1 /*
2 ** This file contains all sources (including headers) to the LEMON
3 ** LALR(1) parser generator. The sources have been combined into a
4 ** single file to make it easy to include LEMON in the source tree
5 ** and Makefile of another program.
6 **
7 ** The author of this program disclaims copyright.
8 */
9 #include <stdio.h>
10 #include <stdarg.h>
11 #include <string.h>
12 #include <ctype.h>
13 #include <stdlib.h>
14 #include <assert.h>
15
16 #define ISSPACE(X) isspace((unsigned char)(X))
17 #define ISDIGIT(X) isdigit((unsigned char)(X))
18 #define ISALNUM(X) isalnum((unsigned char)(X))
19 #define ISALPHA(X) isalpha((unsigned char)(X))
20 #define ISUPPER(X) isupper((unsigned char)(X))
21 #define ISLOWER(X) islower((unsigned char)(X))
22
23
24 #ifndef __WIN32__
25 # if defined(_WIN32) || defined(WIN32)
26 # define __WIN32__
27 # endif
28 #endif
29
30 #ifdef __WIN32__
31 #ifdef __cplusplus
32 extern "C" {
33 #endif
34 extern int access(const char *path, int mode);
35 #ifdef __cplusplus
36 }
37 #endif
38 #else
39 #include <unistd.h>
40 #endif
41
42 /* #define PRIVATE static */
43 #define PRIVATE
44
45 #ifdef TEST
46 #define MAXRHS 5 /* Set low to exercise exception code */
47 #else
48 #define MAXRHS 1000
49 #endif
50
51 extern void memory_error();
52 static int showPrecedenceConflict = 0;
53 static char *msort(char*,char**,int(*)(const char*,const char*));
54
55 /*
56 ** Compilers are getting increasingly pedantic about type conversions
57 ** as C evolves ever closer to Ada.... To work around the latest problems
58 ** we have to define the following variant of strlen().
59 */
60 #define lemonStrlen(X) ((int)strlen(X))
61
62 /*
63 ** Compilers are starting to complain about the use of sprintf() and strcpy(),
64 ** saying they are unsafe. So we define our own versions of those routines too.
65 **
66 ** There are three routines here: lemon_sprintf(), lemon_vsprintf(), and
67 ** lemon_addtext(). The first two are replacements for sprintf() and vsprintf().
68 ** The third is a helper routine for vsnprintf() that adds texts to the end of a
69 ** buffer, making sure the buffer is always zero-terminated.
70 **
71 ** The string formatter is a minimal subset of stdlib sprintf() supporting only
72 ** a few simply conversions:
73 **
74 ** %d
75 ** %s
76 ** %.*s
77 **
78 */
lemon_addtext(char * zBuf,int * pnUsed,const char * zIn,int nIn,int iWidth)79 static void lemon_addtext(
80 char *zBuf, /* The buffer to which text is added */
81 int *pnUsed, /* Slots of the buffer used so far */
82 const char *zIn, /* Text to add */
83 int nIn, /* Bytes of text to add. -1 to use strlen() */
84 int iWidth /* Field width. Negative to left justify */
85 ){
86 if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){}
87 while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; }
88 if( nIn==0 ) return;
89 memcpy(&zBuf[*pnUsed], zIn, nIn);
90 *pnUsed += nIn;
91 while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; }
92 zBuf[*pnUsed] = 0;
93 }
lemon_vsprintf(char * str,const char * zFormat,va_list ap)94 static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){
95 int i, j, k, c;
96 int nUsed = 0;
97 const char *z;
98 char zTemp[50];
99 str[0] = 0;
100 for(i=j=0; (c = zFormat[i])!=0; i++){
101 if( c=='%' ){
102 int iWidth = 0;
103 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
104 c = zFormat[++i];
105 if( ISDIGIT(c) || (c=='-' && ISDIGIT(zFormat[i+1])) ){
106 if( c=='-' ) i++;
107 while( ISDIGIT(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0';
108 if( c=='-' ) iWidth = -iWidth;
109 c = zFormat[i];
110 }
111 if( c=='d' ){
112 int v = va_arg(ap, int);
113 if( v<0 ){
114 lemon_addtext(str, &nUsed, "-", 1, iWidth);
115 v = -v;
116 }else if( v==0 ){
117 lemon_addtext(str, &nUsed, "0", 1, iWidth);
118 }
119 k = 0;
120 while( v>0 ){
121 k++;
122 zTemp[sizeof(zTemp)-k] = (v%10) + '0';
123 v /= 10;
124 }
125 lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth);
126 }else if( c=='s' ){
127 z = va_arg(ap, const char*);
128 lemon_addtext(str, &nUsed, z, -1, iWidth);
129 }else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){
130 i += 2;
131 k = va_arg(ap, int);
132 z = va_arg(ap, const char*);
133 lemon_addtext(str, &nUsed, z, k, iWidth);
134 }else if( c=='%' ){
135 lemon_addtext(str, &nUsed, "%", 1, 0);
136 }else{
137 fprintf(stderr, "illegal format\n");
138 exit(1);
139 }
140 j = i+1;
141 }
142 }
143 lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
144 return nUsed;
145 }
lemon_sprintf(char * str,const char * format,...)146 static int lemon_sprintf(char *str, const char *format, ...){
147 va_list ap;
148 int rc;
149 va_start(ap, format);
150 rc = lemon_vsprintf(str, format, ap);
151 va_end(ap);
152 return rc;
153 }
lemon_strcpy(char * dest,const char * src)154 static void lemon_strcpy(char *dest, const char *src){
155 while( (*(dest++) = *(src++))!=0 ){}
156 }
lemon_strcat(char * dest,const char * src)157 static void lemon_strcat(char *dest, const char *src){
158 while( *dest ) dest++;
159 lemon_strcpy(dest, src);
160 }
161
162
163 /* a few forward declarations... */
164 struct rule;
165 struct lemon;
166 struct action;
167
168 static struct action *Action_new(void);
169 static struct action *Action_sort(struct action *);
170
171 /********** From the file "build.h" ************************************/
172 void FindRulePrecedences(struct lemon*);
173 void FindFirstSets(struct lemon*);
174 void FindStates(struct lemon*);
175 void FindLinks(struct lemon*);
176 void FindFollowSets(struct lemon*);
177 void FindActions(struct lemon*);
178
179 /********* From the file "configlist.h" *********************************/
180 void Configlist_init(void);
181 struct config *Configlist_add(struct rule *, int);
182 struct config *Configlist_addbasis(struct rule *, int);
183 void Configlist_closure(struct lemon *);
184 void Configlist_sort(void);
185 void Configlist_sortbasis(void);
186 struct config *Configlist_return(void);
187 struct config *Configlist_basis(void);
188 void Configlist_eat(struct config *);
189 void Configlist_reset(void);
190
191 /********* From the file "error.h" ***************************************/
192 void ErrorMsg(const char *, int,const char *, ...);
193
194 /****** From the file "option.h" ******************************************/
195 enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
196 OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR};
197 struct s_options {
198 enum option_type type;
199 const char *label;
200 char *arg;
201 void(*fn)();
202 const char *message;
203 };
204 int OptInit(char**,struct s_options*,FILE*);
205 int OptNArgs(void);
206 char *OptArg(int);
207 void OptErr(int);
208 void OptPrint(void);
209
210 /******** From the file "parse.h" *****************************************/
211 void Parse(struct lemon *lemp);
212
213 /********* From the file "plink.h" ***************************************/
214 struct plink *Plink_new(void);
215 void Plink_add(struct plink **, struct config *);
216 void Plink_copy(struct plink **, struct plink *);
217 void Plink_delete(struct plink *);
218
219 /********** From the file "report.h" *************************************/
220 void Reprint(struct lemon *);
221 void ReportOutput(struct lemon *);
222 void ReportTable(struct lemon *, int, int);
223 void ReportHeader(struct lemon *);
224 void CompressTables(struct lemon *);
225 void ResortStates(struct lemon *);
226
227 /********** From the file "set.h" ****************************************/
228 void SetSize(int); /* All sets will be of size N */
229 char *SetNew(void); /* A new set for element 0..N */
230 void SetFree(char*); /* Deallocate a set */
231 int SetAdd(char*,int); /* Add element to a set */
232 int SetUnion(char *,char *); /* A <- A U B, thru element N */
233 #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
234
235 /********** From the file "struct.h" *************************************/
236 /*
237 ** Principal data structures for the LEMON parser generator.
238 */
239
240 typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
241
242 /* Symbols (terminals and nonterminals) of the grammar are stored
243 ** in the following: */
244 enum symbol_type {
245 TERMINAL,
246 NONTERMINAL,
247 MULTITERMINAL
248 };
249 enum e_assoc {
250 LEFT,
251 RIGHT,
252 NONE,
253 UNK
254 };
255 struct symbol {
256 const char *name; /* Name of the symbol */
257 int index; /* Index number for this symbol */
258 enum symbol_type type; /* Symbols are all either TERMINALS or NTs */
259 struct rule *rule; /* Linked list of rules of this (if an NT) */
260 struct symbol *fallback; /* fallback token in case this token doesn't parse */
261 int prec; /* Precedence if defined (-1 otherwise) */
262 enum e_assoc assoc; /* Associativity if precedence is defined */
263 char *firstset; /* First-set for all rules of this symbol */
264 Boolean lambda; /* True if NT and can generate an empty string */
265 int useCnt; /* Number of times used */
266 char *destructor; /* Code which executes whenever this symbol is
267 ** popped from the stack during error processing */
268 int destLineno; /* Line number for start of destructor. Set to
269 ** -1 for duplicate destructors. */
270 char *datatype; /* The data type of information held by this
271 ** object. Only used if type==NONTERMINAL */
272 int dtnum; /* The data type number. In the parser, the value
273 ** stack is a union. The .yy%d element of this
274 ** union is the correct data type for this object */
275 int bContent; /* True if this symbol ever carries content - if
276 ** it is ever more than just syntax */
277 /* The following fields are used by MULTITERMINALs only */
278 int nsubsym; /* Number of constituent symbols in the MULTI */
279 struct symbol **subsym; /* Array of constituent symbols */
280 };
281
282 /* Each production rule in the grammar is stored in the following
283 ** structure. */
284 struct rule {
285 struct symbol *lhs; /* Left-hand side of the rule */
286 const char *lhsalias; /* Alias for the LHS (NULL if none) */
287 int lhsStart; /* True if left-hand side is the start symbol */
288 int ruleline; /* Line number for the rule */
289 int nrhs; /* Number of RHS symbols */
290 struct symbol **rhs; /* The RHS symbols */
291 const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
292 int line; /* Line number at which code begins */
293 const char *code; /* The code executed when this rule is reduced */
294 const char *codePrefix; /* Setup code before code[] above */
295 const char *codeSuffix; /* Breakdown code after code[] above */
296 struct symbol *precsym; /* Precedence symbol for this rule */
297 int index; /* An index number for this rule */
298 int iRule; /* Rule number as used in the generated tables */
299 Boolean noCode; /* True if this rule has no associated C code */
300 Boolean codeEmitted; /* True if the code has been emitted already */
301 Boolean canReduce; /* True if this rule is ever reduced */
302 Boolean doesReduce; /* Reduce actions occur after optimization */
303 Boolean neverReduce; /* Reduce is theoretically possible, but prevented
304 ** by actions or other outside implementation */
305 struct rule *nextlhs; /* Next rule with the same LHS */
306 struct rule *next; /* Next rule in the global list */
307 };
308
309 /* A configuration is a production rule of the grammar together with
310 ** a mark (dot) showing how much of that rule has been processed so far.
311 ** Configurations also contain a follow-set which is a list of terminal
312 ** symbols which are allowed to immediately follow the end of the rule.
313 ** Every configuration is recorded as an instance of the following: */
314 enum cfgstatus {
315 COMPLETE,
316 INCOMPLETE
317 };
318 struct config {
319 struct rule *rp; /* The rule upon which the configuration is based */
320 int dot; /* The parse point */
321 char *fws; /* Follow-set for this configuration only */
322 struct plink *fplp; /* Follow-set forward propagation links */
323 struct plink *bplp; /* Follow-set backwards propagation links */
324 struct state *stp; /* Pointer to state which contains this */
325 enum cfgstatus status; /* used during followset and shift computations */
326 struct config *next; /* Next configuration in the state */
327 struct config *bp; /* The next basis configuration */
328 };
329
330 enum e_action {
331 SHIFT,
332 ACCEPT,
333 REDUCE,
334 ERROR,
335 SSCONFLICT, /* A shift/shift conflict */
336 SRCONFLICT, /* Was a reduce, but part of a conflict */
337 RRCONFLICT, /* Was a reduce, but part of a conflict */
338 SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
339 RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
340 NOT_USED, /* Deleted by compression */
341 SHIFTREDUCE /* Shift first, then reduce */
342 };
343
344 /* Every shift or reduce operation is stored as one of the following */
345 struct action {
346 struct symbol *sp; /* The look-ahead symbol */
347 enum e_action type;
348 union {
349 struct state *stp; /* The new state, if a shift */
350 struct rule *rp; /* The rule, if a reduce */
351 } x;
352 struct symbol *spOpt; /* SHIFTREDUCE optimization to this symbol */
353 struct action *next; /* Next action for this state */
354 struct action *collide; /* Next action with the same hash */
355 };
356
357 /* Each state of the generated parser's finite state machine
358 ** is encoded as an instance of the following structure. */
359 struct state {
360 struct config *bp; /* The basis configurations for this state */
361 struct config *cfp; /* All configurations in this set */
362 int statenum; /* Sequential number for this state */
363 struct action *ap; /* List of actions for this state */
364 int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
365 int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
366 int iDfltReduce; /* Default action is to REDUCE by this rule */
367 struct rule *pDfltReduce;/* The default REDUCE rule. */
368 int autoReduce; /* True if this is an auto-reduce state */
369 };
370 #define NO_OFFSET (-2147483647)
371
372 /* A followset propagation link indicates that the contents of one
373 ** configuration followset should be propagated to another whenever
374 ** the first changes. */
375 struct plink {
376 struct config *cfp; /* The configuration to which linked */
377 struct plink *next; /* The next propagate link */
378 };
379
380 /* The state vector for the entire parser generator is recorded as
381 ** follows. (LEMON uses no global variables and makes little use of
382 ** static variables. Fields in the following structure can be thought
383 ** of as begin global variables in the program.) */
384 struct lemon {
385 struct state **sorted; /* Table of states sorted by state number */
386 struct rule *rule; /* List of all rules */
387 struct rule *startRule; /* First rule */
388 int nstate; /* Number of states */
389 int nxstate; /* nstate with tail degenerate states removed */
390 int nrule; /* Number of rules */
391 int nruleWithAction; /* Number of rules with actions */
392 int nsymbol; /* Number of terminal and nonterminal symbols */
393 int nterminal; /* Number of terminal symbols */
394 int minShiftReduce; /* Minimum shift-reduce action value */
395 int errAction; /* Error action value */
396 int accAction; /* Accept action value */
397 int noAction; /* No-op action value */
398 int minReduce; /* Minimum reduce action */
399 int maxAction; /* Maximum action value of any kind */
400 struct symbol **symbols; /* Sorted array of pointers to symbols */
401 int errorcnt; /* Number of errors */
402 struct symbol *errsym; /* The error symbol */
403 struct symbol *wildcard; /* Token that matches anything */
404 char *name; /* Name of the generated parser */
405 char *arg; /* Declaration of the 3rd argument to parser */
406 char *ctx; /* Declaration of 2nd argument to constructor */
407 char *tokentype; /* Type of terminal symbols in the parser stack */
408 char *vartype; /* The default type of non-terminal symbols */
409 char *start; /* Name of the start symbol for the grammar */
410 char *stacksize; /* Size of the parser stack */
411 char *include; /* Code to put at the start of the C file */
412 char *error; /* Code to execute when an error is seen */
413 char *overflow; /* Code to execute on a stack overflow */
414 char *failure; /* Code to execute on parser failure */
415 char *accept; /* Code to execute when the parser excepts */
416 char *extracode; /* Code appended to the generated file */
417 char *tokendest; /* Code to execute to destroy token data */
418 char *vardest; /* Code for the default non-terminal destructor */
419 char *filename; /* Name of the input file */
420 char *outname; /* Name of the current output file */
421 char *tokenprefix; /* A prefix added to token names in the .h file */
422 int nconflict; /* Number of parsing conflicts */
423 int nactiontab; /* Number of entries in the yy_action[] table */
424 int nlookaheadtab; /* Number of entries in yy_lookahead[] */
425 int tablesize; /* Total table size of all tables in bytes */
426 int basisflag; /* Print only basis configurations */
427 int printPreprocessed; /* Show preprocessor output on stdout */
428 int has_fallback; /* True if any %fallback is seen in the grammar */
429 int nolinenosflag; /* True if #line statements should not be printed */
430 char *argv0; /* Name of the program */
431 };
432
433 #define MemoryCheck(X) if((X)==0){ \
434 extern void memory_error(); \
435 memory_error(); \
436 }
437
438 /**************** From the file "table.h" *********************************/
439 /*
440 ** All code in this file has been automatically generated
441 ** from a specification in the file
442 ** "table.q"
443 ** by the associative array code building program "aagen".
444 ** Do not edit this file! Instead, edit the specification
445 ** file, then rerun aagen.
446 */
447 /*
448 ** Code for processing tables in the LEMON parser generator.
449 */
450 /* Routines for handling a strings */
451
452 const char *Strsafe(const char *);
453
454 void Strsafe_init(void);
455 int Strsafe_insert(const char *);
456 const char *Strsafe_find(const char *);
457
458 /* Routines for handling symbols of the grammar */
459
460 struct symbol *Symbol_new(const char *);
461 int Symbolcmpp(const void *, const void *);
462 void Symbol_init(void);
463 int Symbol_insert(struct symbol *, const char *);
464 struct symbol *Symbol_find(const char *);
465 struct symbol *Symbol_Nth(int);
466 int Symbol_count(void);
467 struct symbol **Symbol_arrayof(void);
468
469 /* Routines to manage the state table */
470
471 int Configcmp(const char *, const char *);
472 struct state *State_new(void);
473 void State_init(void);
474 int State_insert(struct state *, struct config *);
475 struct state *State_find(struct config *);
476 struct state **State_arrayof(void);
477
478 /* Routines used for efficiency in Configlist_add */
479
480 void Configtable_init(void);
481 int Configtable_insert(struct config *);
482 struct config *Configtable_find(struct config *);
483 void Configtable_clear(int(*)(struct config *));
484
485 /****************** From the file "action.c" *******************************/
486 /*
487 ** Routines processing parser actions in the LEMON parser generator.
488 */
489
490 /* Allocate a new parser action */
Action_new(void)491 static struct action *Action_new(void){
492 static struct action *actionfreelist = 0;
493 struct action *newaction;
494
495 if( actionfreelist==0 ){
496 int i;
497 int amt = 100;
498 actionfreelist = (struct action *)calloc(amt, sizeof(struct action));
499 if( actionfreelist==0 ){
500 fprintf(stderr,"Unable to allocate memory for a new parser action.");
501 exit(1);
502 }
503 for(i=0; i<amt-1; i++) actionfreelist[i].next = &actionfreelist[i+1];
504 actionfreelist[amt-1].next = 0;
505 }
506 newaction = actionfreelist;
507 actionfreelist = actionfreelist->next;
508 return newaction;
509 }
510
511 /* Compare two actions for sorting purposes. Return negative, zero, or
512 ** positive if the first action is less than, equal to, or greater than
513 ** the first
514 */
actioncmp(struct action * ap1,struct action * ap2)515 static int actioncmp(
516 struct action *ap1,
517 struct action *ap2
518 ){
519 int rc;
520 rc = ap1->sp->index - ap2->sp->index;
521 if( rc==0 ){
522 rc = (int)ap1->type - (int)ap2->type;
523 }
524 if( rc==0 && (ap1->type==REDUCE || ap1->type==SHIFTREDUCE) ){
525 rc = ap1->x.rp->index - ap2->x.rp->index;
526 }
527 if( rc==0 ){
528 rc = (int) (ap2 - ap1);
529 }
530 return rc;
531 }
532
533 /* Sort parser actions */
Action_sort(struct action * ap)534 static struct action *Action_sort(
535 struct action *ap
536 ){
537 ap = (struct action *)msort((char *)ap,(char **)&ap->next,
538 (int(*)(const char*,const char*))actioncmp);
539 return ap;
540 }
541
Action_add(struct action ** app,enum e_action type,struct symbol * sp,char * arg)542 void Action_add(
543 struct action **app,
544 enum e_action type,
545 struct symbol *sp,
546 char *arg
547 ){
548 struct action *newaction;
549 newaction = Action_new();
550 newaction->next = *app;
551 *app = newaction;
552 newaction->type = type;
553 newaction->sp = sp;
554 newaction->spOpt = 0;
555 if( type==SHIFT ){
556 newaction->x.stp = (struct state *)arg;
557 }else{
558 newaction->x.rp = (struct rule *)arg;
559 }
560 }
561 /********************** New code to implement the "acttab" module ***********/
562 /*
563 ** This module implements routines use to construct the yy_action[] table.
564 */
565
566 /*
567 ** The state of the yy_action table under construction is an instance of
568 ** the following structure.
569 **
570 ** The yy_action table maps the pair (state_number, lookahead) into an
571 ** action_number. The table is an array of integers pairs. The state_number
572 ** determines an initial offset into the yy_action array. The lookahead
573 ** value is then added to this initial offset to get an index X into the
574 ** yy_action array. If the aAction[X].lookahead equals the value of the
575 ** of the lookahead input, then the value of the action_number output is
576 ** aAction[X].action. If the lookaheads do not match then the
577 ** default action for the state_number is returned.
578 **
579 ** All actions associated with a single state_number are first entered
580 ** into aLookahead[] using multiple calls to acttab_action(). Then the
581 ** actions for that single state_number are placed into the aAction[]
582 ** array with a single call to acttab_insert(). The acttab_insert() call
583 ** also resets the aLookahead[] array in preparation for the next
584 ** state number.
585 */
586 struct lookahead_action {
587 int lookahead; /* Value of the lookahead token */
588 int action; /* Action to take on the given lookahead */
589 };
590 typedef struct acttab acttab;
591 struct acttab {
592 int nAction; /* Number of used slots in aAction[] */
593 int nActionAlloc; /* Slots allocated for aAction[] */
594 struct lookahead_action
595 *aAction, /* The yy_action[] table under construction */
596 *aLookahead; /* A single new transaction set */
597 int mnLookahead; /* Minimum aLookahead[].lookahead */
598 int mnAction; /* Action associated with mnLookahead */
599 int mxLookahead; /* Maximum aLookahead[].lookahead */
600 int nLookahead; /* Used slots in aLookahead[] */
601 int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
602 int nterminal; /* Number of terminal symbols */
603 int nsymbol; /* total number of symbols */
604 };
605
606 /* Return the number of entries in the yy_action table */
607 #define acttab_lookahead_size(X) ((X)->nAction)
608
609 /* The value for the N-th entry in yy_action */
610 #define acttab_yyaction(X,N) ((X)->aAction[N].action)
611
612 /* The value for the N-th entry in yy_lookahead */
613 #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
614
615 /* Free all memory associated with the given acttab */
acttab_free(acttab * p)616 void acttab_free(acttab *p){
617 free( p->aAction );
618 free( p->aLookahead );
619 free( p );
620 }
621
622 /* Allocate a new acttab structure */
acttab_alloc(int nsymbol,int nterminal)623 acttab *acttab_alloc(int nsymbol, int nterminal){
624 acttab *p = (acttab *) calloc( 1, sizeof(*p) );
625 if( p==0 ){
626 fprintf(stderr,"Unable to allocate memory for a new acttab.");
627 exit(1);
628 }
629 memset(p, 0, sizeof(*p));
630 p->nsymbol = nsymbol;
631 p->nterminal = nterminal;
632 return p;
633 }
634
635 /* Add a new action to the current transaction set.
636 **
637 ** This routine is called once for each lookahead for a particular
638 ** state.
639 */
acttab_action(acttab * p,int lookahead,int action)640 void acttab_action(acttab *p, int lookahead, int action){
641 if( p->nLookahead>=p->nLookaheadAlloc ){
642 p->nLookaheadAlloc += 25;
643 p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead,
644 sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
645 if( p->aLookahead==0 ){
646 fprintf(stderr,"malloc failed\n");
647 exit(1);
648 }
649 }
650 if( p->nLookahead==0 ){
651 p->mxLookahead = lookahead;
652 p->mnLookahead = lookahead;
653 p->mnAction = action;
654 }else{
655 if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
656 if( p->mnLookahead>lookahead ){
657 p->mnLookahead = lookahead;
658 p->mnAction = action;
659 }
660 }
661 p->aLookahead[p->nLookahead].lookahead = lookahead;
662 p->aLookahead[p->nLookahead].action = action;
663 p->nLookahead++;
664 }
665
666 /*
667 ** Add the transaction set built up with prior calls to acttab_action()
668 ** into the current action table. Then reset the transaction set back
669 ** to an empty set in preparation for a new round of acttab_action() calls.
670 **
671 ** Return the offset into the action table of the new transaction.
672 **
673 ** If the makeItSafe parameter is true, then the offset is chosen so that
674 ** it is impossible to overread the yy_lookaside[] table regardless of
675 ** the lookaside token. This is done for the terminal symbols, as they
676 ** come from external inputs and can contain syntax errors. When makeItSafe
677 ** is false, there is more flexibility in selecting offsets, resulting in
678 ** a smaller table. For non-terminal symbols, which are never syntax errors,
679 ** makeItSafe can be false.
680 */
acttab_insert(acttab * p,int makeItSafe)681 int acttab_insert(acttab *p, int makeItSafe){
682 int i, j, k, n, end;
683 assert( p->nLookahead>0 );
684
685 /* Make sure we have enough space to hold the expanded action table
686 ** in the worst case. The worst case occurs if the transaction set
687 ** must be appended to the current action table
688 */
689 n = p->nsymbol + 1;
690 if( p->nAction + n >= p->nActionAlloc ){
691 int oldAlloc = p->nActionAlloc;
692 p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
693 p->aAction = (struct lookahead_action *) realloc( p->aAction,
694 sizeof(p->aAction[0])*p->nActionAlloc);
695 if( p->aAction==0 ){
696 fprintf(stderr,"malloc failed\n");
697 exit(1);
698 }
699 for(i=oldAlloc; i<p->nActionAlloc; i++){
700 p->aAction[i].lookahead = -1;
701 p->aAction[i].action = -1;
702 }
703 }
704
705 /* Scan the existing action table looking for an offset that is a
706 ** duplicate of the current transaction set. Fall out of the loop
707 ** if and when the duplicate is found.
708 **
709 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
710 */
711 end = makeItSafe ? p->mnLookahead : 0;
712 for(i=p->nAction-1; i>=end; i--){
713 if( p->aAction[i].lookahead==p->mnLookahead ){
714 /* All lookaheads and actions in the aLookahead[] transaction
715 ** must match against the candidate aAction[i] entry. */
716 if( p->aAction[i].action!=p->mnAction ) continue;
717 for(j=0; j<p->nLookahead; j++){
718 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
719 if( k<0 || k>=p->nAction ) break;
720 if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
721 if( p->aLookahead[j].action!=p->aAction[k].action ) break;
722 }
723 if( j<p->nLookahead ) continue;
724
725 /* No possible lookahead value that is not in the aLookahead[]
726 ** transaction is allowed to match aAction[i] */
727 n = 0;
728 for(j=0; j<p->nAction; j++){
729 if( p->aAction[j].lookahead<0 ) continue;
730 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
731 }
732 if( n==p->nLookahead ){
733 break; /* An exact match is found at offset i */
734 }
735 }
736 }
737
738 /* If no existing offsets exactly match the current transaction, find an
739 ** an empty offset in the aAction[] table in which we can add the
740 ** aLookahead[] transaction.
741 */
742 if( i<end ){
743 /* Look for holes in the aAction[] table that fit the current
744 ** aLookahead[] transaction. Leave i set to the offset of the hole.
745 ** If no holes are found, i is left at p->nAction, which means the
746 ** transaction will be appended. */
747 i = makeItSafe ? p->mnLookahead : 0;
748 for(; i<p->nActionAlloc - p->mxLookahead; i++){
749 if( p->aAction[i].lookahead<0 ){
750 for(j=0; j<p->nLookahead; j++){
751 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
752 if( k<0 ) break;
753 if( p->aAction[k].lookahead>=0 ) break;
754 }
755 if( j<p->nLookahead ) continue;
756 for(j=0; j<p->nAction; j++){
757 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
758 }
759 if( j==p->nAction ){
760 break; /* Fits in empty slots */
761 }
762 }
763 }
764 }
765 /* Insert transaction set at index i. */
766 #if 0
767 printf("Acttab:");
768 for(j=0; j<p->nLookahead; j++){
769 printf(" %d", p->aLookahead[j].lookahead);
770 }
771 printf(" inserted at %d\n", i);
772 #endif
773 for(j=0; j<p->nLookahead; j++){
774 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
775 p->aAction[k] = p->aLookahead[j];
776 if( k>=p->nAction ) p->nAction = k+1;
777 }
778 if( makeItSafe && i+p->nterminal>=p->nAction ) p->nAction = i+p->nterminal+1;
779 p->nLookahead = 0;
780
781 /* Return the offset that is added to the lookahead in order to get the
782 ** index into yy_action of the action */
783 return i - p->mnLookahead;
784 }
785
786 /*
787 ** Return the size of the action table without the trailing syntax error
788 ** entries.
789 */
acttab_action_size(acttab * p)790 int acttab_action_size(acttab *p){
791 int n = p->nAction;
792 while( n>0 && p->aAction[n-1].lookahead<0 ){ n--; }
793 return n;
794 }
795
796 /********************** From the file "build.c" *****************************/
797 /*
798 ** Routines to construction the finite state machine for the LEMON
799 ** parser generator.
800 */
801
802 /* Find a precedence symbol of every rule in the grammar.
803 **
804 ** Those rules which have a precedence symbol coded in the input
805 ** grammar using the "[symbol]" construct will already have the
806 ** rp->precsym field filled. Other rules take as their precedence
807 ** symbol the first RHS symbol with a defined precedence. If there
808 ** are not RHS symbols with a defined precedence, the precedence
809 ** symbol field is left blank.
810 */
FindRulePrecedences(struct lemon * xp)811 void FindRulePrecedences(struct lemon *xp)
812 {
813 struct rule *rp;
814 for(rp=xp->rule; rp; rp=rp->next){
815 if( rp->precsym==0 ){
816 int i, j;
817 for(i=0; i<rp->nrhs && rp->precsym==0; i++){
818 struct symbol *sp = rp->rhs[i];
819 if( sp->type==MULTITERMINAL ){
820 for(j=0; j<sp->nsubsym; j++){
821 if( sp->subsym[j]->prec>=0 ){
822 rp->precsym = sp->subsym[j];
823 break;
824 }
825 }
826 }else if( sp->prec>=0 ){
827 rp->precsym = rp->rhs[i];
828 }
829 }
830 }
831 }
832 return;
833 }
834
835 /* Find all nonterminals which will generate the empty string.
836 ** Then go back and compute the first sets of every nonterminal.
837 ** The first set is the set of all terminal symbols which can begin
838 ** a string generated by that nonterminal.
839 */
FindFirstSets(struct lemon * lemp)840 void FindFirstSets(struct lemon *lemp)
841 {
842 int i, j;
843 struct rule *rp;
844 int progress;
845
846 for(i=0; i<lemp->nsymbol; i++){
847 lemp->symbols[i]->lambda = LEMON_FALSE;
848 }
849 for(i=lemp->nterminal; i<lemp->nsymbol; i++){
850 lemp->symbols[i]->firstset = SetNew();
851 }
852
853 /* First compute all lambdas */
854 do{
855 progress = 0;
856 for(rp=lemp->rule; rp; rp=rp->next){
857 if( rp->lhs->lambda ) continue;
858 for(i=0; i<rp->nrhs; i++){
859 struct symbol *sp = rp->rhs[i];
860 assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE );
861 if( sp->lambda==LEMON_FALSE ) break;
862 }
863 if( i==rp->nrhs ){
864 rp->lhs->lambda = LEMON_TRUE;
865 progress = 1;
866 }
867 }
868 }while( progress );
869
870 /* Now compute all first sets */
871 do{
872 struct symbol *s1, *s2;
873 progress = 0;
874 for(rp=lemp->rule; rp; rp=rp->next){
875 s1 = rp->lhs;
876 for(i=0; i<rp->nrhs; i++){
877 s2 = rp->rhs[i];
878 if( s2->type==TERMINAL ){
879 progress += SetAdd(s1->firstset,s2->index);
880 break;
881 }else if( s2->type==MULTITERMINAL ){
882 for(j=0; j<s2->nsubsym; j++){
883 progress += SetAdd(s1->firstset,s2->subsym[j]->index);
884 }
885 break;
886 }else if( s1==s2 ){
887 if( s1->lambda==LEMON_FALSE ) break;
888 }else{
889 progress += SetUnion(s1->firstset,s2->firstset);
890 if( s2->lambda==LEMON_FALSE ) break;
891 }
892 }
893 }
894 }while( progress );
895 return;
896 }
897
898 /* Compute all LR(0) states for the grammar. Links
899 ** are added to between some states so that the LR(1) follow sets
900 ** can be computed later.
901 */
902 PRIVATE struct state *getstate(struct lemon *); /* forward reference */
FindStates(struct lemon * lemp)903 void FindStates(struct lemon *lemp)
904 {
905 struct symbol *sp;
906 struct rule *rp;
907
908 Configlist_init();
909
910 /* Find the start symbol */
911 if( lemp->start ){
912 sp = Symbol_find(lemp->start);
913 if( sp==0 ){
914 ErrorMsg(lemp->filename,0,
915 "The specified start symbol \"%s\" is not "
916 "in a nonterminal of the grammar. \"%s\" will be used as the start "
917 "symbol instead.",lemp->start,lemp->startRule->lhs->name);
918 lemp->errorcnt++;
919 sp = lemp->startRule->lhs;
920 }
921 }else if( lemp->startRule ){
922 sp = lemp->startRule->lhs;
923 }else{
924 ErrorMsg(lemp->filename,0,"Internal error - no start rule\n");
925 exit(1);
926 }
927
928 /* Make sure the start symbol doesn't occur on the right-hand side of
929 ** any rule. Report an error if it does. (YACC would generate a new
930 ** start symbol in this case.) */
931 for(rp=lemp->rule; rp; rp=rp->next){
932 int i;
933 for(i=0; i<rp->nrhs; i++){
934 if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
935 ErrorMsg(lemp->filename,0,
936 "The start symbol \"%s\" occurs on the "
937 "right-hand side of a rule. This will result in a parser which "
938 "does not work properly.",sp->name);
939 lemp->errorcnt++;
940 }
941 }
942 }
943
944 /* The basis configuration set for the first state
945 ** is all rules which have the start symbol as their
946 ** left-hand side */
947 for(rp=sp->rule; rp; rp=rp->nextlhs){
948 struct config *newcfp;
949 rp->lhsStart = 1;
950 newcfp = Configlist_addbasis(rp,0);
951 SetAdd(newcfp->fws,0);
952 }
953
954 /* Compute the first state. All other states will be
955 ** computed automatically during the computation of the first one.
956 ** The returned pointer to the first state is not used. */
957 (void)getstate(lemp);
958 return;
959 }
960
961 /* Return a pointer to a state which is described by the configuration
962 ** list which has been built from calls to Configlist_add.
963 */
964 PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */
getstate(struct lemon * lemp)965 PRIVATE struct state *getstate(struct lemon *lemp)
966 {
967 struct config *cfp, *bp;
968 struct state *stp;
969
970 /* Extract the sorted basis of the new state. The basis was constructed
971 ** by prior calls to "Configlist_addbasis()". */
972 Configlist_sortbasis();
973 bp = Configlist_basis();
974
975 /* Get a state with the same basis */
976 stp = State_find(bp);
977 if( stp ){
978 /* A state with the same basis already exists! Copy all the follow-set
979 ** propagation links from the state under construction into the
980 ** preexisting state, then return a pointer to the preexisting state */
981 struct config *x, *y;
982 for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
983 Plink_copy(&y->bplp,x->bplp);
984 Plink_delete(x->fplp);
985 x->fplp = x->bplp = 0;
986 }
987 cfp = Configlist_return();
988 Configlist_eat(cfp);
989 }else{
990 /* This really is a new state. Construct all the details */
991 Configlist_closure(lemp); /* Compute the configuration closure */
992 Configlist_sort(); /* Sort the configuration closure */
993 cfp = Configlist_return(); /* Get a pointer to the config list */
994 stp = State_new(); /* A new state structure */
995 MemoryCheck(stp);
996 stp->bp = bp; /* Remember the configuration basis */
997 stp->cfp = cfp; /* Remember the configuration closure */
998 stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
999 stp->ap = 0; /* No actions, yet. */
1000 State_insert(stp,stp->bp); /* Add to the state table */
1001 buildshifts(lemp,stp); /* Recursively compute successor states */
1002 }
1003 return stp;
1004 }
1005
1006 /*
1007 ** Return true if two symbols are the same.
1008 */
same_symbol(struct symbol * a,struct symbol * b)1009 int same_symbol(struct symbol *a, struct symbol *b)
1010 {
1011 int i;
1012 if( a==b ) return 1;
1013 if( a->type!=MULTITERMINAL ) return 0;
1014 if( b->type!=MULTITERMINAL ) return 0;
1015 if( a->nsubsym!=b->nsubsym ) return 0;
1016 for(i=0; i<a->nsubsym; i++){
1017 if( a->subsym[i]!=b->subsym[i] ) return 0;
1018 }
1019 return 1;
1020 }
1021
1022 /* Construct all successor states to the given state. A "successor"
1023 ** state is any state which can be reached by a shift action.
1024 */
buildshifts(struct lemon * lemp,struct state * stp)1025 PRIVATE void buildshifts(struct lemon *lemp, struct state *stp)
1026 {
1027 struct config *cfp; /* For looping thru the config closure of "stp" */
1028 struct config *bcfp; /* For the inner loop on config closure of "stp" */
1029 struct config *newcfg; /* */
1030 struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
1031 struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
1032 struct state *newstp; /* A pointer to a successor state */
1033
1034 /* Each configuration becomes complete after it contributes to a successor
1035 ** state. Initially, all configurations are incomplete */
1036 for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
1037
1038 /* Loop through all configurations of the state "stp" */
1039 for(cfp=stp->cfp; cfp; cfp=cfp->next){
1040 if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
1041 if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
1042 Configlist_reset(); /* Reset the new config set */
1043 sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
1044
1045 /* For every configuration in the state "stp" which has the symbol "sp"
1046 ** following its dot, add the same configuration to the basis set under
1047 ** construction but with the dot shifted one symbol to the right. */
1048 for(bcfp=cfp; bcfp; bcfp=bcfp->next){
1049 if( bcfp->status==COMPLETE ) continue; /* Already used */
1050 if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
1051 bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
1052 if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
1053 bcfp->status = COMPLETE; /* Mark this config as used */
1054 newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
1055 Plink_add(&newcfg->bplp,bcfp);
1056 }
1057
1058 /* Get a pointer to the state described by the basis configuration set
1059 ** constructed in the preceding loop */
1060 newstp = getstate(lemp);
1061
1062 /* The state "newstp" is reached from the state "stp" by a shift action
1063 ** on the symbol "sp" */
1064 if( sp->type==MULTITERMINAL ){
1065 int i;
1066 for(i=0; i<sp->nsubsym; i++){
1067 Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
1068 }
1069 }else{
1070 Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
1071 }
1072 }
1073 }
1074
1075 /*
1076 ** Construct the propagation links
1077 */
FindLinks(struct lemon * lemp)1078 void FindLinks(struct lemon *lemp)
1079 {
1080 int i;
1081 struct config *cfp, *other;
1082 struct state *stp;
1083 struct plink *plp;
1084
1085 /* Housekeeping detail:
1086 ** Add to every propagate link a pointer back to the state to
1087 ** which the link is attached. */
1088 for(i=0; i<lemp->nstate; i++){
1089 stp = lemp->sorted[i];
1090 for(cfp=stp?stp->cfp:0; cfp; cfp=cfp->next){
1091 cfp->stp = stp;
1092 }
1093 }
1094
1095 /* Convert all backlinks into forward links. Only the forward
1096 ** links are used in the follow-set computation. */
1097 for(i=0; i<lemp->nstate; i++){
1098 stp = lemp->sorted[i];
1099 for(cfp=stp?stp->cfp:0; cfp; cfp=cfp->next){
1100 for(plp=cfp->bplp; plp; plp=plp->next){
1101 other = plp->cfp;
1102 Plink_add(&other->fplp,cfp);
1103 }
1104 }
1105 }
1106 }
1107
1108 /* Compute all followsets.
1109 **
1110 ** A followset is the set of all symbols which can come immediately
1111 ** after a configuration.
1112 */
FindFollowSets(struct lemon * lemp)1113 void FindFollowSets(struct lemon *lemp)
1114 {
1115 int i;
1116 struct config *cfp;
1117 struct plink *plp;
1118 int progress;
1119 int change;
1120
1121 for(i=0; i<lemp->nstate; i++){
1122 assert( lemp->sorted[i]!=0 );
1123 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1124 cfp->status = INCOMPLETE;
1125 }
1126 }
1127
1128 do{
1129 progress = 0;
1130 for(i=0; i<lemp->nstate; i++){
1131 assert( lemp->sorted[i]!=0 );
1132 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1133 if( cfp->status==COMPLETE ) continue;
1134 for(plp=cfp->fplp; plp; plp=plp->next){
1135 change = SetUnion(plp->cfp->fws,cfp->fws);
1136 if( change ){
1137 plp->cfp->status = INCOMPLETE;
1138 progress = 1;
1139 }
1140 }
1141 cfp->status = COMPLETE;
1142 }
1143 }
1144 }while( progress );
1145 }
1146
1147 static int resolve_conflict(struct action *,struct action *);
1148
1149 /* Compute the reduce actions, and resolve conflicts.
1150 */
FindActions(struct lemon * lemp)1151 void FindActions(struct lemon *lemp)
1152 {
1153 int i,j;
1154 struct config *cfp;
1155 struct state *stp;
1156 struct symbol *sp;
1157 struct rule *rp;
1158
1159 /* Add all of the reduce actions
1160 ** A reduce action is added for each element of the followset of
1161 ** a configuration which has its dot at the extreme right.
1162 */
1163 for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
1164 stp = lemp->sorted[i];
1165 for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
1166 if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
1167 for(j=0; j<lemp->nterminal; j++){
1168 if( SetFind(cfp->fws,j) ){
1169 /* Add a reduce action to the state "stp" which will reduce by the
1170 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
1171 Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
1172 }
1173 }
1174 }
1175 }
1176 }
1177
1178 /* Add the accepting token */
1179 if( lemp->start ){
1180 sp = Symbol_find(lemp->start);
1181 if( sp==0 ){
1182 if( lemp->startRule==0 ){
1183 fprintf(stderr, "internal error on source line %d: no start rule\n",
1184 __LINE__);
1185 exit(1);
1186 }
1187 sp = lemp->startRule->lhs;
1188 }
1189 }else{
1190 sp = lemp->startRule->lhs;
1191 }
1192 /* Add to the first state (which is always the starting state of the
1193 ** finite state machine) an action to ACCEPT if the lookahead is the
1194 ** start nonterminal. */
1195 Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
1196
1197 /* Resolve conflicts */
1198 for(i=0; i<lemp->nstate; i++){
1199 struct action *ap, *nap;
1200 stp = lemp->sorted[i];
1201 /* assert( stp->ap ); */
1202 stp->ap = Action_sort(stp->ap);
1203 for(ap=stp->ap; ap && ap->next; ap=ap->next){
1204 for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
1205 /* The two actions "ap" and "nap" have the same lookahead.
1206 ** Figure out which one should be used */
1207 lemp->nconflict += resolve_conflict(ap,nap);
1208 }
1209 }
1210 }
1211
1212 /* Report an error for each rule that can never be reduced. */
1213 for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
1214 for(i=0; i<lemp->nstate; i++){
1215 struct action *ap;
1216 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
1217 if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
1218 }
1219 }
1220 for(rp=lemp->rule; rp; rp=rp->next){
1221 if( rp->canReduce ) continue;
1222 ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
1223 lemp->errorcnt++;
1224 }
1225 }
1226
1227 /* Resolve a conflict between the two given actions. If the
1228 ** conflict can't be resolved, return non-zero.
1229 **
1230 ** NO LONGER TRUE:
1231 ** To resolve a conflict, first look to see if either action
1232 ** is on an error rule. In that case, take the action which
1233 ** is not associated with the error rule. If neither or both
1234 ** actions are associated with an error rule, then try to
1235 ** use precedence to resolve the conflict.
1236 **
1237 ** If either action is a SHIFT, then it must be apx. This
1238 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1239 */
resolve_conflict(struct action * apx,struct action * apy)1240 static int resolve_conflict(
1241 struct action *apx,
1242 struct action *apy
1243 ){
1244 struct symbol *spx, *spy;
1245 int errcnt = 0;
1246 assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
1247 if( apx->type==SHIFT && apy->type==SHIFT ){
1248 apy->type = SSCONFLICT;
1249 errcnt++;
1250 }
1251 if( apx->type==SHIFT && apy->type==REDUCE ){
1252 spx = apx->sp;
1253 spy = apy->x.rp->precsym;
1254 if( spy==0 || spx->prec<0 || spy->prec<0 ){
1255 /* Not enough precedence information. */
1256 apy->type = SRCONFLICT;
1257 errcnt++;
1258 }else if( spx->prec>spy->prec ){ /* higher precedence wins */
1259 apy->type = RD_RESOLVED;
1260 }else if( spx->prec<spy->prec ){
1261 apx->type = SH_RESOLVED;
1262 }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1263 apy->type = RD_RESOLVED; /* associativity */
1264 }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
1265 apx->type = SH_RESOLVED;
1266 }else{
1267 assert( spx->prec==spy->prec && spx->assoc==NONE );
1268 apx->type = ERROR;
1269 }
1270 }else if( apx->type==REDUCE && apy->type==REDUCE ){
1271 spx = apx->x.rp->precsym;
1272 spy = apy->x.rp->precsym;
1273 if( spx==0 || spy==0 || spx->prec<0 ||
1274 spy->prec<0 || spx->prec==spy->prec ){
1275 apy->type = RRCONFLICT;
1276 errcnt++;
1277 }else if( spx->prec>spy->prec ){
1278 apy->type = RD_RESOLVED;
1279 }else if( spx->prec<spy->prec ){
1280 apx->type = RD_RESOLVED;
1281 }
1282 }else{
1283 assert(
1284 apx->type==SH_RESOLVED ||
1285 apx->type==RD_RESOLVED ||
1286 apx->type==SSCONFLICT ||
1287 apx->type==SRCONFLICT ||
1288 apx->type==RRCONFLICT ||
1289 apy->type==SH_RESOLVED ||
1290 apy->type==RD_RESOLVED ||
1291 apy->type==SSCONFLICT ||
1292 apy->type==SRCONFLICT ||
1293 apy->type==RRCONFLICT
1294 );
1295 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1296 ** REDUCEs on the list. If we reach this point it must be because
1297 ** the parser conflict had already been resolved. */
1298 }
1299 return errcnt;
1300 }
1301 /********************* From the file "configlist.c" *************************/
1302 /*
1303 ** Routines to processing a configuration list and building a state
1304 ** in the LEMON parser generator.
1305 */
1306
1307 static struct config *freelist = 0; /* List of free configurations */
1308 static struct config *current = 0; /* Top of list of configurations */
1309 static struct config **currentend = 0; /* Last on list of configs */
1310 static struct config *basis = 0; /* Top of list of basis configs */
1311 static struct config **basisend = 0; /* End of list of basis configs */
1312
1313 /* Return a pointer to a new configuration */
newconfig(void)1314 PRIVATE struct config *newconfig(void){
1315 return (struct config*)calloc(1, sizeof(struct config));
1316 }
1317
1318 /* The configuration "old" is no longer used */
deleteconfig(struct config * old)1319 PRIVATE void deleteconfig(struct config *old)
1320 {
1321 old->next = freelist;
1322 freelist = old;
1323 }
1324
1325 /* Initialized the configuration list builder */
Configlist_init(void)1326 void Configlist_init(void){
1327 current = 0;
1328 currentend = ¤t;
1329 basis = 0;
1330 basisend = &basis;
1331 Configtable_init();
1332 return;
1333 }
1334
1335 /* Initialized the configuration list builder */
Configlist_reset(void)1336 void Configlist_reset(void){
1337 current = 0;
1338 currentend = ¤t;
1339 basis = 0;
1340 basisend = &basis;
1341 Configtable_clear(0);
1342 return;
1343 }
1344
1345 /* Add another configuration to the configuration list */
Configlist_add(struct rule * rp,int dot)1346 struct config *Configlist_add(
1347 struct rule *rp, /* The rule */
1348 int dot /* Index into the RHS of the rule where the dot goes */
1349 ){
1350 struct config *cfp, model;
1351
1352 assert( currentend!=0 );
1353 model.rp = rp;
1354 model.dot = dot;
1355 cfp = Configtable_find(&model);
1356 if( cfp==0 ){
1357 cfp = newconfig();
1358 cfp->rp = rp;
1359 cfp->dot = dot;
1360 cfp->fws = SetNew();
1361 cfp->stp = 0;
1362 cfp->fplp = cfp->bplp = 0;
1363 cfp->next = 0;
1364 cfp->bp = 0;
1365 *currentend = cfp;
1366 currentend = &cfp->next;
1367 Configtable_insert(cfp);
1368 }
1369 return cfp;
1370 }
1371
1372 /* Add a basis configuration to the configuration list */
Configlist_addbasis(struct rule * rp,int dot)1373 struct config *Configlist_addbasis(struct rule *rp, int dot)
1374 {
1375 struct config *cfp, model;
1376
1377 assert( basisend!=0 );
1378 assert( currentend!=0 );
1379 model.rp = rp;
1380 model.dot = dot;
1381 cfp = Configtable_find(&model);
1382 if( cfp==0 ){
1383 cfp = newconfig();
1384 cfp->rp = rp;
1385 cfp->dot = dot;
1386 cfp->fws = SetNew();
1387 cfp->stp = 0;
1388 cfp->fplp = cfp->bplp = 0;
1389 cfp->next = 0;
1390 cfp->bp = 0;
1391 *currentend = cfp;
1392 currentend = &cfp->next;
1393 *basisend = cfp;
1394 basisend = &cfp->bp;
1395 Configtable_insert(cfp);
1396 }
1397 return cfp;
1398 }
1399
1400 /* Compute the closure of the configuration list */
Configlist_closure(struct lemon * lemp)1401 void Configlist_closure(struct lemon *lemp)
1402 {
1403 struct config *cfp, *newcfp;
1404 struct rule *rp, *newrp;
1405 struct symbol *sp, *xsp;
1406 int i, dot;
1407
1408 assert( currentend!=0 );
1409 for(cfp=current; cfp; cfp=cfp->next){
1410 rp = cfp->rp;
1411 dot = cfp->dot;
1412 if( dot>=rp->nrhs ) continue;
1413 sp = rp->rhs[dot];
1414 if( sp->type==NONTERMINAL ){
1415 if( sp->rule==0 && sp!=lemp->errsym ){
1416 ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1417 sp->name);
1418 lemp->errorcnt++;
1419 }
1420 for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1421 newcfp = Configlist_add(newrp,0);
1422 for(i=dot+1; i<rp->nrhs; i++){
1423 xsp = rp->rhs[i];
1424 if( xsp->type==TERMINAL ){
1425 SetAdd(newcfp->fws,xsp->index);
1426 break;
1427 }else if( xsp->type==MULTITERMINAL ){
1428 int k;
1429 for(k=0; k<xsp->nsubsym; k++){
1430 SetAdd(newcfp->fws, xsp->subsym[k]->index);
1431 }
1432 break;
1433 }else{
1434 SetUnion(newcfp->fws,xsp->firstset);
1435 if( xsp->lambda==LEMON_FALSE ) break;
1436 }
1437 }
1438 if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1439 }
1440 }
1441 }
1442 return;
1443 }
1444
1445 /* Sort the configuration list */
Configlist_sort(void)1446 void Configlist_sort(void){
1447 current = (struct config*)msort((char*)current,(char**)&(current->next),
1448 Configcmp);
1449 currentend = 0;
1450 return;
1451 }
1452
1453 /* Sort the basis configuration list */
Configlist_sortbasis(void)1454 void Configlist_sortbasis(void){
1455 basis = (struct config*)msort((char*)current,(char**)&(current->bp),
1456 Configcmp);
1457 basisend = 0;
1458 return;
1459 }
1460
1461 /* Return a pointer to the head of the configuration list and
1462 ** reset the list */
Configlist_return(void)1463 struct config *Configlist_return(void){
1464 struct config *old;
1465 old = current;
1466 current = 0;
1467 currentend = 0;
1468 return old;
1469 }
1470
1471 /* Return a pointer to the head of the configuration list and
1472 ** reset the list */
Configlist_basis(void)1473 struct config *Configlist_basis(void){
1474 struct config *old;
1475 old = basis;
1476 basis = 0;
1477 basisend = 0;
1478 return old;
1479 }
1480
1481 /* Free all elements of the given configuration list */
Configlist_eat(struct config * cfp)1482 void Configlist_eat(struct config *cfp)
1483 {
1484 struct config *nextcfp;
1485 for(; cfp; cfp=nextcfp){
1486 nextcfp = cfp->next;
1487 assert( cfp->fplp==0 );
1488 assert( cfp->bplp==0 );
1489 if( cfp->fws ) SetFree(cfp->fws);
1490 deleteconfig(cfp);
1491 }
1492 return;
1493 }
1494 /***************** From the file "error.c" *********************************/
1495 /*
1496 ** Code for printing error message.
1497 */
1498
ErrorMsg(const char * filename,int lineno,const char * format,...)1499 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1500 va_list ap;
1501 fprintf(stderr, "%s:%d: ", filename, lineno);
1502 va_start(ap, format);
1503 vfprintf(stderr,format,ap);
1504 va_end(ap);
1505 fprintf(stderr, "\n");
1506 }
1507 /**************** From the file "main.c" ************************************/
1508 /*
1509 ** Main program file for the LEMON parser generator.
1510 */
1511
1512 /* Report an out-of-memory condition and abort. This function
1513 ** is used mostly by the "MemoryCheck" macro in struct.h
1514 */
memory_error(void)1515 void memory_error(void){
1516 fprintf(stderr,"Out of memory. Aborting...\n");
1517 exit(1);
1518 }
1519
1520 static int nDefine = 0; /* Number of -D options on the command line */
1521 static char **azDefine = 0; /* Name of the -D macros */
1522
1523 /* This routine is called with the argument to each -D command-line option.
1524 ** Add the macro defined to the azDefine array.
1525 */
handle_D_option(char * z)1526 static void handle_D_option(char *z){
1527 char **paz;
1528 nDefine++;
1529 azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine);
1530 if( azDefine==0 ){
1531 fprintf(stderr,"out of memory\n");
1532 exit(1);
1533 }
1534 paz = &azDefine[nDefine-1];
1535 *paz = (char *) malloc( lemonStrlen(z)+1 );
1536 if( *paz==0 ){
1537 fprintf(stderr,"out of memory\n");
1538 exit(1);
1539 }
1540 lemon_strcpy(*paz, z);
1541 for(z=*paz; *z && *z!='='; z++){}
1542 *z = 0;
1543 }
1544
1545 /* Rember the name of the output directory
1546 */
1547 static char *outputDir = NULL;
handle_d_option(char * z)1548 static void handle_d_option(char *z){
1549 outputDir = (char *) malloc( lemonStrlen(z)+1 );
1550 if( outputDir==0 ){
1551 fprintf(stderr,"out of memory\n");
1552 exit(1);
1553 }
1554 lemon_strcpy(outputDir, z);
1555 }
1556
1557 static char *user_templatename = NULL;
handle_T_option(char * z)1558 static void handle_T_option(char *z){
1559 user_templatename = (char *) malloc( lemonStrlen(z)+1 );
1560 if( user_templatename==0 ){
1561 memory_error();
1562 }
1563 lemon_strcpy(user_templatename, z);
1564 }
1565
1566 /* Merge together to lists of rules ordered by rule.iRule */
Rule_merge(struct rule * pA,struct rule * pB)1567 static struct rule *Rule_merge(struct rule *pA, struct rule *pB){
1568 struct rule *pFirst = 0;
1569 struct rule **ppPrev = &pFirst;
1570 while( pA && pB ){
1571 if( pA->iRule<pB->iRule ){
1572 *ppPrev = pA;
1573 ppPrev = &pA->next;
1574 pA = pA->next;
1575 }else{
1576 *ppPrev = pB;
1577 ppPrev = &pB->next;
1578 pB = pB->next;
1579 }
1580 }
1581 if( pA ){
1582 *ppPrev = pA;
1583 }else{
1584 *ppPrev = pB;
1585 }
1586 return pFirst;
1587 }
1588
1589 /*
1590 ** Sort a list of rules in order of increasing iRule value
1591 */
Rule_sort(struct rule * rp)1592 static struct rule *Rule_sort(struct rule *rp){
1593 unsigned int i;
1594 struct rule *pNext;
1595 struct rule *x[32];
1596 memset(x, 0, sizeof(x));
1597 while( rp ){
1598 pNext = rp->next;
1599 rp->next = 0;
1600 for(i=0; i<sizeof(x)/sizeof(x[0])-1 && x[i]; i++){
1601 rp = Rule_merge(x[i], rp);
1602 x[i] = 0;
1603 }
1604 x[i] = rp;
1605 rp = pNext;
1606 }
1607 rp = 0;
1608 for(i=0; i<sizeof(x)/sizeof(x[0]); i++){
1609 rp = Rule_merge(x[i], rp);
1610 }
1611 return rp;
1612 }
1613
1614 /* forward reference */
1615 static const char *minimum_size_type(int lwr, int upr, int *pnByte);
1616
1617 /* Print a single line of the "Parser Stats" output
1618 */
stats_line(const char * zLabel,int iValue)1619 static void stats_line(const char *zLabel, int iValue){
1620 int nLabel = lemonStrlen(zLabel);
1621 printf(" %s%.*s %5d\n", zLabel,
1622 35-nLabel, "................................",
1623 iValue);
1624 }
1625
1626 /* The main program. Parse the command line and do it... */
main(int argc,char ** argv)1627 int main(int argc, char **argv){
1628 static int version = 0;
1629 static int rpflag = 0;
1630 static int basisflag = 0;
1631 static int compress = 0;
1632 static int quiet = 0;
1633 static int statistics = 0;
1634 static int mhflag = 0;
1635 static int nolinenosflag = 0;
1636 static int noResort = 0;
1637 static int sqlFlag = 0;
1638 static int printPP = 0;
1639
1640 static struct s_options options[] = {
1641 {OPT_FLAG, "b", (char*)&basisflag, 0, "Print only the basis in report."},
1642 {OPT_FLAG, "c", (char*)&compress, 0, "Don't compress the action table."},
1643 {OPT_FSTR, "d", 0, handle_d_option, "Output directory. Default '.'"},
1644 {OPT_FSTR, "D", 0, handle_D_option, "Define an %ifdef macro."},
1645 {OPT_FLAG, "E", (char*)&printPP, 0, "Print input file after preprocessing."},
1646 {OPT_FSTR, "f", 0, 0, "Ignored. (Placeholder for -f compiler options.)"},
1647 {OPT_FLAG, "g", (char*)&rpflag, 0, "Print grammar without actions."},
1648 {OPT_FSTR, "I", 0, 0, "Ignored. (Placeholder for '-I' compiler options.)"},
1649 {OPT_FLAG, "m", (char*)&mhflag, 0, "Output a makeheaders compatible file."},
1650 {OPT_FLAG, "l", (char*)&nolinenosflag, 0, "Do not print #line statements."},
1651 {OPT_FSTR, "O", 0, 0, "Ignored. (Placeholder for '-O' compiler options.)"},
1652 {OPT_FLAG, "p", (char*)&showPrecedenceConflict, 0,
1653 "Show conflicts resolved by precedence rules"},
1654 {OPT_FLAG, "q", (char*)&quiet, 0, "(Quiet) Don't print the report file."},
1655 {OPT_FLAG, "r", (char*)&noResort, 0, "Do not sort or renumber states"},
1656 {OPT_FLAG, "s", (char*)&statistics, 0,
1657 "Print parser stats to standard output."},
1658 {OPT_FLAG, "S", (char*)&sqlFlag, 0,
1659 "Generate the *.sql file describing the parser tables."},
1660 {OPT_FLAG, "x", (char*)&version, 0, "Print the version number."},
1661 {OPT_FSTR, "T", 0, handle_T_option, "Specify a template file."},
1662 {OPT_FSTR, "W", 0, 0, "Ignored. (Placeholder for '-W' compiler options.)"},
1663 {OPT_FLAG,0,0,0,0}
1664 };
1665 int i;
1666 int exitcode;
1667 struct lemon lem;
1668 struct rule *rp;
1669
1670 (void)argc;
1671 OptInit(argv,options,stderr);
1672 if( version ){
1673 printf("Lemon version 1.0\n");
1674 exit(0);
1675 }
1676 if( OptNArgs()!=1 ){
1677 fprintf(stderr,"Exactly one filename argument is required.\n");
1678 exit(1);
1679 }
1680 memset(&lem, 0, sizeof(lem));
1681 lem.errorcnt = 0;
1682
1683 /* Initialize the machine */
1684 Strsafe_init();
1685 Symbol_init();
1686 State_init();
1687 lem.argv0 = argv[0];
1688 lem.filename = OptArg(0);
1689 lem.basisflag = basisflag;
1690 lem.nolinenosflag = nolinenosflag;
1691 lem.printPreprocessed = printPP;
1692 Symbol_new("$");
1693
1694 /* Parse the input file */
1695 Parse(&lem);
1696 if( lem.printPreprocessed || lem.errorcnt ) exit(lem.errorcnt);
1697 if( lem.nrule==0 ){
1698 fprintf(stderr,"Empty grammar.\n");
1699 exit(1);
1700 }
1701 lem.errsym = Symbol_find("error");
1702
1703 /* Count and index the symbols of the grammar */
1704 Symbol_new("{default}");
1705 lem.nsymbol = Symbol_count();
1706 lem.symbols = Symbol_arrayof();
1707 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1708 qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp);
1709 for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1710 while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; }
1711 assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 );
1712 lem.nsymbol = i - 1;
1713 for(i=1; ISUPPER(lem.symbols[i]->name[0]); i++);
1714 lem.nterminal = i;
1715
1716 /* Assign sequential rule numbers. Start with 0. Put rules that have no
1717 ** reduce action C-code associated with them last, so that the switch()
1718 ** statement that selects reduction actions will have a smaller jump table.
1719 */
1720 for(i=0, rp=lem.rule; rp; rp=rp->next){
1721 rp->iRule = rp->code ? i++ : -1;
1722 }
1723 lem.nruleWithAction = i;
1724 for(rp=lem.rule; rp; rp=rp->next){
1725 if( rp->iRule<0 ) rp->iRule = i++;
1726 }
1727 lem.startRule = lem.rule;
1728 lem.rule = Rule_sort(lem.rule);
1729
1730 /* Generate a reprint of the grammar, if requested on the command line */
1731 if( rpflag ){
1732 Reprint(&lem);
1733 }else{
1734 /* Initialize the size for all follow and first sets */
1735 SetSize(lem.nterminal+1);
1736
1737 /* Find the precedence for every production rule (that has one) */
1738 FindRulePrecedences(&lem);
1739
1740 /* Compute the lambda-nonterminals and the first-sets for every
1741 ** nonterminal */
1742 FindFirstSets(&lem);
1743
1744 /* Compute all LR(0) states. Also record follow-set propagation
1745 ** links so that the follow-set can be computed later */
1746 lem.nstate = 0;
1747 FindStates(&lem);
1748 lem.sorted = State_arrayof();
1749
1750 /* Tie up loose ends on the propagation links */
1751 FindLinks(&lem);
1752
1753 /* Compute the follow set of every reducible configuration */
1754 FindFollowSets(&lem);
1755
1756 /* Compute the action tables */
1757 FindActions(&lem);
1758
1759 /* Compress the action tables */
1760 if( compress==0 ) CompressTables(&lem);
1761
1762 /* Reorder and renumber the states so that states with fewer choices
1763 ** occur at the end. This is an optimization that helps make the
1764 ** generated parser tables smaller. */
1765 if( noResort==0 ) ResortStates(&lem);
1766
1767 /* Generate a report of the parser generated. (the "y.output" file) */
1768 if( !quiet ) ReportOutput(&lem);
1769
1770 /* Generate the source code for the parser */
1771 ReportTable(&lem, mhflag, sqlFlag);
1772
1773 /* Produce a header file for use by the scanner. (This step is
1774 ** omitted if the "-m" option is used because makeheaders will
1775 ** generate the file for us.) */
1776 if( !mhflag ) ReportHeader(&lem);
1777 }
1778 if( statistics ){
1779 printf("Parser statistics:\n");
1780 stats_line("terminal symbols", lem.nterminal);
1781 stats_line("non-terminal symbols", lem.nsymbol - lem.nterminal);
1782 stats_line("total symbols", lem.nsymbol);
1783 stats_line("rules", lem.nrule);
1784 stats_line("states", lem.nxstate);
1785 stats_line("conflicts", lem.nconflict);
1786 stats_line("action table entries", lem.nactiontab);
1787 stats_line("lookahead table entries", lem.nlookaheadtab);
1788 stats_line("total table size (bytes)", lem.tablesize);
1789 }
1790 if( lem.nconflict > 0 ){
1791 fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1792 }
1793
1794 /* return 0 on success, 1 on failure. */
1795 exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0;
1796 exit(exitcode);
1797 return (exitcode);
1798 }
1799 /******************** From the file "msort.c" *******************************/
1800 /*
1801 ** A generic merge-sort program.
1802 **
1803 ** USAGE:
1804 ** Let "ptr" be a pointer to some structure which is at the head of
1805 ** a null-terminated list. Then to sort the list call:
1806 **
1807 ** ptr = msort(ptr,&(ptr->next),cmpfnc);
1808 **
1809 ** In the above, "cmpfnc" is a pointer to a function which compares
1810 ** two instances of the structure and returns an integer, as in
1811 ** strcmp. The second argument is a pointer to the pointer to the
1812 ** second element of the linked list. This address is used to compute
1813 ** the offset to the "next" field within the structure. The offset to
1814 ** the "next" field must be constant for all structures in the list.
1815 **
1816 ** The function returns a new pointer which is the head of the list
1817 ** after sorting.
1818 **
1819 ** ALGORITHM:
1820 ** Merge-sort.
1821 */
1822
1823 /*
1824 ** Return a pointer to the next structure in the linked list.
1825 */
1826 #define NEXT(A) (*(char**)(((char*)A)+offset))
1827
1828 /*
1829 ** Inputs:
1830 ** a: A sorted, null-terminated linked list. (May be null).
1831 ** b: A sorted, null-terminated linked list. (May be null).
1832 ** cmp: A pointer to the comparison function.
1833 ** offset: Offset in the structure to the "next" field.
1834 **
1835 ** Return Value:
1836 ** A pointer to the head of a sorted list containing the elements
1837 ** of both a and b.
1838 **
1839 ** Side effects:
1840 ** The "next" pointers for elements in the lists a and b are
1841 ** changed.
1842 */
merge(char * a,char * b,int (* cmp)(const char *,const char *),int offset)1843 static char *merge(
1844 char *a,
1845 char *b,
1846 int (*cmp)(const char*,const char*),
1847 int offset
1848 ){
1849 char *ptr, *head;
1850
1851 if( a==0 ){
1852 head = b;
1853 }else if( b==0 ){
1854 head = a;
1855 }else{
1856 if( (*cmp)(a,b)<=0 ){
1857 ptr = a;
1858 a = NEXT(a);
1859 }else{
1860 ptr = b;
1861 b = NEXT(b);
1862 }
1863 head = ptr;
1864 while( a && b ){
1865 if( (*cmp)(a,b)<=0 ){
1866 NEXT(ptr) = a;
1867 ptr = a;
1868 a = NEXT(a);
1869 }else{
1870 NEXT(ptr) = b;
1871 ptr = b;
1872 b = NEXT(b);
1873 }
1874 }
1875 if( a ) NEXT(ptr) = a;
1876 else NEXT(ptr) = b;
1877 }
1878 return head;
1879 }
1880
1881 /*
1882 ** Inputs:
1883 ** list: Pointer to a singly-linked list of structures.
1884 ** next: Pointer to pointer to the second element of the list.
1885 ** cmp: A comparison function.
1886 **
1887 ** Return Value:
1888 ** A pointer to the head of a sorted list containing the elements
1889 ** originally in list.
1890 **
1891 ** Side effects:
1892 ** The "next" pointers for elements in list are changed.
1893 */
1894 #define LISTSIZE 30
msort(char * list,char ** next,int (* cmp)(const char *,const char *))1895 static char *msort(
1896 char *list,
1897 char **next,
1898 int (*cmp)(const char*,const char*)
1899 ){
1900 unsigned long offset;
1901 char *ep;
1902 char *set[LISTSIZE];
1903 int i;
1904 offset = (unsigned long)((char*)next - (char*)list);
1905 for(i=0; i<LISTSIZE; i++) set[i] = 0;
1906 while( list ){
1907 ep = list;
1908 list = NEXT(list);
1909 NEXT(ep) = 0;
1910 for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1911 ep = merge(ep,set[i],cmp,offset);
1912 set[i] = 0;
1913 }
1914 set[i] = ep;
1915 }
1916 ep = 0;
1917 for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset);
1918 return ep;
1919 }
1920 /************************ From the file "option.c" **************************/
1921 static char **g_argv;
1922 static struct s_options *op;
1923 static FILE *errstream;
1924
1925 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1926
1927 /*
1928 ** Print the command line with a carrot pointing to the k-th character
1929 ** of the n-th field.
1930 */
errline(int n,int k,FILE * err)1931 static void errline(int n, int k, FILE *err)
1932 {
1933 int spcnt, i;
1934 if( g_argv[0] ){
1935 fprintf(err,"%s",g_argv[0]);
1936 spcnt = lemonStrlen(g_argv[0]) + 1;
1937 }else{
1938 spcnt = 0;
1939 }
1940 for(i=1; i<n && g_argv[i]; i++){
1941 fprintf(err," %s",g_argv[i]);
1942 spcnt += lemonStrlen(g_argv[i])+1;
1943 }
1944 spcnt += k;
1945 for(; g_argv[i]; i++) fprintf(err," %s",g_argv[i]);
1946 if( spcnt<20 ){
1947 fprintf(err,"\n%*s^-- here\n",spcnt,"");
1948 }else{
1949 fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1950 }
1951 }
1952
1953 /*
1954 ** Return the index of the N-th non-switch argument. Return -1
1955 ** if N is out of range.
1956 */
argindex(int n)1957 static int argindex(int n)
1958 {
1959 int i;
1960 int dashdash = 0;
1961 if( g_argv!=0 && *g_argv!=0 ){
1962 for(i=1; g_argv[i]; i++){
1963 if( dashdash || !ISOPT(g_argv[i]) ){
1964 if( n==0 ) return i;
1965 n--;
1966 }
1967 if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
1968 }
1969 }
1970 return -1;
1971 }
1972
1973 static char emsg[] = "Command line syntax error: ";
1974
1975 /*
1976 ** Process a flag command line argument.
1977 */
handleflags(int i,FILE * err)1978 static int handleflags(int i, FILE *err)
1979 {
1980 int v;
1981 int errcnt = 0;
1982 int j;
1983 for(j=0; op[j].label; j++){
1984 if( strncmp(&g_argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
1985 }
1986 v = g_argv[i][0]=='-' ? 1 : 0;
1987 if( op[j].label==0 ){
1988 if( err ){
1989 fprintf(err,"%sundefined option.\n",emsg);
1990 errline(i,1,err);
1991 }
1992 errcnt++;
1993 }else if( op[j].arg==0 && op[j].fn==0 ){
1994 /* Ignore this option */
1995 }else if( op[j].type==OPT_FLAG ){
1996 *((int*)op[j].arg) = v;
1997 }else if( op[j].type==OPT_FFLAG ){
1998 op[j].fn(v);
1999 }else if( op[j].type==OPT_FSTR ){
2000 op[j].fn(&g_argv[i][2]);
2001 }else{
2002 if( err ){
2003 fprintf(err,"%smissing argument on switch.\n",emsg);
2004 errline(i,1,err);
2005 }
2006 errcnt++;
2007 }
2008 return errcnt;
2009 }
2010
2011 /*
2012 ** Process a command line switch which has an argument.
2013 */
handleswitch(int i,FILE * err)2014 static int handleswitch(int i, FILE *err)
2015 {
2016 int lv = 0;
2017 double dv = 0.0;
2018 char *sv = 0, *end;
2019 char *cp;
2020 int j;
2021 int errcnt = 0;
2022 cp = strchr(g_argv[i],'=');
2023 assert( cp!=0 );
2024 *cp = 0;
2025 for(j=0; op[j].label; j++){
2026 if( strcmp(g_argv[i],op[j].label)==0 ) break;
2027 }
2028 *cp = '=';
2029 if( op[j].label==0 ){
2030 if( err ){
2031 fprintf(err,"%sundefined option.\n",emsg);
2032 errline(i,0,err);
2033 }
2034 errcnt++;
2035 }else{
2036 cp++;
2037 switch( op[j].type ){
2038 case OPT_FLAG:
2039 case OPT_FFLAG:
2040 if( err ){
2041 fprintf(err,"%soption requires an argument.\n",emsg);
2042 errline(i,0,err);
2043 }
2044 errcnt++;
2045 break;
2046 case OPT_DBL:
2047 case OPT_FDBL:
2048 dv = strtod(cp,&end);
2049 if( *end ){
2050 if( err ){
2051 fprintf(err,
2052 "%sillegal character in floating-point argument.\n",emsg);
2053 errline(i,(int)((char*)end-(char*)g_argv[i]),err);
2054 }
2055 errcnt++;
2056 }
2057 break;
2058 case OPT_INT:
2059 case OPT_FINT:
2060 lv = strtol(cp,&end,0);
2061 if( *end ){
2062 if( err ){
2063 fprintf(err,"%sillegal character in integer argument.\n",emsg);
2064 errline(i,(int)((char*)end-(char*)g_argv[i]),err);
2065 }
2066 errcnt++;
2067 }
2068 break;
2069 case OPT_STR:
2070 case OPT_FSTR:
2071 sv = cp;
2072 break;
2073 }
2074 switch( op[j].type ){
2075 case OPT_FLAG:
2076 case OPT_FFLAG:
2077 break;
2078 case OPT_DBL:
2079 *(double*)(op[j].arg) = dv;
2080 break;
2081 case OPT_FDBL:
2082 op[j].fn(dv);
2083 break;
2084 case OPT_INT:
2085 *(int*)(op[j].arg) = lv;
2086 break;
2087 case OPT_FINT:
2088 op[j].fn((int)lv);
2089 break;
2090 case OPT_STR:
2091 *(char**)(op[j].arg) = sv;
2092 break;
2093 case OPT_FSTR:
2094 op[j].fn(sv);
2095 break;
2096 }
2097 }
2098 return errcnt;
2099 }
2100
OptInit(char ** a,struct s_options * o,FILE * err)2101 int OptInit(char **a, struct s_options *o, FILE *err)
2102 {
2103 int errcnt = 0;
2104 g_argv = a;
2105 op = o;
2106 errstream = err;
2107 if( g_argv && *g_argv && op ){
2108 int i;
2109 for(i=1; g_argv[i]; i++){
2110 if( g_argv[i][0]=='+' || g_argv[i][0]=='-' ){
2111 errcnt += handleflags(i,err);
2112 }else if( strchr(g_argv[i],'=') ){
2113 errcnt += handleswitch(i,err);
2114 }
2115 }
2116 }
2117 if( errcnt>0 ){
2118 fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
2119 OptPrint();
2120 exit(1);
2121 }
2122 return 0;
2123 }
2124
OptNArgs(void)2125 int OptNArgs(void){
2126 int cnt = 0;
2127 int dashdash = 0;
2128 int i;
2129 if( g_argv!=0 && g_argv[0]!=0 ){
2130 for(i=1; g_argv[i]; i++){
2131 if( dashdash || !ISOPT(g_argv[i]) ) cnt++;
2132 if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
2133 }
2134 }
2135 return cnt;
2136 }
2137
OptArg(int n)2138 char *OptArg(int n)
2139 {
2140 int i;
2141 i = argindex(n);
2142 return i>=0 ? g_argv[i] : 0;
2143 }
2144
OptErr(int n)2145 void OptErr(int n)
2146 {
2147 int i;
2148 i = argindex(n);
2149 if( i>=0 ) errline(i,0,errstream);
2150 }
2151
OptPrint(void)2152 void OptPrint(void){
2153 int i;
2154 int max, len;
2155 max = 0;
2156 for(i=0; op[i].label; i++){
2157 len = lemonStrlen(op[i].label) + 1;
2158 switch( op[i].type ){
2159 case OPT_FLAG:
2160 case OPT_FFLAG:
2161 break;
2162 case OPT_INT:
2163 case OPT_FINT:
2164 len += 9; /* length of "<integer>" */
2165 break;
2166 case OPT_DBL:
2167 case OPT_FDBL:
2168 len += 6; /* length of "<real>" */
2169 break;
2170 case OPT_STR:
2171 case OPT_FSTR:
2172 len += 8; /* length of "<string>" */
2173 break;
2174 }
2175 if( len>max ) max = len;
2176 }
2177 for(i=0; op[i].label; i++){
2178 switch( op[i].type ){
2179 case OPT_FLAG:
2180 case OPT_FFLAG:
2181 fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
2182 break;
2183 case OPT_INT:
2184 case OPT_FINT:
2185 fprintf(errstream," -%s<integer>%*s %s\n",op[i].label,
2186 (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
2187 break;
2188 case OPT_DBL:
2189 case OPT_FDBL:
2190 fprintf(errstream," -%s<real>%*s %s\n",op[i].label,
2191 (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
2192 break;
2193 case OPT_STR:
2194 case OPT_FSTR:
2195 fprintf(errstream," -%s<string>%*s %s\n",op[i].label,
2196 (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
2197 break;
2198 }
2199 }
2200 }
2201 /*********************** From the file "parse.c" ****************************/
2202 /*
2203 ** Input file parser for the LEMON parser generator.
2204 */
2205
2206 /* The state of the parser */
2207 enum e_state {
2208 INITIALIZE,
2209 WAITING_FOR_DECL_OR_RULE,
2210 WAITING_FOR_DECL_KEYWORD,
2211 WAITING_FOR_DECL_ARG,
2212 WAITING_FOR_PRECEDENCE_SYMBOL,
2213 WAITING_FOR_ARROW,
2214 IN_RHS,
2215 LHS_ALIAS_1,
2216 LHS_ALIAS_2,
2217 LHS_ALIAS_3,
2218 RHS_ALIAS_1,
2219 RHS_ALIAS_2,
2220 PRECEDENCE_MARK_1,
2221 PRECEDENCE_MARK_2,
2222 RESYNC_AFTER_RULE_ERROR,
2223 RESYNC_AFTER_DECL_ERROR,
2224 WAITING_FOR_DESTRUCTOR_SYMBOL,
2225 WAITING_FOR_DATATYPE_SYMBOL,
2226 WAITING_FOR_FALLBACK_ID,
2227 WAITING_FOR_WILDCARD_ID,
2228 WAITING_FOR_CLASS_ID,
2229 WAITING_FOR_CLASS_TOKEN,
2230 WAITING_FOR_TOKEN_NAME
2231 };
2232 struct pstate {
2233 char *filename; /* Name of the input file */
2234 int tokenlineno; /* Linenumber at which current token starts */
2235 int errorcnt; /* Number of errors so far */
2236 char *tokenstart; /* Text of current token */
2237 struct lemon *gp; /* Global state vector */
2238 enum e_state state; /* The state of the parser */
2239 struct symbol *fallback; /* The fallback token */
2240 struct symbol *tkclass; /* Token class symbol */
2241 struct symbol *lhs; /* Left-hand side of current rule */
2242 const char *lhsalias; /* Alias for the LHS */
2243 int nrhs; /* Number of right-hand side symbols seen */
2244 struct symbol *rhs[MAXRHS]; /* RHS symbols */
2245 const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
2246 struct rule *prevrule; /* Previous rule parsed */
2247 const char *declkeyword; /* Keyword of a declaration */
2248 char **declargslot; /* Where the declaration argument should be put */
2249 int insertLineMacro; /* Add #line before declaration insert */
2250 int *decllinenoslot; /* Where to write declaration line number */
2251 enum e_assoc declassoc; /* Assign this association to decl arguments */
2252 int preccounter; /* Assign this precedence to decl arguments */
2253 struct rule *firstrule; /* Pointer to first rule in the grammar */
2254 struct rule *lastrule; /* Pointer to the most recently parsed rule */
2255 };
2256
2257 /* Parse a single token */
parseonetoken(struct pstate * psp)2258 static void parseonetoken(struct pstate *psp)
2259 {
2260 const char *x;
2261 x = Strsafe(psp->tokenstart); /* Save the token permanently */
2262 #if 0
2263 printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
2264 x,psp->state);
2265 #endif
2266 switch( psp->state ){
2267 case INITIALIZE:
2268 psp->prevrule = 0;
2269 psp->preccounter = 0;
2270 psp->firstrule = psp->lastrule = 0;
2271 psp->gp->nrule = 0;
2272 /* fall through */
2273 case WAITING_FOR_DECL_OR_RULE:
2274 if( x[0]=='%' ){
2275 psp->state = WAITING_FOR_DECL_KEYWORD;
2276 }else if( ISLOWER(x[0]) ){
2277 psp->lhs = Symbol_new(x);
2278 psp->nrhs = 0;
2279 psp->lhsalias = 0;
2280 psp->state = WAITING_FOR_ARROW;
2281 }else if( x[0]=='{' ){
2282 if( psp->prevrule==0 ){
2283 ErrorMsg(psp->filename,psp->tokenlineno,
2284 "There is no prior rule upon which to attach the code "
2285 "fragment which begins on this line.");
2286 psp->errorcnt++;
2287 }else if( psp->prevrule->code!=0 ){
2288 ErrorMsg(psp->filename,psp->tokenlineno,
2289 "Code fragment beginning on this line is not the first "
2290 "to follow the previous rule.");
2291 psp->errorcnt++;
2292 }else if( strcmp(x, "{NEVER-REDUCE")==0 ){
2293 psp->prevrule->neverReduce = 1;
2294 }else{
2295 psp->prevrule->line = psp->tokenlineno;
2296 psp->prevrule->code = &x[1];
2297 psp->prevrule->noCode = 0;
2298 }
2299 }else if( x[0]=='[' ){
2300 psp->state = PRECEDENCE_MARK_1;
2301 }else{
2302 ErrorMsg(psp->filename,psp->tokenlineno,
2303 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2304 x);
2305 psp->errorcnt++;
2306 }
2307 break;
2308 case PRECEDENCE_MARK_1:
2309 if( !ISUPPER(x[0]) ){
2310 ErrorMsg(psp->filename,psp->tokenlineno,
2311 "The precedence symbol must be a terminal.");
2312 psp->errorcnt++;
2313 }else if( psp->prevrule==0 ){
2314 ErrorMsg(psp->filename,psp->tokenlineno,
2315 "There is no prior rule to assign precedence \"[%s]\".",x);
2316 psp->errorcnt++;
2317 }else if( psp->prevrule->precsym!=0 ){
2318 ErrorMsg(psp->filename,psp->tokenlineno,
2319 "Precedence mark on this line is not the first "
2320 "to follow the previous rule.");
2321 psp->errorcnt++;
2322 }else{
2323 psp->prevrule->precsym = Symbol_new(x);
2324 }
2325 psp->state = PRECEDENCE_MARK_2;
2326 break;
2327 case PRECEDENCE_MARK_2:
2328 if( x[0]!=']' ){
2329 ErrorMsg(psp->filename,psp->tokenlineno,
2330 "Missing \"]\" on precedence mark.");
2331 psp->errorcnt++;
2332 }
2333 psp->state = WAITING_FOR_DECL_OR_RULE;
2334 break;
2335 case WAITING_FOR_ARROW:
2336 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2337 psp->state = IN_RHS;
2338 }else if( x[0]=='(' ){
2339 psp->state = LHS_ALIAS_1;
2340 }else{
2341 ErrorMsg(psp->filename,psp->tokenlineno,
2342 "Expected to see a \":\" following the LHS symbol \"%s\".",
2343 psp->lhs->name);
2344 psp->errorcnt++;
2345 psp->state = RESYNC_AFTER_RULE_ERROR;
2346 }
2347 break;
2348 case LHS_ALIAS_1:
2349 if( ISALPHA(x[0]) ){
2350 psp->lhsalias = x;
2351 psp->state = LHS_ALIAS_2;
2352 }else{
2353 ErrorMsg(psp->filename,psp->tokenlineno,
2354 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2355 x,psp->lhs->name);
2356 psp->errorcnt++;
2357 psp->state = RESYNC_AFTER_RULE_ERROR;
2358 }
2359 break;
2360 case LHS_ALIAS_2:
2361 if( x[0]==')' ){
2362 psp->state = LHS_ALIAS_3;
2363 }else{
2364 ErrorMsg(psp->filename,psp->tokenlineno,
2365 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2366 psp->errorcnt++;
2367 psp->state = RESYNC_AFTER_RULE_ERROR;
2368 }
2369 break;
2370 case LHS_ALIAS_3:
2371 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2372 psp->state = IN_RHS;
2373 }else{
2374 ErrorMsg(psp->filename,psp->tokenlineno,
2375 "Missing \"->\" following: \"%s(%s)\".",
2376 psp->lhs->name,psp->lhsalias);
2377 psp->errorcnt++;
2378 psp->state = RESYNC_AFTER_RULE_ERROR;
2379 }
2380 break;
2381 case IN_RHS:
2382 if( x[0]=='.' ){
2383 struct rule *rp;
2384 rp = (struct rule *)calloc( sizeof(struct rule) +
2385 sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
2386 if( rp==0 ){
2387 ErrorMsg(psp->filename,psp->tokenlineno,
2388 "Can't allocate enough memory for this rule.");
2389 psp->errorcnt++;
2390 psp->prevrule = 0;
2391 }else{
2392 int i;
2393 rp->ruleline = psp->tokenlineno;
2394 rp->rhs = (struct symbol**)&rp[1];
2395 rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]);
2396 for(i=0; i<psp->nrhs; i++){
2397 rp->rhs[i] = psp->rhs[i];
2398 rp->rhsalias[i] = psp->alias[i];
2399 if( rp->rhsalias[i]!=0 ){ rp->rhs[i]->bContent = 1; }
2400 }
2401 rp->lhs = psp->lhs;
2402 rp->lhsalias = psp->lhsalias;
2403 rp->nrhs = psp->nrhs;
2404 rp->code = 0;
2405 rp->noCode = 1;
2406 rp->precsym = 0;
2407 rp->index = psp->gp->nrule++;
2408 rp->nextlhs = rp->lhs->rule;
2409 rp->lhs->rule = rp;
2410 rp->next = 0;
2411 if( psp->firstrule==0 ){
2412 psp->firstrule = psp->lastrule = rp;
2413 }else{
2414 psp->lastrule->next = rp;
2415 psp->lastrule = rp;
2416 }
2417 psp->prevrule = rp;
2418 }
2419 psp->state = WAITING_FOR_DECL_OR_RULE;
2420 }else if( ISALPHA(x[0]) ){
2421 if( psp->nrhs>=MAXRHS ){
2422 ErrorMsg(psp->filename,psp->tokenlineno,
2423 "Too many symbols on RHS of rule beginning at \"%s\".",
2424 x);
2425 psp->errorcnt++;
2426 psp->state = RESYNC_AFTER_RULE_ERROR;
2427 }else{
2428 psp->rhs[psp->nrhs] = Symbol_new(x);
2429 psp->alias[psp->nrhs] = 0;
2430 psp->nrhs++;
2431 }
2432 }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 && ISUPPER(x[1]) ){
2433 struct symbol *msp = psp->rhs[psp->nrhs-1];
2434 if( msp->type!=MULTITERMINAL ){
2435 struct symbol *origsp = msp;
2436 msp = (struct symbol *) calloc(1,sizeof(*msp));
2437 memset(msp, 0, sizeof(*msp));
2438 msp->type = MULTITERMINAL;
2439 msp->nsubsym = 1;
2440 msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*));
2441 msp->subsym[0] = origsp;
2442 msp->name = origsp->name;
2443 psp->rhs[psp->nrhs-1] = msp;
2444 }
2445 msp->nsubsym++;
2446 msp->subsym = (struct symbol **) realloc(msp->subsym,
2447 sizeof(struct symbol*)*msp->nsubsym);
2448 msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
2449 if( ISLOWER(x[1]) || ISLOWER(msp->subsym[0]->name[0]) ){
2450 ErrorMsg(psp->filename,psp->tokenlineno,
2451 "Cannot form a compound containing a non-terminal");
2452 psp->errorcnt++;
2453 }
2454 }else if( x[0]=='(' && psp->nrhs>0 ){
2455 psp->state = RHS_ALIAS_1;
2456 }else{
2457 ErrorMsg(psp->filename,psp->tokenlineno,
2458 "Illegal character on RHS of rule: \"%s\".",x);
2459 psp->errorcnt++;
2460 psp->state = RESYNC_AFTER_RULE_ERROR;
2461 }
2462 break;
2463 case RHS_ALIAS_1:
2464 if( ISALPHA(x[0]) ){
2465 psp->alias[psp->nrhs-1] = x;
2466 psp->state = RHS_ALIAS_2;
2467 }else{
2468 ErrorMsg(psp->filename,psp->tokenlineno,
2469 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2470 x,psp->rhs[psp->nrhs-1]->name);
2471 psp->errorcnt++;
2472 psp->state = RESYNC_AFTER_RULE_ERROR;
2473 }
2474 break;
2475 case RHS_ALIAS_2:
2476 if( x[0]==')' ){
2477 psp->state = IN_RHS;
2478 }else{
2479 ErrorMsg(psp->filename,psp->tokenlineno,
2480 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2481 psp->errorcnt++;
2482 psp->state = RESYNC_AFTER_RULE_ERROR;
2483 }
2484 break;
2485 case WAITING_FOR_DECL_KEYWORD:
2486 if( ISALPHA(x[0]) ){
2487 psp->declkeyword = x;
2488 psp->declargslot = 0;
2489 psp->decllinenoslot = 0;
2490 psp->insertLineMacro = 1;
2491 psp->state = WAITING_FOR_DECL_ARG;
2492 if( strcmp(x,"name")==0 ){
2493 psp->declargslot = &(psp->gp->name);
2494 psp->insertLineMacro = 0;
2495 }else if( strcmp(x,"include")==0 ){
2496 psp->declargslot = &(psp->gp->include);
2497 }else if( strcmp(x,"code")==0 ){
2498 psp->declargslot = &(psp->gp->extracode);
2499 }else if( strcmp(x,"token_destructor")==0 ){
2500 psp->declargslot = &psp->gp->tokendest;
2501 }else if( strcmp(x,"default_destructor")==0 ){
2502 psp->declargslot = &psp->gp->vardest;
2503 }else if( strcmp(x,"token_prefix")==0 ){
2504 psp->declargslot = &psp->gp->tokenprefix;
2505 psp->insertLineMacro = 0;
2506 }else if( strcmp(x,"syntax_error")==0 ){
2507 psp->declargslot = &(psp->gp->error);
2508 }else if( strcmp(x,"parse_accept")==0 ){
2509 psp->declargslot = &(psp->gp->accept);
2510 }else if( strcmp(x,"parse_failure")==0 ){
2511 psp->declargslot = &(psp->gp->failure);
2512 }else if( strcmp(x,"stack_overflow")==0 ){
2513 psp->declargslot = &(psp->gp->overflow);
2514 }else if( strcmp(x,"extra_argument")==0 ){
2515 psp->declargslot = &(psp->gp->arg);
2516 psp->insertLineMacro = 0;
2517 }else if( strcmp(x,"extra_context")==0 ){
2518 psp->declargslot = &(psp->gp->ctx);
2519 psp->insertLineMacro = 0;
2520 }else if( strcmp(x,"token_type")==0 ){
2521 psp->declargslot = &(psp->gp->tokentype);
2522 psp->insertLineMacro = 0;
2523 }else if( strcmp(x,"default_type")==0 ){
2524 psp->declargslot = &(psp->gp->vartype);
2525 psp->insertLineMacro = 0;
2526 }else if( strcmp(x,"stack_size")==0 ){
2527 psp->declargslot = &(psp->gp->stacksize);
2528 psp->insertLineMacro = 0;
2529 }else if( strcmp(x,"start_symbol")==0 ){
2530 psp->declargslot = &(psp->gp->start);
2531 psp->insertLineMacro = 0;
2532 }else if( strcmp(x,"left")==0 ){
2533 psp->preccounter++;
2534 psp->declassoc = LEFT;
2535 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2536 }else if( strcmp(x,"right")==0 ){
2537 psp->preccounter++;
2538 psp->declassoc = RIGHT;
2539 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2540 }else if( strcmp(x,"nonassoc")==0 ){
2541 psp->preccounter++;
2542 psp->declassoc = NONE;
2543 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2544 }else if( strcmp(x,"destructor")==0 ){
2545 psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2546 }else if( strcmp(x,"type")==0 ){
2547 psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2548 }else if( strcmp(x,"fallback")==0 ){
2549 psp->fallback = 0;
2550 psp->state = WAITING_FOR_FALLBACK_ID;
2551 }else if( strcmp(x,"token")==0 ){
2552 psp->state = WAITING_FOR_TOKEN_NAME;
2553 }else if( strcmp(x,"wildcard")==0 ){
2554 psp->state = WAITING_FOR_WILDCARD_ID;
2555 }else if( strcmp(x,"token_class")==0 ){
2556 psp->state = WAITING_FOR_CLASS_ID;
2557 }else{
2558 ErrorMsg(psp->filename,psp->tokenlineno,
2559 "Unknown declaration keyword: \"%%%s\".",x);
2560 psp->errorcnt++;
2561 psp->state = RESYNC_AFTER_DECL_ERROR;
2562 }
2563 }else{
2564 ErrorMsg(psp->filename,psp->tokenlineno,
2565 "Illegal declaration keyword: \"%s\".",x);
2566 psp->errorcnt++;
2567 psp->state = RESYNC_AFTER_DECL_ERROR;
2568 }
2569 break;
2570 case WAITING_FOR_DESTRUCTOR_SYMBOL:
2571 if( !ISALPHA(x[0]) ){
2572 ErrorMsg(psp->filename,psp->tokenlineno,
2573 "Symbol name missing after %%destructor keyword");
2574 psp->errorcnt++;
2575 psp->state = RESYNC_AFTER_DECL_ERROR;
2576 }else{
2577 struct symbol *sp = Symbol_new(x);
2578 psp->declargslot = &sp->destructor;
2579 psp->decllinenoslot = &sp->destLineno;
2580 psp->insertLineMacro = 1;
2581 psp->state = WAITING_FOR_DECL_ARG;
2582 }
2583 break;
2584 case WAITING_FOR_DATATYPE_SYMBOL:
2585 if( !ISALPHA(x[0]) ){
2586 ErrorMsg(psp->filename,psp->tokenlineno,
2587 "Symbol name missing after %%type keyword");
2588 psp->errorcnt++;
2589 psp->state = RESYNC_AFTER_DECL_ERROR;
2590 }else{
2591 struct symbol *sp = Symbol_find(x);
2592 if((sp) && (sp->datatype)){
2593 ErrorMsg(psp->filename,psp->tokenlineno,
2594 "Symbol %%type \"%s\" already defined", x);
2595 psp->errorcnt++;
2596 psp->state = RESYNC_AFTER_DECL_ERROR;
2597 }else{
2598 if (!sp){
2599 sp = Symbol_new(x);
2600 }
2601 psp->declargslot = &sp->datatype;
2602 psp->insertLineMacro = 0;
2603 psp->state = WAITING_FOR_DECL_ARG;
2604 }
2605 }
2606 break;
2607 case WAITING_FOR_PRECEDENCE_SYMBOL:
2608 if( x[0]=='.' ){
2609 psp->state = WAITING_FOR_DECL_OR_RULE;
2610 }else if( ISUPPER(x[0]) ){
2611 struct symbol *sp;
2612 sp = Symbol_new(x);
2613 if( sp->prec>=0 ){
2614 ErrorMsg(psp->filename,psp->tokenlineno,
2615 "Symbol \"%s\" has already be given a precedence.",x);
2616 psp->errorcnt++;
2617 }else{
2618 sp->prec = psp->preccounter;
2619 sp->assoc = psp->declassoc;
2620 }
2621 }else{
2622 ErrorMsg(psp->filename,psp->tokenlineno,
2623 "Can't assign a precedence to \"%s\".",x);
2624 psp->errorcnt++;
2625 }
2626 break;
2627 case WAITING_FOR_DECL_ARG:
2628 if( x[0]=='{' || x[0]=='\"' || ISALNUM(x[0]) ){
2629 const char *zOld, *zNew;
2630 char *zBuf, *z;
2631 int nOld, n, nLine = 0, nNew, nBack;
2632 int addLineMacro;
2633 char zLine[50];
2634 zNew = x;
2635 if( zNew[0]=='"' || zNew[0]=='{' ) zNew++;
2636 nNew = lemonStrlen(zNew);
2637 if( *psp->declargslot ){
2638 zOld = *psp->declargslot;
2639 }else{
2640 zOld = "";
2641 }
2642 nOld = lemonStrlen(zOld);
2643 n = nOld + nNew + 20;
2644 addLineMacro = !psp->gp->nolinenosflag
2645 && psp->insertLineMacro
2646 && psp->tokenlineno>1
2647 && (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0);
2648 if( addLineMacro ){
2649 for(z=psp->filename, nBack=0; *z; z++){
2650 if( *z=='\\' ) nBack++;
2651 }
2652 lemon_sprintf(zLine, "#line %d ", psp->tokenlineno);
2653 nLine = lemonStrlen(zLine);
2654 n += nLine + lemonStrlen(psp->filename) + nBack;
2655 }
2656 *psp->declargslot = (char *) realloc(*psp->declargslot, n);
2657 zBuf = *psp->declargslot + nOld;
2658 if( addLineMacro ){
2659 if( nOld && zBuf[-1]!='\n' ){
2660 *(zBuf++) = '\n';
2661 }
2662 memcpy(zBuf, zLine, nLine);
2663 zBuf += nLine;
2664 *(zBuf++) = '"';
2665 for(z=psp->filename; *z; z++){
2666 if( *z=='\\' ){
2667 *(zBuf++) = '\\';
2668 }
2669 *(zBuf++) = *z;
2670 }
2671 *(zBuf++) = '"';
2672 *(zBuf++) = '\n';
2673 }
2674 if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){
2675 psp->decllinenoslot[0] = psp->tokenlineno;
2676 }
2677 memcpy(zBuf, zNew, nNew);
2678 zBuf += nNew;
2679 *zBuf = 0;
2680 psp->state = WAITING_FOR_DECL_OR_RULE;
2681 }else{
2682 ErrorMsg(psp->filename,psp->tokenlineno,
2683 "Illegal argument to %%%s: %s",psp->declkeyword,x);
2684 psp->errorcnt++;
2685 psp->state = RESYNC_AFTER_DECL_ERROR;
2686 }
2687 break;
2688 case WAITING_FOR_FALLBACK_ID:
2689 if( x[0]=='.' ){
2690 psp->state = WAITING_FOR_DECL_OR_RULE;
2691 }else if( !ISUPPER(x[0]) ){
2692 ErrorMsg(psp->filename, psp->tokenlineno,
2693 "%%fallback argument \"%s\" should be a token", x);
2694 psp->errorcnt++;
2695 }else{
2696 struct symbol *sp = Symbol_new(x);
2697 if( psp->fallback==0 ){
2698 psp->fallback = sp;
2699 }else if( sp->fallback ){
2700 ErrorMsg(psp->filename, psp->tokenlineno,
2701 "More than one fallback assigned to token %s", x);
2702 psp->errorcnt++;
2703 }else{
2704 sp->fallback = psp->fallback;
2705 psp->gp->has_fallback = 1;
2706 }
2707 }
2708 break;
2709 case WAITING_FOR_TOKEN_NAME:
2710 /* Tokens do not have to be declared before use. But they can be
2711 ** in order to control their assigned integer number. The number for
2712 ** each token is assigned when it is first seen. So by including
2713 **
2714 ** %token ONE TWO THREE.
2715 **
2716 ** early in the grammar file, that assigns small consecutive values
2717 ** to each of the tokens ONE TWO and THREE.
2718 */
2719 if( x[0]=='.' ){
2720 psp->state = WAITING_FOR_DECL_OR_RULE;
2721 }else if( !ISUPPER(x[0]) ){
2722 ErrorMsg(psp->filename, psp->tokenlineno,
2723 "%%token argument \"%s\" should be a token", x);
2724 psp->errorcnt++;
2725 }else{
2726 (void)Symbol_new(x);
2727 }
2728 break;
2729 case WAITING_FOR_WILDCARD_ID:
2730 if( x[0]=='.' ){
2731 psp->state = WAITING_FOR_DECL_OR_RULE;
2732 }else if( !ISUPPER(x[0]) ){
2733 ErrorMsg(psp->filename, psp->tokenlineno,
2734 "%%wildcard argument \"%s\" should be a token", x);
2735 psp->errorcnt++;
2736 }else{
2737 struct symbol *sp = Symbol_new(x);
2738 if( psp->gp->wildcard==0 ){
2739 psp->gp->wildcard = sp;
2740 }else{
2741 ErrorMsg(psp->filename, psp->tokenlineno,
2742 "Extra wildcard to token: %s", x);
2743 psp->errorcnt++;
2744 }
2745 }
2746 break;
2747 case WAITING_FOR_CLASS_ID:
2748 if( !ISLOWER(x[0]) ){
2749 ErrorMsg(psp->filename, psp->tokenlineno,
2750 "%%token_class must be followed by an identifier: %s", x);
2751 psp->errorcnt++;
2752 psp->state = RESYNC_AFTER_DECL_ERROR;
2753 }else if( Symbol_find(x) ){
2754 ErrorMsg(psp->filename, psp->tokenlineno,
2755 "Symbol \"%s\" already used", x);
2756 psp->errorcnt++;
2757 psp->state = RESYNC_AFTER_DECL_ERROR;
2758 }else{
2759 psp->tkclass = Symbol_new(x);
2760 psp->tkclass->type = MULTITERMINAL;
2761 psp->state = WAITING_FOR_CLASS_TOKEN;
2762 }
2763 break;
2764 case WAITING_FOR_CLASS_TOKEN:
2765 if( x[0]=='.' ){
2766 psp->state = WAITING_FOR_DECL_OR_RULE;
2767 }else if( ISUPPER(x[0]) || ((x[0]=='|' || x[0]=='/') && ISUPPER(x[1])) ){
2768 struct symbol *msp = psp->tkclass;
2769 msp->nsubsym++;
2770 msp->subsym = (struct symbol **) realloc(msp->subsym,
2771 sizeof(struct symbol*)*msp->nsubsym);
2772 if( !ISUPPER(x[0]) ) x++;
2773 msp->subsym[msp->nsubsym-1] = Symbol_new(x);
2774 }else{
2775 ErrorMsg(psp->filename, psp->tokenlineno,
2776 "%%token_class argument \"%s\" should be a token", x);
2777 psp->errorcnt++;
2778 psp->state = RESYNC_AFTER_DECL_ERROR;
2779 }
2780 break;
2781 case RESYNC_AFTER_RULE_ERROR:
2782 /* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2783 ** break; */
2784 case RESYNC_AFTER_DECL_ERROR:
2785 if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2786 if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2787 break;
2788 }
2789 }
2790
2791 /* The text in the input is part of the argument to an %ifdef or %ifndef.
2792 ** Evaluate the text as a boolean expression. Return true or false.
2793 */
eval_preprocessor_boolean(char * z,int lineno)2794 static int eval_preprocessor_boolean(char *z, int lineno){
2795 int neg = 0;
2796 int res = 0;
2797 int okTerm = 1;
2798 int i;
2799 for(i=0; z[i]!=0; i++){
2800 if( ISSPACE(z[i]) ) continue;
2801 if( z[i]=='!' ){
2802 if( !okTerm ) goto pp_syntax_error;
2803 neg = !neg;
2804 continue;
2805 }
2806 if( z[i]=='|' && z[i+1]=='|' ){
2807 if( okTerm ) goto pp_syntax_error;
2808 if( res ) return 1;
2809 i++;
2810 okTerm = 1;
2811 continue;
2812 }
2813 if( z[i]=='&' && z[i+1]=='&' ){
2814 if( okTerm ) goto pp_syntax_error;
2815 if( !res ) return 0;
2816 i++;
2817 okTerm = 1;
2818 continue;
2819 }
2820 if( z[i]=='(' ){
2821 int k;
2822 int n = 1;
2823 if( !okTerm ) goto pp_syntax_error;
2824 for(k=i+1; z[k]; k++){
2825 if( z[k]==')' ){
2826 n--;
2827 if( n==0 ){
2828 z[k] = 0;
2829 res = eval_preprocessor_boolean(&z[i+1], -1);
2830 z[k] = ')';
2831 if( res<0 ){
2832 i = i-res;
2833 goto pp_syntax_error;
2834 }
2835 i = k;
2836 break;
2837 }
2838 }else if( z[k]=='(' ){
2839 n++;
2840 }else if( z[k]==0 ){
2841 i = k;
2842 goto pp_syntax_error;
2843 }
2844 }
2845 if( neg ){
2846 res = !res;
2847 neg = 0;
2848 }
2849 okTerm = 0;
2850 continue;
2851 }
2852 if( ISALPHA(z[i]) ){
2853 int j, k, n;
2854 if( !okTerm ) goto pp_syntax_error;
2855 for(k=i+1; ISALNUM(z[k]) || z[k]=='_'; k++){}
2856 n = k - i;
2857 res = 0;
2858 for(j=0; j<nDefine; j++){
2859 if( strncmp(azDefine[j],&z[i],n)==0 && azDefine[j][n]==0 ){
2860 res = 1;
2861 break;
2862 }
2863 }
2864 i = k-1;
2865 if( neg ){
2866 res = !res;
2867 neg = 0;
2868 }
2869 okTerm = 0;
2870 continue;
2871 }
2872 goto pp_syntax_error;
2873 }
2874 return res;
2875
2876 pp_syntax_error:
2877 if( lineno>0 ){
2878 fprintf(stderr, "%%if syntax error on line %d.\n", lineno);
2879 fprintf(stderr, " %.*s <-- syntax error here\n", i+1, z);
2880 exit(1);
2881 }else{
2882 return -(i+1);
2883 }
2884 }
2885
2886 /* Run the preprocessor over the input file text. The global variables
2887 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2888 ** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2889 ** comments them out. Text in between is also commented out as appropriate.
2890 */
preprocess_input(char * z)2891 static void preprocess_input(char *z){
2892 int i, j, k;
2893 int exclude = 0;
2894 int start = 0;
2895 int lineno = 1;
2896 int start_lineno = 1;
2897 for(i=0; z[i]; i++){
2898 if( z[i]=='\n' ) lineno++;
2899 if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue;
2900 if( strncmp(&z[i],"%endif",6)==0 && ISSPACE(z[i+6]) ){
2901 if( exclude ){
2902 exclude--;
2903 if( exclude==0 ){
2904 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2905 }
2906 }
2907 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2908 }else if( strncmp(&z[i],"%else",5)==0 && ISSPACE(z[i+5]) ){
2909 if( exclude==1){
2910 exclude = 0;
2911 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2912 }else if( exclude==0 ){
2913 exclude = 1;
2914 start = i;
2915 start_lineno = lineno;
2916 }
2917 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2918 }else if( strncmp(&z[i],"%ifdef ",7)==0
2919 || strncmp(&z[i],"%if ",4)==0
2920 || strncmp(&z[i],"%ifndef ",8)==0 ){
2921 if( exclude ){
2922 exclude++;
2923 }else{
2924 int isNot;
2925 int iBool;
2926 for(j=i; z[j] && !ISSPACE(z[j]); j++){}
2927 iBool = j;
2928 isNot = (j==i+7);
2929 while( z[j] && z[j]!='\n' ){ j++; }
2930 k = z[j];
2931 z[j] = 0;
2932 exclude = eval_preprocessor_boolean(&z[iBool], lineno);
2933 z[j] = k;
2934 if( !isNot ) exclude = !exclude;
2935 if( exclude ){
2936 start = i;
2937 start_lineno = lineno;
2938 }
2939 }
2940 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2941 }
2942 }
2943 if( exclude ){
2944 fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno);
2945 exit(1);
2946 }
2947 }
2948
2949 /* In spite of its name, this function is really a scanner. It read
2950 ** in the entire input file (all at once) then tokenizes it. Each
2951 ** token is passed to the function "parseonetoken" which builds all
2952 ** the appropriate data structures in the global state vector "gp".
2953 */
Parse(struct lemon * gp)2954 void Parse(struct lemon *gp)
2955 {
2956 struct pstate ps;
2957 FILE *fp;
2958 char *filebuf;
2959 unsigned int filesize;
2960 int lineno;
2961 int c;
2962 char *cp, *nextcp;
2963 int startline = 0;
2964
2965 memset(&ps, '\0', sizeof(ps));
2966 ps.gp = gp;
2967 ps.filename = gp->filename;
2968 ps.errorcnt = 0;
2969 ps.state = INITIALIZE;
2970
2971 /* Begin by reading the input file */
2972 fp = fopen(ps.filename,"rb");
2973 if( fp==0 ){
2974 ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2975 gp->errorcnt++;
2976 return;
2977 }
2978 fseek(fp,0,2);
2979 filesize = ftell(fp);
2980 rewind(fp);
2981 filebuf = (char *)malloc( filesize+1 );
2982 if( filesize>100000000 || filebuf==0 ){
2983 ErrorMsg(ps.filename,0,"Input file too large.");
2984 free(filebuf);
2985 gp->errorcnt++;
2986 fclose(fp);
2987 return;
2988 }
2989 if( fread(filebuf,1,filesize,fp)!=filesize ){
2990 ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2991 filesize);
2992 free(filebuf);
2993 gp->errorcnt++;
2994 fclose(fp);
2995 return;
2996 }
2997 fclose(fp);
2998 filebuf[filesize] = 0;
2999
3000 /* Make an initial pass through the file to handle %ifdef and %ifndef */
3001 preprocess_input(filebuf);
3002 if( gp->printPreprocessed ){
3003 printf("%s\n", filebuf);
3004 return;
3005 }
3006
3007 /* Now scan the text of the input file */
3008 lineno = 1;
3009 for(cp=filebuf; (c= *cp)!=0; ){
3010 if( c=='\n' ) lineno++; /* Keep track of the line number */
3011 if( ISSPACE(c) ){ cp++; continue; } /* Skip all white space */
3012 if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */
3013 cp+=2;
3014 while( (c= *cp)!=0 && c!='\n' ) cp++;
3015 continue;
3016 }
3017 if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */
3018 cp+=2;
3019 if( (*cp)=='/' ) cp++;
3020 while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
3021 if( c=='\n' ) lineno++;
3022 cp++;
3023 }
3024 if( c ) cp++;
3025 continue;
3026 }
3027 ps.tokenstart = cp; /* Mark the beginning of the token */
3028 ps.tokenlineno = lineno; /* Linenumber on which token begins */
3029 if( c=='\"' ){ /* String literals */
3030 cp++;
3031 while( (c= *cp)!=0 && c!='\"' ){
3032 if( c=='\n' ) lineno++;
3033 cp++;
3034 }
3035 if( c==0 ){
3036 ErrorMsg(ps.filename,startline,
3037 "String starting on this line is not terminated before "
3038 "the end of the file.");
3039 ps.errorcnt++;
3040 nextcp = cp;
3041 }else{
3042 nextcp = cp+1;
3043 }
3044 }else if( c=='{' ){ /* A block of C code */
3045 int level;
3046 cp++;
3047 for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
3048 if( c=='\n' ) lineno++;
3049 else if( c=='{' ) level++;
3050 else if( c=='}' ) level--;
3051 else if( c=='/' && cp[1]=='*' ){ /* Skip comments */
3052 int prevc;
3053 cp = &cp[2];
3054 prevc = 0;
3055 while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
3056 if( c=='\n' ) lineno++;
3057 prevc = c;
3058 cp++;
3059 }
3060 }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */
3061 cp = &cp[2];
3062 while( (c= *cp)!=0 && c!='\n' ) cp++;
3063 if( c ) lineno++;
3064 }else if( c=='\'' || c=='\"' ){ /* String a character literals */
3065 int startchar, prevc;
3066 startchar = c;
3067 prevc = 0;
3068 for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
3069 if( c=='\n' ) lineno++;
3070 if( prevc=='\\' ) prevc = 0;
3071 else prevc = c;
3072 }
3073 }
3074 }
3075 if( c==0 ){
3076 ErrorMsg(ps.filename,ps.tokenlineno,
3077 "C code starting on this line is not terminated before "
3078 "the end of the file.");
3079 ps.errorcnt++;
3080 nextcp = cp;
3081 }else{
3082 nextcp = cp+1;
3083 }
3084 }else if( ISALNUM(c) ){ /* Identifiers */
3085 while( (c= *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
3086 nextcp = cp;
3087 }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
3088 cp += 3;
3089 nextcp = cp;
3090 }else if( (c=='/' || c=='|') && ISALPHA(cp[1]) ){
3091 cp += 2;
3092 while( (c = *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
3093 nextcp = cp;
3094 }else{ /* All other (one character) operators */
3095 cp++;
3096 nextcp = cp;
3097 }
3098 c = *cp;
3099 *cp = 0; /* Null terminate the token */
3100 parseonetoken(&ps); /* Parse the token */
3101 *cp = (char)c; /* Restore the buffer */
3102 cp = nextcp;
3103 }
3104 free(filebuf); /* Release the buffer after parsing */
3105 gp->rule = ps.firstrule;
3106 gp->errorcnt = ps.errorcnt;
3107 }
3108 /*************************** From the file "plink.c" *********************/
3109 /*
3110 ** Routines processing configuration follow-set propagation links
3111 ** in the LEMON parser generator.
3112 */
3113 static struct plink *plink_freelist = 0;
3114
3115 /* Allocate a new plink */
Plink_new(void)3116 struct plink *Plink_new(void){
3117 struct plink *newlink;
3118
3119 if( plink_freelist==0 ){
3120 int i;
3121 int amt = 100;
3122 plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) );
3123 if( plink_freelist==0 ){
3124 fprintf(stderr,
3125 "Unable to allocate memory for a new follow-set propagation link.\n");
3126 exit(1);
3127 }
3128 for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
3129 plink_freelist[amt-1].next = 0;
3130 }
3131 newlink = plink_freelist;
3132 plink_freelist = plink_freelist->next;
3133 return newlink;
3134 }
3135
3136 /* Add a plink to a plink list */
Plink_add(struct plink ** plpp,struct config * cfp)3137 void Plink_add(struct plink **plpp, struct config *cfp)
3138 {
3139 struct plink *newlink;
3140 newlink = Plink_new();
3141 newlink->next = *plpp;
3142 *plpp = newlink;
3143 newlink->cfp = cfp;
3144 }
3145
3146 /* Transfer every plink on the list "from" to the list "to" */
Plink_copy(struct plink ** to,struct plink * from)3147 void Plink_copy(struct plink **to, struct plink *from)
3148 {
3149 struct plink *nextpl;
3150 while( from ){
3151 nextpl = from->next;
3152 from->next = *to;
3153 *to = from;
3154 from = nextpl;
3155 }
3156 }
3157
3158 /* Delete every plink on the list */
Plink_delete(struct plink * plp)3159 void Plink_delete(struct plink *plp)
3160 {
3161 struct plink *nextpl;
3162
3163 while( plp ){
3164 nextpl = plp->next;
3165 plp->next = plink_freelist;
3166 plink_freelist = plp;
3167 plp = nextpl;
3168 }
3169 }
3170 /*********************** From the file "report.c" **************************/
3171 /*
3172 ** Procedures for generating reports and tables in the LEMON parser generator.
3173 */
3174
3175 /* Generate a filename with the given suffix. Space to hold the
3176 ** name comes from malloc() and must be freed by the calling
3177 ** function.
3178 */
file_makename(struct lemon * lemp,const char * suffix)3179 PRIVATE char *file_makename(struct lemon *lemp, const char *suffix)
3180 {
3181 char *name;
3182 char *cp;
3183 char *filename = lemp->filename;
3184 int sz;
3185
3186 if( outputDir ){
3187 cp = strrchr(filename, '/');
3188 if( cp ) filename = cp + 1;
3189 }
3190 sz = lemonStrlen(filename);
3191 sz += lemonStrlen(suffix);
3192 if( outputDir ) sz += lemonStrlen(outputDir) + 1;
3193 sz += 5;
3194 name = (char*)malloc( sz );
3195 if( name==0 ){
3196 fprintf(stderr,"Can't allocate space for a filename.\n");
3197 exit(1);
3198 }
3199 name[0] = 0;
3200 if( outputDir ){
3201 lemon_strcpy(name, outputDir);
3202 lemon_strcat(name, "/");
3203 }
3204 lemon_strcat(name,filename);
3205 cp = strrchr(name,'.');
3206 if( cp ) *cp = 0;
3207 lemon_strcat(name,suffix);
3208 return name;
3209 }
3210
3211 /* Open a file with a name based on the name of the input file,
3212 ** but with a different (specified) suffix, and return a pointer
3213 ** to the stream */
file_open(struct lemon * lemp,const char * suffix,const char * mode)3214 PRIVATE FILE *file_open(
3215 struct lemon *lemp,
3216 const char *suffix,
3217 const char *mode
3218 ){
3219 FILE *fp;
3220
3221 if( lemp->outname ) free(lemp->outname);
3222 lemp->outname = file_makename(lemp, suffix);
3223 fp = fopen(lemp->outname,mode);
3224 if( fp==0 && *mode=='w' ){
3225 fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
3226 lemp->errorcnt++;
3227 return 0;
3228 }
3229 return fp;
3230 }
3231
3232 /* Print the text of a rule
3233 */
rule_print(FILE * out,struct rule * rp)3234 void rule_print(FILE *out, struct rule *rp){
3235 int i, j;
3236 fprintf(out, "%s",rp->lhs->name);
3237 /* if( rp->lhsalias ) fprintf(out,"(%s)",rp->lhsalias); */
3238 fprintf(out," ::=");
3239 for(i=0; i<rp->nrhs; i++){
3240 struct symbol *sp = rp->rhs[i];
3241 if( sp->type==MULTITERMINAL ){
3242 fprintf(out," %s", sp->subsym[0]->name);
3243 for(j=1; j<sp->nsubsym; j++){
3244 fprintf(out,"|%s", sp->subsym[j]->name);
3245 }
3246 }else{
3247 fprintf(out," %s", sp->name);
3248 }
3249 /* if( rp->rhsalias[i] ) fprintf(out,"(%s)",rp->rhsalias[i]); */
3250 }
3251 }
3252
3253 /* Duplicate the input file without comments and without actions
3254 ** on rules */
Reprint(struct lemon * lemp)3255 void Reprint(struct lemon *lemp)
3256 {
3257 struct rule *rp;
3258 struct symbol *sp;
3259 int i, j, maxlen, len, ncolumns, skip;
3260 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
3261 maxlen = 10;
3262 for(i=0; i<lemp->nsymbol; i++){
3263 sp = lemp->symbols[i];
3264 len = lemonStrlen(sp->name);
3265 if( len>maxlen ) maxlen = len;
3266 }
3267 ncolumns = 76/(maxlen+5);
3268 if( ncolumns<1 ) ncolumns = 1;
3269 skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
3270 for(i=0; i<skip; i++){
3271 printf("//");
3272 for(j=i; j<lemp->nsymbol; j+=skip){
3273 sp = lemp->symbols[j];
3274 assert( sp->index==j );
3275 printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
3276 }
3277 printf("\n");
3278 }
3279 for(rp=lemp->rule; rp; rp=rp->next){
3280 rule_print(stdout, rp);
3281 printf(".");
3282 if( rp->precsym ) printf(" [%s]",rp->precsym->name);
3283 /* if( rp->code ) printf("\n %s",rp->code); */
3284 printf("\n");
3285 }
3286 }
3287
3288 /* Print a single rule.
3289 */
RulePrint(FILE * fp,struct rule * rp,int iCursor)3290 void RulePrint(FILE *fp, struct rule *rp, int iCursor){
3291 struct symbol *sp;
3292 int i, j;
3293 fprintf(fp,"%s ::=",rp->lhs->name);
3294 for(i=0; i<=rp->nrhs; i++){
3295 if( i==iCursor ) fprintf(fp," *");
3296 if( i==rp->nrhs ) break;
3297 sp = rp->rhs[i];
3298 if( sp->type==MULTITERMINAL ){
3299 fprintf(fp," %s", sp->subsym[0]->name);
3300 for(j=1; j<sp->nsubsym; j++){
3301 fprintf(fp,"|%s",sp->subsym[j]->name);
3302 }
3303 }else{
3304 fprintf(fp," %s", sp->name);
3305 }
3306 }
3307 }
3308
3309 /* Print the rule for a configuration.
3310 */
ConfigPrint(FILE * fp,struct config * cfp)3311 void ConfigPrint(FILE *fp, struct config *cfp){
3312 RulePrint(fp, cfp->rp, cfp->dot);
3313 }
3314
3315 /* #define TEST */
3316 #if 0
3317 /* Print a set */
3318 PRIVATE void SetPrint(out,set,lemp)
3319 FILE *out;
3320 char *set;
3321 struct lemon *lemp;
3322 {
3323 int i;
3324 char *spacer;
3325 spacer = "";
3326 fprintf(out,"%12s[","");
3327 for(i=0; i<lemp->nterminal; i++){
3328 if( SetFind(set,i) ){
3329 fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
3330 spacer = " ";
3331 }
3332 }
3333 fprintf(out,"]\n");
3334 }
3335
3336 /* Print a plink chain */
3337 PRIVATE void PlinkPrint(out,plp,tag)
3338 FILE *out;
3339 struct plink *plp;
3340 char *tag;
3341 {
3342 while( plp ){
3343 fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum);
3344 ConfigPrint(out,plp->cfp);
3345 fprintf(out,"\n");
3346 plp = plp->next;
3347 }
3348 }
3349 #endif
3350
3351 /* Print an action to the given file descriptor. Return FALSE if
3352 ** nothing was actually printed.
3353 */
PrintAction(struct action * ap,FILE * fp,int indent)3354 int PrintAction(
3355 struct action *ap, /* The action to print */
3356 FILE *fp, /* Print the action here */
3357 int indent /* Indent by this amount */
3358 ){
3359 int result = 1;
3360 switch( ap->type ){
3361 case SHIFT: {
3362 struct state *stp = ap->x.stp;
3363 fprintf(fp,"%*s shift %-7d",indent,ap->sp->name,stp->statenum);
3364 break;
3365 }
3366 case REDUCE: {
3367 struct rule *rp = ap->x.rp;
3368 fprintf(fp,"%*s reduce %-7d",indent,ap->sp->name,rp->iRule);
3369 RulePrint(fp, rp, -1);
3370 break;
3371 }
3372 case SHIFTREDUCE: {
3373 struct rule *rp = ap->x.rp;
3374 fprintf(fp,"%*s shift-reduce %-7d",indent,ap->sp->name,rp->iRule);
3375 RulePrint(fp, rp, -1);
3376 break;
3377 }
3378 case ACCEPT:
3379 fprintf(fp,"%*s accept",indent,ap->sp->name);
3380 break;
3381 case ERROR:
3382 fprintf(fp,"%*s error",indent,ap->sp->name);
3383 break;
3384 case SRCONFLICT:
3385 case RRCONFLICT:
3386 fprintf(fp,"%*s reduce %-7d ** Parsing conflict **",
3387 indent,ap->sp->name,ap->x.rp->iRule);
3388 break;
3389 case SSCONFLICT:
3390 fprintf(fp,"%*s shift %-7d ** Parsing conflict **",
3391 indent,ap->sp->name,ap->x.stp->statenum);
3392 break;
3393 case SH_RESOLVED:
3394 if( showPrecedenceConflict ){
3395 fprintf(fp,"%*s shift %-7d -- dropped by precedence",
3396 indent,ap->sp->name,ap->x.stp->statenum);
3397 }else{
3398 result = 0;
3399 }
3400 break;
3401 case RD_RESOLVED:
3402 if( showPrecedenceConflict ){
3403 fprintf(fp,"%*s reduce %-7d -- dropped by precedence",
3404 indent,ap->sp->name,ap->x.rp->iRule);
3405 }else{
3406 result = 0;
3407 }
3408 break;
3409 case NOT_USED:
3410 result = 0;
3411 break;
3412 }
3413 if( result && ap->spOpt ){
3414 fprintf(fp," /* because %s==%s */", ap->sp->name, ap->spOpt->name);
3415 }
3416 return result;
3417 }
3418
3419 /* Generate the "*.out" log file */
ReportOutput(struct lemon * lemp)3420 void ReportOutput(struct lemon *lemp)
3421 {
3422 int i, n;
3423 struct state *stp;
3424 struct config *cfp;
3425 struct action *ap;
3426 struct rule *rp;
3427 FILE *fp;
3428
3429 fp = file_open(lemp,".out","wb");
3430 if( fp==0 ) return;
3431 for(i=0; i<lemp->nxstate; i++){
3432 stp = lemp->sorted[i];
3433 fprintf(fp,"State %d:\n",stp->statenum);
3434 if( lemp->basisflag ) cfp=stp->bp;
3435 else cfp=stp->cfp;
3436 while( cfp ){
3437 char buf[20];
3438 if( cfp->dot==cfp->rp->nrhs ){
3439 lemon_sprintf(buf,"(%d)",cfp->rp->iRule);
3440 fprintf(fp," %5s ",buf);
3441 }else{
3442 fprintf(fp," ");
3443 }
3444 ConfigPrint(fp,cfp);
3445 fprintf(fp,"\n");
3446 #if 0
3447 SetPrint(fp,cfp->fws,lemp);
3448 PlinkPrint(fp,cfp->fplp,"To ");
3449 PlinkPrint(fp,cfp->bplp,"From");
3450 #endif
3451 if( lemp->basisflag ) cfp=cfp->bp;
3452 else cfp=cfp->next;
3453 }
3454 fprintf(fp,"\n");
3455 for(ap=stp->ap; ap; ap=ap->next){
3456 if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
3457 }
3458 fprintf(fp,"\n");
3459 }
3460 fprintf(fp, "----------------------------------------------------\n");
3461 fprintf(fp, "Symbols:\n");
3462 fprintf(fp, "The first-set of non-terminals is shown after the name.\n\n");
3463 for(i=0; i<lemp->nsymbol; i++){
3464 int j;
3465 struct symbol *sp;
3466
3467 sp = lemp->symbols[i];
3468 fprintf(fp, " %3d: %s", i, sp->name);
3469 if( sp->type==NONTERMINAL ){
3470 fprintf(fp, ":");
3471 if( sp->lambda ){
3472 fprintf(fp, " <lambda>");
3473 }
3474 for(j=0; j<lemp->nterminal; j++){
3475 if( sp->firstset && SetFind(sp->firstset, j) ){
3476 fprintf(fp, " %s", lemp->symbols[j]->name);
3477 }
3478 }
3479 }
3480 if( sp->prec>=0 ) fprintf(fp," (precedence=%d)", sp->prec);
3481 fprintf(fp, "\n");
3482 }
3483 fprintf(fp, "----------------------------------------------------\n");
3484 fprintf(fp, "Syntax-only Symbols:\n");
3485 fprintf(fp, "The following symbols never carry semantic content.\n\n");
3486 for(i=n=0; i<lemp->nsymbol; i++){
3487 int w;
3488 struct symbol *sp = lemp->symbols[i];
3489 if( sp->bContent ) continue;
3490 w = (int)strlen(sp->name);
3491 if( n>0 && n+w>75 ){
3492 fprintf(fp,"\n");
3493 n = 0;
3494 }
3495 if( n>0 ){
3496 fprintf(fp, " ");
3497 n++;
3498 }
3499 fprintf(fp, "%s", sp->name);
3500 n += w;
3501 }
3502 if( n>0 ) fprintf(fp, "\n");
3503 fprintf(fp, "----------------------------------------------------\n");
3504 fprintf(fp, "Rules:\n");
3505 for(rp=lemp->rule; rp; rp=rp->next){
3506 fprintf(fp, "%4d: ", rp->iRule);
3507 rule_print(fp, rp);
3508 fprintf(fp,".");
3509 if( rp->precsym ){
3510 fprintf(fp," [%s precedence=%d]",
3511 rp->precsym->name, rp->precsym->prec);
3512 }
3513 fprintf(fp,"\n");
3514 }
3515 fclose(fp);
3516 return;
3517 }
3518
3519 /* Search for the file "name" which is in the same directory as
3520 ** the executable */
pathsearch(char * argv0,char * name,int modemask)3521 PRIVATE char *pathsearch(char *argv0, char *name, int modemask)
3522 {
3523 const char *pathlist;
3524 char *pathbufptr = 0;
3525 char *pathbuf = 0;
3526 char *path,*cp;
3527 char c;
3528
3529 #ifdef __WIN32__
3530 cp = strrchr(argv0,'\\');
3531 #else
3532 cp = strrchr(argv0,'/');
3533 #endif
3534 if( cp ){
3535 c = *cp;
3536 *cp = 0;
3537 path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 );
3538 if( path ) lemon_sprintf(path,"%s/%s",argv0,name);
3539 *cp = c;
3540 }else{
3541 pathlist = getenv("PATH");
3542 if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
3543 pathbuf = pathbufptr = (char *) malloc( lemonStrlen(pathlist) + 1 );
3544 path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 );
3545 if( (pathbuf != 0) && (path!=0) ){
3546 lemon_strcpy(pathbuf, pathlist);
3547 while( *pathbuf ){
3548 cp = strchr(pathbuf,':');
3549 if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)];
3550 c = *cp;
3551 *cp = 0;
3552 lemon_sprintf(path,"%s/%s",pathbuf,name);
3553 *cp = c;
3554 if( c==0 ) pathbuf[0] = 0;
3555 else pathbuf = &cp[1];
3556 if( access(path,modemask)==0 ) break;
3557 }
3558 }
3559 free(pathbufptr);
3560 }
3561 return path;
3562 }
3563
3564 /* Given an action, compute the integer value for that action
3565 ** which is to be put in the action table of the generated machine.
3566 ** Return negative if no action should be generated.
3567 */
compute_action(struct lemon * lemp,struct action * ap)3568 PRIVATE int compute_action(struct lemon *lemp, struct action *ap)
3569 {
3570 int act;
3571 switch( ap->type ){
3572 case SHIFT: act = ap->x.stp->statenum; break;
3573 case SHIFTREDUCE: {
3574 /* Since a SHIFT is inherient after a prior REDUCE, convert any
3575 ** SHIFTREDUCE action with a nonterminal on the LHS into a simple
3576 ** REDUCE action: */
3577 if( ap->sp->index>=lemp->nterminal
3578 && (lemp->errsym==0 || ap->sp->index!=lemp->errsym->index)
3579 ){
3580 act = lemp->minReduce + ap->x.rp->iRule;
3581 }else{
3582 act = lemp->minShiftReduce + ap->x.rp->iRule;
3583 }
3584 break;
3585 }
3586 case REDUCE: act = lemp->minReduce + ap->x.rp->iRule; break;
3587 case ERROR: act = lemp->errAction; break;
3588 case ACCEPT: act = lemp->accAction; break;
3589 default: act = -1; break;
3590 }
3591 return act;
3592 }
3593
3594 #define LINESIZE 1000
3595 /* The next cluster of routines are for reading the template file
3596 ** and writing the results to the generated parser */
3597 /* The first function transfers data from "in" to "out" until
3598 ** a line is seen which begins with "%%". The line number is
3599 ** tracked.
3600 **
3601 ** if name!=0, then any word that begin with "Parse" is changed to
3602 ** begin with *name instead.
3603 */
tplt_xfer(char * name,FILE * in,FILE * out,int * lineno)3604 PRIVATE void tplt_xfer(char *name, FILE *in, FILE *out, int *lineno)
3605 {
3606 int i, iStart;
3607 char line[LINESIZE];
3608 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3609 (*lineno)++;
3610 iStart = 0;
3611 if( name ){
3612 for(i=0; line[i]; i++){
3613 if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
3614 && (i==0 || !ISALPHA(line[i-1]))
3615 ){
3616 if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
3617 fprintf(out,"%s",name);
3618 i += 4;
3619 iStart = i+1;
3620 }
3621 }
3622 }
3623 fprintf(out,"%s",&line[iStart]);
3624 }
3625 }
3626
3627 /* Skip forward past the header of the template file to the first "%%"
3628 */
tplt_skip_header(FILE * in,int * lineno)3629 PRIVATE void tplt_skip_header(FILE *in, int *lineno)
3630 {
3631 char line[LINESIZE];
3632 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3633 (*lineno)++;
3634 }
3635 }
3636
3637 /* The next function finds the template file and opens it, returning
3638 ** a pointer to the opened file. */
tplt_open(struct lemon * lemp)3639 PRIVATE FILE *tplt_open(struct lemon *lemp)
3640 {
3641 static char templatename[] = "lempar.c";
3642 char buf[1000];
3643 FILE *in;
3644 char *tpltname;
3645 char *toFree = 0;
3646 char *cp;
3647
3648 /* first, see if user specified a template filename on the command line. */
3649 if (user_templatename != 0) {
3650 if( access(user_templatename,004)==-1 ){
3651 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3652 user_templatename);
3653 lemp->errorcnt++;
3654 return 0;
3655 }
3656 in = fopen(user_templatename,"rb");
3657 if( in==0 ){
3658 fprintf(stderr,"Can't open the template file \"%s\".\n",
3659 user_templatename);
3660 lemp->errorcnt++;
3661 return 0;
3662 }
3663 return in;
3664 }
3665
3666 cp = strrchr(lemp->filename,'.');
3667 if( cp ){
3668 lemon_sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
3669 }else{
3670 lemon_sprintf(buf,"%s.lt",lemp->filename);
3671 }
3672 if( access(buf,004)==0 ){
3673 tpltname = buf;
3674 }else if( access(templatename,004)==0 ){
3675 tpltname = templatename;
3676 }else{
3677 toFree = tpltname = pathsearch(lemp->argv0,templatename,0);
3678 }
3679 if( tpltname==0 ){
3680 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3681 templatename);
3682 lemp->errorcnt++;
3683 return 0;
3684 }
3685 in = fopen(tpltname,"rb");
3686 if( in==0 ){
3687 fprintf(stderr,"Can't open the template file \"%s\".\n",tpltname);
3688 lemp->errorcnt++;
3689 }
3690 free(toFree);
3691 return in;
3692 }
3693
3694 /* Print a #line directive line to the output file. */
tplt_linedir(FILE * out,int lineno,char * filename)3695 PRIVATE void tplt_linedir(FILE *out, int lineno, char *filename)
3696 {
3697 fprintf(out,"#line %d \"",lineno);
3698 while( *filename ){
3699 if( *filename == '\\' ) putc('\\',out);
3700 putc(*filename,out);
3701 filename++;
3702 }
3703 fprintf(out,"\"\n");
3704 }
3705
3706 /* Print a string to the file and keep the linenumber up to date */
tplt_print(FILE * out,struct lemon * lemp,char * str,int * lineno)3707 PRIVATE void tplt_print(FILE *out, struct lemon *lemp, char *str, int *lineno)
3708 {
3709 if( str==0 ) return;
3710 while( *str ){
3711 putc(*str,out);
3712 if( *str=='\n' ) (*lineno)++;
3713 str++;
3714 }
3715 if( str[-1]!='\n' ){
3716 putc('\n',out);
3717 (*lineno)++;
3718 }
3719 if (!lemp->nolinenosflag) {
3720 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3721 }
3722 return;
3723 }
3724
3725 /*
3726 ** The following routine emits code for the destructor for the
3727 ** symbol sp
3728 */
emit_destructor_code(FILE * out,struct symbol * sp,struct lemon * lemp,int * lineno)3729 void emit_destructor_code(
3730 FILE *out,
3731 struct symbol *sp,
3732 struct lemon *lemp,
3733 int *lineno
3734 ){
3735 char *cp = 0;
3736
3737 if( sp->type==TERMINAL ){
3738 cp = lemp->tokendest;
3739 if( cp==0 ) return;
3740 fprintf(out,"{\n"); (*lineno)++;
3741 }else if( sp->destructor ){
3742 cp = sp->destructor;
3743 fprintf(out,"{\n"); (*lineno)++;
3744 if( !lemp->nolinenosflag ){
3745 (*lineno)++;
3746 tplt_linedir(out,sp->destLineno,lemp->filename);
3747 }
3748 }else if( lemp->vardest ){
3749 cp = lemp->vardest;
3750 if( cp==0 ) return;
3751 fprintf(out,"{\n"); (*lineno)++;
3752 }else{
3753 assert( 0 ); /* Cannot happen */
3754 }
3755 for(; *cp; cp++){
3756 if( *cp=='$' && cp[1]=='$' ){
3757 fprintf(out,"(yypminor->yy%d)",sp->dtnum);
3758 cp++;
3759 continue;
3760 }
3761 if( *cp=='\n' ) (*lineno)++;
3762 fputc(*cp,out);
3763 }
3764 fprintf(out,"\n"); (*lineno)++;
3765 if (!lemp->nolinenosflag) {
3766 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3767 }
3768 fprintf(out,"}\n"); (*lineno)++;
3769 return;
3770 }
3771
3772 /*
3773 ** Return TRUE (non-zero) if the given symbol has a destructor.
3774 */
has_destructor(struct symbol * sp,struct lemon * lemp)3775 int has_destructor(struct symbol *sp, struct lemon *lemp)
3776 {
3777 int ret;
3778 if( sp->type==TERMINAL ){
3779 ret = lemp->tokendest!=0;
3780 }else{
3781 ret = lemp->vardest!=0 || sp->destructor!=0;
3782 }
3783 return ret;
3784 }
3785
3786 /*
3787 ** Append text to a dynamically allocated string. If zText is 0 then
3788 ** reset the string to be empty again. Always return the complete text
3789 ** of the string (which is overwritten with each call).
3790 **
3791 ** n bytes of zText are stored. If n==0 then all of zText up to the first
3792 ** \000 terminator is stored. zText can contain up to two instances of
3793 ** %d. The values of p1 and p2 are written into the first and second
3794 ** %d.
3795 **
3796 ** If n==-1, then the previous character is overwritten.
3797 */
append_str(const char * zText,int n,int p1,int p2)3798 PRIVATE char *append_str(const char *zText, int n, int p1, int p2){
3799 static char empty[1] = { 0 };
3800 static char *z = 0;
3801 static int alloced = 0;
3802 static int used = 0;
3803 int c;
3804 char zInt[40];
3805 if( zText==0 ){
3806 if( used==0 && z!=0 ) z[0] = 0;
3807 used = 0;
3808 return z;
3809 }
3810 if( n<=0 ){
3811 if( n<0 ){
3812 used += n;
3813 assert( used>=0 );
3814 }
3815 n = lemonStrlen(zText);
3816 }
3817 if( (int) (n+sizeof(zInt)*2+used) >= alloced ){
3818 alloced = n + sizeof(zInt)*2 + used + 200;
3819 z = (char *) realloc(z, alloced);
3820 }
3821 if( z==0 ) return empty;
3822 while( n-- > 0 ){
3823 c = *(zText++);
3824 if( c=='%' && n>0 && zText[0]=='d' ){
3825 lemon_sprintf(zInt, "%d", p1);
3826 p1 = p2;
3827 lemon_strcpy(&z[used], zInt);
3828 used += lemonStrlen(&z[used]);
3829 zText++;
3830 n--;
3831 }else{
3832 z[used++] = (char)c;
3833 }
3834 }
3835 z[used] = 0;
3836 return z;
3837 }
3838
3839 /*
3840 ** Write and transform the rp->code string so that symbols are expanded.
3841 ** Populate the rp->codePrefix and rp->codeSuffix strings, as appropriate.
3842 **
3843 ** Return 1 if the expanded code requires that "yylhsminor" local variable
3844 ** to be defined.
3845 */
translate_code(struct lemon * lemp,struct rule * rp)3846 PRIVATE int translate_code(struct lemon *lemp, struct rule *rp){
3847 char *cp, *xp;
3848 int i;
3849 int rc = 0; /* True if yylhsminor is used */
3850 int dontUseRhs0 = 0; /* If true, use of left-most RHS label is illegal */
3851 const char *zSkip = 0; /* The zOvwrt comment within rp->code, or NULL */
3852 char lhsused = 0; /* True if the LHS element has been used */
3853 char lhsdirect; /* True if LHS writes directly into stack */
3854 char used[MAXRHS]; /* True for each RHS element which is used */
3855 char zLhs[50]; /* Convert the LHS symbol into this string */
3856 char zOvwrt[900]; /* Comment that to allow LHS to overwrite RHS */
3857
3858 for(i=0; i<rp->nrhs; i++) used[i] = 0;
3859 lhsused = 0;
3860
3861 if( rp->code==0 ){
3862 static char newlinestr[2] = { '\n', '\0' };
3863 rp->code = newlinestr;
3864 rp->line = rp->ruleline;
3865 rp->noCode = 1;
3866 }else{
3867 rp->noCode = 0;
3868 }
3869
3870
3871 if( rp->nrhs==0 ){
3872 /* If there are no RHS symbols, then writing directly to the LHS is ok */
3873 lhsdirect = 1;
3874 }else if( rp->rhsalias[0]==0 ){
3875 /* The left-most RHS symbol has no value. LHS direct is ok. But
3876 ** we have to call the destructor on the RHS symbol first. */
3877 lhsdirect = 1;
3878 if( has_destructor(rp->rhs[0],lemp) ){
3879 append_str(0,0,0,0);
3880 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3881 rp->rhs[0]->index,1-rp->nrhs);
3882 rp->codePrefix = Strsafe(append_str(0,0,0,0));
3883 rp->noCode = 0;
3884 }
3885 }else if( rp->lhsalias==0 ){
3886 /* There is no LHS value symbol. */
3887 lhsdirect = 1;
3888 }else if( strcmp(rp->lhsalias,rp->rhsalias[0])==0 ){
3889 /* The LHS symbol and the left-most RHS symbol are the same, so
3890 ** direct writing is allowed */
3891 lhsdirect = 1;
3892 lhsused = 1;
3893 used[0] = 1;
3894 if( rp->lhs->dtnum!=rp->rhs[0]->dtnum ){
3895 ErrorMsg(lemp->filename,rp->ruleline,
3896 "%s(%s) and %s(%s) share the same label but have "
3897 "different datatypes.",
3898 rp->lhs->name, rp->lhsalias, rp->rhs[0]->name, rp->rhsalias[0]);
3899 lemp->errorcnt++;
3900 }
3901 }else{
3902 lemon_sprintf(zOvwrt, "/*%s-overwrites-%s*/",
3903 rp->lhsalias, rp->rhsalias[0]);
3904 zSkip = strstr(rp->code, zOvwrt);
3905 if( zSkip!=0 ){
3906 /* The code contains a special comment that indicates that it is safe
3907 ** for the LHS label to overwrite left-most RHS label. */
3908 lhsdirect = 1;
3909 }else{
3910 lhsdirect = 0;
3911 }
3912 }
3913 if( lhsdirect ){
3914 sprintf(zLhs, "yymsp[%d].minor.yy%d",1-rp->nrhs,rp->lhs->dtnum);
3915 }else{
3916 rc = 1;
3917 sprintf(zLhs, "yylhsminor.yy%d",rp->lhs->dtnum);
3918 }
3919
3920 append_str(0,0,0,0);
3921
3922 /* This const cast is wrong but harmless, if we're careful. */
3923 for(cp=(char *)rp->code; *cp; cp++){
3924 if( cp==zSkip ){
3925 append_str(zOvwrt,0,0,0);
3926 cp += lemonStrlen(zOvwrt)-1;
3927 dontUseRhs0 = 1;
3928 continue;
3929 }
3930 if( ISALPHA(*cp) && (cp==rp->code || (!ISALNUM(cp[-1]) && cp[-1]!='_')) ){
3931 char saved;
3932 for(xp= &cp[1]; ISALNUM(*xp) || *xp=='_'; xp++);
3933 saved = *xp;
3934 *xp = 0;
3935 if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
3936 append_str(zLhs,0,0,0);
3937 cp = xp;
3938 lhsused = 1;
3939 }else{
3940 for(i=0; i<rp->nrhs; i++){
3941 if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
3942 if( i==0 && dontUseRhs0 ){
3943 ErrorMsg(lemp->filename,rp->ruleline,
3944 "Label %s used after '%s'.",
3945 rp->rhsalias[0], zOvwrt);
3946 lemp->errorcnt++;
3947 }else if( cp!=rp->code && cp[-1]=='@' ){
3948 /* If the argument is of the form @X then substituted
3949 ** the token number of X, not the value of X */
3950 append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
3951 }else{
3952 struct symbol *sp = rp->rhs[i];
3953 int dtnum;
3954 if( sp->type==MULTITERMINAL ){
3955 dtnum = sp->subsym[0]->dtnum;
3956 }else{
3957 dtnum = sp->dtnum;
3958 }
3959 append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);
3960 }
3961 cp = xp;
3962 used[i] = 1;
3963 break;
3964 }
3965 }
3966 }
3967 *xp = saved;
3968 }
3969 append_str(cp, 1, 0, 0);
3970 } /* End loop */
3971
3972 /* Main code generation completed */
3973 cp = append_str(0,0,0,0);
3974 if( cp && cp[0] ) rp->code = Strsafe(cp);
3975 append_str(0,0,0,0);
3976
3977 /* Check to make sure the LHS has been used */
3978 if( rp->lhsalias && !lhsused ){
3979 ErrorMsg(lemp->filename,rp->ruleline,
3980 "Label \"%s\" for \"%s(%s)\" is never used.",
3981 rp->lhsalias,rp->lhs->name,rp->lhsalias);
3982 lemp->errorcnt++;
3983 }
3984
3985 /* Generate destructor code for RHS minor values which are not referenced.
3986 ** Generate error messages for unused labels and duplicate labels.
3987 */
3988 for(i=0; i<rp->nrhs; i++){
3989 if( rp->rhsalias[i] ){
3990 if( i>0 ){
3991 int j;
3992 if( rp->lhsalias && strcmp(rp->lhsalias,rp->rhsalias[i])==0 ){
3993 ErrorMsg(lemp->filename,rp->ruleline,
3994 "%s(%s) has the same label as the LHS but is not the left-most "
3995 "symbol on the RHS.",
3996 rp->rhs[i]->name, rp->rhsalias[i]);
3997 lemp->errorcnt++;
3998 }
3999 for(j=0; j<i; j++){
4000 if( rp->rhsalias[j] && strcmp(rp->rhsalias[j],rp->rhsalias[i])==0 ){
4001 ErrorMsg(lemp->filename,rp->ruleline,
4002 "Label %s used for multiple symbols on the RHS of a rule.",
4003 rp->rhsalias[i]);
4004 lemp->errorcnt++;
4005 break;
4006 }
4007 }
4008 }
4009 if( !used[i] ){
4010 ErrorMsg(lemp->filename,rp->ruleline,
4011 "Label %s for \"%s(%s)\" is never used.",
4012 rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
4013 lemp->errorcnt++;
4014 }
4015 }else if( i>0 && has_destructor(rp->rhs[i],lemp) ){
4016 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
4017 rp->rhs[i]->index,i-rp->nrhs+1);
4018 }
4019 }
4020
4021 /* If unable to write LHS values directly into the stack, write the
4022 ** saved LHS value now. */
4023 if( lhsdirect==0 ){
4024 append_str(" yymsp[%d].minor.yy%d = ", 0, 1-rp->nrhs, rp->lhs->dtnum);
4025 append_str(zLhs, 0, 0, 0);
4026 append_str(";\n", 0, 0, 0);
4027 }
4028
4029 /* Suffix code generation complete */
4030 cp = append_str(0,0,0,0);
4031 if( cp && cp[0] ){
4032 rp->codeSuffix = Strsafe(cp);
4033 rp->noCode = 0;
4034 }
4035
4036 return rc;
4037 }
4038
4039 /*
4040 ** Generate code which executes when the rule "rp" is reduced. Write
4041 ** the code to "out". Make sure lineno stays up-to-date.
4042 */
emit_code(FILE * out,struct rule * rp,struct lemon * lemp,int * lineno)4043 PRIVATE void emit_code(
4044 FILE *out,
4045 struct rule *rp,
4046 struct lemon *lemp,
4047 int *lineno
4048 ){
4049 const char *cp;
4050
4051 /* Setup code prior to the #line directive */
4052 if( rp->codePrefix && rp->codePrefix[0] ){
4053 fprintf(out, "{%s", rp->codePrefix);
4054 for(cp=rp->codePrefix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4055 }
4056
4057 /* Generate code to do the reduce action */
4058 if( rp->code ){
4059 if( !lemp->nolinenosflag ){
4060 (*lineno)++;
4061 tplt_linedir(out,rp->line,lemp->filename);
4062 }
4063 fprintf(out,"{%s",rp->code);
4064 for(cp=rp->code; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4065 fprintf(out,"}\n"); (*lineno)++;
4066 if( !lemp->nolinenosflag ){
4067 (*lineno)++;
4068 tplt_linedir(out,*lineno,lemp->outname);
4069 }
4070 }
4071
4072 /* Generate breakdown code that occurs after the #line directive */
4073 if( rp->codeSuffix && rp->codeSuffix[0] ){
4074 fprintf(out, "%s", rp->codeSuffix);
4075 for(cp=rp->codeSuffix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4076 }
4077
4078 if( rp->codePrefix ){
4079 fprintf(out, "}\n"); (*lineno)++;
4080 }
4081
4082 return;
4083 }
4084
4085 /*
4086 ** Print the definition of the union used for the parser's data stack.
4087 ** This union contains fields for every possible data type for tokens
4088 ** and nonterminals. In the process of computing and printing this
4089 ** union, also set the ".dtnum" field of every terminal and nonterminal
4090 ** symbol.
4091 */
print_stack_union(FILE * out,struct lemon * lemp,int * plineno,int mhflag)4092 void print_stack_union(
4093 FILE *out, /* The output stream */
4094 struct lemon *lemp, /* The main info structure for this parser */
4095 int *plineno, /* Pointer to the line number */
4096 int mhflag /* True if generating makeheaders output */
4097 ){
4098 int lineno; /* The line number of the output */
4099 char **types; /* A hash table of datatypes */
4100 int arraysize; /* Size of the "types" array */
4101 int maxdtlength; /* Maximum length of any ".datatype" field. */
4102 char *stddt; /* Standardized name for a datatype */
4103 int i,j; /* Loop counters */
4104 unsigned hash; /* For hashing the name of a type */
4105 const char *name; /* Name of the parser */
4106
4107 /* Allocate and initialize types[] and allocate stddt[] */
4108 arraysize = lemp->nsymbol * 2;
4109 types = (char**)calloc( arraysize, sizeof(char*) );
4110 if( types==0 ){
4111 fprintf(stderr,"Out of memory.\n");
4112 exit(1);
4113 }
4114 for(i=0; i<arraysize; i++) types[i] = 0;
4115 maxdtlength = 0;
4116 if( lemp->vartype ){
4117 maxdtlength = lemonStrlen(lemp->vartype);
4118 }
4119 for(i=0; i<lemp->nsymbol; i++){
4120 int len;
4121 struct symbol *sp = lemp->symbols[i];
4122 if( sp->datatype==0 ) continue;
4123 len = lemonStrlen(sp->datatype);
4124 if( len>maxdtlength ) maxdtlength = len;
4125 }
4126 stddt = (char*)malloc( maxdtlength*2 + 1 );
4127 if( stddt==0 ){
4128 fprintf(stderr,"Out of memory.\n");
4129 exit(1);
4130 }
4131
4132 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
4133 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
4134 ** used for terminal symbols. If there is no %default_type defined then
4135 ** 0 is also used as the .dtnum value for nonterminals which do not specify
4136 ** a datatype using the %type directive.
4137 */
4138 for(i=0; i<lemp->nsymbol; i++){
4139 struct symbol *sp = lemp->symbols[i];
4140 char *cp;
4141 if( sp==lemp->errsym ){
4142 sp->dtnum = arraysize+1;
4143 continue;
4144 }
4145 if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
4146 sp->dtnum = 0;
4147 continue;
4148 }
4149 cp = sp->datatype;
4150 if( cp==0 ) cp = lemp->vartype;
4151 j = 0;
4152 while( ISSPACE(*cp) ) cp++;
4153 while( *cp ) stddt[j++] = *cp++;
4154 while( j>0 && ISSPACE(stddt[j-1]) ) j--;
4155 stddt[j] = 0;
4156 if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){
4157 sp->dtnum = 0;
4158 continue;
4159 }
4160 hash = 0;
4161 for(j=0; stddt[j]; j++){
4162 hash = hash*53 + stddt[j];
4163 }
4164 hash = (hash & 0x7fffffff)%arraysize;
4165 while( types[hash] ){
4166 if( strcmp(types[hash],stddt)==0 ){
4167 sp->dtnum = hash + 1;
4168 break;
4169 }
4170 hash++;
4171 if( hash>=(unsigned)arraysize ) hash = 0;
4172 }
4173 if( types[hash]==0 ){
4174 sp->dtnum = hash + 1;
4175 types[hash] = (char*)malloc( lemonStrlen(stddt)+1 );
4176 if( types[hash]==0 ){
4177 fprintf(stderr,"Out of memory.\n");
4178 exit(1);
4179 }
4180 lemon_strcpy(types[hash],stddt);
4181 }
4182 }
4183
4184 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
4185 name = lemp->name ? lemp->name : "Parse";
4186 lineno = *plineno;
4187 if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
4188 fprintf(out,"#define %sTOKENTYPE %s\n",name,
4189 lemp->tokentype?lemp->tokentype:"void*"); lineno++;
4190 if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
4191 fprintf(out,"typedef union {\n"); lineno++;
4192 fprintf(out," int yyinit;\n"); lineno++;
4193 fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++;
4194 for(i=0; i<arraysize; i++){
4195 if( types[i]==0 ) continue;
4196 fprintf(out," %s yy%d;\n",types[i],i+1); lineno++;
4197 free(types[i]);
4198 }
4199 if( lemp->errsym && lemp->errsym->useCnt ){
4200 fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++;
4201 }
4202 free(stddt);
4203 free(types);
4204 fprintf(out,"} YYMINORTYPE;\n"); lineno++;
4205 *plineno = lineno;
4206 }
4207
4208 /*
4209 ** Return the name of a C datatype able to represent values between
4210 ** lwr and upr, inclusive. If pnByte!=NULL then also write the sizeof
4211 ** for that type (1, 2, or 4) into *pnByte.
4212 */
minimum_size_type(int lwr,int upr,int * pnByte)4213 static const char *minimum_size_type(int lwr, int upr, int *pnByte){
4214 const char *zType = "int";
4215 int nByte = 4;
4216 if( lwr>=0 ){
4217 if( upr<=255 ){
4218 zType = "unsigned char";
4219 nByte = 1;
4220 }else if( upr<65535 ){
4221 zType = "unsigned short int";
4222 nByte = 2;
4223 }else{
4224 zType = "unsigned int";
4225 nByte = 4;
4226 }
4227 }else if( lwr>=-127 && upr<=127 ){
4228 zType = "signed char";
4229 nByte = 1;
4230 }else if( lwr>=-32767 && upr<32767 ){
4231 zType = "short";
4232 nByte = 2;
4233 }
4234 if( pnByte ) *pnByte = nByte;
4235 return zType;
4236 }
4237
4238 /*
4239 ** Each state contains a set of token transaction and a set of
4240 ** nonterminal transactions. Each of these sets makes an instance
4241 ** of the following structure. An array of these structures is used
4242 ** to order the creation of entries in the yy_action[] table.
4243 */
4244 struct axset {
4245 struct state *stp; /* A pointer to a state */
4246 int isTkn; /* True to use tokens. False for non-terminals */
4247 int nAction; /* Number of actions */
4248 int iOrder; /* Original order of action sets */
4249 };
4250
4251 /*
4252 ** Compare to axset structures for sorting purposes
4253 */
axset_compare(const void * a,const void * b)4254 static int axset_compare(const void *a, const void *b){
4255 struct axset *p1 = (struct axset*)a;
4256 struct axset *p2 = (struct axset*)b;
4257 int c;
4258 c = p2->nAction - p1->nAction;
4259 if( c==0 ){
4260 c = p1->iOrder - p2->iOrder;
4261 }
4262 assert( c!=0 || p1==p2 );
4263 return c;
4264 }
4265
4266 /*
4267 ** Write text on "out" that describes the rule "rp".
4268 */
writeRuleText(FILE * out,struct rule * rp)4269 static void writeRuleText(FILE *out, struct rule *rp){
4270 int j;
4271 fprintf(out,"%s ::=", rp->lhs->name);
4272 for(j=0; j<rp->nrhs; j++){
4273 struct symbol *sp = rp->rhs[j];
4274 if( sp->type!=MULTITERMINAL ){
4275 fprintf(out," %s", sp->name);
4276 }else{
4277 int k;
4278 fprintf(out," %s", sp->subsym[0]->name);
4279 for(k=1; k<sp->nsubsym; k++){
4280 fprintf(out,"|%s",sp->subsym[k]->name);
4281 }
4282 }
4283 }
4284 }
4285
4286
4287 /* Generate C source code for the parser */
ReportTable(struct lemon * lemp,int mhflag,int sqlFlag)4288 void ReportTable(
4289 struct lemon *lemp,
4290 int mhflag, /* Output in makeheaders format if true */
4291 int sqlFlag /* Generate the *.sql file too */
4292 ){
4293 FILE *out, *in, *sql;
4294 int lineno;
4295 struct state *stp;
4296 struct action *ap;
4297 struct rule *rp;
4298 struct acttab *pActtab;
4299 int i, j, n, sz;
4300 int nLookAhead;
4301 int szActionType; /* sizeof(YYACTIONTYPE) */
4302 int szCodeType; /* sizeof(YYCODETYPE) */
4303 const char *name;
4304 int mnTknOfst, mxTknOfst;
4305 int mnNtOfst, mxNtOfst;
4306 struct axset *ax;
4307 char *prefix;
4308
4309 lemp->minShiftReduce = lemp->nstate;
4310 lemp->errAction = lemp->minShiftReduce + lemp->nrule;
4311 lemp->accAction = lemp->errAction + 1;
4312 lemp->noAction = lemp->accAction + 1;
4313 lemp->minReduce = lemp->noAction + 1;
4314 lemp->maxAction = lemp->minReduce + lemp->nrule;
4315
4316 in = tplt_open(lemp);
4317 if( in==0 ) return;
4318 out = file_open(lemp,".c","wb");
4319 if( out==0 ){
4320 fclose(in);
4321 return;
4322 }
4323 if( sqlFlag==0 ){
4324 sql = 0;
4325 }else{
4326 sql = file_open(lemp, ".sql", "wb");
4327 if( sql==0 ){
4328 fclose(in);
4329 fclose(out);
4330 return;
4331 }
4332 fprintf(sql,
4333 "BEGIN;\n"
4334 "CREATE TABLE symbol(\n"
4335 " id INTEGER PRIMARY KEY,\n"
4336 " name TEXT NOT NULL,\n"
4337 " isTerminal BOOLEAN NOT NULL,\n"
4338 " fallback INTEGER REFERENCES symbol"
4339 " DEFERRABLE INITIALLY DEFERRED\n"
4340 ");\n"
4341 );
4342 for(i=0; i<lemp->nsymbol; i++){
4343 fprintf(sql,
4344 "INSERT INTO symbol(id,name,isTerminal,fallback)"
4345 "VALUES(%d,'%s',%s",
4346 i, lemp->symbols[i]->name,
4347 i<lemp->nterminal ? "TRUE" : "FALSE"
4348 );
4349 if( lemp->symbols[i]->fallback ){
4350 fprintf(sql, ",%d);\n", lemp->symbols[i]->fallback->index);
4351 }else{
4352 fprintf(sql, ",NULL);\n");
4353 }
4354 }
4355 fprintf(sql,
4356 "CREATE TABLE rule(\n"
4357 " ruleid INTEGER PRIMARY KEY,\n"
4358 " lhs INTEGER REFERENCES symbol(id),\n"
4359 " txt TEXT\n"
4360 ");\n"
4361 "CREATE TABLE rulerhs(\n"
4362 " ruleid INTEGER REFERENCES rule(ruleid),\n"
4363 " pos INTEGER,\n"
4364 " sym INTEGER REFERENCES symbol(id)\n"
4365 ");\n"
4366 );
4367 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4368 assert( i==rp->iRule );
4369 fprintf(sql,
4370 "INSERT INTO rule(ruleid,lhs,txt)VALUES(%d,%d,'",
4371 rp->iRule, rp->lhs->index
4372 );
4373 writeRuleText(sql, rp);
4374 fprintf(sql,"');\n");
4375 for(j=0; j<rp->nrhs; j++){
4376 struct symbol *sp = rp->rhs[j];
4377 if( sp->type!=MULTITERMINAL ){
4378 fprintf(sql,
4379 "INSERT INTO rulerhs(ruleid,pos,sym)VALUES(%d,%d,%d);\n",
4380 i,j,sp->index
4381 );
4382 }else{
4383 int k;
4384 for(k=0; k<sp->nsubsym; k++){
4385 fprintf(sql,
4386 "INSERT INTO rulerhs(ruleid,pos,sym)VALUES(%d,%d,%d);\n",
4387 i,j,sp->subsym[k]->index
4388 );
4389 }
4390 }
4391 }
4392 }
4393 fprintf(sql, "COMMIT;\n");
4394 }
4395 lineno = 1;
4396
4397 fprintf(out,
4398 "/* This file is automatically generated by Lemon from input grammar\n"
4399 "** source file \"%s\". */\n", lemp->filename); lineno += 2;
4400
4401 /* The first %include directive begins with a C-language comment,
4402 ** then skip over the header comment of the template file
4403 */
4404 if( lemp->include==0 ) lemp->include = "";
4405 for(i=0; ISSPACE(lemp->include[i]); i++){
4406 if( lemp->include[i]=='\n' ){
4407 lemp->include += i+1;
4408 i = -1;
4409 }
4410 }
4411 if( lemp->include[0]=='/' ){
4412 tplt_skip_header(in,&lineno);
4413 }else{
4414 tplt_xfer(lemp->name,in,out,&lineno);
4415 }
4416
4417 /* Generate the include code, if any */
4418 tplt_print(out,lemp,lemp->include,&lineno);
4419 if( mhflag ){
4420 char *incName = file_makename(lemp, ".h");
4421 fprintf(out,"#include \"%s\"\n", incName); lineno++;
4422 free(incName);
4423 }
4424 tplt_xfer(lemp->name,in,out,&lineno);
4425
4426 /* Generate #defines for all tokens */
4427 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4428 else prefix = "";
4429 if( mhflag ){
4430 fprintf(out,"#if INTERFACE\n"); lineno++;
4431 }else{
4432 fprintf(out,"#ifndef %s%s\n", prefix, lemp->symbols[1]->name);
4433 }
4434 for(i=1; i<lemp->nterminal; i++){
4435 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
4436 lineno++;
4437 }
4438 fprintf(out,"#endif\n"); lineno++;
4439 tplt_xfer(lemp->name,in,out,&lineno);
4440
4441 /* Generate the defines */
4442 fprintf(out,"#define YYCODETYPE %s\n",
4443 minimum_size_type(0, lemp->nsymbol, &szCodeType)); lineno++;
4444 fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol); lineno++;
4445 fprintf(out,"#define YYACTIONTYPE %s\n",
4446 minimum_size_type(0,lemp->maxAction,&szActionType)); lineno++;
4447 if( lemp->wildcard ){
4448 fprintf(out,"#define YYWILDCARD %d\n",
4449 lemp->wildcard->index); lineno++;
4450 }
4451 print_stack_union(out,lemp,&lineno,mhflag);
4452 fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++;
4453 if( lemp->stacksize ){
4454 fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++;
4455 }else{
4456 fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++;
4457 }
4458 fprintf(out, "#endif\n"); lineno++;
4459 if( mhflag ){
4460 fprintf(out,"#if INTERFACE\n"); lineno++;
4461 }
4462 name = lemp->name ? lemp->name : "Parse";
4463 if( lemp->arg && lemp->arg[0] ){
4464 i = lemonStrlen(lemp->arg);
4465 while( i>=1 && ISSPACE(lemp->arg[i-1]) ) i--;
4466 while( i>=1 && (ISALNUM(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
4467 fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++;
4468 fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++;
4469 fprintf(out,"#define %sARG_PARAM ,%s\n",name,&lemp->arg[i]); lineno++;
4470 fprintf(out,"#define %sARG_FETCH %s=yypParser->%s;\n",
4471 name,lemp->arg,&lemp->arg[i]); lineno++;
4472 fprintf(out,"#define %sARG_STORE yypParser->%s=%s;\n",
4473 name,&lemp->arg[i],&lemp->arg[i]); lineno++;
4474 }else{
4475 fprintf(out,"#define %sARG_SDECL\n",name); lineno++;
4476 fprintf(out,"#define %sARG_PDECL\n",name); lineno++;
4477 fprintf(out,"#define %sARG_PARAM\n",name); lineno++;
4478 fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
4479 fprintf(out,"#define %sARG_STORE\n",name); lineno++;
4480 }
4481 if( lemp->ctx && lemp->ctx[0] ){
4482 i = lemonStrlen(lemp->ctx);
4483 while( i>=1 && ISSPACE(lemp->ctx[i-1]) ) i--;
4484 while( i>=1 && (ISALNUM(lemp->ctx[i-1]) || lemp->ctx[i-1]=='_') ) i--;
4485 fprintf(out,"#define %sCTX_SDECL %s;\n",name,lemp->ctx); lineno++;
4486 fprintf(out,"#define %sCTX_PDECL ,%s\n",name,lemp->ctx); lineno++;
4487 fprintf(out,"#define %sCTX_PARAM ,%s\n",name,&lemp->ctx[i]); lineno++;
4488 fprintf(out,"#define %sCTX_FETCH %s=yypParser->%s;\n",
4489 name,lemp->ctx,&lemp->ctx[i]); lineno++;
4490 fprintf(out,"#define %sCTX_STORE yypParser->%s=%s;\n",
4491 name,&lemp->ctx[i],&lemp->ctx[i]); lineno++;
4492 }else{
4493 fprintf(out,"#define %sCTX_SDECL\n",name); lineno++;
4494 fprintf(out,"#define %sCTX_PDECL\n",name); lineno++;
4495 fprintf(out,"#define %sCTX_PARAM\n",name); lineno++;
4496 fprintf(out,"#define %sCTX_FETCH\n",name); lineno++;
4497 fprintf(out,"#define %sCTX_STORE\n",name); lineno++;
4498 }
4499 if( mhflag ){
4500 fprintf(out,"#endif\n"); lineno++;
4501 }
4502 if( lemp->errsym && lemp->errsym->useCnt ){
4503 fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++;
4504 fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++;
4505 }
4506 if( lemp->has_fallback ){
4507 fprintf(out,"#define YYFALLBACK 1\n"); lineno++;
4508 }
4509
4510 /* Compute the action table, but do not output it yet. The action
4511 ** table must be computed before generating the YYNSTATE macro because
4512 ** we need to know how many states can be eliminated.
4513 */
4514 ax = (struct axset *) calloc(lemp->nxstate*2, sizeof(ax[0]));
4515 if( ax==0 ){
4516 fprintf(stderr,"malloc failed\n");
4517 exit(1);
4518 }
4519 for(i=0; i<lemp->nxstate; i++){
4520 stp = lemp->sorted[i];
4521 ax[i*2].stp = stp;
4522 ax[i*2].isTkn = 1;
4523 ax[i*2].nAction = stp->nTknAct;
4524 ax[i*2+1].stp = stp;
4525 ax[i*2+1].isTkn = 0;
4526 ax[i*2+1].nAction = stp->nNtAct;
4527 }
4528 mxTknOfst = mnTknOfst = 0;
4529 mxNtOfst = mnNtOfst = 0;
4530 /* In an effort to minimize the action table size, use the heuristic
4531 ** of placing the largest action sets first */
4532 for(i=0; i<lemp->nxstate*2; i++) ax[i].iOrder = i;
4533 qsort(ax, lemp->nxstate*2, sizeof(ax[0]), axset_compare);
4534 pActtab = acttab_alloc(lemp->nsymbol, lemp->nterminal);
4535 for(i=0; i<lemp->nxstate*2 && ax[i].nAction>0; i++){
4536 stp = ax[i].stp;
4537 if( ax[i].isTkn ){
4538 for(ap=stp->ap; ap; ap=ap->next){
4539 int action;
4540 if( ap->sp->index>=lemp->nterminal ) continue;
4541 action = compute_action(lemp, ap);
4542 if( action<0 ) continue;
4543 acttab_action(pActtab, ap->sp->index, action);
4544 }
4545 stp->iTknOfst = acttab_insert(pActtab, 1);
4546 if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
4547 if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
4548 }else{
4549 for(ap=stp->ap; ap; ap=ap->next){
4550 int action;
4551 if( ap->sp->index<lemp->nterminal ) continue;
4552 if( ap->sp->index==lemp->nsymbol ) continue;
4553 action = compute_action(lemp, ap);
4554 if( action<0 ) continue;
4555 acttab_action(pActtab, ap->sp->index, action);
4556 }
4557 stp->iNtOfst = acttab_insert(pActtab, 0);
4558 if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
4559 if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
4560 }
4561 #if 0 /* Uncomment for a trace of how the yy_action[] table fills out */
4562 { int jj, nn;
4563 for(jj=nn=0; jj<pActtab->nAction; jj++){
4564 if( pActtab->aAction[jj].action<0 ) nn++;
4565 }
4566 printf("%4d: State %3d %s n: %2d size: %5d freespace: %d\n",
4567 i, stp->statenum, ax[i].isTkn ? "Token" : "Var ",
4568 ax[i].nAction, pActtab->nAction, nn);
4569 }
4570 #endif
4571 }
4572 free(ax);
4573
4574 /* Mark rules that are actually used for reduce actions after all
4575 ** optimizations have been applied
4576 */
4577 for(rp=lemp->rule; rp; rp=rp->next) rp->doesReduce = LEMON_FALSE;
4578 for(i=0; i<lemp->nxstate; i++){
4579 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
4580 if( ap->type==REDUCE || ap->type==SHIFTREDUCE ){
4581 ap->x.rp->doesReduce = 1;
4582 }
4583 }
4584 }
4585
4586 /* Finish rendering the constants now that the action table has
4587 ** been computed */
4588 fprintf(out,"#define YYNSTATE %d\n",lemp->nxstate); lineno++;
4589 fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++;
4590 fprintf(out,"#define YYNRULE_WITH_ACTION %d\n",lemp->nruleWithAction);
4591 lineno++;
4592 fprintf(out,"#define YYNTOKEN %d\n",lemp->nterminal); lineno++;
4593 fprintf(out,"#define YY_MAX_SHIFT %d\n",lemp->nxstate-1); lineno++;
4594 i = lemp->minShiftReduce;
4595 fprintf(out,"#define YY_MIN_SHIFTREDUCE %d\n",i); lineno++;
4596 i += lemp->nrule;
4597 fprintf(out,"#define YY_MAX_SHIFTREDUCE %d\n", i-1); lineno++;
4598 fprintf(out,"#define YY_ERROR_ACTION %d\n", lemp->errAction); lineno++;
4599 fprintf(out,"#define YY_ACCEPT_ACTION %d\n", lemp->accAction); lineno++;
4600 fprintf(out,"#define YY_NO_ACTION %d\n", lemp->noAction); lineno++;
4601 fprintf(out,"#define YY_MIN_REDUCE %d\n", lemp->minReduce); lineno++;
4602 i = lemp->minReduce + lemp->nrule;
4603 fprintf(out,"#define YY_MAX_REDUCE %d\n", i-1); lineno++;
4604 tplt_xfer(lemp->name,in,out,&lineno);
4605
4606 /* Now output the action table and its associates:
4607 **
4608 ** yy_action[] A single table containing all actions.
4609 ** yy_lookahead[] A table containing the lookahead for each entry in
4610 ** yy_action. Used to detect hash collisions.
4611 ** yy_shift_ofst[] For each state, the offset into yy_action for
4612 ** shifting terminals.
4613 ** yy_reduce_ofst[] For each state, the offset into yy_action for
4614 ** shifting non-terminals after a reduce.
4615 ** yy_default[] Default action for each state.
4616 */
4617
4618 /* Output the yy_action table */
4619 lemp->nactiontab = n = acttab_action_size(pActtab);
4620 lemp->tablesize += n*szActionType;
4621 fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++;
4622 fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++;
4623 for(i=j=0; i<n; i++){
4624 int action = acttab_yyaction(pActtab, i);
4625 if( action<0 ) action = lemp->noAction;
4626 if( j==0 ) fprintf(out," /* %5d */ ", i);
4627 fprintf(out, " %4d,", action);
4628 if( j==9 || i==n-1 ){
4629 fprintf(out, "\n"); lineno++;
4630 j = 0;
4631 }else{
4632 j++;
4633 }
4634 }
4635 fprintf(out, "};\n"); lineno++;
4636
4637 /* Output the yy_lookahead table */
4638 lemp->nlookaheadtab = n = acttab_lookahead_size(pActtab);
4639 lemp->tablesize += n*szCodeType;
4640 fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++;
4641 for(i=j=0; i<n; i++){
4642 int la = acttab_yylookahead(pActtab, i);
4643 if( la<0 ) la = lemp->nsymbol;
4644 if( j==0 ) fprintf(out," /* %5d */ ", i);
4645 fprintf(out, " %4d,", la);
4646 if( j==9 ){
4647 fprintf(out, "\n"); lineno++;
4648 j = 0;
4649 }else{
4650 j++;
4651 }
4652 }
4653 /* Add extra entries to the end of the yy_lookahead[] table so that
4654 ** yy_shift_ofst[]+iToken will always be a valid index into the array,
4655 ** even for the largest possible value of yy_shift_ofst[] and iToken. */
4656 nLookAhead = lemp->nterminal + lemp->nactiontab;
4657 while( i<nLookAhead ){
4658 if( j==0 ) fprintf(out," /* %5d */ ", i);
4659 fprintf(out, " %4d,", lemp->nterminal);
4660 if( j==9 ){
4661 fprintf(out, "\n"); lineno++;
4662 j = 0;
4663 }else{
4664 j++;
4665 }
4666 i++;
4667 }
4668 if( j>0 ){ fprintf(out, "\n"); lineno++; }
4669 fprintf(out, "};\n"); lineno++;
4670
4671 /* Output the yy_shift_ofst[] table */
4672 n = lemp->nxstate;
4673 while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--;
4674 fprintf(out, "#define YY_SHIFT_COUNT (%d)\n", n-1); lineno++;
4675 fprintf(out, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst); lineno++;
4676 fprintf(out, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst); lineno++;
4677 fprintf(out, "static const %s yy_shift_ofst[] = {\n",
4678 minimum_size_type(mnTknOfst, lemp->nterminal+lemp->nactiontab, &sz));
4679 lineno++;
4680 lemp->tablesize += n*sz;
4681 for(i=j=0; i<n; i++){
4682 int ofst;
4683 stp = lemp->sorted[i];
4684 ofst = stp->iTknOfst;
4685 if( ofst==NO_OFFSET ) ofst = lemp->nactiontab;
4686 if( j==0 ) fprintf(out," /* %5d */ ", i);
4687 fprintf(out, " %4d,", ofst);
4688 if( j==9 || i==n-1 ){
4689 fprintf(out, "\n"); lineno++;
4690 j = 0;
4691 }else{
4692 j++;
4693 }
4694 }
4695 fprintf(out, "};\n"); lineno++;
4696
4697 /* Output the yy_reduce_ofst[] table */
4698 n = lemp->nxstate;
4699 while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--;
4700 fprintf(out, "#define YY_REDUCE_COUNT (%d)\n", n-1); lineno++;
4701 fprintf(out, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst); lineno++;
4702 fprintf(out, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst); lineno++;
4703 fprintf(out, "static const %s yy_reduce_ofst[] = {\n",
4704 minimum_size_type(mnNtOfst-1, mxNtOfst, &sz)); lineno++;
4705 lemp->tablesize += n*sz;
4706 for(i=j=0; i<n; i++){
4707 int ofst;
4708 stp = lemp->sorted[i];
4709 ofst = stp->iNtOfst;
4710 if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
4711 if( j==0 ) fprintf(out," /* %5d */ ", i);
4712 fprintf(out, " %4d,", ofst);
4713 if( j==9 || i==n-1 ){
4714 fprintf(out, "\n"); lineno++;
4715 j = 0;
4716 }else{
4717 j++;
4718 }
4719 }
4720 fprintf(out, "};\n"); lineno++;
4721
4722 /* Output the default action table */
4723 fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++;
4724 n = lemp->nxstate;
4725 lemp->tablesize += n*szActionType;
4726 for(i=j=0; i<n; i++){
4727 stp = lemp->sorted[i];
4728 if( j==0 ) fprintf(out," /* %5d */ ", i);
4729 if( stp->iDfltReduce<0 ){
4730 fprintf(out, " %4d,", lemp->errAction);
4731 }else{
4732 fprintf(out, " %4d,", stp->iDfltReduce + lemp->minReduce);
4733 }
4734 if( j==9 || i==n-1 ){
4735 fprintf(out, "\n"); lineno++;
4736 j = 0;
4737 }else{
4738 j++;
4739 }
4740 }
4741 fprintf(out, "};\n"); lineno++;
4742 tplt_xfer(lemp->name,in,out,&lineno);
4743
4744 /* Generate the table of fallback tokens.
4745 */
4746 if( lemp->has_fallback ){
4747 int mx = lemp->nterminal - 1;
4748 /* 2019-08-28: Generate fallback entries for every token to avoid
4749 ** having to do a range check on the index */
4750 /* while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; } */
4751 lemp->tablesize += (mx+1)*szCodeType;
4752 for(i=0; i<=mx; i++){
4753 struct symbol *p = lemp->symbols[i];
4754 if( p->fallback==0 ){
4755 fprintf(out, " 0, /* %10s => nothing */\n", p->name);
4756 }else{
4757 fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index,
4758 p->name, p->fallback->name);
4759 }
4760 lineno++;
4761 }
4762 }
4763 tplt_xfer(lemp->name, in, out, &lineno);
4764
4765 /* Generate a table containing the symbolic name of every symbol
4766 */
4767 for(i=0; i<lemp->nsymbol; i++){
4768 fprintf(out," /* %4d */ \"%s\",\n",i, lemp->symbols[i]->name); lineno++;
4769 }
4770 tplt_xfer(lemp->name,in,out,&lineno);
4771
4772 /* Generate a table containing a text string that describes every
4773 ** rule in the rule set of the grammar. This information is used
4774 ** when tracing REDUCE actions.
4775 */
4776 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4777 assert( rp->iRule==i );
4778 fprintf(out," /* %3d */ \"", i);
4779 writeRuleText(out, rp);
4780 fprintf(out,"\",\n"); lineno++;
4781 }
4782 tplt_xfer(lemp->name,in,out,&lineno);
4783
4784 /* Generate code which executes every time a symbol is popped from
4785 ** the stack while processing errors or while destroying the parser.
4786 ** (In other words, generate the %destructor actions)
4787 */
4788 if( lemp->tokendest ){
4789 int once = 1;
4790 for(i=0; i<lemp->nsymbol; i++){
4791 struct symbol *sp = lemp->symbols[i];
4792 if( sp==0 || sp->type!=TERMINAL ) continue;
4793 if( once ){
4794 fprintf(out, " /* TERMINAL Destructor */\n"); lineno++;
4795 once = 0;
4796 }
4797 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4798 }
4799 for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
4800 if( i<lemp->nsymbol ){
4801 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4802 fprintf(out," break;\n"); lineno++;
4803 }
4804 }
4805 if( lemp->vardest ){
4806 struct symbol *dflt_sp = 0;
4807 int once = 1;
4808 for(i=0; i<lemp->nsymbol; i++){
4809 struct symbol *sp = lemp->symbols[i];
4810 if( sp==0 || sp->type==TERMINAL ||
4811 sp->index<=0 || sp->destructor!=0 ) continue;
4812 if( once ){
4813 fprintf(out, " /* Default NON-TERMINAL Destructor */\n");lineno++;
4814 once = 0;
4815 }
4816 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4817 dflt_sp = sp;
4818 }
4819 if( dflt_sp!=0 ){
4820 emit_destructor_code(out,dflt_sp,lemp,&lineno);
4821 }
4822 fprintf(out," break;\n"); lineno++;
4823 }
4824 for(i=0; i<lemp->nsymbol; i++){
4825 struct symbol *sp = lemp->symbols[i];
4826 if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
4827 if( sp->destLineno<0 ) continue; /* Already emitted */
4828 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
4829
4830 /* Combine duplicate destructors into a single case */
4831 for(j=i+1; j<lemp->nsymbol; j++){
4832 struct symbol *sp2 = lemp->symbols[j];
4833 if( sp2 && sp2->type!=TERMINAL && sp2->destructor
4834 && sp2->dtnum==sp->dtnum
4835 && strcmp(sp->destructor,sp2->destructor)==0 ){
4836 fprintf(out," case %d: /* %s */\n",
4837 sp2->index, sp2->name); lineno++;
4838 sp2->destLineno = -1; /* Avoid emitting this destructor again */
4839 }
4840 }
4841
4842 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4843 fprintf(out," break;\n"); lineno++;
4844 }
4845 tplt_xfer(lemp->name,in,out,&lineno);
4846
4847 /* Generate code which executes whenever the parser stack overflows */
4848 tplt_print(out,lemp,lemp->overflow,&lineno);
4849 tplt_xfer(lemp->name,in,out,&lineno);
4850
4851 /* Generate the tables of rule information. yyRuleInfoLhs[] and
4852 ** yyRuleInfoNRhs[].
4853 **
4854 ** Note: This code depends on the fact that rules are number
4855 ** sequentially beginning with 0.
4856 */
4857 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4858 fprintf(out," %4d, /* (%d) ", rp->lhs->index, i);
4859 rule_print(out, rp);
4860 fprintf(out," */\n"); lineno++;
4861 }
4862 tplt_xfer(lemp->name,in,out,&lineno);
4863 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4864 fprintf(out," %3d, /* (%d) ", -rp->nrhs, i);
4865 rule_print(out, rp);
4866 fprintf(out," */\n"); lineno++;
4867 }
4868 tplt_xfer(lemp->name,in,out,&lineno);
4869
4870 /* Generate code which execution during each REDUCE action */
4871 i = 0;
4872 for(rp=lemp->rule; rp; rp=rp->next){
4873 i += translate_code(lemp, rp);
4874 }
4875 if( i ){
4876 fprintf(out," YYMINORTYPE yylhsminor;\n"); lineno++;
4877 }
4878 /* First output rules other than the default: rule */
4879 for(rp=lemp->rule; rp; rp=rp->next){
4880 struct rule *rp2; /* Other rules with the same action */
4881 if( rp->codeEmitted ) continue;
4882 if( rp->noCode ){
4883 /* No C code actions, so this will be part of the "default:" rule */
4884 continue;
4885 }
4886 fprintf(out," case %d: /* ", rp->iRule);
4887 writeRuleText(out, rp);
4888 fprintf(out, " */\n"); lineno++;
4889 for(rp2=rp->next; rp2; rp2=rp2->next){
4890 if( rp2->code==rp->code && rp2->codePrefix==rp->codePrefix
4891 && rp2->codeSuffix==rp->codeSuffix ){
4892 fprintf(out," case %d: /* ", rp2->iRule);
4893 writeRuleText(out, rp2);
4894 fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->iRule); lineno++;
4895 rp2->codeEmitted = 1;
4896 }
4897 }
4898 emit_code(out,rp,lemp,&lineno);
4899 fprintf(out," break;\n"); lineno++;
4900 rp->codeEmitted = 1;
4901 }
4902 /* Finally, output the default: rule. We choose as the default: all
4903 ** empty actions. */
4904 fprintf(out," default:\n"); lineno++;
4905 for(rp=lemp->rule; rp; rp=rp->next){
4906 if( rp->codeEmitted ) continue;
4907 assert( rp->noCode );
4908 fprintf(out," /* (%d) ", rp->iRule);
4909 writeRuleText(out, rp);
4910 if( rp->neverReduce ){
4911 fprintf(out, " (NEVER REDUCES) */ assert(yyruleno!=%d);\n",
4912 rp->iRule); lineno++;
4913 }else if( rp->doesReduce ){
4914 fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->iRule); lineno++;
4915 }else{
4916 fprintf(out, " (OPTIMIZED OUT) */ assert(yyruleno!=%d);\n",
4917 rp->iRule); lineno++;
4918 }
4919 }
4920 fprintf(out," break;\n"); lineno++;
4921 tplt_xfer(lemp->name,in,out,&lineno);
4922
4923 /* Generate code which executes if a parse fails */
4924 tplt_print(out,lemp,lemp->failure,&lineno);
4925 tplt_xfer(lemp->name,in,out,&lineno);
4926
4927 /* Generate code which executes when a syntax error occurs */
4928 tplt_print(out,lemp,lemp->error,&lineno);
4929 tplt_xfer(lemp->name,in,out,&lineno);
4930
4931 /* Generate code which executes when the parser accepts its input */
4932 tplt_print(out,lemp,lemp->accept,&lineno);
4933 tplt_xfer(lemp->name,in,out,&lineno);
4934
4935 /* Append any addition code the user desires */
4936 tplt_print(out,lemp,lemp->extracode,&lineno);
4937
4938 acttab_free(pActtab);
4939 fclose(in);
4940 fclose(out);
4941 if( sql ) fclose(sql);
4942 return;
4943 }
4944
4945 /* Generate a header file for the parser */
ReportHeader(struct lemon * lemp)4946 void ReportHeader(struct lemon *lemp)
4947 {
4948 FILE *out, *in;
4949 const char *prefix;
4950 char line[LINESIZE];
4951 char pattern[LINESIZE];
4952 int i;
4953
4954 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4955 else prefix = "";
4956 in = file_open(lemp,".h","rb");
4957 if( in ){
4958 int nextChar;
4959 for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
4960 lemon_sprintf(pattern,"#define %s%-30s %3d\n",
4961 prefix,lemp->symbols[i]->name,i);
4962 if( strcmp(line,pattern) ) break;
4963 }
4964 nextChar = fgetc(in);
4965 fclose(in);
4966 if( i==lemp->nterminal && nextChar==EOF ){
4967 /* No change in the file. Don't rewrite it. */
4968 return;
4969 }
4970 }
4971 out = file_open(lemp,".h","wb");
4972 if( out ){
4973 for(i=1; i<lemp->nterminal; i++){
4974 fprintf(out,"#define %s%-30s %3d\n",prefix,lemp->symbols[i]->name,i);
4975 }
4976 fclose(out);
4977 }
4978 return;
4979 }
4980
4981 /* Reduce the size of the action tables, if possible, by making use
4982 ** of defaults.
4983 **
4984 ** In this version, we take the most frequent REDUCE action and make
4985 ** it the default. Except, there is no default if the wildcard token
4986 ** is a possible look-ahead.
4987 */
CompressTables(struct lemon * lemp)4988 void CompressTables(struct lemon *lemp)
4989 {
4990 struct state *stp;
4991 struct action *ap, *ap2, *nextap;
4992 struct rule *rp, *rp2, *rbest;
4993 int nbest, n;
4994 int i;
4995 int usesWildcard;
4996
4997 for(i=0; i<lemp->nstate; i++){
4998 stp = lemp->sorted[i];
4999 nbest = 0;
5000 rbest = 0;
5001 usesWildcard = 0;
5002
5003 for(ap=stp->ap; ap; ap=ap->next){
5004 if( ap->type==SHIFT && ap->sp==lemp->wildcard ){
5005 usesWildcard = 1;
5006 }
5007 if( ap->type!=REDUCE ) continue;
5008 rp = ap->x.rp;
5009 if( rp->lhsStart ) continue;
5010 if( rp==rbest ) continue;
5011 n = 1;
5012 for(ap2=ap->next; ap2; ap2=ap2->next){
5013 if( ap2->type!=REDUCE ) continue;
5014 rp2 = ap2->x.rp;
5015 if( rp2==rbest ) continue;
5016 if( rp2==rp ) n++;
5017 }
5018 if( n>nbest ){
5019 nbest = n;
5020 rbest = rp;
5021 }
5022 }
5023
5024 /* Do not make a default if the number of rules to default
5025 ** is not at least 1 or if the wildcard token is a possible
5026 ** lookahead.
5027 */
5028 if( nbest<1 || usesWildcard ) continue;
5029
5030
5031 /* Combine matching REDUCE actions into a single default */
5032 for(ap=stp->ap; ap; ap=ap->next){
5033 if( ap->type==REDUCE && ap->x.rp==rbest ) break;
5034 }
5035 assert( ap );
5036 ap->sp = Symbol_new("{default}");
5037 for(ap=ap->next; ap; ap=ap->next){
5038 if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
5039 }
5040 stp->ap = Action_sort(stp->ap);
5041
5042 for(ap=stp->ap; ap; ap=ap->next){
5043 if( ap->type==SHIFT ) break;
5044 if( ap->type==REDUCE && ap->x.rp!=rbest ) break;
5045 }
5046 if( ap==0 ){
5047 stp->autoReduce = 1;
5048 stp->pDfltReduce = rbest;
5049 }
5050 }
5051
5052 /* Make a second pass over all states and actions. Convert
5053 ** every action that is a SHIFT to an autoReduce state into
5054 ** a SHIFTREDUCE action.
5055 */
5056 for(i=0; i<lemp->nstate; i++){
5057 stp = lemp->sorted[i];
5058 for(ap=stp->ap; ap; ap=ap->next){
5059 struct state *pNextState;
5060 if( ap->type!=SHIFT ) continue;
5061 pNextState = ap->x.stp;
5062 if( pNextState->autoReduce && pNextState->pDfltReduce!=0 ){
5063 ap->type = SHIFTREDUCE;
5064 ap->x.rp = pNextState->pDfltReduce;
5065 }
5066 }
5067 }
5068
5069 /* If a SHIFTREDUCE action specifies a rule that has a single RHS term
5070 ** (meaning that the SHIFTREDUCE will land back in the state where it
5071 ** started) and if there is no C-code associated with the reduce action,
5072 ** then we can go ahead and convert the action to be the same as the
5073 ** action for the RHS of the rule.
5074 */
5075 for(i=0; i<lemp->nstate; i++){
5076 stp = lemp->sorted[i];
5077 for(ap=stp->ap; ap; ap=nextap){
5078 nextap = ap->next;
5079 if( ap->type!=SHIFTREDUCE ) continue;
5080 rp = ap->x.rp;
5081 if( rp->noCode==0 ) continue;
5082 if( rp->nrhs!=1 ) continue;
5083 #if 1
5084 /* Only apply this optimization to non-terminals. It would be OK to
5085 ** apply it to terminal symbols too, but that makes the parser tables
5086 ** larger. */
5087 if( ap->sp->index<lemp->nterminal ) continue;
5088 #endif
5089 /* If we reach this point, it means the optimization can be applied */
5090 nextap = ap;
5091 for(ap2=stp->ap; ap2 && (ap2==ap || ap2->sp!=rp->lhs); ap2=ap2->next){}
5092 assert( ap2!=0 );
5093 ap->spOpt = ap2->sp;
5094 ap->type = ap2->type;
5095 ap->x = ap2->x;
5096 }
5097 }
5098 }
5099
5100
5101 /*
5102 ** Compare two states for sorting purposes. The smaller state is the
5103 ** one with the most non-terminal actions. If they have the same number
5104 ** of non-terminal actions, then the smaller is the one with the most
5105 ** token actions.
5106 */
stateResortCompare(const void * a,const void * b)5107 static int stateResortCompare(const void *a, const void *b){
5108 const struct state *pA = *(const struct state**)a;
5109 const struct state *pB = *(const struct state**)b;
5110 int n;
5111
5112 n = pB->nNtAct - pA->nNtAct;
5113 if( n==0 ){
5114 n = pB->nTknAct - pA->nTknAct;
5115 if( n==0 ){
5116 n = pB->statenum - pA->statenum;
5117 }
5118 }
5119 assert( n!=0 );
5120 return n;
5121 }
5122
5123
5124 /*
5125 ** Renumber and resort states so that states with fewer choices
5126 ** occur at the end. Except, keep state 0 as the first state.
5127 */
ResortStates(struct lemon * lemp)5128 void ResortStates(struct lemon *lemp)
5129 {
5130 int i;
5131 struct state *stp;
5132 struct action *ap;
5133
5134 for(i=0; i<lemp->nstate; i++){
5135 stp = lemp->sorted[i];
5136 stp->nTknAct = stp->nNtAct = 0;
5137 stp->iDfltReduce = -1; /* Init dflt action to "syntax error" */
5138 stp->iTknOfst = NO_OFFSET;
5139 stp->iNtOfst = NO_OFFSET;
5140 for(ap=stp->ap; ap; ap=ap->next){
5141 int iAction = compute_action(lemp,ap);
5142 if( iAction>=0 ){
5143 if( ap->sp->index<lemp->nterminal ){
5144 stp->nTknAct++;
5145 }else if( ap->sp->index<lemp->nsymbol ){
5146 stp->nNtAct++;
5147 }else{
5148 assert( stp->autoReduce==0 || stp->pDfltReduce==ap->x.rp );
5149 stp->iDfltReduce = iAction;
5150 }
5151 }
5152 }
5153 }
5154 qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]),
5155 stateResortCompare);
5156 for(i=0; i<lemp->nstate; i++){
5157 lemp->sorted[i]->statenum = i;
5158 }
5159 lemp->nxstate = lemp->nstate;
5160 while( lemp->nxstate>1 && lemp->sorted[lemp->nxstate-1]->autoReduce ){
5161 lemp->nxstate--;
5162 }
5163 }
5164
5165
5166 /***************** From the file "set.c" ************************************/
5167 /*
5168 ** Set manipulation routines for the LEMON parser generator.
5169 */
5170
5171 static int size = 0;
5172
5173 /* Set the set size */
SetSize(int n)5174 void SetSize(int n)
5175 {
5176 size = n+1;
5177 }
5178
5179 /* Allocate a new set */
SetNew(void)5180 char *SetNew(void){
5181 char *s;
5182 s = (char*)calloc( size, 1);
5183 if( s==0 ){
5184 memory_error();
5185 }
5186 return s;
5187 }
5188
5189 /* Deallocate a set */
SetFree(char * s)5190 void SetFree(char *s)
5191 {
5192 free(s);
5193 }
5194
5195 /* Add a new element to the set. Return TRUE if the element was added
5196 ** and FALSE if it was already there. */
SetAdd(char * s,int e)5197 int SetAdd(char *s, int e)
5198 {
5199 int rv;
5200 assert( e>=0 && e<size );
5201 rv = s[e];
5202 s[e] = 1;
5203 return !rv;
5204 }
5205
5206 /* Add every element of s2 to s1. Return TRUE if s1 changes. */
SetUnion(char * s1,char * s2)5207 int SetUnion(char *s1, char *s2)
5208 {
5209 int i, progress;
5210 progress = 0;
5211 for(i=0; i<size; i++){
5212 if( s2[i]==0 ) continue;
5213 if( s1[i]==0 ){
5214 progress = 1;
5215 s1[i] = 1;
5216 }
5217 }
5218 return progress;
5219 }
5220 /********************** From the file "table.c" ****************************/
5221 /*
5222 ** All code in this file has been automatically generated
5223 ** from a specification in the file
5224 ** "table.q"
5225 ** by the associative array code building program "aagen".
5226 ** Do not edit this file! Instead, edit the specification
5227 ** file, then rerun aagen.
5228 */
5229 /*
5230 ** Code for processing tables in the LEMON parser generator.
5231 */
5232
strhash(const char * x)5233 PRIVATE unsigned strhash(const char *x)
5234 {
5235 unsigned h = 0;
5236 while( *x ) h = h*13 + *(x++);
5237 return h;
5238 }
5239
5240 /* Works like strdup, sort of. Save a string in malloced memory, but
5241 ** keep strings in a table so that the same string is not in more
5242 ** than one place.
5243 */
Strsafe(const char * y)5244 const char *Strsafe(const char *y)
5245 {
5246 const char *z;
5247 char *cpy;
5248
5249 if( y==0 ) return 0;
5250 z = Strsafe_find(y);
5251 if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){
5252 lemon_strcpy(cpy,y);
5253 z = cpy;
5254 Strsafe_insert(z);
5255 }
5256 MemoryCheck(z);
5257 return z;
5258 }
5259
5260 /* There is one instance of the following structure for each
5261 ** associative array of type "x1".
5262 */
5263 struct s_x1 {
5264 int size; /* The number of available slots. */
5265 /* Must be a power of 2 greater than or */
5266 /* equal to 1 */
5267 int count; /* Number of currently slots filled */
5268 struct s_x1node *tbl; /* The data stored here */
5269 struct s_x1node **ht; /* Hash table for lookups */
5270 };
5271
5272 /* There is one instance of this structure for every data element
5273 ** in an associative array of type "x1".
5274 */
5275 typedef struct s_x1node {
5276 const char *data; /* The data */
5277 struct s_x1node *next; /* Next entry with the same hash */
5278 struct s_x1node **from; /* Previous link */
5279 } x1node;
5280
5281 /* There is only one instance of the array, which is the following */
5282 static struct s_x1 *x1a;
5283
5284 /* Allocate a new associative array */
Strsafe_init(void)5285 void Strsafe_init(void){
5286 if( x1a ) return;
5287 x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
5288 if( x1a ){
5289 x1a->size = 1024;
5290 x1a->count = 0;
5291 x1a->tbl = (x1node*)calloc(1024, sizeof(x1node) + sizeof(x1node*));
5292 if( x1a->tbl==0 ){
5293 free(x1a);
5294 x1a = 0;
5295 }else{
5296 int i;
5297 x1a->ht = (x1node**)&(x1a->tbl[1024]);
5298 for(i=0; i<1024; i++) x1a->ht[i] = 0;
5299 }
5300 }
5301 }
5302 /* Insert a new record into the array. Return TRUE if successful.
5303 ** Prior data with the same key is NOT overwritten */
Strsafe_insert(const char * data)5304 int Strsafe_insert(const char *data)
5305 {
5306 x1node *np;
5307 unsigned h;
5308 unsigned ph;
5309
5310 if( x1a==0 ) return 0;
5311 ph = strhash(data);
5312 h = ph & (x1a->size-1);
5313 np = x1a->ht[h];
5314 while( np ){
5315 if( strcmp(np->data,data)==0 ){
5316 /* An existing entry with the same key is found. */
5317 /* Fail because overwrite is not allows. */
5318 return 0;
5319 }
5320 np = np->next;
5321 }
5322 if( x1a->count>=x1a->size ){
5323 /* Need to make the hash table bigger */
5324 int i,arrSize;
5325 struct s_x1 array;
5326 array.size = arrSize = x1a->size*2;
5327 array.count = x1a->count;
5328 array.tbl = (x1node*)calloc(arrSize, sizeof(x1node) + sizeof(x1node*));
5329 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5330 array.ht = (x1node**)&(array.tbl[arrSize]);
5331 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5332 for(i=0; i<x1a->count; i++){
5333 x1node *oldnp, *newnp;
5334 oldnp = &(x1a->tbl[i]);
5335 h = strhash(oldnp->data) & (arrSize-1);
5336 newnp = &(array.tbl[i]);
5337 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5338 newnp->next = array.ht[h];
5339 newnp->data = oldnp->data;
5340 newnp->from = &(array.ht[h]);
5341 array.ht[h] = newnp;
5342 }
5343 /* free(x1a->tbl); // This program was originally for 16-bit machines.
5344 ** Don't worry about freeing memory on modern platforms. */
5345 *x1a = array;
5346 }
5347 /* Insert the new data */
5348 h = ph & (x1a->size-1);
5349 np = &(x1a->tbl[x1a->count++]);
5350 np->data = data;
5351 if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
5352 np->next = x1a->ht[h];
5353 x1a->ht[h] = np;
5354 np->from = &(x1a->ht[h]);
5355 return 1;
5356 }
5357
5358 /* Return a pointer to data assigned to the given key. Return NULL
5359 ** if no such key. */
Strsafe_find(const char * key)5360 const char *Strsafe_find(const char *key)
5361 {
5362 unsigned h;
5363 x1node *np;
5364
5365 if( x1a==0 ) return 0;
5366 h = strhash(key) & (x1a->size-1);
5367 np = x1a->ht[h];
5368 while( np ){
5369 if( strcmp(np->data,key)==0 ) break;
5370 np = np->next;
5371 }
5372 return np ? np->data : 0;
5373 }
5374
5375 /* Return a pointer to the (terminal or nonterminal) symbol "x".
5376 ** Create a new symbol if this is the first time "x" has been seen.
5377 */
Symbol_new(const char * x)5378 struct symbol *Symbol_new(const char *x)
5379 {
5380 struct symbol *sp;
5381
5382 sp = Symbol_find(x);
5383 if( sp==0 ){
5384 sp = (struct symbol *)calloc(1, sizeof(struct symbol) );
5385 MemoryCheck(sp);
5386 sp->name = Strsafe(x);
5387 sp->type = ISUPPER(*x) ? TERMINAL : NONTERMINAL;
5388 sp->rule = 0;
5389 sp->fallback = 0;
5390 sp->prec = -1;
5391 sp->assoc = UNK;
5392 sp->firstset = 0;
5393 sp->lambda = LEMON_FALSE;
5394 sp->destructor = 0;
5395 sp->destLineno = 0;
5396 sp->datatype = 0;
5397 sp->useCnt = 0;
5398 Symbol_insert(sp,sp->name);
5399 }
5400 sp->useCnt++;
5401 return sp;
5402 }
5403
5404 /* Compare two symbols for sorting purposes. Return negative,
5405 ** zero, or positive if a is less then, equal to, or greater
5406 ** than b.
5407 **
5408 ** Symbols that begin with upper case letters (terminals or tokens)
5409 ** must sort before symbols that begin with lower case letters
5410 ** (non-terminals). And MULTITERMINAL symbols (created using the
5411 ** %token_class directive) must sort at the very end. Other than
5412 ** that, the order does not matter.
5413 **
5414 ** We find experimentally that leaving the symbols in their original
5415 ** order (the order they appeared in the grammar file) gives the
5416 ** smallest parser tables in SQLite.
5417 */
Symbolcmpp(const void * _a,const void * _b)5418 int Symbolcmpp(const void *_a, const void *_b)
5419 {
5420 const struct symbol *a = *(const struct symbol **) _a;
5421 const struct symbol *b = *(const struct symbol **) _b;
5422 int i1 = a->type==MULTITERMINAL ? 3 : a->name[0]>'Z' ? 2 : 1;
5423 int i2 = b->type==MULTITERMINAL ? 3 : b->name[0]>'Z' ? 2 : 1;
5424 return i1==i2 ? a->index - b->index : i1 - i2;
5425 }
5426
5427 /* There is one instance of the following structure for each
5428 ** associative array of type "x2".
5429 */
5430 struct s_x2 {
5431 int size; /* The number of available slots. */
5432 /* Must be a power of 2 greater than or */
5433 /* equal to 1 */
5434 int count; /* Number of currently slots filled */
5435 struct s_x2node *tbl; /* The data stored here */
5436 struct s_x2node **ht; /* Hash table for lookups */
5437 };
5438
5439 /* There is one instance of this structure for every data element
5440 ** in an associative array of type "x2".
5441 */
5442 typedef struct s_x2node {
5443 struct symbol *data; /* The data */
5444 const char *key; /* The key */
5445 struct s_x2node *next; /* Next entry with the same hash */
5446 struct s_x2node **from; /* Previous link */
5447 } x2node;
5448
5449 /* There is only one instance of the array, which is the following */
5450 static struct s_x2 *x2a;
5451
5452 /* Allocate a new associative array */
Symbol_init(void)5453 void Symbol_init(void){
5454 if( x2a ) return;
5455 x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
5456 if( x2a ){
5457 x2a->size = 128;
5458 x2a->count = 0;
5459 x2a->tbl = (x2node*)calloc(128, sizeof(x2node) + sizeof(x2node*));
5460 if( x2a->tbl==0 ){
5461 free(x2a);
5462 x2a = 0;
5463 }else{
5464 int i;
5465 x2a->ht = (x2node**)&(x2a->tbl[128]);
5466 for(i=0; i<128; i++) x2a->ht[i] = 0;
5467 }
5468 }
5469 }
5470 /* Insert a new record into the array. Return TRUE if successful.
5471 ** Prior data with the same key is NOT overwritten */
Symbol_insert(struct symbol * data,const char * key)5472 int Symbol_insert(struct symbol *data, const char *key)
5473 {
5474 x2node *np;
5475 unsigned h;
5476 unsigned ph;
5477
5478 if( x2a==0 ) return 0;
5479 ph = strhash(key);
5480 h = ph & (x2a->size-1);
5481 np = x2a->ht[h];
5482 while( np ){
5483 if( strcmp(np->key,key)==0 ){
5484 /* An existing entry with the same key is found. */
5485 /* Fail because overwrite is not allows. */
5486 return 0;
5487 }
5488 np = np->next;
5489 }
5490 if( x2a->count>=x2a->size ){
5491 /* Need to make the hash table bigger */
5492 int i,arrSize;
5493 struct s_x2 array;
5494 array.size = arrSize = x2a->size*2;
5495 array.count = x2a->count;
5496 array.tbl = (x2node*)calloc(arrSize, sizeof(x2node) + sizeof(x2node*));
5497 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5498 array.ht = (x2node**)&(array.tbl[arrSize]);
5499 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5500 for(i=0; i<x2a->count; i++){
5501 x2node *oldnp, *newnp;
5502 oldnp = &(x2a->tbl[i]);
5503 h = strhash(oldnp->key) & (arrSize-1);
5504 newnp = &(array.tbl[i]);
5505 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5506 newnp->next = array.ht[h];
5507 newnp->key = oldnp->key;
5508 newnp->data = oldnp->data;
5509 newnp->from = &(array.ht[h]);
5510 array.ht[h] = newnp;
5511 }
5512 /* free(x2a->tbl); // This program was originally written for 16-bit
5513 ** machines. Don't worry about freeing this trivial amount of memory
5514 ** on modern platforms. Just leak it. */
5515 *x2a = array;
5516 }
5517 /* Insert the new data */
5518 h = ph & (x2a->size-1);
5519 np = &(x2a->tbl[x2a->count++]);
5520 np->key = key;
5521 np->data = data;
5522 if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
5523 np->next = x2a->ht[h];
5524 x2a->ht[h] = np;
5525 np->from = &(x2a->ht[h]);
5526 return 1;
5527 }
5528
5529 /* Return a pointer to data assigned to the given key. Return NULL
5530 ** if no such key. */
Symbol_find(const char * key)5531 struct symbol *Symbol_find(const char *key)
5532 {
5533 unsigned h;
5534 x2node *np;
5535
5536 if( x2a==0 ) return 0;
5537 h = strhash(key) & (x2a->size-1);
5538 np = x2a->ht[h];
5539 while( np ){
5540 if( strcmp(np->key,key)==0 ) break;
5541 np = np->next;
5542 }
5543 return np ? np->data : 0;
5544 }
5545
5546 /* Return the n-th data. Return NULL if n is out of range. */
Symbol_Nth(int n)5547 struct symbol *Symbol_Nth(int n)
5548 {
5549 struct symbol *data;
5550 if( x2a && n>0 && n<=x2a->count ){
5551 data = x2a->tbl[n-1].data;
5552 }else{
5553 data = 0;
5554 }
5555 return data;
5556 }
5557
5558 /* Return the size of the array */
Symbol_count()5559 int Symbol_count()
5560 {
5561 return x2a ? x2a->count : 0;
5562 }
5563
5564 /* Return an array of pointers to all data in the table.
5565 ** The array is obtained from malloc. Return NULL if memory allocation
5566 ** problems, or if the array is empty. */
Symbol_arrayof()5567 struct symbol **Symbol_arrayof()
5568 {
5569 struct symbol **array;
5570 int i,arrSize;
5571 if( x2a==0 ) return 0;
5572 arrSize = x2a->count;
5573 array = (struct symbol **)calloc(arrSize, sizeof(struct symbol *));
5574 if( array ){
5575 for(i=0; i<arrSize; i++) array[i] = x2a->tbl[i].data;
5576 }
5577 return array;
5578 }
5579
5580 /* Compare two configurations */
Configcmp(const char * _a,const char * _b)5581 int Configcmp(const char *_a,const char *_b)
5582 {
5583 const struct config *a = (struct config *) _a;
5584 const struct config *b = (struct config *) _b;
5585 int x;
5586 x = a->rp->index - b->rp->index;
5587 if( x==0 ) x = a->dot - b->dot;
5588 return x;
5589 }
5590
5591 /* Compare two states */
statecmp(struct config * a,struct config * b)5592 PRIVATE int statecmp(struct config *a, struct config *b)
5593 {
5594 int rc;
5595 for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){
5596 rc = a->rp->index - b->rp->index;
5597 if( rc==0 ) rc = a->dot - b->dot;
5598 }
5599 if( rc==0 ){
5600 if( a ) rc = 1;
5601 if( b ) rc = -1;
5602 }
5603 return rc;
5604 }
5605
5606 /* Hash a state */
statehash(struct config * a)5607 PRIVATE unsigned statehash(struct config *a)
5608 {
5609 unsigned h=0;
5610 while( a ){
5611 h = h*571 + a->rp->index*37 + a->dot;
5612 a = a->bp;
5613 }
5614 return h;
5615 }
5616
5617 /* Allocate a new state structure */
State_new()5618 struct state *State_new()
5619 {
5620 struct state *newstate;
5621 newstate = (struct state *)calloc(1, sizeof(struct state) );
5622 MemoryCheck(newstate);
5623 return newstate;
5624 }
5625
5626 /* There is one instance of the following structure for each
5627 ** associative array of type "x3".
5628 */
5629 struct s_x3 {
5630 int size; /* The number of available slots. */
5631 /* Must be a power of 2 greater than or */
5632 /* equal to 1 */
5633 int count; /* Number of currently slots filled */
5634 struct s_x3node *tbl; /* The data stored here */
5635 struct s_x3node **ht; /* Hash table for lookups */
5636 };
5637
5638 /* There is one instance of this structure for every data element
5639 ** in an associative array of type "x3".
5640 */
5641 typedef struct s_x3node {
5642 struct state *data; /* The data */
5643 struct config *key; /* The key */
5644 struct s_x3node *next; /* Next entry with the same hash */
5645 struct s_x3node **from; /* Previous link */
5646 } x3node;
5647
5648 /* There is only one instance of the array, which is the following */
5649 static struct s_x3 *x3a;
5650
5651 /* Allocate a new associative array */
State_init(void)5652 void State_init(void){
5653 if( x3a ) return;
5654 x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
5655 if( x3a ){
5656 x3a->size = 128;
5657 x3a->count = 0;
5658 x3a->tbl = (x3node*)calloc(128, sizeof(x3node) + sizeof(x3node*));
5659 if( x3a->tbl==0 ){
5660 free(x3a);
5661 x3a = 0;
5662 }else{
5663 int i;
5664 x3a->ht = (x3node**)&(x3a->tbl[128]);
5665 for(i=0; i<128; i++) x3a->ht[i] = 0;
5666 }
5667 }
5668 }
5669 /* Insert a new record into the array. Return TRUE if successful.
5670 ** Prior data with the same key is NOT overwritten */
State_insert(struct state * data,struct config * key)5671 int State_insert(struct state *data, struct config *key)
5672 {
5673 x3node *np;
5674 unsigned h;
5675 unsigned ph;
5676
5677 if( x3a==0 ) return 0;
5678 ph = statehash(key);
5679 h = ph & (x3a->size-1);
5680 np = x3a->ht[h];
5681 while( np ){
5682 if( statecmp(np->key,key)==0 ){
5683 /* An existing entry with the same key is found. */
5684 /* Fail because overwrite is not allows. */
5685 return 0;
5686 }
5687 np = np->next;
5688 }
5689 if( x3a->count>=x3a->size ){
5690 /* Need to make the hash table bigger */
5691 int i,arrSize;
5692 struct s_x3 array;
5693 array.size = arrSize = x3a->size*2;
5694 array.count = x3a->count;
5695 array.tbl = (x3node*)calloc(arrSize, sizeof(x3node) + sizeof(x3node*));
5696 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5697 array.ht = (x3node**)&(array.tbl[arrSize]);
5698 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5699 for(i=0; i<x3a->count; i++){
5700 x3node *oldnp, *newnp;
5701 oldnp = &(x3a->tbl[i]);
5702 h = statehash(oldnp->key) & (arrSize-1);
5703 newnp = &(array.tbl[i]);
5704 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5705 newnp->next = array.ht[h];
5706 newnp->key = oldnp->key;
5707 newnp->data = oldnp->data;
5708 newnp->from = &(array.ht[h]);
5709 array.ht[h] = newnp;
5710 }
5711 free(x3a->tbl);
5712 *x3a = array;
5713 }
5714 /* Insert the new data */
5715 h = ph & (x3a->size-1);
5716 np = &(x3a->tbl[x3a->count++]);
5717 np->key = key;
5718 np->data = data;
5719 if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
5720 np->next = x3a->ht[h];
5721 x3a->ht[h] = np;
5722 np->from = &(x3a->ht[h]);
5723 return 1;
5724 }
5725
5726 /* Return a pointer to data assigned to the given key. Return NULL
5727 ** if no such key. */
State_find(struct config * key)5728 struct state *State_find(struct config *key)
5729 {
5730 unsigned h;
5731 x3node *np;
5732
5733 if( x3a==0 ) return 0;
5734 h = statehash(key) & (x3a->size-1);
5735 np = x3a->ht[h];
5736 while( np ){
5737 if( statecmp(np->key,key)==0 ) break;
5738 np = np->next;
5739 }
5740 return np ? np->data : 0;
5741 }
5742
5743 /* Return an array of pointers to all data in the table.
5744 ** The array is obtained from malloc. Return NULL if memory allocation
5745 ** problems, or if the array is empty. */
State_arrayof(void)5746 struct state **State_arrayof(void)
5747 {
5748 struct state **array;
5749 int i,arrSize;
5750 if( x3a==0 ) return 0;
5751 arrSize = x3a->count;
5752 array = (struct state **)calloc(arrSize, sizeof(struct state *));
5753 if( array ){
5754 for(i=0; i<arrSize; i++) array[i] = x3a->tbl[i].data;
5755 }
5756 return array;
5757 }
5758
5759 /* Hash a configuration */
confighash(struct config * a)5760 PRIVATE unsigned confighash(struct config *a)
5761 {
5762 unsigned h=0;
5763 h = h*571 + a->rp->index*37 + a->dot;
5764 return h;
5765 }
5766
5767 /* There is one instance of the following structure for each
5768 ** associative array of type "x4".
5769 */
5770 struct s_x4 {
5771 int size; /* The number of available slots. */
5772 /* Must be a power of 2 greater than or */
5773 /* equal to 1 */
5774 int count; /* Number of currently slots filled */
5775 struct s_x4node *tbl; /* The data stored here */
5776 struct s_x4node **ht; /* Hash table for lookups */
5777 };
5778
5779 /* There is one instance of this structure for every data element
5780 ** in an associative array of type "x4".
5781 */
5782 typedef struct s_x4node {
5783 struct config *data; /* The data */
5784 struct s_x4node *next; /* Next entry with the same hash */
5785 struct s_x4node **from; /* Previous link */
5786 } x4node;
5787
5788 /* There is only one instance of the array, which is the following */
5789 static struct s_x4 *x4a;
5790
5791 /* Allocate a new associative array */
Configtable_init(void)5792 void Configtable_init(void){
5793 if( x4a ) return;
5794 x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
5795 if( x4a ){
5796 x4a->size = 64;
5797 x4a->count = 0;
5798 x4a->tbl = (x4node*)calloc(64, sizeof(x4node) + sizeof(x4node*));
5799 if( x4a->tbl==0 ){
5800 free(x4a);
5801 x4a = 0;
5802 }else{
5803 int i;
5804 x4a->ht = (x4node**)&(x4a->tbl[64]);
5805 for(i=0; i<64; i++) x4a->ht[i] = 0;
5806 }
5807 }
5808 }
5809 /* Insert a new record into the array. Return TRUE if successful.
5810 ** Prior data with the same key is NOT overwritten */
Configtable_insert(struct config * data)5811 int Configtable_insert(struct config *data)
5812 {
5813 x4node *np;
5814 unsigned h;
5815 unsigned ph;
5816
5817 if( x4a==0 ) return 0;
5818 ph = confighash(data);
5819 h = ph & (x4a->size-1);
5820 np = x4a->ht[h];
5821 while( np ){
5822 if( Configcmp((const char *) np->data,(const char *) data)==0 ){
5823 /* An existing entry with the same key is found. */
5824 /* Fail because overwrite is not allows. */
5825 return 0;
5826 }
5827 np = np->next;
5828 }
5829 if( x4a->count>=x4a->size ){
5830 /* Need to make the hash table bigger */
5831 int i,arrSize;
5832 struct s_x4 array;
5833 array.size = arrSize = x4a->size*2;
5834 array.count = x4a->count;
5835 array.tbl = (x4node*)calloc(arrSize, sizeof(x4node) + sizeof(x4node*));
5836 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
5837 array.ht = (x4node**)&(array.tbl[arrSize]);
5838 for(i=0; i<arrSize; i++) array.ht[i] = 0;
5839 for(i=0; i<x4a->count; i++){
5840 x4node *oldnp, *newnp;
5841 oldnp = &(x4a->tbl[i]);
5842 h = confighash(oldnp->data) & (arrSize-1);
5843 newnp = &(array.tbl[i]);
5844 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5845 newnp->next = array.ht[h];
5846 newnp->data = oldnp->data;
5847 newnp->from = &(array.ht[h]);
5848 array.ht[h] = newnp;
5849 }
5850 /* free(x4a->tbl); // This code was originall written for 16-bit machines.
5851 ** on modern machines, don't worry about freeing this trival amount of
5852 ** memory. */
5853 *x4a = array;
5854 }
5855 /* Insert the new data */
5856 h = ph & (x4a->size-1);
5857 np = &(x4a->tbl[x4a->count++]);
5858 np->data = data;
5859 if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
5860 np->next = x4a->ht[h];
5861 x4a->ht[h] = np;
5862 np->from = &(x4a->ht[h]);
5863 return 1;
5864 }
5865
5866 /* Return a pointer to data assigned to the given key. Return NULL
5867 ** if no such key. */
Configtable_find(struct config * key)5868 struct config *Configtable_find(struct config *key)
5869 {
5870 int h;
5871 x4node *np;
5872
5873 if( x4a==0 ) return 0;
5874 h = confighash(key) & (x4a->size-1);
5875 np = x4a->ht[h];
5876 while( np ){
5877 if( Configcmp((const char *) np->data,(const char *) key)==0 ) break;
5878 np = np->next;
5879 }
5880 return np ? np->data : 0;
5881 }
5882
5883 /* Remove all data from the table. Pass each data to the function "f"
5884 ** as it is removed. ("f" may be null to avoid this step.) */
Configtable_clear(int (* f)(struct config *))5885 void Configtable_clear(int(*f)(struct config *))
5886 {
5887 int i;
5888 if( x4a==0 || x4a->count==0 ) return;
5889 if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
5890 for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;
5891 x4a->count = 0;
5892 return;
5893 }
5894