xref: /illumos-gate/usr/src/lib/libsqlite/tool/lemon.c (revision 88e1588b)
1 /*
2 ** This file contains all sources (including headers) to the LEMON
3 ** LALR(1) parser generator.  The sources have been combined into a
4 ** single file to make it easy to include LEMON in the source tree
5 ** and Makefile of another program.
6 **
7 ** The author of this program disclaims copyright.
8 */
9 #include <stdio.h>
10 #include <stdarg.h>
11 #include <string.h>
12 #include <ctype.h>
13 #include <stdlib.h>
14 
15 #ifndef __WIN32__
16 #   if defined(_WIN32) || defined(WIN32)
17 #	define __WIN32__
18 #   endif
19 #endif
20 
21 /* #define PRIVATE static */
22 #define PRIVATE
23 
24 #ifdef TEST
25 #define MAXRHS 5       /* Set low to exercise exception code */
26 #else
27 #define MAXRHS 1000
28 #endif
29 
30 char *msort();
31 extern void *malloc();
32 
33 /******** From the file "action.h" *************************************/
34 struct action *Action_new();
35 struct action *Action_sort();
36 
37 /********* From the file "assert.h" ************************************/
38 void myassert();
39 #ifndef NDEBUG
40 #  define assert(X) if(!(X))myassert(__FILE__,__LINE__)
41 #else
42 #  define assert(X)
43 #endif
44 
45 /********** From the file "build.h" ************************************/
46 void FindRulePrecedences();
47 void FindFirstSets();
48 void FindStates();
49 void FindLinks();
50 void FindFollowSets();
51 void FindActions();
52 
53 /********* From the file "configlist.h" *********************************/
54 void Configlist_init(/* void */);
55 struct config *Configlist_add(/* struct rule *, int */);
56 struct config *Configlist_addbasis(/* struct rule *, int */);
57 void Configlist_closure(/* void */);
58 void Configlist_sort(/* void */);
59 void Configlist_sortbasis(/* void */);
60 struct config *Configlist_return(/* void */);
61 struct config *Configlist_basis(/* void */);
62 void Configlist_eat(/* struct config * */);
63 void Configlist_reset(/* void */);
64 
65 /********* From the file "error.h" ***************************************/
66 void ErrorMsg(const char *, int,const char *, ...);
67 
68 /****** From the file "option.h" ******************************************/
69 struct s_options {
70   enum { OPT_FLAG=1,  OPT_INT,  OPT_DBL,  OPT_STR,
71          OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR} type;
72   char *label;
73   char *arg;
74   char *message;
75 };
76 int    OptInit(/* char**,struct s_options*,FILE* */);
77 int    OptNArgs(/* void */);
78 char  *OptArg(/* int */);
79 void   OptErr(/* int */);
80 void   OptPrint(/* void */);
81 
82 /******** From the file "parse.h" *****************************************/
83 void Parse(/* struct lemon *lemp */);
84 
85 /********* From the file "plink.h" ***************************************/
86 struct plink *Plink_new(/* void */);
87 void Plink_add(/* struct plink **, struct config * */);
88 void Plink_copy(/* struct plink **, struct plink * */);
89 void Plink_delete(/* struct plink * */);
90 
91 /********** From the file "report.h" *************************************/
92 void Reprint(/* struct lemon * */);
93 void ReportOutput(/* struct lemon * */);
94 void ReportTable(/* struct lemon * */);
95 void ReportHeader(/* struct lemon * */);
96 void CompressTables(/* struct lemon * */);
97 
98 /********** From the file "set.h" ****************************************/
99 void  SetSize(/* int N */);             /* All sets will be of size N */
100 char *SetNew(/* void */);               /* A new set for element 0..N */
101 void  SetFree(/* char* */);             /* Deallocate a set */
102 
103 int SetAdd(/* char*,int */);            /* Add element to a set */
104 int SetUnion(/* char *A,char *B */);    /* A <- A U B, thru element N */
105 
106 #define SetFind(X,Y) (X[Y])       /* True if Y is in set X */
107 
108 /********** From the file "struct.h" *************************************/
109 /*
110 ** Principal data structures for the LEMON parser generator.
111 */
112 
113 typedef enum {B_FALSE=0, B_TRUE} Boolean;
114 
115 /* Symbols (terminals and nonterminals) of the grammar are stored
116 ** in the following: */
117 struct symbol {
118   char *name;              /* Name of the symbol */
119   int index;               /* Index number for this symbol */
120   enum {
121     TERMINAL,
122     NONTERMINAL
123   } type;                  /* Symbols are all either TERMINALS or NTs */
124   struct rule *rule;       /* Linked list of rules of this (if an NT) */
125   struct symbol *fallback; /* fallback token in case this token doesn't parse */
126   int prec;                /* Precedence if defined (-1 otherwise) */
127   enum e_assoc {
128     LEFT,
129     RIGHT,
130     NONE,
131     UNK
132   } assoc;                 /* Associativity if predecence is defined */
133   char *firstset;          /* First-set for all rules of this symbol */
134   Boolean lambda;          /* True if NT and can generate an empty string */
135   char *destructor;        /* Code which executes whenever this symbol is
136                            ** popped from the stack during error processing */
137   int destructorln;        /* Line number of destructor code */
138   char *datatype;          /* The data type of information held by this
139                            ** object. Only used if type==NONTERMINAL */
140   int dtnum;               /* The data type number.  In the parser, the value
141                            ** stack is a union.  The .yy%d element of this
142                            ** union is the correct data type for this object */
143 };
144 
145 /* Each production rule in the grammar is stored in the following
146 ** structure.  */
147 struct rule {
148   struct symbol *lhs;      /* Left-hand side of the rule */
149   char *lhsalias;          /* Alias for the LHS (NULL if none) */
150   int ruleline;            /* Line number for the rule */
151   int nrhs;                /* Number of RHS symbols */
152   struct symbol **rhs;     /* The RHS symbols */
153   char **rhsalias;         /* An alias for each RHS symbol (NULL if none) */
154   int line;                /* Line number at which code begins */
155   char *code;              /* The code executed when this rule is reduced */
156   struct symbol *precsym;  /* Precedence symbol for this rule */
157   int index;               /* An index number for this rule */
158   Boolean canReduce;       /* True if this rule is ever reduced */
159   struct rule *nextlhs;    /* Next rule with the same LHS */
160   struct rule *next;       /* Next rule in the global list */
161 };
162 
163 /* A configuration is a production rule of the grammar together with
164 ** a mark (dot) showing how much of that rule has been processed so far.
165 ** Configurations also contain a follow-set which is a list of terminal
166 ** symbols which are allowed to immediately follow the end of the rule.
167 ** Every configuration is recorded as an instance of the following: */
168 struct config {
169   struct rule *rp;         /* The rule upon which the configuration is based */
170   int dot;                 /* The parse point */
171   char *fws;               /* Follow-set for this configuration only */
172   struct plink *fplp;      /* Follow-set forward propagation links */
173   struct plink *bplp;      /* Follow-set backwards propagation links */
174   struct state *stp;       /* Pointer to state which contains this */
175   enum {
176     COMPLETE,              /* The status is used during followset and */
177     INCOMPLETE             /*    shift computations */
178   } status;
179   struct config *next;     /* Next configuration in the state */
180   struct config *bp;       /* The next basis configuration */
181 };
182 
183 /* Every shift or reduce operation is stored as one of the following */
184 struct action {
185   struct symbol *sp;       /* The look-ahead symbol */
186   enum e_action {
187     SHIFT,
188     ACCEPT,
189     REDUCE,
190     ERROR,
191     CONFLICT,                /* Was a reduce, but part of a conflict */
192     SH_RESOLVED,             /* Was a shift.  Precedence resolved conflict */
193     RD_RESOLVED,             /* Was reduce.  Precedence resolved conflict */
194     NOT_USED                 /* Deleted by compression */
195   } type;
196   union {
197     struct state *stp;     /* The new state, if a shift */
198     struct rule *rp;       /* The rule, if a reduce */
199   } x;
200   struct action *next;     /* Next action for this state */
201   struct action *collide;  /* Next action with the same hash */
202 };
203 
204 /* Each state of the generated parser's finite state machine
205 ** is encoded as an instance of the following structure. */
206 struct state {
207   struct config *bp;       /* The basis configurations for this state */
208   struct config *cfp;      /* All configurations in this set */
209   int index;               /* Sequencial number for this state */
210   struct action *ap;       /* Array of actions for this state */
211   int nTknAct, nNtAct;     /* Number of actions on terminals and nonterminals */
212   int iTknOfst, iNtOfst;   /* yy_action[] offset for terminals and nonterms */
213   int iDflt;               /* Default action */
214 };
215 #define NO_OFFSET (-2147483647)
216 
217 /* A followset propagation link indicates that the contents of one
218 ** configuration followset should be propagated to another whenever
219 ** the first changes. */
220 struct plink {
221   struct config *cfp;      /* The configuration to which linked */
222   struct plink *next;      /* The next propagate link */
223 };
224 
225 /* The state vector for the entire parser generator is recorded as
226 ** follows.  (LEMON uses no global variables and makes little use of
227 ** static variables.  Fields in the following structure can be thought
228 ** of as begin global variables in the program.) */
229 struct lemon {
230   struct state **sorted;   /* Table of states sorted by state number */
231   struct rule *rule;       /* List of all rules */
232   int nstate;              /* Number of states */
233   int nrule;               /* Number of rules */
234   int nsymbol;             /* Number of terminal and nonterminal symbols */
235   int nterminal;           /* Number of terminal symbols */
236   struct symbol **symbols; /* Sorted array of pointers to symbols */
237   int errorcnt;            /* Number of errors */
238   struct symbol *errsym;   /* The error symbol */
239   char *name;              /* Name of the generated parser */
240   char *arg;               /* Declaration of the 3th argument to parser */
241   char *tokentype;         /* Type of terminal symbols in the parser stack */
242   char *vartype;           /* The default type of non-terminal symbols */
243   char *start;             /* Name of the start symbol for the grammar */
244   char *stacksize;         /* Size of the parser stack */
245   char *include;           /* Code to put at the start of the C file */
246   int  includeln;          /* Line number for start of include code */
247   char *error;             /* Code to execute when an error is seen */
248   int  errorln;            /* Line number for start of error code */
249   char *overflow;          /* Code to execute on a stack overflow */
250   int  overflowln;         /* Line number for start of overflow code */
251   char *failure;           /* Code to execute on parser failure */
252   int  failureln;          /* Line number for start of failure code */
253   char *accept;            /* Code to execute when the parser excepts */
254   int  acceptln;           /* Line number for the start of accept code */
255   char *extracode;         /* Code appended to the generated file */
256   int  extracodeln;        /* Line number for the start of the extra code */
257   char *tokendest;         /* Code to execute to destroy token data */
258   int  tokendestln;        /* Line number for token destroyer code */
259   char *vardest;           /* Code for the default non-terminal destructor */
260   int  vardestln;          /* Line number for default non-term destructor code*/
261   char *filename;          /* Name of the input file */
262   char *outname;           /* Name of the current output file */
263   char *tokenprefix;       /* A prefix added to token names in the .h file */
264   int nconflict;           /* Number of parsing conflicts */
265   int tablesize;           /* Size of the parse tables */
266   int basisflag;           /* Print only basis configurations */
267   int has_fallback;        /* True if any %fallback is seen in the grammer */
268   char *argv0;             /* Name of the program */
269 };
270 
271 #define MemoryCheck(X) if((X)==0){ \
272   extern void memory_error(); \
273   memory_error(); \
274 }
275 
276 /**************** From the file "table.h" *********************************/
277 /*
278 ** All code in this file has been automatically generated
279 ** from a specification in the file
280 **              "table.q"
281 ** by the associative array code building program "aagen".
282 ** Do not edit this file!  Instead, edit the specification
283 ** file, then rerun aagen.
284 */
285 /*
286 ** Code for processing tables in the LEMON parser generator.
287 */
288 
289 /* Routines for handling a strings */
290 
291 char *Strsafe();
292 
293 void Strsafe_init(/* void */);
294 int Strsafe_insert(/* char * */);
295 char *Strsafe_find(/* char * */);
296 
297 /* Routines for handling symbols of the grammar */
298 
299 struct symbol *Symbol_new();
300 int Symbolcmpp(/* struct symbol **, struct symbol ** */);
301 void Symbol_init(/* void */);
302 int Symbol_insert(/* struct symbol *, char * */);
303 struct symbol *Symbol_find(/* char * */);
304 struct symbol *Symbol_Nth(/* int */);
305 int Symbol_count(/*  */);
306 struct symbol **Symbol_arrayof(/*  */);
307 
308 /* Routines to manage the state table */
309 
310 int Configcmp(/* struct config *, struct config * */);
311 struct state *State_new();
312 void State_init(/* void */);
313 int State_insert(/* struct state *, struct config * */);
314 struct state *State_find(/* struct config * */);
315 struct state **State_arrayof(/*  */);
316 
317 /* Routines used for efficiency in Configlist_add */
318 
319 void Configtable_init(/* void */);
320 int Configtable_insert(/* struct config * */);
321 struct config *Configtable_find(/* struct config * */);
322 void Configtable_clear(/* int(*)(struct config *) */);
323 /****************** From the file "action.c" *******************************/
324 /*
325 ** Routines processing parser actions in the LEMON parser generator.
326 */
327 
328 /* Allocate a new parser action */
Action_new()329 struct action *Action_new(){
330   static struct action *freelist = 0;
331   struct action *new;
332 
333   if( freelist==0 ){
334     int i;
335     int amt = 100;
336     freelist = (struct action *)malloc( sizeof(struct action)*amt );
337     if( freelist==0 ){
338       fprintf(stderr,"Unable to allocate memory for a new parser action.");
339       exit(1);
340     }
341     for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
342     freelist[amt-1].next = 0;
343   }
344   new = freelist;
345   freelist = freelist->next;
346   return new;
347 }
348 
349 /* Compare two actions */
actioncmp(ap1,ap2)350 static int actioncmp(ap1,ap2)
351 struct action *ap1;
352 struct action *ap2;
353 {
354   int rc;
355   rc = ap1->sp->index - ap2->sp->index;
356   if( rc==0 ) rc = (int)ap1->type - (int)ap2->type;
357   if( rc==0 ){
358     assert( ap1->type==REDUCE || ap1->type==RD_RESOLVED || ap1->type==CONFLICT);
359     assert( ap2->type==REDUCE || ap2->type==RD_RESOLVED || ap2->type==CONFLICT);
360     rc = ap1->x.rp->index - ap2->x.rp->index;
361   }
362   return rc;
363 }
364 
365 /* Sort parser actions */
Action_sort(ap)366 struct action *Action_sort(ap)
367 struct action *ap;
368 {
369   ap = (struct action *)msort((char *)ap,(char **)&ap->next,actioncmp);
370   return ap;
371 }
372 
Action_add(app,type,sp,arg)373 void Action_add(app,type,sp,arg)
374 struct action **app;
375 enum e_action type;
376 struct symbol *sp;
377 char *arg;
378 {
379   struct action *new;
380   new = Action_new();
381   new->next = *app;
382   *app = new;
383   new->type = type;
384   new->sp = sp;
385   if( type==SHIFT ){
386     new->x.stp = (struct state *)arg;
387   }else{
388     new->x.rp = (struct rule *)arg;
389   }
390 }
391 /********************** New code to implement the "acttab" module ***********/
392 /*
393 ** This module implements routines use to construct the yy_action[] table.
394 */
395 
396 /*
397 ** The state of the yy_action table under construction is an instance of
398 ** the following structure
399 */
400 typedef struct acttab acttab;
401 struct acttab {
402   int nAction;                 /* Number of used slots in aAction[] */
403   int nActionAlloc;            /* Slots allocated for aAction[] */
404   struct {
405     int lookahead;             /* Value of the lookahead token */
406     int action;                /* Action to take on the given lookahead */
407   } *aAction,                  /* The yy_action[] table under construction */
408     *aLookahead;               /* A single new transaction set */
409   int mnLookahead;             /* Minimum aLookahead[].lookahead */
410   int mnAction;                /* Action associated with mnLookahead */
411   int mxLookahead;             /* Maximum aLookahead[].lookahead */
412   int nLookahead;              /* Used slots in aLookahead[] */
413   int nLookaheadAlloc;         /* Slots allocated in aLookahead[] */
414 };
415 
416 /* Return the number of entries in the yy_action table */
417 #define acttab_size(X) ((X)->nAction)
418 
419 /* The value for the N-th entry in yy_action */
420 #define acttab_yyaction(X,N)  ((X)->aAction[N].action)
421 
422 /* The value for the N-th entry in yy_lookahead */
423 #define acttab_yylookahead(X,N)  ((X)->aAction[N].lookahead)
424 
425 /* Free all memory associated with the given acttab */
acttab_free(acttab * p)426 void acttab_free(acttab *p){
427   free( p->aAction );
428   free( p->aLookahead );
429   free( p );
430 }
431 
432 /* Allocate a new acttab structure */
acttab_alloc(void)433 acttab *acttab_alloc(void){
434   acttab *p = malloc( sizeof(*p) );
435   if( p==0 ){
436     fprintf(stderr,"Unable to allocate memory for a new acttab.");
437     exit(1);
438   }
439   memset(p, 0, sizeof(*p));
440   return p;
441 }
442 
443 /* Add a new action to the current transaction set
444 */
acttab_action(acttab * p,int lookahead,int action)445 void acttab_action(acttab *p, int lookahead, int action){
446   if( p->nLookahead>=p->nLookaheadAlloc ){
447     p->nLookaheadAlloc += 25;
448     p->aLookahead = realloc( p->aLookahead,
449                              sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
450     if( p->aLookahead==0 ){
451       fprintf(stderr,"malloc failed\n");
452       exit(1);
453     }
454   }
455   if( p->nLookahead==0 ){
456     p->mxLookahead = lookahead;
457     p->mnLookahead = lookahead;
458     p->mnAction = action;
459   }else{
460     if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
461     if( p->mnLookahead>lookahead ){
462       p->mnLookahead = lookahead;
463       p->mnAction = action;
464     }
465   }
466   p->aLookahead[p->nLookahead].lookahead = lookahead;
467   p->aLookahead[p->nLookahead].action = action;
468   p->nLookahead++;
469 }
470 
471 /*
472 ** Add the transaction set built up with prior calls to acttab_action()
473 ** into the current action table.  Then reset the transaction set back
474 ** to an empty set in preparation for a new round of acttab_action() calls.
475 **
476 ** Return the offset into the action table of the new transaction.
477 */
acttab_insert(acttab * p)478 int acttab_insert(acttab *p){
479   int i, j, k, n;
480   assert( p->nLookahead>0 );
481 
482   /* Make sure we have enough space to hold the expanded action table
483   ** in the worst case.  The worst case occurs if the transaction set
484   ** must be appended to the current action table
485   */
486   n = p->mxLookahead + 1;
487   if( p->nAction + n >= p->nActionAlloc ){
488     int oldAlloc = p->nActionAlloc;
489     p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
490     p->aAction = realloc( p->aAction,
491                           sizeof(p->aAction[0])*p->nActionAlloc);
492     if( p->aAction==0 ){
493       fprintf(stderr,"malloc failed\n");
494       exit(1);
495     }
496     for(i=oldAlloc; i<p->nActionAlloc; i++){
497       p->aAction[i].lookahead = -1;
498       p->aAction[i].action = -1;
499     }
500   }
501 
502   /* Scan the existing action table looking for an offset where we can
503   ** insert the current transaction set.  Fall out of the loop when that
504   ** offset is found.  In the worst case, we fall out of the loop when
505   ** i reaches p->nAction, which means we append the new transaction set.
506   **
507   ** i is the index in p->aAction[] where p->mnLookahead is inserted.
508   */
509   for(i=0; i<p->nAction+p->mnLookahead; i++){
510     if( p->aAction[i].lookahead<0 ){
511       for(j=0; j<p->nLookahead; j++){
512         k = p->aLookahead[j].lookahead - p->mnLookahead + i;
513         if( k<0 ) break;
514         if( p->aAction[k].lookahead>=0 ) break;
515       }
516       if( j<p->nLookahead ) continue;
517       for(j=0; j<p->nAction; j++){
518         if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
519       }
520       if( j==p->nAction ){
521         break;  /* Fits in empty slots */
522       }
523     }else if( p->aAction[i].lookahead==p->mnLookahead ){
524       if( p->aAction[i].action!=p->mnAction ) continue;
525       for(j=0; j<p->nLookahead; j++){
526         k = p->aLookahead[j].lookahead - p->mnLookahead + i;
527         if( k<0 || k>=p->nAction ) break;
528         if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
529         if( p->aLookahead[j].action!=p->aAction[k].action ) break;
530       }
531       if( j<p->nLookahead ) continue;
532       n = 0;
533       for(j=0; j<p->nAction; j++){
534         if( p->aAction[j].lookahead<0 ) continue;
535         if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
536       }
537       if( n==p->nLookahead ){
538         break;  /* Same as a prior transaction set */
539       }
540     }
541   }
542   /* Insert transaction set at index i. */
543   for(j=0; j<p->nLookahead; j++){
544     k = p->aLookahead[j].lookahead - p->mnLookahead + i;
545     p->aAction[k] = p->aLookahead[j];
546     if( k>=p->nAction ) p->nAction = k+1;
547   }
548   p->nLookahead = 0;
549 
550   /* Return the offset that is added to the lookahead in order to get the
551   ** index into yy_action of the action */
552   return i - p->mnLookahead;
553 }
554 
555 /********************** From the file "assert.c" ****************************/
556 /*
557 ** A more efficient way of handling assertions.
558 */
myassert(file,line)559 void myassert(file,line)
560 char *file;
561 int line;
562 {
563   fprintf(stderr,"Assertion failed on line %d of file \"%s\"\n",line,file);
564   exit(1);
565 }
566 /********************** From the file "build.c" *****************************/
567 /*
568 ** Routines to construction the finite state machine for the LEMON
569 ** parser generator.
570 */
571 
572 /* Find a precedence symbol of every rule in the grammar.
573 **
574 ** Those rules which have a precedence symbol coded in the input
575 ** grammar using the "[symbol]" construct will already have the
576 ** rp->precsym field filled.  Other rules take as their precedence
577 ** symbol the first RHS symbol with a defined precedence.  If there
578 ** are not RHS symbols with a defined precedence, the precedence
579 ** symbol field is left blank.
580 */
FindRulePrecedences(xp)581 void FindRulePrecedences(xp)
582 struct lemon *xp;
583 {
584   struct rule *rp;
585   for(rp=xp->rule; rp; rp=rp->next){
586     if( rp->precsym==0 ){
587       int i;
588       for(i=0; i<rp->nrhs; i++){
589         if( rp->rhs[i]->prec>=0 ){
590           rp->precsym = rp->rhs[i];
591           break;
592 	}
593       }
594     }
595   }
596   return;
597 }
598 
599 /* Find all nonterminals which will generate the empty string.
600 ** Then go back and compute the first sets of every nonterminal.
601 ** The first set is the set of all terminal symbols which can begin
602 ** a string generated by that nonterminal.
603 */
FindFirstSets(lemp)604 void FindFirstSets(lemp)
605 struct lemon *lemp;
606 {
607   int i;
608   struct rule *rp;
609   int progress;
610 
611   for(i=0; i<lemp->nsymbol; i++){
612     lemp->symbols[i]->lambda = B_FALSE;
613   }
614   for(i=lemp->nterminal; i<lemp->nsymbol; i++){
615     lemp->symbols[i]->firstset = SetNew();
616   }
617 
618   /* First compute all lambdas */
619   do{
620     progress = 0;
621     for(rp=lemp->rule; rp; rp=rp->next){
622       if( rp->lhs->lambda ) continue;
623       for(i=0; i<rp->nrhs; i++){
624          if( rp->rhs[i]->lambda==B_FALSE ) break;
625       }
626       if( i==rp->nrhs ){
627         rp->lhs->lambda = B_TRUE;
628         progress = 1;
629       }
630     }
631   }while( progress );
632 
633   /* Now compute all first sets */
634   do{
635     struct symbol *s1, *s2;
636     progress = 0;
637     for(rp=lemp->rule; rp; rp=rp->next){
638       s1 = rp->lhs;
639       for(i=0; i<rp->nrhs; i++){
640         s2 = rp->rhs[i];
641         if( s2->type==TERMINAL ){
642           progress += SetAdd(s1->firstset,s2->index);
643           break;
644 	}else if( s1==s2 ){
645           if( s1->lambda==B_FALSE ) break;
646 	}else{
647           progress += SetUnion(s1->firstset,s2->firstset);
648           if( s2->lambda==B_FALSE ) break;
649 	}
650       }
651     }
652   }while( progress );
653   return;
654 }
655 
656 /* Compute all LR(0) states for the grammar.  Links
657 ** are added to between some states so that the LR(1) follow sets
658 ** can be computed later.
659 */
660 PRIVATE struct state *getstate(/* struct lemon * */);  /* forward reference */
FindStates(lemp)661 void FindStates(lemp)
662 struct lemon *lemp;
663 {
664   struct symbol *sp;
665   struct rule *rp;
666 
667   Configlist_init();
668 
669   /* Find the start symbol */
670   if( lemp->start ){
671     sp = Symbol_find(lemp->start);
672     if( sp==0 ){
673       ErrorMsg(lemp->filename,0,
674 "The specified start symbol \"%s\" is not \
675 in a nonterminal of the grammar.  \"%s\" will be used as the start \
676 symbol instead.",lemp->start,lemp->rule->lhs->name);
677       lemp->errorcnt++;
678       sp = lemp->rule->lhs;
679     }
680   }else{
681     sp = lemp->rule->lhs;
682   }
683 
684   /* Make sure the start symbol doesn't occur on the right-hand side of
685   ** any rule.  Report an error if it does.  (YACC would generate a new
686   ** start symbol in this case.) */
687   for(rp=lemp->rule; rp; rp=rp->next){
688     int i;
689     for(i=0; i<rp->nrhs; i++){
690       if( rp->rhs[i]==sp ){
691         ErrorMsg(lemp->filename,0,
692 "The start symbol \"%s\" occurs on the \
693 right-hand side of a rule. This will result in a parser which \
694 does not work properly.",sp->name);
695         lemp->errorcnt++;
696       }
697     }
698   }
699 
700   /* The basis configuration set for the first state
701   ** is all rules which have the start symbol as their
702   ** left-hand side */
703   for(rp=sp->rule; rp; rp=rp->nextlhs){
704     struct config *newcfp;
705     newcfp = Configlist_addbasis(rp,0);
706     SetAdd(newcfp->fws,0);
707   }
708 
709   /* Compute the first state.  All other states will be
710   ** computed automatically during the computation of the first one.
711   ** The returned pointer to the first state is not used. */
712   (void)getstate(lemp);
713   return;
714 }
715 
716 /* Return a pointer to a state which is described by the configuration
717 ** list which has been built from calls to Configlist_add.
718 */
719 PRIVATE void buildshifts(/* struct lemon *, struct state * */); /* Forwd ref */
getstate(lemp)720 PRIVATE struct state *getstate(lemp)
721 struct lemon *lemp;
722 {
723   struct config *cfp, *bp;
724   struct state *stp;
725 
726   /* Extract the sorted basis of the new state.  The basis was constructed
727   ** by prior calls to "Configlist_addbasis()". */
728   Configlist_sortbasis();
729   bp = Configlist_basis();
730 
731   /* Get a state with the same basis */
732   stp = State_find(bp);
733   if( stp ){
734     /* A state with the same basis already exists!  Copy all the follow-set
735     ** propagation links from the state under construction into the
736     ** preexisting state, then return a pointer to the preexisting state */
737     struct config *x, *y;
738     for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
739       Plink_copy(&y->bplp,x->bplp);
740       Plink_delete(x->fplp);
741       x->fplp = x->bplp = 0;
742     }
743     cfp = Configlist_return();
744     Configlist_eat(cfp);
745   }else{
746     /* This really is a new state.  Construct all the details */
747     Configlist_closure(lemp);    /* Compute the configuration closure */
748     Configlist_sort();           /* Sort the configuration closure */
749     cfp = Configlist_return();   /* Get a pointer to the config list */
750     stp = State_new();           /* A new state structure */
751     MemoryCheck(stp);
752     stp->bp = bp;                /* Remember the configuration basis */
753     stp->cfp = cfp;              /* Remember the configuration closure */
754     stp->index = lemp->nstate++; /* Every state gets a sequence number */
755     stp->ap = 0;                 /* No actions, yet. */
756     State_insert(stp,stp->bp);   /* Add to the state table */
757     buildshifts(lemp,stp);       /* Recursively compute successor states */
758   }
759   return stp;
760 }
761 
762 /* Construct all successor states to the given state.  A "successor"
763 ** state is any state which can be reached by a shift action.
764 */
buildshifts(lemp,stp)765 PRIVATE void buildshifts(lemp,stp)
766 struct lemon *lemp;
767 struct state *stp;     /* The state from which successors are computed */
768 {
769   struct config *cfp;  /* For looping thru the config closure of "stp" */
770   struct config *bcfp; /* For the inner loop on config closure of "stp" */
771   struct config *new;  /* */
772   struct symbol *sp;   /* Symbol following the dot in configuration "cfp" */
773   struct symbol *bsp;  /* Symbol following the dot in configuration "bcfp" */
774   struct state *newstp; /* A pointer to a successor state */
775 
776   /* Each configuration becomes complete after it contibutes to a successor
777   ** state.  Initially, all configurations are incomplete */
778   for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
779 
780   /* Loop through all configurations of the state "stp" */
781   for(cfp=stp->cfp; cfp; cfp=cfp->next){
782     if( cfp->status==COMPLETE ) continue;    /* Already used by inner loop */
783     if( cfp->dot>=cfp->rp->nrhs ) continue;  /* Can't shift this config */
784     Configlist_reset();                      /* Reset the new config set */
785     sp = cfp->rp->rhs[cfp->dot];             /* Symbol after the dot */
786 
787     /* For every configuration in the state "stp" which has the symbol "sp"
788     ** following its dot, add the same configuration to the basis set under
789     ** construction but with the dot shifted one symbol to the right. */
790     for(bcfp=cfp; bcfp; bcfp=bcfp->next){
791       if( bcfp->status==COMPLETE ) continue;    /* Already used */
792       if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
793       bsp = bcfp->rp->rhs[bcfp->dot];           /* Get symbol after dot */
794       if( bsp!=sp ) continue;                   /* Must be same as for "cfp" */
795       bcfp->status = COMPLETE;                  /* Mark this config as used */
796       new = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
797       Plink_add(&new->bplp,bcfp);
798     }
799 
800     /* Get a pointer to the state described by the basis configuration set
801     ** constructed in the preceding loop */
802     newstp = getstate(lemp);
803 
804     /* The state "newstp" is reached from the state "stp" by a shift action
805     ** on the symbol "sp" */
806     Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
807   }
808 }
809 
810 /*
811 ** Construct the propagation links
812 */
FindLinks(lemp)813 void FindLinks(lemp)
814 struct lemon *lemp;
815 {
816   int i;
817   struct config *cfp, *other;
818   struct state *stp;
819   struct plink *plp;
820 
821   /* Housekeeping detail:
822   ** Add to every propagate link a pointer back to the state to
823   ** which the link is attached. */
824   for(i=0; i<lemp->nstate; i++){
825     stp = lemp->sorted[i];
826     for(cfp=stp->cfp; cfp; cfp=cfp->next){
827       cfp->stp = stp;
828     }
829   }
830 
831   /* Convert all backlinks into forward links.  Only the forward
832   ** links are used in the follow-set computation. */
833   for(i=0; i<lemp->nstate; i++){
834     stp = lemp->sorted[i];
835     for(cfp=stp->cfp; cfp; cfp=cfp->next){
836       for(plp=cfp->bplp; plp; plp=plp->next){
837         other = plp->cfp;
838         Plink_add(&other->fplp,cfp);
839       }
840     }
841   }
842 }
843 
844 /* Compute all followsets.
845 **
846 ** A followset is the set of all symbols which can come immediately
847 ** after a configuration.
848 */
FindFollowSets(lemp)849 void FindFollowSets(lemp)
850 struct lemon *lemp;
851 {
852   int i;
853   struct config *cfp;
854   struct plink *plp;
855   int progress;
856   int change;
857 
858   for(i=0; i<lemp->nstate; i++){
859     for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
860       cfp->status = INCOMPLETE;
861     }
862   }
863 
864   do{
865     progress = 0;
866     for(i=0; i<lemp->nstate; i++){
867       for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
868         if( cfp->status==COMPLETE ) continue;
869         for(plp=cfp->fplp; plp; plp=plp->next){
870           change = SetUnion(plp->cfp->fws,cfp->fws);
871           if( change ){
872             plp->cfp->status = INCOMPLETE;
873             progress = 1;
874 	  }
875 	}
876         cfp->status = COMPLETE;
877       }
878     }
879   }while( progress );
880 }
881 
882 static int resolve_conflict();
883 
884 /* Compute the reduce actions, and resolve conflicts.
885 */
FindActions(lemp)886 void FindActions(lemp)
887 struct lemon *lemp;
888 {
889   int i,j;
890   struct config *cfp;
891   struct state *stp;
892   struct symbol *sp;
893   struct rule *rp;
894 
895   /* Add all of the reduce actions
896   ** A reduce action is added for each element of the followset of
897   ** a configuration which has its dot at the extreme right.
898   */
899   for(i=0; i<lemp->nstate; i++){   /* Loop over all states */
900     stp = lemp->sorted[i];
901     for(cfp=stp->cfp; cfp; cfp=cfp->next){  /* Loop over all configurations */
902       if( cfp->rp->nrhs==cfp->dot ){        /* Is dot at extreme right? */
903         for(j=0; j<lemp->nterminal; j++){
904           if( SetFind(cfp->fws,j) ){
905             /* Add a reduce action to the state "stp" which will reduce by the
906             ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
907             Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
908           }
909 	}
910       }
911     }
912   }
913 
914   /* Add the accepting token */
915   if( lemp->start ){
916     sp = Symbol_find(lemp->start);
917     if( sp==0 ) sp = lemp->rule->lhs;
918   }else{
919     sp = lemp->rule->lhs;
920   }
921   /* Add to the first state (which is always the starting state of the
922   ** finite state machine) an action to ACCEPT if the lookahead is the
923   ** start nonterminal.  */
924   Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
925 
926   /* Resolve conflicts */
927   for(i=0; i<lemp->nstate; i++){
928     struct action *ap, *nap;
929     struct state *stp;
930     stp = lemp->sorted[i];
931     assert( stp->ap );
932     stp->ap = Action_sort(stp->ap);
933     for(ap=stp->ap; ap && ap->next; ap=ap->next){
934       for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
935          /* The two actions "ap" and "nap" have the same lookahead.
936          ** Figure out which one should be used */
937          lemp->nconflict += resolve_conflict(ap,nap,lemp->errsym);
938       }
939     }
940   }
941 
942   /* Report an error for each rule that can never be reduced. */
943   for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = B_FALSE;
944   for(i=0; i<lemp->nstate; i++){
945     struct action *ap;
946     for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
947       if( ap->type==REDUCE ) ap->x.rp->canReduce = B_TRUE;
948     }
949   }
950   for(rp=lemp->rule; rp; rp=rp->next){
951     if( rp->canReduce ) continue;
952     ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
953     lemp->errorcnt++;
954   }
955 }
956 
957 /* Resolve a conflict between the two given actions.  If the
958 ** conflict can't be resolve, return non-zero.
959 **
960 ** NO LONGER TRUE:
961 **   To resolve a conflict, first look to see if either action
962 **   is on an error rule.  In that case, take the action which
963 **   is not associated with the error rule.  If neither or both
964 **   actions are associated with an error rule, then try to
965 **   use precedence to resolve the conflict.
966 **
967 ** If either action is a SHIFT, then it must be apx.  This
968 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
969 */
resolve_conflict(apx,apy,errsym)970 static int resolve_conflict(apx,apy,errsym)
971 struct action *apx;
972 struct action *apy;
973 struct symbol *errsym;   /* The error symbol (if defined.  NULL otherwise) */
974 {
975   struct symbol *spx, *spy;
976   int errcnt = 0;
977   assert( apx->sp==apy->sp );  /* Otherwise there would be no conflict */
978   if( apx->type==SHIFT && apy->type==REDUCE ){
979     spx = apx->sp;
980     spy = apy->x.rp->precsym;
981     if( spy==0 || spx->prec<0 || spy->prec<0 ){
982       /* Not enough precedence information. */
983       apy->type = CONFLICT;
984       errcnt++;
985     }else if( spx->prec>spy->prec ){    /* Lower precedence wins */
986       apy->type = RD_RESOLVED;
987     }else if( spx->prec<spy->prec ){
988       apx->type = SH_RESOLVED;
989     }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
990       apy->type = RD_RESOLVED;                             /* associativity */
991     }else if( spx->prec==spy->prec && spx->assoc==LEFT ){  /* to break tie */
992       apx->type = SH_RESOLVED;
993     }else{
994       assert( spx->prec==spy->prec && spx->assoc==NONE );
995       apy->type = CONFLICT;
996       errcnt++;
997     }
998   }else if( apx->type==REDUCE && apy->type==REDUCE ){
999     spx = apx->x.rp->precsym;
1000     spy = apy->x.rp->precsym;
1001     if( spx==0 || spy==0 || spx->prec<0 ||
1002     spy->prec<0 || spx->prec==spy->prec ){
1003       apy->type = CONFLICT;
1004       errcnt++;
1005     }else if( spx->prec>spy->prec ){
1006       apy->type = RD_RESOLVED;
1007     }else if( spx->prec<spy->prec ){
1008       apx->type = RD_RESOLVED;
1009     }
1010   }else{
1011     assert(
1012       apx->type==SH_RESOLVED ||
1013       apx->type==RD_RESOLVED ||
1014       apx->type==CONFLICT ||
1015       apy->type==SH_RESOLVED ||
1016       apy->type==RD_RESOLVED ||
1017       apy->type==CONFLICT
1018     );
1019     /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1020     ** REDUCEs on the list.  If we reach this point it must be because
1021     ** the parser conflict had already been resolved. */
1022   }
1023   return errcnt;
1024 }
1025 /********************* From the file "configlist.c" *************************/
1026 /*
1027 ** Routines to processing a configuration list and building a state
1028 ** in the LEMON parser generator.
1029 */
1030 
1031 static struct config *freelist = 0;      /* List of free configurations */
1032 static struct config *current = 0;       /* Top of list of configurations */
1033 static struct config **currentend = 0;   /* Last on list of configs */
1034 static struct config *basis = 0;         /* Top of list of basis configs */
1035 static struct config **basisend = 0;     /* End of list of basis configs */
1036 
1037 /* Return a pointer to a new configuration */
newconfig()1038 PRIVATE struct config *newconfig(){
1039   struct config *new;
1040   if( freelist==0 ){
1041     int i;
1042     int amt = 3;
1043     freelist = (struct config *)malloc( sizeof(struct config)*amt );
1044     if( freelist==0 ){
1045       fprintf(stderr,"Unable to allocate memory for a new configuration.");
1046       exit(1);
1047     }
1048     for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1049     freelist[amt-1].next = 0;
1050   }
1051   new = freelist;
1052   freelist = freelist->next;
1053   return new;
1054 }
1055 
1056 /* The configuration "old" is no longer used */
deleteconfig(old)1057 PRIVATE void deleteconfig(old)
1058 struct config *old;
1059 {
1060   old->next = freelist;
1061   freelist = old;
1062 }
1063 
1064 /* Initialized the configuration list builder */
Configlist_init()1065 void Configlist_init(){
1066   current = 0;
1067   currentend = &current;
1068   basis = 0;
1069   basisend = &basis;
1070   Configtable_init();
1071   return;
1072 }
1073 
1074 /* Initialized the configuration list builder */
Configlist_reset()1075 void Configlist_reset(){
1076   current = 0;
1077   currentend = &current;
1078   basis = 0;
1079   basisend = &basis;
1080   Configtable_clear(0);
1081   return;
1082 }
1083 
1084 /* Add another configuration to the configuration list */
Configlist_add(rp,dot)1085 struct config *Configlist_add(rp,dot)
1086 struct rule *rp;    /* The rule */
1087 int dot;            /* Index into the RHS of the rule where the dot goes */
1088 {
1089   struct config *cfp, model;
1090 
1091   assert( currentend!=0 );
1092   model.rp = rp;
1093   model.dot = dot;
1094   cfp = Configtable_find(&model);
1095   if( cfp==0 ){
1096     cfp = newconfig();
1097     cfp->rp = rp;
1098     cfp->dot = dot;
1099     cfp->fws = SetNew();
1100     cfp->stp = 0;
1101     cfp->fplp = cfp->bplp = 0;
1102     cfp->next = 0;
1103     cfp->bp = 0;
1104     *currentend = cfp;
1105     currentend = &cfp->next;
1106     Configtable_insert(cfp);
1107   }
1108   return cfp;
1109 }
1110 
1111 /* Add a basis configuration to the configuration list */
Configlist_addbasis(rp,dot)1112 struct config *Configlist_addbasis(rp,dot)
1113 struct rule *rp;
1114 int dot;
1115 {
1116   struct config *cfp, model;
1117 
1118   assert( basisend!=0 );
1119   assert( currentend!=0 );
1120   model.rp = rp;
1121   model.dot = dot;
1122   cfp = Configtable_find(&model);
1123   if( cfp==0 ){
1124     cfp = newconfig();
1125     cfp->rp = rp;
1126     cfp->dot = dot;
1127     cfp->fws = SetNew();
1128     cfp->stp = 0;
1129     cfp->fplp = cfp->bplp = 0;
1130     cfp->next = 0;
1131     cfp->bp = 0;
1132     *currentend = cfp;
1133     currentend = &cfp->next;
1134     *basisend = cfp;
1135     basisend = &cfp->bp;
1136     Configtable_insert(cfp);
1137   }
1138   return cfp;
1139 }
1140 
1141 /* Compute the closure of the configuration list */
Configlist_closure(lemp)1142 void Configlist_closure(lemp)
1143 struct lemon *lemp;
1144 {
1145   struct config *cfp, *newcfp;
1146   struct rule *rp, *newrp;
1147   struct symbol *sp, *xsp;
1148   int i, dot;
1149 
1150   assert( currentend!=0 );
1151   for(cfp=current; cfp; cfp=cfp->next){
1152     rp = cfp->rp;
1153     dot = cfp->dot;
1154     if( dot>=rp->nrhs ) continue;
1155     sp = rp->rhs[dot];
1156     if( sp->type==NONTERMINAL ){
1157       if( sp->rule==0 && sp!=lemp->errsym ){
1158         ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1159           sp->name);
1160         lemp->errorcnt++;
1161       }
1162       for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1163         newcfp = Configlist_add(newrp,0);
1164         for(i=dot+1; i<rp->nrhs; i++){
1165           xsp = rp->rhs[i];
1166           if( xsp->type==TERMINAL ){
1167             SetAdd(newcfp->fws,xsp->index);
1168             break;
1169 	  }else{
1170             SetUnion(newcfp->fws,xsp->firstset);
1171             if( xsp->lambda==B_FALSE ) break;
1172 	  }
1173 	}
1174         if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1175       }
1176     }
1177   }
1178   return;
1179 }
1180 
1181 /* Sort the configuration list */
Configlist_sort()1182 void Configlist_sort(){
1183   current = (struct config *)msort((char *)current,(char **)&(current->next),Configcmp);
1184   currentend = 0;
1185   return;
1186 }
1187 
1188 /* Sort the basis configuration list */
Configlist_sortbasis()1189 void Configlist_sortbasis(){
1190   basis = (struct config *)msort((char *)current,(char **)&(current->bp),Configcmp);
1191   basisend = 0;
1192   return;
1193 }
1194 
1195 /* Return a pointer to the head of the configuration list and
1196 ** reset the list */
Configlist_return()1197 struct config *Configlist_return(){
1198   struct config *old;
1199   old = current;
1200   current = 0;
1201   currentend = 0;
1202   return old;
1203 }
1204 
1205 /* Return a pointer to the head of the configuration list and
1206 ** reset the list */
Configlist_basis()1207 struct config *Configlist_basis(){
1208   struct config *old;
1209   old = basis;
1210   basis = 0;
1211   basisend = 0;
1212   return old;
1213 }
1214 
1215 /* Free all elements of the given configuration list */
Configlist_eat(cfp)1216 void Configlist_eat(cfp)
1217 struct config *cfp;
1218 {
1219   struct config *nextcfp;
1220   for(; cfp; cfp=nextcfp){
1221     nextcfp = cfp->next;
1222     assert( cfp->fplp==0 );
1223     assert( cfp->bplp==0 );
1224     if( cfp->fws ) SetFree(cfp->fws);
1225     deleteconfig(cfp);
1226   }
1227   return;
1228 }
1229 /***************** From the file "error.c" *********************************/
1230 /*
1231 ** Code for printing error message.
1232 */
1233 
1234 /* Find a good place to break "msg" so that its length is at least "min"
1235 ** but no more than "max".  Make the point as close to max as possible.
1236 */
findbreak(msg,min,max)1237 static int findbreak(msg,min,max)
1238 char *msg;
1239 int min;
1240 int max;
1241 {
1242   int i,spot;
1243   char c;
1244   for(i=spot=min; i<=max; i++){
1245     c = msg[i];
1246     if( c=='\t' ) msg[i] = ' ';
1247     if( c=='\n' ){ msg[i] = ' '; spot = i; break; }
1248     if( c==0 ){ spot = i; break; }
1249     if( c=='-' && i<max-1 ) spot = i+1;
1250     if( c==' ' ) spot = i;
1251   }
1252   return spot;
1253 }
1254 
1255 /*
1256 ** The error message is split across multiple lines if necessary.  The
1257 ** splits occur at a space, if there is a space available near the end
1258 ** of the line.
1259 */
1260 #define ERRMSGSIZE  10000 /* Hope this is big enough.  No way to error check */
1261 #define LINEWIDTH      79 /* Max width of any output line */
1262 #define PREFIXLIMIT    30 /* Max width of the prefix on each line */
ErrorMsg(const char * filename,int lineno,const char * format,...)1263 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1264   char errmsg[ERRMSGSIZE];
1265   char prefix[PREFIXLIMIT+10];
1266   int errmsgsize;
1267   int prefixsize;
1268   int availablewidth;
1269   va_list ap;
1270   int end, restart, base;
1271 
1272   va_start(ap, format);
1273   /* Prepare a prefix to be prepended to every output line */
1274   if( lineno>0 ){
1275     sprintf(prefix,"%.*s:%d: ",PREFIXLIMIT-10,filename,lineno);
1276   }else{
1277     sprintf(prefix,"%.*s: ",PREFIXLIMIT-10,filename);
1278   }
1279   prefixsize = strlen(prefix);
1280   availablewidth = LINEWIDTH - prefixsize;
1281 
1282   /* Generate the error message */
1283   vsprintf(errmsg,format,ap);
1284   va_end(ap);
1285   errmsgsize = strlen(errmsg);
1286   /* Remove trailing '\n's from the error message. */
1287   while( errmsgsize>0 && errmsg[errmsgsize-1]=='\n' ){
1288      errmsg[--errmsgsize] = 0;
1289   }
1290 
1291   /* Print the error message */
1292   base = 0;
1293   while( errmsg[base]!=0 ){
1294     end = restart = findbreak(&errmsg[base],0,availablewidth);
1295     restart += base;
1296     while( errmsg[restart]==' ' ) restart++;
1297     fprintf(stdout,"%s%.*s\n",prefix,end,&errmsg[base]);
1298     base = restart;
1299   }
1300 }
1301 /**************** From the file "main.c" ************************************/
1302 /*
1303 ** Main program file for the LEMON parser generator.
1304 */
1305 
1306 /* Report an out-of-memory condition and abort.  This function
1307 ** is used mostly by the "MemoryCheck" macro in struct.h
1308 */
memory_error()1309 void memory_error(){
1310   fprintf(stderr,"Out of memory.  Aborting...\n");
1311   exit(1);
1312 }
1313 
1314 
1315 /* The main program.  Parse the command line and do it... */
main(argc,argv)1316 int main(argc,argv)
1317 int argc;
1318 char **argv;
1319 {
1320   static int version = 0;
1321   static int rpflag = 0;
1322   static int basisflag = 0;
1323   static int compress = 0;
1324   static int quiet = 0;
1325   static int statistics = 0;
1326   static int mhflag = 0;
1327   static struct s_options options[] = {
1328     {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1329     {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1330     {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1331     {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file"},
1332     {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1333     {OPT_FLAG, "s", (char*)&statistics, "Print parser stats to standard output."},
1334     {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1335     {OPT_FLAG,0,0,0}
1336   };
1337   int i;
1338   struct lemon lem;
1339 
1340   OptInit(argv,options,stderr);
1341   if( version ){
1342      printf("Lemon version 1.0\n");
1343      exit(0);
1344   }
1345   if( OptNArgs()!=1 ){
1346     fprintf(stderr,"Exactly one filename argument is required.\n");
1347     exit(1);
1348   }
1349   lem.errorcnt = 0;
1350 
1351   /* Initialize the machine */
1352   Strsafe_init();
1353   Symbol_init();
1354   State_init();
1355   lem.argv0 = argv[0];
1356   lem.filename = OptArg(0);
1357   lem.basisflag = basisflag;
1358   lem.has_fallback = 0;
1359   lem.nconflict = 0;
1360   lem.name = lem.include = lem.arg = lem.tokentype = lem.start = 0;
1361   lem.vartype = 0;
1362   lem.stacksize = 0;
1363   lem.error = lem.overflow = lem.failure = lem.accept = lem.tokendest =
1364      lem.tokenprefix = lem.outname = lem.extracode = 0;
1365   lem.vardest = 0;
1366   lem.tablesize = 0;
1367   Symbol_new("$");
1368   lem.errsym = Symbol_new("error");
1369 
1370   /* Parse the input file */
1371   Parse(&lem);
1372   if( lem.errorcnt ) exit(lem.errorcnt);
1373   if( lem.rule==0 ){
1374     fprintf(stderr,"Empty grammar.\n");
1375     exit(1);
1376   }
1377 
1378   /* Count and index the symbols of the grammar */
1379   lem.nsymbol = Symbol_count();
1380   Symbol_new("{default}");
1381   lem.symbols = Symbol_arrayof();
1382   for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
1383   qsort(lem.symbols,lem.nsymbol+1,sizeof(struct symbol*),
1384         (int(*)())Symbolcmpp);
1385   for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
1386   for(i=1; isupper(lem.symbols[i]->name[0]); i++);
1387   lem.nterminal = i;
1388 
1389   /* Generate a reprint of the grammar, if requested on the command line */
1390   if( rpflag ){
1391     Reprint(&lem);
1392   }else{
1393     /* Initialize the size for all follow and first sets */
1394     SetSize(lem.nterminal);
1395 
1396     /* Find the precedence for every production rule (that has one) */
1397     FindRulePrecedences(&lem);
1398 
1399     /* Compute the lambda-nonterminals and the first-sets for every
1400     ** nonterminal */
1401     FindFirstSets(&lem);
1402 
1403     /* Compute all LR(0) states.  Also record follow-set propagation
1404     ** links so that the follow-set can be computed later */
1405     lem.nstate = 0;
1406     FindStates(&lem);
1407     lem.sorted = State_arrayof();
1408 
1409     /* Tie up loose ends on the propagation links */
1410     FindLinks(&lem);
1411 
1412     /* Compute the follow set of every reducible configuration */
1413     FindFollowSets(&lem);
1414 
1415     /* Compute the action tables */
1416     FindActions(&lem);
1417 
1418     /* Compress the action tables */
1419     if( compress==0 ) CompressTables(&lem);
1420 
1421     /* Generate a report of the parser generated.  (the "y.output" file) */
1422     if( !quiet ) ReportOutput(&lem);
1423 
1424     /* Generate the source code for the parser */
1425     ReportTable(&lem, mhflag);
1426 
1427     /* Produce a header file for use by the scanner.  (This step is
1428     ** omitted if the "-m" option is used because makeheaders will
1429     ** generate the file for us.) */
1430     if( !mhflag ) ReportHeader(&lem);
1431   }
1432   if( statistics ){
1433     printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
1434       lem.nterminal, lem.nsymbol - lem.nterminal, lem.nrule);
1435     printf("                   %d states, %d parser table entries, %d conflicts\n",
1436       lem.nstate, lem.tablesize, lem.nconflict);
1437   }
1438   if( lem.nconflict ){
1439     fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1440   }
1441   exit(lem.errorcnt + lem.nconflict);
1442   return (lem.errorcnt + lem.nconflict);
1443 }
1444 /******************** From the file "msort.c" *******************************/
1445 /*
1446 ** A generic merge-sort program.
1447 **
1448 ** USAGE:
1449 ** Let "ptr" be a pointer to some structure which is at the head of
1450 ** a null-terminated list.  Then to sort the list call:
1451 **
1452 **     ptr = msort(ptr,&(ptr->next),cmpfnc);
1453 **
1454 ** In the above, "cmpfnc" is a pointer to a function which compares
1455 ** two instances of the structure and returns an integer, as in
1456 ** strcmp.  The second argument is a pointer to the pointer to the
1457 ** second element of the linked list.  This address is used to compute
1458 ** the offset to the "next" field within the structure.  The offset to
1459 ** the "next" field must be constant for all structures in the list.
1460 **
1461 ** The function returns a new pointer which is the head of the list
1462 ** after sorting.
1463 **
1464 ** ALGORITHM:
1465 ** Merge-sort.
1466 */
1467 
1468 /*
1469 ** Return a pointer to the next structure in the linked list.
1470 */
1471 #define NEXT(A) (*(char**)(((unsigned long)A)+offset))
1472 
1473 /*
1474 ** Inputs:
1475 **   a:       A sorted, null-terminated linked list.  (May be null).
1476 **   b:       A sorted, null-terminated linked list.  (May be null).
1477 **   cmp:     A pointer to the comparison function.
1478 **   offset:  Offset in the structure to the "next" field.
1479 **
1480 ** Return Value:
1481 **   A pointer to the head of a sorted list containing the elements
1482 **   of both a and b.
1483 **
1484 ** Side effects:
1485 **   The "next" pointers for elements in the lists a and b are
1486 **   changed.
1487 */
merge(a,b,cmp,offset)1488 static char *merge(a,b,cmp,offset)
1489 char *a;
1490 char *b;
1491 int (*cmp)();
1492 int offset;
1493 {
1494   char *ptr, *head;
1495 
1496   if( a==0 ){
1497     head = b;
1498   }else if( b==0 ){
1499     head = a;
1500   }else{
1501     if( (*cmp)(a,b)<0 ){
1502       ptr = a;
1503       a = NEXT(a);
1504     }else{
1505       ptr = b;
1506       b = NEXT(b);
1507     }
1508     head = ptr;
1509     while( a && b ){
1510       if( (*cmp)(a,b)<0 ){
1511         NEXT(ptr) = a;
1512         ptr = a;
1513         a = NEXT(a);
1514       }else{
1515         NEXT(ptr) = b;
1516         ptr = b;
1517         b = NEXT(b);
1518       }
1519     }
1520     if( a ) NEXT(ptr) = a;
1521     else    NEXT(ptr) = b;
1522   }
1523   return head;
1524 }
1525 
1526 /*
1527 ** Inputs:
1528 **   list:      Pointer to a singly-linked list of structures.
1529 **   next:      Pointer to pointer to the second element of the list.
1530 **   cmp:       A comparison function.
1531 **
1532 ** Return Value:
1533 **   A pointer to the head of a sorted list containing the elements
1534 **   orginally in list.
1535 **
1536 ** Side effects:
1537 **   The "next" pointers for elements in list are changed.
1538 */
1539 #define LISTSIZE 30
msort(list,next,cmp)1540 char *msort(list,next,cmp)
1541 char *list;
1542 char **next;
1543 int (*cmp)();
1544 {
1545   unsigned long offset;
1546   char *ep;
1547   char *set[LISTSIZE];
1548   int i;
1549   offset = (unsigned long)next - (unsigned long)list;
1550   for(i=0; i<LISTSIZE; i++) set[i] = 0;
1551   while( list ){
1552     ep = list;
1553     list = NEXT(list);
1554     NEXT(ep) = 0;
1555     for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1556       ep = merge(ep,set[i],cmp,offset);
1557       set[i] = 0;
1558     }
1559     set[i] = ep;
1560   }
1561   ep = 0;
1562   for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(ep,set[i],cmp,offset);
1563   return ep;
1564 }
1565 /************************ From the file "option.c" **************************/
1566 static char **argv;
1567 static struct s_options *op;
1568 static FILE *errstream;
1569 
1570 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1571 
1572 /*
1573 ** Print the command line with a carrot pointing to the k-th character
1574 ** of the n-th field.
1575 */
errline(n,k,err)1576 static void errline(n,k,err)
1577 int n;
1578 int k;
1579 FILE *err;
1580 {
1581   int spcnt, i;
1582   spcnt = 0;
1583   if( argv[0] ) fprintf(err,"%s",argv[0]);
1584   spcnt = strlen(argv[0]) + 1;
1585   for(i=1; i<n && argv[i]; i++){
1586     fprintf(err," %s",argv[i]);
1587     spcnt += strlen(argv[i]+1);
1588   }
1589   spcnt += k;
1590   for(; argv[i]; i++) fprintf(err," %s",argv[i]);
1591   if( spcnt<20 ){
1592     fprintf(err,"\n%*s^-- here\n",spcnt,"");
1593   }else{
1594     fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1595   }
1596 }
1597 
1598 /*
1599 ** Return the index of the N-th non-switch argument.  Return -1
1600 ** if N is out of range.
1601 */
argindex(n)1602 static int argindex(n)
1603 int n;
1604 {
1605   int i;
1606   int dashdash = 0;
1607   if( argv!=0 && *argv!=0 ){
1608     for(i=1; argv[i]; i++){
1609       if( dashdash || !ISOPT(argv[i]) ){
1610         if( n==0 ) return i;
1611         n--;
1612       }
1613       if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1614     }
1615   }
1616   return -1;
1617 }
1618 
1619 static char emsg[] = "Command line syntax error: ";
1620 
1621 /*
1622 ** Process a flag command line argument.
1623 */
handleflags(i,err)1624 static int handleflags(i,err)
1625 int i;
1626 FILE *err;
1627 {
1628   int v;
1629   int errcnt = 0;
1630   int j;
1631   for(j=0; op[j].label; j++){
1632     if( strcmp(&argv[i][1],op[j].label)==0 ) break;
1633   }
1634   v = argv[i][0]=='-' ? 1 : 0;
1635   if( op[j].label==0 ){
1636     if( err ){
1637       fprintf(err,"%sundefined option.\n",emsg);
1638       errline(i,1,err);
1639     }
1640     errcnt++;
1641   }else if( op[j].type==OPT_FLAG ){
1642     *((int*)op[j].arg) = v;
1643   }else if( op[j].type==OPT_FFLAG ){
1644     (*(void(*)())(op[j].arg))(v);
1645   }else{
1646     if( err ){
1647       fprintf(err,"%smissing argument on switch.\n",emsg);
1648       errline(i,1,err);
1649     }
1650     errcnt++;
1651   }
1652   return errcnt;
1653 }
1654 
1655 /*
1656 ** Process a command line switch which has an argument.
1657 */
handleswitch(i,err)1658 static int handleswitch(i,err)
1659 int i;
1660 FILE *err;
1661 {
1662   int lv = 0;
1663   double dv = 0.0;
1664   char *sv = 0, *end;
1665   char *cp;
1666   int j;
1667   int errcnt = 0;
1668   cp = strchr(argv[i],'=');
1669   *cp = 0;
1670   for(j=0; op[j].label; j++){
1671     if( strcmp(argv[i],op[j].label)==0 ) break;
1672   }
1673   *cp = '=';
1674   if( op[j].label==0 ){
1675     if( err ){
1676       fprintf(err,"%sundefined option.\n",emsg);
1677       errline(i,0,err);
1678     }
1679     errcnt++;
1680   }else{
1681     cp++;
1682     switch( op[j].type ){
1683       case OPT_FLAG:
1684       case OPT_FFLAG:
1685         if( err ){
1686           fprintf(err,"%soption requires an argument.\n",emsg);
1687           errline(i,0,err);
1688         }
1689         errcnt++;
1690         break;
1691       case OPT_DBL:
1692       case OPT_FDBL:
1693         dv = strtod(cp,&end);
1694         if( *end ){
1695           if( err ){
1696             fprintf(err,"%sillegal character in floating-point argument.\n",emsg);
1697             errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1698           }
1699           errcnt++;
1700         }
1701         break;
1702       case OPT_INT:
1703       case OPT_FINT:
1704         lv = strtol(cp,&end,0);
1705         if( *end ){
1706           if( err ){
1707             fprintf(err,"%sillegal character in integer argument.\n",emsg);
1708             errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1709           }
1710           errcnt++;
1711         }
1712         break;
1713       case OPT_STR:
1714       case OPT_FSTR:
1715         sv = cp;
1716         break;
1717     }
1718     switch( op[j].type ){
1719       case OPT_FLAG:
1720       case OPT_FFLAG:
1721         break;
1722       case OPT_DBL:
1723         *(double*)(op[j].arg) = dv;
1724         break;
1725       case OPT_FDBL:
1726         (*(void(*)())(op[j].arg))(dv);
1727         break;
1728       case OPT_INT:
1729         *(int*)(op[j].arg) = lv;
1730         break;
1731       case OPT_FINT:
1732         (*(void(*)())(op[j].arg))((int)lv);
1733         break;
1734       case OPT_STR:
1735         *(char**)(op[j].arg) = sv;
1736         break;
1737       case OPT_FSTR:
1738         (*(void(*)())(op[j].arg))(sv);
1739         break;
1740     }
1741   }
1742   return errcnt;
1743 }
1744 
OptInit(a,o,err)1745 int OptInit(a,o,err)
1746 char **a;
1747 struct s_options *o;
1748 FILE *err;
1749 {
1750   int errcnt = 0;
1751   argv = a;
1752   op = o;
1753   errstream = err;
1754   if( argv && *argv && op ){
1755     int i;
1756     for(i=1; argv[i]; i++){
1757       if( argv[i][0]=='+' || argv[i][0]=='-' ){
1758         errcnt += handleflags(i,err);
1759       }else if( strchr(argv[i],'=') ){
1760         errcnt += handleswitch(i,err);
1761       }
1762     }
1763   }
1764   if( errcnt>0 ){
1765     fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
1766     OptPrint();
1767     exit(1);
1768   }
1769   return 0;
1770 }
1771 
OptNArgs()1772 int OptNArgs(){
1773   int cnt = 0;
1774   int dashdash = 0;
1775   int i;
1776   if( argv!=0 && argv[0]!=0 ){
1777     for(i=1; argv[i]; i++){
1778       if( dashdash || !ISOPT(argv[i]) ) cnt++;
1779       if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1780     }
1781   }
1782   return cnt;
1783 }
1784 
OptArg(n)1785 char *OptArg(n)
1786 int n;
1787 {
1788   int i;
1789   i = argindex(n);
1790   return i>=0 ? argv[i] : 0;
1791 }
1792 
OptErr(n)1793 void OptErr(n)
1794 int n;
1795 {
1796   int i;
1797   i = argindex(n);
1798   if( i>=0 ) errline(i,0,errstream);
1799 }
1800 
OptPrint()1801 void OptPrint(){
1802   int i;
1803   int max, len;
1804   max = 0;
1805   for(i=0; op[i].label; i++){
1806     len = strlen(op[i].label) + 1;
1807     switch( op[i].type ){
1808       case OPT_FLAG:
1809       case OPT_FFLAG:
1810         break;
1811       case OPT_INT:
1812       case OPT_FINT:
1813         len += 9;       /* length of "<integer>" */
1814         break;
1815       case OPT_DBL:
1816       case OPT_FDBL:
1817         len += 6;       /* length of "<real>" */
1818         break;
1819       case OPT_STR:
1820       case OPT_FSTR:
1821         len += 8;       /* length of "<string>" */
1822         break;
1823     }
1824     if( len>max ) max = len;
1825   }
1826   for(i=0; op[i].label; i++){
1827     switch( op[i].type ){
1828       case OPT_FLAG:
1829       case OPT_FFLAG:
1830         fprintf(errstream,"  -%-*s  %s\n",max,op[i].label,op[i].message);
1831         break;
1832       case OPT_INT:
1833       case OPT_FINT:
1834         fprintf(errstream,"  %s=<integer>%*s  %s\n",op[i].label,
1835           (int)(max-strlen(op[i].label)-9),"",op[i].message);
1836         break;
1837       case OPT_DBL:
1838       case OPT_FDBL:
1839         fprintf(errstream,"  %s=<real>%*s  %s\n",op[i].label,
1840           (int)(max-strlen(op[i].label)-6),"",op[i].message);
1841         break;
1842       case OPT_STR:
1843       case OPT_FSTR:
1844         fprintf(errstream,"  %s=<string>%*s  %s\n",op[i].label,
1845           (int)(max-strlen(op[i].label)-8),"",op[i].message);
1846         break;
1847     }
1848   }
1849 }
1850 /*********************** From the file "parse.c" ****************************/
1851 /*
1852 ** Input file parser for the LEMON parser generator.
1853 */
1854 
1855 /* The state of the parser */
1856 struct pstate {
1857   char *filename;       /* Name of the input file */
1858   int tokenlineno;      /* Linenumber at which current token starts */
1859   int errorcnt;         /* Number of errors so far */
1860   char *tokenstart;     /* Text of current token */
1861   struct lemon *gp;     /* Global state vector */
1862   enum e_state {
1863     INITIALIZE,
1864     WAITING_FOR_DECL_OR_RULE,
1865     WAITING_FOR_DECL_KEYWORD,
1866     WAITING_FOR_DECL_ARG,
1867     WAITING_FOR_PRECEDENCE_SYMBOL,
1868     WAITING_FOR_ARROW,
1869     IN_RHS,
1870     LHS_ALIAS_1,
1871     LHS_ALIAS_2,
1872     LHS_ALIAS_3,
1873     RHS_ALIAS_1,
1874     RHS_ALIAS_2,
1875     PRECEDENCE_MARK_1,
1876     PRECEDENCE_MARK_2,
1877     RESYNC_AFTER_RULE_ERROR,
1878     RESYNC_AFTER_DECL_ERROR,
1879     WAITING_FOR_DESTRUCTOR_SYMBOL,
1880     WAITING_FOR_DATATYPE_SYMBOL,
1881     WAITING_FOR_FALLBACK_ID
1882   } state;                   /* The state of the parser */
1883   struct symbol *fallback;   /* The fallback token */
1884   struct symbol *lhs;        /* Left-hand side of current rule */
1885   char *lhsalias;            /* Alias for the LHS */
1886   int nrhs;                  /* Number of right-hand side symbols seen */
1887   struct symbol *rhs[MAXRHS];  /* RHS symbols */
1888   char *alias[MAXRHS];       /* Aliases for each RHS symbol (or NULL) */
1889   struct rule *prevrule;     /* Previous rule parsed */
1890   char *declkeyword;         /* Keyword of a declaration */
1891   char **declargslot;        /* Where the declaration argument should be put */
1892   int *decllnslot;           /* Where the declaration linenumber is put */
1893   enum e_assoc declassoc;    /* Assign this association to decl arguments */
1894   int preccounter;           /* Assign this precedence to decl arguments */
1895   struct rule *firstrule;    /* Pointer to first rule in the grammar */
1896   struct rule *lastrule;     /* Pointer to the most recently parsed rule */
1897 };
1898 
1899 /* Parse a single token */
parseonetoken(psp)1900 static void parseonetoken(psp)
1901 struct pstate *psp;
1902 {
1903   char *x;
1904   x = Strsafe(psp->tokenstart);     /* Save the token permanently */
1905 #if 0
1906   printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
1907     x,psp->state);
1908 #endif
1909   switch( psp->state ){
1910     case INITIALIZE:
1911       psp->prevrule = 0;
1912       psp->preccounter = 0;
1913       psp->firstrule = psp->lastrule = 0;
1914       psp->gp->nrule = 0;
1915       /* FALLTHROUGH */
1916     case WAITING_FOR_DECL_OR_RULE:
1917       if( x[0]=='%' ){
1918         psp->state = WAITING_FOR_DECL_KEYWORD;
1919       }else if( islower(x[0]) ){
1920         psp->lhs = Symbol_new(x);
1921         psp->nrhs = 0;
1922         psp->lhsalias = 0;
1923         psp->state = WAITING_FOR_ARROW;
1924       }else if( x[0]=='{' ){
1925         if( psp->prevrule==0 ){
1926           ErrorMsg(psp->filename,psp->tokenlineno,
1927 "There is not prior rule opon which to attach the code \
1928 fragment which begins on this line.");
1929           psp->errorcnt++;
1930 	}else if( psp->prevrule->code!=0 ){
1931           ErrorMsg(psp->filename,psp->tokenlineno,
1932 "Code fragment beginning on this line is not the first \
1933 to follow the previous rule.");
1934           psp->errorcnt++;
1935         }else{
1936           psp->prevrule->line = psp->tokenlineno;
1937           psp->prevrule->code = &x[1];
1938 	}
1939       }else if( x[0]=='[' ){
1940         psp->state = PRECEDENCE_MARK_1;
1941       }else{
1942         ErrorMsg(psp->filename,psp->tokenlineno,
1943           "Token \"%s\" should be either \"%%\" or a nonterminal name.",
1944           x);
1945         psp->errorcnt++;
1946       }
1947       break;
1948     case PRECEDENCE_MARK_1:
1949       if( !isupper(x[0]) ){
1950         ErrorMsg(psp->filename,psp->tokenlineno,
1951           "The precedence symbol must be a terminal.");
1952         psp->errorcnt++;
1953       }else if( psp->prevrule==0 ){
1954         ErrorMsg(psp->filename,psp->tokenlineno,
1955           "There is no prior rule to assign precedence \"[%s]\".",x);
1956         psp->errorcnt++;
1957       }else if( psp->prevrule->precsym!=0 ){
1958         ErrorMsg(psp->filename,psp->tokenlineno,
1959 "Precedence mark on this line is not the first \
1960 to follow the previous rule.");
1961         psp->errorcnt++;
1962       }else{
1963         psp->prevrule->precsym = Symbol_new(x);
1964       }
1965       psp->state = PRECEDENCE_MARK_2;
1966       break;
1967     case PRECEDENCE_MARK_2:
1968       if( x[0]!=']' ){
1969         ErrorMsg(psp->filename,psp->tokenlineno,
1970           "Missing \"]\" on precedence mark.");
1971         psp->errorcnt++;
1972       }
1973       psp->state = WAITING_FOR_DECL_OR_RULE;
1974       break;
1975     case WAITING_FOR_ARROW:
1976       if( x[0]==':' && x[1]==':' && x[2]=='=' ){
1977         psp->state = IN_RHS;
1978       }else if( x[0]=='(' ){
1979         psp->state = LHS_ALIAS_1;
1980       }else{
1981         ErrorMsg(psp->filename,psp->tokenlineno,
1982           "Expected to see a \":\" following the LHS symbol \"%s\".",
1983           psp->lhs->name);
1984         psp->errorcnt++;
1985         psp->state = RESYNC_AFTER_RULE_ERROR;
1986       }
1987       break;
1988     case LHS_ALIAS_1:
1989       if( isalpha(x[0]) ){
1990         psp->lhsalias = x;
1991         psp->state = LHS_ALIAS_2;
1992       }else{
1993         ErrorMsg(psp->filename,psp->tokenlineno,
1994           "\"%s\" is not a valid alias for the LHS \"%s\"\n",
1995           x,psp->lhs->name);
1996         psp->errorcnt++;
1997         psp->state = RESYNC_AFTER_RULE_ERROR;
1998       }
1999       break;
2000     case LHS_ALIAS_2:
2001       if( x[0]==')' ){
2002         psp->state = LHS_ALIAS_3;
2003       }else{
2004         ErrorMsg(psp->filename,psp->tokenlineno,
2005           "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2006         psp->errorcnt++;
2007         psp->state = RESYNC_AFTER_RULE_ERROR;
2008       }
2009       break;
2010     case LHS_ALIAS_3:
2011       if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2012         psp->state = IN_RHS;
2013       }else{
2014         ErrorMsg(psp->filename,psp->tokenlineno,
2015           "Missing \"->\" following: \"%s(%s)\".",
2016            psp->lhs->name,psp->lhsalias);
2017         psp->errorcnt++;
2018         psp->state = RESYNC_AFTER_RULE_ERROR;
2019       }
2020       break;
2021     case IN_RHS:
2022       if( x[0]=='.' ){
2023         struct rule *rp;
2024         rp = (struct rule *)malloc( sizeof(struct rule) +
2025              sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs );
2026         if( rp==0 ){
2027           ErrorMsg(psp->filename,psp->tokenlineno,
2028             "Can't allocate enough memory for this rule.");
2029           psp->errorcnt++;
2030           psp->prevrule = 0;
2031 	}else{
2032           int i;
2033           rp->ruleline = psp->tokenlineno;
2034           rp->rhs = (struct symbol**)&rp[1];
2035           rp->rhsalias = (char**)&(rp->rhs[psp->nrhs]);
2036           for(i=0; i<psp->nrhs; i++){
2037             rp->rhs[i] = psp->rhs[i];
2038             rp->rhsalias[i] = psp->alias[i];
2039 	  }
2040           rp->lhs = psp->lhs;
2041           rp->lhsalias = psp->lhsalias;
2042           rp->nrhs = psp->nrhs;
2043           rp->code = 0;
2044           rp->precsym = 0;
2045           rp->index = psp->gp->nrule++;
2046           rp->nextlhs = rp->lhs->rule;
2047           rp->lhs->rule = rp;
2048           rp->next = 0;
2049           if( psp->firstrule==0 ){
2050             psp->firstrule = psp->lastrule = rp;
2051 	  }else{
2052             psp->lastrule->next = rp;
2053             psp->lastrule = rp;
2054 	  }
2055           psp->prevrule = rp;
2056 	}
2057         psp->state = WAITING_FOR_DECL_OR_RULE;
2058       }else if( isalpha(x[0]) ){
2059         if( psp->nrhs>=MAXRHS ){
2060           ErrorMsg(psp->filename,psp->tokenlineno,
2061             "Too many symbol on RHS or rule beginning at \"%s\".",
2062             x);
2063           psp->errorcnt++;
2064           psp->state = RESYNC_AFTER_RULE_ERROR;
2065 	}else{
2066           psp->rhs[psp->nrhs] = Symbol_new(x);
2067           psp->alias[psp->nrhs] = 0;
2068           psp->nrhs++;
2069 	}
2070       }else if( x[0]=='(' && psp->nrhs>0 ){
2071         psp->state = RHS_ALIAS_1;
2072       }else{
2073         ErrorMsg(psp->filename,psp->tokenlineno,
2074           "Illegal character on RHS of rule: \"%s\".",x);
2075         psp->errorcnt++;
2076         psp->state = RESYNC_AFTER_RULE_ERROR;
2077       }
2078       break;
2079     case RHS_ALIAS_1:
2080       if( isalpha(x[0]) ){
2081         psp->alias[psp->nrhs-1] = x;
2082         psp->state = RHS_ALIAS_2;
2083       }else{
2084         ErrorMsg(psp->filename,psp->tokenlineno,
2085           "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2086           x,psp->rhs[psp->nrhs-1]->name);
2087         psp->errorcnt++;
2088         psp->state = RESYNC_AFTER_RULE_ERROR;
2089       }
2090       break;
2091     case RHS_ALIAS_2:
2092       if( x[0]==')' ){
2093         psp->state = IN_RHS;
2094       }else{
2095         ErrorMsg(psp->filename,psp->tokenlineno,
2096           "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2097         psp->errorcnt++;
2098         psp->state = RESYNC_AFTER_RULE_ERROR;
2099       }
2100       break;
2101     case WAITING_FOR_DECL_KEYWORD:
2102       if( isalpha(x[0]) ){
2103         psp->declkeyword = x;
2104         psp->declargslot = 0;
2105         psp->decllnslot = 0;
2106         psp->state = WAITING_FOR_DECL_ARG;
2107         if( strcmp(x,"name")==0 ){
2108           psp->declargslot = &(psp->gp->name);
2109 	}else if( strcmp(x,"include")==0 ){
2110           psp->declargslot = &(psp->gp->include);
2111           psp->decllnslot = &psp->gp->includeln;
2112 	}else if( strcmp(x,"code")==0 ){
2113           psp->declargslot = &(psp->gp->extracode);
2114           psp->decllnslot = &psp->gp->extracodeln;
2115 	}else if( strcmp(x,"token_destructor")==0 ){
2116           psp->declargslot = &psp->gp->tokendest;
2117           psp->decllnslot = &psp->gp->tokendestln;
2118 	}else if( strcmp(x,"default_destructor")==0 ){
2119           psp->declargslot = &psp->gp->vardest;
2120           psp->decllnslot = &psp->gp->vardestln;
2121 	}else if( strcmp(x,"token_prefix")==0 ){
2122           psp->declargslot = &psp->gp->tokenprefix;
2123 	}else if( strcmp(x,"syntax_error")==0 ){
2124           psp->declargslot = &(psp->gp->error);
2125           psp->decllnslot = &psp->gp->errorln;
2126 	}else if( strcmp(x,"parse_accept")==0 ){
2127           psp->declargslot = &(psp->gp->accept);
2128           psp->decllnslot = &psp->gp->acceptln;
2129 	}else if( strcmp(x,"parse_failure")==0 ){
2130           psp->declargslot = &(psp->gp->failure);
2131           psp->decllnslot = &psp->gp->failureln;
2132 	}else if( strcmp(x,"stack_overflow")==0 ){
2133           psp->declargslot = &(psp->gp->overflow);
2134           psp->decllnslot = &psp->gp->overflowln;
2135         }else if( strcmp(x,"extra_argument")==0 ){
2136           psp->declargslot = &(psp->gp->arg);
2137         }else if( strcmp(x,"token_type")==0 ){
2138           psp->declargslot = &(psp->gp->tokentype);
2139         }else if( strcmp(x,"default_type")==0 ){
2140           psp->declargslot = &(psp->gp->vartype);
2141         }else if( strcmp(x,"stack_size")==0 ){
2142           psp->declargslot = &(psp->gp->stacksize);
2143         }else if( strcmp(x,"start_symbol")==0 ){
2144           psp->declargslot = &(psp->gp->start);
2145         }else if( strcmp(x,"left")==0 ){
2146           psp->preccounter++;
2147           psp->declassoc = LEFT;
2148           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2149         }else if( strcmp(x,"right")==0 ){
2150           psp->preccounter++;
2151           psp->declassoc = RIGHT;
2152           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2153         }else if( strcmp(x,"nonassoc")==0 ){
2154           psp->preccounter++;
2155           psp->declassoc = NONE;
2156           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2157 	}else if( strcmp(x,"destructor")==0 ){
2158           psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2159 	}else if( strcmp(x,"type")==0 ){
2160           psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2161         }else if( strcmp(x,"fallback")==0 ){
2162           psp->fallback = 0;
2163           psp->state = WAITING_FOR_FALLBACK_ID;
2164         }else{
2165           ErrorMsg(psp->filename,psp->tokenlineno,
2166             "Unknown declaration keyword: \"%%%s\".",x);
2167           psp->errorcnt++;
2168           psp->state = RESYNC_AFTER_DECL_ERROR;
2169 	}
2170       }else{
2171         ErrorMsg(psp->filename,psp->tokenlineno,
2172           "Illegal declaration keyword: \"%s\".",x);
2173         psp->errorcnt++;
2174         psp->state = RESYNC_AFTER_DECL_ERROR;
2175       }
2176       break;
2177     case WAITING_FOR_DESTRUCTOR_SYMBOL:
2178       if( !isalpha(x[0]) ){
2179         ErrorMsg(psp->filename,psp->tokenlineno,
2180           "Symbol name missing after %destructor keyword");
2181         psp->errorcnt++;
2182         psp->state = RESYNC_AFTER_DECL_ERROR;
2183       }else{
2184         struct symbol *sp = Symbol_new(x);
2185         psp->declargslot = &sp->destructor;
2186         psp->decllnslot = &sp->destructorln;
2187         psp->state = WAITING_FOR_DECL_ARG;
2188       }
2189       break;
2190     case WAITING_FOR_DATATYPE_SYMBOL:
2191       if( !isalpha(x[0]) ){
2192         ErrorMsg(psp->filename,psp->tokenlineno,
2193           "Symbol name missing after %destructor keyword");
2194         psp->errorcnt++;
2195         psp->state = RESYNC_AFTER_DECL_ERROR;
2196       }else{
2197         struct symbol *sp = Symbol_new(x);
2198         psp->declargslot = &sp->datatype;
2199         psp->decllnslot = 0;
2200         psp->state = WAITING_FOR_DECL_ARG;
2201       }
2202       break;
2203     case WAITING_FOR_PRECEDENCE_SYMBOL:
2204       if( x[0]=='.' ){
2205         psp->state = WAITING_FOR_DECL_OR_RULE;
2206       }else if( isupper(x[0]) ){
2207         struct symbol *sp;
2208         sp = Symbol_new(x);
2209         if( sp->prec>=0 ){
2210           ErrorMsg(psp->filename,psp->tokenlineno,
2211             "Symbol \"%s\" has already be given a precedence.",x);
2212           psp->errorcnt++;
2213 	}else{
2214           sp->prec = psp->preccounter;
2215           sp->assoc = psp->declassoc;
2216 	}
2217       }else{
2218         ErrorMsg(psp->filename,psp->tokenlineno,
2219           "Can't assign a precedence to \"%s\".",x);
2220         psp->errorcnt++;
2221       }
2222       break;
2223     case WAITING_FOR_DECL_ARG:
2224       if( (x[0]=='{' || x[0]=='\"' || isalnum(x[0])) ){
2225         if( *(psp->declargslot)!=0 ){
2226           ErrorMsg(psp->filename,psp->tokenlineno,
2227             "The argument \"%s\" to declaration \"%%%s\" is not the first.",
2228             x[0]=='\"' ? &x[1] : x,psp->declkeyword);
2229           psp->errorcnt++;
2230           psp->state = RESYNC_AFTER_DECL_ERROR;
2231 	}else{
2232           *(psp->declargslot) = (x[0]=='\"' || x[0]=='{') ? &x[1] : x;
2233           if( psp->decllnslot ) *psp->decllnslot = psp->tokenlineno;
2234           psp->state = WAITING_FOR_DECL_OR_RULE;
2235 	}
2236       }else{
2237         ErrorMsg(psp->filename,psp->tokenlineno,
2238           "Illegal argument to %%%s: %s",psp->declkeyword,x);
2239         psp->errorcnt++;
2240         psp->state = RESYNC_AFTER_DECL_ERROR;
2241       }
2242       break;
2243     case WAITING_FOR_FALLBACK_ID:
2244       if( x[0]=='.' ){
2245         psp->state = WAITING_FOR_DECL_OR_RULE;
2246       }else if( !isupper(x[0]) ){
2247         ErrorMsg(psp->filename, psp->tokenlineno,
2248           "%%fallback argument \"%s\" should be a token", x);
2249         psp->errorcnt++;
2250       }else{
2251         struct symbol *sp = Symbol_new(x);
2252         if( psp->fallback==0 ){
2253           psp->fallback = sp;
2254         }else if( sp->fallback ){
2255           ErrorMsg(psp->filename, psp->tokenlineno,
2256             "More than one fallback assigned to token %s", x);
2257           psp->errorcnt++;
2258         }else{
2259           sp->fallback = psp->fallback;
2260           psp->gp->has_fallback = 1;
2261         }
2262       }
2263       break;
2264     case RESYNC_AFTER_RULE_ERROR:
2265 /*      if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2266 **      break; */
2267     case RESYNC_AFTER_DECL_ERROR:
2268       if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2269       if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2270       break;
2271   }
2272 }
2273 
2274 /* In spite of its name, this function is really a scanner.  It read
2275 ** in the entire input file (all at once) then tokenizes it.  Each
2276 ** token is passed to the function "parseonetoken" which builds all
2277 ** the appropriate data structures in the global state vector "gp".
2278 */
Parse(gp)2279 void Parse(gp)
2280 struct lemon *gp;
2281 {
2282   struct pstate ps;
2283   FILE *fp;
2284   char *filebuf;
2285   int filesize;
2286   int lineno;
2287   int c;
2288   char *cp, *nextcp;
2289   int startline = 0;
2290 
2291   ps.gp = gp;
2292   ps.filename = gp->filename;
2293   ps.errorcnt = 0;
2294   ps.state = INITIALIZE;
2295 
2296   /* Begin by reading the input file */
2297   fp = fopen(ps.filename,"rb");
2298   if( fp==0 ){
2299     ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2300     gp->errorcnt++;
2301     return;
2302   }
2303   fseek(fp,0,2);
2304   filesize = ftell(fp);
2305   rewind(fp);
2306   filebuf = (char *)malloc( filesize+1 );
2307   if( filebuf==0 ){
2308     ErrorMsg(ps.filename,0,"Can't allocate %d of memory to hold this file.",
2309       filesize+1);
2310     gp->errorcnt++;
2311     return;
2312   }
2313   if( fread(filebuf,1,filesize,fp)!=filesize ){
2314     ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2315       filesize);
2316     free(filebuf);
2317     gp->errorcnt++;
2318     return;
2319   }
2320   fclose(fp);
2321   filebuf[filesize] = 0;
2322 
2323   /* Now scan the text of the input file */
2324   lineno = 1;
2325   for(cp=filebuf; (c= *cp)!=0; ){
2326     if( c=='\n' ) lineno++;              /* Keep track of the line number */
2327     if( isspace(c) ){ cp++; continue; }  /* Skip all white space */
2328     if( c=='/' && cp[1]=='/' ){          /* Skip C++ style comments */
2329       cp+=2;
2330       while( (c= *cp)!=0 && c!='\n' ) cp++;
2331       continue;
2332     }
2333     if( c=='/' && cp[1]=='*' ){          /* Skip C style comments */
2334       cp+=2;
2335       while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
2336         if( c=='\n' ) lineno++;
2337         cp++;
2338       }
2339       if( c ) cp++;
2340       continue;
2341     }
2342     ps.tokenstart = cp;                /* Mark the beginning of the token */
2343     ps.tokenlineno = lineno;           /* Linenumber on which token begins */
2344     if( c=='\"' ){                     /* String literals */
2345       cp++;
2346       while( (c= *cp)!=0 && c!='\"' ){
2347         if( c=='\n' ) lineno++;
2348         cp++;
2349       }
2350       if( c==0 ){
2351         ErrorMsg(ps.filename,startline,
2352 "String starting on this line is not terminated before the end of the file.");
2353         ps.errorcnt++;
2354         nextcp = cp;
2355       }else{
2356         nextcp = cp+1;
2357       }
2358     }else if( c=='{' ){               /* A block of C code */
2359       int level;
2360       cp++;
2361       for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
2362         if( c=='\n' ) lineno++;
2363         else if( c=='{' ) level++;
2364         else if( c=='}' ) level--;
2365         else if( c=='/' && cp[1]=='*' ){  /* Skip comments */
2366           int prevc;
2367           cp = &cp[2];
2368           prevc = 0;
2369           while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
2370             if( c=='\n' ) lineno++;
2371             prevc = c;
2372             cp++;
2373 	  }
2374 	}else if( c=='/' && cp[1]=='/' ){  /* Skip C++ style comments too */
2375           cp = &cp[2];
2376           while( (c= *cp)!=0 && c!='\n' ) cp++;
2377           if( c ) lineno++;
2378 	}else if( c=='\'' || c=='\"' ){    /* String a character literals */
2379           int startchar, prevc;
2380           startchar = c;
2381           prevc = 0;
2382           for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
2383             if( c=='\n' ) lineno++;
2384             if( prevc=='\\' ) prevc = 0;
2385             else              prevc = c;
2386 	  }
2387 	}
2388       }
2389       if( c==0 ){
2390         ErrorMsg(ps.filename,ps.tokenlineno,
2391 "C code starting on this line is not terminated before the end of the file.");
2392         ps.errorcnt++;
2393         nextcp = cp;
2394       }else{
2395         nextcp = cp+1;
2396       }
2397     }else if( isalnum(c) ){          /* Identifiers */
2398       while( (c= *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
2399       nextcp = cp;
2400     }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
2401       cp += 3;
2402       nextcp = cp;
2403     }else{                          /* All other (one character) operators */
2404       cp++;
2405       nextcp = cp;
2406     }
2407     c = *cp;
2408     *cp = 0;                        /* Null terminate the token */
2409     parseonetoken(&ps);             /* Parse the token */
2410     *cp = c;                        /* Restore the buffer */
2411     cp = nextcp;
2412   }
2413   free(filebuf);                    /* Release the buffer after parsing */
2414   gp->rule = ps.firstrule;
2415   gp->errorcnt = ps.errorcnt;
2416 }
2417 /*************************** From the file "plink.c" *********************/
2418 /*
2419 ** Routines processing configuration follow-set propagation links
2420 ** in the LEMON parser generator.
2421 */
2422 static struct plink *plink_freelist = 0;
2423 
2424 /* Allocate a new plink */
Plink_new()2425 struct plink *Plink_new(){
2426   struct plink *new;
2427 
2428   if( plink_freelist==0 ){
2429     int i;
2430     int amt = 100;
2431     plink_freelist = (struct plink *)malloc( sizeof(struct plink)*amt );
2432     if( plink_freelist==0 ){
2433       fprintf(stderr,
2434       "Unable to allocate memory for a new follow-set propagation link.\n");
2435       exit(1);
2436     }
2437     for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
2438     plink_freelist[amt-1].next = 0;
2439   }
2440   new = plink_freelist;
2441   plink_freelist = plink_freelist->next;
2442   return new;
2443 }
2444 
2445 /* Add a plink to a plink list */
Plink_add(plpp,cfp)2446 void Plink_add(plpp,cfp)
2447 struct plink **plpp;
2448 struct config *cfp;
2449 {
2450   struct plink *new;
2451   new = Plink_new();
2452   new->next = *plpp;
2453   *plpp = new;
2454   new->cfp = cfp;
2455 }
2456 
2457 /* Transfer every plink on the list "from" to the list "to" */
Plink_copy(to,from)2458 void Plink_copy(to,from)
2459 struct plink **to;
2460 struct plink *from;
2461 {
2462   struct plink *nextpl;
2463   while( from ){
2464     nextpl = from->next;
2465     from->next = *to;
2466     *to = from;
2467     from = nextpl;
2468   }
2469 }
2470 
2471 /* Delete every plink on the list */
Plink_delete(plp)2472 void Plink_delete(plp)
2473 struct plink *plp;
2474 {
2475   struct plink *nextpl;
2476 
2477   while( plp ){
2478     nextpl = plp->next;
2479     plp->next = plink_freelist;
2480     plink_freelist = plp;
2481     plp = nextpl;
2482   }
2483 }
2484 /*********************** From the file "report.c" **************************/
2485 /*
2486 ** Procedures for generating reports and tables in the LEMON parser generator.
2487 */
2488 
2489 /* Generate a filename with the given suffix.  Space to hold the
2490 ** name comes from malloc() and must be freed by the calling
2491 ** function.
2492 */
file_makename(lemp,suffix)2493 PRIVATE char *file_makename(lemp,suffix)
2494 struct lemon *lemp;
2495 char *suffix;
2496 {
2497   char *name;
2498   char *cp;
2499 
2500   name = malloc( strlen(lemp->filename) + strlen(suffix) + 5 );
2501   if( name==0 ){
2502     fprintf(stderr,"Can't allocate space for a filename.\n");
2503     exit(1);
2504   }
2505   strcpy(name,lemp->filename);
2506   cp = strrchr(name,'.');
2507   if( cp ) *cp = 0;
2508   strcat(name,suffix);
2509   return name;
2510 }
2511 
2512 /* Open a file with a name based on the name of the input file,
2513 ** but with a different (specified) suffix, and return a pointer
2514 ** to the stream */
file_open(lemp,suffix,mode)2515 PRIVATE FILE *file_open(lemp,suffix,mode)
2516 struct lemon *lemp;
2517 char *suffix;
2518 char *mode;
2519 {
2520   FILE *fp;
2521 
2522   if( lemp->outname ) free(lemp->outname);
2523   lemp->outname = file_makename(lemp, suffix);
2524   fp = fopen(lemp->outname,mode);
2525   if( fp==0 && *mode=='w' ){
2526     fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
2527     lemp->errorcnt++;
2528     return 0;
2529   }
2530   return fp;
2531 }
2532 
2533 /* Duplicate the input file without comments and without actions
2534 ** on rules */
Reprint(lemp)2535 void Reprint(lemp)
2536 struct lemon *lemp;
2537 {
2538   struct rule *rp;
2539   struct symbol *sp;
2540   int i, j, maxlen, len, ncolumns, skip;
2541   printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
2542   maxlen = 10;
2543   for(i=0; i<lemp->nsymbol; i++){
2544     sp = lemp->symbols[i];
2545     len = strlen(sp->name);
2546     if( len>maxlen ) maxlen = len;
2547   }
2548   ncolumns = 76/(maxlen+5);
2549   if( ncolumns<1 ) ncolumns = 1;
2550   skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
2551   for(i=0; i<skip; i++){
2552     printf("//");
2553     for(j=i; j<lemp->nsymbol; j+=skip){
2554       sp = lemp->symbols[j];
2555       assert( sp->index==j );
2556       printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
2557     }
2558     printf("\n");
2559   }
2560   for(rp=lemp->rule; rp; rp=rp->next){
2561     printf("%s",rp->lhs->name);
2562 /*    if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
2563     printf(" ::=");
2564     for(i=0; i<rp->nrhs; i++){
2565       printf(" %s",rp->rhs[i]->name);
2566 /*      if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
2567     }
2568     printf(".");
2569     if( rp->precsym ) printf(" [%s]",rp->precsym->name);
2570 /*    if( rp->code ) printf("\n    %s",rp->code); */
2571     printf("\n");
2572   }
2573 }
2574 
ConfigPrint(fp,cfp)2575 void ConfigPrint(fp,cfp)
2576 FILE *fp;
2577 struct config *cfp;
2578 {
2579   struct rule *rp;
2580   int i;
2581   rp = cfp->rp;
2582   fprintf(fp,"%s ::=",rp->lhs->name);
2583   for(i=0; i<=rp->nrhs; i++){
2584     if( i==cfp->dot ) fprintf(fp," *");
2585     if( i==rp->nrhs ) break;
2586     fprintf(fp," %s",rp->rhs[i]->name);
2587   }
2588 }
2589 
2590 /* #define TEST */
2591 #ifdef TEST
2592 /* Print a set */
SetPrint(out,set,lemp)2593 PRIVATE void SetPrint(out,set,lemp)
2594 FILE *out;
2595 char *set;
2596 struct lemon *lemp;
2597 {
2598   int i;
2599   char *spacer;
2600   spacer = "";
2601   fprintf(out,"%12s[","");
2602   for(i=0; i<lemp->nterminal; i++){
2603     if( SetFind(set,i) ){
2604       fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
2605       spacer = " ";
2606     }
2607   }
2608   fprintf(out,"]\n");
2609 }
2610 
2611 /* Print a plink chain */
PlinkPrint(out,plp,tag)2612 PRIVATE void PlinkPrint(out,plp,tag)
2613 FILE *out;
2614 struct plink *plp;
2615 char *tag;
2616 {
2617   while( plp ){
2618     fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->index);
2619     ConfigPrint(out,plp->cfp);
2620     fprintf(out,"\n");
2621     plp = plp->next;
2622   }
2623 }
2624 #endif
2625 
2626 /* Print an action to the given file descriptor.  Return FALSE if
2627 ** nothing was actually printed.
2628 */
PrintAction(struct action * ap,FILE * fp,int indent)2629 int PrintAction(struct action *ap, FILE *fp, int indent){
2630   int result = 1;
2631   switch( ap->type ){
2632     case SHIFT:
2633       fprintf(fp,"%*s shift  %d",indent,ap->sp->name,ap->x.stp->index);
2634       break;
2635     case REDUCE:
2636       fprintf(fp,"%*s reduce %d",indent,ap->sp->name,ap->x.rp->index);
2637       break;
2638     case ACCEPT:
2639       fprintf(fp,"%*s accept",indent,ap->sp->name);
2640       break;
2641     case ERROR:
2642       fprintf(fp,"%*s error",indent,ap->sp->name);
2643       break;
2644     case CONFLICT:
2645       fprintf(fp,"%*s reduce %-3d ** Parsing conflict **",
2646         indent,ap->sp->name,ap->x.rp->index);
2647       break;
2648     case SH_RESOLVED:
2649     case RD_RESOLVED:
2650     case NOT_USED:
2651       result = 0;
2652       break;
2653   }
2654   return result;
2655 }
2656 
2657 /* Generate the "y.output" log file */
ReportOutput(lemp)2658 void ReportOutput(lemp)
2659 struct lemon *lemp;
2660 {
2661   int i;
2662   struct state *stp;
2663   struct config *cfp;
2664   struct action *ap;
2665   FILE *fp;
2666 
2667   fp = file_open(lemp,".out","w");
2668   if( fp==0 ) return;
2669   fprintf(fp," \b");
2670   for(i=0; i<lemp->nstate; i++){
2671     stp = lemp->sorted[i];
2672     fprintf(fp,"State %d:\n",stp->index);
2673     if( lemp->basisflag ) cfp=stp->bp;
2674     else                  cfp=stp->cfp;
2675     while( cfp ){
2676       char buf[20];
2677       if( cfp->dot==cfp->rp->nrhs ){
2678         sprintf(buf,"(%d)",cfp->rp->index);
2679         fprintf(fp,"    %5s ",buf);
2680       }else{
2681         fprintf(fp,"          ");
2682       }
2683       ConfigPrint(fp,cfp);
2684       fprintf(fp,"\n");
2685 #ifdef TEST
2686       SetPrint(fp,cfp->fws,lemp);
2687       PlinkPrint(fp,cfp->fplp,"To  ");
2688       PlinkPrint(fp,cfp->bplp,"From");
2689 #endif
2690       if( lemp->basisflag ) cfp=cfp->bp;
2691       else                  cfp=cfp->next;
2692     }
2693     fprintf(fp,"\n");
2694     for(ap=stp->ap; ap; ap=ap->next){
2695       if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
2696     }
2697     fprintf(fp,"\n");
2698   }
2699   fclose(fp);
2700   return;
2701 }
2702 
2703 /* Search for the file "name" which is in the same directory as
2704 ** the exacutable */
pathsearch(argv0,name,modemask)2705 PRIVATE char *pathsearch(argv0,name,modemask)
2706 char *argv0;
2707 char *name;
2708 int modemask;
2709 {
2710   char *pathlist;
2711   char *path,*cp;
2712   char c;
2713   extern int access();
2714 
2715 #ifdef __WIN32__
2716   cp = strrchr(argv0,'\\');
2717 #else
2718   cp = strrchr(argv0,'/');
2719 #endif
2720   if( cp ){
2721     c = *cp;
2722     *cp = 0;
2723     path = (char *)malloc( strlen(argv0) + strlen(name) + 2 );
2724     if( path ) sprintf(path,"%s/%s",argv0,name);
2725     *cp = c;
2726   }else{
2727     extern char *getenv();
2728     pathlist = getenv("PATH");
2729     if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
2730     path = (char *)malloc( strlen(pathlist)+strlen(name)+2 );
2731     if( path!=0 ){
2732       while( *pathlist ){
2733         cp = strchr(pathlist,':');
2734         if( cp==0 ) cp = &pathlist[strlen(pathlist)];
2735         c = *cp;
2736         *cp = 0;
2737         sprintf(path,"%s/%s",pathlist,name);
2738         *cp = c;
2739         if( c==0 ) pathlist = "";
2740         else pathlist = &cp[1];
2741         if( access(path,modemask)==0 ) break;
2742       }
2743     }
2744   }
2745   return path;
2746 }
2747 
2748 /* Given an action, compute the integer value for that action
2749 ** which is to be put in the action table of the generated machine.
2750 ** Return negative if no action should be generated.
2751 */
compute_action(lemp,ap)2752 PRIVATE int compute_action(lemp,ap)
2753 struct lemon *lemp;
2754 struct action *ap;
2755 {
2756   int act;
2757   switch( ap->type ){
2758     case SHIFT:  act = ap->x.stp->index;               break;
2759     case REDUCE: act = ap->x.rp->index + lemp->nstate; break;
2760     case ERROR:  act = lemp->nstate + lemp->nrule;     break;
2761     case ACCEPT: act = lemp->nstate + lemp->nrule + 1; break;
2762     default:     act = -1; break;
2763   }
2764   return act;
2765 }
2766 
2767 #define LINESIZE 1000
2768 /* The next cluster of routines are for reading the template file
2769 ** and writing the results to the generated parser */
2770 /* The first function transfers data from "in" to "out" until
2771 ** a line is seen which begins with "%%".  The line number is
2772 ** tracked.
2773 **
2774 ** if name!=0, then any word that begin with "Parse" is changed to
2775 ** begin with *name instead.
2776 */
tplt_xfer(name,in,out,lineno)2777 PRIVATE void tplt_xfer(name,in,out,lineno)
2778 char *name;
2779 FILE *in;
2780 FILE *out;
2781 int *lineno;
2782 {
2783   int i, iStart;
2784   char line[LINESIZE];
2785   while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
2786     (*lineno)++;
2787     iStart = 0;
2788     if( name ){
2789       for(i=0; line[i]; i++){
2790         if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
2791           && (i==0 || !isalpha(line[i-1]))
2792         ){
2793           if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
2794           fprintf(out,"%s",name);
2795           i += 4;
2796           iStart = i+1;
2797         }
2798       }
2799     }
2800     fprintf(out,"%s",&line[iStart]);
2801   }
2802 }
2803 
2804 /* The next function finds the template file and opens it, returning
2805 ** a pointer to the opened file. */
tplt_open(lemp)2806 PRIVATE FILE *tplt_open(lemp)
2807 struct lemon *lemp;
2808 {
2809   static char templatename[] = "lempar.c";
2810   char buf[1000];
2811   FILE *in;
2812   char *tpltname;
2813   char *cp;
2814 
2815   cp = strrchr(lemp->filename,'.');
2816   if( cp ){
2817     sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
2818   }else{
2819     sprintf(buf,"%s.lt",lemp->filename);
2820   }
2821   if( access(buf,004)==0 ){
2822     tpltname = buf;
2823   }else if( access(templatename,004)==0 ){
2824     tpltname = templatename;
2825   }else{
2826     tpltname = pathsearch(lemp->argv0,templatename,0);
2827   }
2828   if( tpltname==0 ){
2829     fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
2830     templatename);
2831     lemp->errorcnt++;
2832     return 0;
2833   }
2834   in = fopen(tpltname,"r");
2835   if( in==0 ){
2836     fprintf(stderr,"Can't open the template file \"%s\".\n",templatename);
2837     lemp->errorcnt++;
2838     return 0;
2839   }
2840   return in;
2841 }
2842 
2843 /* Print a string to the file and keep the linenumber up to date */
tplt_print(out,lemp,str,strln,lineno)2844 PRIVATE void tplt_print(out,lemp,str,strln,lineno)
2845 FILE *out;
2846 struct lemon *lemp;
2847 char *str;
2848 int strln;
2849 int *lineno;
2850 {
2851   if( str==0 ) return;
2852   fprintf(out,"#line %d \"%s\"\n",strln,lemp->filename); (*lineno)++;
2853   while( *str ){
2854     if( *str=='\n' ) (*lineno)++;
2855     putc(*str,out);
2856     str++;
2857   }
2858   fprintf(out,"\n#line %d \"%s\"\n",*lineno+2,lemp->outname); (*lineno)+=2;
2859   return;
2860 }
2861 
2862 /*
2863 ** The following routine emits code for the destructor for the
2864 ** symbol sp
2865 */
emit_destructor_code(out,sp,lemp,lineno)2866 void emit_destructor_code(out,sp,lemp,lineno)
2867 FILE *out;
2868 struct symbol *sp;
2869 struct lemon *lemp;
2870 int *lineno;
2871 {
2872  char *cp = 0;
2873 
2874  int linecnt = 0;
2875  if( sp->type==TERMINAL ){
2876    cp = lemp->tokendest;
2877    if( cp==0 ) return;
2878    fprintf(out,"#line %d \"%s\"\n{",lemp->tokendestln,lemp->filename);
2879  }else if( sp->destructor ){
2880    cp = sp->destructor;
2881    fprintf(out,"#line %d \"%s\"\n{",sp->destructorln,lemp->filename);
2882  }else if( lemp->vardest ){
2883    cp = lemp->vardest;
2884    if( cp==0 ) return;
2885    fprintf(out,"#line %d \"%s\"\n{",lemp->vardestln,lemp->filename);
2886  }else{
2887    assert( 0 );  /* Cannot happen */
2888  }
2889  for(; *cp; cp++){
2890    if( *cp=='$' && cp[1]=='$' ){
2891      fprintf(out,"(yypminor->yy%d)",sp->dtnum);
2892      cp++;
2893      continue;
2894    }
2895    if( *cp=='\n' ) linecnt++;
2896    fputc(*cp,out);
2897  }
2898  (*lineno) += 3 + linecnt;
2899  fprintf(out,"}\n#line %d \"%s\"\n",*lineno,lemp->outname);
2900  return;
2901 }
2902 
2903 /*
2904 ** Return TRUE (non-zero) if the given symbol has a destructor.
2905 */
has_destructor(sp,lemp)2906 int has_destructor(sp, lemp)
2907 struct symbol *sp;
2908 struct lemon *lemp;
2909 {
2910   int ret;
2911   if( sp->type==TERMINAL ){
2912     ret = lemp->tokendest!=0;
2913   }else{
2914     ret = lemp->vardest!=0 || sp->destructor!=0;
2915   }
2916   return ret;
2917 }
2918 
2919 /*
2920 ** Generate code which executes when the rule "rp" is reduced.  Write
2921 ** the code to "out".  Make sure lineno stays up-to-date.
2922 */
emit_code(out,rp,lemp,lineno)2923 PRIVATE void emit_code(out,rp,lemp,lineno)
2924 FILE *out;
2925 struct rule *rp;
2926 struct lemon *lemp;
2927 int *lineno;
2928 {
2929  char *cp, *xp;
2930  int linecnt = 0;
2931  int i;
2932  char lhsused = 0;    /* True if the LHS element has been used */
2933  char used[MAXRHS];   /* True for each RHS element which is used */
2934 
2935  for(i=0; i<rp->nrhs; i++) used[i] = 0;
2936  lhsused = 0;
2937 
2938  /* Generate code to do the reduce action */
2939  if( rp->code ){
2940    fprintf(out,"#line %d \"%s\"\n{",rp->line,lemp->filename);
2941    for(cp=rp->code; *cp; cp++){
2942      if( isalpha(*cp) && (cp==rp->code || (!isalnum(cp[-1]) && cp[-1]!='_')) ){
2943        char saved;
2944        for(xp= &cp[1]; isalnum(*xp) || *xp=='_'; xp++);
2945        saved = *xp;
2946        *xp = 0;
2947        if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
2948          fprintf(out,"yygotominor.yy%d",rp->lhs->dtnum);
2949          cp = xp;
2950          lhsused = 1;
2951        }else{
2952          for(i=0; i<rp->nrhs; i++){
2953            if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
2954              fprintf(out,"yymsp[%d].minor.yy%d",i-rp->nrhs+1,rp->rhs[i]->dtnum);
2955              cp = xp;
2956              used[i] = 1;
2957              break;
2958            }
2959          }
2960        }
2961        *xp = saved;
2962      }
2963      if( *cp=='\n' ) linecnt++;
2964      fputc(*cp,out);
2965    } /* End loop */
2966    (*lineno) += 3 + linecnt;
2967    fprintf(out,"}\n#line %d \"%s\"\n",*lineno,lemp->outname);
2968  } /* End if( rp->code ) */
2969 
2970  /* Check to make sure the LHS has been used */
2971  if( rp->lhsalias && !lhsused ){
2972    ErrorMsg(lemp->filename,rp->ruleline,
2973      "Label \"%s\" for \"%s(%s)\" is never used.",
2974        rp->lhsalias,rp->lhs->name,rp->lhsalias);
2975    lemp->errorcnt++;
2976  }
2977 
2978  /* Generate destructor code for RHS symbols which are not used in the
2979  ** reduce code */
2980  for(i=0; i<rp->nrhs; i++){
2981    if( rp->rhsalias[i] && !used[i] ){
2982      ErrorMsg(lemp->filename,rp->ruleline,
2983        "Label %s for \"%s(%s)\" is never used.",
2984        rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
2985      lemp->errorcnt++;
2986    }else if( rp->rhsalias[i]==0 ){
2987      if( has_destructor(rp->rhs[i],lemp) ){
2988        fprintf(out,"  yy_destructor(%d,&yymsp[%d].minor);\n",
2989           rp->rhs[i]->index,i-rp->nrhs+1); (*lineno)++;
2990      }else{
2991        fprintf(out,"        /* No destructor defined for %s */\n",
2992         rp->rhs[i]->name);
2993         (*lineno)++;
2994      }
2995    }
2996  }
2997  return;
2998 }
2999 
3000 /*
3001 ** Print the definition of the union used for the parser's data stack.
3002 ** This union contains fields for every possible data type for tokens
3003 ** and nonterminals.  In the process of computing and printing this
3004 ** union, also set the ".dtnum" field of every terminal and nonterminal
3005 ** symbol.
3006 */
print_stack_union(out,lemp,plineno,mhflag)3007 void print_stack_union(out,lemp,plineno,mhflag)
3008 FILE *out;                  /* The output stream */
3009 struct lemon *lemp;         /* The main info structure for this parser */
3010 int *plineno;               /* Pointer to the line number */
3011 int mhflag;                 /* True if generating makeheaders output */
3012 {
3013   int lineno = *plineno;    /* The line number of the output */
3014   char **types;             /* A hash table of datatypes */
3015   int arraysize;            /* Size of the "types" array */
3016   int maxdtlength;          /* Maximum length of any ".datatype" field. */
3017   char *stddt;              /* Standardized name for a datatype */
3018   int i,j;                  /* Loop counters */
3019   int hash;                 /* For hashing the name of a type */
3020   char *name;               /* Name of the parser */
3021 
3022   /* Allocate and initialize types[] and allocate stddt[] */
3023   arraysize = lemp->nsymbol * 2;
3024   types = (char**)malloc( arraysize * sizeof(char*) );
3025   for(i=0; i<arraysize; i++) types[i] = 0;
3026   maxdtlength = 0;
3027   if( lemp->vartype ){
3028     maxdtlength = strlen(lemp->vartype);
3029   }
3030   for(i=0; i<lemp->nsymbol; i++){
3031     int len;
3032     struct symbol *sp = lemp->symbols[i];
3033     if( sp->datatype==0 ) continue;
3034     len = strlen(sp->datatype);
3035     if( len>maxdtlength ) maxdtlength = len;
3036   }
3037   stddt = (char*)malloc( maxdtlength*2 + 1 );
3038   if( types==0 || stddt==0 ){
3039     fprintf(stderr,"Out of memory.\n");
3040     exit(1);
3041   }
3042 
3043   /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3044   ** is filled in with the hash index plus 1.  A ".dtnum" value of 0 is
3045   ** used for terminal symbols.  If there is no %default_type defined then
3046   ** 0 is also used as the .dtnum value for nonterminals which do not specify
3047   ** a datatype using the %type directive.
3048   */
3049   for(i=0; i<lemp->nsymbol; i++){
3050     struct symbol *sp = lemp->symbols[i];
3051     char *cp;
3052     if( sp==lemp->errsym ){
3053       sp->dtnum = arraysize+1;
3054       continue;
3055     }
3056     if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
3057       sp->dtnum = 0;
3058       continue;
3059     }
3060     cp = sp->datatype;
3061     if( cp==0 ) cp = lemp->vartype;
3062     j = 0;
3063     while( isspace(*cp) ) cp++;
3064     while( *cp ) stddt[j++] = *cp++;
3065     while( j>0 && isspace(stddt[j-1]) ) j--;
3066     stddt[j] = 0;
3067     hash = 0;
3068     for(j=0; stddt[j]; j++){
3069       hash = hash*53 + stddt[j];
3070     }
3071     hash = (hash & 0x7fffffff)%arraysize;
3072     while( types[hash] ){
3073       if( strcmp(types[hash],stddt)==0 ){
3074         sp->dtnum = hash + 1;
3075