 ** omitted if the "-m" option is used because makeheaders will
1504 ** generate the file for us.) */
1505 if( !mhflag
) ReportHeader(&lem
);
1508 printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
1509 lem
.nterminal
, lem
.nsymbol
- lem
.nterminal
, lem
.nrule
);
1510 printf(" %d states, %d parser table entries, %d conflicts\n",
1511 lem
.nstate
, lem
.tablesize
, lem
.nconflict
);
1513 if( lem
.nconflict
> 0 ){
1514 fprintf(stderr
,"%d parsing conflicts.\n",lem
.nconflict
);
1517 /* return 0 on success, 1 on failure. */
1518 exitcode
= ((lem
.errorcnt
> 0) || (lem
.nconflict
> 0)) ? 1 : 0;
1522 /******************** From the file "msort.c" *******************************/
1524 ** A generic merge-sort program.
1527 ** Let "ptr" be a pointer to some structure which is at the head of
1528 ** a null-terminated list. Then to sort the list call:
1530 ** ptr = msort(ptr,&(ptr->next),cmpfnc);
1532 ** In the above, "cmpfnc" is a pointer to a function which compares
1533 ** two instances of the structure and returns an integer, as in
1534 ** strcmp. The second argument is a pointer to the pointer to the
1535 ** second element of the linked list. This address is used to compute
1536 ** the offset to the "next" field within the structure. The offset to
1537 ** the "next" field must be constant for all structures in the list.
1539 ** The function returns a new pointer which is the head of the list
1547 ** Return a pointer to the next structure in the linked list.
1549 #define NEXT(A) (*(char**)(((char*)A)+offset))
1553 ** a: A sorted, null-terminated linked list. (May be null).
1554 ** b: A sorted, null-terminated linked list. (May be null).
1555 ** cmp: A pointer to the comparison function.
1556 ** offset: Offset in the structure to the "next" field.
1559 ** A pointer to the head of a sorted list containing the elements
1563 ** The "next" pointers for elements in the lists a and b are
1569 int (*cmp
)(const char*,const char*),
1579 if( (*cmp
)(a
,b
)<=0 ){
1588 if( (*cmp
)(a
,b
)<=0 ){
1598 if( a
) NEXT(ptr
) = a
;
1606 ** list: Pointer to a singly-linked list of structures.
1607 ** next: Pointer to pointer to the second element of the list.
1608 ** cmp: A comparison function.
1611 ** A pointer to the head of a sorted list containing the elements
1612 ** orginally in list.
1615 ** The "next" pointers for elements in list are changed.
1621 int (*cmp
)(const char*,const char*)
1623 unsigned long offset
;
1625 char *set
[LISTSIZE
];
1627 offset
= (unsigned long)next
- (unsigned long)list
;
1628 for(i
=0; i
<LISTSIZE
; i
++) set
[i
] = 0;
1633 for(i
=0; i
<LISTSIZE
-1 && set
[i
]!=0; i
++){
1634 ep
= merge(ep
,set
[i
],cmp
,offset
);
1640 for(i
=0; i
<LISTSIZE
; i
++) if( set
[i
] ) ep
= merge(set
[i
],ep
,cmp
,offset
);
1643 /************************ From the file "option.c" **************************/
1645 static struct s_options
*op
;
1646 static FILE *errstream
;
1648 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1651 ** Print the command line with a carrot pointing to the k-th character
1652 ** of the n-th field.
1654 static void errline(int n
, int k
, FILE *err
)
1657 if( argv
[0] ) fprintf(err
,"%s",argv
[0]);
1658 spcnt
= lemonStrlen(argv
[0]) + 1;
1659 for(i
=1; i
<n
&& argv
[i
]; i
++){
1660 fprintf(err
," %s",argv
[i
]);
1661 spcnt
+= lemonStrlen(argv
[i
])+1;
1664 for(; argv
[i
]; i
++) fprintf(err
," %s",argv
[i
]);
1666 fprintf(err
,"\n%*s^-- here\n",spcnt
,"");
1668 fprintf(err
,"\n%*shere --^\n",spcnt
-7,"");
1673 ** Return the index of the N-th non-switch argument. Return -1
1674 ** if N is out of range.
1676 static int argindex(int n
)
1680 if( argv
!=0 && *argv
!=0 ){
1681 for(i
=1; argv
[i
]; i
++){
1682 if( dashdash
|| !ISOPT(argv
[i
]) ){
1683 if( n
==0 ) return i
;
1686 if( strcmp(argv
[i
],"--")==0 ) dashdash
= 1;
1692 static char emsg
[] = "Command line syntax error: ";
1695 ** Process a flag command line argument.
1697 static int handleflags(int i
, FILE *err
)
1702 for(j
=0; op
[j
].label
; j
++){
1703 if( strncmp(&argv
[i
][1],op
[j
].label
,lemonStrlen(op
[j
].label
))==0 ) break;
1705 v
= argv
[i
][0]=='-' ? 1 : 0;
1706 if( op
[j
].label
==0 ){
1708 fprintf(err
,"%sundefined option.\n",emsg
);
1712 }else if( op
[j
].type
==OPT_FLAG
){
1713 *((int*)op
[j
].arg
) = v
;
1714 }else if( op
[j
].type
==OPT_FFLAG
){
1715 (*(void(*)(int))(op
[j
].arg
))(v
);
1716 }else if( op
[j
].type
==OPT_FSTR
){
1717 (*(void(*)(char *))(op
[j
].arg
))(&argv
[i
][2]);
1720 fprintf(err
,"%smissing argument on switch.\n",emsg
);
1729 ** Process a command line switch which has an argument.
1731 static int handleswitch(int i
, FILE *err
)
1739 cp
= strchr(argv
[i
],'=');
1742 for(j
=0; op
[j
].label
; j
++){
1743 if( strcmp(argv
[i
],op
[j
].label
)==0 ) break;
1746 if( op
[j
].label
==0 ){
1748 fprintf(err
,"%sundefined option.\n",emsg
);
1754 switch( op
[j
].type
){
1758 fprintf(err
,"%soption requires an argument.\n",emsg
);
1765 dv
= strtod(cp
,&end
);
1768 fprintf(err
,"%sillegal character in floating-point argument.\n",emsg
);
1769 errline(i
,((unsigned long)end
)-(unsigned long)argv
[i
],err
);
1776 lv
= strtol(cp
,&end
,0);
1779 fprintf(err
,"%sillegal character in integer argument.\n",emsg
);
1780 errline(i
,((unsigned long)end
)-(unsigned long)argv
[i
],err
);
1790 switch( op
[j
].type
){
1795 *(double*)(op
[j
].arg
) = dv
;
1798 (*(void(*)(double))(op
[j
].arg
))(dv
);
1801 *(int*)(op
[j
].arg
) = lv
;
1804 (*(void(*)(int))(op
[j
].arg
))((int)lv
);
1807 *(char**)(op
[j
].arg
) = sv
;
1810 (*(void(*)(char *))(op
[j
].arg
))(sv
);
1817 int OptInit(char **a
, struct s_options
*o
, FILE *err
)
1823 if( argv
&& *argv
&& op
){
1825 for(i
=1; argv
[i
]; i
++){
1826 if( argv
[i
][0]=='+' || argv
[i
][0]=='-' ){
1827 errcnt
+= handleflags(i
,err
);
1828 }else if( strchr(argv
[i
],'=') ){
1829 errcnt
+= handleswitch(i
,err
);
1834 fprintf(err
,"Valid command line options for \"%s\" are:\n",*a
);
1845 if( argv
!=0 && argv
[0]!=0 ){
1846 for(i
=1; argv
[i
]; i
++){
1847 if( dashdash
|| !ISOPT(argv
[i
]) ) cnt
++;
1848 if( strcmp(argv
[i
],"--")==0 ) dashdash
= 1;
1858 return i
>=0 ? argv
[i
] : 0;
1865 if( i
>=0 ) errline(i
,0,errstream
);
1872 for(i
=0; op
[i
].label
; i
++){
1873 len
= lemonStrlen(op
[i
].label
) + 1;
1874 switch( op
[i
].type
){
1880 len
+= 9; /* length of "<integer>" */
1884 len
+= 6; /* length of "<real>" */
1888 len
+= 8; /* length of "<string>" */
1891 if( len
>max
) max
= len
;
1893 for(i
=0; op
[i
].label
; i
++){
1894 switch( op
[i
].type
){
1897 fprintf(errstream
," -%-*s %s\n",max
,op
[i
].label
,op
[i
].message
);
1901 fprintf(errstream
," %s=<integer>%*s %s\n",op
[i
].label
,
1902 (int)(max
-lemonStrlen(op
[i
].label
)-9),"",op
[i
].message
);
1906 fprintf(errstream
," %s=<real>%*s %s\n",op
[i
].label
,
1907 (int)(max
-lemonStrlen(op
[i
].label
)-6),"",op
[i
].message
);
1911 fprintf(errstream
," %s=<string>%*s %s\n",op
[i
].label
,
1912 (int)(max
-lemonStrlen(op
[i
].label
)-8),"",op
[i
].message
);
1917 /*********************** From the file "parse.c" ****************************/
1919 ** Input file parser for the LEMON parser generator.
1922 /* The state of the parser */
1925 WAITING_FOR_DECL_OR_RULE
,
1926 WAITING_FOR_DECL_KEYWORD
,
1927 WAITING_FOR_DECL_ARG
,
1928 WAITING_FOR_PRECEDENCE_SYMBOL
,
1938 RESYNC_AFTER_RULE_ERROR
,
1939 RESYNC_AFTER_DECL_ERROR
,
1940 WAITING_FOR_DESTRUCTOR_SYMBOL
,
1941 WAITING_FOR_DATATYPE_SYMBOL
,
1942 WAITING_FOR_FALLBACK_ID
,
1943 WAITING_FOR_WILDCARD_ID
1946 char *filename
; /* Name of the input file */
1947 int tokenlineno
; /* Linenumber at which current token starts */
1948 int errorcnt
; /* Number of errors so far */
1949 char *tokenstart
; /* Text of current token */
1950 struct lemon
*gp
; /* Global state vector */
1951 enum e_state state
; /* The state of the parser */
1952 struct symbol
*fallback
; /* The fallback token */
1953 struct symbol
*lhs
; /* Left-hand side of current rule */
1954 const char *lhsalias
; /* Alias for the LHS */
1955 int nrhs
; /* Number of right-hand side symbols seen */
1956 struct symbol
*rhs
[MAXRHS
]; /* RHS symbols */
1957 const char *alias
[MAXRHS
]; /* Aliases for each RHS symbol (or NULL) */
1958 struct rule
*prevrule
; /* Previous rule parsed */
1959 const char *declkeyword
; /* Keyword of a declaration */
1960 char **declargslot
; /* Where the declaration argument should be put */
1961 int insertLineMacro
; /* Add #line before declaration insert */
1962 int *decllinenoslot
; /* Where to write declaration line number */
1963 enum e_assoc declassoc
; /* Assign this association to decl arguments */
1964 int preccounter
; /* Assign this precedence to decl arguments */
1965 struct rule
*firstrule
; /* Pointer to first rule in the grammar */
1966 struct rule
*lastrule
; /* Pointer to the most recently parsed rule */
1969 /* Parse a single token */
1970 static void parseonetoken(struct pstate
*psp
)
1973 x
= Strsafe(psp
->tokenstart
); /* Save the token permanently */
1975 printf("%s:%d: Token=[%s] state=%d\n",psp
->filename
,psp
->tokenlineno
,
1978 switch( psp
->state
){
1981 psp
->preccounter
= 0;
1982 psp
->firstrule
= psp
->lastrule
= 0;
1984 /* Fall thru to next case */
1985 case WAITING_FOR_DECL_OR_RULE
:
1987 psp
->state
= WAITING_FOR_DECL_KEYWORD
;
1988 }else if( islower(x
[0]) ){
1989 psp
->lhs
= Symbol_new(x
);
1992 psp
->state
= WAITING_FOR_ARROW
;
1993 }else if( x
[0]=='{' ){
1994 if( psp
->prevrule
==0 ){
1995 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
1996 "There is no prior rule upon which to attach the code \
1997 fragment which begins on this line.");
1999 }else if( psp
->prevrule
->code
!=0 ){
2000 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2001 "Code fragment beginning on this line is not the first \
2002 to follow the previous rule.");
2005 psp
->prevrule
->line
= psp
->tokenlineno
;
2006 psp
->prevrule
->code
= &x
[1];
2008 }else if( x
[0]=='[' ){
2009 psp
->state
= PRECEDENCE_MARK_1
;
2011 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2012 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2017 case PRECEDENCE_MARK_1
:
2018 if( !isupper(x
[0]) ){
2019 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2020 "The precedence symbol must be a terminal.");
2022 }else if( psp
->prevrule
==0 ){
2023 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2024 "There is no prior rule to assign precedence \"[%s]\".",x
);
2026 }else if( psp
->prevrule
->precsym
!=0 ){
2027 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2028 "Precedence mark on this line is not the first \
2029 to follow the previous rule.");
2032 psp
->prevrule
->precsym
= Symbol_new(x
);
2034 psp
->state
= PRECEDENCE_MARK_2
;
2036 case PRECEDENCE_MARK_2
:
2038 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2039 "Missing \"]\" on precedence mark.");
2042 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2044 case WAITING_FOR_ARROW
:
2045 if( x
[0]==':' && x
[1]==':' && x
[2]=='=' ){
2046 psp
->state
= IN_RHS
;
2047 }else if( x
[0]=='(' ){
2048 psp
->state
= LHS_ALIAS_1
;
2050 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2051 "Expected to see a \":\" following the LHS symbol \"%s\".",
2054 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2058 if( isalpha(x
[0]) ){
2060 psp
->state
= LHS_ALIAS_2
;
2062 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2063 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2066 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2071 psp
->state
= LHS_ALIAS_3
;
2073 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2074 "Missing \")\" following LHS alias name \"%s\".",psp
->lhsalias
);
2076 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2080 if( x
[0]==':' && x
[1]==':' && x
[2]=='=' ){
2081 psp
->state
= IN_RHS
;
2083 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2084 "Missing \"->\" following: \"%s(%s)\".",
2085 psp
->lhs
->name
,psp
->lhsalias
);
2087 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2093 rp
= (struct rule
*)calloc( sizeof(struct rule
) +
2094 sizeof(struct symbol
*)*psp
->nrhs
+ sizeof(char*)*psp
->nrhs
, 1);
2096 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2097 "Can't allocate enough memory for this rule.");
2102 rp
->ruleline
= psp
->tokenlineno
;
2103 rp
->rhs
= (struct symbol
**)&rp
[1];
2104 rp
->rhsalias
= (const char**)&(rp
->rhs
[psp
->nrhs
]);
2105 for(i
=0; i
<psp
->nrhs
; i
++){
2106 rp
->rhs
[i
] = psp
->rhs
[i
];
2107 rp
->rhsalias
[i
] = psp
->alias
[i
];
2110 rp
->lhsalias
= psp
->lhsalias
;
2111 rp
->nrhs
= psp
->nrhs
;
2114 rp
->index
= psp
->gp
->nrule
++;
2115 rp
->nextlhs
= rp
->lhs
->rule
;
2118 if( psp
->firstrule
==0 ){
2119 psp
->firstrule
= psp
->lastrule
= rp
;
2121 psp
->lastrule
->next
= rp
;
2126 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2127 }else if( isalpha(x
[0]) ){
2128 if( psp
->nrhs
>=MAXRHS
){
2129 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2130 "Too many symbols on RHS of rule beginning at \"%s\".",
2133 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2135 psp
->rhs
[psp
->nrhs
] = Symbol_new(x
);
2136 psp
->alias
[psp
->nrhs
] = 0;
2139 }else if( (x
[0]=='|' || x
[0]=='/') && psp
->nrhs
>0 ){
2140 struct symbol
*msp
= psp
->rhs
[psp
->nrhs
-1];
2141 if( msp
->type
!=MULTITERMINAL
){
2142 struct symbol
*origsp
= msp
;
2143 msp
= (struct symbol
*) calloc(1,sizeof(*msp
));
2144 memset(msp
, 0, sizeof(*msp
));
2145 msp
->type
= MULTITERMINAL
;
2147 msp
->subsym
= (struct symbol
**) calloc(1,sizeof(struct symbol
*));
2148 msp
->subsym
[0] = origsp
;
2149 msp
->name
= origsp
->name
;
2150 psp
->rhs
[psp
->nrhs
-1] = msp
;
2153 msp
->subsym
= (struct symbol
**) realloc(msp
->subsym
,
2154 sizeof(struct symbol
*)*msp
->nsubsym
);
2155 msp
->subsym
[msp
->nsubsym
-1] = Symbol_new(&x
[1]);
2156 if( islower(x
[1]) || islower(msp
->subsym
[0]->name
[0]) ){
2157 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2158 "Cannot form a compound containing a non-terminal");
2161 }else if( x
[0]=='(' && psp
->nrhs
>0 ){
2162 psp
->state
= RHS_ALIAS_1
;
2164 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2165 "Illegal character on RHS of rule: \"%s\".",x
);
2167 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2171 if( isalpha(x
[0]) ){
2172 psp
->alias
[psp
->nrhs
-1] = x
;
2173 psp
->state
= RHS_ALIAS_2
;
2175 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2176 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2177 x
,psp
->rhs
[psp
->nrhs
-1]->name
);
2179 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2184 psp
->state
= IN_RHS
;
2186 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2187 "Missing \")\" following LHS alias name \"%s\".",psp
->lhsalias
);
2189 psp
->state
= RESYNC_AFTER_RULE_ERROR
;
2192 case WAITING_FOR_DECL_KEYWORD
:
2193 if( isalpha(x
[0]) ){
2194 psp
->declkeyword
= x
;
2195 psp
->declargslot
= 0;
2196 psp
->decllinenoslot
= 0;
2197 psp
->insertLineMacro
= 1;
2198 psp
->state
= WAITING_FOR_DECL_ARG
;
2199 if( strcmp(x
,"name")==0 ){
2200 psp
->declargslot
= &(psp
->gp
->name
);
2201 psp
->insertLineMacro
= 0;
2202 }else if( strcmp(x
,"include")==0 ){
2203 psp
->declargslot
= &(psp
->gp
->include
);
2204 }else if( strcmp(x
,"code")==0 ){
2205 psp
->declargslot
= &(psp
->gp
->extracode
);
2206 }else if( strcmp(x
,"token_destructor")==0 ){
2207 psp
->declargslot
= &psp
->gp
->tokendest
;
2208 }else if( strcmp(x
,"default_destructor")==0 ){
2209 psp
->declargslot
= &psp
->gp
->vardest
;
2210 }else if( strcmp(x
,"token_prefix")==0 ){
2211 psp
->declargslot
= &psp
->gp
->tokenprefix
;
2212 psp
->insertLineMacro
= 0;
2213 }else if( strcmp(x
,"syntax_error")==0 ){
2214 psp
->declargslot
= &(psp
->gp
->error
);
2215 }else if( strcmp(x
,"parse_accept")==0 ){
2216 psp
->declargslot
= &(psp
->gp
->accept
);
2217 }else if( strcmp(x
,"parse_failure")==0 ){
2218 psp
->declargslot
= &(psp
->gp
->failure
);
2219 }else if( strcmp(x
,"stack_overflow")==0 ){
2220 psp
->declargslot
= &(psp
->gp
->overflow
);
2221 }else if( strcmp(x
,"extra_argument")==0 ){
2222 psp
->declargslot
= &(psp
->gp
->arg
);
2223 psp
->insertLineMacro
= 0;
2224 }else if( strcmp(x
,"token_type")==0 ){
2225 psp
->declargslot
= &(psp
->gp
->tokentype
);
2226 psp
->insertLineMacro
= 0;
2227 }else if( strcmp(x
,"default_type")==0 ){
2228 psp
->declargslot
= &(psp
->gp
->vartype
);
2229 psp
->insertLineMacro
= 0;
2230 }else if( strcmp(x
,"stack_size")==0 ){
2231 psp
->declargslot
= &(psp
->gp
->stacksize
);
2232 psp
->insertLineMacro
= 0;
2233 }else if( strcmp(x
,"start_symbol")==0 ){
2234 psp
->declargslot
= &(psp
->gp
->start
);
2235 psp
->insertLineMacro
= 0;
2236 }else if( strcmp(x
,"left")==0 ){
2238 psp
->declassoc
= LEFT
;
2239 psp
->state
= WAITING_FOR_PRECEDENCE_SYMBOL
;
2240 }else if( strcmp(x
,"right")==0 ){
2242 psp
->declassoc
= RIGHT
;
2243 psp
->state
= WAITING_FOR_PRECEDENCE_SYMBOL
;
2244 }else if( strcmp(x
,"nonassoc")==0 ){
2246 psp
->declassoc
= NONE
;
2247 psp
->state
= WAITING_FOR_PRECEDENCE_SYMBOL
;
2248 }else if( strcmp(x
,"destructor")==0 ){
2249 psp
->state
= WAITING_FOR_DESTRUCTOR_SYMBOL
;
2250 }else if( strcmp(x
,"type")==0 ){
2251 psp
->state
= WAITING_FOR_DATATYPE_SYMBOL
;
2252 }else if( strcmp(x
,"fallback")==0 ){
2254 psp
->state
= WAITING_FOR_FALLBACK_ID
;
2255 }else if( strcmp(x
,"wildcard")==0 ){
2256 psp
->state
= WAITING_FOR_WILDCARD_ID
;
2258 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2259 "Unknown declaration keyword: \"%%%s\".",x
);
2261 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2264 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2265 "Illegal declaration keyword: \"%s\".",x
);
2267 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2270 case WAITING_FOR_DESTRUCTOR_SYMBOL
:
2271 if( !isalpha(x
[0]) ){
2272 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2273 "Symbol name missing after %%destructor keyword");
2275 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2277 struct symbol
*sp
= Symbol_new(x
);
2278 psp
->declargslot
= &sp
->destructor
;
2279 psp
->decllinenoslot
= &sp
->destLineno
;
2280 psp
->insertLineMacro
= 1;
2281 psp
->state
= WAITING_FOR_DECL_ARG
;
2284 case WAITING_FOR_DATATYPE_SYMBOL
:
2285 if( !isalpha(x
[0]) ){
2286 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2287 "Symbol name missing after %%type keyword");
2289 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2291 struct symbol
*sp
= Symbol_find(x
);
2292 if((sp
) && (sp
->datatype
)){
2293 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2294 "Symbol %%type \"%s\" already defined", x
);
2296 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2301 psp
->declargslot
= &sp
->datatype
;
2302 psp
->insertLineMacro
= 0;
2303 psp
->state
= WAITING_FOR_DECL_ARG
;
2307 case WAITING_FOR_PRECEDENCE_SYMBOL
:
2309 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2310 }else if( isupper(x
[0]) ){
2314 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2315 "Symbol \"%s\" has already be given a precedence.",x
);
2318 sp
->prec
= psp
->preccounter
;
2319 sp
->assoc
= psp
->declassoc
;
2322 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2323 "Can't assign a precedence to \"%s\".",x
);
2327 case WAITING_FOR_DECL_ARG
:
2328 if( x
[0]=='{' || x
[0]=='\"' || isalnum(x
[0]) ){
2329 const char *zOld
, *zNew
;
2331 int nOld
, n
, nLine
, nNew
, nBack
;
2335 if( zNew
[0]=='"' || zNew
[0]=='{' ) zNew
++;
2336 nNew
= lemonStrlen(zNew
);
2337 if( *psp
->declargslot
){
2338 zOld
= *psp
->declargslot
;
2342 nOld
= lemonStrlen(zOld
);
2343 n
= nOld
+ nNew
+ 20;
2344 addLineMacro
= !psp
->gp
->nolinenosflag
&& psp
->insertLineMacro
&&
2345 (psp
->decllinenoslot
==0 || psp
->decllinenoslot
[0]!=0);
2347 for(z
=psp
->filename
, nBack
=0; *z
; z
++){
2348 if( *z
=='\\' ) nBack
++;
2350 sprintf(zLine
, "#line %d ", psp
->tokenlineno
);
2351 nLine
= lemonStrlen(zLine
);
2352 n
+= nLine
+ lemonStrlen(psp
->filename
) + nBack
;
2354 *psp
->declargslot
= (char *) realloc(*psp
->declargslot
, n
);
2355 zBuf
= *psp
->declargslot
+ nOld
;
2357 if( nOld
&& zBuf
[-1]!='\n' ){
2360 memcpy(zBuf
, zLine
, nLine
);
2363 for(z
=psp
->filename
; *z
; z
++){
2372 if( psp
->decllinenoslot
&& psp
->decllinenoslot
[0]==0 ){
2373 psp
->decllinenoslot
[0] = psp
->tokenlineno
;
2375 memcpy(zBuf
, zNew
, nNew
);
2378 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2380 ErrorMsg(psp
->filename
,psp
->tokenlineno
,
2381 "Illegal argument to %%%s: %s",psp
->declkeyword
,x
);
2383 psp
->state
= RESYNC_AFTER_DECL_ERROR
;
2386 case WAITING_FOR_FALLBACK_ID
:
2388 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2389 }else if( !isupper(x
[0]) ){
2390 ErrorMsg(psp
->filename
, psp
->tokenlineno
,
2391 "%%fallback argument \"%s\" should be a token", x
);
2394 struct symbol
*sp
= Symbol_new(x
);
2395 if( psp
->fallback
==0 ){
2397 }else if( sp
->fallback
){
2398 ErrorMsg(psp
->filename
, psp
->tokenlineno
,
2399 "More than one fallback assigned to token %s", x
);
2402 sp
->fallback
= psp
->fallback
;
2403 psp
->gp
->has_fallback
= 1;
2407 case WAITING_FOR_WILDCARD_ID
:
2409 psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2410 }else if( !isupper(x
[0]) ){
2411 ErrorMsg(psp
->filename
, psp
->tokenlineno
,
2412 "%%wildcard argument \"%s\" should be a token", x
);
2415 struct symbol
*sp
= Symbol_new(x
);
2416 if( psp
->gp
->wildcard
==0 ){
2417 psp
->gp
->wildcard
= sp
;
2419 ErrorMsg(psp
->filename
, psp
->tokenlineno
,
2420 "Extra wildcard to token: %s", x
);
2425 case RESYNC_AFTER_RULE_ERROR
:
2426 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2428 case RESYNC_AFTER_DECL_ERROR
:
2429 if( x
[0]=='.' ) psp
->state
= WAITING_FOR_DECL_OR_RULE
;
2430 if( x
[0]=='%' ) psp
->state
= WAITING_FOR_DECL_KEYWORD
;
2435 /* Run the preprocessor over the input file text. The global variables
2436 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2437 ** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2438 ** comments them out. Text in between is also commented out as appropriate.
2440 static void preprocess_input(char *z
){
2445 int start_lineno
= 1;
2446 for(i
=0; z
[i
]; i
++){
2447 if( z
[i
]=='\n' ) lineno
++;
2448 if( z
[i
]!='%' || (i
>0 && z
[i
-1]!='\n') ) continue;
2449 if( strncmp(&z
[i
],"%endif",6)==0 && isspace(z
[i
+6]) ){
2453 for(j
=start
; j
<i
; j
++) if( z
[j
]!='\n' ) z
[j
] = ' ';
2456 for(j
=i
; z
[j
] && z
[j
]!='\n'; j
++) z
[j
] = ' ';
2457 }else if( (strncmp(&z
[i
],"%ifdef",6)==0 && isspace(z
[i
+6]))
2458 || (strncmp(&z
[i
],"%ifndef",7)==0 && isspace(z
[i
+7])) ){
2462 for(j
=i
+7; isspace(z
[j
]); j
++){}
2463 for(n
=0; z
[j
+n
] && !isspace(z
[j
+n
]); n
++){}
2465 for(k
=0; k
<nDefine
; k
++){
2466 if( strncmp(azDefine
[k
],&z
[j
],n
)==0 && lemonStrlen(azDefine
[k
])==n
){
2471 if( z
[i
+3]=='n' ) exclude
= !exclude
;
2474 start_lineno
= lineno
;
2477 for(j
=i
; z
[j
] && z
[j
]!='\n'; j
++) z
[j
] = ' ';
2481 fprintf(stderr
,"unterminated %%ifdef starting on line %d\n", start_lineno
);
2486 /* In spite of its name, this function is really a scanner. It read
2487 ** in the entire input file (all at once) then tokenizes it. Each
2488 ** token is passed to the function "parseonetoken" which builds all
2489 ** the appropriate data structures in the global state vector "gp".
2491 void Parse(struct lemon
*gp
)
2502 memset(&ps
, '\0', sizeof(ps
));
2504 ps
.filename
= gp
->filename
;
2506 ps
.state
= INITIALIZE
;
2508 /* Begin by reading the input file */
2509 fp
= fopen(ps
.filename
,"rb");
2511 ErrorMsg(ps
.filename
,0,"Can't open this file for reading.");
2516 filesize
= ftell(fp
);
2518 filebuf
= (char *)malloc( filesize
+1 );
2520 ErrorMsg(ps
.filename
,0,"Can't allocate %d of memory to hold this file.",
2526 if( fread(filebuf
,1,filesize
,fp
)!=filesize
){
2527 ErrorMsg(ps
.filename
,0,"Can't read in all %d bytes of this file.",
2535 filebuf
[filesize
] = 0;
2537 /* Make an initial pass through the file to handle %ifdef and %ifndef */
2538 preprocess_input(filebuf
);
2540 /* Now scan the text of the input file */
2542 for(cp
=filebuf
; (c
= *cp
)!=0; ){
2543 if( c
=='\n' ) lineno
++; /* Keep track of the line number */
2544 if( isspace(c
) ){ cp
++; continue; } /* Skip all white space */
2545 if( c
=='/' && cp
[1]=='/' ){ /* Skip C++ style comments */
2547 while( (c
= *cp
)!=0 && c
!='\n' ) cp
++;
2550 if( c
=='/' && cp
[1]=='*' ){ /* Skip C style comments */
2552 while( (c
= *cp
)!=0 && (c
!='/' || cp
[-1]!='*') ){
2553 if( c
=='\n' ) lineno
++;
2559 ps
.tokenstart
= cp
; /* Mark the beginning of the token */
2560 ps
.tokenlineno
= lineno
; /* Linenumber on which token begins */
2561 if( c
=='\"' ){ /* String literals */
2563 while( (c
= *cp
)!=0 && c
!='\"' ){
2564 if( c
=='\n' ) lineno
++;
2568 ErrorMsg(ps
.filename
,startline
,
2569 "String starting on this line is not terminated before the end of the file.");
2575 }else if( c
=='{' ){ /* A block of C code */
2578 for(level
=1; (c
= *cp
)!=0 && (level
>1 || c
!='}'); cp
++){
2579 if( c
=='\n' ) lineno
++;
2580 else if( c
=='{' ) level
++;
2581 else if( c
=='}' ) level
--;
2582 else if( c
=='/' && cp
[1]=='*' ){ /* Skip comments */
2586 while( (c
= *cp
)!=0 && (c
!='/' || prevc
!='*') ){
2587 if( c
=='\n' ) lineno
++;
2591 }else if( c
=='/' && cp
[1]=='/' ){ /* Skip C++ style comments too */
2593 while( (c
= *cp
)!=0 && c
!='\n' ) cp
++;
2595 }else if( c
=='\'' || c
=='\"' ){ /* String a character literals */
2596 int startchar
, prevc
;
2599 for(cp
++; (c
= *cp
)!=0 && (c
!=startchar
|| prevc
=='\\'); cp
++){
2600 if( c
=='\n' ) lineno
++;
2601 if( prevc
=='\\' ) prevc
= 0;
2607 ErrorMsg(ps
.filename
,ps
.tokenlineno
,
2608 "C code starting on this line is not terminated before the end of the file.");
2614 }else if( isalnum(c
) ){ /* Identifiers */
2615 while( (c
= *cp
)!=0 && (isalnum(c
) || c
=='_') ) cp
++;
2617 }else if( c
==':' && cp
[1]==':' && cp
[2]=='=' ){ /* The operator "::=" */
2620 }else if( (c
=='/' || c
=='|') && isalpha(cp
[1]) ){
2622 while( (c
= *cp
)!=0 && (isalnum(c
) || c
=='_') ) cp
++;
2624 }else{ /* All other (one character) operators */
2629 *cp
= 0; /* Null terminate the token */
2630 parseonetoken(&ps
); /* Parse the token */
2631 *cp
= c
; /* Restore the buffer */
2634 free(filebuf
); /* Release the buffer after parsing */
2635 gp
->rule
= ps
.firstrule
;
2636 gp
->errorcnt
= ps
.errorcnt
;
2638 /*************************** From the file "plink.c" *********************/
2640 ** Routines processing configuration follow-set propagation links
2641 ** in the LEMON parser generator.
2643 static struct plink
*plink_freelist
= 0;
2645 /* Allocate a new plink */
2646 struct plink
*Plink_new(){
2647 struct plink
*newlink
;
2649 if( plink_freelist
==0 ){
2652 plink_freelist
= (struct plink
*)calloc( amt
, sizeof(struct plink
) );
2653 if( plink_freelist
==0 ){
2655 "Unable to allocate memory for a new follow-set propagation link.\n");
2658 for(i
=0; i
<amt
-1; i
++) plink_freelist
[i
].next
= &plink_freelist
[i
+1];
2659 plink_freelist
[amt
-1].next
= 0;
2661 newlink
= plink_freelist
;
2662 plink_freelist
= plink_freelist
->next
;
2666 /* Add a plink to a plink list */
2667 void Plink_add(struct plink
**plpp
, struct config
*cfp
)
2669 struct plink
*newlink
;
2670 newlink
= Plink_new();
2671 newlink
->next
= *plpp
;
2676 /* Transfer every plink on the list "from" to the list "to" */
2677 void Plink_copy(struct plink
**to
, struct plink
*from
)
2679 struct plink
*nextpl
;
2681 nextpl
= from
->next
;
2688 /* Delete every plink on the list */
2689 void Plink_delete(struct plink
*plp
)
2691 struct plink
*nextpl
;
2695 plp
->next
= plink_freelist
;
2696 plink_freelist
= plp
;
2700 /*********************** From the file "report.c" **************************/
2702 ** Procedures for generating reports and tables in the LEMON parser generator.
2705 /* Generate a filename with the given suffix. Space to hold the
2706 ** name comes from malloc() and must be freed by the calling
2709 PRIVATE
char *file_makename(struct lemon
*lemp
, const char *suffix
)
2714 name
= (char*)malloc( lemonStrlen(lemp
->filename
) + lemonStrlen(suffix
) + 5 );
2716 fprintf(stderr
,"Can't allocate space for a filename.\n");
2719 strcpy(name
,lemp
->filename
);
2720 cp
= strrchr(name
,'.');
2722 strcat(name
,suffix
);
2726 /* Open a file with a name based on the name of the input file,
2727 ** but with a different (specified) suffix, and return a pointer
2729 PRIVATE
FILE *file_open(
2736 if( lemp
->outname
) free(lemp
->outname
);
2737 lemp
->outname
= file_makename(lemp
, suffix
);
2738 fp
= fopen(lemp
->outname
,mode
);
2739 if( fp
==0 && *mode
=='w' ){
2740 fprintf(stderr
,"Can't open file \"%s\".\n",lemp
->outname
);
2747 /* Duplicate the input file without comments and without actions
2749 void Reprint(struct lemon
*lemp
)
2753 int i
, j
, maxlen
, len
, ncolumns
, skip
;
2754 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp
->filename
);
2756 for(i
=0; i
<lemp
->nsymbol
; i
++){
2757 sp
= lemp
->symbols
[i
];
2758 len
= lemonStrlen(sp
->name
);
2759 if( len
>maxlen
) maxlen
= len
;
2761 ncolumns
= 76/(maxlen
+5);
2762 if( ncolumns
<1 ) ncolumns
= 1;
2763 skip
= (lemp
->nsymbol
+ ncolumns
- 1)/ncolumns
;
2764 for(i
=0; i
<skip
; i
++){
2766 for(j
=i
; j
<lemp
->nsymbol
; j
+=skip
){
2767 sp
= lemp
->symbols
[j
];
2768 assert( sp
->index
==j
);
2769 printf(" %3d %-*.*s",j
,maxlen
,maxlen
,sp
->name
);
2773 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
2774 printf("%s",rp
->lhs
->name
);
2775 /* if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
2777 for(i
=0; i
<rp
->nrhs
; i
++){
2779 printf(" %s", sp
->name
);
2780 if( sp
->type
==MULTITERMINAL
){
2781 for(j
=1; j
<sp
->nsubsym
; j
++){
2782 printf("|%s", sp
->subsym
[j
]->name
);
2785 /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
2788 if( rp
->precsym
) printf(" [%s]",rp
->precsym
->name
);
2789 /* if( rp->code ) printf("\n %s",rp->code); */
2794 void ConfigPrint(FILE *fp
, struct config
*cfp
)
2800 fprintf(fp
,"%s ::=",rp
->lhs
->name
);
2801 for(i
=0; i
<=rp
->nrhs
; i
++){
2802 if( i
==cfp
->dot
) fprintf(fp
," *");
2803 if( i
==rp
->nrhs
) break;
2805 fprintf(fp
," %s", sp
->name
);
2806 if( sp
->type
==MULTITERMINAL
){
2807 for(j
=1; j
<sp
->nsubsym
; j
++){
2808 fprintf(fp
,"|%s",sp
->subsym
[j
]->name
);
2817 PRIVATE
void SetPrint(out
,set
,lemp
)
2825 fprintf(out
,"%12s[","");
2826 for(i
=0; i
<lemp
->nterminal
; i
++){
2827 if( SetFind(set
,i
) ){
2828 fprintf(out
,"%s%s",spacer
,lemp
->symbols
[i
]->name
);
2835 /* Print a plink chain */
2836 PRIVATE
void PlinkPrint(out
,plp
,tag
)
2842 fprintf(out
,"%12s%s (state %2d) ","",tag
,plp
->cfp
->stp
->statenum
);
2843 ConfigPrint(out
,plp
->cfp
);
2850 /* Print an action to the given file descriptor. Return FALSE if
2851 ** nothing was actually printed.
2853 int PrintAction(struct action
*ap
, FILE *fp
, int indent
){
2857 fprintf(fp
,"%*s shift %d",indent
,ap
->sp
->name
,ap
->x
.stp
->statenum
);
2860 fprintf(fp
,"%*s reduce %d",indent
,ap
->sp
->name
,ap
->x
.rp
->index
);
2863 fprintf(fp
,"%*s accept",indent
,ap
->sp
->name
);
2866 fprintf(fp
,"%*s error",indent
,ap
->sp
->name
);
2870 fprintf(fp
,"%*s reduce %-3d ** Parsing conflict **",
2871 indent
,ap
->sp
->name
,ap
->x
.rp
->index
);
2874 fprintf(fp
,"%*s shift %-3d ** Parsing conflict **",
2875 indent
,ap
->sp
->name
,ap
->x
.stp
->statenum
);
2878 if( showPrecedenceConflict
){
2879 fprintf(fp
,"%*s shift %-3d -- dropped by precedence",
2880 indent
,ap
->sp
->name
,ap
->x
.stp
->statenum
);
2886 if( showPrecedenceConflict
){
2887 fprintf(fp
,"%*s reduce %-3d -- dropped by precedence",
2888 indent
,ap
->sp
->name
,ap
->x
.rp
->index
);
2900 /* Generate the "y.output" log file */
2901 void ReportOutput(struct lemon
*lemp
)
2909 fp
= file_open(lemp
,".out","wb");
2911 for(i
=0; i
<lemp
->nstate
; i
++){
2912 stp
= lemp
->sorted
[i
];
2913 fprintf(fp
,"State %d:\n",stp
->statenum
);
2914 if( lemp
->basisflag
) cfp
=stp
->bp
;
2918 if( cfp
->dot
==cfp
->rp
->nrhs
){
2919 sprintf(buf
,"(%d)",cfp
->rp
->index
);
2920 fprintf(fp
," %5s ",buf
);
2924 ConfigPrint(fp
,cfp
);
2927 SetPrint(fp
,cfp
->fws
,lemp
);
2928 PlinkPrint(fp
,cfp
->fplp
,"To ");
2929 PlinkPrint(fp
,cfp
->bplp
,"From");
2931 if( lemp
->basisflag
) cfp
=cfp
->bp
;
2935 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
2936 if( PrintAction(ap
,fp
,30) ) fprintf(fp
,"\n");
2940 fprintf(fp
, "----------------------------------------------------\n");
2941 fprintf(fp
, "Symbols:\n");
2942 for(i
=0; i
<lemp
->nsymbol
; i
++){
2946 sp
= lemp
->symbols
[i
];
2947 fprintf(fp
, " %3d: %s", i
, sp
->name
);
2948 if( sp
->type
==NONTERMINAL
){
2951 fprintf(fp
, " <lambda>");
2953 for(j
=0; j
<lemp
->nterminal
; j
++){
2954 if( sp
->firstset
&& SetFind(sp
->firstset
, j
) ){
2955 fprintf(fp
, " %s", lemp
->symbols
[j
]->name
);
2965 /* Search for the file "name" which is in the same directory as
2966 ** the exacutable */
2967 PRIVATE
char *pathsearch(char *argv0
, char *name
, int modemask
)
2969 const char *pathlist
;
2976 cp
= strrchr(argv0
,'\\');
2978 cp
= strrchr(argv0
,'/');
2983 path
= (char *)malloc( lemonStrlen(argv0
) + lemonStrlen(name
) + 2 );
2984 if( path
) sprintf(path
,"%s/%s",argv0
,name
);
2987 pathlist
= getenv("PATH");
2988 if( pathlist
==0 ) pathlist
= ".:/bin:/usr/bin";
2989 pathbuf
= (char *) malloc( lemonStrlen(pathlist
) + 1 );
2990 path
= (char *)malloc( lemonStrlen(pathlist
)+lemonStrlen(name
)+2 );
2991 if( (pathbuf
!= 0) && (path
!=0) ){
2992 pathbufptr
= pathbuf
;
2993 strcpy(pathbuf
, pathlist
);
2995 cp
= strchr(pathbuf
,':');
2996 if( cp
==0 ) cp
= &pathbuf
[lemonStrlen(pathbuf
)];
2999 sprintf(path
,"%s/%s",pathbuf
,name
);
3001 if( c
==0 ) pathbuf
[0] = 0;
3002 else pathbuf
= &cp
[1];
3003 if( access(path
,modemask
)==0 ) break;
3011 /* Given an action, compute the integer value for that action
3012 ** which is to be put in the action table of the generated machine.
3013 ** Return negative if no action should be generated.
3015 PRIVATE
int compute_action(struct lemon
*lemp
, struct action
*ap
)
3019 case SHIFT
: act
= ap
->x
.stp
->statenum
; break;
3020 case REDUCE
: act
= ap
->x
.rp
->index
+ lemp
->nstate
; break;
3021 case ERROR
: act
= lemp
->nstate
+ lemp
->nrule
; break;
3022 case ACCEPT
: act
= lemp
->nstate
+ lemp
->nrule
+ 1; break;
3023 default: act
= -1; break;
3028 #define LINESIZE 1000
3029 /* The next cluster of routines are for reading the template file
3030 ** and writing the results to the generated parser */
3031 /* The first function transfers data from "in" to "out" until
3032 ** a line is seen which begins with "%%". The line number is
3035 ** if name!=0, then any word that begin with "Parse" is changed to
3036 ** begin with *name instead.
3038 PRIVATE
void tplt_xfer(char *name
, FILE *in
, FILE *out
, int *lineno
)
3041 char line
[LINESIZE
];
3042 while( fgets(line
,LINESIZE
,in
) && (line
[0]!='%' || line
[1]!='%') ){
3046 for(i
=0; line
[i
]; i
++){
3047 if( line
[i
]=='P' && strncmp(&line
[i
],"Parse",5)==0
3048 && (i
==0 || !isalpha(line
[i
-1]))
3050 if( i
>iStart
) fprintf(out
,"%.*s",i
-iStart
,&line
[iStart
]);
3051 fprintf(out
,"%s",name
);
3057 fprintf(out
,"%s",&line
[iStart
]);
3061 /* The next function finds the template file and opens it, returning
3062 ** a pointer to the opened file. */
3063 PRIVATE
FILE *tplt_open(struct lemon
*lemp
)
3065 static char templatename
[] = "lempar.c";
3071 /* first, see if user specified a template filename on the command line. */
3072 if (user_templatename
!= 0) {
3073 if( access(user_templatename
,004)==-1 ){
3074 fprintf(stderr
,"Can't find the parser driver template file \"%s\".\n",
3079 in
= fopen(user_templatename
,"rb");
3081 fprintf(stderr
,"Can't open the template file \"%s\".\n",user_templatename
);
3088 cp
= strrchr(lemp
->filename
,'.');
3090 sprintf(buf
,"%.*s.lt",(int)(cp
-lemp
->filename
),lemp
->filename
);
3092 sprintf(buf
,"%s.lt",lemp
->filename
);
3094 if( access(buf
,004)==0 ){
3096 }else if( access(templatename
,004)==0 ){
3097 tpltname
= templatename
;
3099 tpltname
= pathsearch(lemp
->argv0
,templatename
,0);
3102 fprintf(stderr
,"Can't find the parser driver template file \"%s\".\n",
3107 in
= fopen(tpltname
,"rb");
3109 fprintf(stderr
,"Can't open the template file \"%s\".\n",templatename
);
3116 /* Print a #line directive line to the output file. */
3117 PRIVATE
void tplt_linedir(FILE *out
, int lineno
, char *filename
)
3119 fprintf(out
,"#line %d \"",lineno
);
3121 if( *filename
== '\\' ) putc('\\',out
);
3122 putc(*filename
,out
);
3125 fprintf(out
,"\"\n");
3128 /* Print a string to the file and keep the linenumber up to date */
3129 PRIVATE
void tplt_print(FILE *out
, struct lemon
*lemp
, char *str
, int *lineno
)
3131 if( str
==0 ) return;
3134 if( *str
=='\n' ) (*lineno
)++;
3137 if( str
[-1]!='\n' ){
3141 if (!lemp
->nolinenosflag
) {
3142 (*lineno
)++; tplt_linedir(out
,*lineno
,lemp
->outname
);
3148 ** The following routine emits code for the destructor for the
3151 void emit_destructor_code(
3159 if( sp
->type
==TERMINAL
){
3160 cp
= lemp
->tokendest
;
3162 fprintf(out
,"{\n"); (*lineno
)++;
3163 }else if( sp
->destructor
){
3164 cp
= sp
->destructor
;
3165 fprintf(out
,"{\n"); (*lineno
)++;
3166 if (!lemp
->nolinenosflag
) { (*lineno
)++; tplt_linedir(out
,sp
->destLineno
,lemp
->filename
); }
3167 }else if( lemp
->vardest
){
3170 fprintf(out
,"{\n"); (*lineno
)++;
3172 assert( 0 ); /* Cannot happen */
3175 if( *cp
=='$' && cp
[1]=='$' ){
3176 fprintf(out
,"(yypminor->yy%d)",sp
->dtnum
);
3180 if( *cp
=='\n' ) (*lineno
)++;
3183 fprintf(out
,"\n"); (*lineno
)++;
3184 if (!lemp
->nolinenosflag
) {
3185 (*lineno
)++; tplt_linedir(out
,*lineno
,lemp
->outname
);
3187 fprintf(out
,"}\n"); (*lineno
)++;
3192 ** Return TRUE (non-zero) if the given symbol has a destructor.
3194 int has_destructor(struct symbol
*sp
, struct lemon
*lemp
)
3197 if( sp
->type
==TERMINAL
){
3198 ret
= lemp
->tokendest
!=0;
3200 ret
= lemp
->vardest
!=0 || sp
->destructor
!=0;
3206 ** Append text to a dynamically allocated string. If zText is 0 then
3207 ** reset the string to be empty again. Always return the complete text
3208 ** of the string (which is overwritten with each call).
3210 ** n bytes of zText are stored. If n==0 then all of zText up to the first
3211 ** \000 terminator is stored. zText can contain up to two instances of
3212 ** %d. The values of p1 and p2 are written into the first and second
3215 ** If n==-1, then the previous character is overwritten.
3217 PRIVATE
char *append_str(const char *zText
, int n
, int p1
, int p2
){
3218 static char empty
[1] = { 0 };
3220 static int alloced
= 0;
3221 static int used
= 0;
3233 n
= lemonStrlen(zText
);
3235 if( (int) (n
+sizeof(zInt
)*2+used
) >= alloced
){
3236 alloced
= n
+ sizeof(zInt
)*2 + used
+ 200;
3237 z
= (char *) realloc(z
, alloced
);
3239 if( z
==0 ) return empty
;
3242 if( c
=='%' && n
>0 && zText
[0]=='d' ){
3243 sprintf(zInt
, "%d", p1
);
3245 strcpy(&z
[used
], zInt
);
3246 used
+= lemonStrlen(&z
[used
]);
3258 ** zCode is a string that is the action associated with a rule. Expand
3259 ** the symbols in this string so that the refer to elements of the parser
3262 PRIVATE
void translate_code(struct lemon
*lemp
, struct rule
*rp
){
3265 char lhsused
= 0; /* True if the LHS element has been used */
3266 char used
[MAXRHS
]; /* True for each RHS element which is used */
3268 for(i
=0; i
<rp
->nrhs
; i
++) used
[i
] = 0;
3272 static char newlinestr
[2] = { '\n', '\0' };
3273 rp
->code
= newlinestr
;
3274 rp
->line
= rp
->ruleline
;
3277 append_str(0,0,0,0);
3279 /* This const cast is wrong but harmless, if we're careful. */
3280 for(cp
=(char *)rp
->code
; *cp
; cp
++){
3281 if( isalpha(*cp
) && (cp
==rp
->code
|| (!isalnum(cp
[-1]) && cp
[-1]!='_')) ){
3283 for(xp
= &cp
[1]; isalnum(*xp
) || *xp
=='_'; xp
++);
3286 if( rp
->lhsalias
&& strcmp(cp
,rp
->lhsalias
)==0 ){
3287 append_str("yygotominor.yy%d",0,rp
->lhs
->dtnum
,0);
3291 for(i
=0; i
<rp
->nrhs
; i
++){
3292 if( rp
->rhsalias
[i
] && strcmp(cp
,rp
->rhsalias
[i
])==0 ){
3293 if( cp
!=rp
->code
&& cp
[-1]=='@' ){
3294 /* If the argument is of the form @X then substituted
3295 ** the token number of X, not the value of X */
3296 append_str("yymsp[%d].major",-1,i
-rp
->nrhs
+1,0);
3298 struct symbol
*sp
= rp
->rhs
[i
];
3300 if( sp
->type
==MULTITERMINAL
){
3301 dtnum
= sp
->subsym
[0]->dtnum
;
3305 append_str("yymsp[%d].minor.yy%d",0,i
-rp
->nrhs
+1, dtnum
);
3315 append_str(cp
, 1, 0, 0);
3318 /* Check to make sure the LHS has been used */
3319 if( rp
->lhsalias
&& !lhsused
){
3320 ErrorMsg(lemp
->filename
,rp
->ruleline
,
3321 "Label \"%s\" for \"%s(%s)\" is never used.",
3322 rp
->lhsalias
,rp
->lhs
->name
,rp
->lhsalias
);
3326 /* Generate destructor code for RHS symbols which are not used in the
3328 for(i
=0; i
<rp
->nrhs
; i
++){
3329 if( rp
->rhsalias
[i
] && !used
[i
] ){
3330 ErrorMsg(lemp
->filename
,rp
->ruleline
,
3331 "Label %s for \"%s(%s)\" is never used.",
3332 rp
->rhsalias
[i
],rp
->rhs
[i
]->name
,rp
->rhsalias
[i
]);
3334 }else if( rp
->rhsalias
[i
]==0 ){
3335 if( has_destructor(rp
->rhs
[i
],lemp
) ){
3336 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3337 rp
->rhs
[i
]->index
,i
-rp
->nrhs
+1);
3339 /* No destructor defined for this term */
3344 cp
= append_str(0,0,0,0);
3345 rp
->code
= Strsafe(cp
?cp
:"");
3350 ** Generate code which executes when the rule "rp" is reduced. Write
3351 ** the code to "out". Make sure lineno stays up-to-date.
3353 PRIVATE
void emit_code(
3361 /* Generate code to do the reduce action */
3363 if (!lemp
->nolinenosflag
) { (*lineno
)++; tplt_linedir(out
,rp
->line
,lemp
->filename
); }
3364 fprintf(out
,"{%s",rp
->code
);
3365 for(cp
=rp
->code
; *cp
; cp
++){
3366 if( *cp
=='\n' ) (*lineno
)++;
3368 fprintf(out
,"}\n"); (*lineno
)++;
3369 if (!lemp
->nolinenosflag
) { (*lineno
)++; tplt_linedir(out
,*lineno
,lemp
->outname
); }
3370 } /* End if( rp->code ) */
3376 ** Print the definition of the union used for the parser's data stack.
3377 ** This union contains fields for every possible data type for tokens
3378 ** and nonterminals. In the process of computing and printing this
3379 ** union, also set the ".dtnum" field of every terminal and nonterminal
3382 void print_stack_union(
3383 FILE *out
, /* The output stream */
3384 struct lemon
*lemp
, /* The main info structure for this parser */
3385 int *plineno
, /* Pointer to the line number */
3386 int mhflag
/* True if generating makeheaders output */
3388 int lineno
= *plineno
; /* The line number of the output */
3389 char **types
; /* A hash table of datatypes */
3390 int arraysize
; /* Size of the "types" array */
3391 int maxdtlength
; /* Maximum length of any ".datatype" field. */
3392 char *stddt
; /* Standardized name for a datatype */
3393 int i
,j
; /* Loop counters */
3394 int hash
; /* For hashing the name of a type */
3395 const char *name
; /* Name of the parser */
3397 /* Allocate and initialize types[] and allocate stddt[] */
3398 arraysize
= lemp
->nsymbol
* 2;
3399 types
= (char**)calloc( arraysize
, sizeof(char*) );
3401 fprintf(stderr
,"Out of memory.\n");
3404 for(i
=0; i
<arraysize
; i
++) types
[i
] = 0;
3406 if( lemp
->vartype
){
3407 maxdtlength
= lemonStrlen(lemp
->vartype
);
3409 for(i
=0; i
<lemp
->nsymbol
; i
++){
3411 struct symbol
*sp
= lemp
->symbols
[i
];
3412 if( sp
->datatype
==0 ) continue;
3413 len
= lemonStrlen(sp
->datatype
);
3414 if( len
>maxdtlength
) maxdtlength
= len
;
3416 stddt
= (char*)malloc( maxdtlength
*2 + 1 );
3418 fprintf(stderr
,"Out of memory.\n");
3422 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3423 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
3424 ** used for terminal symbols. If there is no %default_type defined then
3425 ** 0 is also used as the .dtnum value for nonterminals which do not specify
3426 ** a datatype using the %type directive.
3428 for(i
=0; i
<lemp
->nsymbol
; i
++){
3429 struct symbol
*sp
= lemp
->symbols
[i
];
3431 if( sp
==lemp
->errsym
){
3432 sp
->dtnum
= arraysize
+1;
3435 if( sp
->type
!=NONTERMINAL
|| (sp
->datatype
==0 && lemp
->vartype
==0) ){
3440 if( cp
==0 ) cp
= lemp
->vartype
;
3442 while( isspace(*cp
) ) cp
++;
3443 while( *cp
) stddt
[j
++] = *cp
++;
3444 while( j
>0 && isspace(stddt
[j
-1]) ) j
--;
3446 if( lemp
->tokentype
&& strcmp(stddt
, lemp
->tokentype
)==0 ){
3451 for(j
=0; stddt
[j
]; j
++){
3452 hash
= hash
*53 + stddt
[j
];
3454 hash
= (hash
& 0x7fffffff)%arraysize
;
3455 while( types
[hash
] ){
3456 if( strcmp(types
[hash
],stddt
)==0 ){
3457 sp
->dtnum
= hash
+ 1;
3461 if( hash
>=arraysize
) hash
= 0;
3463 if( types
[hash
]==0 ){
3464 sp
->dtnum
= hash
+ 1;
3465 types
[hash
] = (char*)malloc( lemonStrlen(stddt
)+1 );
3466 if( types
[hash
]==0 ){
3467 fprintf(stderr
,"Out of memory.\n");
3470 strcpy(types
[hash
],stddt
);
3474 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
3475 name
= lemp
->name
? lemp
->name
: "Parse";
3477 if( mhflag
){ fprintf(out
,"#if INTERFACE\n"); lineno
++; }
3478 fprintf(out
,"#define %sTOKENTYPE %s\n",name
,
3479 lemp
->tokentype
?lemp
->tokentype
:"void*"); lineno
++;
3480 if( mhflag
){ fprintf(out
,"#endif\n"); lineno
++; }
3481 fprintf(out
,"typedef union {\n"); lineno
++;
3482 fprintf(out
," int yyinit;\n"); lineno
++;
3483 fprintf(out
," %sTOKENTYPE yy0;\n",name
); lineno
++;
3484 for(i
=0; i
<arraysize
; i
++){
3485 if( types
[i
]==0 ) continue;
3486 fprintf(out
," %s yy%d;\n",types
[i
],i
+1); lineno
++;
3489 if( lemp
->errsym
->useCnt
){
3490 fprintf(out
," int yy%d;\n",lemp
->errsym
->dtnum
); lineno
++;
3494 fprintf(out
,"} YYMINORTYPE;\n"); lineno
++;
3499 ** Return the name of a C datatype able to represent values between
3500 ** lwr and upr, inclusive.
3502 static const char *minimum_size_type(int lwr
, int upr
){
3505 return "unsigned char";
3506 }else if( upr
<65535 ){
3507 return "unsigned short int";
3509 return "unsigned int";
3511 }else if( lwr
>=-127 && upr
<=127 ){
3512 return "signed char";
3513 }else if( lwr
>=-32767 && upr
<32767 ){
3521 ** Each state contains a set of token transaction and a set of
3522 ** nonterminal transactions. Each of these sets makes an instance
3523 ** of the following structure. An array of these structures is used
3524 ** to order the creation of entries in the yy_action[] table.
3527 struct state
*stp
; /* A pointer to a state */
3528 int isTkn
; /* True to use tokens. False for non-terminals */
3529 int nAction
; /* Number of actions */
3530 int iOrder
; /* Original order of action sets */
3534 ** Compare to axset structures for sorting purposes
3536 static int axset_compare(const void *a
, const void *b
){
3537 struct axset
*p1
= (struct axset
*)a
;
3538 struct axset
*p2
= (struct axset
*)b
;
3540 c
= p2
->nAction
- p1
->nAction
;
3542 c
= p2
->iOrder
- p1
->iOrder
;
3544 assert( c
!=0 || p1
==p2
);
3549 ** Write text on "out" that describes the rule "rp".
3551 static void writeRuleText(FILE *out
, struct rule
*rp
){
3553 fprintf(out
,"%s ::=", rp
->lhs
->name
);
3554 for(j
=0; j
<rp
->nrhs
; j
++){
3555 struct symbol
*sp
= rp
->rhs
[j
];
3556 fprintf(out
," %s", sp
->name
);
3557 if( sp
->type
==MULTITERMINAL
){
3559 for(k
=1; k
<sp
->nsubsym
; k
++){
3560 fprintf(out
,"|%s",sp
->subsym
[k
]->name
);
3567 /* Generate C source code for the parser */
3570 int mhflag
/* Output in makeheaders format if true */
3573 char line
[LINESIZE
];
3578 struct acttab
*pActtab
;
3581 int mnTknOfst
, mxTknOfst
;
3582 int mnNtOfst
, mxNtOfst
;
3585 in
= tplt_open(lemp
);
3587 out
= file_open(lemp
,".c","wb");
3593 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3595 /* Generate the include code, if any */
3596 tplt_print(out
,lemp
,lemp
->include
,&lineno
);
3598 char *name
= file_makename(lemp
, ".h");
3599 fprintf(out
,"#include \"%s\"\n", name
); lineno
++;
3602 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3604 /* Generate #defines for all tokens */
3607 fprintf(out
,"#if INTERFACE\n"); lineno
++;
3608 if( lemp
->tokenprefix
) prefix
= lemp
->tokenprefix
;
3610 for(i
=1; i
<lemp
->nterminal
; i
++){
3611 fprintf(out
,"#define %s%-30s %2d\n",prefix
,lemp
->symbols
[i
]->name
,i
);
3614 fprintf(out
,"#endif\n"); lineno
++;
3616 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3618 /* Generate the defines */
3619 fprintf(out
,"#define YYCODETYPE %s\n",
3620 minimum_size_type(0, lemp
->nsymbol
+1)); lineno
++;
3621 fprintf(out
,"#define YYNOCODE %d\n",lemp
->nsymbol
+1); lineno
++;
3622 fprintf(out
,"#define YYACTIONTYPE %s\n",
3623 minimum_size_type(0, lemp
->nstate
+lemp
->nrule
+5)); lineno
++;
3624 if( lemp
->wildcard
){
3625 fprintf(out
,"#define YYWILDCARD %d\n",
3626 lemp
->wildcard
->index
); lineno
++;
3628 print_stack_union(out
,lemp
,&lineno
,mhflag
);
3629 fprintf(out
, "#ifndef YYSTACKDEPTH\n"); lineno
++;
3630 if( lemp
->stacksize
){
3631 fprintf(out
,"#define YYSTACKDEPTH %s\n",lemp
->stacksize
); lineno
++;
3633 fprintf(out
,"#define YYSTACKDEPTH 100\n"); lineno
++;
3635 fprintf(out
, "#endif\n"); lineno
++;
3637 fprintf(out
,"#if INTERFACE\n"); lineno
++;
3639 name
= lemp
->name
? lemp
->name
: "Parse";
3640 if( lemp
->arg
&& lemp
->arg
[0] ){
3642 i
= lemonStrlen(lemp
->arg
);
3643 while( i
>=1 && isspace(lemp
->arg
[i
-1]) ) i
--;
3644 while( i
>=1 && (isalnum(lemp
->arg
[i
-1]) || lemp
->arg
[i
-1]=='_') ) i
--;
3645 fprintf(out
,"#define %sARG_SDECL %s;\n",name
,lemp
->arg
); lineno
++;
3646 fprintf(out
,"#define %sARG_PDECL ,%s\n",name
,lemp
->arg
); lineno
++;
3647 fprintf(out
,"#define %sARG_FETCH %s = yypParser->%s\n",
3648 name
,lemp
->arg
,&lemp
->arg
[i
]); lineno
++;
3649 fprintf(out
,"#define %sARG_STORE yypParser->%s = %s\n",
3650 name
,&lemp
->arg
[i
],&lemp
->arg
[i
]); lineno
++;
3652 fprintf(out
,"#define %sARG_SDECL\n",name
); lineno
++;
3653 fprintf(out
,"#define %sARG_PDECL\n",name
); lineno
++;
3654 fprintf(out
,"#define %sARG_FETCH\n",name
); lineno
++;
3655 fprintf(out
,"#define %sARG_STORE\n",name
); lineno
++;
3658 fprintf(out
,"#endif\n"); lineno
++;
3660 fprintf(out
,"#define YYNSTATE %d\n",lemp
->nstate
); lineno
++;
3661 fprintf(out
,"#define YYNRULE %d\n",lemp
->nrule
); lineno
++;
3662 if( lemp
->errsym
->useCnt
){
3663 fprintf(out
,"#define YYERRORSYMBOL %d\n",lemp
->errsym
->index
); lineno
++;
3664 fprintf(out
,"#define YYERRSYMDT yy%d\n",lemp
->errsym
->dtnum
); lineno
++;
3666 if( lemp
->has_fallback
){
3667 fprintf(out
,"#define YYFALLBACK 1\n"); lineno
++;
3669 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3671 /* Generate the action table and its associates:
3673 ** yy_action[] A single table containing all actions.
3674 ** yy_lookahead[] A table containing the lookahead for each entry in
3675 ** yy_action. Used to detect hash collisions.
3676 ** yy_shift_ofst[] For each state, the offset into yy_action for
3677 ** shifting terminals.
3678 ** yy_reduce_ofst[] For each state, the offset into yy_action for
3679 ** shifting non-terminals after a reduce.
3680 ** yy_default[] Default action for each state.
3683 /* Compute the actions on all states and count them up */
3684 ax
= (struct axset
*) calloc(lemp
->nstate
*2, sizeof(ax
[0]));
3686 fprintf(stderr
,"malloc failed\n");
3689 for(i
=0; i
<lemp
->nstate
; i
++){
3690 stp
= lemp
->sorted
[i
];
3693 ax
[i
*2].nAction
= stp
->nTknAct
;
3694 ax
[i
*2+1].stp
= stp
;
3695 ax
[i
*2+1].isTkn
= 0;
3696 ax
[i
*2+1].nAction
= stp
->nNtAct
;
3698 mxTknOfst
= mnTknOfst
= 0;
3699 mxNtOfst
= mnNtOfst
= 0;
3701 /* Compute the action table. In order to try to keep the size of the
3702 ** action table to a minimum, the heuristic of placing the largest action
3703 ** sets first is used.
3705 for(i
=0; i
<lemp
->nstate
*2; i
++) ax
[i
].iOrder
= i
;
3706 qsort(ax
, lemp
->nstate
*2, sizeof(ax
[0]), axset_compare
);
3707 pActtab
= acttab_alloc();
3708 for(i
=0; i
<lemp
->nstate
*2 && ax
[i
].nAction
>0; i
++){
3711 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3713 if( ap
->sp
->index
>=lemp
->nterminal
) continue;
3714 action
= compute_action(lemp
, ap
);
3715 if( action
<0 ) continue;
3716 acttab_action(pActtab
, ap
->sp
->index
, action
);
3718 stp
->iTknOfst
= acttab_insert(pActtab
);
3719 if( stp
->iTknOfst
<mnTknOfst
) mnTknOfst
= stp
->iTknOfst
;
3720 if( stp
->iTknOfst
>mxTknOfst
) mxTknOfst
= stp
->iTknOfst
;
3722 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
3724 if( ap
->sp
->index
<lemp
->nterminal
) continue;
3725 if( ap
->sp
->index
==lemp
->nsymbol
) continue;
3726 action
= compute_action(lemp
, ap
);
3727 if( action
<0 ) continue;
3728 acttab_action(pActtab
, ap
->sp
->index
, action
);
3730 stp
->iNtOfst
= acttab_insert(pActtab
);
3731 if( stp
->iNtOfst
<mnNtOfst
) mnNtOfst
= stp
->iNtOfst
;
3732 if( stp
->iNtOfst
>mxNtOfst
) mxNtOfst
= stp
->iNtOfst
;
3737 /* Output the yy_action table */
3738 n
= acttab_size(pActtab
);
3739 fprintf(out
,"#define YY_ACTTAB_COUNT (%d)\n", n
); lineno
++;
3740 fprintf(out
,"static const YYACTIONTYPE yy_action[] = {\n"); lineno
++;
3741 for(i
=j
=0; i
<n
; i
++){
3742 int action
= acttab_yyaction(pActtab
, i
);
3743 if( action
<0 ) action
= lemp
->nstate
+ lemp
->nrule
+ 2;
3744 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3745 fprintf(out
, " %4d,", action
);
3746 if( j
==9 || i
==n
-1 ){
3747 fprintf(out
, "\n"); lineno
++;
3753 fprintf(out
, "};\n"); lineno
++;
3755 /* Output the yy_lookahead table */
3756 fprintf(out
,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno
++;
3757 for(i
=j
=0; i
<n
; i
++){
3758 int la
= acttab_yylookahead(pActtab
, i
);
3759 if( la
<0 ) la
= lemp
->nsymbol
;
3760 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3761 fprintf(out
, " %4d,", la
);
3762 if( j
==9 || i
==n
-1 ){
3763 fprintf(out
, "\n"); lineno
++;
3769 fprintf(out
, "};\n"); lineno
++;
3771 /* Output the yy_shift_ofst[] table */
3772 fprintf(out
, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst
-1); lineno
++;
3774 while( n
>0 && lemp
->sorted
[n
-1]->iTknOfst
==NO_OFFSET
) n
--;
3775 fprintf(out
, "#define YY_SHIFT_COUNT (%d)\n", n
-1); lineno
++;
3776 fprintf(out
, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst
); lineno
++;
3777 fprintf(out
, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst
); lineno
++;
3778 fprintf(out
, "static const %s yy_shift_ofst[] = {\n",
3779 minimum_size_type(mnTknOfst
-1, mxTknOfst
)); lineno
++;
3780 for(i
=j
=0; i
<n
; i
++){
3782 stp
= lemp
->sorted
[i
];
3783 ofst
= stp
->iTknOfst
;
3784 if( ofst
==NO_OFFSET
) ofst
= mnTknOfst
- 1;
3785 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3786 fprintf(out
, " %4d,", ofst
);
3787 if( j
==9 || i
==n
-1 ){
3788 fprintf(out
, "\n"); lineno
++;
3794 fprintf(out
, "};\n"); lineno
++;
3796 /* Output the yy_reduce_ofst[] table */
3797 fprintf(out
, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst
-1); lineno
++;
3799 while( n
>0 && lemp
->sorted
[n
-1]->iNtOfst
==NO_OFFSET
) n
--;
3800 fprintf(out
, "#define YY_REDUCE_COUNT (%d)\n", n
-1); lineno
++;
3801 fprintf(out
, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst
); lineno
++;
3802 fprintf(out
, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst
); lineno
++;
3803 fprintf(out
, "static const %s yy_reduce_ofst[] = {\n",
3804 minimum_size_type(mnNtOfst
-1, mxNtOfst
)); lineno
++;
3805 for(i
=j
=0; i
<n
; i
++){
3807 stp
= lemp
->sorted
[i
];
3808 ofst
= stp
->iNtOfst
;
3809 if( ofst
==NO_OFFSET
) ofst
= mnNtOfst
- 1;
3810 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3811 fprintf(out
, " %4d,", ofst
);
3812 if( j
==9 || i
==n
-1 ){
3813 fprintf(out
, "\n"); lineno
++;
3819 fprintf(out
, "};\n"); lineno
++;
3821 /* Output the default action table */
3822 fprintf(out
, "static const YYACTIONTYPE yy_default[] = {\n"); lineno
++;
3824 for(i
=j
=0; i
<n
; i
++){
3825 stp
= lemp
->sorted
[i
];
3826 if( j
==0 ) fprintf(out
," /* %5d */ ", i
);
3827 fprintf(out
, " %4d,", stp
->iDflt
);
3828 if( j
==9 || i
==n
-1 ){
3829 fprintf(out
, "\n"); lineno
++;
3835 fprintf(out
, "};\n"); lineno
++;
3836 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3838 /* Generate the table of fallback tokens.
3840 if( lemp
->has_fallback
){
3841 int mx
= lemp
->nterminal
- 1;
3842 while( mx
>0 && lemp
->symbols
[mx
]->fallback
==0 ){ mx
--; }
3843 for(i
=0; i
<=mx
; i
++){
3844 struct symbol
*p
= lemp
->symbols
[i
];
3845 if( p
->fallback
==0 ){
3846 fprintf(out
, " 0, /* %10s => nothing */\n", p
->name
);
3848 fprintf(out
, " %3d, /* %10s => %s */\n", p
->fallback
->index
,
3849 p
->name
, p
->fallback
->name
);
3854 tplt_xfer(lemp
->name
, in
, out
, &lineno
);
3856 /* Generate a table containing the symbolic name of every symbol
3858 for(i
=0; i
<lemp
->nsymbol
; i
++){
3859 sprintf(line
,"\"%s\",",lemp
->symbols
[i
]->name
);
3860 fprintf(out
," %-15s",line
);
3861 if( (i
&3)==3 ){ fprintf(out
,"\n"); lineno
++; }
3863 if( (i
&3)!=0 ){ fprintf(out
,"\n"); lineno
++; }
3864 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3866 /* Generate a table containing a text string that describes every
3867 ** rule in the rule set of the grammar. This information is used
3868 ** when tracing REDUCE actions.
3870 for(i
=0, rp
=lemp
->rule
; rp
; rp
=rp
->next
, i
++){
3871 assert( rp
->index
==i
);
3872 fprintf(out
," /* %3d */ \"", i
);
3873 writeRuleText(out
, rp
);
3874 fprintf(out
,"\",\n"); lineno
++;
3876 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3878 /* Generate code which executes every time a symbol is popped from
3879 ** the stack while processing errors or while destroying the parser.
3880 ** (In other words, generate the %destructor actions)
3882 if( lemp
->tokendest
){
3884 for(i
=0; i
<lemp
->nsymbol
; i
++){
3885 struct symbol
*sp
= lemp
->symbols
[i
];
3886 if( sp
==0 || sp
->type
!=TERMINAL
) continue;
3888 fprintf(out
, " /* TERMINAL Destructor */\n"); lineno
++;
3891 fprintf(out
," case %d: /* %s */\n", sp
->index
, sp
->name
); lineno
++;
3893 for(i
=0; i
<lemp
->nsymbol
&& lemp
->symbols
[i
]->type
!=TERMINAL
; i
++);
3894 if( i
<lemp
->nsymbol
){
3895 emit_destructor_code(out
,lemp
->symbols
[i
],lemp
,&lineno
);
3896 fprintf(out
," break;\n"); lineno
++;
3899 if( lemp
->vardest
){
3900 struct symbol
*dflt_sp
= 0;
3902 for(i
=0; i
<lemp
->nsymbol
; i
++){
3903 struct symbol
*sp
= lemp
->symbols
[i
];
3904 if( sp
==0 || sp
->type
==TERMINAL
||
3905 sp
->index
<=0 || sp
->destructor
!=0 ) continue;
3907 fprintf(out
, " /* Default NON-TERMINAL Destructor */\n"); lineno
++;
3910 fprintf(out
," case %d: /* %s */\n", sp
->index
, sp
->name
); lineno
++;
3914 emit_destructor_code(out
,dflt_sp
,lemp
,&lineno
);
3916 fprintf(out
," break;\n"); lineno
++;
3918 for(i
=0; i
<lemp
->nsymbol
; i
++){
3919 struct symbol
*sp
= lemp
->symbols
[i
];
3920 if( sp
==0 || sp
->type
==TERMINAL
|| sp
->destructor
==0 ) continue;
3921 fprintf(out
," case %d: /* %s */\n", sp
->index
, sp
->name
); lineno
++;
3923 /* Combine duplicate destructors into a single case */
3924 for(j
=i
+1; j
<lemp
->nsymbol
; j
++){
3925 struct symbol
*sp2
= lemp
->symbols
[j
];
3926 if( sp2
&& sp2
->type
!=TERMINAL
&& sp2
->destructor
3927 && sp2
->dtnum
==sp
->dtnum
3928 && strcmp(sp
->destructor
,sp2
->destructor
)==0 ){
3929 fprintf(out
," case %d: /* %s */\n",
3930 sp2
->index
, sp2
->name
); lineno
++;
3931 sp2
->destructor
= 0;
3935 emit_destructor_code(out
,lemp
->symbols
[i
],lemp
,&lineno
);
3936 fprintf(out
," break;\n"); lineno
++;
3938 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3940 /* Generate code which executes whenever the parser stack overflows */
3941 tplt_print(out
,lemp
,lemp
->overflow
,&lineno
);
3942 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3944 /* Generate the table of rule information
3946 ** Note: This code depends on the fact that rules are number
3947 ** sequentually beginning with 0.
3949 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
3950 fprintf(out
," { %d, %d },\n",rp
->lhs
->index
,rp
->nrhs
); lineno
++;
3952 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3954 /* Generate code which execution during each REDUCE action */
3955 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
3956 translate_code(lemp
, rp
);
3958 /* First output rules other than the default: rule */
3959 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
3960 struct rule
*rp2
; /* Other rules with the same action */
3961 if( rp
->code
==0 ) continue;
3962 if( rp
->code
[0]=='\n' && rp
->code
[1]==0 ) continue; /* Will be default: */
3963 fprintf(out
," case %d: /* ", rp
->index
);
3964 writeRuleText(out
, rp
);
3965 fprintf(out
, " */\n"); lineno
++;
3966 for(rp2
=rp
->next
; rp2
; rp2
=rp2
->next
){
3967 if( rp2
->code
==rp
->code
){
3968 fprintf(out
," case %d: /* ", rp2
->index
);
3969 writeRuleText(out
, rp2
);
3970 fprintf(out
," */ yytestcase(yyruleno==%d);\n", rp2
->index
); lineno
++;
3974 emit_code(out
,rp
,lemp
,&lineno
);
3975 fprintf(out
," break;\n"); lineno
++;
3978 /* Finally, output the default: rule. We choose as the default: all
3979 ** empty actions. */
3980 fprintf(out
," default:\n"); lineno
++;
3981 for(rp
=lemp
->rule
; rp
; rp
=rp
->next
){
3982 if( rp
->code
==0 ) continue;
3983 assert( rp
->code
[0]=='\n' && rp
->code
[1]==0 );
3984 fprintf(out
," /* (%d) ", rp
->index
);
3985 writeRuleText(out
, rp
);
3986 fprintf(out
, " */ yytestcase(yyruleno==%d);\n", rp
->index
); lineno
++;
3988 fprintf(out
," break;\n"); lineno
++;
3989 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3991 /* Generate code which executes if a parse fails */
3992 tplt_print(out
,lemp
,lemp
->failure
,&lineno
);
3993 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3995 /* Generate code which executes when a syntax error occurs */
3996 tplt_print(out
,lemp
,lemp
->error
,&lineno
);
3997 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
3999 /* Generate code which executes when the parser accepts its input */
4000 tplt_print(out
,lemp
,lemp
->accept
,&lineno
);
4001 tplt_xfer(lemp
->name
,in
,out
,&lineno
);
4003 /* Append any addition code the user desires */
4004 tplt_print(out
,lemp
,lemp
->extracode
,&lineno
);
4011 /* Generate a header file for the parser */
4012 void ReportHeader(struct lemon
*lemp
)
4016 char line
[LINESIZE
];
4017 char pattern
[LINESIZE
];
4020 if( lemp
->tokenprefix
) prefix
= lemp
->tokenprefix
;
4022 in
= file_open(lemp
,".h","rb");
4025 for(i
=1; i
<lemp
->nterminal
&& fgets(line
,LINESIZE
,in
); i
++){
4026 sprintf(pattern
,"#define %s%-30s %2d\n",prefix
,lemp
->symbols
[i
]->name
,i
);
4027 if( strcmp(line
,pattern
) ) break;
4029 nextChar
= fgetc(in
);
4031 if( i
==lemp
->nterminal
&& nextChar
==EOF
){
4032 /* No change in the file. Don't rewrite it. */
4036 out
= file_open(lemp
,".h","wb");
4038 for(i
=1; i
<lemp
->nterminal
; i
++){
4039 fprintf(out
,"#define %s%-30s %2d\n",prefix
,lemp
->symbols
[i
]->name
,i
);
4046 /* Reduce the size of the action tables, if possible, by making use
4049 ** In this version, we take the most frequent REDUCE action and make
4050 ** it the default. Except, there is no default if the wildcard token
4051 ** is a possible look-ahead.
4053 void CompressTables(struct lemon
*lemp
)
4056 struct action
*ap
, *ap2
;
4057 struct rule
*rp
, *rp2
, *rbest
;
4062 for(i
=0; i
<lemp
->nstate
; i
++){
4063 stp
= lemp
->sorted
[i
];
4068 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
4069 if( ap
->type
==SHIFT
&& ap
->sp
==lemp
->wildcard
){
4072 if( ap
->type
!=REDUCE
) continue;
4074 if( rp
->lhsStart
) continue;
4075 if( rp
==rbest
) continue;
4077 for(ap2
=ap
->next
; ap2
; ap2
=ap2
->next
){
4078 if( ap2
->type
!=REDUCE
) continue;
4080 if( rp2
==rbest
) continue;
4089 /* Do not make a default if the number of rules to default
4090 ** is not at least 1 or if the wildcard token is a possible
4093 if( nbest
<1 || usesWildcard
) continue;
4096 /* Combine matching REDUCE actions into a single default */
4097 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
4098 if( ap
->type
==REDUCE
&& ap
->x
.rp
==rbest
) break;
4101 ap
->sp
= Symbol_new("{default}");
4102 for(ap
=ap
->next
; ap
; ap
=ap
->next
){
4103 if( ap
->type
==REDUCE
&& ap
->x
.rp
==rbest
) ap
->type
= NOT_USED
;
4105 stp
->ap
= Action_sort(stp
->ap
);
4111 ** Compare two states for sorting purposes. The smaller state is the
4112 ** one with the most non-terminal actions. If they have the same number
4113 ** of non-terminal actions, then the smaller is the one with the most
4116 static int stateResortCompare(const void *a
, const void *b
){
4117 const struct state
*pA
= *(const struct state
**)a
;
4118 const struct state
*pB
= *(const struct state
**)b
;
4121 n
= pB
->nNtAct
- pA
->nNtAct
;
4123 n
= pB
->nTknAct
- pA
->nTknAct
;
4125 n
= pB
->statenum
- pA
->statenum
;
4134 ** Renumber and resort states so that states with fewer choices
4135 ** occur at the end. Except, keep state 0 as the first state.
4137 void ResortStates(struct lemon
*lemp
)
4143 for(i
=0; i
<lemp
->nstate
; i
++){
4144 stp
= lemp
->sorted
[i
];
4145 stp
->nTknAct
= stp
->nNtAct
= 0;
4146 stp
->iDflt
= lemp
->nstate
+ lemp
->nrule
;
4147 stp
->iTknOfst
= NO_OFFSET
;
4148 stp
->iNtOfst
= NO_OFFSET
;
4149 for(ap
=stp
->ap
; ap
; ap
=ap
->next
){
4150 if( compute_action(lemp
,ap
)>=0 ){
4151 if( ap
->sp
->index
<lemp
->nterminal
){
4153 }else if( ap
->sp
->index
<lemp
->nsymbol
){
4156 stp
->iDflt
= compute_action(lemp
, ap
);
4161 qsort(&lemp
->sorted
[1], lemp
->nstate
-1, sizeof(lemp
->sorted
[0]),
4162 stateResortCompare
);
4163 for(i
=0; i
<lemp
->nstate
; i
++){
4164 lemp
->sorted
[i
]->statenum
= i
;
4169 /***************** From the file "set.c" ************************************/
4171 ** Set manipulation routines for the LEMON parser generator.
4174 static int size
= 0;
4176 /* Set the set size */
4182 /* Allocate a new set */
4185 s
= (char*)calloc( size
, 1);
4187 extern void memory_error();
4193 /* Deallocate a set */
4194 void SetFree(char *s
)
4199 /* Add a new element to the set. Return TRUE if the element was added
4200 ** and FALSE if it was already there. */
4201 int SetAdd(char *s
, int e
)
4204 assert( e
>=0 && e
<size
);
4210 /* Add every element of s2 to s1. Return TRUE if s1 changes. */
4211 int SetUnion(char *s1
, char *s2
)
4215 for(i
=0; i
<size
; i
++){
4216 if( s2
[i
]==0 ) continue;
4224 /********************** From the file "table.c" ****************************/
4226 ** All code in this file has been automatically generated
4227 ** from a specification in the file
4229 ** by the associative array code building program "aagen".
4230 ** Do not edit this file! Instead, edit the specification
4231 ** file, then rerun aagen.
4234 ** Code for processing tables in the LEMON parser generator.
4237 PRIVATE
int strhash(const char *x
)
4240 while( *x
) h
= h
*13 + *(x
++);
4244 /* Works like strdup, sort of. Save a string in malloced memory, but
4245 ** keep strings in a table so that the same string is not in more
4248 const char *Strsafe(const char *y
)
4253 if( y
==0 ) return 0;
4254 z
= Strsafe_find(y
);
4255 if( z
==0 && (cpy
=(char *)malloc( lemonStrlen(y
)+1 ))!=0 ){
4264 /* There is one instance of the following structure for each
4265 ** associative array of type "x1".
4268 int size
; /* The number of available slots. */
4269 /* Must be a power of 2 greater than or */
4271 int count
; /* Number of currently slots filled */
4272 struct s_x1node
*tbl
; /* The data stored here */
4273 struct s_x1node
**ht
; /* Hash table for lookups */
4276 /* There is one instance of this structure for every data element
4277 ** in an associative array of type "x1".
4279 typedef struct s_x1node
{
4280 const char *data
; /* The data */
4281 struct s_x1node
*next
; /* Next entry with the same hash */
4282 struct s_x1node
**from
; /* Previous link */
4285 /* There is only one instance of the array, which is the following */
4286 static struct s_x1
*x1a
;
4288 /* Allocate a new associative array */
4289 void Strsafe_init(){
4291 x1a
= (struct s_x1
*)malloc( sizeof(struct s_x1
) );
4295 x1a
->tbl
= (x1node
*)malloc(
4296 (sizeof(x1node
) + sizeof(x1node
*))*1024 );
4302 x1a
->ht
= (x1node
**)&(x1a
->tbl
[1024]);
4303 for(i
=0; i
<1024; i
++) x1a
->ht
[i
] = 0;
4307 /* Insert a new record into the array. Return TRUE if successful.
4308 ** Prior data with the same key is NOT overwritten */
4309 int Strsafe_insert(const char *data
)
4315 if( x1a
==0 ) return 0;
4317 h
= ph
& (x1a
->size
-1);
4320 if( strcmp(np
->data
,data
)==0 ){
4321 /* An existing entry with the same key is found. */
4322 /* Fail because overwrite is not allows. */
4327 if( x1a
->count
>=x1a
->size
){
4328 /* Need to make the hash table bigger */
4331 array
.size
= size
= x1a
->size
*2;
4332 array
.count
= x1a
->count
;
4333 array
.tbl
= (x1node
*)malloc(
4334 (sizeof(x1node
) + sizeof(x1node
*))*size
);
4335 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4336 array
.ht
= (x1node
**)&(array
.tbl
[size
]);
4337 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4338 for(i
=0; i
<x1a
->count
; i
++){
4339 x1node
*oldnp
, *newnp
;
4340 oldnp
= &(x1a
->tbl
[i
]);
4341 h
= strhash(oldnp
->data
) & (size
-1);
4342 newnp
= &(array
.tbl
[i
]);
4343 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4344 newnp
->next
= array
.ht
[h
];
4345 newnp
->data
= oldnp
->data
;
4346 newnp
->from
= &(array
.ht
[h
]);
4347 array
.ht
[h
] = newnp
;
4352 /* Insert the new data */
4353 h
= ph
& (x1a
->size
-1);
4354 np
= &(x1a
->tbl
[x1a
->count
++]);
4356 if( x1a
->ht
[h
] ) x1a
->ht
[h
]->from
= &(np
->next
);
4357 np
->next
= x1a
->ht
[h
];
4359 np
->from
= &(x1a
->ht
[h
]);
4363 /* Return a pointer to data assigned to the given key. Return NULL
4364 ** if no such key. */
4365 const char *Strsafe_find(const char *key
)
4370 if( x1a
==0 ) return 0;
4371 h
= strhash(key
) & (x1a
->size
-1);
4374 if( strcmp(np
->data
,key
)==0 ) break;
4377 return np
? np
->data
: 0;
4380 /* Return a pointer to the (terminal or nonterminal) symbol "x".
4381 ** Create a new symbol if this is the first time "x" has been seen.
4383 struct symbol
*Symbol_new(const char *x
)
4387 sp
= Symbol_find(x
);
4389 sp
= (struct symbol
*)calloc(1, sizeof(struct symbol
) );
4391 sp
->name
= Strsafe(x
);
4392 sp
->type
= isupper(*x
) ? TERMINAL
: NONTERMINAL
;
4398 sp
->lambda
= LEMON_FALSE
;
4403 Symbol_insert(sp
,sp
->name
);
4409 /* Compare two symbols for working purposes
4411 ** Symbols that begin with upper case letters (terminals or tokens)
4412 ** must sort before symbols that begin with lower case letters
4413 ** (non-terminals). Other than that, the order does not matter.
4415 ** We find experimentally that leaving the symbols in their original
4416 ** order (the order they appeared in the grammar file) gives the
4417 ** smallest parser tables in SQLite.
4419 int Symbolcmpp(const void *_a
, const void *_b
)
4421 const struct symbol
**a
= (const struct symbol
**) _a
;
4422 const struct symbol
**b
= (const struct symbol
**) _b
;
4423 int i1
= (**a
).index
+ 10000000*((**a
).name
[0]>'Z');
4424 int i2
= (**b
).index
+ 10000000*((**b
).name
[0]>'Z');
4425 assert( i1
!=i2
|| strcmp((**a
).name
,(**b
).name
)==0 );
4429 /* There is one instance of the following structure for each
4430 ** associative array of type "x2".
4433 int size
; /* The number of available slots. */
4434 /* Must be a power of 2 greater than or */
4436 int count
; /* Number of currently slots filled */
4437 struct s_x2node
*tbl
; /* The data stored here */
4438 struct s_x2node
**ht
; /* Hash table for lookups */
4441 /* There is one instance of this structure for every data element
4442 ** in an associative array of type "x2".
4444 typedef struct s_x2node
{
4445 struct symbol
*data
; /* The data */
4446 const char *key
; /* The key */
4447 struct s_x2node
*next
; /* Next entry with the same hash */
4448 struct s_x2node
**from
; /* Previous link */
4451 /* There is only one instance of the array, which is the following */
4452 static struct s_x2
*x2a
;
4454 /* Allocate a new associative array */
4457 x2a
= (struct s_x2
*)malloc( sizeof(struct s_x2
) );
4461 x2a
->tbl
= (x2node
*)malloc(
4462 (sizeof(x2node
) + sizeof(x2node
*))*128 );
4468 x2a
->ht
= (x2node
**)&(x2a
->tbl
[128]);
4469 for(i
=0; i
<128; i
++) x2a
->ht
[i
] = 0;
4473 /* Insert a new record into the array. Return TRUE if successful.
4474 ** Prior data with the same key is NOT overwritten */
4475 int Symbol_insert(struct symbol
*data
, const char *key
)
4481 if( x2a
==0 ) return 0;
4483 h
= ph
& (x2a
->size
-1);
4486 if( strcmp(np
->key
,key
)==0 ){
4487 /* An existing entry with the same key is found. */
4488 /* Fail because overwrite is not allows. */
4493 if( x2a
->count
>=x2a
->size
){
4494 /* Need to make the hash table bigger */
4497 array
.size
= size
= x2a
->size
*2;
4498 array
.count
= x2a
->count
;
4499 array
.tbl
= (x2node
*)malloc(
4500 (sizeof(x2node
) + sizeof(x2node
*))*size
);
4501 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4502 array
.ht
= (x2node
**)&(array
.tbl
[size
]);
4503 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4504 for(i
=0; i
<x2a
->count
; i
++){
4505 x2node
*oldnp
, *newnp
;
4506 oldnp
= &(x2a
->tbl
[i
]);
4507 h
= strhash(oldnp
->key
) & (size
-1);
4508 newnp
= &(array
.tbl
[i
]);
4509 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4510 newnp
->next
= array
.ht
[h
];
4511 newnp
->key
= oldnp
->key
;
4512 newnp
->data
= oldnp
->data
;
4513 newnp
->from
= &(array
.ht
[h
]);
4514 array
.ht
[h
] = newnp
;
4519 /* Insert the new data */
4520 h
= ph
& (x2a
->size
-1);
4521 np
= &(x2a
->tbl
[x2a
->count
++]);
4524 if( x2a
->ht
[h
] ) x2a
->ht
[h
]->from
= &(np
->next
);
4525 np
->next
= x2a
->ht
[h
];
4527 np
->from
= &(x2a
->ht
[h
]);
4531 /* Return a pointer to data assigned to the given key. Return NULL
4532 ** if no such key. */
4533 struct symbol
*Symbol_find(const char *key
)
4538 if( x2a
==0 ) return 0;
4539 h
= strhash(key
) & (x2a
->size
-1);
4542 if( strcmp(np
->key
,key
)==0 ) break;
4545 return np
? np
->data
: 0;
4548 /* Return the n-th data. Return NULL if n is out of range. */
4549 struct symbol
*Symbol_Nth(int n
)
4551 struct symbol
*data
;
4552 if( x2a
&& n
>0 && n
<=x2a
->count
){
4553 data
= x2a
->tbl
[n
-1].data
;
4560 /* Return the size of the array */
4563 return x2a
? x2a
->count
: 0;
4566 /* Return an array of pointers to all data in the table.
4567 ** The array is obtained from malloc. Return NULL if memory allocation
4568 ** problems, or if the array is empty. */
4569 struct symbol
**Symbol_arrayof()
4571 struct symbol
**array
;
4573 if( x2a
==0 ) return 0;
4575 array
= (struct symbol
**)calloc(size
, sizeof(struct symbol
*));
4577 for(i
=0; i
<size
; i
++) array
[i
] = x2a
->tbl
[i
].data
;
4582 /* Compare two configurations */
4583 int Configcmp(const char *_a
,const char *_b
)
4585 const struct config
*a
= (struct config
*) _a
;
4586 const struct config
*b
= (struct config
*) _b
;
4588 x
= a
->rp
->index
- b
->rp
->index
;
4589 if( x
==0 ) x
= a
->dot
- b
->dot
;
4593 /* Compare two states */
4594 PRIVATE
int statecmp(struct config
*a
, struct config
*b
)
4597 for(rc
=0; rc
==0 && a
&& b
; a
=a
->bp
, b
=b
->bp
){
4598 rc
= a
->rp
->index
- b
->rp
->index
;
4599 if( rc
==0 ) rc
= a
->dot
- b
->dot
;
4609 PRIVATE
int statehash(struct config
*a
)
4613 h
= h
*571 + a
->rp
->index
*37 + a
->dot
;
4619 /* Allocate a new state structure */
4620 struct state
*State_new()
4622 struct state
*newstate
;
4623 newstate
= (struct state
*)calloc(1, sizeof(struct state
) );
4624 MemoryCheck(newstate
);
4628 /* There is one instance of the following structure for each
4629 ** associative array of type "x3".
4632 int size
; /* The number of available slots. */
4633 /* Must be a power of 2 greater than or */
4635 int count
; /* Number of currently slots filled */
4636 struct s_x3node
*tbl
; /* The data stored here */
4637 struct s_x3node
**ht
; /* Hash table for lookups */
4640 /* There is one instance of this structure for every data element
4641 ** in an associative array of type "x3".
4643 typedef struct s_x3node
{
4644 struct state
*data
; /* The data */
4645 struct config
*key
; /* The key */
4646 struct s_x3node
*next
; /* Next entry with the same hash */
4647 struct s_x3node
**from
; /* Previous link */
4650 /* There is only one instance of the array, which is the following */
4651 static struct s_x3
*x3a
;
4653 /* Allocate a new associative array */
4656 x3a
= (struct s_x3
*)malloc( sizeof(struct s_x3
) );
4660 x3a
->tbl
= (x3node
*)malloc(
4661 (sizeof(x3node
) + sizeof(x3node
*))*128 );
4667 x3a
->ht
= (x3node
**)&(x3a
->tbl
[128]);
4668 for(i
=0; i
<128; i
++) x3a
->ht
[i
] = 0;
4672 /* Insert a new record into the array. Return TRUE if successful.
4673 ** Prior data with the same key is NOT overwritten */
4674 int State_insert(struct state
*data
, struct config
*key
)
4680 if( x3a
==0 ) return 0;
4681 ph
= statehash(key
);
4682 h
= ph
& (x3a
->size
-1);
4685 if( statecmp(np
->key
,key
)==0 ){
4686 /* An existing entry with the same key is found. */
4687 /* Fail because overwrite is not allows. */
4692 if( x3a
->count
>=x3a
->size
){
4693 /* Need to make the hash table bigger */
4696 array
.size
= size
= x3a
->size
*2;
4697 array
.count
= x3a
->count
;
4698 array
.tbl
= (x3node
*)malloc(
4699 (sizeof(x3node
) + sizeof(x3node
*))*size
);
4700 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4701 array
.ht
= (x3node
**)&(array
.tbl
[size
]);
4702 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4703 for(i
=0; i
<x3a
->count
; i
++){
4704 x3node
*oldnp
, *newnp
;
4705 oldnp
= &(x3a
->tbl
[i
]);
4706 h
= statehash(oldnp
->key
) & (size
-1);
4707 newnp
= &(array
.tbl
[i
]);
4708 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4709 newnp
->next
= array
.ht
[h
];
4710 newnp
->key
= oldnp
->key
;
4711 newnp
->data
= oldnp
->data
;
4712 newnp
->from
= &(array
.ht
[h
]);
4713 array
.ht
[h
] = newnp
;
4718 /* Insert the new data */
4719 h
= ph
& (x3a
->size
-1);
4720 np
= &(x3a
->tbl
[x3a
->count
++]);
4723 if( x3a
->ht
[h
] ) x3a
->ht
[h
]->from
= &(np
->next
);
4724 np
->next
= x3a
->ht
[h
];
4726 np
->from
= &(x3a
->ht
[h
]);
4730 /* Return a pointer to data assigned to the given key. Return NULL
4731 ** if no such key. */
4732 struct state
*State_find(struct config
*key
)
4737 if( x3a
==0 ) return 0;
4738 h
= statehash(key
) & (x3a
->size
-1);
4741 if( statecmp(np
->key
,key
)==0 ) break;
4744 return np
? np
->data
: 0;
4747 /* Return an array of pointers to all data in the table.
4748 ** The array is obtained from malloc. Return NULL if memory allocation
4749 ** problems, or if the array is empty. */
4750 struct state
**State_arrayof()
4752 struct state
**array
;
4754 if( x3a
==0 ) return 0;
4756 array
= (struct state
**)malloc( sizeof(struct state
*)*size
);
4758 for(i
=0; i
<size
; i
++) array
[i
] = x3a
->tbl
[i
].data
;
4763 /* Hash a configuration */
4764 PRIVATE
int confighash(struct config
*a
)
4767 h
= h
*571 + a
->rp
->index
*37 + a
->dot
;
4771 /* There is one instance of the following structure for each
4772 ** associative array of type "x4".
4775 int size
; /* The number of available slots. */
4776 /* Must be a power of 2 greater than or */
4778 int count
; /* Number of currently slots filled */
4779 struct s_x4node
*tbl
; /* The data stored here */
4780 struct s_x4node
**ht
; /* Hash table for lookups */
4783 /* There is one instance of this structure for every data element
4784 ** in an associative array of type "x4".
4786 typedef struct s_x4node
{
4787 struct config
*data
; /* The data */
4788 struct s_x4node
*next
; /* Next entry with the same hash */
4789 struct s_x4node
**from
; /* Previous link */
4792 /* There is only one instance of the array, which is the following */
4793 static struct s_x4
*x4a
;
4795 /* Allocate a new associative array */
4796 void Configtable_init(){
4798 x4a
= (struct s_x4
*)malloc( sizeof(struct s_x4
) );
4802 x4a
->tbl
= (x4node
*)malloc(
4803 (sizeof(x4node
) + sizeof(x4node
*))*64 );
4809 x4a
->ht
= (x4node
**)&(x4a
->tbl
[64]);
4810 for(i
=0; i
<64; i
++) x4a
->ht
[i
] = 0;
4814 /* Insert a new record into the array. Return TRUE if successful.
4815 ** Prior data with the same key is NOT overwritten */
4816 int Configtable_insert(struct config
*data
)
4822 if( x4a
==0 ) return 0;
4823 ph
= confighash(data
);
4824 h
= ph
& (x4a
->size
-1);
4827 if( Configcmp((const char *) np
->data
,(const char *) data
)==0 ){
4828 /* An existing entry with the same key is found. */
4829 /* Fail because overwrite is not allows. */
4834 if( x4a
->count
>=x4a
->size
){
4835 /* Need to make the hash table bigger */
4838 array
.size
= size
= x4a
->size
*2;
4839 array
.count
= x4a
->count
;
4840 array
.tbl
= (x4node
*)malloc(
4841 (sizeof(x4node
) + sizeof(x4node
*))*size
);
4842 if( array
.tbl
==0 ) return 0; /* Fail due to malloc failure */
4843 array
.ht
= (x4node
**)&(array
.tbl
[size
]);
4844 for(i
=0; i
<size
; i
++) array
.ht
[i
] = 0;
4845 for(i
=0; i
<x4a
->count
; i
++){
4846 x4node
*oldnp
, *newnp
;
4847 oldnp
= &(x4a
->tbl
[i
]);
4848 h
= confighash(oldnp
->data
) & (size
-1);
4849 newnp
= &(array
.tbl
[i
]);
4850 if( array
.ht
[h
] ) array
.ht
[h
]->from
= &(newnp
->next
);
4851 newnp
->next
= array
.ht
[h
];
4852 newnp
->data
= oldnp
->data
;
4853 newnp
->from
= &(array
.ht
[h
]);
4854 array
.ht
[h
] = newnp
;
4859 /* Insert the new data */
4860 h
= ph
& (x4a
->size
-1);
4861 np
= &(x4a
->tbl
[x4a
->count
++]);
4863 if( x4a
->ht
[h
] ) x4a
->ht
[h
]->from
= &(np
->next
);
4864 np
->next
= x4a
->ht
[h
];
4866 np
->from
= &(x4a
->ht
[h
]);
4870 /* Return a pointer to data assigned to the given key. Return NULL
4871 ** if no such key. */
4872 struct config
*Configtable_find(struct config
*key
)
4877 if( x4a
==0 ) return 0;
4878 h
= confighash(key
) & (x4a
->size
-1);
4881 if( Configcmp((const char *) np
->data
,(const char *) key
)==0 ) break;
4884 return np
? np
->data
: 0;
4887 /* Remove all data from the table. Pass each data to the function "f"
4888 ** as it is removed. ("f" may be null to avoid this step.) */
4889 void Configtable_clear(int(*f
)(struct config
*))
4892 if( x4a
==0 || x4a
->count
==0 ) return;
4893 if( f
) for(i
=0; i
<x4a
->count
; i
++) (*f
)(x4a
->tbl
[i
].data
);
4894 for(i
=0; i
<x4a
->size
; i
++) x4a
->ht
[i
] = 0;