1 #ifndef _GNU_SOURCE
2 #define _GNU_SOURCE
3 #endif
4 
5 #ifdef __COVERITY__
6 #define _Float128 long double
7 #define _Float64x long double
8 #define _Float64  double
9 #define _Float32x double
10 #define _Float32  float
11 #endif
12 
13 /*
14 ** This file contains all sources (including headers) to the LEMON
15 ** LALR(1) parser generator.  The sources have been combined into a
16 ** single file to make it easy to include LEMON in the source tree
17 ** and Makefile of another program.
18 **
19 ** The author of this program disclaims copyright.
20 */
21 #include <stdio.h>
22 #include <stdarg.h>
23 #include <string.h>
24 #include <ctype.h>
25 #include <stdlib.h>
26 #include <inttypes.h>
27 #include <unistd.h>    /* access() */
28 
29 #define UNUSED(x) ( (void)(x) )
30 
31 #ifndef __WIN32__
32 #   if defined(_WIN32) || defined(WIN32)
33 #	define __WIN32__
34 #   endif
35 #endif
36 
37 #if __GNUC__ > 2
38 #define NORETURN __attribute__ ((__noreturn__))
39 #else
40 #define NORETURN
41 #endif
42 
43 /* #define PRIVATE static */
44 #define PRIVATE static
45 
46 #ifdef TEST
47 #define MAXRHS 5       /* Set low to exercise exception code */
48 #else
49 #define MAXRHS 1000
50 #endif
51 
52 void *msort(void *list, void **next, int(*cmp)(void *, void *));
53 
54 static void memory_error() NORETURN;
55 
56 /******** From the file "action.h" *************************************/
57 struct action *Action_new();
58 struct action *Action_sort();
59 void Action_add();
60 
61 /********* From the file "assert.h" ************************************/
62 void myassert() NORETURN;
63 #ifndef NDEBUG
64 #  define assert(X) if(!(X))myassert(__FILE__,__LINE__)
65 #else
66 #  define assert(X)
67 #endif
68 
69 /********** From the file "build.h" ************************************/
70 void FindRulePrecedences();
71 void FindFirstSets();
72 void FindStates();
73 void FindLinks();
74 void FindFollowSets();
75 void FindActions();
76 
77 /********* From the file "configlist.h" *********************************/
78 void Configlist_init(/* void */);
79 struct config *Configlist_add(/* struct rule *, int */);
80 struct config *Configlist_addbasis(/* struct rule *, int */);
81 void Configlist_closure(/* void */);
82 void Configlist_sort(/* void */);
83 void Configlist_sortbasis(/* void */);
84 struct config *Configlist_return(/* void */);
85 struct config *Configlist_basis(/* void */);
86 void Configlist_eat(/* struct config * */);
87 void Configlist_reset(/* void */);
88 
89 /********* From the file "error.h" ***************************************/
90 void ErrorMsg(const char *, int,const char *, ...);
91 
92 /****** From the file "option.h" ******************************************/
93 struct s_options {
94   enum { OPT_FLAG=1,  OPT_INT,  OPT_DBL,  OPT_STR,
95          OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR} type;
96   char *label;
97   void *arg;
98   const char *message;
99 };
100 int    OptInit(/* char**,struct s_options*,FILE* */);
101 int    OptNArgs(/* void */);
102 char  *OptArg(/* int */);
103 void   OptErr(/* int */);
104 void   OptPrint(/* void */);
105 
106 /******** From the file "parse.h" *****************************************/
107 void Parse(/* struct lemon *lemp */);
108 
109 /********* From the file "plink.h" ***************************************/
110 struct plink *Plink_new(/* void */);
111 void Plink_add(/* struct plink **, struct config * */);
112 void Plink_copy(/* struct plink **, struct plink * */);
113 void Plink_delete(/* struct plink * */);
114 
115 /********** From the file "report.h" *************************************/
116 void Reprint(/* struct lemon * */);
117 void ReportOutput(/* struct lemon * */);
118 void ReportTable(/* struct lemon * */);
119 void ReportHeader(/* struct lemon * */);
120 void CompressTables(/* struct lemon * */);
121 
122 /********** From the file "set.h" ****************************************/
123 void  SetSize(/* int N */);             /* All sets will be of size N */
124 char *SetNew(/* void */);               /* A new set for element 0..N */
125 void  SetFree(/* char* */);             /* Deallocate a set */
126 
127 int SetAdd(/* char*,int */);            /* Add element to a set */
128 int SetUnion(/* char *A,char *B */);    /* A <- A U B, through element N */
129 
130 #define SetFind(X,Y) (X[Y])       /* True if Y is in set X */
131 
132 /********** From the file "struct.h" *************************************/
133 /*
134 ** Principal data structures for the LEMON parser generator.
135 */
136 
137 typedef enum {Bo_FALSE=0, Bo_TRUE} Boolean;
138 
139 /* Symbols (terminals and nonterminals) of the grammar are stored
140 ** in the following: */
141 struct symbol {
142   char *name;              /* Name of the symbol */
143   int index;               /* Index number for this symbol */
144   enum {
145     TERMINAL,
146     NONTERMINAL
147   } type;                  /* Symbols are all either TERMINALS or NTs */
148   struct rule *rule;       /* Linked list of rules of this (if an NT) */
149   struct symbol *fallback; /* fallback token in case this token doesn't parse */
150   int prec;                /* Precedence if defined (-1 otherwise) */
151   enum e_assoc {
152     LEFT,
153     RIGHT,
154     NONE,
155     UNK
156   } assoc;                 /* Associativity if predecence is defined */
157   char *firstset;          /* First-set for all rules of this symbol */
158   Boolean lambda;          /* True if NT and can generate an empty string */
159   char *destructor;        /* Code which executes whenever this symbol is
160                            ** popped from the stack during error processing */
161   int destructorln;        /* Line number of destructor code */
162   char *datatype;          /* The data type of information held by this
163                            ** object. Only used if type==NONTERMINAL */
164   int dtnum;               /* The data type number.  In the parser, the value
165                            ** stack is a union.  The .yy%d element of this
166                            ** union is the correct data type for this object */
167 };
168 
169 /* Each production rule in the grammar is stored in the following
170 ** structure.  */
171 struct rule {
172   struct symbol *lhs;      /* Left-hand side of the rule */
173   char *lhsalias;          /* Alias for the LHS (NULL if none) */
174   int ruleline;            /* Line number for the rule */
175   int nrhs;                /* Number of RHS symbols */
176   struct symbol **rhs;     /* The RHS symbols */
177   char **rhsalias;         /* An alias for each RHS symbol (NULL if none) */
178   int line;                /* Line number at which code begins */
179   char *code;              /* The code executed when this rule is reduced */
180   struct symbol *precsym;  /* Precedence symbol for this rule */
181   int index;               /* An index number for this rule */
182   Boolean canReduce;       /* True if this rule is ever reduced */
183   struct rule *nextlhs;    /* Next rule with the same LHS */
184   struct rule *next;       /* Next rule in the global list */
185 };
186 
187 /* A configuration is a production rule of the grammar together with
188 ** a mark (dot) showing how much of that rule has been processed so far.
189 ** Configurations also contain a follow-set which is a list of terminal
190 ** symbols which are allowed to immediately follow the end of the rule.
191 ** Every configuration is recorded as an instance of the following: */
192 struct config {
193   struct rule *rp;         /* The rule upon which the configuration is based */
194   int dot;                 /* The parse point */
195   char *fws;               /* Follow-set for this configuration only */
196   struct plink *fplp;      /* Follow-set forward propagation links */
197   struct plink *bplp;      /* Follow-set backwards propagation links */
198   struct state *stp;       /* Pointer to state which contains this */
199   enum {
200     COMPLETE,              /* The status is used during followset and */
201     INCOMPLETE             /*    shift computations */
202   } status;
203   struct config *next;     /* Next configuration in the state */
204   struct config *bp;       /* The next basis configuration */
205 };
206 
207 /* Every shift or reduce operation is stored as one of the following */
208 struct action {
209   struct symbol *sp;       /* The look-ahead symbol */
210   enum e_action {
211     SHIFT,
212     ACCEPT,
213     REDUCE,
214     ERROR,
215     CONFLICT,                /* Was a reduce, but part of a conflict */
216     SH_RESOLVED,             /* Was a shift.  Precedence resolved conflict */
217     RD_RESOLVED,             /* Was reduce.  Precedence resolved conflict */
218     NOT_USED                 /* Deleted by compression */
219   } type;
220   union {
221     struct state *stp;     /* The new state, if a shift */
222     struct rule *rp;       /* The rule, if a reduce */
223   } x;
224   struct action *next;     /* Next action for this state */
225   struct action *collide;  /* Next action with the same hash */
226 };
227 
228 /* Each state of the generated parser's finite state machine
229 ** is encoded as an instance of the following structure. */
230 struct state {
231   struct config *bp;       /* The basis configurations for this state */
232   struct config *cfp;      /* All configurations in this set */
233   int index;               /* Sequential number for this state */
234   struct action *ap;       /* Array of actions for this state */
235   int nTknAct, nNtAct;     /* Number of actions on terminals and nonterminals */
236   int iTknOfst, iNtOfst;   /* yy_action[] offset for terminals and nonterms */
237   int iDflt;               /* Default action */
238 };
239 #define NO_OFFSET (-2147483647)
240 
241 /* A followset propagation link indicates that the contents of one
242 ** configuration followset should be propagated to another whenever
243 ** the first changes. */
244 struct plink {
245   struct config *cfp;      /* The configuration to which linked */
246   struct plink *next;      /* The next propagate link */
247 };
248 
249 /* The state vector for the entire parser generator is recorded as
250 ** follows.  (LEMON uses no global variables and makes little use of
251 ** static variables.  Fields in the following structure can be thought
252 ** of as begin global variables in the program.) */
253 struct lemon {
254   struct state **sorted;   /* Table of states sorted by state number */
255   struct rule *rule;       /* List of all rules */
256   int nstate;              /* Number of states */
257   int nrule;               /* Number of rules */
258   int nsymbol;             /* Number of terminal and nonterminal symbols */
259   int nterminal;           /* Number of terminal symbols */
260   struct symbol **symbols; /* Sorted array of pointers to symbols */
261   int errorcnt;            /* Number of errors */
262   struct symbol *errsym;   /* The error symbol */
263   char *name;              /* Name of the generated parser */
264   char *arg;               /* Declaration of the 3th argument to parser */
265   char *tokentype;         /* Type of terminal symbols in the parser stack */
266   char *vartype;           /* The default type of non-terminal symbols */
267   char *start;             /* Name of the start symbol for the grammar */
268   char *stacksize;         /* Size of the parser stack */
269   char *include;           /* Code to put at the start of the C file */
270   int  includeln;          /* Line number for start of include code */
271   char *error;             /* Code to execute when an error is seen */
272   int  errorln;            /* Line number for start of error code */
273   char *overflow;          /* Code to execute on a stack overflow */
274   int  overflowln;         /* Line number for start of overflow code */
275   char *failure;           /* Code to execute on parser failure */
276   int  failureln;          /* Line number for start of failure code */
277   char *accept;            /* Code to execute when the parser excepts */
278   int  acceptln;           /* Line number for the start of accept code */
279   char *extracode;         /* Code appended to the generated file */
280   int  extracodeln;        /* Line number for the start of the extra code */
281   char *tokendest;         /* Code to execute to destroy token data */
282   int  tokendestln;        /* Line number for token destroyer code */
283   char *vardest;           /* Code for the default non-terminal destructor */
284   int  vardestln;          /* Line number for default non-term destructor code*/
285   char *filename;          /* Name of the input file */
286   char *tmplname;          /* Name of the template file */
287   char *outname;           /* Name of the current output file */
288   char *tokenprefix;       /* A prefix added to token names in the .h file */
289   int nconflict;           /* Number of parsing conflicts */
290   int tablesize;           /* Size of the parse tables */
291   int basisflag;           /* Print only basis configurations */
292   int has_fallback;        /* True if any %fallback is seen in the grammar */
293   char *argv0;             /* Name of the program */
294 };
295 
296 #define MemoryCheck(X) if((X)==0){ \
297   memory_error(); \
298 }
299 
300 /**************** From the file "table.h" *********************************/
301 /*
302 ** All code in this file has been automatically generated
303 ** from a specification in the file
304 **              "table.q"
305 ** by the associative array code building program "aagen".
306 ** Do not edit this file!  Instead, edit the specification
307 ** file, then rerun aagen.
308 */
309 /*
310 ** Code for processing tables in the LEMON parser generator.
311 */
312 
313 /* Routines for handling a strings */
314 
315 char *Strsafe();
316 
317 void Strsafe_init(/* void */);
318 int Strsafe_insert(/* char * */);
319 char *Strsafe_find(/* char * */);
320 
321 /* Routines for handling symbols of the grammar */
322 
323 struct symbol *Symbol_new();
324 int Symbolcmpp(/* struct symbol **, struct symbol ** */);
325 void Symbol_init(/* void */);
326 int Symbol_insert(/* struct symbol *, char * */);
327 struct symbol *Symbol_find(/* char * */);
328 struct symbol *Symbol_Nth(/* int */);
329 int Symbol_count(/*  */);
330 int State_count(void);
331 struct symbol **Symbol_arrayof(/*  */);
332 
333 /* Routines to manage the state table */
334 
335 int Configcmp(/* struct config *, struct config * */);
336 struct state *State_new();
337 void State_init(/* void */);
338 int State_insert(/* struct state *, struct config * */);
339 struct state *State_find(/* struct config * */);
340 struct state **State_arrayof(/*  */);
341 
342 /* Routines used for efficiency in Configlist_add */
343 
344 void Configtable_init(/* void */);
345 int Configtable_insert(/* struct config * */);
346 struct config *Configtable_find(/* struct config * */);
347 void Configtable_clear(/* int(*)(struct config *) */);
348 /****************** From the file "action.c" *******************************/
349 /*
350 ** Routines processing parser actions in the LEMON parser generator.
351 */
352 
353 /* Allocate a new parser action */
Action_new()354 struct action *Action_new(){
355   static struct action *freelist = NULL;
356   struct action *new;
357 
358   if( freelist==NULL ){
359     int i;
360     int amt = 100;
361     freelist = (struct action *)malloc( sizeof(struct action)*amt );
362     if( freelist==0 ){
363       fprintf(stderr,"Unable to allocate memory for a new parser action.");
364       exit(1);
365     }
366     for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
367     freelist[amt-1].next = 0;
368   }
369   new = freelist;
370   freelist = freelist->next;
371   return new;
372 }
373 
374 /* Compare two actions */
actioncmp(ap1,ap2)375 static int actioncmp(ap1,ap2)
376 struct action *ap1;
377 struct action *ap2;
378 {
379   int rc;
380   rc = ap1->sp->index - ap2->sp->index;
381   if( rc==0 ) rc = (int)ap1->type - (int)ap2->type;
382   if( rc==0 ){
383     assert( ap1->type==REDUCE || ap1->type==RD_RESOLVED || ap1->type==CONFLICT);
384     assert( ap2->type==REDUCE || ap2->type==RD_RESOLVED || ap2->type==CONFLICT);
385     rc = ap1->x.rp->index - ap2->x.rp->index;
386   }
387   return rc;
388 }
389 
390 /* Sort parser actions */
Action_sort(ap)391 struct action *Action_sort(ap)
392 struct action *ap;
393 {
394   ap = (struct action *)msort(ap,(void **)&ap->next,actioncmp);
395   return ap;
396 }
397 
Action_add(app,type,sp,arg)398 void Action_add(app,type,sp,arg)
399 struct action **app;
400 enum e_action type;
401 struct symbol *sp;
402 void *arg;
403 {
404   struct action *new;
405   new = Action_new();
406   new->next = *app;
407   *app = new;
408   new->type = type;
409   new->sp = sp;
410   if( type==SHIFT ){
411     new->x.stp = (struct state *)arg;
412   }else{
413     new->x.rp = (struct rule *)arg;
414   }
415 }
416 /********************** New code to implement the "acttab" module ***********/
417 /*
418 ** This module implements routines use to construct the yy_action[] table.
419 */
420 
421 /*
422 ** The state of the yy_action table under construction is an instance of
423 ** the following structure
424 */
425 typedef struct acttab acttab;
426 struct acttab {
427   int nAction;                 /* Number of used slots in aAction[] */
428   int nActionAlloc;            /* Slots allocated for aAction[] */
429   struct {
430     int lookahead;             /* Value of the lookahead token */
431     int action;                /* Action to take on the given lookahead */
432   } *aAction,                  /* The yy_action[] table under construction */
433     *aLookahead;               /* A single new transaction set */
434   int mnLookahead;             /* Minimum aLookahead[].lookahead */
435   int mnAction;                /* Action associated with mnLookahead */
436   int mxLookahead;             /* Maximum aLookahead[].lookahead */
437   int nLookahead;              /* Used slots in aLookahead[] */
438   int nLookaheadAlloc;         /* Slots allocated in aLookahead[] */
439 };
440 
441 /* Return the number of entries in the yy_action table */
442 #define acttab_size(X) ((X)->nAction)
443 
444 /* The value for the N-th entry in yy_action */
445 #define acttab_yyaction(X,N)  ((X)->aAction[N].action)
446 
447 /* The value for the N-th entry in yy_lookahead */
448 #define acttab_yylookahead(X,N)  ((X)->aAction[N].lookahead)
449 
450 /* Free all memory associated with the given acttab */
451 /*
452 PRIVATE void acttab_free(acttab *p){
453   free( p->aAction );
454   free( p->aLookahead );
455   free( p );
456 }
457 */
458 
459 /* Allocate a new acttab structure */
acttab_alloc(void)460 PRIVATE acttab *acttab_alloc(void){
461   acttab *p = malloc( sizeof(*p) );
462   if( p==0 ){
463     fprintf(stderr,"Unable to allocate memory for a new acttab.");
464     exit(1);
465   }
466   memset(p, 0, sizeof(*p));
467   return p;
468 }
469 
470 /* Add a new action to the current transaction set
471 */
acttab_action(acttab * p,int lookahead,int action)472 PRIVATE void acttab_action(acttab *p, int lookahead, int action){
473   if( p->nLookahead>=p->nLookaheadAlloc ){
474     p->nLookaheadAlloc += 25;
475     p->aLookahead = realloc( p->aLookahead,
476                              sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
477     if( p->aLookahead==0 ){
478       fprintf(stderr,"malloc failed\n");
479       exit(1);
480     }
481   }
482   if( p->nLookahead==0 ){
483     p->mxLookahead = lookahead;
484     p->mnLookahead = lookahead;
485     p->mnAction = action;
486   }else{
487     if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
488     if( p->mnLookahead>lookahead ){
489       p->mnLookahead = lookahead;
490       p->mnAction = action;
491     }
492   }
493   p->aLookahead[p->nLookahead].lookahead = lookahead;
494   p->aLookahead[p->nLookahead].action = action;
495   p->nLookahead++;
496 }
497 
498 /*
499 ** Add the transaction set built up with prior calls to acttab_action()
500 ** into the current action table.  Then reset the transaction set back
501 ** to an empty set in preparation for a new round of acttab_action() calls.
502 **
503 ** Return the offset into the action table of the new transaction.
504 */
acttab_insert(acttab * p)505 PRIVATE int acttab_insert(acttab *p){
506   int i, j, k, n;
507   assert( p->nLookahead>0 );
508 
509   /* Make sure we have enough space to hold the expanded action table
510   ** in the worst case.  The worst case occurs if the transaction set
511   ** must be appended to the current action table
512   */
513   n = p->mxLookahead + 1;
514   if( p->nAction + n >= p->nActionAlloc ){
515     int oldAlloc = p->nActionAlloc;
516     p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
517     p->aAction = realloc( p->aAction,
518                           sizeof(p->aAction[0])*p->nActionAlloc);
519     if( p->aAction==0 ){
520       fprintf(stderr,"malloc failed\n");
521       exit(1);
522     }
523     for(i=oldAlloc; i<p->nActionAlloc; i++){
524       p->aAction[i].lookahead = -1;
525       p->aAction[i].action = -1;
526     }
527   }
528 
529   /* Scan the existing action table looking for an offset where we can
530   ** insert the current transaction set.  Fall out of the loop when that
531   ** offset is found.  In the worst case, we fall out of the loop when
532   ** i reaches p->nAction, which means we append the new transaction set.
533   **
534   ** i is the index in p->aAction[] where p->mnLookahead is inserted.
535   */
536   for(i=0; i<p->nAction+p->mnLookahead; i++){
537     if( p->aAction[i].lookahead<0 ){
538       for(j=0; j<p->nLookahead; j++){
539         k = p->aLookahead[j].lookahead - p->mnLookahead + i;
540         if( k<0 ) break;
541         if( p->aAction[k].lookahead>=0 ) break;
542       }
543       if( j<p->nLookahead ) continue;
544       for(j=0; j<p->nAction; j++){
545         if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
546       }
547       if( j==p->nAction ){
548         break;  /* Fits in empty slots */
549       }
550     }else if( p->aAction[i].lookahead==p->mnLookahead ){
551       if( p->aAction[i].action!=p->mnAction ) continue;
552       for(j=0; j<p->nLookahead; j++){
553         k = p->aLookahead[j].lookahead - p->mnLookahead + i;
554         if( k<0 || k>=p->nAction ) break;
555         if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
556         if( p->aLookahead[j].action!=p->aAction[k].action ) break;
557       }
558       if( j<p->nLookahead ) continue;
559       n = 0;
560       for(j=0; j<p->nAction; j++){
561         if( p->aAction[j].lookahead<0 ) continue;
562         if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
563       }
564       if( n==p->nLookahead ){
565         break;  /* Same as a prior transaction set */
566       }
567     }
568   }
569   /* Insert transaction set at index i. */
570   for(j=0; j<p->nLookahead; j++){
571     k = p->aLookahead[j].lookahead - p->mnLookahead + i;
572     p->aAction[k] = p->aLookahead[j];
573     if( k>=p->nAction ) p->nAction = k+1;
574   }
575   p->nLookahead = 0;
576 
577   /* Return the offset that is added to the lookahead in order to get the
578   ** index into yy_action of the action */
579   return i - p->mnLookahead;
580 }
581 
582 /********************** From the file "assert.c" ****************************/
583 /*
584 ** A more efficient way of handling assertions.
585 */
myassert(file,line)586 void myassert(file,line)
587 char *file;
588 int line;
589 {
590   fprintf(stderr,"Assertion failed on line %d of file \"%s\"\n",line,file);
591   exit(1);
592 }
593 /********************** From the file "build.c" *****************************/
594 /*
595 ** Routines to construction the finite state machine for the LEMON
596 ** parser generator.
597 */
598 
599 /* Find a precedence symbol of every rule in the grammar.
600 **
601 ** Those rules which have a precedence symbol coded in the input
602 ** grammar using the "[symbol]" construct will already have the
603 ** rp->precsym field filled.  Other rules take as their precedence
604 ** symbol the first RHS symbol with a defined precedence.  If there
605 ** are not RHS symbols with a defined precedence, the precedence
606 ** symbol field is left blank.
607 */
FindRulePrecedences(xp)608 void FindRulePrecedences(xp)
609 struct lemon *xp;
610 {
611   struct rule *rp;
612   for(rp=xp->rule; rp; rp=rp->next){
613     if( rp->precsym==0 ){
614       int i;
615       for(i=0; i<rp->nrhs; i++){
616         if( rp->rhs[i]->prec>=0 ){
617           rp->precsym = rp->rhs[i];
618           break;
619 	}
620       }
621     }
622   }
623   return;
624 }
625 
626 /* Find all nonterminals which will generate the empty string.
627 ** Then go back and compute the first sets of every nonterminal.
628 ** The first set is the set of all terminal symbols which can begin
629 ** a string generated by that nonterminal.
630 */
FindFirstSets(lemp)631 void FindFirstSets(lemp)
632 struct lemon *lemp;
633 {
634   int i;
635   struct rule *rp;
636   int progress;
637 
638   for(i=0; i<lemp->nsymbol; i++){
639     lemp->symbols[i]->lambda = Bo_FALSE;
640   }
641   for(i=lemp->nterminal; i<lemp->nsymbol; i++){
642     lemp->symbols[i]->firstset = SetNew();
643   }
644 
645   /* First compute all lambdas */
646   do{
647     progress = 0;
648     for(rp=lemp->rule; rp; rp=rp->next){
649       if( rp->lhs->lambda ) continue;
650       for(i=0; i<rp->nrhs; i++){
651          if( rp->rhs[i]->lambda==Bo_FALSE ) break;
652       }
653       if( i==rp->nrhs ){
654         rp->lhs->lambda = Bo_TRUE;
655         progress = 1;
656       }
657     }
658   }while( progress );
659 
660   /* Now compute all first sets */
661   do{
662     struct symbol *s1, *s2;
663     progress = 0;
664     for(rp=lemp->rule; rp; rp=rp->next){
665       s1 = rp->lhs;
666       for(i=0; i<rp->nrhs; i++){
667         s2 = rp->rhs[i];
668         if( s2->type==TERMINAL ){
669           progress += SetAdd(s1->firstset,s2->index);
670           break;
671 	}else if( s1==s2 ){
672           if( s1->lambda==Bo_FALSE ) break;
673 	}else{
674           progress += SetUnion(s1->firstset,s2->firstset);
675           if( s2->lambda==Bo_FALSE ) break;
676 	}
677       }
678     }
679   }while( progress );
680   return;
681 }
682 
683 /* Compute all LR(0) states for the grammar.  Links
684 ** are added to between some states so that the LR(1) follow sets
685 ** can be computed later.
686 */
687 PRIVATE struct state *getstate(/* struct lemon * */);  /* forward reference */
FindStates(lemp)688 void FindStates(lemp)
689 struct lemon *lemp;
690 {
691   struct symbol *sp;
692   struct rule *rp;
693 
694   Configlist_init();
695 
696   /* Find the start symbol */
697   if( lemp->start ){
698     sp = Symbol_find(lemp->start);
699     if( sp==0 ){
700       ErrorMsg(lemp->filename,0,
701 "The specified start symbol \"%s\" is not \
702 in a nonterminal of the grammar.  \"%s\" will be used as the start \
703 symbol instead.",lemp->start,lemp->rule->lhs->name);
704       lemp->errorcnt++;
705       sp = lemp->rule->lhs;
706     }
707   }else{
708     sp = lemp->rule->lhs;
709   }
710 
711   /* Make sure the start symbol doesn't occur on the right-hand side of
712   ** any rule.  Report an error if it does.  (YACC would generate a new
713   ** start symbol in this case.) */
714   for(rp=lemp->rule; rp; rp=rp->next){
715     int i;
716     for(i=0; i<rp->nrhs; i++){
717       if( rp->rhs[i]==sp ){
718         ErrorMsg(lemp->filename,0,
719 "The start symbol \"%s\" occurs on the \
720 right-hand side of a rule. This will result in a parser which \
721 does not work properly.",sp->name);
722         lemp->errorcnt++;
723       }
724     }
725   }
726 
727   /* The basis configuration set for the first state
728   ** is all rules which have the start symbol as their
729   ** left-hand side */
730   for(rp=sp->rule; rp; rp=rp->nextlhs){
731     struct config *newcfp;
732     newcfp = Configlist_addbasis(rp,0);
733     SetAdd(newcfp->fws,0);
734   }
735 
736   /* Compute the first state.  All other states will be
737   ** computed automatically during the computation of the first one.
738   ** The returned pointer to the first state is not used. */
739   (void)getstate(lemp);
740   return;
741 }
742 
743 /* Return a pointer to a state which is described by the configuration
744 ** list which has been built from calls to Configlist_add.
745 */
746 PRIVATE void buildshifts(/* struct lemon *, struct state * */); /* Forwd ref */
getstate(lemp)747 PRIVATE struct state *getstate(lemp)
748 struct lemon *lemp;
749 {
750   struct config *cfp, *bp;
751   struct state *stp;
752 
753   /* Extract the sorted basis of the new state.  The basis was constructed
754   ** by prior calls to "Configlist_addbasis()". */
755   Configlist_sortbasis();
756   bp = Configlist_basis();
757 
758   /* Get a state with the same basis */
759   stp = State_find(bp);
760   if( stp ){
761     /* A state with the same basis already exists!  Copy all the follow-set
762     ** propagation links from the state under construction into the
763     ** preexisting state, then return a pointer to the preexisting state */
764     struct config *x, *y;
765     for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
766       Plink_copy(&y->bplp,x->bplp);
767       Plink_delete(x->fplp);
768       x->fplp = x->bplp = 0;
769     }
770     cfp = Configlist_return();
771     Configlist_eat(cfp);
772   }else{
773     /* This really is a new state.  Construct all the details */
774     Configlist_closure(lemp);    /* Compute the configuration closure */
775     Configlist_sort();           /* Sort the configuration closure */
776     cfp = Configlist_return();   /* Get a pointer to the config list */
777     stp = State_new();           /* A new state structure */
778     MemoryCheck(stp);
779     stp->bp = bp;                /* Remember the configuration basis */
780     stp->cfp = cfp;              /* Remember the configuration closure */
781     stp->index = lemp->nstate++; /* Every state gets a sequence number */
782     stp->ap = 0;                 /* No actions, yet. */
783     State_insert(stp,stp->bp);   /* Add to the state table */
784     buildshifts(lemp,stp);       /* Recursively compute successor states */
785   }
786   return stp;
787 }
788 
789 /* Construct all successor states to the given state.  A "successor"
790 ** state is any state which can be reached by a shift action.
791 */
buildshifts(lemp,stp)792 PRIVATE void buildshifts(lemp,stp)
793 struct lemon *lemp;
794 struct state *stp;     /* The state from which successors are computed */
795 {
796   struct config *cfp;  /* For looping through the config closure of "stp" */
797   struct config *bcfp; /* For the inner loop on config closure of "stp" */
798   struct config *new;  /* */
799   struct symbol *sp;   /* Symbol following the dot in configuration "cfp" */
800   struct symbol *bsp;  /* Symbol following the dot in configuration "bcfp" */
801   struct state *newstp; /* A pointer to a successor state */
802 
803   /* Each configuration becomes complete after it contributes to a successor
804   ** state.  Initially, all configurations are incomplete */
805   for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
806 
807   /* Loop through all configurations of the state "stp" */
808   for(cfp=stp->cfp; cfp; cfp=cfp->next){
809     if( cfp->status==COMPLETE ) continue;    /* Already used by inner loop */
810     if( cfp->dot>=cfp->rp->nrhs ) continue;  /* Can't shift this config */
811     Configlist_reset();                      /* Reset the new config set */
812     sp = cfp->rp->rhs[cfp->dot];             /* Symbol after the dot */
813 
814     /* For every configuration in the state "stp" which has the symbol "sp"
815     ** following its dot, add the same configuration to the basis set under
816     ** construction but with the dot shifted one symbol to the right. */
817     for(bcfp=cfp; bcfp; bcfp=bcfp->next){
818       if( bcfp->status==COMPLETE ) continue;    /* Already used */
819       if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
820       bsp = bcfp->rp->rhs[bcfp->dot];           /* Get symbol after dot */
821       if( bsp!=sp ) continue;                   /* Must be same as for "cfp" */
822       bcfp->status = COMPLETE;                  /* Mark this config as used */
823       new = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
824       Plink_add(&new->bplp,bcfp);
825     }
826 
827     /* Get a pointer to the state described by the basis configuration set
828     ** constructed in the preceding loop */
829     newstp = getstate(lemp);
830 
831     /* The state "newstp" is reached from the state "stp" by a shift action
832     ** on the symbol "sp" */
833     Action_add(&stp->ap,SHIFT,sp,newstp);
834   }
835 }
836 
837 /*
838 ** Construct the propagation links
839 */
FindLinks(lemp)840 void FindLinks(lemp)
841 struct lemon *lemp;
842 {
843   int i;
844   struct config *cfp, *other;
845   struct state *stp;
846   struct plink *plp;
847 
848   /* Housekeeping detail:
849   ** Add to every propagate link a pointer back to the state to
850   ** which the link is attached. */
851   for(i=0; i<lemp->nstate; i++){
852     stp = lemp->sorted[i];
853     for(cfp=stp->cfp; cfp; cfp=cfp->next){
854       cfp->stp = stp;
855     }
856   }
857 
858   /* Convert all backlinks into forward links.  Only the forward
859   ** links are used in the follow-set computation. */
860   for(i=0; i<lemp->nstate; i++){
861     stp = lemp->sorted[i];
862     for(cfp=stp->cfp; cfp; cfp=cfp->next){
863       for(plp=cfp->bplp; plp; plp=plp->next){
864         other = plp->cfp;
865         Plink_add(&other->fplp,cfp);
866       }
867     }
868   }
869 }
870 
871 /* Compute all followsets.
872 **
873 ** A followset is the set of all symbols which can come immediately
874 ** after a configuration.
875 */
FindFollowSets(lemp)876 void FindFollowSets(lemp)
877 struct lemon *lemp;
878 {
879   int i;
880   struct config *cfp;
881   struct state *stp;
882   struct plink *plp;
883   int progress;
884   int change;
885 
886   for(i=0; i<lemp->nstate; i++){
887     stp = lemp->sorted[i];
888     for(cfp=stp->cfp; cfp; cfp=cfp->next){
889       cfp->status = INCOMPLETE;
890     }
891   }
892 
893   do{
894     progress = 0;
895     for(i=0; i<lemp->nstate; i++){
896       stp = lemp->sorted[i];
897       for(cfp=stp->cfp; cfp; cfp=cfp->next){
898         if( cfp->status==COMPLETE ) continue;
899         for(plp=cfp->fplp; plp; plp=plp->next){
900           change = SetUnion(plp->cfp->fws,cfp->fws);
901           if( change ){
902             plp->cfp->status = INCOMPLETE;
903             progress = 1;
904 	  }
905 	}
906         cfp->status = COMPLETE;
907       }
908     }
909   }while( progress );
910 }
911 
912 static int resolve_conflict();
913 
914 /* Compute the reduce actions, and resolve conflicts.
915 */
FindActions(lemp)916 void FindActions(lemp)
917 struct lemon *lemp;
918 {
919   int i,j;
920   struct config *cfp;
921   struct symbol *sp;
922   struct rule *rp;
923 
924   /* Add all of the reduce actions
925   ** A reduce action is added for each element of the followset of
926   ** a configuration which has its dot at the extreme right.
927   */
928   for(i=0; i<lemp->nstate; i++){   /* Loop over all states */
929     struct state *stp;
930     stp = lemp->sorted[i];
931     for(cfp=stp->cfp; cfp; cfp=cfp->next){  /* Loop over all configurations */
932       if( cfp->rp->nrhs==cfp->dot ){        /* Is dot at extreme right? */
933         for(j=0; j<lemp->nterminal; j++){
934           if( SetFind(cfp->fws,j) ){
935             /* Add a reduce action to the state "stp" which will reduce by the
936             ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
937             Action_add(&stp->ap,REDUCE,lemp->symbols[j],cfp->rp);
938           }
939 	}
940       }
941     }
942   }
943 
944   /* Add the accepting token */
945   if( lemp->start ){
946     sp = Symbol_find(lemp->start);
947     if( sp==0 ) sp = lemp->rule->lhs;
948   }else{
949     sp = lemp->rule->lhs;
950   }
951   /* Add to the first state (which is always the starting state of the
952   ** finite state machine) an action to ACCEPT if the lookahead is the
953   ** start nonterminal.  */
954   if (lemp->nstate) { /*(should always be true)*/
955     struct state *stp;
956     stp = lemp->sorted[0];
957     Action_add(&stp->ap,ACCEPT,sp,0);
958   }
959 
960   /* Resolve conflicts */
961   for(i=0; i<lemp->nstate; i++){
962     struct action *ap, *nap;
963     struct state *stp;
964     stp = lemp->sorted[i];
965     assert( stp->ap );
966     stp->ap = Action_sort(stp->ap);
967     for(ap=stp->ap; ap && ap->next; ap=ap->next){
968       for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
969          /* The two actions "ap" and "nap" have the same lookahead.
970          ** Figure out which one should be used */
971          lemp->nconflict += resolve_conflict(ap,nap,lemp->errsym);
972       }
973     }
974   }
975 
976   /* Report an error for each rule that can never be reduced. */
977   for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = Bo_FALSE;
978   for(i=0; i<lemp->nstate; i++){
979     struct action *ap;
980     for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
981       if( ap->type==REDUCE ) ap->x.rp->canReduce = Bo_TRUE;
982     }
983   }
984   for(rp=lemp->rule; rp; rp=rp->next){
985     if( rp->canReduce ) continue;
986     ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
987     lemp->errorcnt++;
988   }
989 }
990 
991 /* Resolve a conflict between the two given actions.  If the
992 ** conflict can't be resolve, return non-zero.
993 **
994 ** NO LONGER TRUE:
995 **   To resolve a conflict, first look to see if either action
996 **   is on an error rule.  In that case, take the action which
997 **   is not associated with the error rule.  If neither or both
998 **   actions are associated with an error rule, then try to
999 **   use precedence to resolve the conflict.
1000 **
1001 ** If either action is a SHIFT, then it must be apx.  This
1002 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1003 */
resolve_conflict(apx,apy,errsym)1004 static int resolve_conflict(apx,apy,errsym)
1005 struct action *apx;
1006 struct action *apy;
1007 struct symbol *errsym;   /* The error symbol (if defined.  NULL otherwise) */
1008 {
1009   struct symbol *spx, *spy;
1010   int errcnt = 0;
1011   UNUSED(errsym);
1012   assert( apx->sp==apy->sp );  /* Otherwise there would be no conflict */
1013   if( apx->type==SHIFT && apy->type==REDUCE ){
1014     spx = apx->sp;
1015     spy = apy->x.rp->precsym;
1016     if( spy==0 || spx->prec<0 || spy->prec<0 ){
1017       /* Not enough precedence information. */
1018       apy->type = CONFLICT;
1019       errcnt++;
1020     }else if( spx->prec>spy->prec ){    /* Lower precedence wins */
1021       apy->type = RD_RESOLVED;
1022     }else if( spx->prec<spy->prec ){
1023       apx->type = SH_RESOLVED;
1024     }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1025       apy->type = RD_RESOLVED;                             /* associativity */
1026     }else if( spx->prec==spy->prec && spx->assoc==LEFT ){  /* to break tie */
1027       apx->type = SH_RESOLVED;
1028     }else{
1029       assert( spx->prec==spy->prec && spx->assoc==NONE );
1030       apy->type = CONFLICT;
1031       errcnt++;
1032     }
1033   }else if( apx->type==REDUCE && apy->type==REDUCE ){
1034     spx = apx->x.rp->precsym;
1035     spy = apy->x.rp->precsym;
1036     if( spx==0 || spy==0 || spx->prec<0 ||
1037     spy->prec<0 || spx->prec==spy->prec ){
1038       apy->type = CONFLICT;
1039       errcnt++;
1040     }else if( spx->prec>spy->prec ){
1041       apy->type = RD_RESOLVED;
1042     }else if( spx->prec<spy->prec ){
1043       apx->type = RD_RESOLVED;
1044     }
1045   }else{
1046     assert(
1047       apx->type==SH_RESOLVED ||
1048       apx->type==RD_RESOLVED ||
1049       apx->type==CONFLICT ||
1050       apy->type==SH_RESOLVED ||
1051       apy->type==RD_RESOLVED ||
1052       apy->type==CONFLICT
1053     );
1054     /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1055     ** REDUCEs on the list.  If we reach this point it must be because
1056     ** the parser conflict had already been resolved. */
1057   }
1058   return errcnt;
1059 }
1060 /********************* From the file "configlist.c" *************************/
1061 /*
1062 ** Routines to processing a configuration list and building a state
1063 ** in the LEMON parser generator.
1064 */
1065 
1066 static struct config *freelist = 0;      /* List of free configurations */
1067 static struct config *current = 0;       /* Top of list of configurations */
1068 static struct config **currentend = 0;   /* Last on list of configs */
1069 static struct config *basis = 0;         /* Top of list of basis configs */
1070 static struct config **basisend = 0;     /* End of list of basis configs */
1071 
1072 /* Return a pointer to a new configuration */
newconfig()1073 PRIVATE struct config *newconfig(){
1074   struct config *new;
1075   if( freelist==0 ){
1076     int i;
1077     int amt = 3;
1078     freelist = (struct config *)malloc( sizeof(struct config)*amt );
1079     if( freelist==0 ){
1080       fprintf(stderr,"Unable to allocate memory for a new configuration.");
1081       exit(1);
1082     }
1083     for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1084     freelist[amt-1].next = 0;
1085   }
1086   new = freelist;
1087   freelist = freelist->next;
1088   return new;
1089 }
1090 
1091 /* The configuration "old" is no longer used */
deleteconfig(old)1092 PRIVATE void deleteconfig(old)
1093 struct config *old;
1094 {
1095   old->next = freelist;
1096   freelist = old;
1097 }
1098 
1099 /* Initialized the configuration list builder */
Configlist_init()1100 void Configlist_init(){
1101   current = 0;
1102   currentend = &current;
1103   basis = 0;
1104   basisend = &basis;
1105   Configtable_init();
1106   return;
1107 }
1108 
1109 /* Initialized the configuration list builder */
Configlist_reset()1110 void Configlist_reset(){
1111   current = 0;
1112   currentend = &current;
1113   basis = 0;
1114   basisend = &basis;
1115   Configtable_clear(0);
1116   return;
1117 }
1118 
1119 /* Add another configuration to the configuration list */
Configlist_add(rp,dot)1120 struct config *Configlist_add(rp,dot)
1121 struct rule *rp;    /* The rule */
1122 int dot;            /* Index into the RHS of the rule where the dot goes */
1123 {
1124   struct config *cfp, model;
1125 
1126   assert( currentend!=0 );
1127   model.rp = rp;
1128   model.dot = dot;
1129   cfp = Configtable_find(&model);
1130   if( cfp==0 ){
1131     cfp = newconfig();
1132     cfp->rp = rp;
1133     cfp->dot = dot;
1134     cfp->fws = SetNew();
1135     cfp->stp = 0;
1136     cfp->fplp = cfp->bplp = 0;
1137     cfp->next = 0;
1138     cfp->bp = 0;
1139     *currentend = cfp;
1140     currentend = &cfp->next;
1141     Configtable_insert(cfp);
1142   }
1143   return cfp;
1144 }
1145 
1146 /* Add a basis configuration to the configuration list */
Configlist_addbasis(rp,dot)1147 struct config *Configlist_addbasis(rp,dot)
1148 struct rule *rp;
1149 int dot;
1150 {
1151   struct config *cfp, model;
1152 
1153   assert( basisend!=0 );
1154   assert( currentend!=0 );
1155   model.rp = rp;
1156   model.dot = dot;
1157   cfp = Configtable_find(&model);
1158   if( cfp==0 ){
1159     cfp = newconfig();
1160     cfp->rp = rp;
1161     cfp->dot = dot;
1162     cfp->fws = SetNew();
1163     cfp->stp = 0;
1164     cfp->fplp = cfp->bplp = 0;
1165     cfp->next = 0;
1166     cfp->bp = 0;
1167     *currentend = cfp;
1168     currentend = &cfp->next;
1169     *basisend = cfp;
1170     basisend = &cfp->bp;
1171     Configtable_insert(cfp);
1172   }
1173   return cfp;
1174 }
1175 
1176 /* Compute the closure of the configuration list */
Configlist_closure(lemp)1177 void Configlist_closure(lemp)
1178 struct lemon *lemp;
1179 {
1180   struct config *cfp, *newcfp;
1181   struct rule *rp, *newrp;
1182   struct symbol *sp, *xsp;
1183   int i, dot;
1184 
1185   assert( currentend!=0 );
1186   for(cfp=current; cfp; cfp=cfp->next){
1187     rp = cfp->rp;
1188     dot = cfp->dot;
1189     if( dot>=rp->nrhs ) continue;
1190     sp = rp->rhs[dot];
1191     if( sp->type==NONTERMINAL ){
1192       if( sp->rule==0 && sp!=lemp->errsym ){
1193         ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1194           sp->name);
1195         lemp->errorcnt++;
1196       }
1197       for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1198         newcfp = Configlist_add(newrp,0);
1199         for(i=dot+1; i<rp->nrhs; i++){
1200           xsp = rp->rhs[i];
1201           if( xsp->type==TERMINAL ){
1202             SetAdd(newcfp->fws,xsp->index);
1203             break;
1204 	  }else{
1205             SetUnion(newcfp->fws,xsp->firstset);
1206             if( xsp->lambda==Bo_FALSE ) break;
1207 	  }
1208 	}
1209         if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1210       }
1211     }
1212   }
1213   return;
1214 }
1215 
1216 /* Sort the configuration list */
Configlist_sort()1217 void Configlist_sort(){
1218   current = (struct config *)msort(current,(void **)&(current->next),Configcmp);
1219   currentend = 0;
1220   return;
1221 }
1222 
1223 /* Sort the basis configuration list */
Configlist_sortbasis()1224 void Configlist_sortbasis(){
1225   basis = (struct config *)msort(current,(void **)&(current->bp),Configcmp);
1226   basisend = 0;
1227   return;
1228 }
1229 
1230 /* Return a pointer to the head of the configuration list and
1231 ** reset the list */
Configlist_return()1232 struct config *Configlist_return(){
1233   struct config *old;
1234   old = current;
1235   current = 0;
1236   currentend = 0;
1237   return old;
1238 }
1239 
1240 /* Return a pointer to the head of the configuration list and
1241 ** reset the list */
Configlist_basis()1242 struct config *Configlist_basis(){
1243   struct config *old;
1244   old = basis;
1245   basis = 0;
1246   basisend = 0;
1247   return old;
1248 }
1249 
1250 /* Free all elements of the given configuration list */
Configlist_eat(cfp)1251 void Configlist_eat(cfp)
1252 struct config *cfp;
1253 {
1254   struct config *nextcfp;
1255   for(; cfp; cfp=nextcfp){
1256     nextcfp = cfp->next;
1257     assert( cfp->fplp==0 );
1258     assert( cfp->bplp==0 );
1259     if( cfp->fws ) SetFree(cfp->fws);
1260     deleteconfig(cfp);
1261   }
1262   return;
1263 }
1264 /***************** From the file "error.c" *********************************/
1265 /*
1266 ** Code for printing error message.
1267 */
1268 
1269 /* Find a good place to break "msg" so that its length is at least "min"
1270 ** but no more than "max".  Make the point as close to max as possible.
1271 */
findbreak(msg,min,max)1272 static int findbreak(msg,min,max)
1273 char *msg;
1274 int min;
1275 int max;
1276 {
1277   int i,spot;
1278   char c;
1279   for(i=spot=min; i<=max; i++){
1280     c = msg[i];
1281     if( c=='\t' ) msg[i] = ' ';
1282     if( c=='\n' ){ msg[i] = ' '; spot = i; break; }
1283     if( c==0 ){ spot = i; break; }
1284     if( c=='-' && i<max-1 ) spot = i+1;
1285     if( c==' ' ) spot = i;
1286   }
1287   return spot;
1288 }
1289 
1290 /*
1291 ** The error message is split across multiple lines if necessary.  The
1292 ** splits occur at a space, if there is a space available near the end
1293 ** of the line.
1294 */
1295 #define ERRMSGSIZE  10000 /* Hope this is big enough.  No way to error check */
1296 #define LINEWIDTH      79 /* Max width of any output line */
1297 #define PREFIXLIMIT    30 /* Max width of the prefix on each line */
ErrorMsg(const char * filename,int lineno,const char * format,...)1298 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1299   char errmsg[ERRMSGSIZE];
1300   char prefix[PREFIXLIMIT+10];
1301   int errmsgsize;
1302   int prefixsize;
1303   int availablewidth;
1304   va_list ap;
1305   int end, restart, base;
1306 
1307   va_start(ap, format);
1308   /* Prepare a prefix to be prepended to every output line */
1309   if( lineno>0 ){
1310     sprintf(prefix,"%.*s:%d: ",PREFIXLIMIT-10,filename,lineno);
1311   }else{
1312     sprintf(prefix,"%.*s: ",PREFIXLIMIT-10,filename);
1313   }
1314   prefixsize = strlen(prefix);
1315   availablewidth = LINEWIDTH - prefixsize;
1316 
1317   /* Generate the error message */
1318   vsprintf(errmsg,format,ap);
1319   va_end(ap);
1320   errmsgsize = strlen(errmsg);
1321   /* Remove trailing '\n's from the error message. */
1322   while( errmsgsize>0 && errmsg[errmsgsize-1]=='\n' ){
1323      errmsg[--errmsgsize] = 0;
1324   }
1325 
1326   /* Print the error message */
1327   base = 0;
1328   while( errmsg[base]!=0 ){
1329     end = restart = findbreak(&errmsg[base],0,availablewidth);
1330     restart += base;
1331     while( errmsg[restart]==' ' ) restart++;
1332     fprintf(stdout,"%s%.*s\n",prefix,end,&errmsg[base]);
1333     base = restart;
1334   }
1335 }
1336 /**************** From the file "main.c" ************************************/
1337 /*
1338 ** Main program file for the LEMON parser generator.
1339 */
1340 
1341 /* Report an out-of-memory condition and abort.  This function
1342 ** is used mostly by the "MemoryCheck" macro in struct.h
1343 */
memory_error()1344 void memory_error() {
1345   fprintf(stderr,"Out of memory.  Aborting...\n");
1346   exit(1);
1347 }
1348 
1349 static const char* out_dir = ".";
1350 /* The main program.  Parse the command line and do it... */
main(argc,argv)1351 int main(argc,argv)
1352 int argc;
1353 char **argv;
1354 {
1355   static int version = 0;
1356   static int rpflag = 0;
1357   static int basisflag = 0;
1358   static int compress = 0;
1359   static int quiet = 0;
1360   static int statistics = 0;
1361   static int mhflag = 0;
1362   static struct s_options options[] = {
1363     {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1364     {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1365     {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1366     {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file"},
1367     {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1368     {OPT_FLAG, "s", (char*)&statistics, "Print parser stats to standard output."},
1369     {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1370     {OPT_STR, "o", (char*)&out_dir, "Customize output directory."},
1371     {OPT_FLAG,0,0,0}
1372   };
1373   int i;
1374   struct lemon lem;
1375   char *def_tmpl_name = "lempar.c";
1376 
1377   UNUSED(argc);
1378   OptInit(argv,options,stderr);
1379   if( version ){
1380      printf("Lemon version 1.0\n");
1381      exit(0);
1382   }
1383   if( OptNArgs() < 1 ){
1384     fprintf(stderr,"Exactly one filename argument is required.\n");
1385     exit(1);
1386   }
1387   lem.errorcnt = 0;
1388 
1389   /* Initialize the machine */
1390   Strsafe_init();
1391   Symbol_init();
1392   State_init();
1393   lem.argv0 = argv[0];
1394   lem.filename = OptArg(0);
1395   lem.tmplname = (OptNArgs() == 2) ? OptArg(1) : def_tmpl_name;
1396   lem.basisflag = basisflag;
1397   lem.has_fallback = 0;
1398   lem.nconflict = 0;
1399   lem.name = lem.include = lem.arg = lem.tokentype = lem.start = 0;
1400   lem.vartype = 0;
1401   lem.stacksize = 0;
1402   lem.error = lem.overflow = lem.failure = lem.accept = lem.tokendest =
1403   lem.tokenprefix = lem.outname = lem.extracode = 0;
1404   lem.vardest = 0;
1405   lem.tablesize = 0;
1406   Symbol_new("$");
1407   lem.errsym = Symbol_new("error");
1408 
1409   /* Parse the input file */
1410   Parse(&lem);
1411   if( lem.errorcnt ) exit(lem.errorcnt);
1412   if( lem.rule==0 ){
1413     fprintf(stderr,"Empty grammar.\n");
1414     exit(1);
1415   }
1416 
1417   /* Count and index the symbols of the grammar */
1418   Symbol_new("{default}");
1419   lem.nsymbol = Symbol_count();
1420   lem.symbols = Symbol_arrayof();
1421   for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1422   qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*),
1423         (int(*)())Symbolcmpp);
1424   for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1425   for(i=1; i<lem.nsymbol && isupper(lem.symbols[i]->name[0]); i++);
1426   lem.nsymbol--; /*(do not count "{default}")*/
1427   lem.nterminal = i;
1428 
1429   /* Generate a reprint of the grammar, if requested on the command line */
1430   if( rpflag ){
1431     Reprint(&lem);
1432   }else{
1433     /* Initialize the size for all follow and first sets */
1434     SetSize(lem.nterminal);
1435 
1436     /* Find the precedence for every production rule (that has one) */
1437     FindRulePrecedences(&lem);
1438 
1439     /* Compute the lambda-nonterminals and the first-sets for every
1440     ** nonterminal */
1441     FindFirstSets(&lem);
1442 
1443     /* Compute all LR(0) states.  Also record follow-set propagation
1444     ** links so that the follow-set can be computed later */
1445     lem.nstate = 0;
1446     FindStates(&lem);
1447     lem.nstate = State_count();
1448     lem.sorted = State_arrayof();
1449 
1450     /* Tie up loose ends on the propagation links */
1451     FindLinks(&lem);
1452 
1453     /* Compute the follow set of every reducible configuration */
1454     FindFollowSets(&lem);
1455 
1456     /* Compute the action tables */
1457     FindActions(&lem);
1458 
1459     /* Compress the action tables */
1460     if( compress==0 ) CompressTables(&lem);
1461 
1462     /* Generate a report of the parser generated.  (the "y.output" file) */
1463     if( !quiet ) ReportOutput(&lem);
1464 
1465     /* Generate the source code for the parser */
1466     ReportTable(&lem, mhflag);
1467 
1468     /* Produce a header file for use by the scanner.  (This step is
1469     ** omitted if the "-m" option is used because makeheaders will
1470     ** generate the file for us.) */
1471     if( !mhflag ) ReportHeader(&lem);
1472   }
1473   if( statistics ){
1474     printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
1475       lem.nterminal, lem.nsymbol - lem.nterminal, lem.nrule);
1476     printf("                   %d states, %d parser table entries, %d conflicts\n",
1477       lem.nstate, lem.tablesize, lem.nconflict);
1478   }
1479   if( lem.nconflict ){
1480     fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1481   }
1482   exit(lem.errorcnt + lem.nconflict);
1483 }
1484 /******************** From the file "msort.c" *******************************/
1485 /*
1486 ** A generic merge-sort program.
1487 **
1488 ** USAGE:
1489 ** Let "ptr" be a pointer to some structure which is at the head of
1490 ** a null-terminated list.  Then to sort the list call:
1491 **
1492 **     ptr = msort(ptr,&(ptr->next),cmpfnc);
1493 **
1494 ** In the above, "cmpfnc" is a pointer to a function which compares
1495 ** two instances of the structure and returns an integer, as in
1496 ** strcmp.  The second argument is a pointer to the pointer to the
1497 ** second element of the linked list.  This address is used to compute
1498 ** the offset to the "next" field within the structure.  The offset to
1499 ** the "next" field must be constant for all structures in the list.
1500 **
1501 ** The function returns a new pointer which is the head of the list
1502 ** after sorting.
1503 **
1504 ** ALGORITHM:
1505 ** Merge-sort.
1506 */
1507 
1508 /*
1509 ** Return a pointer to the next structure in the linked list.
1510 */
1511 #define NEXT(A) (*(char**)(((unsigned long)A)+offset))
1512 
1513 /*
1514 ** Inputs:
1515 **   a:       A sorted, null-terminated linked list.  (May be null).
1516 **   b:       A sorted, null-terminated linked list.  (May be null).
1517 **   cmp:     A pointer to the comparison function.
1518 **   offset:  Offset in the structure to the "next" field.
1519 **
1520 ** Return Value:
1521 **   A pointer to the head of a sorted list containing the elements
1522 **   of both a and b.
1523 **
1524 ** Side effects:
1525 **   The "next" pointers for elements in the lists a and b are
1526 **   changed.
1527 */
merge(a,b,cmp,offset)1528 static char *merge(a,b,cmp,offset)
1529 char *a;
1530 char *b;
1531 int (*cmp)();
1532 int offset;
1533 {
1534   char *ptr, *head;
1535 
1536   if( a==0 ){
1537     head = b;
1538   }else if( b==0 ){
1539     head = a;
1540   }else{
1541     if( (*cmp)(a,b)<0 ){
1542       ptr = a;
1543       a = NEXT(a);
1544     }else{
1545       ptr = b;
1546       b = NEXT(b);
1547     }
1548     head = ptr;
1549     while( a && b ){
1550       if( (*cmp)(a,b)<0 ){
1551         NEXT(ptr) = a;
1552         ptr = a;
1553         a = NEXT(a);
1554       }else{
1555         NEXT(ptr) = b;
1556         ptr = b;
1557         b = NEXT(b);
1558       }
1559     }
1560     if( a ) NEXT(ptr) = a;
1561     else    NEXT(ptr) = b;
1562   }
1563   return head;
1564 }
1565 
1566 /*
1567 ** Inputs:
1568 **   list:      Pointer to a singly-linked list of structures.
1569 **   next:      Pointer to pointer to the second element of the list.
1570 **   cmp:       A comparison function.
1571 **
1572 ** Return Value:
1573 **   A pointer to the head of a sorted list containing the elements
1574 **   originally in list.
1575 **
1576 ** Side effects:
1577 **   The "next" pointers for elements in list are changed.
1578 */
1579 #define LISTSIZE 30
msort(void * list,void ** next,int (* cmp)(void *,void *))1580 void *msort(void *list, void **next, int(*cmp)(void *, void *))
1581 {
1582   unsigned long offset;
1583   char *ep;
1584   char *set[LISTSIZE];
1585   int i;
1586   offset = (unsigned long)next - (unsigned long)list;
1587   for(i=0; i<LISTSIZE; i++) set[i] = 0;
1588   while( list ){
1589     ep = list;
1590     list = NEXT(list);
1591     NEXT(ep) = 0;
1592     for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1593       ep = merge(ep,set[i],cmp,offset);
1594       set[i] = 0;
1595     }
1596     set[i] = ep;
1597   }
1598   ep = 0;
1599   for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(ep,set[i],cmp,offset);
1600   return ep;
1601 }
1602 /************************ From the file "option.c" **************************/
1603 static char **argv;
1604 static struct s_options *op;
1605 static FILE *errstream;
1606 
1607 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1608 
1609 /*
1610 ** Print the command line with a carrot pointing to the k-th character
1611 ** of the n-th field.
1612 */
errline(n,k,err)1613 static void errline(n,k,err)
1614 int n;
1615 int k;
1616 FILE *err;
1617 {
1618   int spcnt = 0, i;
1619   if( argv[0] ) {
1620     fprintf(err,"%s",argv[0]);
1621     spcnt += strlen(argv[0]) + 1;
1622   }
1623   for(i=1; i<n && argv[i]; i++){
1624     fprintf(err," %s",argv[i]);
1625     spcnt += strlen(argv[i]) + 1;
1626   }
1627   spcnt += k;
1628   for(; argv[i]; i++) fprintf(err," %s",argv[i]);
1629   if( spcnt<20 ){
1630     fprintf(err,"\n%*s^-- here\n",spcnt,"");
1631   }else{
1632     fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1633   }
1634 }
1635 
1636 /*
1637 ** Return the index of the N-th non-switch argument.  Return -1
1638 ** if N is out of range.
1639 */
argindex(n)1640 static int argindex(n)
1641 int n;
1642 {
1643   int i;
1644   int dashdash = 0;
1645   if( argv!=0 && *argv!=0 ){
1646     for(i=1; argv[i]; i++){
1647       if( dashdash || !ISOPT(argv[i]) ){
1648         if( n==0 ) return i;
1649         n--;
1650       }
1651       if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1652     }
1653   }
1654   return -1;
1655 }
1656 
1657 static char emsg[] = "Command line syntax error: ";
1658 
1659 /*
1660 ** Process a flag command line argument.
1661 */
handleflags(i,err)1662 static int handleflags(i,err)
1663 int i;
1664 FILE *err;
1665 {
1666   int v;
1667   int errcnt = 0;
1668   int j;
1669   for(j=0; op[j].label; j++){
1670     if( strcmp(&argv[i][1],op[j].label)==0 ) break;
1671   }
1672   v = argv[i][0]=='-' ? 1 : 0;
1673   if( op[j].label==0 ){
1674     if( err ){
1675       fprintf(err,"%sundefined option.\n",emsg);
1676       errline(i,1,err);
1677     }
1678     errcnt++;
1679   }else if( op[j].type==OPT_FLAG ){
1680     *((int*)op[j].arg) = v;
1681   }else if( op[j].type==OPT_FFLAG ){
1682     (*(void(*)())(intptr_t)(op[j].arg))(v);
1683   }else{
1684     if( err ){
1685       fprintf(err,"%smissing argument on switch.\n",emsg);
1686       errline(i,1,err);
1687     }
1688     errcnt++;
1689   }
1690   return errcnt;
1691 }
1692 
1693 /*
1694 ** Process a command line switch which has an argument.
1695 */
handleswitch(i,err)1696 static int handleswitch(i,err)
1697 int i;
1698 FILE *err;
1699 {
1700   int lv = 0;
1701   double dv = 0.0;
1702   char *sv = 0, *end;
1703   char *cp;
1704   int j;
1705   int errcnt = 0;
1706   cp = strchr(argv[i],'=');
1707   *cp = 0;
1708   for(j=0; op[j].label; j++){
1709     if( strcmp(argv[i],op[j].label)==0 ) break;
1710   }
1711   *cp = '=';
1712   if( op[j].label==0 ){
1713     if( err ){
1714       fprintf(err,"%sundefined option.\n",emsg);
1715       errline(i,0,err);
1716     }
1717     errcnt++;
1718   }else{
1719     cp++;
1720     switch( op[j].type ){
1721       case OPT_FLAG:
1722       case OPT_FFLAG:
1723         if( err ){
1724           fprintf(err,"%soption requires an argument.\n",emsg);
1725           errline(i,0,err);
1726         }
1727         errcnt++;
1728         break;
1729       case OPT_DBL:
1730       case OPT_FDBL:
1731         dv = strtod(cp,&end);
1732         if( *end ){
1733           if( err ){
1734             fprintf(err,"%sillegal character in floating-point argument.\n",emsg);
1735             errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1736           }
1737           errcnt++;
1738         }
1739         break;
1740       case OPT_INT:
1741       case OPT_FINT:
1742         lv = strtol(cp,&end,0);
1743         if( *end ){
1744           if( err ){
1745             fprintf(err,"%sillegal character in integer argument.\n",emsg);
1746             errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1747           }
1748           errcnt++;
1749         }
1750         break;
1751       case OPT_STR:
1752       case OPT_FSTR:
1753         sv = cp;
1754         break;
1755     }
1756     switch( op[j].type ){
1757       case OPT_FLAG:
1758       case OPT_FFLAG:
1759         break;
1760       case OPT_DBL:
1761         *(double*)(op[j].arg) = dv;
1762         break;
1763       case OPT_FDBL:
1764         (*(void(*)())(intptr_t)(op[j].arg))(dv);
1765         break;
1766       case OPT_INT:
1767         *(int*)(op[j].arg) = lv;
1768         break;
1769       case OPT_FINT:
1770         (*(void(*)())(intptr_t)(op[j].arg))((int)lv);
1771         break;
1772       case OPT_STR:
1773         *(char**)(op[j].arg) = sv;
1774         break;
1775       case OPT_FSTR:
1776         (*(void(*)())(intptr_t)(op[j].arg))(sv);
1777         break;
1778     }
1779   }
1780   return errcnt;
1781 }
1782 
OptInit(a,o,err)1783 int OptInit(a,o,err)
1784 char **a;
1785 struct s_options *o;
1786 FILE *err;
1787 {
1788   int errcnt = 0;
1789   argv = a;
1790   op = o;
1791   errstream = err;
1792   if( argv && *argv && op ){
1793     int i;
1794     for(i=1; argv[i]; i++){
1795       if( argv[i][0]=='+' || argv[i][0]=='-' ){
1796         errcnt += handleflags(i,err);
1797       }else if( strchr(argv[i],'=') ){
1798         errcnt += handleswitch(i,err);
1799       }
1800     }
1801   }
1802   if( errcnt>0 ){
1803     fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
1804     OptPrint();
1805     exit(1);
1806   }
1807   return 0;
1808 }
1809 
OptNArgs()1810 int OptNArgs(){
1811   int cnt = 0;
1812   int dashdash = 0;
1813   int i;
1814   if( argv!=0 && argv[0]!=0 ){
1815     for(i=1; argv[i]; i++){
1816       if( dashdash || !ISOPT(argv[i]) ) cnt++;
1817       if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1818     }
1819   }
1820   return cnt;
1821 }
1822 
OptArg(n)1823 char *OptArg(n)
1824 int n;
1825 {
1826   int i;
1827   i = argindex(n);
1828   return i>=0 ? argv[i] : 0;
1829 }
1830 
OptErr(n)1831 void OptErr(n)
1832 int n;
1833 {
1834   int i;
1835   i = argindex(n);
1836   if( i>=0 ) errline(i,0,errstream);
1837 }
1838 
OptPrint()1839 void OptPrint(){
1840   int i;
1841   int max, len;
1842   max = 0;
1843   for(i=0; op[i].label; i++){
1844     len = strlen(op[i].label) + 1;
1845     switch( op[i].type ){
1846       case OPT_FLAG:
1847       case OPT_FFLAG:
1848         break;
1849       case OPT_INT:
1850       case OPT_FINT:
1851         len += 9;       /* length of "<integer>" */
1852         break;
1853       case OPT_DBL:
1854       case OPT_FDBL:
1855         len += 6;       /* length of "<real>" */
1856         break;
1857       case OPT_STR:
1858       case OPT_FSTR:
1859         len += 8;       /* length of "<string>" */
1860         break;
1861     }
1862     if( len>max ) max = len;
1863   }
1864   for(i=0; op[i].label; i++){
1865     switch( op[i].type ){
1866       case OPT_FLAG:
1867       case OPT_FFLAG:
1868         fprintf(errstream,"  -%-*s  %s\n",max,op[i].label,op[i].message);
1869         break;
1870       case OPT_INT:
1871       case OPT_FINT:
1872         fprintf(errstream,"  %s=<integer>%*s  %s\n",op[i].label,
1873           (int)(max-strlen(op[i].label)-9),"",op[i].message);
1874         break;
1875       case OPT_DBL:
1876       case OPT_FDBL:
1877         fprintf(errstream,"  %s=<real>%*s  %s\n",op[i].label,
1878           (int)(max-strlen(op[i].label)-6),"",op[i].message);
1879         break;
1880       case OPT_STR:
1881       case OPT_FSTR:
1882         fprintf(errstream,"  %s=<string>%*s  %s\n",op[i].label,
1883           (int)(max-strlen(op[i].label)-8),"",op[i].message);
1884         break;
1885     }
1886   }
1887 }
1888 /*********************** From the file "parse.c" ****************************/
1889 /*
1890 ** Input file parser for the LEMON parser generator.
1891 */
1892 
1893 /* The state of the parser */
1894 struct pstate {
1895   char *filename;       /* Name of the input file */
1896   int tokenlineno;      /* Linenumber at which current token starts */
1897   int errorcnt;         /* Number of errors so far */
1898   char *tokenstart;     /* Text of current token */
1899   struct lemon *gp;     /* Global state vector */
1900   enum e_state {
1901     INITIALIZE,
1902     WAITING_FOR_DECL_OR_RULE,
1903     WAITING_FOR_DECL_KEYWORD,
1904     WAITING_FOR_DECL_ARG,
1905     WAITING_FOR_PRECEDENCE_SYMBOL,
1906     WAITING_FOR_ARROW,
1907     IN_RHS,
1908     LHS_ALIAS_1,
1909     LHS_ALIAS_2,
1910     LHS_ALIAS_3,
1911     RHS_ALIAS_1,
1912     RHS_ALIAS_2,
1913     PRECEDENCE_MARK_1,
1914     PRECEDENCE_MARK_2,
1915     RESYNC_AFTER_RULE_ERROR,
1916     RESYNC_AFTER_DECL_ERROR,
1917     WAITING_FOR_DESTRUCTOR_SYMBOL,
1918     WAITING_FOR_DATATYPE_SYMBOL,
1919     WAITING_FOR_FALLBACK_ID
1920   } state;                   /* The state of the parser */
1921   struct symbol *fallback;   /* The fallback token */
1922   struct symbol *lhs;        /* Left-hand side of current rule */
1923   char *lhsalias;            /* Alias for the LHS */
1924   int nrhs;                  /* Number of right-hand side symbols seen */
1925   struct symbol *rhs[MAXRHS];  /* RHS symbols */
1926   char *alias[MAXRHS];       /* Aliases for each RHS symbol (or NULL) */
1927   struct rule *prevrule;     /* Previous rule parsed */
1928   char *declkeyword;         /* Keyword of a declaration */
1929   char **declargslot;        /* Where the declaration argument should be put */
1930   int *decllnslot;           /* Where the declaration linenumber is put */
1931   enum e_assoc declassoc;    /* Assign this association to decl arguments */
1932   int preccounter;           /* Assign this precedence to decl arguments */
1933   struct rule *firstrule;    /* Pointer to first rule in the grammar */
1934   struct rule *lastrule;     /* Pointer to the most recently parsed rule */
1935 };
1936 
1937 /* Parse a single token */
parseonetoken(psp)1938 static void parseonetoken(psp)
1939 struct pstate *psp;
1940 {
1941   char *x;
1942   x = Strsafe(psp->tokenstart);     /* Save the token permanently */
1943 #if 0
1944   printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
1945     x,psp->state);
1946 #endif
1947   switch( psp->state ){
1948     case INITIALIZE:
1949       psp->prevrule = 0;
1950       psp->preccounter = 0;
1951       psp->firstrule = psp->lastrule = 0;
1952       psp->gp->nrule = 0;
1953       /* Fall through */
1954     case WAITING_FOR_DECL_OR_RULE:
1955       if( x[0]=='%' ){
1956         psp->state = WAITING_FOR_DECL_KEYWORD;
1957       }else if( islower(x[0]) ){
1958         psp->lhs = Symbol_new(x);
1959         psp->nrhs = 0;
1960         psp->lhsalias = 0;
1961         psp->state = WAITING_FOR_ARROW;
1962       }else if( x[0]=='{' ){
1963         if( psp->prevrule==0 ){
1964           ErrorMsg(psp->filename,psp->tokenlineno,
1965 "There is not prior rule opon which to attach the code \
1966 fragment which begins on this line.");
1967           psp->errorcnt++;
1968 	}else if( psp->prevrule->code!=0 ){
1969           ErrorMsg(psp->filename,psp->tokenlineno,
1970 "Code fragment beginning on this line is not the first \
1971 to follow the previous rule.");
1972           psp->errorcnt++;
1973         }else{
1974           psp->prevrule->line = psp->tokenlineno;
1975           psp->prevrule->code = &x[1];
1976 	}
1977       }else if( x[0]=='[' ){
1978         psp->state = PRECEDENCE_MARK_1;
1979       }else{
1980         ErrorMsg(psp->filename,psp->tokenlineno,
1981           "Token \"%s\" should be either \"%%\" or a nonterminal name.",
1982           x);
1983         psp->errorcnt++;
1984       }
1985       break;
1986     case PRECEDENCE_MARK_1:
1987       if( !isupper(x[0]) ){
1988         ErrorMsg(psp->filename,psp->tokenlineno,
1989           "The precedence symbol must be a terminal.");
1990         psp->errorcnt++;
1991       }else if( psp->prevrule==0 ){
1992         ErrorMsg(psp->filename,psp->tokenlineno,
1993           "There is no prior rule to assign precedence \"[%s]\".",x);
1994         psp->errorcnt++;
1995       }else if( psp->prevrule->precsym!=0 ){
1996         ErrorMsg(psp->filename,psp->tokenlineno,
1997 "Precedence mark on this line is not the first \
1998 to follow the previous rule.");
1999         psp->errorcnt++;
2000       }else{
2001         psp->prevrule->precsym = Symbol_new(x);
2002       }
2003       psp->state = PRECEDENCE_MARK_2;
2004       break;
2005     case PRECEDENCE_MARK_2:
2006       if( x[0]!=']' ){
2007         ErrorMsg(psp->filename,psp->tokenlineno,
2008           "Missing \"]\" on precedence mark.");
2009         psp->errorcnt++;
2010       }
2011       psp->state = WAITING_FOR_DECL_OR_RULE;
2012       break;
2013     case WAITING_FOR_ARROW:
2014       if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2015         psp->state = IN_RHS;
2016       }else if( x[0]=='(' ){
2017         psp->state = LHS_ALIAS_1;
2018       }else{
2019         ErrorMsg(psp->filename,psp->tokenlineno,
2020           "Expected to see a \":\" following the LHS symbol \"%s\".",
2021           psp->lhs->name);
2022         psp->errorcnt++;
2023         psp->state = RESYNC_AFTER_RULE_ERROR;
2024       }
2025       break;
2026     case LHS_ALIAS_1:
2027       if( isalpha(x[0]) ){
2028         psp->lhsalias = x;
2029         psp->state = LHS_ALIAS_2;
2030       }else{
2031         ErrorMsg(psp->filename,psp->tokenlineno,
2032           "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2033           x,psp->lhs->name);
2034         psp->errorcnt++;
2035         psp->state = RESYNC_AFTER_RULE_ERROR;
2036       }
2037       break;
2038     case LHS_ALIAS_2:
2039       if( x[0]==')' ){
2040         psp->state = LHS_ALIAS_3;
2041       }else{
2042         ErrorMsg(psp->filename,psp->tokenlineno,
2043           "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2044         psp->errorcnt++;
2045         psp->state = RESYNC_AFTER_RULE_ERROR;
2046       }
2047       break;
2048     case LHS_ALIAS_3:
2049       if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2050         psp->state = IN_RHS;
2051       }else{
2052         ErrorMsg(psp->filename,psp->tokenlineno,
2053           "Missing \"->\" following: \"%s(%s)\".",
2054            psp->lhs->name,psp->lhsalias);
2055         psp->errorcnt++;
2056         psp->state = RESYNC_AFTER_RULE_ERROR;
2057       }
2058       break;
2059     case IN_RHS:
2060       if( x[0]=='.' ){
2061         struct rule *rp;
2062         rp = (struct rule *)malloc( sizeof(struct rule) +
2063              sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs );
2064         if( rp==0 ){
2065           ErrorMsg(psp->filename,psp->tokenlineno,
2066             "Can't allocate enough memory for this rule.");
2067           psp->errorcnt++;
2068           psp->prevrule = 0;
2069 	}else{
2070           int i;
2071           rp->ruleline = psp->tokenlineno;
2072           rp->rhs = (struct symbol**)&rp[1];
2073           rp->rhsalias = (char**)&(rp->rhs[psp->nrhs]);
2074           for(i=0; i<psp->nrhs; i++){
2075             rp->rhs[i] = psp->rhs[i];
2076             rp->rhsalias[i] = psp->alias[i];
2077 	  }
2078           rp->lhs = psp->lhs;
2079           rp->lhsalias = psp->lhsalias;
2080           rp->nrhs = psp->nrhs;
2081           rp->code = 0;
2082           rp->precsym = 0;
2083           rp->index = psp->gp->nrule++;
2084           rp->nextlhs = rp->lhs->rule;
2085           rp->lhs->rule = rp;
2086           rp->next = 0;
2087           if( psp->firstrule==0 ){
2088             psp->firstrule = psp->lastrule = rp;
2089 	  }else{
2090             psp->lastrule->next = rp;
2091             psp->lastrule = rp;
2092 	  }
2093           psp->prevrule = rp;
2094 	}
2095         psp->state = WAITING_FOR_DECL_OR_RULE;
2096       }else if( isalpha(x[0]) ){
2097         if( psp->nrhs>=MAXRHS ){
2098           ErrorMsg(psp->filename,psp->tokenlineno,
2099             "Too many symbol on RHS or rule beginning at \"%s\".",
2100             x);
2101           psp->errorcnt++;
2102           psp->state = RESYNC_AFTER_RULE_ERROR;
2103 	}else{
2104           psp->rhs[psp->nrhs] = Symbol_new(x);
2105           psp->alias[psp->nrhs] = 0;
2106           psp->nrhs++;
2107 	}
2108       }else if( x[0]=='(' && psp->nrhs>0 ){
2109         psp->state = RHS_ALIAS_1;
2110       }else{
2111         ErrorMsg(psp->filename,psp->tokenlineno,
2112           "Illegal character on RHS of rule: \"%s\".",x);
2113         psp->errorcnt++;
2114         psp->state = RESYNC_AFTER_RULE_ERROR;
2115       }
2116       break;
2117     case RHS_ALIAS_1:
2118       if( isalpha(x[0]) ){
2119         psp->alias[psp->nrhs-1] = x;
2120         psp->state = RHS_ALIAS_2;
2121       }else{
2122         ErrorMsg(psp->filename,psp->tokenlineno,
2123           "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2124           x,psp->rhs[psp->nrhs-1]->name);
2125         psp->errorcnt++;
2126         psp->state = RESYNC_AFTER_RULE_ERROR;
2127       }
2128       break;
2129     case RHS_ALIAS_2:
2130       if( x[0]==')' ){
2131         psp->state = IN_RHS;
2132       }else{
2133         ErrorMsg(psp->filename,psp->tokenlineno,
2134           "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2135         psp->errorcnt++;
2136         psp->state = RESYNC_AFTER_RULE_ERROR;
2137       }
2138       break;
2139     case WAITING_FOR_DECL_KEYWORD:
2140       if( isalpha(x[0]) ){
2141         psp->declkeyword = x;
2142         psp->declargslot = 0;
2143         psp->decllnslot = 0;
2144         psp->state = WAITING_FOR_DECL_ARG;
2145         if( strcmp(x,"name")==0 ){
2146           psp->declargslot = &(psp->gp->name);
2147 	}else if( strcmp(x,"include")==0 ){
2148           psp->declargslot = &(psp->gp->include);
2149           psp->decllnslot = &psp->gp->includeln;
2150 	}else if( strcmp(x,"code")==0 ){
2151           psp->declargslot = &(psp->gp->extracode);
2152           psp->decllnslot = &psp->gp->extracodeln;
2153 	}else if( strcmp(x,"token_destructor")==0 ){
2154           psp->declargslot = &psp->gp->tokendest;
2155           psp->decllnslot = &psp->gp->tokendestln;
2156 	}else if( strcmp(x,"default_destructor")==0 ){
2157           psp->declargslot = &psp->gp->vardest;
2158           psp->decllnslot = &psp->gp->vardestln;
2159 	}else if( strcmp(x,"token_prefix")==0 ){
2160           psp->declargslot = &psp->gp->tokenprefix;
2161 	}else if( strcmp(x,"syntax_error")==0 ){
2162           psp->declargslot = &(psp->gp->error);
2163           psp->decllnslot = &psp->gp->errorln;
2164 	}else if( strcmp(x,"parse_accept")==0 ){
2165           psp->declargslot = &(psp->gp->accept);
2166           psp->decllnslot = &psp->gp->acceptln;
2167 	}else if( strcmp(x,"parse_failure")==0 ){
2168           psp->declargslot = &(psp->gp->failure);
2169           psp->decllnslot = &psp->gp->failureln;
2170 	}else if( strcmp(x,"stack_overflow")==0 ){
2171           psp->declargslot = &(psp->gp->overflow);
2172           psp->decllnslot = &psp->gp->overflowln;
2173         }else if( strcmp(x,"extra_argument")==0 ){
2174           psp->declargslot = &(psp->gp->arg);
2175         }else if( strcmp(x,"token_type")==0 ){
2176           psp->declargslot = &(psp->gp->tokentype);
2177         }else if( strcmp(x,"default_type")==0 ){
2178           psp->declargslot = &(psp->gp->vartype);
2179         }else if( strcmp(x,"stack_size")==0 ){
2180           psp->declargslot = &(psp->gp->stacksize);
2181         }else if( strcmp(x,"start_symbol")==0 ){
2182           psp->declargslot = &(psp->gp->start);
2183         }else if( strcmp(x,"left")==0 ){
2184           psp->preccounter++;
2185           psp->declassoc = LEFT;
2186           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2187         }else if( strcmp(x,"right")==0 ){
2188           psp->preccounter++;
2189           psp->declassoc = RIGHT;
2190           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2191         }else if( strcmp(x,"nonassoc")==0 ){
2192           psp->preccounter++;
2193           psp->declassoc = NONE;
2194           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2195 	}else if( strcmp(x,"destructor")==0 ){
2196           psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2197 	}else if( strcmp(x,"type")==0 ){
2198           psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2199         }else if( strcmp(x,"fallback")==0 ){
2200           psp->fallback = 0;
2201           psp->state = WAITING_FOR_FALLBACK_ID;
2202         }else{
2203           ErrorMsg(psp->filename,psp->tokenlineno,
2204             "Unknown declaration keyword: \"%%%s\".",x);
2205           psp->errorcnt++;
2206           psp->state = RESYNC_AFTER_DECL_ERROR;
2207 	}
2208       }else{
2209         ErrorMsg(psp->filename,psp->tokenlineno,
2210           "Illegal declaration keyword: \"%s\".",x);
2211         psp->errorcnt++;
2212         psp->state = RESYNC_AFTER_DECL_ERROR;
2213       }
2214       break;
2215     case WAITING_FOR_DESTRUCTOR_SYMBOL:
2216       if( !isalpha(x[0]) ){
2217         ErrorMsg(psp->filename,psp->tokenlineno,
2218           "Symbol name missing after %%destructor keyword");
2219         psp->errorcnt++;
2220         psp->state = RESYNC_AFTER_DECL_ERROR;
2221       }else{
2222         struct symbol *sp = Symbol_new(x);
2223         psp->declargslot = &sp->destructor;
2224         psp->decllnslot = &sp->destructorln;
2225         psp->state = WAITING_FOR_DECL_ARG;
2226       }
2227       break;
2228     case WAITING_FOR_DATATYPE_SYMBOL:
2229       if( !isalpha(x[0]) ){
2230         ErrorMsg(psp->filename,psp->tokenlineno,
2231           "Symbol name missing after %%destructor keyword");
2232         psp->errorcnt++;
2233         psp->state = RESYNC_AFTER_DECL_ERROR;
2234       }else{
2235         struct symbol *sp = Symbol_new(x);
2236         psp->declargslot = &sp->datatype;
2237         psp->decllnslot = 0;
2238         psp->state = WAITING_FOR_DECL_ARG;
2239       }
2240       break;
2241     case WAITING_FOR_PRECEDENCE_SYMBOL:
2242       if( x[0]=='.' ){
2243         psp->state = WAITING_FOR_DECL_OR_RULE;
2244       }else if( isupper(x[0]) ){
2245         struct symbol *sp;
2246         sp = Symbol_new(x);
2247         if( sp->prec>=0 ){
2248           ErrorMsg(psp->filename,psp->tokenlineno,
2249             "Symbol \"%s\" has already be given a precedence.",x);
2250           psp->errorcnt++;
2251 	}else{
2252           sp->prec = psp->preccounter;
2253           sp->assoc = psp->declassoc;
2254 	}
2255       }else{
2256         ErrorMsg(psp->filename,psp->tokenlineno,
2257           "Can't assign a precedence to \"%s\".",x);
2258         psp->errorcnt++;
2259       }
2260       break;
2261     case WAITING_FOR_DECL_ARG:
2262       if( (x[0]=='{' || x[0]=='\"' || isalnum(x[0])) ){
2263         if( *(psp->declargslot)!=0 ){
2264           ErrorMsg(psp->filename,psp->tokenlineno,
2265             "The argument \"%s\" to declaration \"%%%s\" is not the first.",
2266             x[0]=='\"' ? &x[1] : x,psp->declkeyword);
2267           psp->errorcnt++;
2268           psp->state = RESYNC_AFTER_DECL_ERROR;
2269 	}else{
2270           *(psp->declargslot) = (x[0]=='\"' || x[0]=='{') ? &x[1] : x;
2271           if( psp->decllnslot ) *psp->decllnslot = psp->tokenlineno;
2272           psp->state = WAITING_FOR_DECL_OR_RULE;
2273 	}
2274       }else{
2275         ErrorMsg(psp->filename,psp->tokenlineno,
2276           "Illegal argument to %%%s: %s",psp->declkeyword,x);
2277         psp->errorcnt++;
2278         psp->state = RESYNC_AFTER_DECL_ERROR;
2279       }
2280       break;
2281     case WAITING_FOR_FALLBACK_ID:
2282       if( x[0]=='.' ){
2283         psp->state = WAITING_FOR_DECL_OR_RULE;
2284       }else if( !isupper(x[0]) ){
2285         ErrorMsg(psp->filename, psp->tokenlineno,
2286           "%%fallback argument \"%s\" should be a token", x);
2287         psp->errorcnt++;
2288       }else{
2289         struct symbol *sp = Symbol_new(x);
2290         if( psp->fallback==0 ){
2291           psp->fallback = sp;
2292         }else if( sp->fallback ){
2293           ErrorMsg(psp->filename, psp->tokenlineno,
2294             "More than one fallback assigned to token %s", x);
2295           psp->errorcnt++;
2296         }else{
2297           sp->fallback = psp->fallback;
2298           psp->gp->has_fallback = 1;
2299         }
2300       }
2301       break;
2302     case RESYNC_AFTER_RULE_ERROR:
2303 /*      if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2304 **      break; */
2305     case RESYNC_AFTER_DECL_ERROR:
2306       if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2307       if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2308       break;
2309   }
2310 }
2311 
2312 /* In spite of its name, this function is really a scanner.  It read
2313 ** in the entire input file (all at once) then tokenizes it.  Each
2314 ** token is passed to the function "parseonetoken" which builds all
2315 ** the appropriate data structures in the global state vector "gp".
2316 */
2317 struct pstate ps;
Parse(gp)2318 void Parse(gp)
2319 struct lemon *gp;
2320 {
2321   FILE *fp;
2322   char *filebuf;
2323   size_t filesize;
2324   int lineno;
2325   int c;
2326   char *cp, *nextcp;
2327   int startline = 0;
2328 
2329   ps.gp = gp;
2330   ps.filename = gp->filename;
2331   ps.errorcnt = 0;
2332   ps.state = INITIALIZE;
2333 
2334   /* Begin by reading the input file */
2335   fp = fopen(ps.filename,"rb");
2336   if( fp==0 ){
2337     ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2338     gp->errorcnt++;
2339     return;
2340   }
2341   fseek(fp,0,2);
2342   filesize = ftell(fp);
2343   rewind(fp);
2344   filebuf = (char *)malloc( filesize+1 );
2345   if( filebuf==0 ){
2346     ErrorMsg(ps.filename,0,"Can't allocate %zu of memory to hold this file.",
2347       filesize+1);
2348     fclose(fp);
2349     gp->errorcnt++;
2350     return;
2351   }
2352   if( fread(filebuf,1,filesize,fp)!=filesize ){
2353     ErrorMsg(ps.filename,0,"Can't read in all %zu bytes of this file.",
2354       filesize);
2355     free(filebuf);
2356     fclose(fp);
2357     gp->errorcnt++;
2358     return;
2359   }
2360   fclose(fp);
2361   filebuf[filesize] = 0;
2362 
2363   /* Now scan the text of the input file */
2364   lineno = 1;
2365   for(cp=filebuf; (c= *cp)!=0; ){
2366     if( c=='\n' ) lineno++;              /* Keep track of the line number */
2367     if( isspace(c) ){ cp++; continue; }  /* Skip all white space */
2368     if( c=='/' && cp[1]=='/' ){          /* Skip C++ style comments */
2369       cp+=2;
2370       while( (c= *cp)!=0 && c!='\n' ) cp++;
2371       continue;
2372     }
2373     if( c=='/' && cp[1]=='*' ){          /* Skip C style comments */
2374       cp+=2;
2375       while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
2376         if( c=='\n' ) lineno++;
2377         cp++;
2378       }
2379       if( c ) cp++;
2380       continue;
2381     }
2382     ps.tokenstart = cp;                /* Mark the beginning of the token */
2383     ps.tokenlineno = lineno;           /* Linenumber on which token begins */
2384     if( c=='\"' ){                     /* String literals */
2385       cp++;
2386       while( (c= *cp)!=0 && c!='\"' ){
2387         if( c=='\n' ) lineno++;
2388         cp++;
2389       }
2390       if( c==0 ){
2391         ErrorMsg(ps.filename,startline,
2392 "String starting on this line is not terminated before the end of the file.");
2393         ps.errorcnt++;
2394         nextcp = cp;
2395       }else{
2396         nextcp = cp+1;
2397       }
2398     }else if( c=='{' ){               /* A block of C code */
2399       int level;
2400       cp++;
2401       for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
2402         if( c=='\n' ) lineno++;
2403         else if( c=='{' ) level++;
2404         else if( c=='}' ) level--;
2405         else if( c=='/' && cp[1]=='*' ){  /* Skip comments */
2406           int prevc;
2407           cp = &cp[2];
2408           prevc = 0;
2409           while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
2410             if( c=='\n' ) lineno++;
2411             prevc = c;
2412             cp++;
2413 	  }
2414 	}else if( c=='/' && cp[1]=='/' ){  /* Skip C++ style comments too */
2415           cp = &cp[2];
2416           while( (c= *cp)!=0 && c!='\n' ) cp++;
2417           if( c ) lineno++;
2418 	}else if( c=='\'' || c=='\"' ){    /* String a character literals */
2419           int startchar, prevc;
2420           startchar = c;
2421           prevc = 0;
2422           for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
2423             if( c=='\n' ) lineno++;
2424             if( prevc=='\\' ) prevc = 0;
2425             else              prevc = c;
2426 	  }
2427 	}
2428       }
2429       if( c==0 ){
2430         ErrorMsg(ps.filename,ps.tokenlineno,
2431 "C code starting on this line is not terminated before the end of the file.");
2432         ps.errorcnt++;
2433         nextcp = cp;
2434       }else{
2435         nextcp = cp+1;
2436       }
2437     }else if( isalnum(c) ){          /* Identifiers */
2438       while( (c= *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
2439       nextcp = cp;
2440     }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
2441       cp += 3;
2442       nextcp = cp;
2443     }else{                          /* All other (one character) operators */
2444       cp++;
2445       nextcp = cp;
2446     }
2447     c = *cp;
2448     *cp = 0;                        /* Null terminate the token */
2449     parseonetoken(&ps);             /* Parse the token */
2450     *cp = c;                        /* Restore the buffer */
2451     cp = nextcp;
2452   }
2453   free(filebuf);                    /* Release the buffer after parsing */
2454   gp->rule = ps.firstrule;
2455   gp->errorcnt = ps.errorcnt;
2456 }
2457 /*************************** From the file "plink.c" *********************/
2458 /*
2459 ** Routines processing configuration follow-set propagation links
2460 ** in the LEMON parser generator.
2461 */
2462 static struct plink *plink_freelist = 0;
2463 
2464 /* Allocate a new plink */
Plink_new()2465 struct plink *Plink_new(){
2466   struct plink *new;
2467 
2468   if( plink_freelist==0 ){
2469     int i;
2470     int amt = 100;
2471     plink_freelist = (struct plink *)malloc( sizeof(struct plink)*amt );
2472     if( plink_freelist==0 ){
2473       fprintf(stderr,
2474       "Unable to allocate memory for a new follow-set propagation link.\n");
2475       exit(1);
2476     }
2477     for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
2478     plink_freelist[amt-1].next = 0;
2479   }
2480   new = plink_freelist;
2481   plink_freelist = plink_freelist->next;
2482   return new;
2483 }
2484 
2485 /* Add a plink to a plink list */
Plink_add(plpp,cfp)2486 void Plink_add(plpp,cfp)
2487 struct plink **plpp;
2488 struct config *cfp;
2489 {
2490   struct plink *new;
2491   new = Plink_new();
2492   new->next = *plpp;
2493   *plpp = new;
2494   new->cfp = cfp;
2495 }
2496 
2497 /* Transfer every plink on the list "from" to the list "to" */
Plink_copy(to,from)2498 void Plink_copy(to,from)
2499 struct plink **to;
2500 struct plink *from;
2501 {
2502   struct plink *nextpl;
2503   while( from ){
2504     nextpl = from->next;
2505     from->next = *to;
2506     *to = from;
2507     from = nextpl;
2508   }
2509 }
2510 
2511 /* Delete every plink on the list */
Plink_delete(plp)2512 void Plink_delete(plp)
2513 struct plink *plp;
2514 {
2515   struct plink *nextpl;
2516 
2517   while( plp ){
2518     nextpl = plp->next;
2519     plp->next = plink_freelist;
2520     plink_freelist = plp;
2521     plp = nextpl;
2522   }
2523 }
2524 /*********************** From the file "report.c" **************************/
2525 /*
2526 ** Procedures for generating reports and tables in the LEMON parser generator.
2527 */
2528 
2529 /* Generate a filename with the given suffix.  Space to hold the
2530 ** name comes from malloc() and must be freed by the calling
2531 ** function.
2532 */
file_makename(lemp,suffix)2533 PRIVATE char *file_makename(lemp,suffix)
2534 struct lemon *lemp;
2535 char *suffix;
2536 {
2537   char *name;
2538   char *cp;
2539 
2540   name = malloc( strlen(out_dir) + strlen(lemp->filename) + strlen(suffix) + 6 );
2541   if( name==0 ){
2542     fprintf(stderr,"Can't allocate space for a filename.\n");
2543     exit(1);
2544   }
2545 	/* skip directory, JK */
2546 	if (NULL == (cp = strrchr(lemp->filename, '/'))) {
2547 		cp = lemp->filename;
2548 	} else {
2549 		cp++;
2550 	}
2551   strcpy(name,out_dir);
2552   strcat(name,"/");
2553   strcat(name,cp);
2554   cp = strrchr(name,'.');
2555   if( cp ) *cp = 0;
2556   strcat(name,suffix);
2557   return name;
2558 }
2559 
2560 /* Open a file with a name based on the name of the input file,
2561 ** but with a different (specified) suffix, and return a pointer
2562 ** to the stream */
file_open(lemp,suffix,mode)2563 PRIVATE FILE *file_open(lemp,suffix,mode)
2564 struct lemon *lemp;
2565 char *suffix;
2566 char *mode;
2567 {
2568   FILE *fp;
2569 
2570   if( lemp->outname ) free(lemp->outname);
2571   lemp->outname = file_makename(lemp, suffix);
2572   fp = fopen(lemp->outname,mode);
2573   if( fp==0 && *mode=='w' ){
2574     fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
2575     lemp->errorcnt++;
2576     return 0;
2577   }
2578   return fp;
2579 }
2580 
2581 /* Duplicate the input file without comments and without actions
2582 ** on rules */
Reprint(lemp)2583 void Reprint(lemp)
2584 struct lemon *lemp;
2585 {
2586   struct rule *rp;
2587   struct symbol *sp;
2588   int i, j, maxlen, len, ncolumns, skip;
2589   printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
2590   maxlen = 10;
2591   for(i=0; i<lemp->nsymbol; i++){
2592     sp = lemp->symbols[i];
2593     len = strlen(sp->name);
2594     if( len>maxlen ) maxlen = len;
2595   }
2596   ncolumns = 76/(maxlen+5);
2597   if( ncolumns<1 ) ncolumns = 1;
2598   skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
2599   for(i=0; i<skip; i++){
2600     printf("//");
2601     for(j=i; j<lemp->nsymbol; j+=skip){
2602       sp = lemp->symbols[j];
2603       assert( sp->index==j );
2604       printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
2605     }
2606     printf("\n");
2607   }
2608   for(rp=lemp->rule; rp; rp=rp->next){
2609     printf("%s",rp->lhs->name);
2610 /*    if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
2611     printf(" ::=");
2612     for(i=0; i<rp->nrhs; i++){
2613       printf(" %s",rp->rhs[i]->name);
2614 /*      if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
2615     }
2616     printf(".");
2617     if( rp->precsym ) printf(" [%s]",rp->precsym->name);
2618 /*    if( rp->code ) printf("\n    %s",rp->code); */
2619     printf("\n");
2620   }
2621 }
2622 
ConfigPrint(fp,cfp)2623 PRIVATE void ConfigPrint(fp,cfp)
2624 FILE *fp;
2625 struct config *cfp;
2626 {
2627   struct rule *rp;
2628   int i;
2629   rp = cfp->rp;
2630   fprintf(fp,"%s ::=",rp->lhs->name);
2631   for(i=0; i<=rp->nrhs; i++){
2632     if( i==cfp->dot ) fprintf(fp," *");
2633     if( i==rp->nrhs ) break;
2634     fprintf(fp," %s",rp->rhs[i]->name);
2635   }
2636 }
2637 
2638 /* #define TEST */
2639 #ifdef TEST
2640 /* Print a set */
SetPrint(out,set,lemp)2641 PRIVATE void SetPrint(out,set,lemp)
2642 FILE *out;
2643 char *set;
2644 struct lemon *lemp;
2645 {
2646   int i;
2647   char *spacer;
2648   spacer = "";
2649   fprintf(out,"%12s[","");
2650   for(i=0; i<lemp->nterminal; i++){
2651     if( SetFind(set,i) ){
2652       fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
2653       spacer = " ";
2654     }
2655   }
2656   fprintf(out,"]\n");
2657 }
2658 
2659 /* Print a plink chain */
PlinkPrint(out,plp,tag)2660 void PlinkPrint(out,plp,tag)
2661 FILE *out;
2662 struct plink *plp;
2663 char *tag;
2664 {
2665   while( plp ){
2666     fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->index);
2667     ConfigPrint(out,plp->cfp);
2668     fprintf(out,"\n");
2669     plp = plp->next;
2670   }
2671 }
2672 #endif
2673 
2674 /* Print an action to the given file descriptor.  Return FALSE if
2675 ** nothing was actually printed.
2676 */
PrintAction(struct action * ap,FILE * fp,int indent)2677 PRIVATE int PrintAction(struct action *ap, FILE *fp, int indent){
2678   int result = 1;
2679   switch( ap->type ){
2680     case SHIFT:
2681       fprintf(fp,"%*s shift  %d",indent,ap->sp->name,ap->x.stp->index);
2682       break;
2683     case REDUCE:
2684       fprintf(fp,"%*s reduce %d",indent,ap->sp->name,ap->x.rp->index);
2685       break;
2686     case ACCEPT:
2687       fprintf(fp,"%*s accept",indent,ap->sp->name);
2688       break;
2689     case ERROR:
2690       fprintf(fp,"%*s error",indent,ap->sp->name);
2691       break;
2692     case CONFLICT:
2693       fprintf(fp,"%*s reduce %-3d ** Parsing conflict **",
2694         indent,ap->sp->name,ap->x.rp->index);
2695       break;
2696     case SH_RESOLVED:
2697     case RD_RESOLVED:
2698     case NOT_USED:
2699       result = 0;
2700       break;
2701   }
2702   return result;
2703 }
2704 
2705 /* Generate the "y.output" log file */
ReportOutput(lemp)2706 void ReportOutput(lemp)
2707 struct lemon *lemp;
2708 {
2709   int i;
2710   struct state *stp;
2711   struct config *cfp;
2712   struct action *ap;
2713   FILE *fp;
2714 
2715   fp = file_open(lemp,".out","w");
2716   if( fp==0 ) return;
2717   fprintf(fp," \b");
2718   for(i=0; i<lemp->nstate; i++){
2719     stp = lemp->sorted[i];
2720     fprintf(fp,"State %d:\n",stp->index);
2721     if( lemp->basisflag ) cfp=stp->bp;
2722     else                  cfp=stp->cfp;
2723     while( cfp ){
2724       char buf[20];
2725       if( cfp->dot==cfp->rp->nrhs ){
2726         sprintf(buf,"(%d)",cfp->rp->index);
2727         fprintf(fp,"    %5s ",buf);
2728       }else{
2729         fprintf(fp,"          ");
2730       }
2731       ConfigPrint(fp,cfp);
2732       fprintf(fp,"\n");
2733 #ifdef TEST
2734       SetPrint(fp,cfp->fws,lemp);
2735       PlinkPrint(fp,cfp->fplp,"To  ");
2736       PlinkPrint(fp,cfp->bplp,"From");
2737 #endif
2738       if( lemp->basisflag ) cfp=cfp->bp;
2739       else                  cfp=cfp->next;
2740     }
2741     fprintf(fp,"\n");
2742     for(ap=stp->ap; ap; ap=ap->next){
2743       if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
2744     }
2745     fprintf(fp,"\n");
2746   }
2747   fclose(fp);
2748   return;
2749 }
2750 
2751 /* Search for the file "name" which is in the same directory as
2752 ** the executable */
pathsearch(argv0,name,modemask)2753 PRIVATE char *pathsearch(argv0,name,modemask)
2754 char *argv0;
2755 char *name;
2756 int modemask;
2757 {
2758   char *pathlist;
2759   char *path,*cp;
2760   char c;
2761 
2762 #ifdef __WIN32__
2763   cp = strrchr(argv0,'\\');
2764 #else
2765   cp = strrchr(argv0,'/');
2766 #endif
2767   if( cp ){
2768     c = *cp;
2769     *cp = 0;
2770     path = (char *)malloc( strlen(argv0) + strlen(name) + 2 );
2771     if( path ) sprintf(path,"%s/%s",argv0,name);
2772     *cp = c;
2773   }else{
2774     pathlist = getenv("PATH");
2775     if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
2776     path = (char *)malloc( strlen(pathlist)+strlen(name)+2 );
2777     if( path!=0 ){
2778       while( *pathlist ){
2779         cp = strchr(pathlist,':');
2780         if( cp==0 ) cp = &pathlist[strlen(pathlist)];
2781         c = *cp;
2782         *cp = 0;
2783         sprintf(path,"%s/%s",pathlist,name);
2784         *cp = c;
2785         if( c==0 ) pathlist = "";
2786         else pathlist = &cp[1];
2787         if( access(path,modemask)==0 ) break;
2788       }
2789     }
2790   }
2791   return path;
2792 }
2793 
2794 /* Given an action, compute the integer value for that action
2795 ** which is to be put in the action table of the generated machine.
2796 ** Return negative if no action should be generated.
2797 */
compute_action(lemp,ap)2798 PRIVATE int compute_action(lemp,ap)
2799 struct lemon *lemp;
2800 struct action *ap;
2801 {
2802   int act;
2803   switch( ap->type ){
2804     case SHIFT:  act = ap->x.stp->index;               break;
2805     case REDUCE: act = ap->x.rp->index + lemp->nstate; break;
2806     case ERROR:  act = lemp->nstate + lemp->nrule;     break;
2807     case ACCEPT: act = lemp->nstate + lemp->nrule + 1; break;
2808     default:     act = -1; break;
2809   }
2810   return act;
2811 }
2812 
2813 #define LINESIZE 1000
2814 /* The next cluster of routines are for reading the template file
2815 ** and writing the results to the generated parser */
2816 /* The first function transfers data from "in" to "out" until
2817 ** a line is seen which begins with "%%".  The line number is
2818 ** tracked.
2819 **
2820 ** if name!=0, then any word that begin with "Parse" is changed to
2821 ** begin with *name instead.
2822 */
tplt_xfer(name,in,out,lineno)2823 PRIVATE void tplt_xfer(name,in,out,lineno)
2824 char *name;
2825 FILE *in;
2826 FILE *out;
2827 int *lineno;
2828 {
2829   int i, iStart;
2830   char line[LINESIZE];
2831   while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
2832     (*lineno)++;
2833     iStart = 0;
2834     if( name ){
2835       for(i=0; line[i]; i++){
2836         if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
2837           && (i==0 || !isalpha(line[i-1]))
2838         ){
2839           if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
2840           fprintf(out,"%s",name);
2841           i += 4;
2842           iStart = i+1;
2843         }
2844       }
2845     }
2846     fprintf(out,"%s",&line[iStart]);
2847   }
2848 }
2849 
2850 /* The next function finds the template file and opens it, returning
2851 ** a pointer to the opened file. */
tplt_open(lemp)2852 PRIVATE FILE *tplt_open(lemp)
2853 struct lemon *lemp;
2854 {
2855 
2856   char buf[1000];
2857   FILE *in;
2858   char *tpltname;
2859   char *tpltname_alloc = NULL;
2860   char *cp;
2861 
2862   cp = strrchr(lemp->filename,'.');
2863   if( cp ){
2864     sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
2865   }else{
2866     sprintf(buf,"%s.lt",lemp->filename);
2867   }
2868   if( access(buf,004)==0 ){
2869     tpltname = buf;
2870   }else if( access(lemp->tmplname,004)==0 ){
2871     tpltname = lemp->tmplname;
2872   }else{
2873     tpltname = tpltname_alloc = pathsearch(lemp->argv0,lemp->tmplname,0);
2874   }
2875   if( tpltname==0 ){
2876     fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
2877     lemp->tmplname);
2878     lemp->errorcnt++;
2879     return 0;
2880   }
2881   in = fopen(tpltname,"r");
2882   if( in==0 ){
2883     fprintf(stderr,"Can't open the template file \"%s\".\n",tpltname);
2884     lemp->errorcnt++;
2885   }
2886   if (tpltname_alloc) free(tpltname_alloc);
2887   return in;
2888 }
2889 
2890 /* Print a string to the file and keep the linenumber up to date */
tplt_print(out,lemp,str,strln,lineno)2891 PRIVATE void tplt_print(out,lemp,str,strln,lineno)
2892 FILE *out;
2893 struct lemon *lemp;
2894 char *str;
2895 int strln;
2896 int *lineno;
2897 {
2898   if( str==0 ) return;
2899   fprintf(out,"#line %d \"%s\"\n",strln,lemp->filename); (*lineno)++;
2900   while( *str ){
2901     if( *str=='\n' ) (*lineno)++;
2902     putc(*str,out);
2903     str++;
2904   }
2905   fprintf(out,"\n#line %d \"%s\"\n",*lineno+2,lemp->outname); (*lineno)+=2;
2906   return;
2907 }
2908 
2909 /*
2910 ** The following routine emits code for the destructor for the
2911 ** symbol sp
2912 */
emit_destructor_code(out,sp,lemp,lineno)2913 PRIVATE void emit_destructor_code(out,sp,lemp,lineno)
2914 FILE *out;
2915 struct symbol *sp;
2916 struct lemon *lemp;
2917 int *lineno;
2918 {
2919  char *cp = 0;
2920 
2921  int linecnt = 0;
2922  if( sp->type==TERMINAL ){
2923    cp = lemp->tokendest;
2924    if( cp==0 ) return;
2925    fprintf(out,"#line %d \"%s\"\n{",lemp->tokendestln,lemp->filename);
2926  }else if( sp->destructor ){
2927    cp = sp->destructor;
2928    fprintf(out,"#line %d \"%s\"\n{",sp->destructorln,lemp->filename);
2929  }else{
2930    cp = lemp->vardest;
2931    if( cp==0 ) return;
2932    fprintf(out,"#line %d \"%s\"\n{",lemp->vardestln,lemp->filename);
2933  }
2934  for(; *cp; cp++){
2935    if( *cp=='$' && cp[1]=='$' ){
2936      fprintf(out,"(yypminor->yy%d)",sp->dtnum);
2937      cp++;
2938      continue;
2939    }
2940    if( *cp=='\n' ) linecnt++;
2941    fputc(*cp,out);
2942  }
2943  (*lineno) += 3 + linecnt;
2944  fprintf(out,"}\n#line %d \"%s\"\n",*lineno,lemp->outname);
2945  return;
2946 }
2947 
2948 /*
2949 ** Return TRUE (non-zero) if the given symbol has a destructor.
2950 */
has_destructor(sp,lemp)2951 PRIVATE int has_destructor(sp, lemp)
2952 struct symbol *sp;
2953 struct lemon *lemp;
2954 {
2955   int ret;
2956   if( sp->type==TERMINAL ){
2957     ret = lemp->tokendest!=0;
2958   }else{
2959     ret = lemp->vardest!=0 || sp->destructor!=0;
2960   }
2961   return ret;
2962 }
2963 
2964 /*
2965 ** Generate code which executes when the rule "rp" is reduced.  Write
2966 ** the code to "out".  Make sure lineno stays up-to-date.
2967 */
emit_code(out,rp,lemp,lineno)2968 PRIVATE void emit_code(out,rp,lemp,lineno)
2969 FILE *out;
2970 struct rule *rp;
2971 struct lemon *lemp;
2972 int *lineno;
2973 {
2974  char *cp, *xp;
2975  int linecnt = 0;
2976  int i;
2977  char lhsused = 0;    /* True if the LHS element has been used */
2978  char used[MAXRHS];   /* True for each RHS element which is used */
2979 
2980  for(i=0; i<rp->nrhs; i++) used[i] = 0;
2981  lhsused = 0;
2982 
2983  /* Generate code to do the reduce action */
2984  if( rp->code ){
2985    fprintf(out,"#line %d \"%s\"\n{",rp->line,lemp->filename);
2986    for(cp=rp->code; *cp; cp++){
2987      if( isalpha(*cp) && (cp==rp->code || (!isalnum(cp[-1]) && cp[-1]!='_')) ){
2988        char saved;
2989        for(xp= &cp[1]; isalnum(*xp) || *xp=='_'; xp++);
2990        saved = *xp;
2991        *xp = 0;
2992        if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
2993          fprintf(out,"yygotominor.yy%d",rp->lhs->dtnum);
2994          cp = xp;
2995          lhsused = 1;
2996        }else{
2997          for(i=0; i<rp->nrhs; i++){
2998            if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
2999              fprintf(out,"yymsp[%d].minor.yy%d",i-rp->nrhs+1,rp->rhs[i]->dtnum);
3000              cp = xp;
3001              used[i] = 1;
3002              break;
3003            }
3004          }
3005        }
3006        *xp = saved;
3007      }
3008      if( *cp=='\n' ) linecnt++;
3009      fputc(*cp,out);
3010    } /* End loop */
3011    (*lineno) += 3 + linecnt;
3012    fprintf(out,"}\n#line %d \"%s\"\n",*lineno,lemp->outname);
3013  } /* End if( rp->code ) */
3014 
3015  /* Check to make sure the LHS has been used */
3016  if( rp->lhsalias && !lhsused ){
3017    ErrorMsg(lemp->filename,rp->ruleline,
3018      "Label \"%s\" for \"%s(%s)\" is never used.",
3019        rp->lhsalias,rp->lhs->name,rp->lhsalias);
3020    lemp->errorcnt++;
3021  }
3022 
3023  /* Generate destructor code for RHS symbols which are not used in the
3024  ** reduce code */
3025  for(i=0; i<rp->nrhs; i++){
3026    if( rp->rhsalias[i] && !used[i] ){
3027      ErrorMsg(lemp->filename,rp->ruleline,
3028        "Label %s for \"%s(%s)\" is never used.",
3029        rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
3030      lemp->errorcnt++;
3031    }else if( rp->rhsalias[i]==0 ){
3032      if( has_destructor(rp->rhs[i],lemp) ){
3033        fprintf(out,"  yy_destructor(%d,&yymsp[%d].minor);\n",
3034           rp->rhs[i]->index,i-rp->nrhs+1); (*lineno)++;
3035      }else{
3036        fprintf(out,"        /* No destructor defined for %s */\n",
3037         rp->rhs[i]->name);
3038         (*lineno)++;
3039      }
3040    }
3041  }
3042  return;
3043 }
3044 
3045 /*
3046 ** Print the definition of the union used for the parser's data stack.
3047 ** This union contains fields for every possible data type for tokens
3048 ** and nonterminals.  In the process of computing and printing this
3049 ** union, also set the ".dtnum" field of every terminal and nonterminal
3050 ** symbol.
3051 */
print_stack_union(out,lemp,plineno,mhflag)3052 PRIVATE void print_stack_union(out,lemp,plineno,mhflag)
3053 FILE *out;                  /* The output stream */
3054 struct lemon *lemp;         /* The main info structure for this parser */
3055 int *plineno;               /* Pointer to the line number */
3056 int mhflag;                 /* True if generating makeheaders output */
3057 {
3058   int lineno;               /* The line number of the output */
3059   char **types;             /* A hash table of datatypes */
3060   int arraysize;            /* Size of the "types" array */
3061   int maxdtlength;          /* Maximum length of any ".datatype" field. */
3062   char *stddt;              /* Standardized name for a datatype */
3063   int i,j;                  /* Loop counters */
3064   int hash;                 /* For hashing the name of a type */
3065   char *name;               /* Name of the parser */
3066 
3067   /* Allocate and initialize types[] and allocate stddt[] */
3068   arraysize = lemp->nsymbol * 2;
3069   types = (char**)malloc( arraysize * sizeof(char*) );
3070   for(i=0; i<arraysize; i++) types[i] = 0;
3071   maxdtlength = 0;
3072   if( lemp->vartype ){
3073     maxdtlength = strlen(lemp->vartype);
3074   }
3075   for(i=0; i<lemp->nsymbol; i++){
3076     int len;
3077     struct symbol *sp = lemp->symbols[i];
3078     if( sp->datatype==0 ) continue;
3079     len = strlen(sp->datatype);
3080     if( len>maxdtlength ) maxdtlength = len;
3081   }
3082   stddt = (char*)malloc( maxdtlength*2 + 1 );
3083   if( types==0 || stddt==0 ){
3084     fprintf(stderr,"Out of memory.\n");
3085     exit(1);
3086   }
3087 
3088   /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3089   ** is filled in with the hash index plus 1.  A ".dtnum" value of 0 is
3090   ** used for terminal symbols.  If there is no %default_type defined then
3091   ** 0 is also used as the .dtnum value for nonterminals which do not specify
3092   ** a datatype using the %type directive.
3093   */
3094   for(i=0; i<lemp->nsymbol; i++){
3095     struct symbol *sp = lemp->symbols[i];
3096     char *cp;
3097     if( sp==lemp->errsym ){
3098       sp->dtnum = arraysize+1;
3099       continue;
3100     }
3101     if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
3102       sp->dtnum = 0;
3103       continue;
3104     }
3105     cp = sp->datatype;
3106     if( cp==0 ) cp = lemp->vartype;
3107     j = 0;
3108     while( isspace(*cp) ) cp++;
3109     while( *cp ) stddt[j++] = *cp++;
3110     while( j>0 && isspace(stddt[j-1]) ) j--;
3111     stddt[j] = 0;
3112     hash = 0;
3113     for(j=0; stddt[j]; j++){
3114       hash = (unsigned int)hash*53u + (unsigned int) stddt[j];
3115     }
3116     hash = (hash & 0x7fffffff)%arraysize;
3117     while( types[hash] ){
3118       if( strcmp(types[hash],stddt)==0 ){
3119         sp->dtnum = hash + 1;
3120         break;
3121       }
3122       hash++;
3123       if( hash>=arraysize ) hash = 0;
3124     }
3125     if( types[hash]==0 ){
3126       sp->dtnum = hash + 1;
3127       types[hash] = (char*)malloc( strlen(stddt)+1 );
3128       if( types[hash]==0 ){
3129         fprintf(stderr,"Out of memory.\n");
3130         exit(1);
3131       }
3132       strcpy(types[hash],stddt);
3133     }
3134   }
3135 
3136   /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
3137   name = lemp->name ? lemp->name : "Parse";
3138   lineno = *plineno;
3139   if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
3140   fprintf(out,"#define %sTOKENTYPE %s\n",name,
3141     lemp->tokentype?lemp->tokentype:"void*");  lineno++;
3142   if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
3143   fprintf(out,"typedef union {\n"); lineno++;
3144   fprintf(out,"  %sTOKENTYPE yy0;\n",name); lineno++;
3145   for(i=0; i<arraysize; i++){
3146     if( types[i]==0 ) continue;
3147     fprintf(out,"  %s yy%d;\n",types[i],i+1); lineno++;
3148     free(types[i]);
3149   }
3150   fprintf(out,"  int yy%d;\n",lemp->errsym->dtnum); lineno++;
3151   free(stddt);
3152   free(types);
3153   fprintf(out,"} YYMINORTYPE;\n"); lineno++;
3154   *plineno = lineno;
3155 }
3156 
3157 /*
3158 ** Return the name of a C datatype able to represent values between
3159 ** lwr and upr, inclusive.
3160 */
minimum_size_type(int lwr,int upr)3161 static const char *minimum_size_type(int lwr, int upr){
3162   if( lwr>=0 ){
3163     if( upr<=255 ){
3164       return "unsigned char";
3165     }else if( upr<65535 ){
3166       return "unsigned short int";
3167     }else{
3168       return "unsigned int";
3169     }
3170   }else if( lwr>=-127 && upr<=127 ){
3171     return "signed char";
3172   }else if( lwr>=-32767 && upr<32767 ){
3173     return "short";
3174   }else{
3175     return "int";
3176   }
3177 }
3178 
3179 /*
3180 ** Each state contains a set of token transaction and a set of
3181 ** nonterminal transactions.  Each of these sets makes an instance
3182 ** of the following structure.  An array of these structures is used
3183 ** to order the creation of entries in the yy_action[] table.
3184 */
3185 struct axset {
3186   struct state *stp;   /* A pointer to a state */
3187   int isTkn;           /* True to use tokens.  False for non-terminals */
3188   int nAction;         /* Number of actions */
3189 };
3190 
3191 /*
3192 ** Compare to axset structures for sorting purposes
3193 */
axset_compare(const void * a,const void * b)3194 static int axset_compare(const void *a, const void *b){
3195   struct axset *p1 = (struct axset*)a;
3196   struct axset *p2 = (struct axset*)b;
3197   return p2->nAction - p1->nAction;
3198 }
3199 
3200 /* Generate C source code for the parser */
ReportTable(lemp,mhflag)3201 void ReportTable(lemp, mhflag)
3202 struct lemon *lemp;
3203 int mhflag;     /* Output in makeheaders format if true */
3204 {
3205   FILE *out, *in;
3206   char line[LINESIZE];
3207   int  lineno;
3208   struct state *stp;
3209   struct action *ap;
3210   struct rule *rp;
3211   struct acttab *pActtab;
3212   int i, j, n;
3213   int mnTknOfst, mxTknOfst;
3214   int mnNtOfst, mxNtOfst;
3215   struct axset *ax;
3216   char *name;
3217 
3218   in = tplt_open(lemp);
3219   if( in==0 ) return;
3220   out = file_open(lemp,".c","w");
3221   if( out==0 ){
3222     fclose(in);
3223     return;
3224   }
3225   lineno = 1;
3226   tplt_xfer(lemp->name,in,out,&lineno);
3227 
3228   /* Generate the include code, if any */
3229   tplt_print(out,lemp,lemp->include,lemp->includeln,&lineno);
3230   if( mhflag ){
3231     name = file_makename(lemp, ".h");
3232     fprintf(out,"#include \"%s\"\n", name); lineno++;
3233     free(name);
3234   }
3235   tplt_xfer(lemp->name,in,out,&lineno);
3236 
3237   /* Generate #defines for all tokens */
3238   if( mhflag ){
3239     char *prefix;
3240     fprintf(out,"#if INTERFACE\n"); lineno++;
3241     if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
3242     else                    prefix = "";
3243     for(i=1; i<lemp->nterminal; i++){
3244       fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3245       lineno++;
3246     }
3247     fprintf(out,"#endif\n"); lineno++;
3248   }
3249   tplt_xfer(lemp->name,in,out,&lineno);
3250 
3251   /* Generate the defines */
3252   fprintf(out,"/* \001 */\n");
3253   fprintf(out,"#define YYCODETYPE %s\n",
3254     minimum_size_type(0, lemp->nsymbol+5)); lineno++;
3255   fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol+1);  lineno++;
3256   fprintf(out,"#define YYACTIONTYPE %s\n",
3257     minimum_size_type(0, lemp->nstate+lemp->nrule+5));  lineno++;
3258   print_stack_union(out,lemp,&lineno,mhflag);
3259   if( lemp->stacksize ){
3260     if( atoi(lemp->stacksize)<=0 ){
3261       ErrorMsg(lemp->filename,0,
3262 "Illegal stack size: [%s].  The stack size should be an integer constant.",
3263         lemp->stacksize);
3264       lemp->errorcnt++;
3265       lemp->stacksize = "100";
3266     }
3267     fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize);  lineno++;
3268   }else{
3269     fprintf(out,"#define YYSTACKDEPTH 100\n");  lineno++;
3270   }
3271   if( mhflag ){
3272     fprintf(out,"#if INTERFACE\n"); lineno++;
3273   }
3274   name = lemp->name ? lemp->name : "Parse";
3275   if( lemp->arg && lemp->arg[0] ){
3276     i = strlen(lemp->arg);
3277     while( i>=1 && isspace(lemp->arg[i-1]) ) i--;
3278     while( i>=1 && (isalnum(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
3279     fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg);  lineno++;
3280     fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg);  lineno++;
3281     fprintf(out,"#define %sARG_FETCH %s = yypParser->%s\n",
3282                  name,lemp->arg,&lemp->arg[i]);  lineno++;
3283     fprintf(out,"#define %sARG_STORE yypParser->%s = %s\n",
3284                  name,&lemp->arg[i],&lemp->arg[i]);  lineno++;
3285   }else{
3286     fprintf(out,"#define %sARG_SDECL\n",name);  lineno++;
3287     fprintf(out,"#define %sARG_PDECL\n",name);  lineno++;
3288     fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
3289     fprintf(out,"#define %sARG_STORE\n",name); lineno++;
3290   }
3291   if( mhflag ){
3292     fprintf(out,"#endif\n"); lineno++;
3293   }
3294   fprintf(out,"#define YYNSTATE %d\n",lemp->nstate);  lineno++;
3295   fprintf(out,"#define YYNRULE %d\n",lemp->nrule);  lineno++;
3296   fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index);  lineno++;
3297   fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum);  lineno++;
3298   if( lemp->has_fallback ){
3299     fprintf(out,"#define YYFALLBACK 1\n");  lineno++;
3300   }
3301   tplt_xfer(lemp->name,in,out,&lineno);
3302 
3303   /* Generate the action table and its associates:
3304   **
3305   **  yy_action[]        A single table containing all actions.
3306   **  yy_lookahead[]     A table containing the lookahead for each entry in
3307   **                     yy_action.  Used to detect hash collisions.
3308   **  yy_shift_ofst[]    For each state, the offset into yy_action for
3309   **                     shifting terminals.
3310   **  yy_reduce_ofst[]   For each state, the offset into yy_action for
3311   **                     shifting non-terminals after a reduce.
3312   **  yy_default[]       Default action for each state.
3313   */
3314 
3315   /* Compute the actions on all states and count them up */
3316   ax = malloc( sizeof(ax[0])*lemp->nstate*2 );
3317   if( ax==0 ){
3318     fprintf(stderr,"malloc failed\n");
3319     exit(1);
3320   }
3321   for(i=0; i<lemp->nstate; i++){
3322     stp = lemp->sorted[i];
3323     stp->nTknAct = stp->nNtAct = 0;
3324     stp->iDflt = lemp->nstate + lemp->nrule;
3325     stp->iTknOfst = NO_OFFSET;
3326     stp->iNtOfst = NO_OFFSET;
3327     for(ap=stp->ap; ap; ap=ap->next){
3328       if( compute_action(lemp,ap)>=0 ){
3329         if( ap->sp->index<lemp->nterminal ){
3330           stp->nTknAct++;
3331         }else if( ap->sp->index<lemp->nsymbol ){
3332           stp->nNtAct++;
3333         }else{
3334           stp->iDflt = compute_action(lemp, ap);
3335         }
3336       }
3337     }
3338     ax[i*2].stp = stp;
3339     ax[i*2].isTkn = 1;
3340     ax[i*2].nAction = stp->nTknAct;
3341     ax[i*2+1].stp = stp;
3342     ax[i*2+1].isTkn = 0;
3343     ax[i*2+1].nAction = stp->nNtAct;
3344   }
3345   mxTknOfst = mnTknOfst = 0;
3346   mxNtOfst = mnNtOfst = 0;
3347 
3348   /* Compute the action table.  In order to try to keep the size of the
3349   ** action table to a minimum, the heuristic of placing the largest action
3350   ** sets first is used.
3351   */
3352   qsort(ax, lemp->nstate*2, sizeof(ax[0]), axset_compare);
3353   pActtab = acttab_alloc();
3354   for(i=0; i<lemp->nstate*2 && ax[i].nAction>0; i++){
3355     stp = ax[i].stp;
3356     if( ax[i].isTkn ){
3357       for(ap=stp->ap; ap; ap=ap->next){
3358         int action;
3359         if( ap->sp->index>=lemp->nterminal ) continue;
3360         action = compute_action(lemp, ap);
3361         if( action<0 ) continue;
3362         acttab_action(pActtab, ap->sp->index, action);
3363       }
3364       stp->iTknOfst = acttab_insert(pActtab);
3365       if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
3366       if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
3367     }else{
3368       for(ap=stp->ap; ap; ap=ap->next){
3369         int action;
3370         if( ap->sp->index<lemp->nterminal ) continue;
3371         if( ap->sp->index==lemp->nsymbol ) continue;
3372         action = compute_action(lemp, ap);
3373         if( action<0 ) continue;
3374         acttab_action(pActtab, ap->sp->index, action);
3375       }
3376       stp->iNtOfst = acttab_insert(pActtab);
3377       if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
3378       if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
3379     }
3380   }
3381   free(ax);
3382 
3383   /* Output the yy_action table */
3384   fprintf(out,"static YYACTIONTYPE yy_action[] = {\n"); lineno++;
3385   n = acttab_size(pActtab);
3386   for(i=j=0; i<n; i++){
3387     int action = acttab_yyaction(pActtab, i);
3388     if( action<0 ) action = lemp->nsymbol + lemp->nrule + 2;
3389     if( j==0 ) fprintf(out," /* %5d */ ", i);
3390     fprintf(out, " %4d,", action);
3391     if( j==9 || i==n-1 ){
3392       fprintf(out, "\n"); lineno++;
3393       j = 0;
3394     }else{
3395       j++;
3396     }
3397   }
3398   fprintf(out, "};\n"); lineno++;
3399 
3400   /* Output the yy_lookahead table */
3401   fprintf(out,"static YYCODETYPE yy_lookahead[] = {\n"); lineno++;
3402   for(i=j=0; i<n; i++){
3403     int la = acttab_yylookahead(pActtab, i);
3404     if( la<0 ) la = lemp->nsymbol;
3405     if( j==0 ) fprintf(out," /* %5d */ ", i);
3406     fprintf(out, " %4d,", la);
3407     if( j==9 || i==n-1 ){
3408       fprintf(out, "\n"); lineno++;
3409       j = 0;
3410     }else{
3411       j++;
3412     }
3413   }
3414   fprintf(out, "};\n"); lineno++;
3415 
3416   /* Output the yy_shift_ofst[] table */
3417   fprintf(out, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst-1); lineno++;
3418   fprintf(out, "static %s yy_shift_ofst[] = {\n",
3419           minimum_size_type(mnTknOfst-1, mxTknOfst)); lineno++;
3420   n = lemp->nstate;
3421   for(i=j=0; i<n; i++){
3422     int ofst;
3423     stp = lemp->sorted[i];
3424     ofst = stp->iTknOfst;
3425     if( ofst==NO_OFFSET ) ofst = mnTknOfst - 1;
3426     if( j==0 ) fprintf(out," /* %5d */ ", i);
3427     fprintf(out, " %4d,", ofst);
3428     if( j==9 || i==n-1 ){
3429       fprintf(out, "\n"); lineno++;
3430       j = 0;
3431     }else{
3432       j++;
3433     }
3434   }
3435   fprintf(out, "};\n"); lineno++;
3436 
3437   /* Output the yy_reduce_ofst[] table */
3438   fprintf(out, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst-1); lineno++;
3439   fprintf(out, "static %s yy_reduce_ofst[] = {\n",
3440           minimum_size_type(mnNtOfst-1, mxNtOfst)); lineno++;
3441   n = lemp->nstate;
3442   for(i=j=0; i<n; i++){
3443     int ofst;
3444     stp = lemp->sorted[i];
3445     ofst = stp->iNtOfst;
3446     if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
3447     if( j==0 ) fprintf(out," /* %5d */ ", i);
3448     fprintf(out, " %4d,", ofst);
3449     if( j==9 || i==n-1 ){
3450       fprintf(out, "\n"); lineno++;
3451       j = 0;
3452     }else{
3453       j++;
3454     }
3455   }
3456   fprintf(out, "};\n"); lineno++;
3457 
3458   /* Output the default action table */
3459   fprintf(out, "static YYACTIONTYPE yy_default[] = {\n"); lineno++;
3460   n = lemp->nstate;
3461   for(i=j=0; i<n; i++){
3462     stp = lemp->sorted[i];
3463     if( j==0 ) fprintf(out," /* %5d */ ", i);
3464     fprintf(out, " %4d,", stp->iDflt);
3465     if( j==9 || i==n-1 ){
3466       fprintf(out, "\n"); lineno++;
3467       j = 0;
3468     }else{
3469       j++;
3470     }
3471   }
3472   fprintf(out, "};\n"); lineno++;
3473   tplt_xfer(lemp->name,in,out,&lineno);
3474 
3475   /* Generate the table of fallback tokens.
3476   */
3477   if( lemp->has_fallback ){
3478     for(i=0; i<lemp->nterminal; i++){
3479       struct symbol *p = lemp->symbols[i];
3480       if( p->fallback==0 ){
3481         fprintf(out, "    0,  /* %10s => nothing */\n", p->name);
3482       }else{
3483         fprintf(out, "  %3d,  /* %10s => %s */\n", p->fallback->index,
3484           p->name, p->fallback->name);
3485       }
3486       lineno++;
3487     }
3488   }
3489   tplt_xfer(lemp->name, in, out, &lineno);
3490 
3491   /* Generate a table containing the symbolic name of every symbol
3492   */
3493   for(i=0; i<lemp->nsymbol; i++){
3494     sprintf(line,"\"%s\",",lemp->symbols[i]->name);
3495     fprintf(out,"  %-15s",line);
3496     if( (i&3)==3 ){ fprintf(out,"\n"); lineno++; }
3497   }
3498   if( (i&3)!=0 ){ fprintf(out,"\n"); lineno++; }
3499   tplt_xfer(lemp->name,in,out,&lineno);
3500 
3501   /* Generate a table containing a text string that describes every
3502   ** rule in the rule set of the grammar.  This information is used
3503   ** when tracing REDUCE actions.
3504   */
3505   for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
3506     assert( rp->index==i );
3507     fprintf(out," /* %3d */ \"%s ::=", i, rp->lhs->name);
3508     for(j=0; j<rp->nrhs; j++) fprintf(out," %s",rp->rhs[j]->name);
3509     fprintf(out,"\",\n"); lineno++;
3510   }
3511   tplt_xfer(lemp->name,in,out,&lineno);
3512 
3513   /* Generate code which executes every time a symbol is popped from
3514   ** the stack while processing errors or while destroying the parser.
3515   ** (In other words, generate the %destructor actions)
3516   */
3517   if( lemp->tokendest ){
3518     for(i=0; i<lemp->nsymbol; i++){
3519       struct symbol *sp = lemp->symbols[i];
3520       if( sp==0 || sp->type!=TERMINAL ) continue;
3521       fprintf(out,"    case %d:\n",sp->index); lineno++;
3522     }
3523     for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
3524     if( i<lemp->nsymbol ){
3525       emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
3526       fprintf(out,"      break;\n"); lineno++;
3527     }
3528   }
3529   for(i=0; i<lemp->nsymbol; i++){
3530     struct symbol *sp = lemp->symbols[i];
3531     if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
3532     fprintf(out,"    case %d:\n",sp->index); lineno++;
3533     emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
3534     fprintf(out,"      break;\n"); lineno++;
3535   }
3536   if( lemp->vardest ){
3537     struct symbol *dflt_sp = 0;
3538     for(i=0; i<lemp->nsymbol; i++){
3539       struct symbol *sp = lemp->symbols[i];
3540       if( sp==0 || sp->type==TERMINAL ||
3541           sp->index<=0 || sp->destructor!=0 ) continue;
3542       fprintf(out,"    case %d:\n",sp->index); lineno++;
3543       dflt_sp = sp;
3544     }
3545     if( dflt_sp!=0 ){
3546       emit_destructor_code(out,dflt_sp,lemp,&lineno);
3547       fprintf(out,"      break;\n"); lineno++;
3548     }
3549   }
3550   tplt_xfer(lemp->name,in,out,&lineno);
3551 
3552   /* Generate code which executes whenever the parser stack overflows */
3553   tplt_print(out,lemp,lemp->overflow,lemp->overflowln,&lineno);
3554   tplt_xfer(lemp->name,in,out,&lineno);
3555 
3556   /* Generate the table of rule information
3557   **
3558   ** Note: This code depends on the fact that rules are number
3559   ** sequentually beginning with 0.
3560   */
3561   for(rp=lemp->rule; rp; rp=rp->next){
3562     fprintf(out,"  { %d, %d },\n",rp->lhs->index,rp->nrhs); lineno++;
3563   }
3564   tplt_xfer(lemp->name,in,out,&lineno);
3565 
3566   /* Generate code which execution during each REDUCE action */
3567   for(rp=lemp->rule; rp; rp=rp->next){
3568     fprintf(out,"      case %d:\n",rp->index); lineno++;
3569     emit_code(out,rp,lemp,&lineno);
3570     fprintf(out,"        break;\n"); lineno++;
3571   }
3572   tplt_xfer(lemp->name,in,out,&lineno);
3573 
3574   /* Generate code which executes if a parse fails */
3575   tplt_print(out,lemp,lemp->failure,lemp->failureln,&lineno);
3576   tplt_xfer(lemp->name,in,out,&lineno);
3577 
3578   /* Generate code which executes when a syntax error occurs */
3579   tplt_print(out,lemp,lemp->error,lemp->errorln,&lineno);
3580   tplt_xfer(lemp->name,in,out,&lineno);
3581 
3582   /* Generate code which executes when the parser accepts its input */
3583   tplt_print(out,lemp,lemp->accept,lemp->acceptln,&lineno);
3584   tplt_xfer(lemp->name,in,out,&lineno);
3585 
3586   /* Append any addition code the user desires */
3587   tplt_print(out,lemp,lemp->extracode,lemp->extracodeln,&lineno);
3588 
3589   fclose(in);
3590   fclose(out);
3591   return;
3592 }
3593 
3594 /* Generate a header file for the parser */
ReportHeader(lemp)3595 void ReportHeader(lemp)
3596 struct lemon *lemp;
3597 {
3598   FILE *out, *in;
3599   char *prefix;
3600   char line[LINESIZE];
3601   char pattern[LINESIZE];
3602   int i;
3603 
3604   if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
3605   else                    prefix = "";
3606   in = file_open(lemp,".h","r");
3607   if( in ){
3608     for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
3609       sprintf(pattern,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3610       if( strcmp(line,pattern) ) break;
3611     }
3612     fclose(in);
3613     if( i==lemp->nterminal ){
3614       /* No change in the file.  Don't rewrite it. */
3615       return;
3616     }
3617   }
3618   out = file_open(lemp,".h","w");
3619   if( out ){
3620     for(i=1; i<lemp->nterminal; i++){
3621       fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3622     }
3623     fclose(out);
3624   }
3625   return;
3626 }
3627 
3628 /* Reduce the size of the action tables, if possible, by making use
3629 ** of defaults.
3630 **
3631 ** In this version, we take the most frequent REDUCE action and make
3632 ** it the default.  Only default a reduce if there are more than one.
3633 */
CompressTables(lemp)3634 void CompressTables(lemp)
3635 struct lemon *lemp;
3636 {
3637   struct state *stp;
3638   struct action *ap, *ap2;
3639   struct rule *rp, *rp2, *rbest;
3640   int nbest, n;
3641   int i;
3642 
3643   for(i=0; i<lemp->nstate; i++){
3644     stp = lemp->sorted[i];
3645     nbest = 0;
3646     rbest = 0;
3647 
3648     for(ap=stp->ap; ap; ap=ap->next){
3649       if( ap->type!=REDUCE ) continue;
3650       rp = ap->x.rp;
3651       if( rp==rbest ) continue;
3652       n = 1;
3653       for(ap2=ap->next; ap2; ap2=ap2->next){
3654         if( ap2->type!=REDUCE ) continue;
3655         rp2 = ap2->x.rp;
3656         if( rp2==rbest ) continue;
3657         if( rp2==rp ) n++;
3658       }
3659       if( n>nbest ){
3660         nbest = n;
3661         rbest = rp;
3662       }
3663     }
3664 
3665     /* Do not make a default if the number of rules to default
3666     ** is not at least 2 */
3667     if( nbest<2 ) continue;
3668 
3669 
3670     /* Combine matching REDUCE actions into a single default */
3671     for(ap=stp->ap; ap; ap=ap->next){
3672       if( ap->type==REDUCE && ap->x.rp==rbest ) break;
3673     }
3674     assert( ap );
3675     ap->sp = Symbol_new("{default}");
3676     for(ap=ap->next; ap; ap=ap->next){
3677       if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
3678     }
3679     stp->ap = Action_sort(stp->ap);
3680   }
3681 }
3682 
3683 /***************** From the file "set.c" ************************************/
3684 /*
3685 ** Set manipulation routines for the LEMON parser generator.
3686 */
3687 
3688 static int global_size = 0;
3689 
3690 /* Set the set size */
SetSize(n)3691 void SetSize(n)
3692 int n;
3693 {
3694   global_size = n+1;
3695 }
3696 
3697 /* Allocate a new set */
SetNew()3698 char *SetNew(){
3699   char *s;
3700   int i;
3701   s = (char*)malloc( global_size );
3702   if( s==0 ){
3703     memory_error();
3704   }
3705   for(i=0; i<global_size; i++) s[i] = 0;
3706   return s;
3707 }
3708 
3709 /* Deallocate a set */
SetFree(s)3710 void SetFree(s)
3711 char *s;
3712 {
3713   free(s);
3714 }
3715 
3716 /* Add a new element to the set.  Return TRUE if the element was added
3717 ** and FALSE if it was already there. */
SetAdd(s,e)3718 int SetAdd(s,e)
3719 char *s;
3720 int e;
3721 {
3722   int rv;
3723   rv = s[e];
3724   s[e] = 1;
3725   return !rv;
3726 }
3727 
3728 /* Add every element of s2 to s1.  Return TRUE if s1 changes. */
SetUnion(s1,s2)3729 int SetUnion(s1,s2)
3730 char *s1;
3731 char *s2;
3732 {
3733   int i, progress;
3734   progress = 0;
3735   for(i=0; i<global_size; i++){
3736     if( s2[i]==0 ) continue;
3737     if( s1[i]==0 ){
3738       progress = 1;
3739       s1[i] = 1;
3740     }
3741   }
3742   return progress;
3743 }
3744 /********************** From the file "table.c" ****************************/
3745 /*
3746 ** All code in this file has been automatically generated
3747 ** from a specification in the file
3748 **              "table.q"
3749 ** by the associative array code building program "aagen".
3750 ** Do not edit this file!  Instead, edit the specification
3751 ** file, then rerun aagen.
3752 */
3753 /*
3754 ** Code for processing tables in the LEMON parser generator.
3755 */
3756 
strhash(x)3757 PRIVATE int strhash(x)
3758 char *x;
3759 {
3760   unsigned int h = 0;
3761   while( *x) h = h*13u + (unsigned int) *(x++);
3762   return h;
3763 }
3764 
3765 /* Works like strdup, sort of.  Save a string in malloced memory, but
3766 ** keep strings in a table so that the same string is not in more
3767 ** than one place.
3768 */
Strsafe(y)3769 char *Strsafe(y)
3770 char *y;
3771 {
3772   char *z;
3773 
3774   z = Strsafe_find(y);
3775   if( z==0 && (z=malloc( strlen(y)+1 ))!=0 ){
3776     strcpy(z,y);
3777     Strsafe_insert(z);
3778   }
3779   MemoryCheck(z);
3780   return z;
3781 }
3782 
3783 /* There is one instance of the following structure for each
3784 ** associative array of type "x1".
3785 */
3786 struct s_x1 {
3787   int size;               /* The number of available slots. */
3788                           /*   Must be a power of 2 greater than or */
3789                           /*   equal to 1 */
3790   int count;              /* Number of currently slots filled */
3791   struct s_x1node *tbl;  /* The data stored here */
3792   struct s_x1node **ht;  /* Hash table for lookups */
3793 };
3794 
3795 /* There is one instance of this structure for every data element
3796 ** in an associative array of type "x1".
3797 */
3798 typedef struct s_x1node {
3799   char *data;                  /* The data */
3800   struct s_x1node *next;   /* Next entry with the same hash */
3801   struct s_x1node **from;  /* Previous link */
3802 } x1node;
3803 
3804 /* There is only one instance of the array, which is the following */
3805 static struct s_x1 *x1a;
3806 
3807 /* Allocate a new associative array */
Strsafe_init()3808 void Strsafe_init(){
3809   if( x1a ) return;
3810   x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
3811   if( x1a ){
3812     x1a->size = 1024;
3813     x1a->count = 0;
3814     x1a->tbl = (x1node*)malloc(
3815       (sizeof(x1node) + sizeof(x1node*))*1024 );
3816     if( x1a->tbl==0 ){
3817       free(x1a);
3818       x1a = 0;
3819     }else{
3820       int i;
3821       x1a->ht = (x1node**)&(x1a->tbl[1024]);
3822       for(i=0; i<1024; i++) x1a->ht[i] = 0;
3823     }
3824   }
3825 }
3826 /* Insert a new record into the array.  Return TRUE if successful.
3827 ** Prior data with the same key is NOT overwritten */
Strsafe_insert(data)3828 int Strsafe_insert(data)
3829 char *data;
3830 {
3831   x1node *np;
3832   int h;
3833   int ph;
3834 
3835   if( x1a==0 ) return 0;
3836   ph = strhash(data);
3837   h = ph & (x1a->size-1);
3838   np = x1a->ht[h];
3839   while( np ){
3840     if( strcmp(np->data,data)==0 ){
3841       /* An existing entry with the same key is found. */
3842       /* Fail because overwrite is not allows. */
3843       return 0;
3844     }
3845     np = np->next;
3846   }
3847   if( x1a->count>=x1a->size ){
3848     /* Need to make the hash table bigger */
3849     int i,size;
3850     struct s_x1 array;
3851     array.size = size = x1a->size*2;
3852     array.count = x1a->count;
3853     array.tbl = (x1node*)malloc(
3854       (sizeof(x1node) + sizeof(x1node*))*size );
3855     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
3856     array.ht = (x1node**)&(array.tbl[size]);
3857     for(i=0; i<size; i++) array.ht[i] = 0;
3858     for(i=0; i<x1a->count; i++){
3859       x1node *oldnp, *newnp;
3860       oldnp = &(x1a->tbl[i]);
3861       h = strhash(oldnp->data) & (size-1);
3862       newnp = &(array.tbl[i]);
3863       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
3864       newnp->next = array.ht[h];
3865       newnp->data = oldnp->data;
3866       newnp->from = &(array.ht[h]);
3867       array.ht[h] = newnp;
3868     }
3869     free(x1a->tbl);
3870     /* *x1a = array; *//* copy 'array' */
3871     memcpy(x1a, &array, sizeof(array));
3872   }
3873   /* Insert the new data */
3874   h = ph & (x1a->size-1);
3875   np = &(x1a->tbl[x1a->count++]);
3876   np->data = data;
3877   if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
3878   np->next = x1a->ht[h];
3879   x1a->ht[h] = np;
3880   np->from = &(x1a->ht[h]);
3881   return 1;
3882 }
3883 
3884 /* Return a pointer to data assigned to the given key.  Return NULL
3885 ** if no such key. */
Strsafe_find(key)3886 char *Strsafe_find(key)
3887 char *key;
3888 {
3889   int h;
3890   x1node *np;
3891 
3892   if( x1a==0 ) return 0;
3893   h = strhash(key) & (x1a->size-1);
3894   np = x1a->ht[h];
3895   while( np ){
3896     if( strcmp(np->data,key)==0 ) break;
3897     np = np->next;
3898   }
3899   return np ? np->data : 0;
3900 }
3901 
3902 /* Return a pointer to the (terminal or nonterminal) symbol "x".
3903 ** Create a new symbol if this is the first time "x" has been seen.
3904 */
Symbol_new(x)3905 struct symbol *Symbol_new(x)
3906 char *x;
3907 {
3908   struct symbol *sp;
3909 
3910   sp = Symbol_find(x);
3911   if( sp==0 ){
3912     sp = (struct symbol *)malloc( sizeof(struct symbol) );
3913     MemoryCheck(sp);
3914     sp->name = Strsafe(x);
3915     sp->type = isupper(*x) ? TERMINAL : NONTERMINAL;
3916     sp->rule = 0;
3917     sp->fallback = 0;
3918     sp->prec = -1;
3919     sp->assoc = UNK;
3920     sp->firstset = 0;
3921     sp->lambda = Bo_FALSE;
3922     sp->destructor = 0;
3923     sp->datatype = 0;
3924     Symbol_insert(sp,sp->name);
3925   }
3926   return sp;
3927 }
3928 
3929 /* Compare two symbols for working purposes
3930 **
3931 ** Symbols that begin with upper case letters (terminals or tokens)
3932 ** must sort before symbols that begin with lower case letters
3933 ** (non-terminals).  Other than that, the order does not matter.
3934 **
3935 ** We find experimentally that leaving the symbols in their original
3936 ** order (the order they appeared in the grammar file) gives the
3937 ** smallest parser tables in SQLite.
3938 */
Symbolcmpp(struct symbol ** a,struct symbol ** b)3939 int Symbolcmpp(struct symbol **a, struct symbol **b){
3940   int i1 = (**a).index + 10000000*((**a).name[0]>'Z');
3941   int i2 = (**b).index + 10000000*((**b).name[0]>'Z');
3942   return i1-i2;
3943 }
3944 
3945 /* There is one instance of the following structure for each
3946 ** associative array of type "x2".
3947 */
3948 struct s_x2 {
3949   int size;               /* The number of available slots. */
3950                           /*   Must be a power of 2 greater than or */
3951                           /*   equal to 1 */
3952   int count;              /* Number of currently slots filled */
3953   struct s_x2node *tbl;  /* The data stored here */
3954   struct s_x2node **ht;  /* Hash table for lookups */
3955 };
3956 
3957 /* There is one instance of this structure for every data element
3958 ** in an associative array of type "x2".
3959 */
3960 typedef struct s_x2node {
3961   struct symbol *data;                  /* The data */
3962   char *key;                   /* The key */
3963   struct s_x2node *next;   /* Next entry with the same hash */
3964   struct s_x2node **from;  /* Previous link */
3965 } x2node;
3966 
3967 /* There is only one instance of the array, which is the following */
3968 static struct s_x2 *x2a;
3969 
3970 /* Allocate a new associative array */
Symbol_init()3971 void Symbol_init(){
3972   if( x2a ) return;
3973   x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
3974   if( x2a ){
3975     x2a->size = 128;
3976     x2a->count = 0;
3977     x2a->tbl = (x2node*)malloc(
3978       (sizeof(x2node) + sizeof(x2node*))*128 );
3979     if( x2a->tbl==0 ){
3980       free(x2a);
3981       x2a = 0;
3982     }else{
3983       int i;
3984       x2a->ht = (x2node**)&(x2a->tbl[128]);
3985       for(i=0; i<128; i++) x2a->ht[i] = 0;
3986     }
3987   }
3988 }
3989 /* Insert a new record into the array.  Return TRUE if successful.
3990 ** Prior data with the same key is NOT overwritten */
Symbol_insert(data,key)3991 int Symbol_insert(data,key)
3992 struct symbol *data;
3993 char *key;
3994 {
3995   x2node *np;
3996   int h;
3997   int ph;
3998 
3999   if( x2a==0 ) return 0;
4000   ph = strhash(key);
4001   h = ph & (x2a->size-1);
4002   np = x2a->ht[h];
4003   while( np ){
4004     if( strcmp(np->key,key)==0 ){
4005       /* An existing entry with the same key is found. */
4006       /* Fail because overwrite is not allows. */
4007       return 0;
4008     }
4009     np = np->next;
4010   }
4011   if( x2a->count>=x2a->size ){
4012     /* Need to make the hash table bigger */
4013     int i,size;
4014     struct s_x2 array;
4015     array.size = size = x2a->size*2;
4016     array.count = x2a->count;
4017     array.tbl = (x2node*)malloc(
4018       (sizeof(x2node) + sizeof(x2node*))*size );
4019     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
4020     array.ht = (x2node**)&(array.tbl[size]);
4021     for(i=0; i<size; i++) array.ht[i] = 0;
4022     for(i=0; i<x2a->count; i++){
4023       x2node *oldnp, *newnp;
4024       oldnp = &(x2a->tbl[i]);
4025       h = strhash(oldnp->key) & (size-1);
4026       newnp = &(array.tbl[i]);
4027       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4028       newnp->next = array.ht[h];
4029       newnp->key = oldnp->key;
4030       newnp->data = oldnp->data;
4031       newnp->from = &(array.ht[h]);
4032       array.ht[h] = newnp;
4033     }
4034     free(x2a->tbl);
4035     /* *x2a = array; *//* copy 'array' */
4036     memcpy(x2a, &array, sizeof(array));
4037   }
4038   /* Insert the new data */
4039   h = ph & (x2a->size-1);
4040   np = &(x2a->tbl[x2a->count++]);
4041   np->key = key;
4042   np->data = data;
4043   if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
4044   np->next = x2a->ht[h];
4045   x2a->ht[h] = np;
4046   np->from = &(x2a->ht[h]);
4047   return 1;
4048 }
4049 
4050 /* Return a pointer to data assigned to the given key.  Return NULL
4051 ** if no such key. */
Symbol_find(key)4052 struct symbol *Symbol_find(key)
4053 char *key;
4054 {
4055   int h;
4056   x2node *np;
4057 
4058   if( x2a==0 ) return 0;
4059   h = strhash(key) & (x2a->size-1);
4060   np = x2a->ht[h];
4061   while( np ){
4062     if( strcmp(np->key,key)==0 ) break;
4063     np = np->next;
4064   }
4065   return np ? np->data : 0;
4066 }
4067 
4068 /* Return the n-th data.  Return NULL if n is out of range. */
Symbol_Nth(n)4069 struct symbol *Symbol_Nth(n)
4070 int n;
4071 {
4072   struct symbol *data;
4073   if( x2a && n>0 && n<=x2a->count ){
4074     data = x2a->tbl[n-1].data;
4075   }else{
4076     data = 0;
4077   }
4078   return data;
4079 }
4080 
4081 /* Return the size of the array */
Symbol_count()4082 int Symbol_count()
4083 {
4084   return x2a ? x2a->count : 0;
4085 }
4086 
4087 /* Return an array of pointers to all data in the table.
4088 ** The array is obtained from malloc.  Return NULL if memory allocation
4089 ** problems, or if the array is empty. */
Symbol_arrayof()4090 struct symbol **Symbol_arrayof()
4091 {
4092   struct symbol **array;
4093   int i,size;
4094   if( x2a==0 ) return 0;
4095   size = x2a->count;
4096   array = (struct symbol **)malloc( sizeof(struct symbol *)*size );
4097   if( array ){
4098     for(i=0; i<size; i++) array[i] = x2a->tbl[i].data;
4099   }
4100   return array;
4101 }
4102 
4103 /* Compare two configurations */
Configcmp(a,b)4104 int Configcmp(a,b)
4105 struct config *a;
4106 struct config *b;
4107 {
4108   int x;
4109   x = a->rp->index - b->rp->index;
4110   if( x==0 ) x = a->dot - b->dot;
4111   return x;
4112 }
4113 
4114 /* Compare two states */
statecmp(a,b)4115 PRIVATE int statecmp(a,b)
4116 struct config *a;
4117 struct config *b;
4118 {
4119   int rc;
4120   for(rc=0; rc==0 && a && b;  a=a->bp, b=b->bp){
4121     rc = a->rp->index - b->rp->index;
4122     if( rc==0 ) rc = a->dot - b->dot;
4123   }
4124   if( rc==0 ){
4125     if( a ) rc = 1;
4126     if( b ) rc = -1;
4127   }
4128   return rc;
4129 }
4130 
4131 /* Hash a state */
statehash(a)4132 PRIVATE int statehash(a)
4133 struct config *a;
4134 {
4135   unsigned int h=0;
4136   while( a ){
4137     h = h*571u + (unsigned int)a->rp->index*37u + (unsigned int)a->dot;
4138     a = a->bp;
4139   }
4140   return h;
4141 }
4142 
4143 /* Allocate a new state structure */
State_new()4144 struct state *State_new()
4145 {
4146   struct state *new;
4147   new = (struct state *)malloc( sizeof(struct state) );
4148   MemoryCheck(new);
4149   return new;
4150 }
4151 
4152 /* There is one instance of the following structure for each
4153 ** associative array of type "x3".
4154 */
4155 struct s_x3 {
4156   int size;               /* The number of available slots. */
4157                           /*   Must be a power of 2 greater than or */
4158                           /*   equal to 1 */
4159   int count;              /* Number of currently slots filled */
4160   struct s_x3node *tbl;  /* The data stored here */
4161   struct s_x3node **ht;  /* Hash table for lookups */
4162 };
4163 
4164 /* There is one instance of this structure for every data element
4165 ** in an associative array of type "x3".
4166 */
4167 typedef struct s_x3node {
4168   struct state *data;                  /* The data */
4169   struct config *key;                   /* The key */
4170   struct s_x3node *next;   /* Next entry with the same hash */
4171   struct s_x3node **from;  /* Previous link */
4172 } x3node;
4173 
4174 /* There is only one instance of the array, which is the following */
4175 static struct s_x3 *x3a;
4176 
4177 /* Allocate a new associative array */
State_init()4178 void State_init(){
4179   if( x3a ) return;
4180   x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
4181   if( x3a ){
4182     x3a->size = 128;
4183     x3a->count = 0;
4184     x3a->tbl = (x3node*)malloc(
4185       (sizeof(x3node) + sizeof(x3node*))*128 );
4186     if( x3a->tbl==0 ){
4187       free(x3a);
4188       x3a = 0;
4189     }else{
4190       int i;
4191       x3a->ht = (x3node**)&(x3a->tbl[128]);
4192       for(i=0; i<128; i++) x3a->ht[i] = 0;
4193     }
4194   }
4195 }
4196 /* Insert a new record into the array.  Return TRUE if successful.
4197 ** Prior data with the same key is NOT overwritten */
State_insert(data,key)4198 int State_insert(data,key)
4199 struct state *data;
4200 struct config *key;
4201 {
4202   x3node *np;
4203   int h;
4204   int ph;
4205 
4206   if( x3a==0 ) return 0;
4207   ph = statehash(key);
4208   h = ph & (x3a->size-1);
4209   np = x3a->ht[h];
4210   while( np ){
4211     if( statecmp(np->key,key)==0 ){
4212       /* An existing entry with the same key is found. */
4213       /* Fail because overwrite is not allows. */
4214       return 0;
4215     }
4216     np = np->next;
4217   }
4218   if( x3a->count>=x3a->size ){
4219     /* Need to make the hash table bigger */
4220     int i,size;
4221     struct s_x3 array;
4222     array.size = size = x3a->size*2;
4223     array.count = x3a->count;
4224     array.tbl = (x3node*)malloc(
4225       (sizeof(x3node) + sizeof(x3node*))*size );
4226     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
4227     array.ht = (x3node**)&(array.tbl[size]);
4228     for(i=0; i<size; i++) array.ht[i] = 0;
4229     for(i=0; i<x3a->count; i++){
4230       x3node *oldnp, *newnp;
4231       oldnp = &(x3a->tbl[i]);
4232       h = statehash(oldnp->key) & (size-1);
4233       newnp = &(array.tbl[i]);
4234       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4235       newnp->next = array.ht[h];
4236       newnp->key = oldnp->key;
4237       newnp->data = oldnp->data;
4238       newnp->from = &(array.ht[h]);
4239       array.ht[h] = newnp;
4240     }
4241     free(x3a->tbl);
4242     /* *x3a = array; *//* copy 'array' */
4243     memcpy(x3a, &array, sizeof(array));
4244   }
4245   /* Insert the new data */
4246   h = ph & (x3a->size-1);
4247   np = &(x3a->tbl[x3a->count++]);
4248   np->key = key;
4249   np->data = data;
4250   if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
4251   np->next = x3a->ht[h];
4252   x3a->ht[h] = np;
4253   np->from = &(x3a->ht[h]);
4254   return 1;
4255 }
4256 
4257 /* Return a pointer to data assigned to the given key.  Return NULL
4258 ** if no such key. */
State_find(key)4259 struct state *State_find(key)
4260 struct config *key;
4261 {
4262   int h;
4263   x3node *np;
4264 
4265   if( x3a==0 ) return 0;
4266   h = statehash(key) & (x3a->size-1);
4267   np = x3a->ht[h];
4268   while( np ){
4269     if( statecmp(np->key,key)==0 ) break;
4270     np = np->next;
4271   }
4272   return np ? np->data : 0;
4273 }
4274 
4275 /* Return the size of the array */
State_count(void)4276 int State_count(void)
4277 {
4278   return x3a ? x3a->count : 0;
4279 }
4280 
4281 /* Return an array of pointers to all data in the table.
4282 ** The array is obtained from malloc.  Return NULL if memory allocation
4283 ** problems, or if the array is empty. */
State_arrayof()4284 struct state **State_arrayof()
4285 {
4286   struct state **array;
4287   int i,size;
4288   if( x3a==0 ) return 0;
4289   size = x3a->count;
4290   array = (struct state **)malloc( sizeof(struct state *)*size );
4291   if( array ){
4292     for(i=0; i<size; i++) array[i] = x3a->tbl[i].data;
4293   }
4294   return array;
4295 }
4296 
4297 /* Hash a configuration */
confighash(a)4298 PRIVATE int confighash(a)
4299 struct config *a;
4300 {
4301   int h=0;
4302   h = h*571 + a->rp->index*37 + a->dot;
4303   return h;
4304 }
4305 
4306 /* There is one instance of the following structure for each
4307 ** associative array of type "x4".
4308 */
4309 struct s_x4 {
4310   int size;               /* The number of available slots. */
4311                           /*   Must be a power of 2 greater than or */
4312                           /*   equal to 1 */
4313   int count;              /* Number of currently slots filled */
4314   struct s_x4node *tbl;  /* The data stored here */
4315   struct s_x4node **ht;  /* Hash table for lookups */
4316 };
4317 
4318 /* There is one instance of this structure for every data element
4319 ** in an associative array of type "x4".
4320 */
4321 typedef struct s_x4node {
4322   struct config *data;                  /* The data */
4323   struct s_x4node *next;   /* Next entry with the same hash */
4324   struct s_x4node **from;  /* Previous link */
4325 } x4node;
4326 
4327 /* There is only one instance of the array, which is the following */
4328 static struct s_x4 *x4a;
4329 
4330 /* Allocate a new associative array */
Configtable_init()4331 void Configtable_init(){
4332   if( x4a ) return;
4333   x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
4334   if( x4a ){
4335     x4a->size = 64;
4336     x4a->count = 0;
4337     x4a->tbl = (x4node*)malloc(
4338       (sizeof(x4node) + sizeof(x4node*))*64 );
4339     if( x4a->tbl==0 ){
4340       free(x4a);
4341       x4a = 0;
4342     }else{
4343       int i;
4344       x4a->ht = (x4node**)&(x4a->tbl[64]);
4345       for(i=0; i<64; i++) x4a->ht[i] = 0;
4346     }
4347   }
4348 }
4349 /* Insert a new record into the array.  Return TRUE if successful.
4350 ** Prior data with the same key is NOT overwritten */
Configtable_insert(data)4351 int Configtable_insert(data)
4352 struct config *data;
4353 {
4354   x4node *np;
4355   int h;
4356   int ph;
4357 
4358   if( x4a==0 ) return 0;
4359   ph = confighash(data);
4360   h = ph & (x4a->size-1);
4361   np = x4a->ht[h];
4362   while( np ){
4363     if( Configcmp(np->data,data)==0 ){
4364       /* An existing entry with the same key is found. */
4365       /* Fail because overwrite is not allows. */
4366       return 0;
4367     }
4368     np = np->next;
4369   }
4370   if( x4a->count>=x4a->size ){
4371     /* Need to make the hash table bigger */
4372     int i,size;
4373     struct s_x4 array;
4374     array.size = size = x4a->size*2;
4375     array.count = x4a->count;
4376     array.tbl = (x4node*)malloc(
4377       (sizeof(x4node) + sizeof(x4node*))*size );
4378     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
4379     array.ht = (x4node**)&(array.tbl[size]);
4380     for(i=0; i<size; i++) array.ht[i] = 0;
4381     for(i=0; i<x4a->count; i++){
4382       x4node *oldnp, *newnp;
4383       oldnp = &(x4a->tbl[i]);
4384       h = confighash(oldnp->data) & (size-1);
4385       newnp = &(array.tbl[i]);
4386       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4387       newnp->next = array.ht[h];
4388       newnp->data = oldnp->data;
4389       newnp->from = &(array.ht[h]);
4390       array.ht[h] = newnp;
4391     }
4392     free(x4a->tbl);
4393     /* *x4a = array; *//* copy 'array' */
4394     memcpy(x4a, &array, sizeof(array));
4395   }
4396   /* Insert the new data */
4397   h = ph & (x4a->size-1);
4398   np = &(x4a->tbl[x4a->count++]);
4399   np->data = data;
4400   if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
4401   np->next = x4a->ht[h];
4402   x4a->ht[h] = np;
4403   np->from = &(x4a->ht[h]);
4404   return 1;
4405 }
4406 
4407 /* Return a pointer to data assigned to the given key.  Return NULL
4408 ** if no such key. */
Configtable_find(key)4409 struct config *Configtable_find(key)
4410 struct config *key;
4411 {
4412   int h;
4413   x4node *np;
4414 
4415   if( x4a==0 ) return 0;
4416   h = confighash(key) & (x4a->size-1);
4417   np = x4a->ht[h];
4418   while( np ){
4419     if( Configcmp(np->data,key)==0 ) break;
4420     np = np->next;
4421   }
4422   return np ? np->data : 0;
4423 }
4424 
4425 /* Remove all data from the table.  Pass each data to the function "f"
4426 ** as it is removed.  ("f" may be null to avoid this step.) */
4427 void Configtable_clear(f)
4428 int(*f)(/* struct config * */);
4429 {
4430   int i;
4431   if( x4a==0 || x4a->count==0 ) return;
4432   if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
4433   for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;
4434   x4a->count = 0;
4435   return;
4436 }
4437