1 /*
2 ** This file contains all sources (including headers) to the LEMON
3 ** LALR(1) parser generator.  The sources have been combined into a
4 ** single file to make it easy to include LEMON in the source tree
5 ** and Makefile of another program.
6 **
7 ** The author of this program disclaims copyright.
8 */
9 #include <stdio.h>
10 #include <stdarg.h>
11 #include <string.h>
12 #include <ctype.h>
13 #include <stdlib.h>
14 #include <assert.h>
15 
16 #define ISSPACE(X) isspace((unsigned char)(X))
17 #define ISDIGIT(X) isdigit((unsigned char)(X))
18 #define ISALNUM(X) isalnum((unsigned char)(X))
19 #define ISALPHA(X) isalpha((unsigned char)(X))
20 #define ISUPPER(X) isupper((unsigned char)(X))
21 #define ISLOWER(X) islower((unsigned char)(X))
22 
23 
24 #ifndef __WIN32__
25 #   if defined(_WIN32) || defined(WIN32)
26 #       define __WIN32__
27 #   endif
28 #endif
29 
30 #ifdef __WIN32__
31 #ifdef __cplusplus
32 extern "C" {
33 #endif
34 extern int access(const char *path, int mode);
35 #ifdef __cplusplus
36 }
37 #endif
38 #else
39 #include <unistd.h>
40 #endif
41 
42 /* #define PRIVATE static */
43 #define PRIVATE
44 
45 #ifdef TEST
46 #define MAXRHS 5       /* Set low to exercise exception code */
47 #else
48 #define MAXRHS 1000
49 #endif
50 
51 extern void memory_error(void);
52 static int showPrecedenceConflict = 0;
53 static char *msort(char*,char**,int(*)(const char*,const char*));
54 
55 /*
56 ** Compilers are getting increasingly pedantic about type conversions
57 ** as C evolves ever closer to Ada....  To work around the latest problems
58 ** we have to define the following variant of strlen().
59 */
60 #define lemonStrlen(X)   ((int)strlen(X))
61 
62 /*
63 ** Compilers are starting to complain about the use of sprintf() and strcpy(),
64 ** saying they are unsafe.  So we define our own versions of those routines too.
65 **
66 ** There are three routines here:  lemon_sprintf(), lemon_vsprintf(), and
67 ** lemon_addtext(). The first two are replacements for sprintf() and vsprintf().
68 ** The third is a helper routine for vsnprintf() that adds texts to the end of a
69 ** buffer, making sure the buffer is always zero-terminated.
70 **
71 ** The string formatter is a minimal subset of stdlib sprintf() supporting only
72 ** a few simply conversions:
73 **
74 **   %d
75 **   %s
76 **   %.*s
77 **
78 */
lemon_addtext(char * zBuf,int * pnUsed,const char * zIn,int nIn,int iWidth)79 static void lemon_addtext(
80   char *zBuf,           /* The buffer to which text is added */
81   int *pnUsed,          /* Slots of the buffer used so far */
82   const char *zIn,      /* Text to add */
83   int nIn,              /* Bytes of text to add.  -1 to use strlen() */
84   int iWidth            /* Field width.  Negative to left justify */
85 ){
86   if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){}
87   while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; }
88   if( nIn==0 ) return;
89   memcpy(&zBuf[*pnUsed], zIn, nIn);
90   *pnUsed += nIn;
91   while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; }
92   zBuf[*pnUsed] = 0;
93 }
lemon_vsprintf(char * str,const char * zFormat,va_list ap)94 static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){
95   int i, j, k, c;
96   int nUsed = 0;
97   const char *z;
98   char zTemp[50];
99   str[0] = 0;
100   for(i=j=0; (c = zFormat[i])!=0; i++){
101     if( c=='%' ){
102       int iWidth = 0;
103       lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
104       c = zFormat[++i];
105       if( ISDIGIT(c) || (c=='-' && ISDIGIT(zFormat[i+1])) ){
106         if( c=='-' ) i++;
107         while( ISDIGIT(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0';
108         if( c=='-' ) iWidth = -iWidth;
109         c = zFormat[i];
110       }
111       if( c=='d' ){
112         int v = va_arg(ap, int);
113         if( v<0 ){
114           lemon_addtext(str, &nUsed, "-", 1, iWidth);
115           v = -v;
116         }else if( v==0 ){
117           lemon_addtext(str, &nUsed, "0", 1, iWidth);
118         }
119         k = 0;
120         while( v>0 ){
121           k++;
122           zTemp[sizeof(zTemp)-k] = (v%10) + '0';
123           v /= 10;
124         }
125         lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth);
126       }else if( c=='s' ){
127         z = va_arg(ap, const char*);
128         lemon_addtext(str, &nUsed, z, -1, iWidth);
129       }else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){
130         i += 2;
131         k = va_arg(ap, int);
132         z = va_arg(ap, const char*);
133         lemon_addtext(str, &nUsed, z, k, iWidth);
134       }else if( c=='%' ){
135         lemon_addtext(str, &nUsed, "%", 1, 0);
136       }else{
137         fprintf(stderr, "illegal format\n");
138         exit(1);
139       }
140       j = i+1;
141     }
142   }
143   lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
144   return nUsed;
145 }
lemon_sprintf(char * str,const char * format,...)146 static int lemon_sprintf(char *str, const char *format, ...){
147   va_list ap;
148   int rc;
149   va_start(ap, format);
150   rc = lemon_vsprintf(str, format, ap);
151   va_end(ap);
152   return rc;
153 }
lemon_strcpy(char * dest,const char * src)154 static void lemon_strcpy(char *dest, const char *src){
155   while( (*(dest++) = *(src++))!=0 ){}
156 }
lemon_strcat(char * dest,const char * src)157 static void lemon_strcat(char *dest, const char *src){
158   while( *dest ) dest++;
159   lemon_strcpy(dest, src);
160 }
161 
162 
163 /* a few forward declarations... */
164 struct rule;
165 struct lemon;
166 struct action;
167 
168 static struct action *Action_new(void);
169 static struct action *Action_sort(struct action *);
170 
171 /********** From the file "build.h" ************************************/
172 void FindRulePrecedences(struct lemon*);
173 void FindFirstSets(struct lemon*);
174 void FindStates(struct lemon*);
175 void FindLinks(struct lemon*);
176 void FindFollowSets(struct lemon*);
177 void FindActions(struct lemon*);
178 
179 /********* From the file "configlist.h" *********************************/
180 void Configlist_init(void);
181 struct config *Configlist_add(struct rule *, int);
182 struct config *Configlist_addbasis(struct rule *, int);
183 void Configlist_closure(struct lemon *);
184 void Configlist_sort(void);
185 void Configlist_sortbasis(void);
186 struct config *Configlist_return(void);
187 struct config *Configlist_basis(void);
188 void Configlist_eat(struct config *);
189 void Configlist_reset(void);
190 
191 /********* From the file "error.h" ***************************************/
192 void ErrorMsg(const char *, int,const char *, ...);
193 
194 /****** From the file "option.h" ******************************************/
195 enum option_type { OPT_FLAG=1,  OPT_INT,  OPT_DBL,  OPT_STR,
196          OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR};
197 struct s_options {
198   enum option_type type;
199   const char *label;
200   char *arg;
201   const char *message;
202 };
203 int    OptInit(char**,struct s_options*,FILE*);
204 int    OptNArgs(void);
205 char  *OptArg(int);
206 void   OptErr(int);
207 void   OptPrint(void);
208 
209 /******** From the file "parse.h" *****************************************/
210 void Parse(struct lemon *lemp);
211 
212 /********* From the file "plink.h" ***************************************/
213 struct plink *Plink_new(void);
214 void Plink_add(struct plink **, struct config *);
215 void Plink_copy(struct plink **, struct plink *);
216 void Plink_delete(struct plink *);
217 
218 /********** From the file "report.h" *************************************/
219 void Reprint(struct lemon *);
220 void ReportOutput(struct lemon *);
221 void ReportTable(struct lemon *, int, int);
222 void ReportHeader(struct lemon *);
223 void CompressTables(struct lemon *);
224 void ResortStates(struct lemon *);
225 
226 /********** From the file "set.h" ****************************************/
227 void  SetSize(int);             /* All sets will be of size N */
228 char *SetNew(void);               /* A new set for element 0..N */
229 void  SetFree(char*);             /* Deallocate a set */
230 int SetAdd(char*,int);            /* Add element to a set */
231 int SetUnion(char *,char *);    /* A <- A U B, thru element N */
232 #define SetFind(X,Y) (X[Y])       /* True if Y is in set X */
233 
234 /********** From the file "struct.h" *************************************/
235 /*
236 ** Principal data structures for the LEMON parser generator.
237 */
238 
239 typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
240 
241 /* Symbols (terminals and nonterminals) of the grammar are stored
242 ** in the following: */
243 enum symbol_type {
244   TERMINAL,
245   NONTERMINAL,
246   MULTITERMINAL
247 };
248 enum e_assoc {
249     LEFT,
250     RIGHT,
251     NONE,
252     UNK
253 };
254 struct symbol {
255   const char *name;        /* Name of the symbol */
256   int index;               /* Index number for this symbol */
257   enum symbol_type type;   /* Symbols are all either TERMINALS or NTs */
258   struct rule *rule;       /* Linked list of rules of this (if an NT) */
259   struct symbol *fallback; /* fallback token in case this token doesn't parse */
260   int prec;                /* Precedence if defined (-1 otherwise) */
261   enum e_assoc assoc;      /* Associativity if precedence is defined */
262   char *firstset;          /* First-set for all rules of this symbol */
263   Boolean lambda;          /* True if NT and can generate an empty string */
264   int useCnt;              /* Number of times used */
265   char *destructor;        /* Code which executes whenever this symbol is
266                            ** popped from the stack during error processing */
267   int destLineno;          /* Line number for start of destructor.  Set to
268                            ** -1 for duplicate destructors. */
269   char *datatype;          /* The data type of information held by this
270                            ** object. Only used if type==NONTERMINAL */
271   int dtnum;               /* The data type number.  In the parser, the value
272                            ** stack is a union.  The .yy%d element of this
273                            ** union is the correct data type for this object */
274   int bContent;            /* True if this symbol ever carries content - if
275                            ** it is ever more than just syntax */
276   /* The following fields are used by MULTITERMINALs only */
277   int nsubsym;             /* Number of constituent symbols in the MULTI */
278   struct symbol **subsym;  /* Array of constituent symbols */
279 };
280 
281 /* Each production rule in the grammar is stored in the following
282 ** structure.  */
283 struct rule {
284   struct symbol *lhs;      /* Left-hand side of the rule */
285   const char *lhsalias;    /* Alias for the LHS (NULL if none) */
286   int lhsStart;            /* True if left-hand side is the start symbol */
287   int ruleline;            /* Line number for the rule */
288   int nrhs;                /* Number of RHS symbols */
289   struct symbol **rhs;     /* The RHS symbols */
290   const char **rhsalias;   /* An alias for each RHS symbol (NULL if none) */
291   int line;                /* Line number at which code begins */
292   const char *code;        /* The code executed when this rule is reduced */
293   const char *codePrefix;  /* Setup code before code[] above */
294   const char *codeSuffix;  /* Breakdown code after code[] above */
295   struct symbol *precsym;  /* Precedence symbol for this rule */
296   int index;               /* An index number for this rule */
297   int iRule;               /* Rule number as used in the generated tables */
298   Boolean noCode;          /* True if this rule has no associated C code */
299   Boolean codeEmitted;     /* True if the code has been emitted already */
300   Boolean canReduce;       /* True if this rule is ever reduced */
301   Boolean doesReduce;      /* Reduce actions occur after optimization */
302   Boolean neverReduce;     /* Reduce is theoretically possible, but prevented
303                            ** by actions or other outside implementation */
304   struct rule *nextlhs;    /* Next rule with the same LHS */
305   struct rule *next;       /* Next rule in the global list */
306 };
307 
308 /* A configuration is a production rule of the grammar together with
309 ** a mark (dot) showing how much of that rule has been processed so far.
310 ** Configurations also contain a follow-set which is a list of terminal
311 ** symbols which are allowed to immediately follow the end of the rule.
312 ** Every configuration is recorded as an instance of the following: */
313 enum cfgstatus {
314   COMPLETE,
315   INCOMPLETE
316 };
317 struct config {
318   struct rule *rp;         /* The rule upon which the configuration is based */
319   int dot;                 /* The parse point */
320   char *fws;               /* Follow-set for this configuration only */
321   struct plink *fplp;      /* Follow-set forward propagation links */
322   struct plink *bplp;      /* Follow-set backwards propagation links */
323   struct state *stp;       /* Pointer to state which contains this */
324   enum cfgstatus status;   /* used during followset and shift computations */
325   struct config *next;     /* Next configuration in the state */
326   struct config *bp;       /* The next basis configuration */
327 };
328 
329 enum e_action {
330   SHIFT,
331   ACCEPT,
332   REDUCE,
333   ERROR,
334   SSCONFLICT,              /* A shift/shift conflict */
335   SRCONFLICT,              /* Was a reduce, but part of a conflict */
336   RRCONFLICT,              /* Was a reduce, but part of a conflict */
337   SH_RESOLVED,             /* Was a shift.  Precedence resolved conflict */
338   RD_RESOLVED,             /* Was reduce.  Precedence resolved conflict */
339   NOT_USED,                /* Deleted by compression */
340   SHIFTREDUCE              /* Shift first, then reduce */
341 };
342 
343 /* Every shift or reduce operation is stored as one of the following */
344 struct action {
345   struct symbol *sp;       /* The look-ahead symbol */
346   enum e_action type;
347   union {
348     struct state *stp;     /* The new state, if a shift */
349     struct rule *rp;       /* The rule, if a reduce */
350   } x;
351   struct symbol *spOpt;    /* SHIFTREDUCE optimization to this symbol */
352   struct action *next;     /* Next action for this state */
353   struct action *collide;  /* Next action with the same hash */
354 };
355 
356 /* Each state of the generated parser's finite state machine
357 ** is encoded as an instance of the following structure. */
358 struct state {
359   struct config *bp;       /* The basis configurations for this state */
360   struct config *cfp;      /* All configurations in this set */
361   int statenum;            /* Sequential number for this state */
362   struct action *ap;       /* List of actions for this state */
363   int nTknAct, nNtAct;     /* Number of actions on terminals and nonterminals */
364   int iTknOfst, iNtOfst;   /* yy_action[] offset for terminals and nonterms */
365   int iDfltReduce;         /* Default action is to REDUCE by this rule */
366   struct rule *pDfltReduce;/* The default REDUCE rule. */
367   int autoReduce;          /* True if this is an auto-reduce state */
368 };
369 #define NO_OFFSET (-2147483647)
370 
371 /* A followset propagation link indicates that the contents of one
372 ** configuration followset should be propagated to another whenever
373 ** the first changes. */
374 struct plink {
375   struct config *cfp;      /* The configuration to which linked */
376   struct plink *next;      /* The next propagate link */
377 };
378 
379 /* The state vector for the entire parser generator is recorded as
380 ** follows.  (LEMON uses no global variables and makes little use of
381 ** static variables.  Fields in the following structure can be thought
382 ** of as begin global variables in the program.) */
383 struct lemon {
384   struct state **sorted;   /* Table of states sorted by state number */
385   struct rule *rule;       /* List of all rules */
386   struct rule *startRule;  /* First rule */
387   int nstate;              /* Number of states */
388   int nxstate;             /* nstate with tail degenerate states removed */
389   int nrule;               /* Number of rules */
390   int nruleWithAction;     /* Number of rules with actions */
391   int nsymbol;             /* Number of terminal and nonterminal symbols */
392   int nterminal;           /* Number of terminal symbols */
393   int minShiftReduce;      /* Minimum shift-reduce action value */
394   int errAction;           /* Error action value */
395   int accAction;           /* Accept action value */
396   int noAction;            /* No-op action value */
397   int minReduce;           /* Minimum reduce action */
398   int maxAction;           /* Maximum action value of any kind */
399   struct symbol **symbols; /* Sorted array of pointers to symbols */
400   int errorcnt;            /* Number of errors */
401   struct symbol *errsym;   /* The error symbol */
402   struct symbol *wildcard; /* Token that matches anything */
403   char *name;              /* Name of the generated parser */
404   char *arg;               /* Declaration of the 3th argument to parser */
405   char *ctx;               /* Declaration of 2nd argument to constructor */
406   char *tokentype;         /* Type of terminal symbols in the parser stack */
407   char *vartype;           /* The default type of non-terminal symbols */
408   char *start;             /* Name of the start symbol for the grammar */
409   char *stacksize;         /* Size of the parser stack */
410   char *include;           /* Code to put at the start of the C file */
411   char *error;             /* Code to execute when an error is seen */
412   char *overflow;          /* Code to execute on a stack overflow */
413   char *failure;           /* Code to execute on parser failure */
414   char *accept;            /* Code to execute when the parser excepts */
415   char *extracode;         /* Code appended to the generated file */
416   char *tokendest;         /* Code to execute to destroy token data */
417   char *vardest;           /* Code for the default non-terminal destructor */
418   char *filename;          /* Name of the input file */
419   char *outname;           /* Name of the current output file */
420   char *tokenprefix;       /* A prefix added to token names in the .h file */
421   int nconflict;           /* Number of parsing conflicts */
422   int nactiontab;          /* Number of entries in the yy_action[] table */
423   int nlookaheadtab;       /* Number of entries in yy_lookahead[] */
424   int tablesize;           /* Total table size of all tables in bytes */
425   int basisflag;           /* Print only basis configurations */
426   int printPreprocessed;   /* Show preprocessor output on stdout */
427   int has_fallback;        /* True if any %fallback is seen in the grammar */
428   int nolinenosflag;       /* True if #line statements should not be printed */
429   char *argv0;             /* Name of the program */
430 };
431 
432 #define MemoryCheck(X) if((X)==0){ \
433   memory_error(); \
434 }
435 
436 /**************** From the file "table.h" *********************************/
437 /*
438 ** All code in this file has been automatically generated
439 ** from a specification in the file
440 **              "table.q"
441 ** by the associative array code building program "aagen".
442 ** Do not edit this file!  Instead, edit the specification
443 ** file, then rerun aagen.
444 */
445 /*
446 ** Code for processing tables in the LEMON parser generator.
447 */
448 /* Routines for handling a strings */
449 
450 const char *Strsafe(const char *);
451 
452 void Strsafe_init(void);
453 int Strsafe_insert(const char *);
454 const char *Strsafe_find(const char *);
455 
456 /* Routines for handling symbols of the grammar */
457 
458 struct symbol *Symbol_new(const char *);
459 int Symbolcmpp(const void *, const void *);
460 void Symbol_init(void);
461 int Symbol_insert(struct symbol *, const char *);
462 struct symbol *Symbol_find(const char *);
463 struct symbol *Symbol_Nth(int);
464 int Symbol_count(void);
465 struct symbol **Symbol_arrayof(void);
466 
467 /* Routines to manage the state table */
468 
469 int Configcmp(const char *, const char *);
470 struct state *State_new(void);
471 void State_init(void);
472 int State_insert(struct state *, struct config *);
473 struct state *State_find(struct config *);
474 struct state **State_arrayof(void);
475 
476 /* Routines used for efficiency in Configlist_add */
477 
478 void Configtable_init(void);
479 int Configtable_insert(struct config *);
480 struct config *Configtable_find(struct config *);
481 void Configtable_clear(int(*)(struct config *));
482 
483 /****************** From the file "action.c" *******************************/
484 /*
485 ** Routines processing parser actions in the LEMON parser generator.
486 */
487 
488 /* Allocate a new parser action */
Action_new(void)489 static struct action *Action_new(void){
490   static struct action *actionfreelist = 0;
491   struct action *newaction;
492 
493   if( actionfreelist==0 ){
494     int i;
495     int amt = 100;
496     actionfreelist = (struct action *)calloc(amt, sizeof(struct action));
497     if( actionfreelist==0 ){
498       fprintf(stderr,"Unable to allocate memory for a new parser action.");
499       exit(1);
500     }
501     for(i=0; i<amt-1; i++) actionfreelist[i].next = &actionfreelist[i+1];
502     actionfreelist[amt-1].next = 0;
503   }
504   newaction = actionfreelist;
505   actionfreelist = actionfreelist->next;
506   return newaction;
507 }
508 
509 /* Compare two actions for sorting purposes.  Return negative, zero, or
510 ** positive if the first action is less than, equal to, or greater than
511 ** the first
512 */
actioncmp(struct action * ap1,struct action * ap2)513 static int actioncmp(
514   struct action *ap1,
515   struct action *ap2
516 ){
517   int rc;
518   rc = ap1->sp->index - ap2->sp->index;
519   if( rc==0 ){
520     rc = (int)ap1->type - (int)ap2->type;
521   }
522   if( rc==0 && (ap1->type==REDUCE || ap1->type==SHIFTREDUCE) ){
523     rc = ap1->x.rp->index - ap2->x.rp->index;
524   }
525   if( rc==0 ){
526     rc = (int) (ap2 - ap1);
527   }
528   return rc;
529 }
530 
531 /* Sort parser actions */
Action_sort(struct action * ap)532 static struct action *Action_sort(
533   struct action *ap
534 ){
535   ap = (struct action *)msort((char *)ap,(char **)&ap->next,
536                               (int(*)(const char*,const char*))actioncmp);
537   return ap;
538 }
539 
Action_add(struct action ** app,enum e_action type,struct symbol * sp,char * arg)540 void Action_add(
541   struct action **app,
542   enum e_action type,
543   struct symbol *sp,
544   char *arg
545 ){
546   struct action *newaction;
547   newaction = Action_new();
548   newaction->next = *app;
549   *app = newaction;
550   newaction->type = type;
551   newaction->sp = sp;
552   newaction->spOpt = 0;
553   if( type==SHIFT ){
554     newaction->x.stp = (struct state *)arg;
555   }else{
556     newaction->x.rp = (struct rule *)arg;
557   }
558 }
559 /********************** New code to implement the "acttab" module ***********/
560 /*
561 ** This module implements routines use to construct the yy_action[] table.
562 */
563 
564 /*
565 ** The state of the yy_action table under construction is an instance of
566 ** the following structure.
567 **
568 ** The yy_action table maps the pair (state_number, lookahead) into an
569 ** action_number.  The table is an array of integers pairs.  The state_number
570 ** determines an initial offset into the yy_action array.  The lookahead
571 ** value is then added to this initial offset to get an index X into the
572 ** yy_action array. If the aAction[X].lookahead equals the value of the
573 ** of the lookahead input, then the value of the action_number output is
574 ** aAction[X].action.  If the lookaheads do not match then the
575 ** default action for the state_number is returned.
576 **
577 ** All actions associated with a single state_number are first entered
578 ** into aLookahead[] using multiple calls to acttab_action().  Then the
579 ** actions for that single state_number are placed into the aAction[]
580 ** array with a single call to acttab_insert().  The acttab_insert() call
581 ** also resets the aLookahead[] array in preparation for the next
582 ** state number.
583 */
584 struct lookahead_action {
585   int lookahead;             /* Value of the lookahead token */
586   int action;                /* Action to take on the given lookahead */
587 };
588 typedef struct acttab acttab;
589 struct acttab {
590   int nAction;                 /* Number of used slots in aAction[] */
591   int nActionAlloc;            /* Slots allocated for aAction[] */
592   struct lookahead_action
593     *aAction,                  /* The yy_action[] table under construction */
594     *aLookahead;               /* A single new transaction set */
595   int mnLookahead;             /* Minimum aLookahead[].lookahead */
596   int mnAction;                /* Action associated with mnLookahead */
597   int mxLookahead;             /* Maximum aLookahead[].lookahead */
598   int nLookahead;              /* Used slots in aLookahead[] */
599   int nLookaheadAlloc;         /* Slots allocated in aLookahead[] */
600   int nterminal;               /* Number of terminal symbols */
601   int nsymbol;                 /* total number of symbols */
602 };
603 
604 /* Return the number of entries in the yy_action table */
605 #define acttab_lookahead_size(X) ((X)->nAction)
606 
607 /* The value for the N-th entry in yy_action */
608 #define acttab_yyaction(X,N)  ((X)->aAction[N].action)
609 
610 /* The value for the N-th entry in yy_lookahead */
611 #define acttab_yylookahead(X,N)  ((X)->aAction[N].lookahead)
612 
613 /* Free all memory associated with the given acttab */
acttab_free(acttab * p)614 void acttab_free(acttab *p){
615   free( p->aAction );
616   free( p->aLookahead );
617   free( p );
618 }
619 
620 /* Allocate a new acttab structure */
acttab_alloc(int nsymbol,int nterminal)621 acttab *acttab_alloc(int nsymbol, int nterminal){
622   acttab *p = (acttab *) calloc( 1, sizeof(*p) );
623   if( p==0 ){
624     fprintf(stderr,"Unable to allocate memory for a new acttab.");
625     exit(1);
626   }
627   memset(p, 0, sizeof(*p));
628   p->nsymbol = nsymbol;
629   p->nterminal = nterminal;
630   return p;
631 }
632 
633 /* Add a new action to the current transaction set.
634 **
635 ** This routine is called once for each lookahead for a particular
636 ** state.
637 */
acttab_action(acttab * p,int lookahead,int action)638 void acttab_action(acttab *p, int lookahead, int action){
639   if( p->nLookahead>=p->nLookaheadAlloc ){
640     p->nLookaheadAlloc += 25;
641     p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead,
642                              sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
643     if( p->aLookahead==0 ){
644       fprintf(stderr,"malloc failed\n");
645       exit(1);
646     }
647   }
648   if( p->nLookahead==0 ){
649     p->mxLookahead = lookahead;
650     p->mnLookahead = lookahead;
651     p->mnAction = action;
652   }else{
653     if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
654     if( p->mnLookahead>lookahead ){
655       p->mnLookahead = lookahead;
656       p->mnAction = action;
657     }
658   }
659   p->aLookahead[p->nLookahead].lookahead = lookahead;
660   p->aLookahead[p->nLookahead].action = action;
661   p->nLookahead++;
662 }
663 
664 /*
665 ** Add the transaction set built up with prior calls to acttab_action()
666 ** into the current action table.  Then reset the transaction set back
667 ** to an empty set in preparation for a new round of acttab_action() calls.
668 **
669 ** Return the offset into the action table of the new transaction.
670 **
671 ** If the makeItSafe parameter is true, then the offset is chosen so that
672 ** it is impossible to overread the yy_lookaside[] table regardless of
673 ** the lookaside token.  This is done for the terminal symbols, as they
674 ** come from external inputs and can contain syntax errors.  When makeItSafe
675 ** is false, there is more flexibility in selecting offsets, resulting in
676 ** a smaller table.  For non-terminal symbols, which are never syntax errors,
677 ** makeItSafe can be false.
678 */
acttab_insert(acttab * p,int makeItSafe)679 int acttab_insert(acttab *p, int makeItSafe){
680   int i, j, k, n, end;
681   assert( p->nLookahead>0 );
682 
683   /* Make sure we have enough space to hold the expanded action table
684   ** in the worst case.  The worst case occurs if the transaction set
685   ** must be appended to the current action table
686   */
687   n = p->nsymbol + 1;
688   if( p->nAction + n >= p->nActionAlloc ){
689     int oldAlloc = p->nActionAlloc;
690     p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
691     p->aAction = (struct lookahead_action *) realloc( p->aAction,
692                           sizeof(p->aAction[0])*p->nActionAlloc);
693     if( p->aAction==0 ){
694       fprintf(stderr,"malloc failed\n");
695       exit(1);
696     }
697     assert(oldAlloc < p->nActionAlloc);     /* hint for CSA */
698     for(i=oldAlloc; i<p->nActionAlloc; i++){
699       p->aAction[i].lookahead = -1;
700       p->aAction[i].action = -1;
701     }
702   }
703   assert(p->aAction);  /* Hint for CSA (for p->aAction[i] below) */
704 
705   /* Scan the existing action table looking for an offset that is a
706   ** duplicate of the current transaction set.  Fall out of the loop
707   ** if and when the duplicate is found.
708   **
709   ** i is the index in p->aAction[] where p->mnLookahead is inserted.
710   */
711   end = makeItSafe ? p->mnLookahead : 0;
712   for(i=p->nAction-1; i>=end; i--){
713     if( p->aAction[i].lookahead==p->mnLookahead ){
714       /* All lookaheads and actions in the aLookahead[] transaction
715       ** must match against the candidate aAction[i] entry. */
716       if( p->aAction[i].action!=p->mnAction ) continue;
717       for(j=0; j<p->nLookahead; j++){
718         k = p->aLookahead[j].lookahead - p->mnLookahead + i;
719         if( k<0 || k>=p->nAction ) break;
720         if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
721         if( p->aLookahead[j].action!=p->aAction[k].action ) break;
722       }
723       if( j<p->nLookahead ) continue;
724 
725       /* No possible lookahead value that is not in the aLookahead[]
726       ** transaction is allowed to match aAction[i] */
727       n = 0;
728       for(j=0; j<p->nAction; j++){
729         if( p->aAction[j].lookahead<0 ) continue;
730         if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
731       }
732       if( n==p->nLookahead ){
733         break;  /* An exact match is found at offset i */
734       }
735     }
736   }
737 
738   /* If no existing offsets exactly match the current transaction, find an
739   ** an empty offset in the aAction[] table in which we can add the
740   ** aLookahead[] transaction.
741   */
742   if( i<end ){
743     /* Look for holes in the aAction[] table that fit the current
744     ** aLookahead[] transaction.  Leave i set to the offset of the hole.
745     ** If no holes are found, i is left at p->nAction, which means the
746     ** transaction will be appended. */
747     i = makeItSafe ? p->mnLookahead : 0;
748     for(; i<p->nActionAlloc - p->mxLookahead; i++){
749       if( p->aAction[i].lookahead<0 ){
750         for(j=0; j<p->nLookahead; j++){
751           k = p->aLookahead[j].lookahead - p->mnLookahead + i;
752           if( k<0 ) break;
753           if( p->aAction[k].lookahead>=0 ) break;
754         }
755         if( j<p->nLookahead ) continue;
756         for(j=0; j<p->nAction; j++){
757           if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
758         }
759         if( j==p->nAction ){
760           break;  /* Fits in empty slots */
761         }
762       }
763     }
764   }
765   /* Insert transaction set at index i. */
766 #if 0
767   printf("Acttab:");
768   for(j=0; j<p->nLookahead; j++){
769     printf(" %d", p->aLookahead[j].lookahead);
770   }
771   printf(" inserted at %d\n", i);
772 #endif
773   for(j=0; j<p->nLookahead; j++){
774     k = p->aLookahead[j].lookahead - p->mnLookahead + i;
775     p->aAction[k] = p->aLookahead[j];
776     if( k>=p->nAction ) p->nAction = k+1;
777   }
778   if( makeItSafe && i+p->nterminal>=p->nAction ) p->nAction = i+p->nterminal+1;
779   p->nLookahead = 0;
780 
781   /* Return the offset that is added to the lookahead in order to get the
782   ** index into yy_action of the action */
783   return i - p->mnLookahead;
784 }
785 
786 /*
787 ** Return the size of the action table without the trailing syntax error
788 ** entries.
789 */
acttab_action_size(acttab * p)790 int acttab_action_size(acttab *p){
791   int n = p->nAction;
792   while( n>0 && p->aAction[n-1].lookahead<0 ){ n--; }
793   return n;
794 }
795 
796 /********************** From the file "build.c" *****************************/
797 /*
798 ** Routines to construction the finite state machine for the LEMON
799 ** parser generator.
800 */
801 
802 /* Find a precedence symbol of every rule in the grammar.
803 **
804 ** Those rules which have a precedence symbol coded in the input
805 ** grammar using the "[symbol]" construct will already have the
806 ** rp->precsym field filled.  Other rules take as their precedence
807 ** symbol the first RHS symbol with a defined precedence.  If there
808 ** are not RHS symbols with a defined precedence, the precedence
809 ** symbol field is left blank.
810 */
FindRulePrecedences(struct lemon * xp)811 void FindRulePrecedences(struct lemon *xp)
812 {
813   struct rule *rp;
814   for(rp=xp->rule; rp; rp=rp->next){
815     if( rp->precsym==0 ){
816       int i, j;
817       for(i=0; i<rp->nrhs && rp->precsym==0; i++){
818         struct symbol *sp = rp->rhs[i];
819         if( sp->type==MULTITERMINAL ){
820           for(j=0; j<sp->nsubsym; j++){
821             if( sp->subsym[j]->prec>=0 ){
822               rp->precsym = sp->subsym[j];
823               break;
824             }
825           }
826         }else if( sp->prec>=0 ){
827           rp->precsym = rp->rhs[i];
828         }
829       }
830     }
831   }
832   return;
833 }
834 
835 /* Find all nonterminals which will generate the empty string.
836 ** Then go back and compute the first sets of every nonterminal.
837 ** The first set is the set of all terminal symbols which can begin
838 ** a string generated by that nonterminal.
839 */
FindFirstSets(struct lemon * lemp)840 void FindFirstSets(struct lemon *lemp)
841 {
842   int i, j;
843   struct rule *rp;
844   int progress;
845 
846   for(i=0; i<lemp->nsymbol; i++){
847     lemp->symbols[i]->lambda = LEMON_FALSE;
848   }
849   for(i=lemp->nterminal; i<lemp->nsymbol; i++){
850     lemp->symbols[i]->firstset = SetNew();
851   }
852 
853   /* First compute all lambdas */
854   do{
855     progress = 0;
856     for(rp=lemp->rule; rp; rp=rp->next){
857       if( rp->lhs->lambda ) continue;
858       for(i=0; i<rp->nrhs; i++){
859         struct symbol *sp = rp->rhs[i];
860         assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE );
861         if( sp->lambda==LEMON_FALSE ) break;
862       }
863       if( i==rp->nrhs ){
864         rp->lhs->lambda = LEMON_TRUE;
865         progress = 1;
866       }
867     }
868   }while( progress );
869 
870   /* Now compute all first sets */
871   do{
872     struct symbol *s1, *s2;
873     progress = 0;
874     for(rp=lemp->rule; rp; rp=rp->next){
875       s1 = rp->lhs;
876       for(i=0; i<rp->nrhs; i++){
877         s2 = rp->rhs[i];
878         if( s2->type==TERMINAL ){
879           progress += SetAdd(s1->firstset,s2->index);
880           break;
881         }else if( s2->type==MULTITERMINAL ){
882           for(j=0; j<s2->nsubsym; j++){
883             progress += SetAdd(s1->firstset,s2->subsym[j]->index);
884           }
885           break;
886         }else if( s1==s2 ){
887           if( s1->lambda==LEMON_FALSE ) break;
888         }else{
889           progress += SetUnion(s1->firstset,s2->firstset);
890           if( s2->lambda==LEMON_FALSE ) break;
891         }
892       }
893     }
894   }while( progress );
895   return;
896 }
897 
898 /* Compute all LR(0) states for the grammar.  Links
899 ** are added to between some states so that the LR(1) follow sets
900 ** can be computed later.
901 */
902 PRIVATE struct state *getstate(struct lemon *);  /* forward reference */
FindStates(struct lemon * lemp)903 void FindStates(struct lemon *lemp)
904 {
905   struct symbol *sp;
906   struct rule *rp;
907 
908   Configlist_init();
909 
910   /* Find the start symbol */
911   if( lemp->start ){
912     sp = Symbol_find(lemp->start);
913     if( sp==0 ){
914       ErrorMsg(lemp->filename,0,
915         "The specified start symbol \"%s\" is not "
916         "in a nonterminal of the grammar.  \"%s\" will be used as the start "
917         "symbol instead.",lemp->start,lemp->startRule->lhs->name);
918       lemp->errorcnt++;
919       sp = lemp->startRule->lhs;
920     }
921   }else{
922     sp = lemp->startRule->lhs;
923   }
924 
925   /* Make sure the start symbol doesn't occur on the right-hand side of
926   ** any rule.  Report an error if it does.  (YACC would generate a new
927   ** start symbol in this case.) */
928   for(rp=lemp->rule; rp; rp=rp->next){
929     int i;
930     for(i=0; i<rp->nrhs; i++){
931       if( rp->rhs[i]==sp ){   /* FIX ME:  Deal with multiterminals */
932         ErrorMsg(lemp->filename,0,
933           "The start symbol \"%s\" occurs on the "
934           "right-hand side of a rule. This will result in a parser which "
935           "does not work properly.",sp->name);
936         lemp->errorcnt++;
937       }
938     }
939   }
940 
941   /* The basis configuration set for the first state
942   ** is all rules which have the start symbol as their
943   ** left-hand side */
944   for(rp=sp->rule; rp; rp=rp->nextlhs){
945     struct config *newcfp;
946     rp->lhsStart = 1;
947     newcfp = Configlist_addbasis(rp,0);
948     SetAdd(newcfp->fws,0);
949   }
950 
951   /* Compute the first state.  All other states will be
952   ** computed automatically during the computation of the first one.
953   ** The returned pointer to the first state is not used. */
954   (void)getstate(lemp);
955   return;
956 }
957 
958 /* Return a pointer to a state which is described by the configuration
959 ** list which has been built from calls to Configlist_add.
960 */
961 PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */
getstate(struct lemon * lemp)962 PRIVATE struct state *getstate(struct lemon *lemp)
963 {
964   struct config *cfp, *bp;
965   struct state *stp;
966 
967   /* Extract the sorted basis of the new state.  The basis was constructed
968   ** by prior calls to "Configlist_addbasis()". */
969   Configlist_sortbasis();
970   bp = Configlist_basis();
971 
972   /* Get a state with the same basis */
973   stp = State_find(bp);
974   if( stp ){
975     /* A state with the same basis already exists!  Copy all the follow-set
976     ** propagation links from the state under construction into the
977     ** preexisting state, then return a pointer to the preexisting state */
978     struct config *x, *y;
979     for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
980       Plink_copy(&y->bplp,x->bplp);
981       Plink_delete(x->fplp);
982       x->fplp = x->bplp = 0;
983     }
984     cfp = Configlist_return();
985     Configlist_eat(cfp);
986   }else{
987     /* This really is a new state.  Construct all the details */
988     Configlist_closure(lemp);    /* Compute the configuration closure */
989     Configlist_sort();           /* Sort the configuration closure */
990     cfp = Configlist_return();   /* Get a pointer to the config list */
991     stp = State_new();           /* A new state structure */
992     MemoryCheck(stp);
993     stp->bp = bp;                /* Remember the configuration basis */
994     stp->cfp = cfp;              /* Remember the configuration closure */
995     stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
996     stp->ap = 0;                 /* No actions, yet. */
997 #ifndef NDEBUG
998     int ret =
999 #endif
1000     State_insert(stp,stp->bp);   /* Add to the state table */
1001 #ifndef NDEBUG
1002     assert(ret == 1);  /* CSA hint: stp did not leak, it has escaped. */
1003 #endif
1004     buildshifts(lemp,stp);       /* Recursively compute successor states */
1005   }
1006   return stp;
1007 }
1008 
1009 /*
1010 ** Return true if two symbols are the same.
1011 */
same_symbol(struct symbol * a,struct symbol * b)1012 int same_symbol(struct symbol *a, struct symbol *b)
1013 {
1014   int i;
1015   if( a==b ) return 1;
1016   if( a->type!=MULTITERMINAL ) return 0;
1017   if( b->type!=MULTITERMINAL ) return 0;
1018   if( a->nsubsym!=b->nsubsym ) return 0;
1019   for(i=0; i<a->nsubsym; i++){
1020     if( a->subsym[i]!=b->subsym[i] ) return 0;
1021   }
1022   return 1;
1023 }
1024 
1025 /* Construct all successor states to the given state.  A "successor"
1026 ** state is any state which can be reached by a shift action.
1027 */
buildshifts(struct lemon * lemp,struct state * stp)1028 PRIVATE void buildshifts(struct lemon *lemp, struct state *stp)
1029 {
1030   struct config *cfp;  /* For looping thru the config closure of "stp" */
1031   struct config *bcfp; /* For the inner loop on config closure of "stp" */
1032   struct config *newcfg;  /* */
1033   struct symbol *sp;   /* Symbol following the dot in configuration "cfp" */
1034   struct symbol *bsp;  /* Symbol following the dot in configuration "bcfp" */
1035   struct state *newstp; /* A pointer to a successor state */
1036 
1037   /* Each configuration becomes complete after it contibutes to a successor
1038   ** state.  Initially, all configurations are incomplete */
1039   for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
1040 
1041   /* Loop through all configurations of the state "stp" */
1042   for(cfp=stp->cfp; cfp; cfp=cfp->next){
1043     if( cfp->status==COMPLETE ) continue;    /* Already used by inner loop */
1044     if( cfp->dot>=cfp->rp->nrhs ) continue;  /* Can't shift this config */
1045     Configlist_reset();                      /* Reset the new config set */
1046     sp = cfp->rp->rhs[cfp->dot];             /* Symbol after the dot */
1047 
1048     /* For every configuration in the state "stp" which has the symbol "sp"
1049     ** following its dot, add the same configuration to the basis set under
1050     ** construction but with the dot shifted one symbol to the right. */
1051     for(bcfp=cfp; bcfp; bcfp=bcfp->next){
1052       if( bcfp->status==COMPLETE ) continue;    /* Already used */
1053       if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
1054       bsp = bcfp->rp->rhs[bcfp->dot];           /* Get symbol after dot */
1055       if( !same_symbol(bsp,sp) ) continue;      /* Must be same as for "cfp" */
1056       bcfp->status = COMPLETE;                  /* Mark this config as used */
1057       newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
1058       Plink_add(&newcfg->bplp,bcfp);
1059     }
1060 
1061     /* Get a pointer to the state described by the basis configuration set
1062     ** constructed in the preceding loop */
1063     newstp = getstate(lemp);
1064 
1065     /* The state "newstp" is reached from the state "stp" by a shift action
1066     ** on the symbol "sp" */
1067     if( sp->type==MULTITERMINAL ){
1068       int i;
1069       for(i=0; i<sp->nsubsym; i++){
1070         Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
1071       }
1072     }else{
1073       Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
1074     }
1075   }
1076 }
1077 
1078 /*
1079 ** Construct the propagation links
1080 */
FindLinks(struct lemon * lemp)1081 void FindLinks(struct lemon *lemp)
1082 {
1083   int i;
1084   struct config *cfp, *other;
1085   struct state *stp;
1086   struct plink *plp;
1087 
1088   /* Housekeeping detail:
1089   ** Add to every propagate link a pointer back to the state to
1090   ** which the link is attached. */
1091   for(i=0; i<lemp->nstate; i++){
1092     stp = lemp->sorted[i];
1093     assert(stp);    /* Hint for CSA */
1094     for(cfp=stp->cfp; cfp; cfp=cfp->next){
1095       cfp->stp = stp;
1096     }
1097   }
1098 
1099   /* Convert all backlinks into forward links.  Only the forward
1100   ** links are used in the follow-set computation. */
1101   for(i=0; i<lemp->nstate; i++){
1102     stp = lemp->sorted[i];
1103     for(cfp=stp->cfp; cfp; cfp=cfp->next){
1104       for(plp=cfp->bplp; plp; plp=plp->next){
1105         other = plp->cfp;
1106         Plink_add(&other->fplp,cfp);
1107       }
1108     }
1109   }
1110 }
1111 
1112 /* Compute all followsets.
1113 **
1114 ** A followset is the set of all symbols which can come immediately
1115 ** after a configuration.
1116 */
FindFollowSets(struct lemon * lemp)1117 void FindFollowSets(struct lemon *lemp)
1118 {
1119   int i;
1120   struct config *cfp;
1121   struct plink *plp;
1122   int progress;
1123   int change;
1124 
1125   for(i=0; i<lemp->nstate; i++){
1126     for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1127       cfp->status = INCOMPLETE;
1128     }
1129   }
1130 
1131   do{
1132     progress = 0;
1133     for(i=0; i<lemp->nstate; i++){
1134       for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
1135         if( cfp->status==COMPLETE ) continue;
1136         for(plp=cfp->fplp; plp; plp=plp->next){
1137           change = SetUnion(plp->cfp->fws,cfp->fws);
1138           if( change ){
1139             plp->cfp->status = INCOMPLETE;
1140             progress = 1;
1141           }
1142         }
1143         cfp->status = COMPLETE;
1144       }
1145     }
1146   }while( progress );
1147 }
1148 
1149 static int resolve_conflict(struct action *,struct action *);
1150 
1151 /* Compute the reduce actions, and resolve conflicts.
1152 */
FindActions(struct lemon * lemp)1153 void FindActions(struct lemon *lemp)
1154 {
1155   int i,j;
1156   struct config *cfp;
1157   struct state *stp;
1158   struct symbol *sp;
1159   struct rule *rp;
1160 
1161   /* Add all of the reduce actions
1162   ** A reduce action is added for each element of the followset of
1163   ** a configuration which has its dot at the extreme right.
1164   */
1165   for(i=0; i<lemp->nstate; i++){   /* Loop over all states */
1166     stp = lemp->sorted[i];
1167     for(cfp=stp->cfp; cfp; cfp=cfp->next){  /* Loop over all configurations */
1168       if( cfp->rp->nrhs==cfp->dot ){        /* Is dot at extreme right? */
1169         for(j=0; j<lemp->nterminal; j++){
1170           if( SetFind(cfp->fws,j) ){
1171             /* Add a reduce action to the state "stp" which will reduce by the
1172             ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
1173             Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
1174           }
1175         }
1176       }
1177     }
1178   }
1179 
1180   /* Add the accepting token */
1181   if( lemp->start ){
1182     sp = Symbol_find(lemp->start);
1183     if( sp==0 ) sp = lemp->startRule->lhs;
1184   }else{
1185     sp = lemp->startRule->lhs;
1186   }
1187   /* Add to the first state (which is always the starting state of the
1188   ** finite state machine) an action to ACCEPT if the lookahead is the
1189   ** start nonterminal.  */
1190   Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
1191 
1192   /* Resolve conflicts */
1193   for(i=0; i<lemp->nstate; i++){
1194     struct action *ap, *nap;
1195     stp = lemp->sorted[i];
1196     /* assert( stp->ap ); */
1197     stp->ap = Action_sort(stp->ap);
1198     for(ap=stp->ap; ap && ap->next; ap=ap->next){
1199       for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
1200          /* The two actions "ap" and "nap" have the same lookahead.
1201          ** Figure out which one should be used */
1202          lemp->nconflict += resolve_conflict(ap,nap);
1203       }
1204     }
1205   }
1206 
1207   /* Report an error for each rule that can never be reduced. */
1208   for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
1209   for(i=0; i<lemp->nstate; i++){
1210     struct action *ap;
1211     for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
1212       if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
1213     }
1214   }
1215   for(rp=lemp->rule; rp; rp=rp->next){
1216     if( rp->canReduce ) continue;
1217     ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
1218     lemp->errorcnt++;
1219   }
1220 }
1221 
1222 /* Resolve a conflict between the two given actions.  If the
1223 ** conflict can't be resolved, return non-zero.
1224 **
1225 ** NO LONGER TRUE:
1226 **   To resolve a conflict, first look to see if either action
1227 **   is on an error rule.  In that case, take the action which
1228 **   is not associated with the error rule.  If neither or both
1229 **   actions are associated with an error rule, then try to
1230 **   use precedence to resolve the conflict.
1231 **
1232 ** If either action is a SHIFT, then it must be apx.  This
1233 ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1234 */
resolve_conflict(struct action * apx,struct action * apy)1235 static int resolve_conflict(
1236   struct action *apx,
1237   struct action *apy
1238 ){
1239   struct symbol *spx, *spy;
1240   int errcnt = 0;
1241   assert( apx->sp==apy->sp );  /* Otherwise there would be no conflict */
1242   if( apx->type==SHIFT && apy->type==SHIFT ){
1243     apy->type = SSCONFLICT;
1244     errcnt++;
1245   }
1246   if( apx->type==SHIFT && apy->type==REDUCE ){
1247     spx = apx->sp;
1248     spy = apy->x.rp->precsym;
1249     if( spy==0 || spx->prec<0 || spy->prec<0 ){
1250       /* Not enough precedence information. */
1251       apy->type = SRCONFLICT;
1252       errcnt++;
1253     }else if( spx->prec>spy->prec ){    /* higher precedence wins */
1254       apy->type = RD_RESOLVED;
1255     }else if( spx->prec<spy->prec ){
1256       apx->type = SH_RESOLVED;
1257     }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1258       apy->type = RD_RESOLVED;                             /* associativity */
1259     }else if( spx->prec==spy->prec && spx->assoc==LEFT ){  /* to break tie */
1260       apx->type = SH_RESOLVED;
1261     }else{
1262       assert( spx->prec==spy->prec && spx->assoc==NONE );
1263       apx->type = ERROR;
1264     }
1265   }else if( apx->type==REDUCE && apy->type==REDUCE ){
1266     spx = apx->x.rp->precsym;
1267     spy = apy->x.rp->precsym;
1268     if( spx==0 || spy==0 || spx->prec<0 ||
1269     spy->prec<0 || spx->prec==spy->prec ){
1270       apy->type = RRCONFLICT;
1271       errcnt++;
1272     }else if( spx->prec>spy->prec ){
1273       apy->type = RD_RESOLVED;
1274     }else if( spx->prec<spy->prec ){
1275       apx->type = RD_RESOLVED;
1276     }
1277   }else{
1278     assert(
1279       apx->type==SH_RESOLVED ||
1280       apx->type==RD_RESOLVED ||
1281       apx->type==SSCONFLICT ||
1282       apx->type==SRCONFLICT ||
1283       apx->type==RRCONFLICT ||
1284       apy->type==SH_RESOLVED ||
1285       apy->type==RD_RESOLVED ||
1286       apy->type==SSCONFLICT ||
1287       apy->type==SRCONFLICT ||
1288       apy->type==RRCONFLICT
1289     );
1290     /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1291     ** REDUCEs on the list.  If we reach this point it must be because
1292     ** the parser conflict had already been resolved. */
1293   }
1294   return errcnt;
1295 }
1296 /********************* From the file "configlist.c" *************************/
1297 /*
1298 ** Routines to processing a configuration list and building a state
1299 ** in the LEMON parser generator.
1300 */
1301 
1302 static struct config *freelist = 0;      /* List of free configurations */
1303 static struct config *current = 0;       /* Top of list of configurations */
1304 static struct config **currentend = 0;   /* Last on list of configs */
1305 static struct config *basis = 0;         /* Top of list of basis configs */
1306 static struct config **basisend = 0;     /* End of list of basis configs */
1307 
1308 /* Return a pointer to a new configuration */
newconfig(void)1309 PRIVATE struct config *newconfig(void){
1310   struct config *newcfg;
1311   if( freelist==0 ){
1312     int i;
1313     int amt = 3;
1314     freelist = (struct config *)calloc( amt, sizeof(struct config) );
1315     if( freelist==0 ){
1316       fprintf(stderr,"Unable to allocate memory for a new configuration.");
1317       exit(1);
1318     }
1319     for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1320     freelist[amt-1].next = 0;
1321   }
1322   newcfg = freelist;
1323   freelist = freelist->next;
1324   return newcfg;
1325 }
1326 
1327 /* The configuration "old" is no longer used */
deleteconfig(struct config * old)1328 PRIVATE void deleteconfig(struct config *old)
1329 {
1330   old->next = freelist;
1331   freelist = old;
1332 }
1333 
1334 /* Initialized the configuration list builder */
Configlist_init(void)1335 void Configlist_init(void){
1336   current = 0;
1337   currentend = &current;
1338   basis = 0;
1339   basisend = &basis;
1340   Configtable_init();
1341   return;
1342 }
1343 
1344 /* Initialized the configuration list builder */
Configlist_reset(void)1345 void Configlist_reset(void){
1346   current = 0;
1347   currentend = &current;
1348   basis = 0;
1349   basisend = &basis;
1350   Configtable_clear(0);
1351   return;
1352 }
1353 
1354 /* Add another configuration to the configuration list */
Configlist_add(struct rule * rp,int dot)1355 struct config *Configlist_add(
1356   struct rule *rp,    /* The rule */
1357   int dot             /* Index into the RHS of the rule where the dot goes */
1358 ){
1359   struct config *cfp, model;
1360 
1361   assert( currentend!=0 );
1362   model.rp = rp;
1363   model.dot = dot;
1364   cfp = Configtable_find(&model);
1365   if( cfp==0 ){
1366     cfp = newconfig();
1367     cfp->rp = rp;
1368     cfp->dot = dot;
1369     cfp->fws = SetNew();
1370     cfp->stp = 0;
1371     cfp->fplp = cfp->bplp = 0;
1372     cfp->next = 0;
1373     cfp->bp = 0;
1374     *currentend = cfp;
1375     currentend = &cfp->next;
1376     Configtable_insert(cfp);
1377   }
1378   return cfp;
1379 }
1380 
1381 /* Add a basis configuration to the configuration list */
Configlist_addbasis(struct rule * rp,int dot)1382 struct config *Configlist_addbasis(struct rule *rp, int dot)
1383 {
1384   struct config *cfp, model;
1385 
1386   assert( basisend!=0 );
1387   assert( currentend!=0 );
1388   model.rp = rp;
1389   model.dot = dot;
1390   cfp = Configtable_find(&model);
1391   if( cfp==0 ){
1392     cfp = newconfig();
1393     cfp->rp = rp;
1394     cfp->dot = dot;
1395     cfp->fws = SetNew();
1396     cfp->stp = 0;
1397     cfp->fplp = cfp->bplp = 0;
1398     cfp->next = 0;
1399     cfp->bp = 0;
1400     *currentend = cfp;
1401     currentend = &cfp->next;
1402     *basisend = cfp;
1403     basisend = &cfp->bp;
1404     Configtable_insert(cfp);
1405   }
1406   return cfp;
1407 }
1408 
1409 /* Compute the closure of the configuration list */
Configlist_closure(struct lemon * lemp)1410 void Configlist_closure(struct lemon *lemp)
1411 {
1412   struct config *cfp, *newcfp;
1413   struct rule *rp, *newrp;
1414   struct symbol *sp, *xsp;
1415   int i, dot;
1416 
1417   assert( currentend!=0 );
1418   for(cfp=current; cfp; cfp=cfp->next){
1419     rp = cfp->rp;
1420     dot = cfp->dot;
1421     if( dot>=rp->nrhs ) continue;
1422     sp = rp->rhs[dot];
1423     if( sp->type==NONTERMINAL ){
1424       if( sp->rule==0 && sp!=lemp->errsym ){
1425         ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1426           sp->name);
1427         lemp->errorcnt++;
1428       }
1429       for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1430         newcfp = Configlist_add(newrp,0);
1431         for(i=dot+1; i<rp->nrhs; i++){
1432           xsp = rp->rhs[i];
1433           if( xsp->type==TERMINAL ){
1434             SetAdd(newcfp->fws,xsp->index);
1435             break;
1436           }else if( xsp->type==MULTITERMINAL ){
1437             int k;
1438             for(k=0; k<xsp->nsubsym; k++){
1439               SetAdd(newcfp->fws, xsp->subsym[k]->index);
1440             }
1441             break;
1442           }else{
1443             SetUnion(newcfp->fws,xsp->firstset);
1444             if( xsp->lambda==LEMON_FALSE ) break;
1445           }
1446         }
1447         if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1448       }
1449     }
1450   }
1451   return;
1452 }
1453 
1454 /* Sort the configuration list */
Configlist_sort(void)1455 void Configlist_sort(void){
1456   current = (struct config*)msort((char*)current,(char**)&(current->next),
1457                                   Configcmp);
1458   currentend = 0;
1459   return;
1460 }
1461 
1462 /* Sort the basis configuration list */
Configlist_sortbasis(void)1463 void Configlist_sortbasis(void){
1464   basis = (struct config*)msort((char*)current,(char**)&(current->bp),
1465                                 Configcmp);
1466   basisend = 0;
1467   return;
1468 }
1469 
1470 /* Return a pointer to the head of the configuration list and
1471 ** reset the list */
Configlist_return(void)1472 struct config *Configlist_return(void){
1473   struct config *old;
1474   old = current;
1475   current = 0;
1476   currentend = 0;
1477   return old;
1478 }
1479 
1480 /* Return a pointer to the head of the configuration list and
1481 ** reset the list */
Configlist_basis(void)1482 struct config *Configlist_basis(void){
1483   struct config *old;
1484   old = basis;
1485   basis = 0;
1486   basisend = 0;
1487   return old;
1488 }
1489 
1490 /* Free all elements of the given configuration list */
Configlist_eat(struct config * cfp)1491 void Configlist_eat(struct config *cfp)
1492 {
1493   struct config *nextcfp;
1494   for(; cfp; cfp=nextcfp){
1495     nextcfp = cfp->next;
1496     assert( cfp->fplp==0 );
1497     assert( cfp->bplp==0 );
1498     if( cfp->fws ) SetFree(cfp->fws);
1499     deleteconfig(cfp);
1500   }
1501   return;
1502 }
1503 /***************** From the file "error.c" *********************************/
1504 /*
1505 ** Code for printing error message.
1506 */
1507 
ErrorMsg(const char * filename,int lineno,const char * format,...)1508 void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1509   va_list ap;
1510   fprintf(stderr, "%s:%d: ", filename, lineno);
1511   va_start(ap, format);
1512   vfprintf(stderr,format,ap);
1513   va_end(ap);
1514   fprintf(stderr, "\n");
1515 }
1516 /**************** From the file "main.c" ************************************/
1517 /*
1518 ** Main program file for the LEMON parser generator.
1519 */
1520 
1521 /* Report an out-of-memory condition and abort.  This function
1522 ** is used mostly by the "MemoryCheck" macro in struct.h
1523 */
memory_error(void)1524 void memory_error(void){
1525   fprintf(stderr,"Out of memory.  Aborting...\n");
1526   exit(1);
1527 }
1528 
1529 static int nDefine = 0;      /* Number of -D options on the command line */
1530 static char **azDefine = 0;  /* Name of the -D macros */
1531 
1532 /* This routine is called with the argument to each -D command-line option.
1533 ** Add the macro defined to the azDefine array.
1534 */
handle_D_option(char * z)1535 static void handle_D_option(char *z){
1536   char **paz;
1537   nDefine++;
1538   azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine);
1539   if( azDefine==0 ){
1540     fprintf(stderr,"out of memory\n");
1541     exit(1);
1542   }
1543   paz = &azDefine[nDefine-1];
1544   *paz = (char *) malloc( lemonStrlen(z)+1 );
1545   if( *paz==0 ){
1546     fprintf(stderr,"out of memory\n");
1547     exit(1);
1548   }
1549   lemon_strcpy(*paz, z);
1550   for(z=*paz; *z && *z!='='; z++){}
1551   *z = 0;
1552 }
1553 
1554 /* Rember the name of the output directory
1555 */
1556 static char *outputDir = NULL;
handle_d_option(char * z)1557 static void handle_d_option(char *z){
1558   outputDir = (char *) malloc( lemonStrlen(z)+1 );
1559   if( outputDir==0 ){
1560     fprintf(stderr,"out of memory\n");
1561     exit(1);
1562   }
1563   lemon_strcpy(outputDir, z);
1564 }
1565 
1566 static char *user_templatename = NULL;
handle_T_option(char * z)1567 static void handle_T_option(char *z){
1568   user_templatename = (char *) malloc( lemonStrlen(z)+1 );
1569   if( user_templatename==0 ){
1570     memory_error();
1571   }
1572   lemon_strcpy(user_templatename, z);
1573 }
1574 
1575 /* Merge together to lists of rules ordered by rule.iRule */
Rule_merge(struct rule * pA,struct rule * pB)1576 static struct rule *Rule_merge(struct rule *pA, struct rule *pB){
1577   struct rule *pFirst = 0;
1578   struct rule **ppPrev = &pFirst;
1579   while( pA && pB ){
1580     if( pA->iRule<pB->iRule ){
1581       *ppPrev = pA;
1582       ppPrev = &pA->next;
1583       pA = pA->next;
1584     }else{
1585       *ppPrev = pB;
1586       ppPrev = &pB->next;
1587       pB = pB->next;
1588     }
1589   }
1590   if( pA ){
1591     *ppPrev = pA;
1592   }else{
1593     *ppPrev = pB;
1594   }
1595   return pFirst;
1596 }
1597 
1598 /*
1599 ** Sort a list of rules in order of increasing iRule value
1600 */
Rule_sort(struct rule * rp)1601 static struct rule *Rule_sort(struct rule *rp){
1602   unsigned int i;
1603   struct rule *pNext;
1604   struct rule *x[32];
1605   memset(x, 0, sizeof(x));
1606   while( rp ){
1607     pNext = rp->next;
1608     rp->next = 0;
1609     for(i=0; i<sizeof(x)/sizeof(x[0])-1 && x[i]; i++){
1610       rp = Rule_merge(x[i], rp);
1611       x[i] = 0;
1612     }
1613     x[i] = rp;
1614     rp = pNext;
1615   }
1616   rp = 0;
1617   for(i=0; i<sizeof(x)/sizeof(x[0]); i++){
1618     rp = Rule_merge(x[i], rp);
1619   }
1620   return rp;
1621 }
1622 
1623 /* forward reference */
1624 static const char *minimum_size_type(int lwr, int upr, int *pnByte);
1625 
1626 /* Print a single line of the "Parser Stats" output
1627 */
stats_line(const char * zLabel,int iValue)1628 static void stats_line(const char *zLabel, int iValue){
1629   int nLabel = lemonStrlen(zLabel);
1630   printf("  %s%.*s %5d\n", zLabel,
1631          35-nLabel, "................................",
1632          iValue);
1633 }
1634 
1635 /* The main program.  Parse the command line and do it... */
main(int argc,char ** argv)1636 int main(int argc, char **argv){
1637   static int version = 0;
1638   static int rpflag = 0;
1639   static int basisflag = 0;
1640   static int compress = 0;
1641   static int quiet = 0;
1642   static int statistics = 0;
1643   static int mhflag = 0;
1644   static int nolinenosflag = 0;
1645   static int noResort = 0;
1646   static int sqlFlag = 0;
1647   static int printPP = 0;
1648 
1649   static struct s_options options[] = {
1650     {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1651     {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1652     {OPT_FSTR, "d", (char*)&handle_d_option, "Output directory.  Default '.'"},
1653     {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
1654     {OPT_FLAG, "E", (char*)&printPP, "Print input file after preprocessing."},
1655     {OPT_FSTR, "f", 0, "Ignored.  (Placeholder for -f compiler options.)"},
1656     {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1657     {OPT_FSTR, "I", 0, "Ignored.  (Placeholder for '-I' compiler options.)"},
1658     {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."},
1659     {OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."},
1660     {OPT_FSTR, "O", 0, "Ignored.  (Placeholder for '-O' compiler options.)"},
1661     {OPT_FLAG, "p", (char*)&showPrecedenceConflict,
1662                     "Show conflicts resolved by precedence rules"},
1663     {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1664     {OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"},
1665     {OPT_FLAG, "s", (char*)&statistics,
1666                                    "Print parser stats to standard output."},
1667     {OPT_FLAG, "S", (char*)&sqlFlag,
1668                     "Generate the *.sql file describing the parser tables."},
1669     {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1670     {OPT_FSTR, "T", (char*)handle_T_option, "Specify a template file."},
1671     {OPT_FSTR, "W", 0, "Ignored.  (Placeholder for '-W' compiler options.)"},
1672     {OPT_FLAG,0,0,0}
1673   };
1674   int i;
1675   int exitcode;
1676   struct lemon lem;
1677   struct rule *rp;
1678 
1679   (void)argc;
1680   OptInit(argv,options,stderr);
1681   if( version ){
1682      printf("Lemon version 1.0\n");
1683      exit(0);
1684   }
1685   if( OptNArgs()!=1 ){
1686     fprintf(stderr,"Exactly one filename argument is required.\n");
1687     exit(1);
1688   }
1689   memset(&lem, 0, sizeof(lem));
1690   lem.errorcnt = 0;
1691 
1692   /* Initialize the machine */
1693   Strsafe_init();
1694   Symbol_init();
1695   State_init();
1696   lem.argv0 = argv[0];
1697   lem.filename = OptArg(0);
1698   lem.basisflag = basisflag;
1699   lem.nolinenosflag = nolinenosflag;
1700   lem.printPreprocessed = printPP;
1701   Symbol_new("$");
1702 
1703   /* Parse the input file */
1704   Parse(&lem);
1705   if( lem.printPreprocessed || lem.errorcnt ) exit(lem.errorcnt);
1706   assert(lem.rule);  /* Hint for CSA (no errors => rule found). */
1707   if( lem.nrule==0 ){
1708     fprintf(stderr,"Empty grammar.\n");
1709     exit(1);
1710   }
1711   lem.errsym = Symbol_find("error");
1712 
1713   /* Count and index the symbols of the grammar */
1714   Symbol_new("{default}");
1715   lem.nsymbol = Symbol_count();
1716   lem.symbols = Symbol_arrayof();
1717   for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1718   qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp);
1719   for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
1720   while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; }
1721   assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 );
1722   lem.nsymbol = i - 1;
1723   for(i=1; ISUPPER(lem.symbols[i]->name[0]); i++);
1724   lem.nterminal = i;
1725 
1726   /* Assign sequential rule numbers.  Start with 0.  Put rules that have no
1727   ** reduce action C-code associated with them last, so that the switch()
1728   ** statement that selects reduction actions will have a smaller jump table.
1729   */
1730   for(i=0, rp=lem.rule; rp; rp=rp->next){
1731     rp->iRule = rp->code ? i++ : -1;
1732   }
1733   lem.nruleWithAction = i;
1734   for(rp=lem.rule; rp; rp=rp->next){
1735     if( rp->iRule<0 ) rp->iRule = i++;
1736   }
1737   lem.startRule = lem.rule;
1738   lem.rule = Rule_sort(lem.rule);
1739 
1740   /* Generate a reprint of the grammar, if requested on the command line */
1741   if( rpflag ){
1742     Reprint(&lem);
1743   }else{
1744     /* Initialize the size for all follow and first sets */
1745     SetSize(lem.nterminal+1);
1746 
1747     /* Find the precedence for every production rule (that has one) */
1748     FindRulePrecedences(&lem);
1749 
1750     /* Compute the lambda-nonterminals and the first-sets for every
1751     ** nonterminal */
1752     FindFirstSets(&lem);
1753 
1754     /* Compute all LR(0) states.  Also record follow-set propagation
1755     ** links so that the follow-set can be computed later */
1756     lem.nstate = 0;
1757     FindStates(&lem);
1758     lem.sorted = State_arrayof();
1759 
1760     /* Tie up loose ends on the propagation links */
1761     FindLinks(&lem);
1762 
1763     /* Compute the follow set of every reducible configuration */
1764     FindFollowSets(&lem);
1765 
1766     /* Compute the action tables */
1767     FindActions(&lem);
1768 
1769     /* Compress the action tables */
1770     if( compress==0 ) CompressTables(&lem);
1771 
1772     /* Reorder and renumber the states so that states with fewer choices
1773     ** occur at the end.  This is an optimization that helps make the
1774     ** generated parser tables smaller. */
1775     if( noResort==0 ) ResortStates(&lem);
1776 
1777     /* Generate a report of the parser generated.  (the "y.output" file) */
1778     if( !quiet ) ReportOutput(&lem);
1779 
1780     /* Generate the source code for the parser */
1781     ReportTable(&lem, mhflag, sqlFlag);
1782 
1783     /* Produce a header file for use by the scanner.  (This step is
1784     ** omitted if the "-m" option is used because makeheaders will
1785     ** generate the file for us.) */
1786     if( !mhflag ) ReportHeader(&lem);
1787   }
1788   if( statistics ){
1789     printf("Parser statistics:\n");
1790     stats_line("terminal symbols", lem.nterminal);
1791     stats_line("non-terminal symbols", lem.nsymbol - lem.nterminal);
1792     stats_line("total symbols", lem.nsymbol);
1793     stats_line("rules", lem.nrule);
1794     stats_line("states", lem.nxstate);
1795     stats_line("conflicts", lem.nconflict);
1796     stats_line("action table entries", lem.nactiontab);
1797     stats_line("lookahead table entries", lem.nlookaheadtab);
1798     stats_line("total table size (bytes)", lem.tablesize);
1799   }
1800   if( lem.nconflict > 0 ){
1801     fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1802   }
1803 
1804   /* return 0 on success, 1 on failure. */
1805   exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0;
1806   exit(exitcode);
1807   return (exitcode);
1808 }
1809 /******************** From the file "msort.c" *******************************/
1810 /*
1811 ** A generic merge-sort program.
1812 **
1813 ** USAGE:
1814 ** Let "ptr" be a pointer to some structure which is at the head of
1815 ** a null-terminated list.  Then to sort the list call:
1816 **
1817 **     ptr = msort(ptr,&(ptr->next),cmpfnc);
1818 **
1819 ** In the above, "cmpfnc" is a pointer to a function which compares
1820 ** two instances of the structure and returns an integer, as in
1821 ** strcmp.  The second argument is a pointer to the pointer to the
1822 ** second element of the linked list.  This address is used to compute
1823 ** the offset to the "next" field within the structure.  The offset to
1824 ** the "next" field must be constant for all structures in the list.
1825 **
1826 ** The function returns a new pointer which is the head of the list
1827 ** after sorting.
1828 **
1829 ** ALGORITHM:
1830 ** Merge-sort.
1831 */
1832 
1833 /*
1834 ** Return a pointer to the next structure in the linked list.
1835 */
1836 #define NEXT(A) (*(char**)(((char*)A)+offset))
1837 
1838 /*
1839 ** Inputs:
1840 **   a:       A sorted, null-terminated linked list.  (May be null).
1841 **   b:       A sorted, null-terminated linked list.  (May be null).
1842 **   cmp:     A pointer to the comparison function.
1843 **   offset:  Offset in the structure to the "next" field.
1844 **
1845 ** Return Value:
1846 **   A pointer to the head of a sorted list containing the elements
1847 **   of both a and b.
1848 **
1849 ** Side effects:
1850 **   The "next" pointers for elements in the lists a and b are
1851 **   changed.
1852 */
merge(char * a,char * b,int (* cmp)(const char *,const char *),int offset)1853 static char *merge(
1854   char *a,
1855   char *b,
1856   int (*cmp)(const char*,const char*),
1857   int offset
1858 ){
1859   char *ptr, *head;
1860 
1861   if( a==0 ){
1862     head = b;
1863   }else if( b==0 ){
1864     head = a;
1865   }else{
1866     if( (*cmp)(a,b)<=0 ){
1867       ptr = a;
1868       a = NEXT(a);
1869     }else{
1870       ptr = b;
1871       b = NEXT(b);
1872     }
1873     head = ptr;
1874     while( a && b ){
1875       if( (*cmp)(a,b)<=0 ){
1876         NEXT(ptr) = a;
1877         ptr = a;
1878         a = NEXT(a);
1879       }else{
1880         NEXT(ptr) = b;
1881         ptr = b;
1882         b = NEXT(b);
1883       }
1884     }
1885     if( a ) NEXT(ptr) = a;
1886     else    NEXT(ptr) = b;
1887   }
1888   return head;
1889 }
1890 
1891 /*
1892 ** Inputs:
1893 **   list:      Pointer to a singly-linked list of structures.
1894 **   next:      Pointer to pointer to the second element of the list.
1895 **   cmp:       A comparison function.
1896 **
1897 ** Return Value:
1898 **   A pointer to the head of a sorted list containing the elements
1899 **   orginally in list.
1900 **
1901 ** Side effects:
1902 **   The "next" pointers for elements in list are changed.
1903 */
1904 #define LISTSIZE 30
msort(char * list,char ** next,int (* cmp)(const char *,const char *))1905 static char *msort(
1906   char *list,
1907   char **next,
1908   int (*cmp)(const char*,const char*)
1909 ){
1910   unsigned long offset;
1911   char *ep;
1912   char *set[LISTSIZE];
1913   int i;
1914   offset = (unsigned long)((char*)next - (char*)list);
1915   for(i=0; i<LISTSIZE; i++) set[i] = 0;
1916   while( list ){
1917     ep = list;
1918     list = NEXT(list);
1919     NEXT(ep) = 0;
1920     for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1921       ep = merge(ep,set[i],cmp,offset);
1922       set[i] = 0;
1923     }
1924     set[i] = ep;
1925   }
1926   ep = 0;
1927   for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset);
1928   return ep;
1929 }
1930 /************************ From the file "option.c" **************************/
1931 static char **g_argv;
1932 static struct s_options *op;
1933 static FILE *errstream;
1934 
1935 #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1936 
1937 /*
1938 ** Print the command line with a carrot pointing to the k-th character
1939 ** of the n-th field.
1940 */
errline(int n,int k,FILE * err)1941 static void errline(int n, int k, FILE *err)
1942 {
1943   int spcnt, i;
1944   if( g_argv[0] ) fprintf(err,"%s",g_argv[0]);
1945   spcnt = lemonStrlen(g_argv[0]) + 1;
1946   for(i=1; i<n && g_argv[i]; i++){
1947     fprintf(err," %s",g_argv[i]);
1948     spcnt += lemonStrlen(g_argv[i])+1;
1949   }
1950   spcnt += k;
1951   for(; g_argv[i]; i++) fprintf(err," %s",g_argv[i]);
1952   if( spcnt<20 ){
1953     fprintf(err,"\n%*s^-- here\n",spcnt,"");
1954   }else{
1955     fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1956   }
1957 }
1958 
1959 /*
1960 ** Return the index of the N-th non-switch argument.  Return -1
1961 ** if N is out of range.
1962 */
argindex(int n)1963 static int argindex(int n)
1964 {
1965   int i;
1966   int dashdash = 0;
1967   if( g_argv!=0 && *g_argv!=0 ){
1968     for(i=1; g_argv[i]; i++){
1969       if( dashdash || !ISOPT(g_argv[i]) ){
1970         if( n==0 ) return i;
1971         n--;
1972       }
1973       if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
1974     }
1975   }
1976   return -1;
1977 }
1978 
1979 static char emsg[] = "Command line syntax error: ";
1980 
1981 /*
1982 ** Process a flag command line argument.
1983 */
handleflags(int i,FILE * err)1984 static int handleflags(int i, FILE *err)
1985 {
1986   int v;
1987   int errcnt = 0;
1988   int j;
1989   for(j=0; op[j].label; j++){
1990     if( strncmp(&g_argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
1991   }
1992   v = g_argv[i][0]=='-' ? 1 : 0;
1993   if( op[j].label==0 ){
1994     if( err ){
1995       fprintf(err,"%sundefined option.\n",emsg);
1996       errline(i,1,err);
1997     }
1998     errcnt++;
1999   }else if( op[j].arg==0 ){
2000     /* Ignore this option */
2001   }else if( op[j].type==OPT_FLAG ){
2002     *((int*)op[j].arg) = v;
2003   }else if( op[j].type==OPT_FFLAG ){
2004     (*(void(*)(int))(op[j].arg))(v);
2005   }else if( op[j].type==OPT_FSTR ){
2006     (*(void(*)(char *))(op[j].arg))(&g_argv[i][2]);
2007   }else{
2008     if( err ){
2009       fprintf(err,"%smissing argument on switch.\n",emsg);
2010       errline(i,1,err);
2011     }
2012     errcnt++;
2013   }
2014   return errcnt;
2015 }
2016 
2017 /*
2018 ** Process a command line switch which has an argument.
2019 */
handleswitch(int i,FILE * err)2020 static int handleswitch(int i, FILE *err)
2021 {
2022   int lv = 0;
2023   double dv = 0.0;
2024   char *sv = 0, *end;
2025   char *cp;
2026   int j;
2027   int errcnt = 0;
2028   cp = strchr(g_argv[i],'=');
2029   assert( cp!=0 );
2030   *cp = 0;
2031   for(j=0; op[j].label; j++){
2032     if( strcmp(g_argv[i],op[j].label)==0 ) break;
2033   }
2034   *cp = '=';
2035   if( op[j].label==0 ){
2036     if( err ){
2037       fprintf(err,"%sundefined option.\n",emsg);
2038       errline(i,0,err);
2039     }
2040     errcnt++;
2041   }else{
2042     cp++;
2043     switch( op[j].type ){
2044       case OPT_FLAG:
2045       case OPT_FFLAG:
2046         if( err ){
2047           fprintf(err,"%soption requires an argument.\n",emsg);
2048           errline(i,0,err);
2049         }
2050         errcnt++;
2051         break;
2052       case OPT_DBL:
2053       case OPT_FDBL:
2054         dv = strtod(cp,&end);
2055         if( *end ){
2056           if( err ){
2057             fprintf(err,
2058                "%sillegal character in floating-point argument.\n",emsg);
2059             errline(i,(int)((char*)end-(char*)g_argv[i]),err);
2060           }
2061           errcnt++;
2062         }
2063         break;
2064       case OPT_INT:
2065       case OPT_FINT:
2066         lv = strtol(cp,&end,0);
2067         if( *end ){
2068           if( err ){
2069             fprintf(err,"%sillegal character in integer argument.\n",emsg);
2070             errline(i,(int)((char*)end-(char*)g_argv[i]),err);
2071           }
2072           errcnt++;
2073         }
2074         break;
2075       case OPT_STR:
2076       case OPT_FSTR:
2077         sv = cp;
2078         break;
2079     }
2080     switch( op[j].type ){
2081       case OPT_FLAG:
2082       case OPT_FFLAG:
2083         break;
2084       case OPT_DBL:
2085         *(double*)(op[j].arg) = dv;
2086         break;
2087       case OPT_FDBL:
2088         (*(void(*)(double))(op[j].arg))(dv);
2089         break;
2090       case OPT_INT:
2091         *(int*)(op[j].arg) = lv;
2092         break;
2093       case OPT_FINT:
2094         (*(void(*)(int))(op[j].arg))((int)lv);
2095         break;
2096       case OPT_STR:
2097         *(char**)(op[j].arg) = sv;
2098         break;
2099       case OPT_FSTR:
2100         (*(void(*)(char *))(op[j].arg))(sv);
2101         break;
2102     }
2103   }
2104   return errcnt;
2105 }
2106 
OptInit(char ** a,struct s_options * o,FILE * err)2107 int OptInit(char **a, struct s_options *o, FILE *err)
2108 {
2109   int errcnt = 0;
2110   g_argv = a;
2111   op = o;
2112   errstream = err;
2113   if( g_argv && *g_argv && op ){
2114     int i;
2115     for(i=1; g_argv[i]; i++){
2116       if( g_argv[i][0]=='+' || g_argv[i][0]=='-' ){
2117         errcnt += handleflags(i,err);
2118       }else if( strchr(g_argv[i],'=') ){
2119         errcnt += handleswitch(i,err);
2120       }
2121     }
2122   }
2123   if( errcnt>0 ){
2124     fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
2125     OptPrint();
2126     exit(1);
2127   }
2128   return 0;
2129 }
2130 
OptNArgs(void)2131 int OptNArgs(void){
2132   int cnt = 0;
2133   int dashdash = 0;
2134   int i;
2135   if( g_argv!=0 && g_argv[0]!=0 ){
2136     for(i=1; g_argv[i]; i++){
2137       if( dashdash || !ISOPT(g_argv[i]) ) cnt++;
2138       if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
2139     }
2140   }
2141   return cnt;
2142 }
2143 
OptArg(int n)2144 char *OptArg(int n)
2145 {
2146   int i;
2147   i = argindex(n);
2148   return i>=0 ? g_argv[i] : 0;
2149 }
2150 
OptErr(int n)2151 void OptErr(int n)
2152 {
2153   int i;
2154   i = argindex(n);
2155   if( i>=0 ) errline(i,0,errstream);
2156 }
2157 
OptPrint(void)2158 void OptPrint(void){
2159   int i;
2160   int max, len;
2161   max = 0;
2162   for(i=0; op[i].label; i++){
2163     len = lemonStrlen(op[i].label) + 1;
2164     switch( op[i].type ){
2165       case OPT_FLAG:
2166       case OPT_FFLAG:
2167         break;
2168       case OPT_INT:
2169       case OPT_FINT:
2170         len += 9;       /* length of "<integer>" */
2171         break;
2172       case OPT_DBL:
2173       case OPT_FDBL:
2174         len += 6;       /* length of "<real>" */
2175         break;
2176       case OPT_STR:
2177       case OPT_FSTR:
2178         len += 8;       /* length of "<string>" */
2179         break;
2180     }
2181     if( len>max ) max = len;
2182   }
2183   for(i=0; op[i].label; i++){
2184     switch( op[i].type ){
2185       case OPT_FLAG:
2186       case OPT_FFLAG:
2187         fprintf(errstream,"  -%-*s  %s\n",max,op[i].label,op[i].message);
2188         break;
2189       case OPT_INT:
2190       case OPT_FINT:
2191         fprintf(errstream,"  -%s<integer>%*s  %s\n",op[i].label,
2192           (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
2193         break;
2194       case OPT_DBL:
2195       case OPT_FDBL:
2196         fprintf(errstream,"  -%s<real>%*s  %s\n",op[i].label,
2197           (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
2198         break;
2199       case OPT_STR:
2200       case OPT_FSTR:
2201         fprintf(errstream,"  -%s<string>%*s  %s\n",op[i].label,
2202           (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
2203         break;
2204     }
2205   }
2206 }
2207 /*********************** From the file "parse.c" ****************************/
2208 /*
2209 ** Input file parser for the LEMON parser generator.
2210 */
2211 
2212 /* The state of the parser */
2213 enum e_state {
2214   INITIALIZE,
2215   WAITING_FOR_DECL_OR_RULE,
2216   WAITING_FOR_DECL_KEYWORD,
2217   WAITING_FOR_DECL_ARG,
2218   WAITING_FOR_PRECEDENCE_SYMBOL,
2219   WAITING_FOR_ARROW,
2220   IN_RHS,
2221   LHS_ALIAS_1,
2222   LHS_ALIAS_2,
2223   LHS_ALIAS_3,
2224   RHS_ALIAS_1,
2225   RHS_ALIAS_2,
2226   PRECEDENCE_MARK_1,
2227   PRECEDENCE_MARK_2,
2228   RESYNC_AFTER_RULE_ERROR,
2229   RESYNC_AFTER_DECL_ERROR,
2230   WAITING_FOR_DESTRUCTOR_SYMBOL,
2231   WAITING_FOR_DATATYPE_SYMBOL,
2232   WAITING_FOR_FALLBACK_ID,
2233   WAITING_FOR_WILDCARD_ID,
2234   WAITING_FOR_CLASS_ID,
2235   WAITING_FOR_CLASS_TOKEN,
2236   WAITING_FOR_TOKEN_NAME
2237 };
2238 struct pstate {
2239   char *filename;       /* Name of the input file */
2240   int tokenlineno;      /* Linenumber at which current token starts */
2241   int errorcnt;         /* Number of errors so far */
2242   char *tokenstart;     /* Text of current token */
2243   struct lemon *gp;     /* Global state vector */
2244   enum e_state state;        /* The state of the parser */
2245   struct symbol *fallback;   /* The fallback token */
2246   struct symbol *tkclass;    /* Token class symbol */
2247   struct symbol *lhs;        /* Left-hand side of current rule */
2248   const char *lhsalias;      /* Alias for the LHS */
2249   int nrhs;                  /* Number of right-hand side symbols seen */
2250   struct symbol *rhs[MAXRHS];  /* RHS symbols */
2251   const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
2252   struct rule *prevrule;     /* Previous rule parsed */
2253   const char *declkeyword;   /* Keyword of a declaration */
2254   char **declargslot;        /* Where the declaration argument should be put */
2255   int insertLineMacro;       /* Add #line before declaration insert */
2256   int *decllinenoslot;       /* Where to write declaration line number */
2257   enum e_assoc declassoc;    /* Assign this association to decl arguments */
2258   int preccounter;           /* Assign this precedence to decl arguments */
2259   struct rule *firstrule;    /* Pointer to first rule in the grammar */
2260   struct rule *lastrule;     /* Pointer to the most recently parsed rule */
2261 };
2262 
2263 /* Parse a single token */
parseonetoken(struct pstate * psp)2264 static void parseonetoken(struct pstate *psp)
2265 {
2266   const char *x;
2267   x = Strsafe(psp->tokenstart);     /* Save the token permanently */
2268 #if 0
2269   printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
2270     x,psp->state);
2271 #endif
2272   switch( psp->state ){
2273     case INITIALIZE:
2274       psp->prevrule = 0;
2275       psp->preccounter = 0;
2276       psp->firstrule = psp->lastrule = 0;
2277       psp->gp->nrule = 0;
2278       /* fall through */
2279     case WAITING_FOR_DECL_OR_RULE:
2280       if( x[0]=='%' ){
2281         psp->state = WAITING_FOR_DECL_KEYWORD;
2282       }else if( ISLOWER(x[0]) ){
2283         psp->lhs = Symbol_new(x);
2284         psp->nrhs = 0;
2285         psp->lhsalias = 0;
2286         psp->state = WAITING_FOR_ARROW;
2287       }else if( x[0]=='{' ){
2288         if( psp->prevrule==0 ){
2289           ErrorMsg(psp->filename,psp->tokenlineno,
2290             "There is no prior rule upon which to attach the code "
2291             "fragment which begins on this line.");
2292           psp->errorcnt++;
2293         }else if( psp->prevrule->code!=0 ){
2294           ErrorMsg(psp->filename,psp->tokenlineno,
2295             "Code fragment beginning on this line is not the first "
2296             "to follow the previous rule.");
2297           psp->errorcnt++;
2298         }else if( strcmp(x, "{NEVER-REDUCE")==0 ){
2299           psp->prevrule->neverReduce = 1;
2300         }else{
2301           psp->prevrule->line = psp->tokenlineno;
2302           psp->prevrule->code = &x[1];
2303           psp->prevrule->noCode = 0;
2304         }
2305       }else if( x[0]=='[' ){
2306         psp->state = PRECEDENCE_MARK_1;
2307       }else{
2308         ErrorMsg(psp->filename,psp->tokenlineno,
2309           "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2310           x);
2311         psp->errorcnt++;
2312       }
2313       break;
2314     case PRECEDENCE_MARK_1:
2315       if( !ISUPPER(x[0]) ){
2316         ErrorMsg(psp->filename,psp->tokenlineno,
2317           "The precedence symbol must be a terminal.");
2318         psp->errorcnt++;
2319       }else if( psp->prevrule==0 ){
2320         ErrorMsg(psp->filename,psp->tokenlineno,
2321           "There is no prior rule to assign precedence \"[%s]\".",x);
2322         psp->errorcnt++;
2323       }else if( psp->prevrule->precsym!=0 ){
2324         ErrorMsg(psp->filename,psp->tokenlineno,
2325           "Precedence mark on this line is not the first "
2326           "to follow the previous rule.");
2327         psp->errorcnt++;
2328       }else{
2329         psp->prevrule->precsym = Symbol_new(x);
2330       }
2331       psp->state = PRECEDENCE_MARK_2;
2332       break;
2333     case PRECEDENCE_MARK_2:
2334       if( x[0]!=']' ){
2335         ErrorMsg(psp->filename,psp->tokenlineno,
2336           "Missing \"]\" on precedence mark.");
2337         psp->errorcnt++;
2338       }
2339       psp->state = WAITING_FOR_DECL_OR_RULE;
2340       break;
2341     case WAITING_FOR_ARROW:
2342       if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2343         psp->state = IN_RHS;
2344       }else if( x[0]=='(' ){
2345         psp->state = LHS_ALIAS_1;
2346       }else{
2347         ErrorMsg(psp->filename,psp->tokenlineno,
2348           "Expected to see a \":\" following the LHS symbol \"%s\".",
2349           psp->lhs->name);
2350         psp->errorcnt++;
2351         psp->state = RESYNC_AFTER_RULE_ERROR;
2352       }
2353       break;
2354     case LHS_ALIAS_1:
2355       if( ISALPHA(x[0]) ){
2356         psp->lhsalias = x;
2357         psp->state = LHS_ALIAS_2;
2358       }else{
2359         ErrorMsg(psp->filename,psp->tokenlineno,
2360           "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2361           x,psp->lhs->name);
2362         psp->errorcnt++;
2363         psp->state = RESYNC_AFTER_RULE_ERROR;
2364       }
2365       break;
2366     case LHS_ALIAS_2:
2367       if( x[0]==')' ){
2368         psp->state = LHS_ALIAS_3;
2369       }else{
2370         ErrorMsg(psp->filename,psp->tokenlineno,
2371           "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2372         psp->errorcnt++;
2373         psp->state = RESYNC_AFTER_RULE_ERROR;
2374       }
2375       break;
2376     case LHS_ALIAS_3:
2377       if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2378         psp->state = IN_RHS;
2379       }else{
2380         ErrorMsg(psp->filename,psp->tokenlineno,
2381           "Missing \"->\" following: \"%s(%s)\".",
2382            psp->lhs->name,psp->lhsalias);
2383         psp->errorcnt++;
2384         psp->state = RESYNC_AFTER_RULE_ERROR;
2385       }
2386       break;
2387     case IN_RHS:
2388       if( x[0]=='.' ){
2389         struct rule *rp;
2390         rp = (struct rule *)calloc( sizeof(struct rule) +
2391              sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
2392         if( rp==0 ){
2393           ErrorMsg(psp->filename,psp->tokenlineno,
2394             "Can't allocate enough memory for this rule.");
2395           psp->errorcnt++;
2396           psp->prevrule = 0;
2397         }else{
2398           int i;
2399           rp->ruleline = psp->tokenlineno;
2400           rp->rhs = (struct symbol**)&rp[1];
2401           rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]);
2402           for(i=0; i<psp->nrhs; i++){
2403             rp->rhs[i] = psp->rhs[i];
2404             rp->rhsalias[i] = psp->alias[i];
2405             if( rp->rhsalias[i]!=0 ){ rp->rhs[i]->bContent = 1; }
2406           }
2407           rp->lhs = psp->lhs;
2408           rp->lhsalias = psp->lhsalias;
2409           rp->nrhs = psp->nrhs;
2410           rp->code = 0;
2411           rp->noCode = 1;
2412           rp->precsym = 0;
2413           rp->index = psp->gp->nrule++;
2414           rp->nextlhs = rp->lhs->rule;
2415           rp->lhs->rule = rp;
2416           rp->next = 0;
2417           if( psp->firstrule==0 ){
2418             psp->firstrule = psp->lastrule = rp;
2419           }else{
2420             psp->lastrule->next = rp;
2421             psp->lastrule = rp;
2422           }
2423           psp->prevrule = rp;
2424         }
2425         psp->state = WAITING_FOR_DECL_OR_RULE;
2426       }else if( ISALPHA(x[0]) ){
2427         if( psp->nrhs>=MAXRHS ){
2428           ErrorMsg(psp->filename,psp->tokenlineno,
2429             "Too many symbols on RHS of rule beginning at \"%s\".",
2430             x);
2431           psp->errorcnt++;
2432           psp->state = RESYNC_AFTER_RULE_ERROR;
2433         }else{
2434           psp->rhs[psp->nrhs] = Symbol_new(x);
2435           psp->alias[psp->nrhs] = 0;
2436           psp->nrhs++;
2437         }
2438       }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 && ISUPPER(x[1]) ){
2439         struct symbol *msp = psp->rhs[psp->nrhs-1];
2440         if( msp->type!=MULTITERMINAL ){
2441           struct symbol *origsp = msp;
2442           msp = (struct symbol *) calloc(1,sizeof(*msp));
2443           memset(msp, 0, sizeof(*msp));
2444           msp->type = MULTITERMINAL;
2445           msp->nsubsym = 1;
2446           msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*));
2447           msp->subsym[0] = origsp;
2448           msp->name = origsp->name;
2449           psp->rhs[psp->nrhs-1] = msp;
2450         }
2451         msp->nsubsym++;
2452         msp->subsym = (struct symbol **) realloc(msp->subsym,
2453           sizeof(struct symbol*)*msp->nsubsym);
2454         msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
2455         if( ISLOWER(x[1]) || ISLOWER(msp->subsym[0]->name[0]) ){
2456           ErrorMsg(psp->filename,psp->tokenlineno,
2457             "Cannot form a compound containing a non-terminal");
2458           psp->errorcnt++;
2459         }
2460       }else if( x[0]=='(' && psp->nrhs>0 ){
2461         psp->state = RHS_ALIAS_1;
2462       }else{
2463         ErrorMsg(psp->filename,psp->tokenlineno,
2464           "Illegal character on RHS of rule: \"%s\".",x);
2465         psp->errorcnt++;
2466         psp->state = RESYNC_AFTER_RULE_ERROR;
2467       }
2468       break;
2469     case RHS_ALIAS_1:
2470       if( ISALPHA(x[0]) ){
2471         psp->alias[psp->nrhs-1] = x;
2472         psp->state = RHS_ALIAS_2;
2473       }else{
2474         ErrorMsg(psp->filename,psp->tokenlineno,
2475           "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2476           x,psp->rhs[psp->nrhs-1]->name);
2477         psp->errorcnt++;
2478         psp->state = RESYNC_AFTER_RULE_ERROR;
2479       }
2480       break;
2481     case RHS_ALIAS_2:
2482       if( x[0]==')' ){
2483         psp->state = IN_RHS;
2484       }else{
2485         ErrorMsg(psp->filename,psp->tokenlineno,
2486           "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2487         psp->errorcnt++;
2488         psp->state = RESYNC_AFTER_RULE_ERROR;
2489       }
2490       break;
2491     case WAITING_FOR_DECL_KEYWORD:
2492       if( ISALPHA(x[0]) ){
2493         psp->declkeyword = x;
2494         psp->declargslot = 0;
2495         psp->decllinenoslot = 0;
2496         psp->insertLineMacro = 1;
2497         psp->state = WAITING_FOR_DECL_ARG;
2498         if( strcmp(x,"name")==0 ){
2499           psp->declargslot = &(psp->gp->name);
2500           psp->insertLineMacro = 0;
2501         }else if( strcmp(x,"include")==0 ){
2502           psp->declargslot = &(psp->gp->include);
2503         }else if( strcmp(x,"code")==0 ){
2504           psp->declargslot = &(psp->gp->extracode);
2505         }else if( strcmp(x,"token_destructor")==0 ){
2506           psp->declargslot = &psp->gp->tokendest;
2507         }else if( strcmp(x,"default_destructor")==0 ){
2508           psp->declargslot = &psp->gp->vardest;
2509         }else if( strcmp(x,"token_prefix")==0 ){
2510           psp->declargslot = &psp->gp->tokenprefix;
2511           psp->insertLineMacro = 0;
2512         }else if( strcmp(x,"syntax_error")==0 ){
2513           psp->declargslot = &(psp->gp->error);
2514         }else if( strcmp(x,"parse_accept")==0 ){
2515           psp->declargslot = &(psp->gp->accept);
2516         }else if( strcmp(x,"parse_failure")==0 ){
2517           psp->declargslot = &(psp->gp->failure);
2518         }else if( strcmp(x,"stack_overflow")==0 ){
2519           psp->declargslot = &(psp->gp->overflow);
2520         }else if( strcmp(x,"extra_argument")==0 ){
2521           psp->declargslot = &(psp->gp->arg);
2522           psp->insertLineMacro = 0;
2523         }else if( strcmp(x,"extra_context")==0 ){
2524           psp->declargslot = &(psp->gp->ctx);
2525           psp->insertLineMacro = 0;
2526         }else if( strcmp(x,"token_type")==0 ){
2527           psp->declargslot = &(psp->gp->tokentype);
2528           psp->insertLineMacro = 0;
2529         }else if( strcmp(x,"default_type")==0 ){
2530           psp->declargslot = &(psp->gp->vartype);
2531           psp->insertLineMacro = 0;
2532         }else if( strcmp(x,"stack_size")==0 ){
2533           psp->declargslot = &(psp->gp->stacksize);
2534           psp->insertLineMacro = 0;
2535         }else if( strcmp(x,"start_symbol")==0 ){
2536           psp->declargslot = &(psp->gp->start);
2537           psp->insertLineMacro = 0;
2538         }else if( strcmp(x,"left")==0 ){
2539           psp->preccounter++;
2540           psp->declassoc = LEFT;
2541           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2542         }else if( strcmp(x,"right")==0 ){
2543           psp->preccounter++;
2544           psp->declassoc = RIGHT;
2545           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2546         }else if( strcmp(x,"nonassoc")==0 ){
2547           psp->preccounter++;
2548           psp->declassoc = NONE;
2549           psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2550         }else if( strcmp(x,"destructor")==0 ){
2551           psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2552         }else if( strcmp(x,"type")==0 ){
2553           psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2554         }else if( strcmp(x,"fallback")==0 ){
2555           psp->fallback = 0;
2556           psp->state = WAITING_FOR_FALLBACK_ID;
2557         }else if( strcmp(x,"token")==0 ){
2558           psp->state = WAITING_FOR_TOKEN_NAME;
2559         }else if( strcmp(x,"wildcard")==0 ){
2560           psp->state = WAITING_FOR_WILDCARD_ID;
2561         }else if( strcmp(x,"token_class")==0 ){
2562           psp->state = WAITING_FOR_CLASS_ID;
2563         }else{
2564           ErrorMsg(psp->filename,psp->tokenlineno,
2565             "Unknown declaration keyword: \"%%%s\".",x);
2566           psp->errorcnt++;
2567           psp->state = RESYNC_AFTER_DECL_ERROR;
2568         }
2569       }else{
2570         ErrorMsg(psp->filename,psp->tokenlineno,
2571           "Illegal declaration keyword: \"%s\".",x);
2572         psp->errorcnt++;
2573         psp->state = RESYNC_AFTER_DECL_ERROR;
2574       }
2575       break;
2576     case WAITING_FOR_DESTRUCTOR_SYMBOL:
2577       if( !ISALPHA(x[0]) ){
2578         ErrorMsg(psp->filename,psp->tokenlineno,
2579           "Symbol name missing after %%destructor keyword");
2580         psp->errorcnt++;
2581         psp->state = RESYNC_AFTER_DECL_ERROR;
2582       }else{
2583         struct symbol *sp = Symbol_new(x);
2584         psp->declargslot = &sp->destructor;
2585         psp->decllinenoslot = &sp->destLineno;
2586         psp->insertLineMacro = 1;
2587         psp->state = WAITING_FOR_DECL_ARG;
2588       }
2589       break;
2590     case WAITING_FOR_DATATYPE_SYMBOL:
2591       if( !ISALPHA(x[0]) ){
2592         ErrorMsg(psp->filename,psp->tokenlineno,
2593           "Symbol name missing after %%type keyword");
2594         psp->errorcnt++;
2595         psp->state = RESYNC_AFTER_DECL_ERROR;
2596       }else{
2597         struct symbol *sp = Symbol_find(x);
2598         if((sp) && (sp->datatype)){
2599           ErrorMsg(psp->filename,psp->tokenlineno,
2600             "Symbol %%type \"%s\" already defined", x);
2601           psp->errorcnt++;
2602           psp->state = RESYNC_AFTER_DECL_ERROR;
2603         }else{
2604           if (!sp){
2605             sp = Symbol_new(x);
2606           }
2607           psp->declargslot = &sp->datatype;
2608           psp->insertLineMacro = 0;
2609           psp->state = WAITING_FOR_DECL_ARG;
2610         }
2611       }
2612       break;
2613     case WAITING_FOR_PRECEDENCE_SYMBOL:
2614       if( x[0]=='.' ){
2615         psp->state = WAITING_FOR_DECL_OR_RULE;
2616       }else if( ISUPPER(x[0]) ){
2617         struct symbol *sp;
2618         sp = Symbol_new(x);
2619         if( sp->prec>=0 ){
2620           ErrorMsg(psp->filename,psp->tokenlineno,
2621             "Symbol \"%s\" has already be given a precedence.",x);
2622           psp->errorcnt++;
2623         }else{
2624           sp->prec = psp->preccounter;
2625           sp->assoc = psp->declassoc;
2626         }
2627       }else{
2628         ErrorMsg(psp->filename,psp->tokenlineno,
2629           "Can't assign a precedence to \"%s\".",x);
2630         psp->errorcnt++;
2631       }
2632       break;
2633     case WAITING_FOR_DECL_ARG:
2634       if( x[0]=='{' || x[0]=='\"' || ISALNUM(x[0]) ){
2635         const char *zOld, *zNew;
2636         char *zBuf, *z;
2637         int nOld, n, nLine = 0, nNew, nBack;
2638         int addLineMacro;
2639         char zLine[50];
2640         zNew = x;
2641         if( zNew[0]=='"' || zNew[0]=='{' ) zNew++;
2642         nNew = lemonStrlen(zNew);
2643         if( *psp->declargslot ){
2644           zOld = *psp->declargslot;
2645         }else{
2646           zOld = "";
2647         }
2648         nOld = lemonStrlen(zOld);
2649         n = nOld + nNew + 20;
2650         addLineMacro = !psp->gp->nolinenosflag
2651                        && psp->insertLineMacro
2652                        && psp->tokenlineno>1
2653                        && (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0);
2654         if( addLineMacro ){
2655           for(z=psp->filename, nBack=0; *z; z++){
2656             if( *z=='\\' ) nBack++;
2657           }
2658           lemon_sprintf(zLine, "#line %d ", psp->tokenlineno);
2659           nLine = lemonStrlen(zLine);
2660           n += nLine + lemonStrlen(psp->filename) + nBack;
2661         }
2662         *psp->declargslot = (char *) realloc(*psp->declargslot, n);
2663         zBuf = *psp->declargslot + nOld;
2664         if( addLineMacro ){
2665           if( nOld && zBuf[-1]!='\n' ){
2666             *(zBuf++) = '\n';
2667           }
2668           memcpy(zBuf, zLine, nLine);
2669           zBuf += nLine;
2670           *(zBuf++) = '"';
2671           for(z=psp->filename; *z; z++){
2672             if( *z=='\\' ){
2673               *(zBuf++) = '\\';
2674             }
2675             *(zBuf++) = *z;
2676           }
2677           *(zBuf++) = '"';
2678           *(zBuf++) = '\n';
2679         }
2680         if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){
2681           psp->decllinenoslot[0] = psp->tokenlineno;
2682         }
2683         memcpy(zBuf, zNew, nNew);
2684         zBuf += nNew;
2685         *zBuf = 0;
2686         psp->state = WAITING_FOR_DECL_OR_RULE;
2687       }else{
2688         ErrorMsg(psp->filename,psp->tokenlineno,
2689           "Illegal argument to %%%s: %s",psp->declkeyword,x);
2690         psp->errorcnt++;
2691         psp->state = RESYNC_AFTER_DECL_ERROR;
2692       }
2693       break;
2694     case WAITING_FOR_FALLBACK_ID:
2695       if( x[0]=='.' ){
2696         psp->state = WAITING_FOR_DECL_OR_RULE;
2697       }else if( !ISUPPER(x[0]) ){
2698         ErrorMsg(psp->filename, psp->tokenlineno,
2699           "%%fallback argument \"%s\" should be a token", x);
2700         psp->errorcnt++;
2701       }else{
2702         struct symbol *sp = Symbol_new(x);
2703         if( psp->fallback==0 ){
2704           psp->fallback = sp;
2705         }else if( sp->fallback ){
2706           ErrorMsg(psp->filename, psp->tokenlineno,
2707             "More than one fallback assigned to token %s", x);
2708           psp->errorcnt++;
2709         }else{
2710           sp->fallback = psp->fallback;
2711           psp->gp->has_fallback = 1;
2712         }
2713       }
2714       break;
2715     case WAITING_FOR_TOKEN_NAME:
2716       /* Tokens do not have to be declared before use.  But they can be
2717       ** in order to control their assigned integer number.  The number for
2718       ** each token is assigned when it is first seen.  So by including
2719       **
2720       **     %token ONE TWO THREE
2721       **
2722       ** early in the grammar file, that assigns small consecutive values
2723       ** to each of the tokens ONE TWO and THREE.
2724       */
2725       if( x[0]=='.' ){
2726         psp->state = WAITING_FOR_DECL_OR_RULE;
2727       }else if( !ISUPPER(x[0]) ){
2728         ErrorMsg(psp->filename, psp->tokenlineno,
2729           "%%token argument \"%s\" should be a token", x);
2730         psp->errorcnt++;
2731       }else{
2732         (void)Symbol_new(x);
2733       }
2734       break;
2735     case WAITING_FOR_WILDCARD_ID:
2736       if( x[0]=='.' ){
2737         psp->state = WAITING_FOR_DECL_OR_RULE;
2738       }else if( !ISUPPER(x[0]) ){
2739         ErrorMsg(psp->filename, psp->tokenlineno,
2740           "%%wildcard argument \"%s\" should be a token", x);
2741         psp->errorcnt++;
2742       }else{
2743         struct symbol *sp = Symbol_new(x);
2744         if( psp->gp->wildcard==0 ){
2745           psp->gp->wildcard = sp;
2746         }else{
2747           ErrorMsg(psp->filename, psp->tokenlineno,
2748             "Extra wildcard to token: %s", x);
2749           psp->errorcnt++;
2750         }
2751       }
2752       break;
2753     case WAITING_FOR_CLASS_ID:
2754       if( !ISLOWER(x[0]) ){
2755         ErrorMsg(psp->filename, psp->tokenlineno,
2756           "%%token_class must be followed by an identifier: %s", x);
2757         psp->errorcnt++;
2758         psp->state = RESYNC_AFTER_DECL_ERROR;
2759      }else if( Symbol_find(x) ){
2760         ErrorMsg(psp->filename, psp->tokenlineno,
2761           "Symbol \"%s\" already used", x);
2762         psp->errorcnt++;
2763         psp->state = RESYNC_AFTER_DECL_ERROR;
2764       }else{
2765         psp->tkclass = Symbol_new(x);
2766         psp->tkclass->type = MULTITERMINAL;
2767         psp->state = WAITING_FOR_CLASS_TOKEN;
2768       }
2769       break;
2770     case WAITING_FOR_CLASS_TOKEN:
2771       if( x[0]=='.' ){
2772         psp->state = WAITING_FOR_DECL_OR_RULE;
2773       }else if( ISUPPER(x[0]) || ((x[0]=='|' || x[0]=='/') && ISUPPER(x[1])) ){
2774         struct symbol *msp = psp->tkclass;
2775         msp->nsubsym++;
2776         msp->subsym = (struct symbol **) realloc(msp->subsym,
2777           sizeof(struct symbol*)*msp->nsubsym);
2778         if( !ISUPPER(x[0]) ) x++;
2779         msp->subsym[msp->nsubsym-1] = Symbol_new(x);
2780       }else{
2781         ErrorMsg(psp->filename, psp->tokenlineno,
2782           "%%token_class argument \"%s\" should be a token", x);
2783         psp->errorcnt++;
2784         psp->state = RESYNC_AFTER_DECL_ERROR;
2785       }
2786       break;
2787     case RESYNC_AFTER_RULE_ERROR:
2788 /*      if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2789 **      break; */
2790     case RESYNC_AFTER_DECL_ERROR:
2791       if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2792       if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2793       break;
2794   }
2795 }
2796 
2797 /* The text in the input is part of the argument to an %ifdef or %ifndef.
2798 ** Evaluate the text as a boolean expression.  Return true or false.
2799 */
eval_preprocessor_boolean(char * z,int lineno)2800 static int eval_preprocessor_boolean(char *z, int lineno){
2801   int neg = 0;
2802   int res = 0;
2803   int okTerm = 1;
2804   int i;
2805   for(i=0; z[i]!=0; i++){
2806     if( ISSPACE(z[i]) ) continue;
2807     if( z[i]=='!' ){
2808       if( !okTerm ) goto pp_syntax_error;
2809       neg = !neg;
2810       continue;
2811     }
2812     if( z[i]=='|' && z[i+1]=='|' ){
2813       if( okTerm ) goto pp_syntax_error;
2814       if( res ) return 1;
2815       i++;
2816       okTerm = 1;
2817       continue;
2818     }
2819     if( z[i]=='&' && z[i+1]=='&' ){
2820       if( okTerm ) goto pp_syntax_error;
2821       if( !res ) return 0;
2822       i++;
2823       okTerm = 1;
2824       continue;
2825     }
2826     if( z[i]=='(' ){
2827       int k;
2828       int n = 1;
2829       if( !okTerm ) goto pp_syntax_error;
2830       for(k=i+1; z[k]; k++){
2831         if( z[k]==')' ){
2832           n--;
2833           if( n==0 ){
2834             z[k] = 0;
2835             res = eval_preprocessor_boolean(&z[i+1], -1);
2836             z[k] = ')';
2837             if( res<0 ){
2838               i = i-res;
2839               goto pp_syntax_error;
2840             }
2841             i = k;
2842             break;
2843           }
2844         }else if( z[k]=='(' ){
2845           n++;
2846         }else if( z[k]==0 ){
2847           i = k;
2848           goto pp_syntax_error;
2849         }
2850       }
2851       if( neg ){
2852         res = !res;
2853         neg = 0;
2854       }
2855       okTerm = 0;
2856       continue;
2857     }
2858     if( ISALPHA(z[i]) ){
2859       int j, k, n;
2860       if( !okTerm ) goto pp_syntax_error;
2861       for(k=i+1; ISALNUM(z[k]) || z[k]=='_'; k++){}
2862       n = k - i;
2863       res = 0;
2864       for(j=0; j<nDefine; j++){
2865         if( strncmp(azDefine[j],&z[i],n)==0 && azDefine[j][n]==0 ){
2866           res = 1;
2867           break;
2868         }
2869       }
2870       i = k-1;
2871       if( neg ){
2872         res = !res;
2873         neg = 0;
2874       }
2875       okTerm = 0;
2876       continue;
2877     }
2878     goto pp_syntax_error;
2879   }
2880   return res;
2881 
2882 pp_syntax_error:
2883   if( lineno>0 ){
2884     fprintf(stderr, "%%if syntax error on line %d.\n", lineno);
2885     fprintf(stderr, "  %.*s <-- syntax error here\n", i+1, z);
2886     exit(1);
2887   }else{
2888     return -(i+1);
2889   }
2890 }
2891 
2892 /* Run the preprocessor over the input file text.  The global variables
2893 ** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2894 ** macros.  This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2895 ** comments them out.  Text in between is also commented out as appropriate.
2896 */
preprocess_input(char * z)2897 static void preprocess_input(char *z){
2898   int i, j, k;
2899   int exclude = 0;
2900   int start = 0;
2901   int lineno = 1;
2902   int start_lineno = 1;
2903   for(i=0; z[i]; i++){
2904     if( z[i]=='\n' ) lineno++;
2905     if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue;
2906     if( strncmp(&z[i],"%endif",6)==0 && ISSPACE(z[i+6]) ){
2907       if( exclude ){
2908         exclude--;
2909         if( exclude==0 ){
2910           for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2911         }
2912       }
2913       for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2914     }else if( strncmp(&z[i],"%else",5)==0 && ISSPACE(z[i+5]) ){
2915       if( exclude==1){
2916         exclude = 0;
2917         for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2918       }else if( exclude==0 ){
2919         exclude = 1;
2920         start = i;
2921         start_lineno = lineno;
2922       }
2923       for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2924     }else if( strncmp(&z[i],"%ifdef ",7)==0
2925           || strncmp(&z[i],"%if ",4)==0
2926           || strncmp(&z[i],"%ifndef ",8)==0 ){
2927       if( exclude ){
2928         exclude++;
2929       }else{
2930         int isNot;
2931         int iBool;
2932         for(j=i; z[j] && !ISSPACE(z[j]); j++){}
2933         iBool = j;
2934         isNot = (j==i+7);
2935         while( z[j] && z[j]!='\n' ){ j++; }
2936         k = z[j];
2937         z[j] = 0;
2938         exclude = eval_preprocessor_boolean(&z[iBool], lineno);
2939         z[j] = k;
2940         if( !isNot ) exclude = !exclude;
2941         if( exclude ){
2942           start = i;
2943           start_lineno = lineno;
2944         }
2945       }
2946       for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2947     }
2948   }
2949   if( exclude ){
2950     fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno);
2951     exit(1);
2952   }
2953 }
2954 
2955 /* In spite of its name, this function is really a scanner.  It read
2956 ** in the entire input file (all at once) then tokenizes it.  Each
2957 ** token is passed to the function "parseonetoken" which builds all
2958 ** the appropriate data structures in the global state vector "gp".
2959 */
Parse(struct lemon * gp)2960 void Parse(struct lemon *gp)
2961 {
2962   struct pstate ps;
2963   FILE *fp;
2964   char *filebuf;
2965   unsigned int filesize;
2966   int lineno;
2967   int c;
2968   char *cp, *nextcp;
2969   int startline = 0;
2970 
2971   memset(&ps, '\0', sizeof(ps));
2972   ps.gp = gp;
2973   ps.filename = gp->filename;
2974   ps.errorcnt = 0;
2975   ps.state = INITIALIZE;
2976 
2977   /* Begin by reading the input file */
2978   fp = fopen(ps.filename,"rb");
2979   if( fp==0 ){
2980     ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2981     gp->errorcnt++;
2982     return;
2983   }
2984   fseek(fp,0,2);
2985   filesize = ftell(fp);
2986   rewind(fp);
2987   filebuf = (char *)malloc( filesize+1 );
2988   if( filesize>100000000 || filebuf==0 ){
2989     ErrorMsg(ps.filename,0,"Input file too large.");
2990     free(filebuf);
2991     gp->errorcnt++;
2992     fclose(fp);
2993     return;
2994   }
2995   if( fread(filebuf,1,filesize,fp)!=filesize ){
2996     ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2997       filesize);
2998     free(filebuf);
2999     gp->errorcnt++;
3000     fclose(fp);
3001     return;
3002   }
3003   fclose(fp);
3004   filebuf[filesize] = 0;
3005 
3006   /* Make an initial pass through the file to handle %ifdef and %ifndef */
3007   preprocess_input(filebuf);
3008   if( gp->printPreprocessed ){
3009     printf("%s\n", filebuf);
3010     return;
3011   }
3012 
3013   /* Now scan the text of the input file */
3014   lineno = 1;
3015   for(cp=filebuf; (c= *cp)!=0; ){
3016     if( c=='\n' ) lineno++;              /* Keep track of the line number */
3017     if( ISSPACE(c) ){ cp++; continue; }  /* Skip all white space */
3018     if( c=='/' && cp[1]=='/' ){          /* Skip C++ style comments */
3019       cp+=2;
3020       while( (c= *cp)!=0 && c!='\n' ) cp++;
3021       continue;
3022     }
3023     if( c=='/' && cp[1]=='*' ){          /* Skip C style comments */
3024       cp+=2;
3025       while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
3026         if( c=='\n' ) lineno++;
3027         cp++;
3028       }
3029       if( c ) cp++;
3030       continue;
3031     }
3032     ps.tokenstart = cp;                /* Mark the beginning of the token */
3033     ps.tokenlineno = lineno;           /* Linenumber on which token begins */
3034     if( c=='\"' ){                     /* String literals */
3035       cp++;
3036       while( (c= *cp)!=0 && c!='\"' ){
3037         if( c=='\n' ) lineno++;
3038         cp++;
3039       }
3040       if( c==0 ){
3041         ErrorMsg(ps.filename,startline,
3042             "String starting on this line is not terminated before "
3043             "the end of the file.");
3044         ps.errorcnt++;
3045         nextcp = cp;
3046       }else{
3047         nextcp = cp+1;
3048       }
3049     }else if( c=='{' ){               /* A block of C code */
3050       int level;
3051       cp++;
3052       for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
3053         if( c=='\n' ) lineno++;
3054         else if( c=='{' ) level++;
3055         else if( c=='}' ) level--;
3056         else if( c=='/' && cp[1]=='*' ){  /* Skip comments */
3057           int prevc;
3058           cp = &cp[2];
3059           prevc = 0;
3060           while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
3061             if( c=='\n' ) lineno++;
3062             prevc = c;
3063             cp++;
3064           }
3065         }else if( c=='/' && cp[1]=='/' ){  /* Skip C++ style comments too */
3066           cp = &cp[2];
3067           while( (c= *cp)!=0 && c!='\n' ) cp++;
3068           if( c ) lineno++;
3069         }else if( c=='\'' || c=='\"' ){    /* String a character literals */
3070           int startchar, prevc;
3071           startchar = c;
3072           prevc = 0;
3073           for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
3074             if( c=='\n' ) lineno++;
3075             if( prevc=='\\' ) prevc = 0;
3076             else              prevc = c;
3077           }
3078         }
3079       }
3080       if( c==0 ){
3081         ErrorMsg(ps.filename,ps.tokenlineno,
3082           "C code starting on this line is not terminated before "
3083           "the end of the file.");
3084         ps.errorcnt++;
3085         nextcp = cp;
3086       }else{
3087         nextcp = cp+1;
3088       }
3089     }else if( ISALNUM(c) ){          /* Identifiers */
3090       while( (c= *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
3091       nextcp = cp;
3092     }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
3093       cp += 3;
3094       nextcp = cp;
3095     }else if( (c=='/' || c=='|') && ISALPHA(cp[1]) ){
3096       cp += 2;
3097       while( (c = *cp)!=0 && (ISALNUM(c) || c=='_') ) cp++;
3098       nextcp = cp;
3099     }else{                          /* All other (one character) operators */
3100       cp++;
3101       nextcp = cp;
3102     }
3103     c = *cp;
3104     *cp = 0;                        /* Null terminate the token */
3105     parseonetoken(&ps);             /* Parse the token */
3106     *cp = (char)c;                  /* Restore the buffer */
3107     cp = nextcp;
3108   }
3109   free(filebuf);                    /* Release the buffer after parsing */
3110   gp->rule = ps.firstrule;
3111   gp->errorcnt = ps.errorcnt;
3112 }
3113 /*************************** From the file "plink.c" *********************/
3114 /*
3115 ** Routines processing configuration follow-set propagation links
3116 ** in the LEMON parser generator.
3117 */
3118 static struct plink *plink_freelist = 0;
3119 
3120 /* Allocate a new plink */
Plink_new(void)3121 struct plink *Plink_new(void){
3122   struct plink *newlink;
3123 
3124   if( plink_freelist==0 ){
3125     int i;
3126     int amt = 100;
3127     plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) );
3128     if( plink_freelist==0 ){
3129       fprintf(stderr,
3130       "Unable to allocate memory for a new follow-set propagation link.\n");
3131       exit(1);
3132     }
3133     for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
3134     plink_freelist[amt-1].next = 0;
3135   }
3136   newlink = plink_freelist;
3137   plink_freelist = plink_freelist->next;
3138   return newlink;
3139 }
3140 
3141 /* Add a plink to a plink list */
Plink_add(struct plink ** plpp,struct config * cfp)3142 void Plink_add(struct plink **plpp, struct config *cfp)
3143 {
3144   struct plink *newlink;
3145   newlink = Plink_new();
3146   newlink->next = *plpp;
3147   *plpp = newlink;
3148   newlink->cfp = cfp;
3149 }
3150 
3151 /* Transfer every plink on the list "from" to the list "to" */
Plink_copy(struct plink ** to,struct plink * from)3152 void Plink_copy(struct plink **to, struct plink *from)
3153 {
3154   struct plink *nextpl;
3155   while( from ){
3156     nextpl = from->next;
3157     from->next = *to;
3158     *to = from;
3159     from = nextpl;
3160   }
3161 }
3162 
3163 /* Delete every plink on the list */
Plink_delete(struct plink * plp)3164 void Plink_delete(struct plink *plp)
3165 {
3166   struct plink *nextpl;
3167 
3168   while( plp ){
3169     nextpl = plp->next;
3170     plp->next = plink_freelist;
3171     plink_freelist = plp;
3172     plp = nextpl;
3173   }
3174 }
3175 /*********************** From the file "report.c" **************************/
3176 /*
3177 ** Procedures for generating reports and tables in the LEMON parser generator.
3178 */
3179 
3180 /* Generate a filename with the given suffix.  Space to hold the
3181 ** name comes from malloc() and must be freed by the calling
3182 ** function.
3183 */
file_makename(struct lemon * lemp,const char * suffix)3184 PRIVATE char *file_makename(struct lemon *lemp, const char *suffix)
3185 {
3186   char *name;
3187   char *cp;
3188   char *filename = lemp->filename;
3189   int sz;
3190 
3191   if( outputDir ){
3192     cp = strrchr(filename, '/');
3193     if( cp ) filename = cp + 1;
3194   }
3195   sz = lemonStrlen(filename);
3196   sz += lemonStrlen(suffix);
3197   if( outputDir ) sz += lemonStrlen(outputDir) + 1;
3198   sz += 5;
3199   name = (char*)malloc( sz );
3200   if( name==0 ){
3201     fprintf(stderr,"Can't allocate space for a filename.\n");
3202     exit(1);
3203   }
3204   name[0] = 0;
3205   if( outputDir ){
3206     lemon_strcpy(name, outputDir);
3207     lemon_strcat(name, "/");
3208   }
3209   lemon_strcat(name,filename);
3210   cp = strrchr(name,'.');
3211   if( cp ) *cp = 0;
3212   lemon_strcat(name,suffix);
3213   return name;
3214 }
3215 
3216 /* Open a file with a name based on the name of the input file,
3217 ** but with a different (specified) suffix, and return a pointer
3218 ** to the stream */
file_open(struct lemon * lemp,const char * suffix,const char * mode)3219 PRIVATE FILE *file_open(
3220   struct lemon *lemp,
3221   const char *suffix,
3222   const char *mode
3223 ){
3224   FILE *fp;
3225 
3226   if( lemp->outname ) free(lemp->outname);
3227   lemp->outname = file_makename(lemp, suffix);
3228   fp = fopen(lemp->outname,mode);
3229   if( fp==0 && *mode=='w' ){
3230     fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
3231     lemp->errorcnt++;
3232     return 0;
3233   }
3234   return fp;
3235 }
3236 
3237 /* Print the text of a rule
3238 */
rule_print(FILE * out,struct rule * rp)3239 void rule_print(FILE *out, struct rule *rp){
3240   int i, j;
3241   fprintf(out, "%s",rp->lhs->name);
3242   /*    if( rp->lhsalias ) fprintf(out,"(%s)",rp->lhsalias); */
3243   fprintf(out," ::=");
3244   for(i=0; i<rp->nrhs; i++){
3245     struct symbol *sp = rp->rhs[i];
3246     if( sp->type==MULTITERMINAL ){
3247       fprintf(out," %s", sp->subsym[0]->name);
3248       for(j=1; j<sp->nsubsym; j++){
3249         fprintf(out,"|%s", sp->subsym[j]->name);
3250       }
3251     }else{
3252       fprintf(out," %s", sp->name);
3253     }
3254     /* if( rp->rhsalias[i] ) fprintf(out,"(%s)",rp->rhsalias[i]); */
3255   }
3256 }
3257 
3258 /* Duplicate the input file without comments and without actions
3259 ** on rules */
Reprint(struct lemon * lemp)3260 void Reprint(struct lemon *lemp)
3261 {
3262   struct rule *rp;
3263   struct symbol *sp;
3264   int i, j, maxlen, len, ncolumns, skip;
3265   printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
3266   maxlen = 10;
3267   for(i=0; i<lemp->nsymbol; i++){
3268     sp = lemp->symbols[i];
3269     len = lemonStrlen(sp->name);
3270     if( len>maxlen ) maxlen = len;
3271   }
3272   ncolumns = 76/(maxlen+5);
3273   if( ncolumns<1 ) ncolumns = 1;
3274   skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
3275   for(i=0; i<skip; i++){
3276     printf("//");
3277     for(j=i; j<lemp->nsymbol; j+=skip){
3278       sp = lemp->symbols[j];
3279       assert( sp->index==j );
3280       printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
3281     }
3282     printf("\n");
3283   }
3284   for(rp=lemp->rule; rp; rp=rp->next){
3285     rule_print(stdout, rp);
3286     printf(".");
3287     if( rp->precsym ) printf(" [%s]",rp->precsym->name);
3288     /* if( rp->code ) printf("\n    %s",rp->code); */
3289     printf("\n");
3290   }
3291 }
3292 
3293 /* Print a single rule.
3294 */
RulePrint(FILE * fp,struct rule * rp,int iCursor)3295 void RulePrint(FILE *fp, struct rule *rp, int iCursor){
3296   struct symbol *sp;
3297   int i, j;
3298   fprintf(fp,"%s ::=",rp->lhs->name);
3299   for(i=0; i<=rp->nrhs; i++){
3300     if( i==iCursor ) fprintf(fp," *");
3301     if( i==rp->nrhs ) break;
3302     sp = rp->rhs[i];
3303     if( sp->type==MULTITERMINAL ){
3304       fprintf(fp," %s", sp->subsym[0]->name);
3305       for(j=1; j<sp->nsubsym; j++){
3306         fprintf(fp,"|%s",sp->subsym[j]->name);
3307       }
3308     }else{
3309       fprintf(fp," %s", sp->name);
3310     }
3311   }
3312 }
3313 
3314 /* Print the rule for a configuration.
3315 */
ConfigPrint(FILE * fp,struct config * cfp)3316 void ConfigPrint(FILE *fp, struct config *cfp){
3317   RulePrint(fp, cfp->rp, cfp->dot);
3318 }
3319 
3320 /* #define TEST */
3321 #if 0
3322 /* Print a set */
3323 PRIVATE void SetPrint(out,set,lemp)
3324 FILE *out;
3325 char *set;
3326 struct lemon *lemp;
3327 {
3328   int i;
3329   char *spacer;
3330   spacer = "";
3331   fprintf(out,"%12s[","");
3332   for(i=0; i<lemp->nterminal; i++){
3333     if( SetFind(set,i) ){
3334       fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
3335       spacer = " ";
3336     }
3337   }
3338   fprintf(out,"]\n");
3339 }
3340 
3341 /* Print a plink chain */
3342 PRIVATE void PlinkPrint(out,plp,tag)
3343 FILE *out;
3344 struct plink *plp;
3345 char *tag;
3346 {
3347   while( plp ){
3348     fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum);
3349     ConfigPrint(out,plp->cfp);
3350     fprintf(out,"\n");
3351     plp = plp->next;
3352   }
3353 }
3354 #endif
3355 
3356 /* Print an action to the given file descriptor.  Return FALSE if
3357 ** nothing was actually printed.
3358 */
PrintAction(struct action * ap,FILE * fp,int indent)3359 int PrintAction(
3360   struct action *ap,          /* The action to print */
3361   FILE *fp,                   /* Print the action here */
3362   int indent                  /* Indent by this amount */
3363 ){
3364   int result = 1;
3365   switch( ap->type ){
3366     case SHIFT: {
3367       struct state *stp = ap->x.stp;
3368       fprintf(fp,"%*s shift        %-7d",indent,ap->sp->name,stp->statenum);
3369       break;
3370     }
3371     case REDUCE: {
3372       struct rule *rp = ap->x.rp;
3373       fprintf(fp,"%*s reduce       %-7d",indent,ap->sp->name,rp->iRule);
3374       RulePrint(fp, rp, -1);
3375       break;
3376     }
3377     case SHIFTREDUCE: {
3378       struct rule *rp = ap->x.rp;
3379       fprintf(fp,"%*s shift-reduce %-7d",indent,ap->sp->name,rp->iRule);
3380       RulePrint(fp, rp, -1);
3381       break;
3382     }
3383     case ACCEPT:
3384       fprintf(fp,"%*s accept",indent,ap->sp->name);
3385       break;
3386     case ERROR:
3387       fprintf(fp,"%*s error",indent,ap->sp->name);
3388       break;
3389     case SRCONFLICT:
3390     case RRCONFLICT:
3391       fprintf(fp,"%*s reduce       %-7d ** Parsing conflict **",
3392         indent,ap->sp->name,ap->x.rp->iRule);
3393       break;
3394     case SSCONFLICT:
3395       fprintf(fp,"%*s shift        %-7d ** Parsing conflict **",
3396         indent,ap->sp->name,ap->x.stp->statenum);
3397       break;
3398     case SH_RESOLVED:
3399       if( showPrecedenceConflict ){
3400         fprintf(fp,"%*s shift        %-7d -- dropped by precedence",
3401                 indent,ap->sp->name,ap->x.stp->statenum);
3402       }else{
3403         result = 0;
3404       }
3405       break;
3406     case RD_RESOLVED:
3407       if( showPrecedenceConflict ){
3408         fprintf(fp,"%*s reduce %-7d -- dropped by precedence",
3409                 indent,ap->sp->name,ap->x.rp->iRule);
3410       }else{
3411         result = 0;
3412       }
3413       break;
3414     case NOT_USED:
3415       result = 0;
3416       break;
3417   }
3418   if( result && ap->spOpt ){
3419     fprintf(fp,"  /* because %s==%s */", ap->sp->name, ap->spOpt->name);
3420   }
3421   return result;
3422 }
3423 
3424 /* Generate the "*.out" log file */
ReportOutput(struct lemon * lemp)3425 void ReportOutput(struct lemon *lemp)
3426 {
3427   int i, n;
3428   struct state *stp;
3429   struct config *cfp;
3430   struct action *ap;
3431   struct rule *rp;
3432   FILE *fp;
3433 
3434   fp = file_open(lemp,".out","wb");
3435   if( fp==0 ) return;
3436   for(i=0; i<lemp->nxstate; i++){
3437     stp = lemp->sorted[i];
3438     fprintf(fp,"State %d:\n",stp->statenum);
3439     if( lemp->basisflag ) cfp=stp->bp;
3440     else                  cfp=stp->cfp;
3441     while( cfp ){
3442       char buf[20];
3443       if( cfp->dot==cfp->rp->nrhs ){
3444         lemon_sprintf(buf,"(%d)",cfp->rp->iRule);
3445         fprintf(fp,"    %5s ",buf);
3446       }else{
3447         fprintf(fp,"          ");
3448       }
3449       ConfigPrint(fp,cfp);
3450       fprintf(fp,"\n");
3451 #if 0
3452       SetPrint(fp,cfp->fws,lemp);
3453       PlinkPrint(fp,cfp->fplp,"To  ");
3454       PlinkPrint(fp,cfp->bplp,"From");
3455 #endif
3456       if( lemp->basisflag ) cfp=cfp->bp;
3457       else                  cfp=cfp->next;
3458     }
3459     fprintf(fp,"\n");
3460     for(ap=stp->ap; ap; ap=ap->next){
3461       if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
3462     }
3463     fprintf(fp,"\n");
3464   }
3465   fprintf(fp, "----------------------------------------------------\n");
3466   fprintf(fp, "Symbols:\n");
3467   fprintf(fp, "The first-set of non-terminals is shown after the name.\n\n");
3468   for(i=0; i<lemp->nsymbol; i++){
3469     int j;
3470     struct symbol *sp;
3471 
3472     sp = lemp->symbols[i];
3473     fprintf(fp, "  %3d: %s", i, sp->name);
3474     if( sp->type==NONTERMINAL ){
3475       fprintf(fp, ":");
3476       if( sp->lambda ){
3477         fprintf(fp, " <lambda>");
3478       }
3479       for(j=0; j<lemp->nterminal; j++){
3480         if( sp->firstset && SetFind(sp->firstset, j) ){
3481           fprintf(fp, " %s", lemp->symbols[j]->name);
3482         }
3483       }
3484     }
3485     if( sp->prec>=0 ) fprintf(fp," (precedence=%d)", sp->prec);
3486     fprintf(fp, "\n");
3487   }
3488   fprintf(fp, "----------------------------------------------------\n");
3489   fprintf(fp, "Syntax-only Symbols:\n");
3490   fprintf(fp, "The following symbols never carry semantic content.\n\n");
3491   for(i=n=0; i<lemp->nsymbol; i++){
3492     int w;
3493     struct symbol *sp = lemp->symbols[i];
3494     if( sp->bContent ) continue;
3495     w = (int)strlen(sp->name);
3496     if( n>0 && n+w>75 ){
3497       fprintf(fp,"\n");
3498       n = 0;
3499     }
3500     if( n>0 ){
3501       fprintf(fp, " ");
3502       n++;
3503     }
3504     fprintf(fp, "%s", sp->name);
3505     n += w;
3506   }
3507   if( n>0 ) fprintf(fp, "\n");
3508   fprintf(fp, "----------------------------------------------------\n");
3509   fprintf(fp, "Rules:\n");
3510   for(rp=lemp->rule; rp; rp=rp->next){
3511     fprintf(fp, "%4d: ", rp->iRule);
3512     rule_print(fp, rp);
3513     fprintf(fp,".");
3514     if( rp->precsym ){
3515       fprintf(fp," [%s precedence=%d]",
3516               rp->precsym->name, rp->precsym->prec);
3517     }
3518     fprintf(fp,"\n");
3519   }
3520   fclose(fp);
3521   return;
3522 }
3523 
3524 /* Search for the file "name" which is in the same directory as
3525 ** the exacutable */
pathsearch(char * argv0,char * name,int modemask)3526 PRIVATE char *pathsearch(char *argv0, char *name, int modemask)
3527 {
3528   const char *pathlist;
3529   char *pathbufptr = 0;
3530   char *pathbuf = 0;
3531   char *path,*cp;
3532   char c;
3533 
3534 #ifdef __WIN32__
3535   cp = strrchr(argv0,'\\');
3536 #else
3537   cp = strrchr(argv0,'/');
3538 #endif
3539   if( cp ){
3540     c = *cp;
3541     *cp = 0;
3542     path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 );
3543     if( path ) lemon_sprintf(path,"%s/%s",argv0,name);
3544     *cp = c;
3545   }else{
3546     pathlist = getenv("PATH");
3547     if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
3548     pathbuf = (char *) malloc( lemonStrlen(pathlist) + 1 );
3549     path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 );
3550     MemoryCheck(pathbuf); MemoryCheck(path);  /* Fail on allocation failure. */
3551     if( (pathbuf != 0) && (path!=0) ){
3552       pathbufptr = pathbuf;
3553       lemon_strcpy(pathbuf, pathlist);
3554       while( *pathbuf ){
3555         cp = strchr(pathbuf,':');
3556         if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)];
3557         c = *cp;
3558         *cp = 0;
3559         lemon_sprintf(path,"%s/%s",pathbuf,name);
3560         *cp = c;
3561         if( c==0 ) pathbuf[0] = 0;
3562         else pathbuf = &cp[1];
3563         if( access(path,modemask)==0 ) break;
3564       }
3565     }
3566     free(pathbufptr);
3567   }
3568   return path;
3569 }
3570 
3571 /* Given an action, compute the integer value for that action
3572 ** which is to be put in the action table of the generated machine.
3573 ** Return negative if no action should be generated.
3574 */
compute_action(struct lemon * lemp,struct action * ap)3575 PRIVATE int compute_action(struct lemon *lemp, struct action *ap)
3576 {
3577   int act;
3578   switch( ap->type ){
3579     case SHIFT:  act = ap->x.stp->statenum;                        break;
3580     case SHIFTREDUCE: {
3581       /* Since a SHIFT is inherient after a prior REDUCE, convert any
3582       ** SHIFTREDUCE action with a nonterminal on the LHS into a simple
3583       ** REDUCE action: */
3584       if( ap->sp->index>=lemp->nterminal ){
3585         act = lemp->minReduce + ap->x.rp->iRule;
3586       }else{
3587         act = lemp->minShiftReduce + ap->x.rp->iRule;
3588       }
3589       break;
3590     }
3591     case REDUCE: act = lemp->minReduce + ap->x.rp->iRule;          break;
3592     case ERROR:  act = lemp->errAction;                            break;
3593     case ACCEPT: act = lemp->accAction;                            break;
3594     default:     act = -1; break;
3595   }
3596   return act;
3597 }
3598 
3599 #define LINESIZE 1000
3600 /* The next cluster of routines are for reading the template file
3601 ** and writing the results to the generated parser */
3602 /* The first function transfers data from "in" to "out" until
3603 ** a line is seen which begins with "%%".  The line number is
3604 ** tracked.
3605 **
3606 ** if name!=0, then any word that begin with "Parse" is changed to
3607 ** begin with *name instead.
3608 */
tplt_xfer(char * name,FILE * in,FILE * out,int * lineno)3609 PRIVATE void tplt_xfer(char *name, FILE *in, FILE *out, int *lineno)
3610 {
3611   int i, iStart;
3612   char line[LINESIZE];
3613   while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3614     (*lineno)++;
3615     iStart = 0;
3616     if( name ){
3617       for(i=0; line[i]; i++){
3618         if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
3619           && (i==0 || !ISALPHA(line[i-1]))
3620         ){
3621           if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
3622           fprintf(out,"%s",name);
3623           i += 4;
3624           iStart = i+1;
3625         }
3626       }
3627     }
3628     fprintf(out,"%s",&line[iStart]);
3629   }
3630 }
3631 
3632 /* Skip forward past the header of the template file to the first "%%"
3633 */
tplt_skip_header(FILE * in,int * lineno)3634 PRIVATE void tplt_skip_header(FILE *in, int *lineno)
3635 {
3636   char line[LINESIZE];
3637   while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3638     (*lineno)++;
3639   }
3640 }
3641 
3642 /* The next function finds the template file and opens it, returning
3643 ** a pointer to the opened file. */
tplt_open(struct lemon * lemp)3644 PRIVATE FILE *tplt_open(struct lemon *lemp)
3645 {
3646   static char templatename[] = "lempar.c";
3647   char buf[1000];
3648   FILE *in;
3649   char *tpltname;
3650   char *toFree = 0;
3651   char *cp;
3652 
3653   /* first, see if user specified a template filename on the command line. */
3654   if (user_templatename != 0) {
3655     if( access(user_templatename,004)==-1 ){
3656       fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3657         user_templatename);
3658       lemp->errorcnt++;
3659       return 0;
3660     }
3661     in = fopen(user_templatename,"rb");
3662     if( in==0 ){
3663       fprintf(stderr,"Can't open the template file \"%s\".\n",
3664               user_templatename);
3665       lemp->errorcnt++;
3666       return 0;
3667     }
3668     return in;
3669   }
3670 
3671   cp = strrchr(lemp->filename,'.');
3672   if( cp ){
3673     lemon_sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
3674   }else{
3675     lemon_sprintf(buf,"%s.lt",lemp->filename);
3676   }
3677   if( access(buf,004)==0 ){
3678     tpltname = buf;
3679   }else if( access(templatename,004)==0 ){
3680     tpltname = templatename;
3681   }else{
3682     toFree = tpltname = pathsearch(lemp->argv0,templatename,0);
3683   }
3684   if( tpltname==0 ){
3685     fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3686     templatename);
3687     lemp->errorcnt++;
3688     return 0;
3689   }
3690   in = fopen(tpltname,"rb");
3691   if( in==0 ){
3692     fprintf(stderr,"Can't open the template file \"%s\".\n",tpltname);
3693     lemp->errorcnt++;
3694   }
3695   free(toFree);
3696   return in;
3697 }
3698 
3699 /* Print a #line directive line to the output file. */
tplt_linedir(FILE * out,int lineno,char * filename)3700 PRIVATE void tplt_linedir(FILE *out, int lineno, char *filename)
3701 {
3702   fprintf(out,"#line %d \"",lineno);
3703   while( *filename ){
3704     if( *filename == '\\' ) putc('\\',out);
3705     putc(*filename,out);
3706     filename++;
3707   }
3708   fprintf(out,"\"\n");
3709 }
3710 
3711 /* Print a string to the file and keep the linenumber up to date */
tplt_print(FILE * out,struct lemon * lemp,char * str,int * lineno)3712 PRIVATE void tplt_print(FILE *out, struct lemon *lemp, char *str, int *lineno)
3713 {
3714   if( str==0 ) return;
3715   while( *str ){
3716     putc(*str,out);
3717     if( *str=='\n' ) (*lineno)++;
3718     str++;
3719   }
3720   if( str[-1]!='\n' ){
3721     putc('\n',out);
3722     (*lineno)++;
3723   }
3724   if (!lemp->nolinenosflag) {
3725     (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3726   }
3727   return;
3728 }
3729 
3730 /*
3731 ** The following routine emits code for the destructor for the
3732 ** symbol sp
3733 */
emit_destructor_code(FILE * out,struct symbol * sp,struct lemon * lemp,int * lineno)3734 void emit_destructor_code(
3735   FILE *out,
3736   struct symbol *sp,
3737   struct lemon *lemp,
3738   int *lineno
3739 ){
3740  char *cp = 0;
3741 
3742  if( sp->type==TERMINAL ){
3743    cp = lemp->tokendest;
3744    if( cp==0 ) return;
3745    fprintf(out,"{\n"); (*lineno)++;
3746  }else if( sp->destructor ){
3747    cp = sp->destructor;
3748    fprintf(out,"{\n"); (*lineno)++;
3749    if( !lemp->nolinenosflag ){
3750      (*lineno)++;
3751      tplt_linedir(out,sp->destLineno,lemp->filename);
3752    }
3753  }else if( lemp->vardest ){
3754    cp = lemp->vardest;
3755    if( cp==0 ) return;
3756    fprintf(out,"{\n"); (*lineno)++;
3757  }else{
3758    assert( 0 );  /* Cannot happen */
3759  }
3760  for(; *cp; cp++){
3761    if( *cp=='$' && cp[1]=='$' ){
3762      fprintf(out,"(yypminor->yy%d)",sp->dtnum);
3763      cp++;
3764      continue;
3765    }
3766    if( *cp=='\n' ) (*lineno)++;
3767    fputc(*cp,out);
3768  }
3769  fprintf(out,"\n"); (*lineno)++;
3770  if (!lemp->nolinenosflag) {
3771    (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3772  }
3773  fprintf(out,"}\n"); (*lineno)++;
3774  return;
3775 }
3776 
3777 /*
3778 ** Return TRUE (non-zero) if the given symbol has a destructor.
3779 */
has_destructor(struct symbol * sp,struct lemon * lemp)3780 int has_destructor(struct symbol *sp, struct lemon *lemp)
3781 {
3782   int ret;
3783   if( sp->type==TERMINAL ){
3784     ret = lemp->tokendest!=0;
3785   }else{
3786     ret = lemp->vardest!=0 || sp->destructor!=0;
3787   }
3788   return ret;
3789 }
3790 
3791 /*
3792 ** Append text to a dynamically allocated string.  If zText is 0 then
3793 ** reset the string to be empty again.  Always return the complete text
3794 ** of the string (which is overwritten with each call).
3795 **
3796 ** n bytes of zText are stored.  If n==0 then all of zText up to the first
3797 ** \000 terminator is stored.  zText can contain up to two instances of
3798 ** %d.  The values of p1 and p2 are written into the first and second
3799 ** %d.
3800 **
3801 ** If n==-1, then the previous character is overwritten.
3802 */
append_str(const char * zText,int n,int p1,int p2)3803 PRIVATE char *append_str(const char *zText, int n, int p1, int p2){
3804   static char empty[1] = { 0 };
3805   static char *z = 0;
3806   static int alloced = 0;
3807   static int used = 0;
3808   int c;
3809   char zInt[40];
3810   if( zText==0 ){
3811     if( used==0 && z!=0 ) z[0] = 0;
3812     used = 0;
3813     return z;
3814   }
3815   if( n<=0 ){
3816     if( n<0 ){
3817       used += n;
3818       assert( used>=0 );
3819     }
3820     n = lemonStrlen(zText);
3821   }
3822   if( (int) (n+sizeof(zInt)*2+used) >= alloced ){
3823     alloced = n + sizeof(zInt)*2 + used + 200;
3824     z = (char *) realloc(z,  alloced);
3825   }
3826   if( z==0 ) return empty;
3827   while( n-- > 0 ){
3828     c = *(zText++);
3829     if( c=='%' && n>0 && zText[0]=='d' ){
3830       lemon_sprintf(zInt, "%d", p1);
3831       p1 = p2;
3832       lemon_strcpy(&z[used], zInt);
3833       used += lemonStrlen(&z[used]);
3834       zText++;
3835       n--;
3836     }else{
3837       z[used++] = (char)c;
3838     }
3839   }
3840   z[used] = 0;
3841   return z;
3842 }
3843 
3844 /*
3845 ** Write and transform the rp->code string so that symbols are expanded.
3846 ** Populate the rp->codePrefix and rp->codeSuffix strings, as appropriate.
3847 **
3848 ** Return 1 if the expanded code requires that "yylhsminor" local variable
3849 ** to be defined.
3850 */
translate_code(struct lemon * lemp,struct rule * rp)3851 PRIVATE int translate_code(struct lemon *lemp, struct rule *rp){
3852   char *cp, *xp;
3853   int i;
3854   int rc = 0;            /* True if yylhsminor is used */
3855   int dontUseRhs0 = 0;   /* If true, use of left-most RHS label is illegal */
3856   const char *zSkip = 0; /* The zOvwrt comment within rp->code, or NULL */
3857   char lhsused = 0;      /* True if the LHS element has been used */
3858   char lhsdirect;        /* True if LHS writes directly into stack */
3859   char used[MAXRHS];     /* True for each RHS element which is used */
3860   char zLhs[50];         /* Convert the LHS symbol into this string */
3861   char zOvwrt[900];      /* Comment that to allow LHS to overwrite RHS */
3862 
3863   for(i=0; i<rp->nrhs; i++) used[i] = 0;
3864   lhsused = 0;
3865 
3866   if( rp->code==0 ){
3867     static char newlinestr[2] = { '\n', '\0' };
3868     rp->code = newlinestr;
3869     rp->line = rp->ruleline;
3870     rp->noCode = 1;
3871   }else{
3872     rp->noCode = 0;
3873   }
3874 
3875 
3876   if( rp->nrhs==0 ){
3877     /* If there are no RHS symbols, then writing directly to the LHS is ok */
3878     lhsdirect = 1;
3879   }else if( rp->rhsalias[0]==0 ){
3880     /* The left-most RHS symbol has no value.  LHS direct is ok.  But
3881     ** we have to call the distructor on the RHS symbol first. */
3882     lhsdirect = 1;
3883     if( has_destructor(rp->rhs[0],lemp) ){
3884       append_str(0,0,0,0);
3885       append_str("  yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3886                  rp->rhs[0]->index,1-rp->nrhs);
3887       rp->codePrefix = Strsafe(append_str(0,0,0,0));
3888       rp->noCode = 0;
3889     }
3890   }else if( rp->lhsalias==0 ){
3891     /* There is no LHS value symbol. */
3892     lhsdirect = 1;
3893   }else if( strcmp(rp->lhsalias,rp->rhsalias[0])==0 ){
3894     /* The LHS symbol and the left-most RHS symbol are the same, so
3895     ** direct writing is allowed */
3896     lhsdirect = 1;
3897     lhsused = 1;
3898     used[0] = 1;
3899     if( rp->lhs->dtnum!=rp->rhs[0]->dtnum ){
3900       ErrorMsg(lemp->filename,rp->ruleline,
3901         "%s(%s) and %s(%s) share the same label but have "
3902         "different datatypes.",
3903         rp->lhs->name, rp->lhsalias, rp->rhs[0]->name, rp->rhsalias[0]);
3904       lemp->errorcnt++;
3905     }
3906   }else{
3907     lemon_sprintf(zOvwrt, "/*%s-overwrites-%s*/",
3908                   rp->lhsalias, rp->rhsalias[0]);
3909     zSkip = strstr(rp->code, zOvwrt);
3910     if( zSkip!=0 ){
3911       /* The code contains a special comment that indicates that it is safe
3912       ** for the LHS label to overwrite left-most RHS label. */
3913       lhsdirect = 1;
3914     }else{
3915       lhsdirect = 0;
3916     }
3917   }
3918   if( lhsdirect ){
3919     sprintf(zLhs, "yymsp[%d].minor.yy%d",1-rp->nrhs,rp->lhs->dtnum);
3920   }else{
3921     rc = 1;
3922     sprintf(zLhs, "yylhsminor.yy%d",rp->lhs->dtnum);
3923   }
3924 
3925   append_str(0,0,0,0);
3926 
3927   /* This const cast is wrong but harmless, if we're careful. */
3928   for(cp=(char *)rp->code; *cp; cp++){
3929     if( cp==zSkip ){
3930       append_str(zOvwrt,0,0,0);
3931       cp += lemonStrlen(zOvwrt)-1;
3932       dontUseRhs0 = 1;
3933       continue;
3934     }
3935     if( ISALPHA(*cp) && (cp==rp->code || (!ISALNUM(cp[-1]) && cp[-1]!='_')) ){
3936       char saved;
3937       for(xp= &cp[1]; ISALNUM(*xp) || *xp=='_'; xp++);
3938       saved = *xp;
3939       *xp = 0;
3940       if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
3941         append_str(zLhs,0,0,0);
3942         cp = xp;
3943         lhsused = 1;
3944       }else{
3945         for(i=0; i<rp->nrhs; i++){
3946           if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
3947             if( i==0 && dontUseRhs0 ){
3948               ErrorMsg(lemp->filename,rp->ruleline,
3949                  "Label %s used after '%s'.",
3950                  rp->rhsalias[0], zOvwrt);
3951               lemp->errorcnt++;
3952             }else if( cp!=rp->code && cp[-1]=='@' ){
3953               /* If the argument is of the form @X then substituted
3954               ** the token number of X, not the value of X */
3955               append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
3956             }else{
3957               struct symbol *sp = rp->rhs[i];
3958               int dtnum;
3959               if( sp->type==MULTITERMINAL ){
3960                 dtnum = sp->subsym[0]->dtnum;
3961               }else{
3962                 dtnum = sp->dtnum;
3963               }
3964               append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);
3965             }
3966             cp = xp;
3967             used[i] = 1;
3968             break;
3969           }
3970         }
3971       }
3972       *xp = saved;
3973     }
3974     append_str(cp, 1, 0, 0);
3975   } /* End loop */
3976 
3977   /* Main code generation completed */
3978   cp = append_str(0,0,0,0);
3979   if( cp && cp[0] ) rp->code = Strsafe(cp);
3980   append_str(0,0,0,0);
3981 
3982   /* Check to make sure the LHS has been used */
3983   if( rp->lhsalias && !lhsused ){
3984     ErrorMsg(lemp->filename,rp->ruleline,
3985       "Label \"%s\" for \"%s(%s)\" is never used.",
3986         rp->lhsalias,rp->lhs->name,rp->lhsalias);
3987     lemp->errorcnt++;
3988   }
3989 
3990   /* Generate destructor code for RHS minor values which are not referenced.
3991   ** Generate error messages for unused labels and duplicate labels.
3992   */
3993   for(i=0; i<rp->nrhs; i++){
3994     if( rp->rhsalias[i] ){
3995       if( i>0 ){
3996         int j;
3997         if( rp->lhsalias && strcmp(rp->lhsalias,rp->rhsalias[i])==0 ){
3998           ErrorMsg(lemp->filename,rp->ruleline,
3999             "%s(%s) has the same label as the LHS but is not the left-most "
4000             "symbol on the RHS.",
4001             rp->rhs[i]->name, rp->rhsalias[i]);
4002           lemp->errorcnt++;
4003         }
4004         for(j=0; j<i; j++){
4005           if( rp->rhsalias[j] && strcmp(rp->rhsalias[j],rp->rhsalias[i])==0 ){
4006             ErrorMsg(lemp->filename,rp->ruleline,
4007               "Label %s used for multiple symbols on the RHS of a rule.",
4008               rp->rhsalias[i]);
4009             lemp->errorcnt++;
4010             break;
4011           }
4012         }
4013       }
4014       if( !used[i] ){
4015         ErrorMsg(lemp->filename,rp->ruleline,
4016           "Label %s for \"%s(%s)\" is never used.",
4017           rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
4018         lemp->errorcnt++;
4019       }
4020     }else if( i>0 && has_destructor(rp->rhs[i],lemp) ){
4021       append_str("  yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
4022          rp->rhs[i]->index,i-rp->nrhs+1);
4023     }
4024   }
4025 
4026   /* If unable to write LHS values directly into the stack, write the
4027   ** saved LHS value now. */
4028   if( lhsdirect==0 ){
4029     append_str("  yymsp[%d].minor.yy%d = ", 0, 1-rp->nrhs, rp->lhs->dtnum);
4030     append_str(zLhs, 0, 0, 0);
4031     append_str(";\n", 0, 0, 0);
4032   }
4033 
4034   /* Suffix code generation complete */
4035   cp = append_str(0,0,0,0);
4036   if( cp && cp[0] ){
4037     rp->codeSuffix = Strsafe(cp);
4038     rp->noCode = 0;
4039   }
4040 
4041   return rc;
4042 }
4043 
4044 /*
4045 ** Generate code which executes when the rule "rp" is reduced.  Write
4046 ** the code to "out".  Make sure lineno stays up-to-date.
4047 */
emit_code(FILE * out,struct rule * rp,struct lemon * lemp,int * lineno)4048 PRIVATE void emit_code(
4049   FILE *out,
4050   struct rule *rp,
4051   struct lemon *lemp,
4052   int *lineno
4053 ){
4054  const char *cp;
4055 
4056  /* Setup code prior to the #line directive */
4057  if( rp->codePrefix && rp->codePrefix[0] ){
4058    fprintf(out, "{%s", rp->codePrefix);
4059    for(cp=rp->codePrefix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4060  }
4061 
4062  /* Generate code to do the reduce action */
4063  if( rp->code ){
4064    if( !lemp->nolinenosflag ){
4065      (*lineno)++;
4066      tplt_linedir(out,rp->line,lemp->filename);
4067    }
4068    fprintf(out,"{%s",rp->code);
4069    for(cp=rp->code; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4070    fprintf(out,"}\n"); (*lineno)++;
4071    if( !lemp->nolinenosflag ){
4072      (*lineno)++;
4073      tplt_linedir(out,*lineno,lemp->outname);
4074    }
4075  }
4076 
4077  /* Generate breakdown code that occurs after the #line directive */
4078  if( rp->codeSuffix && rp->codeSuffix[0] ){
4079    fprintf(out, "%s", rp->codeSuffix);
4080    for(cp=rp->codeSuffix; *cp; cp++){ if( *cp=='\n' ) (*lineno)++; }
4081  }
4082 
4083  if( rp->codePrefix ){
4084    fprintf(out, "}\n"); (*lineno)++;
4085  }
4086 
4087  return;
4088 }
4089 
4090 /*
4091 ** Print the definition of the union used for the parser's data stack.
4092 ** This union contains fields for every possible data type for tokens
4093 ** and nonterminals.  In the process of computing and printing this
4094 ** union, also set the ".dtnum" field of every terminal and nonterminal
4095 ** symbol.
4096 */
print_stack_union(FILE * out,struct lemon * lemp,int * plineno,int mhflag)4097 void print_stack_union(
4098   FILE *out,                  /* The output stream */
4099   struct lemon *lemp,         /* The main info structure for this parser */
4100   int *plineno,               /* Pointer to the line number */
4101   int mhflag                  /* True if generating makeheaders output */
4102 ){
4103   int lineno;               /* The line number of the output */
4104   char **types;             /* A hash table of datatypes */
4105   int arraysize;            /* Size of the "types" array */
4106   int maxdtlength;          /* Maximum length of any ".datatype" field. */
4107   char *stddt;              /* Standardized name for a datatype */
4108   int i,j;                  /* Loop counters */
4109   unsigned hash;            /* For hashing the name of a type */
4110   const char *name;         /* Name of the parser */
4111 
4112   /* Allocate and initialize types[] and allocate stddt[] */
4113   arraysize = lemp->nsymbol * 2;
4114   types = (char**)calloc( arraysize, sizeof(char*) );
4115   if( types==0 ){
4116     fprintf(stderr,"Out of memory.\n");
4117     exit(1);
4118   }
4119   for(i=0; i<arraysize; i++) types[i] = 0;
4120   maxdtlength = 0;
4121   if( lemp->vartype ){
4122     maxdtlength = lemonStrlen(lemp->vartype);
4123   }
4124   for(i=0; i<lemp->nsymbol; i++){
4125     int len;
4126     struct symbol *sp = lemp->symbols[i];
4127     if( sp->datatype==0 ) continue;
4128     len = lemonStrlen(sp->datatype);
4129     if( len>maxdtlength ) maxdtlength = len;
4130   }
4131   stddt = (char*)malloc( maxdtlength*2 + 1 );
4132   if( stddt==0 ){
4133     fprintf(stderr,"Out of memory.\n");
4134     exit(1);
4135   }
4136 
4137   /* Build a hash table of datatypes. The ".dtnum" field of each symbol
4138   ** is filled in with the hash index plus 1.  A ".dtnum" value of 0 is
4139   ** used for terminal symbols.  If there is no %default_type defined then
4140   ** 0 is also used as the .dtnum value for nonterminals which do not specify
4141   ** a datatype using the %type directive.
4142   */
4143   for(i=0; i<lemp->nsymbol; i++){
4144     struct symbol *sp = lemp->symbols[i];
4145     char *cp;
4146     if( sp==lemp->errsym ){
4147       sp->dtnum = arraysize+1;
4148       continue;
4149     }
4150     if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
4151       sp->dtnum = 0;
4152       continue;
4153     }
4154     cp = sp->datatype;
4155     if( cp==0 ) cp = lemp->vartype;
4156     j = 0;
4157     while( ISSPACE(*cp) ) cp++;
4158     while( *cp ) stddt[j++] = *cp++;
4159     while( j>0 && ISSPACE(stddt[j-1]) ) j--;
4160     stddt[j] = 0;
4161     if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){
4162       sp->dtnum = 0;
4163       continue;
4164     }
4165     hash = 0;
4166     for(j=0; stddt[j]; j++){
4167       hash = hash*53 + stddt[j];
4168     }
4169     hash = (hash & 0x7fffffff)%arraysize;
4170     while( types[hash] ){
4171       if( strcmp(types[hash],stddt)==0 ){
4172         sp->dtnum = hash + 1;
4173         break;
4174       }
4175       hash++;
4176       if( hash>=(unsigned)arraysize ) hash = 0;
4177     }
4178     if( types[hash]==0 ){
4179       sp->dtnum = hash + 1;
4180       types[hash] = (char*)malloc( lemonStrlen(stddt)+1 );
4181       if( types[hash]==0 ){
4182         fprintf(stderr,"Out of memory.\n");
4183         exit(1);
4184       }
4185       lemon_strcpy(types[hash],stddt);
4186     }
4187   }
4188 
4189   /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
4190   name = lemp->name ? lemp->name : "Parse";
4191   lineno = *plineno;
4192   if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
4193   fprintf(out,"#define %sTOKENTYPE %s\n",name,
4194     lemp->tokentype?lemp->tokentype:"void*");  lineno++;
4195   if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
4196   fprintf(out,"typedef union {\n"); lineno++;
4197   fprintf(out,"  int yyinit;\n"); lineno++;
4198   fprintf(out,"  %sTOKENTYPE yy0;\n",name); lineno++;
4199   for(i=0; i<arraysize; i++){
4200     if( types[i]==0 ) continue;
4201     fprintf(out,"  %s yy%d;\n",types[i],i+1); lineno++;
4202     free(types[i]);
4203   }
4204   if( lemp->errsym && lemp->errsym->useCnt ){
4205     fprintf(out,"  int yy%d;\n",lemp->errsym->dtnum); lineno++;
4206   }
4207   free(stddt);
4208   free(types);
4209   fprintf(out,"} YYMINORTYPE;\n"); lineno++;
4210   *plineno = lineno;
4211 }
4212 
4213 /*
4214 ** Return the name of a C datatype able to represent values between
4215 ** lwr and upr, inclusive.  If pnByte!=NULL then also write the sizeof
4216 ** for that type (1, 2, or 4) into *pnByte.
4217 */
minimum_size_type(int lwr,int upr,int * pnByte)4218 static const char *minimum_size_type(int lwr, int upr, int *pnByte){
4219   const char *zType = "int";
4220   int nByte = 4;
4221   if( lwr>=0 ){
4222     if( upr<=255 ){
4223       zType = "unsigned char";
4224       nByte = 1;
4225     }else if( upr<65535 ){
4226       zType = "unsigned short int";
4227       nByte = 2;
4228     }else{
4229       zType = "unsigned int";
4230       nByte = 4;
4231     }
4232   }else if( lwr>=-127 && upr<=127 ){
4233     zType = "signed char";
4234     nByte = 1;
4235   }else if( lwr>=-32767 && upr<32767 ){
4236     zType = "short";
4237     nByte = 2;
4238   }
4239   if( pnByte ) *pnByte = nByte;
4240   return zType;
4241 }
4242 
4243 /*
4244 ** Each state contains a set of token transaction and a set of
4245 ** nonterminal transactions.  Each of these sets makes an instance
4246 ** of the following structure.  An array of these structures is used
4247 ** to order the creation of entries in the yy_action[] table.
4248 */
4249 struct axset {
4250   struct state *stp;   /* A pointer to a state */
4251   int isTkn;           /* True to use tokens.  False for non-terminals */
4252   int nAction;         /* Number of actions */
4253   int iOrder;          /* Original order of action sets */
4254 };
4255 
4256 /*
4257 ** Compare to axset structures for sorting purposes
4258 */
axset_compare(const void * a,const void * b)4259 static int axset_compare(const void *a, const void *b){
4260   struct axset *p1 = (struct axset*)a;
4261   struct axset *p2 = (struct axset*)b;
4262   int c;
4263   c = p2->nAction - p1->nAction;
4264   if( c==0 ){
4265     c = p1->iOrder - p2->iOrder;
4266   }
4267   assert( c!=0 || p1==p2 );
4268   return c;
4269 }
4270 
4271 /*
4272 ** Write text on "out" that describes the rule "rp".
4273 */
writeRuleText(FILE * out,struct rule * rp)4274 static void writeRuleText(FILE *out, struct rule *rp){
4275   int j;
4276   fprintf(out,"%s ::=", rp->lhs->name);
4277   for(j=0; j<rp->nrhs; j++){
4278     struct symbol *sp = rp->rhs[j];
4279     if( sp->type!=MULTITERMINAL ){
4280       fprintf(out," %s", sp->name);
4281     }else{
4282       int k;
4283       fprintf(out," %s", sp->subsym[0]->name);
4284       for(k=1; k<sp->nsubsym; k++){
4285         fprintf(out,"|%s",sp->subsym[k]->name);
4286       }
4287     }
4288   }
4289 }
4290 
4291 
4292 /* Generate C source code for the parser */
ReportTable(struct lemon * lemp,int mhflag,int sqlFlag)4293 void ReportTable(
4294   struct lemon *lemp,
4295   int mhflag,     /* Output in makeheaders format if true */
4296   int sqlFlag     /* Generate the *.sql file too */
4297 ){
4298   FILE *out, *in, *sql;
4299   char line[LINESIZE];
4300   int  lineno;
4301   struct state *stp;
4302   struct action *ap;
4303   struct rule *rp;
4304   struct acttab *pActtab;
4305   int i, j, n, sz;
4306   int nLookAhead;
4307   int szActionType;     /* sizeof(YYACTIONTYPE) */
4308   int szCodeType;       /* sizeof(YYCODETYPE)   */
4309   const char *name;
4310   int mnTknOfst, mxTknOfst;
4311   int mnNtOfst, mxNtOfst;
4312   struct axset *ax;
4313   char *prefix;
4314 
4315   lemp->minShiftReduce = lemp->nstate;
4316   lemp->errAction = lemp->minShiftReduce + lemp->nrule;
4317   lemp->accAction = lemp->errAction + 1;
4318   lemp->noAction = lemp->accAction + 1;
4319   lemp->minReduce = lemp->noAction + 1;
4320   lemp->maxAction = lemp->minReduce + lemp->nrule;
4321 
4322   in = tplt_open(lemp);
4323   if( in==0 ) return;
4324   out = file_open(lemp,".c","wb");
4325   if( out==0 ){
4326     fclose(in);
4327     return;
4328   }
4329   if( sqlFlag==0 ){
4330     sql = 0;
4331   }else{
4332     sql = file_open(lemp, ".sql", "wb");
4333     if( sql==0 ){
4334       fclose(in);
4335       fclose(out);
4336       return;
4337     }
4338     fprintf(sql,
4339        "BEGIN;\n"
4340        "CREATE TABLE symbol(\n"
4341        "  id INTEGER PRIMARY KEY,\n"
4342        "  name TEXT NOT NULL,\n"
4343        "  isTerminal BOOLEAN NOT NULL,\n"
4344        "  fallback INTEGER REFERENCES symbol"
4345                " DEFERRABLE INITIALLY DEFERRED\n"
4346        ");\n"
4347     );
4348     for(i=0; i<lemp->nsymbol; i++){
4349       fprintf(sql,
4350          "INSERT INTO symbol(id,name,isTerminal,fallback)"
4351          "VALUES(%d,'%s',%s",
4352          i, lemp->symbols[i]->name,
4353          i<lemp->nterminal ? "TRUE" : "FALSE"
4354       );
4355       if( lemp->symbols[i]->fallback ){
4356         fprintf(sql, ",%d);\n", lemp->symbols[i]->fallback->index);
4357       }else{
4358         fprintf(sql, ",NULL);\n");
4359       }
4360     }
4361     fprintf(sql,
4362       "CREATE TABLE rule(\n"
4363       "  ruleid INTEGER PRIMARY KEY,\n"
4364       "  lhs INTEGER REFERENCES symbol(id),\n"
4365       "  txt TEXT\n"
4366       ");\n"
4367       "CREATE TABLE rulerhs(\n"
4368       "  ruleid INTEGER REFERENCES rule(ruleid),\n"
4369       "  pos INTEGER,\n"
4370       "  sym INTEGER REFERENCES symbol(id)\n"
4371       ");\n"
4372     );
4373     for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4374       assert( i==rp->iRule );
4375       fprintf(sql,
4376         "INSERT INTO rule(ruleid,lhs,txt)VALUES(%d,%d,'",
4377         rp->iRule, rp->lhs->index
4378       );
4379       writeRuleText(sql, rp);
4380       fprintf(sql,"');\n");
4381       for(j=0; j<rp->nrhs; j++){
4382         struct symbol *sp = rp->rhs[j];
4383         if( sp->type!=MULTITERMINAL ){
4384           fprintf(sql,
4385             "INSERT INTO rulerhs(ruleid,pos,sym)VALUES(%d,%d,%d);\n",
4386             i,j,sp->index
4387           );
4388         }else{
4389           int k;
4390           for(k=0; k<sp->nsubsym; k++){
4391             fprintf(sql,
4392               "INSERT INTO rulerhs(ruleid,pos,sym)VALUES(%d,%d,%d);\n",
4393               i,j,sp->subsym[k]->index
4394             );
4395           }
4396         }
4397       }
4398     }
4399     fprintf(sql, "COMMIT;\n");
4400   }
4401   lineno = 1;
4402 
4403   fprintf(out,
4404      "/* This file is automatically generated by Lemon from input grammar\n"
4405      "** source file \"%s\". */\n", lemp->filename); lineno += 2;
4406 
4407   /* The first %include directive begins with a C-language comment,
4408   ** then skip over the header comment of the template file
4409   */
4410   if( lemp->include==0 ) lemp->include = "";
4411   for(i=0; ISSPACE(lemp->include[i]); i++){
4412     if( lemp->include[i]=='\n' ){
4413       lemp->include += i+1;
4414       i = -1;
4415     }
4416   }
4417   if( lemp->include[0]=='/' ){
4418     tplt_skip_header(in,&lineno);
4419   }else{
4420     tplt_xfer(lemp->name,in,out,&lineno);
4421   }
4422 
4423   /* Generate the include code, if any */
4424   tplt_print(out,lemp,lemp->include,&lineno);
4425   if( mhflag ){
4426     char *incName = file_makename(lemp, ".h");
4427     fprintf(out,"#include \"%s\"\n", incName); lineno++;
4428     free(incName);
4429   }
4430   tplt_xfer(lemp->name,in,out,&lineno);
4431 
4432   /* Generate #defines for all tokens */
4433   if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4434   else                    prefix = "";
4435   if( mhflag ){
4436     fprintf(out,"#if INTERFACE\n"); lineno++;
4437   }else{
4438     fprintf(out,"#ifndef %s%s\n", prefix, lemp->symbols[1]->name);
4439   }
4440   for(i=1; i<lemp->nterminal; i++){
4441     fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
4442     lineno++;
4443   }
4444   fprintf(out,"#endif\n"); lineno++;
4445   tplt_xfer(lemp->name,in,out,&lineno);
4446 
4447   /* Generate the defines */
4448   fprintf(out,"#define YYCODETYPE %s\n",
4449     minimum_size_type(0, lemp->nsymbol, &szCodeType)); lineno++;
4450   fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol);  lineno++;
4451   fprintf(out,"#define YYACTIONTYPE %s\n",
4452     minimum_size_type(0,lemp->maxAction,&szActionType)); lineno++;
4453   if( lemp->wildcard ){
4454     fprintf(out,"#define YYWILDCARD %d\n",
4455        lemp->wildcard->index); lineno++;
4456   }
4457   print_stack_union(out,lemp,&lineno,mhflag);
4458   fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++;
4459   if( lemp->stacksize ){
4460     fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize);  lineno++;
4461   }else{
4462     fprintf(out,"#define YYSTACKDEPTH 100\n");  lineno++;
4463   }
4464   fprintf(out, "#endif\n"); lineno++;
4465   if( mhflag ){
4466     fprintf(out,"#if INTERFACE\n"); lineno++;
4467   }
4468   name = lemp->name ? lemp->name : "Parse";
4469   if( lemp->arg && lemp->arg[0] ){
4470     i = lemonStrlen(lemp->arg);
4471     while( i>=1 && ISSPACE(lemp->arg[i-1]) ) i--;
4472     while( i>=1 && (ISALNUM(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
4473     fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg);  lineno++;
4474     fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg);  lineno++;
4475     fprintf(out,"#define %sARG_PARAM ,%s\n",name,&lemp->arg[i]);  lineno++;
4476     fprintf(out,"#define %sARG_FETCH %s=yypParser->%s;\n",
4477                  name,lemp->arg,&lemp->arg[i]);  lineno++;
4478     fprintf(out,"#define %sARG_STORE yypParser->%s=%s;\n",
4479                  name,&lemp->arg[i],&lemp->arg[i]);  lineno++;
4480   }else{
4481     fprintf(out,"#define %sARG_SDECL\n",name); lineno++;
4482     fprintf(out,"#define %sARG_PDECL\n",name); lineno++;
4483     fprintf(out,"#define %sARG_PARAM\n",name); lineno++;
4484     fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
4485     fprintf(out,"#define %sARG_STORE\n",name); lineno++;
4486   }
4487   if( lemp->ctx && lemp->ctx[0] ){
4488     i = lemonStrlen(lemp->ctx);
4489     while( i>=1 && ISSPACE(lemp->ctx[i-1]) ) i--;
4490     while( i>=1 && (ISALNUM(lemp->ctx[i-1]) || lemp->ctx[i-1]=='_') ) i--;
4491     fprintf(out,"#define %sCTX_SDECL %s;\n",name,lemp->ctx);  lineno++;
4492     fprintf(out,"#define %sCTX_PDECL ,%s\n",name,lemp->ctx);  lineno++;
4493     fprintf(out,"#define %sCTX_PARAM ,%s\n",name,&lemp->ctx[i]);  lineno++;
4494     fprintf(out,"#define %sCTX_FETCH %s=yypParser->%s;\n",
4495                  name,lemp->ctx,&lemp->ctx[i]);  lineno++;
4496     fprintf(out,"#define %sCTX_STORE yypParser->%s=%s;\n",
4497                  name,&lemp->ctx[i],&lemp->ctx[i]);  lineno++;
4498   }else{
4499     fprintf(out,"#define %sCTX_SDECL\n",name); lineno++;
4500     fprintf(out,"#define %sCTX_PDECL\n",name); lineno++;
4501     fprintf(out,"#define %sCTX_PARAM\n",name); lineno++;
4502     fprintf(out,"#define %sCTX_FETCH\n",name); lineno++;
4503     fprintf(out,"#define %sCTX_STORE\n",name); lineno++;
4504   }
4505   if( mhflag ){
4506     fprintf(out,"#endif\n"); lineno++;
4507   }
4508   if( lemp->errsym && lemp->errsym->useCnt ){
4509     fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++;
4510     fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++;
4511   }
4512   if( lemp->has_fallback ){
4513     fprintf(out,"#define YYFALLBACK 1\n");  lineno++;
4514   }
4515 
4516   /* Compute the action table, but do not output it yet.  The action
4517   ** table must be computed before generating the YYNSTATE macro because
4518   ** we need to know how many states can be eliminated.
4519   */
4520   ax = (struct axset *) calloc(lemp->nxstate*2, sizeof(ax[0]));
4521   if( ax==0 ){
4522     fprintf(stderr,"malloc failed\n");
4523     exit(1);
4524   }
4525   for(i=0; i<lemp->nxstate; i++){
4526     stp = lemp->sorted[i];
4527     ax[i*2].stp = stp;
4528     ax[i*2].isTkn = 1;
4529     ax[i*2].nAction = stp->nTknAct;
4530     ax[i*2+1].stp = stp;
4531     ax[i*2+1].isTkn = 0;
4532     ax[i*2+1].nAction = stp->nNtAct;
4533   }
4534   mxTknOfst = mnTknOfst = 0;
4535   mxNtOfst = mnNtOfst = 0;
4536   /* In an effort to minimize the action table size, use the heuristic
4537   ** of placing the largest action sets first */
4538   for(i=0; i<lemp->nxstate*2; i++) ax[i].iOrder = i;
4539   qsort(ax, lemp->nxstate*2, sizeof(ax[0]), axset_compare);
4540   pActtab = acttab_alloc(lemp->nsymbol, lemp->nterminal);
4541   for(i=0; i<lemp->nxstate*2 && ax[i].nAction>0; i++){
4542     stp = ax[i].stp;
4543     if( ax[i].isTkn ){
4544       for(ap=stp->ap; ap; ap=ap->next){
4545         int action;
4546         if( ap->sp->index>=lemp->nterminal ) continue;
4547         action = compute_action(lemp, ap);
4548         if( action<0 ) continue;
4549         acttab_action(pActtab, ap->sp->index, action);
4550       }
4551       stp->iTknOfst = acttab_insert(pActtab, 1);
4552       if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
4553       if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
4554     }else{
4555       for(ap=stp->ap; ap; ap=ap->next){
4556         int action;
4557         if( ap->sp->index<lemp->nterminal ) continue;
4558         if( ap->sp->index==lemp->nsymbol ) continue;
4559         action = compute_action(lemp, ap);
4560         if( action<0 ) continue;
4561         acttab_action(pActtab, ap->sp->index, action);
4562       }
4563       stp->iNtOfst = acttab_insert(pActtab, 0);
4564       if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
4565       if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
4566     }
4567 #if 0  /* Uncomment for a trace of how the yy_action[] table fills out */
4568     { int jj, nn;
4569       for(jj=nn=0; jj<pActtab->nAction; jj++){
4570         if( pActtab->aAction[jj].action<0 ) nn++;
4571       }
4572       printf("%4d: State %3d %s n: %2d size: %5d freespace: %d\n",
4573              i, stp->statenum, ax[i].isTkn ? "Token" : "Var  ",
4574              ax[i].nAction, pActtab->nAction, nn);
4575     }
4576 #endif
4577   }
4578   free(ax);
4579 
4580   /* Mark rules that are actually used for reduce actions after all
4581   ** optimizations have been applied
4582   */
4583   for(rp=lemp->rule; rp; rp=rp->next) rp->doesReduce = LEMON_FALSE;
4584   for(i=0; i<lemp->nxstate; i++){
4585     for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
4586       if( ap->type==REDUCE || ap->type==SHIFTREDUCE ){
4587         ap->x.rp->doesReduce = 1;
4588       }
4589     }
4590   }
4591 
4592   /* Finish rendering the constants now that the action table has
4593   ** been computed */
4594   fprintf(out,"#define YYNSTATE             %d\n",lemp->nxstate);  lineno++;
4595   fprintf(out,"#define YYNRULE              %d\n",lemp->nrule);  lineno++;
4596   fprintf(out,"#define YYNRULE_WITH_ACTION  %d\n",lemp->nruleWithAction);
4597          lineno++;
4598   fprintf(out,"#define YYNTOKEN             %d\n",lemp->nterminal); lineno++;
4599   fprintf(out,"#define YY_MAX_SHIFT         %d\n",lemp->nxstate-1); lineno++;
4600   i = lemp->minShiftReduce;
4601   fprintf(out,"#define YY_MIN_SHIFTREDUCE   %d\n",i); lineno++;
4602   i += lemp->nrule;
4603   fprintf(out,"#define YY_MAX_SHIFTREDUCE   %d\n", i-1); lineno++;
4604   fprintf(out,"#define YY_ERROR_ACTION      %d\n", lemp->errAction); lineno++;
4605   fprintf(out,"#define YY_ACCEPT_ACTION     %d\n", lemp->accAction); lineno++;
4606   fprintf(out,"#define YY_NO_ACTION         %d\n", lemp->noAction); lineno++;
4607   fprintf(out,"#define YY_MIN_REDUCE        %d\n", lemp->minReduce); lineno++;
4608   i = lemp->minReduce + lemp->nrule;
4609   fprintf(out,"#define YY_MAX_REDUCE        %d\n", i-1); lineno++;
4610   tplt_xfer(lemp->name,in,out,&lineno);
4611 
4612   /* Now output the action table and its associates:
4613   **
4614   **  yy_action[]        A single table containing all actions.
4615   **  yy_lookahead[]     A table containing the lookahead for each entry in
4616   **                     yy_action.  Used to detect hash collisions.
4617   **  yy_shift_ofst[]    For each state, the offset into yy_action for
4618   **                     shifting terminals.
4619   **  yy_reduce_ofst[]   For each state, the offset into yy_action for
4620   **                     shifting non-terminals after a reduce.
4621   **  yy_default[]       Default action for each state.
4622   */
4623 
4624   /* Output the yy_action table */
4625   lemp->nactiontab = n = acttab_action_size(pActtab);
4626   lemp->tablesize += n*szActionType;
4627   fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++;
4628   fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++;
4629   for(i=j=0; i<n; i++){
4630     int action = acttab_yyaction(pActtab, i);
4631     if( action<0 ) action = lemp->noAction;
4632     if( j==0 ) fprintf(out," /* %5d */ ", i);
4633     fprintf(out, " %4d,", action);
4634     if( j==9 || i==n-1 ){
4635       fprintf(out, "\n"); lineno++;
4636       j = 0;
4637     }else{
4638       j++;
4639     }
4640   }
4641   fprintf(out, "};\n"); lineno++;
4642 
4643   /* Output the yy_lookahead table */
4644   lemp->nlookaheadtab = n = acttab_lookahead_size(pActtab);
4645   lemp->tablesize += n*szCodeType;
4646   fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++;
4647   for(i=j=0; i<n; i++){
4648     int la = acttab_yylookahead(pActtab, i);
4649     if( la<0 ) la = lemp->nsymbol;
4650     if( j==0 ) fprintf(out," /* %5d */ ", i);
4651     fprintf(out, " %4d,", la);
4652     if( j==9 ){
4653       fprintf(out, "\n"); lineno++;
4654       j = 0;
4655     }else{
4656       j++;
4657     }
4658   }
4659   /* Add extra entries to the end of the yy_lookahead[] table so that
4660   ** yy_shift_ofst[]+iToken will always be a valid index into the array,
4661   ** even for the largest possible value of yy_shift_ofst[] and iToken. */
4662   nLookAhead = lemp->nterminal + lemp->nactiontab;
4663   while( i<nLookAhead ){
4664     if( j==0 ) fprintf(out," /* %5d */ ", i);
4665     fprintf(out, " %4d,", lemp->nterminal);
4666     if( j==9 ){
4667       fprintf(out, "\n"); lineno++;
4668       j = 0;
4669     }else{
4670       j++;
4671     }
4672     i++;
4673   }
4674   if( j>0 ){ fprintf(out, "\n"); lineno++; }
4675   fprintf(out, "};\n"); lineno++;
4676 
4677   /* Output the yy_shift_ofst[] table */
4678   n = lemp->nxstate;
4679   while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--;
4680   fprintf(out, "#define YY_SHIFT_COUNT    (%d)\n", n-1); lineno++;
4681   fprintf(out, "#define YY_SHIFT_MIN      (%d)\n", mnTknOfst); lineno++;
4682   fprintf(out, "#define YY_SHIFT_MAX      (%d)\n", mxTknOfst); lineno++;
4683   fprintf(out, "static const %s yy_shift_ofst[] = {\n",
4684        minimum_size_type(mnTknOfst, lemp->nterminal+lemp->nactiontab, &sz));
4685        lineno++;
4686   lemp->tablesize += n*sz;
4687   for(i=j=0; i<n; i++){
4688     int ofst;
4689     stp = lemp->sorted[i];
4690     ofst = stp->iTknOfst;
4691     if( ofst==NO_OFFSET ) ofst = lemp->nactiontab;
4692     if( j==0 ) fprintf(out," /* %5d */ ", i);
4693     fprintf(out, " %4d,", ofst);
4694     if( j==9 || i==n-1 ){
4695       fprintf(out, "\n"); lineno++;
4696       j = 0;
4697     }else{
4698       j++;
4699     }
4700   }
4701   fprintf(out, "};\n"); lineno++;
4702 
4703   /* Output the yy_reduce_ofst[] table */
4704   n = lemp->nxstate;
4705   while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--;
4706   fprintf(out, "#define YY_REDUCE_COUNT (%d)\n", n-1); lineno++;
4707   fprintf(out, "#define YY_REDUCE_MIN   (%d)\n", mnNtOfst); lineno++;
4708   fprintf(out, "#define YY_REDUCE_MAX   (%d)\n", mxNtOfst); lineno++;
4709   fprintf(out, "static const %s yy_reduce_ofst[] = {\n",
4710           minimum_size_type(mnNtOfst-1, mxNtOfst, &sz)); lineno++;
4711   lemp->tablesize += n*sz;
4712   for(i=j=0; i<n; i++){
4713     int ofst;
4714     stp = lemp->sorted[i];
4715     ofst = stp->iNtOfst;
4716     if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
4717     if( j==0 ) fprintf(out," /* %5d */ ", i);
4718     fprintf(out, " %4d,", ofst);
4719     if( j==9 || i==n-1 ){
4720       fprintf(out, "\n"); lineno++;
4721       j = 0;
4722     }else{
4723       j++;
4724     }
4725   }
4726   fprintf(out, "};\n"); lineno++;
4727 
4728   /* Output the default action table */
4729   fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++;
4730   n = lemp->nxstate;
4731   lemp->tablesize += n*szActionType;
4732   for(i=j=0; i<n; i++){
4733     stp = lemp->sorted[i];
4734     if( j==0 ) fprintf(out," /* %5d */ ", i);
4735     if( stp->iDfltReduce<0 ){
4736       fprintf(out, " %4d,", lemp->errAction);
4737     }else{
4738       fprintf(out, " %4d,", stp->iDfltReduce + lemp->minReduce);
4739     }
4740     if( j==9 || i==n-1 ){
4741       fprintf(out, "\n"); lineno++;
4742       j = 0;
4743     }else{
4744       j++;
4745     }
4746   }
4747   fprintf(out, "};\n"); lineno++;
4748   tplt_xfer(lemp->name,in,out,&lineno);
4749 
4750   /* Generate the table of fallback tokens.
4751   */
4752   if( lemp->has_fallback ){
4753     int mx = lemp->nterminal - 1;
4754     /* 2019-08-28:  Generate fallback entries for every token to avoid
4755     ** having to do a range check on the index */
4756     /* while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; } */
4757     lemp->tablesize += (mx+1)*szCodeType;
4758     for(i=0; i<=mx; i++){
4759       struct symbol *p = lemp->symbols[i];
4760       if( p->fallback==0 ){
4761         fprintf(out, "    0,  /* %10s => nothing */\n", p->name);
4762       }else{
4763         fprintf(out, "  %3d,  /* %10s => %s */\n", p->fallback->index,
4764           p->name, p->fallback->name);
4765       }
4766       lineno++;
4767     }
4768   }
4769   tplt_xfer(lemp->name, in, out, &lineno);
4770 
4771   /* Generate a table containing the symbolic name of every symbol
4772   */
4773   for(i=0; i<lemp->nsymbol; i++){
4774     lemon_sprintf(line,"\"%s\",",lemp->symbols[i]->name);
4775     fprintf(out,"  /* %4d */ \"%s\",\n",i, lemp->symbols[i]->name); lineno++;
4776   }
4777   tplt_xfer(lemp->name,in,out,&lineno);
4778 
4779   /* Generate a table containing a text string that describes every
4780   ** rule in the rule set of the grammar.  This information is used
4781   ** when tracing REDUCE actions.
4782   */
4783   for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4784     assert( rp->iRule==i );
4785     fprintf(out," /* %3d */ \"", i);
4786     writeRuleText(out, rp);
4787     fprintf(out,"\",\n"); lineno++;
4788   }
4789   tplt_xfer(lemp->name,in,out,&lineno);
4790 
4791   /* Generate code which executes every time a symbol is popped from
4792   ** the stack while processing errors or while destroying the parser.
4793   ** (In other words, generate the %destructor actions)
4794   */
4795   if( lemp->tokendest ){
4796     int once = 1;
4797     for(i=0; i<lemp->nsymbol; i++){
4798       struct symbol *sp = lemp->symbols[i];
4799       if( sp==0 || sp->type!=TERMINAL ) continue;
4800       if( once ){
4801         fprintf(out, "      /* TERMINAL Destructor */\n"); lineno++;
4802         once = 0;
4803       }
4804       fprintf(out,"    case %d: /* %s */\n", sp->index, sp->name); lineno++;
4805     }
4806     for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
4807     if( i<lemp->nsymbol ){
4808       emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4809       fprintf(out,"      break;\n"); lineno++;
4810     }
4811   }
4812   if( lemp->vardest ){
4813     struct symbol *dflt_sp = 0;
4814     int once = 1;
4815     for(i=0; i<lemp->nsymbol; i++){
4816       struct symbol *sp = lemp->symbols[i];
4817       if( sp==0 || sp->type==TERMINAL ||
4818           sp->index<=0 || sp->destructor!=0 ) continue;
4819       if( once ){
4820         fprintf(out, "      /* Default NON-TERMINAL Destructor */\n");lineno++;
4821         once = 0;
4822       }
4823       fprintf(out,"    case %d: /* %s */\n", sp->index, sp->name); lineno++;
4824       dflt_sp = sp;
4825     }
4826     if( dflt_sp!=0 ){
4827       emit_destructor_code(out,dflt_sp,lemp,&lineno);
4828     }
4829     fprintf(out,"      break;\n"); lineno++;
4830   }
4831   for(i=0; i<lemp->nsymbol; i++){
4832     struct symbol *sp = lemp->symbols[i];
4833     if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
4834     if( sp->destLineno<0 ) continue;  /* Already emitted */
4835     fprintf(out,"    case %d: /* %s */\n", sp->index, sp->name); lineno++;
4836 
4837     /* Combine duplicate destructors into a single case */
4838     for(j=i+1; j<lemp->nsymbol; j++){
4839       struct symbol *sp2 = lemp->symbols[j];
4840       if( sp2 && sp2->type!=TERMINAL && sp2->destructor
4841           && sp2->dtnum==sp->dtnum
4842           && strcmp(sp->destructor,sp2->destructor)==0 ){
4843          fprintf(out,"    case %d: /* %s */\n",
4844                  sp2->index, sp2->name); lineno++;
4845          sp2->destLineno = -1;  /* Avoid emitting this destructor again */
4846       }
4847     }
4848 
4849     emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
4850     fprintf(out,"      break;\n"); lineno++;
4851   }
4852   tplt_xfer(lemp->name,in,out,&lineno);
4853 
4854   /* Generate code which executes whenever the parser stack overflows */
4855   tplt_print(out,lemp,lemp->overflow,&lineno);
4856   tplt_xfer(lemp->name,in,out,&lineno);
4857 
4858   /* Generate the tables of rule information.  yyRuleInfoLhs[] and
4859   ** yyRuleInfoNRhs[].
4860   **
4861   ** Note: This code depends on the fact that rules are number
4862   ** sequentually beginning with 0.
4863   */
4864   for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4865     fprintf(out,"  %4d,  /* (%d) ", rp->lhs->index, i);
4866      rule_print(out, rp);
4867     fprintf(out," */\n"); lineno++;
4868   }
4869   tplt_xfer(lemp->name,in,out,&lineno);
4870   for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
4871     fprintf(out,"  %3d,  /* (%d) ", -rp->nrhs, i);
4872     rule_print(out, rp);
4873     fprintf(out," */\n"); lineno++;
4874   }
4875   tplt_xfer(lemp->name,in,out,&lineno);
4876 
4877   /* Generate code which execution during each REDUCE action */
4878   i = 0;
4879   for(rp=lemp->rule; rp; rp=rp->next){
4880     i += translate_code(lemp, rp);
4881   }
4882   if( i ){
4883     fprintf(out,"        YYMINORTYPE yylhsminor;\n"); lineno++;
4884   }
4885   /* First output rules other than the default: rule */
4886   for(rp=lemp->rule; rp; rp=rp->next){
4887     struct rule *rp2;               /* Other rules with the same action */
4888     if( rp->codeEmitted ) continue;
4889     if( rp->noCode ){
4890       /* No C code actions, so this will be part of the "default:" rule */
4891       continue;
4892     }
4893     fprintf(out,"      case %d: /* ", rp->iRule);
4894     writeRuleText(out, rp);
4895     fprintf(out, " */\n"); lineno++;
4896     for(rp2=rp->next; rp2; rp2=rp2->next){
4897       if( rp2->code==rp->code && rp2->codePrefix==rp->codePrefix
4898              && rp2->codeSuffix==rp->codeSuffix ){
4899         fprintf(out,"      case %d: /* ", rp2->iRule);
4900         writeRuleText(out, rp2);
4901         fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->iRule); lineno++;
4902         rp2->codeEmitted = 1;
4903       }
4904     }
4905     emit_code(out,rp,lemp,&lineno);
4906     fprintf(out,"        break;\n"); lineno++;
4907     rp->codeEmitted = 1;
4908   }
4909   /* Finally, output the default: rule.  We choose as the default: all
4910   ** empty actions. */
4911   fprintf(out,"      default:\n"); lineno++;
4912   for(rp=lemp->rule; rp; rp=rp->next){
4913     if( rp->codeEmitted ) continue;
4914     assert( rp->noCode );
4915     fprintf(out,"      /* (%d) ", rp->iRule);
4916     writeRuleText(out, rp);
4917     if( rp->neverReduce ){
4918       fprintf(out, " (NEVER REDUCES) */ assert(yyruleno!=%d);\n",
4919               rp->iRule); lineno++;
4920     }else if( rp->doesReduce ){
4921       fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->iRule); lineno++;
4922     }else{
4923       fprintf(out, " (OPTIMIZED OUT) */ assert(yyruleno!=%d);\n",
4924               rp->iRule); lineno++;
4925     }
4926   }
4927   fprintf(out,"        break;\n"); lineno++;
4928   tplt_xfer(lemp->name,in,out,&lineno);
4929 
4930   /* Generate code which executes if a parse fails */
4931   tplt_print(out,lemp,lemp->failure,&lineno);
4932   tplt_xfer(lemp->name,in,out,&lineno);
4933 
4934   /* Generate code which executes when a syntax error occurs */
4935   tplt_print(out,lemp,lemp->error,&lineno);
4936   tplt_xfer(lemp->name,in,out,&lineno);
4937 
4938   /* Generate code which executes when the parser accepts its input */
4939   tplt_print(out,lemp,lemp->accept,&lineno);
4940   tplt_xfer(lemp->name,in,out,&lineno);
4941 
4942   /* Append any addition code the user desires */
4943   tplt_print(out,lemp,lemp->extracode,&lineno);
4944 
4945   acttab_free(pActtab);
4946   fclose(in);
4947   fclose(out);
4948   if( sql ) fclose(sql);
4949   return;
4950 }
4951 
4952 /* Generate a header file for the parser */
ReportHeader(struct lemon * lemp)4953 void ReportHeader(struct lemon *lemp)
4954 {
4955   FILE *out, *in;
4956   const char *prefix;
4957   char line[LINESIZE];
4958   char pattern[LINESIZE];
4959   int i;
4960 
4961   if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4962   else                    prefix = "";
4963   in = file_open(lemp,".h","rb");
4964   if( in ){
4965     int nextChar;
4966     for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
4967       lemon_sprintf(pattern,"#define %s%-30s %3d\n",
4968                     prefix,lemp->symbols[i]->name,i);
4969       if( strcmp(line,pattern) ) break;
4970     }
4971     nextChar = fgetc(in);
4972     fclose(in);
4973     if( i==lemp->nterminal && nextChar==EOF ){
4974       /* No change in the file.  Don't rewrite it. */
4975       return;
4976     }
4977   }
4978   out = file_open(lemp,".h","wb");
4979   if( out ){
4980     for(i=1; i<lemp->nterminal; i++){
4981       fprintf(out,"#define %s%-30s %3d\n",prefix,lemp->symbols[i]->name,i);
4982     }
4983     fclose(out);
4984   }
4985   return;
4986 }
4987 
4988 /* Reduce the size of the action tables, if possible, by making use
4989 ** of defaults.
4990 **
4991 ** In this version, we take the most frequent REDUCE action and make
4992 ** it the default.  Except, there is no default if the wildcard token
4993 ** is a possible look-ahead.
4994 */
CompressTables(struct lemon * lemp)4995 void CompressTables(struct lemon *lemp)
4996 {
4997   struct state *stp;
4998   struct action *ap, *ap2, *nextap;
4999   struct rule *rp, *rp2, *rbest;
5000   int nbest, n;
5001   int i;
5002   int usesWildcard;
5003 
5004   for(i=0; i<lemp->nstate; i++){
5005     stp = lemp->sorted[i];
5006     nbest = 0;
5007     rbest = 0;
5008     usesWildcard = 0;
5009 
5010     for(ap=stp->ap; ap; ap=ap->next){
5011       if( ap->type==SHIFT && ap->sp==lemp->wildcard ){
5012         usesWildcard = 1;
5013       }
5014       if( ap->type!=REDUCE ) continue;
5015       rp = ap->x.rp;
5016       if( rp->lhsStart ) continue;
5017       if( rp==rbest ) continue;
5018       n = 1;
5019       for(ap2=ap->next; ap2; ap2=ap2->next){
5020         if( ap2->type!=REDUCE ) continue;
5021         rp2 = ap2->x.rp;
5022         if( rp2==rbest ) continue;
5023         if( rp2==rp ) n++;
5024       }
5025       if( n>nbest ){
5026         nbest = n;
5027         rbest = rp;
5028       }
5029     }
5030 
5031     /* Do not make a default if the number of rules to default
5032     ** is not at least 1 or if the wildcard token is a possible
5033     ** lookahead.
5034     */
5035     if( nbest<1 || usesWildcard ) continue;
5036 
5037 
5038     /* Combine matching REDUCE actions into a single default */
5039     for(ap=stp->ap; ap; ap=ap->next){
5040       if( ap->type==REDUCE && ap->x.rp==rbest ) break;
5041     }
5042     assert( ap );
5043     ap->sp = Symbol_new("{default}");
5044     for(ap=ap->next; ap; ap=ap->next){
5045       if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
5046     }
5047     stp->ap = Action_sort(stp->ap);
5048 
5049     for(ap=stp->ap; ap; ap=ap->next){
5050       if( ap->type==SHIFT ) break;
5051       if( ap->type==REDUCE && ap->x.rp!=rbest ) break;
5052     }
5053     if( ap==0 ){
5054       stp->autoReduce = 1;
5055       stp->pDfltReduce = rbest;
5056     }
5057   }
5058 
5059   /* Make a second pass over all states and actions.  Convert
5060   ** every action that is a SHIFT to an autoReduce state into
5061   ** a SHIFTREDUCE action.
5062   */
5063   for(i=0; i<lemp->nstate; i++){
5064     stp = lemp->sorted[i];
5065     for(ap=stp->ap; ap; ap=ap->next){
5066       struct state *pNextState;
5067       if( ap->type!=SHIFT ) continue;
5068       pNextState = ap->x.stp;
5069       if( pNextState->autoReduce && pNextState->pDfltReduce!=0 ){
5070         ap->type = SHIFTREDUCE;
5071         ap->x.rp = pNextState->pDfltReduce;
5072       }
5073     }
5074   }
5075 
5076   /* If a SHIFTREDUCE action specifies a rule that has a single RHS term
5077   ** (meaning that the SHIFTREDUCE will land back in the state where it
5078   ** started) and if there is no C-code associated with the reduce action,
5079   ** then we can go ahead and convert the action to be the same as the
5080   ** action for the RHS of the rule.
5081   */
5082   for(i=0; i<lemp->nstate; i++){
5083     stp = lemp->sorted[i];
5084     for(ap=stp->ap; ap; ap=nextap){
5085       nextap = ap->next;
5086       if( ap->type!=SHIFTREDUCE ) continue;
5087       rp = ap->x.rp;
5088       if( rp->noCode==0 ) continue;
5089       if( rp->nrhs!=1 ) continue;
5090 #if 1
5091       /* Only apply this optimization to non-terminals.  It would be OK to
5092       ** apply it to terminal symbols too, but that makes the parser tables
5093       ** larger. */
5094       if( ap->sp->index<lemp->nterminal ) continue;
5095 #endif
5096       /* If we reach this point, it means the optimization can be applied */
5097       nextap = ap;
5098       for(ap2=stp->ap; ap2 && (ap2==ap || ap2->sp!=rp->lhs); ap2=ap2->next){}
5099       assert( ap2!=0 );
5100       ap->spOpt = ap2->sp;
5101       ap->type = ap2->type;
5102       ap->x = ap2->x;
5103     }
5104   }
5105 }
5106 
5107 
5108 /*
5109 ** Compare two states for sorting purposes.  The smaller state is the
5110 ** one with the most non-terminal actions.  If they have the same number
5111 ** of non-terminal actions, then the smaller is the one with the most
5112 ** token actions.
5113 */
stateResortCompare(const void * a,const void * b)5114 static int stateResortCompare(const void *a, const void *b){
5115   const struct state *pA = *(const struct state**)a;
5116   const struct state *pB = *(const struct state**)b;
5117   int n;
5118 
5119   n = pB->nNtAct - pA->nNtAct;
5120   if( n==0 ){
5121     n = pB->nTknAct - pA->nTknAct;
5122     if( n==0 ){
5123       n = pB->statenum - pA->statenum;
5124     }
5125   }
5126   assert( n!=0 );
5127   return n;
5128 }
5129 
5130 
5131 /*
5132 ** Renumber and resort states so that states with fewer choices
5133 ** occur at the end.  Except, keep state 0 as the first state.
5134 */
ResortStates(struct lemon * lemp)5135 void ResortStates(struct lemon *lemp)
5136 {
5137   int i;
5138   struct state *stp;
5139   struct action *ap;
5140 
5141   for(i=0; i<lemp->nstate; i++){
5142     stp = lemp->sorted[i];
5143     stp->nTknAct = stp->nNtAct = 0;
5144     stp->iDfltReduce = -1; /* Init dflt action to "syntax error" */
5145     stp->iTknOfst = NO_OFFSET;
5146     stp->iNtOfst = NO_OFFSET;
5147     for(ap=stp->ap; ap; ap=ap->next){
5148       int iAction = compute_action(lemp,ap);
5149       if( iAction>=0 ){
5150         if( ap->sp->index<lemp->nterminal ){
5151           stp->nTknAct++;
5152         }else if( ap->sp->index<lemp->nsymbol ){
5153           stp->nNtAct++;
5154         }else{
5155           assert( stp->autoReduce==0 || stp->pDfltReduce==ap->x.rp );
5156           stp->iDfltReduce = iAction;
5157         }
5158       }
5159     }
5160   }
5161   qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]),
5162         stateResortCompare);
5163   for(i=0; i<lemp->nstate; i++){
5164     lemp->sorted[i]->statenum = i;
5165   }
5166   lemp->nxstate = lemp->nstate;
5167   while( lemp->nxstate>1 && lemp->sorted[lemp->nxstate-1]->autoReduce ){
5168     lemp->nxstate--;
5169   }
5170 }
5171 
5172 
5173 /***************** From the file "set.c" ************************************/
5174 /*
5175 ** Set manipulation routines for the LEMON parser generator.
5176 */
5177 
5178 static int size = 0;
5179 
5180 /* Set the set size */
SetSize(int n)5181 void SetSize(int n)
5182 {
5183   size = n+1;
5184 }
5185 
5186 /* Allocate a new set */
SetNew(void)5187 char *SetNew(void){
5188   char *s;
5189   s = (char*)calloc( size, 1);
5190   if( s==0 ){
5191     memory_error();
5192   }
5193   return s;
5194 }
5195 
5196 /* Deallocate a set */
SetFree(char * s)5197 void SetFree(char *s)
5198 {
5199   free(s);
5200 }
5201 
5202 /* Add a new element to the set.  Return TRUE if the element was added
5203 ** and FALSE if it was already there. */
SetAdd(char * s,int e)5204 int SetAdd(char *s, int e)
5205 {
5206   int rv;
5207   assert( e>=0 && e<size );
5208   rv = s[e];
5209   s[e] = 1;
5210   return !rv;
5211 }
5212 
5213 /* Add every element of s2 to s1.  Return TRUE if s1 changes. */
SetUnion(char * s1,char * s2)5214 int SetUnion(char *s1, char *s2)
5215 {
5216   int i, progress;
5217   progress = 0;
5218   for(i=0; i<size; i++){
5219     if( s2[i]==0 ) continue;
5220     if( s1[i]==0 ){
5221       progress = 1;
5222       s1[i] = 1;
5223     }
5224   }
5225   return progress;
5226 }
5227 /********************** From the file "table.c" ****************************/
5228 /*
5229 ** All code in this file has been automatically generated
5230 ** from a specification in the file
5231 **              "table.q"
5232 ** by the associative array code building program "aagen".
5233 ** Do not edit this file!  Instead, edit the specification
5234 ** file, then rerun aagen.
5235 */
5236 /*
5237 ** Code for processing tables in the LEMON parser generator.
5238 */
5239 
strhash(const char * x)5240 PRIVATE unsigned strhash(const char *x)
5241 {
5242   unsigned h = 0;
5243   while( *x ) h = h*13 + *(x++);
5244   return h;
5245 }
5246 
5247 /* Works like strdup, sort of.  Save a string in malloced memory, but
5248 ** keep strings in a table so that the same string is not in more
5249 ** than one place.
5250 */
Strsafe(const char * y)5251 const char *Strsafe(const char *y)
5252 {
5253   const char *z;
5254   char *cpy;
5255 
5256   if( y==0 ) return 0;
5257   z = Strsafe_find(y);
5258   if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){
5259     lemon_strcpy(cpy,y);
5260     z = cpy;
5261     Strsafe_insert(z);
5262   }
5263   MemoryCheck(z);
5264   return z;
5265 }
5266 
5267 /* There is one instance of the following structure for each
5268 ** associative array of type "x1".
5269 */
5270 struct s_x1 {
5271   int size;               /* The number of available slots. */
5272                           /*   Must be a power of 2 greater than or */
5273                           /*   equal to 1 */
5274   int count;              /* Number of currently slots filled */
5275   struct s_x1node *tbl;  /* The data stored here */
5276   struct s_x1node **ht;  /* Hash table for lookups */
5277 };
5278 
5279 /* There is one instance of this structure for every data element
5280 ** in an associative array of type "x1".
5281 */
5282 typedef struct s_x1node {
5283   const char *data;        /* The data */
5284   struct s_x1node *next;   /* Next entry with the same hash */
5285   struct s_x1node **from;  /* Previous link */
5286 } x1node;
5287 
5288 /* There is only one instance of the array, which is the following */
5289 static struct s_x1 *x1a;
5290 
5291 /* Allocate a new associative array */
Strsafe_init(void)5292 void Strsafe_init(void){
5293   if( x1a ) return;
5294   x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
5295   if( x1a ){
5296     x1a->size = 1024;
5297     x1a->count = 0;
5298     x1a->tbl = (x1node*)calloc(1024, sizeof(x1node) + sizeof(x1node*));
5299     if( x1a->tbl==0 ){
5300       free(x1a);
5301       x1a = 0;
5302     }else{
5303       int i;
5304       x1a->ht = (x1node**)&(x1a->tbl[1024]);
5305       for(i=0; i<1024; i++) x1a->ht[i] = 0;
5306     }
5307   }
5308 }
5309 /* Insert a new record into the array.  Return TRUE if successful.
5310 ** Prior data with the same key is NOT overwritten */
Strsafe_insert(const char * data)5311 int Strsafe_insert(const char *data)
5312 {
5313   x1node *np;
5314   unsigned h;
5315   unsigned ph;
5316 
5317   if( x1a==0 ) return 0;
5318   ph = strhash(data);
5319   h = ph & (x1a->size-1);
5320   np = x1a->ht[h];
5321   while( np ){
5322     if( strcmp(np->data,data)==0 ){
5323       /* An existing entry with the same key is found. */
5324       /* Fail because overwrite is not allows. */
5325       return 0;
5326     }
5327     np = np->next;
5328   }
5329   if( x1a->count>=x1a->size ){
5330     /* Need to make the hash table bigger */
5331     int i,arrSize;
5332     struct s_x1 array;
5333     array.size = arrSize = x1a->size*2;
5334     array.count = x1a->count;
5335     array.tbl = (x1node*)calloc(arrSize, sizeof(x1node) + sizeof(x1node*));
5336     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
5337     array.ht = (x1node**)&(array.tbl[arrSize]);
5338     for(i=0; i<arrSize; i++) array.ht[i] = 0;
5339     for(i=0; i<x1a->count; i++){
5340       x1node *oldnp, *newnp;
5341       oldnp = &(x1a->tbl[i]);
5342       h = strhash(oldnp->data) & (arrSize-1);
5343       newnp = &(array.tbl[i]);
5344       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5345       newnp->next = array.ht[h];
5346       newnp->data = oldnp->data;
5347       newnp->from = &(array.ht[h]);
5348       array.ht[h] = newnp;
5349     }
5350     free(x1a->tbl);
5351     memcpy(x1a, &array, sizeof(array)); /* *x1a = array; */
5352   }
5353   /* Insert the new data */
5354   h = ph & (x1a->size-1);
5355   np = &(x1a->tbl[x1a->count++]);
5356   np->data = data;
5357   if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
5358   np->next = x1a->ht[h];
5359   x1a->ht[h] = np;
5360   np->from = &(x1a->ht[h]);
5361   return 1;
5362 }
5363 
5364 /* Return a pointer to data assigned to the given key.  Return NULL
5365 ** if no such key. */
Strsafe_find(const char * key)5366 const char *Strsafe_find(const char *key)
5367 {
5368   unsigned h;
5369   x1node *np;
5370 
5371   if( x1a==0 ) return 0;
5372   h = strhash(key) & (x1a->size-1);
5373   np = x1a->ht[h];
5374   while( np ){
5375     if( strcmp(np->data,key)==0 ) break;
5376     np = np->next;
5377   }
5378   return np ? np->data : 0;
5379 }
5380 
5381 /* Return a pointer to the (terminal or nonterminal) symbol "x".
5382 ** Create a new symbol if this is the first time "x" has been seen.
5383 */
Symbol_new(const char * x)5384 struct symbol *Symbol_new(const char *x)
5385 {
5386   struct symbol *sp;
5387 
5388   sp = Symbol_find(x);
5389   if( sp==0 ){
5390     sp = (struct symbol *)calloc(1, sizeof(struct symbol) );
5391     MemoryCheck(sp);
5392     sp->name = Strsafe(x);
5393     sp->type = ISUPPER(*x) ? TERMINAL : NONTERMINAL;
5394     sp->rule = 0;
5395     sp->fallback = 0;
5396     sp->prec = -1;
5397     sp->assoc = UNK;
5398     sp->firstset = 0;
5399     sp->lambda = LEMON_FALSE;
5400     sp->destructor = 0;
5401     sp->destLineno = 0;
5402     sp->datatype = 0;
5403     sp->useCnt = 0;
5404     Symbol_insert(sp,sp->name);
5405   }
5406   sp->useCnt++;
5407   return sp;
5408 }
5409 
5410 /* Compare two symbols for sorting purposes.  Return negative,
5411 ** zero, or positive if a is less then, equal to, or greater
5412 ** than b.
5413 **
5414 ** Symbols that begin with upper case letters (terminals or tokens)
5415 ** must sort before symbols that begin with lower case letters
5416 ** (non-terminals).  And MULTITERMINAL symbols (created using the
5417 ** %token_class directive) must sort at the very end. Other than
5418 ** that, the order does not matter.
5419 **
5420 ** We find experimentally that leaving the symbols in their original
5421 ** order (the order they appeared in the grammar file) gives the
5422 ** smallest parser tables in SQLite.
5423 */
Symbolcmpp(const void * _a,const void * _b)5424 int Symbolcmpp(const void *_a, const void *_b)
5425 {
5426   const struct symbol *a = *(const struct symbol **) _a;
5427   const struct symbol *b = *(const struct symbol **) _b;
5428   int i1 = a->type==MULTITERMINAL ? 3 : a->name[0]>'Z' ? 2 : 1;
5429   int i2 = b->type==MULTITERMINAL ? 3 : b->name[0]>'Z' ? 2 : 1;
5430   return i1==i2 ? a->index - b->index : i1 - i2;
5431 }
5432 
5433 /* There is one instance of the following structure for each
5434 ** associative array of type "x2".
5435 */
5436 struct s_x2 {
5437   int size;               /* The number of available slots. */
5438                           /*   Must be a power of 2 greater than or */
5439                           /*   equal to 1 */
5440   int count;              /* Number of currently slots filled */
5441   struct s_x2node *tbl;  /* The data stored here */
5442   struct s_x2node **ht;  /* Hash table for lookups */
5443 };
5444 
5445 /* There is one instance of this structure for every data element
5446 ** in an associative array of type "x2".
5447 */
5448 typedef struct s_x2node {
5449   struct symbol *data;     /* The data */
5450   const char *key;         /* The key */
5451   struct s_x2node *next;   /* Next entry with the same hash */
5452   struct s_x2node **from;  /* Previous link */
5453 } x2node;
5454 
5455 /* There is only one instance of the array, which is the following */
5456 static struct s_x2 *x2a;
5457 
5458 /* Allocate a new associative array */
Symbol_init(void)5459 void Symbol_init(void){
5460   if( x2a ) return;
5461   x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
5462   if( x2a ){
5463     x2a->size = 128;
5464     x2a->count = 0;
5465     x2a->tbl = (x2node*)calloc(128, sizeof(x2node) + sizeof(x2node*));
5466     if( x2a->tbl==0 ){
5467       free(x2a);
5468       x2a = 0;
5469     }else{
5470       int i;
5471       x2a->ht = (x2node**)&(x2a->tbl[128]);
5472       for(i=0; i<128; i++) x2a->ht[i] = 0;
5473     }
5474   }
5475 }
5476 /* Insert a new record into the array.  Return TRUE if successful.
5477 ** Prior data with the same key is NOT overwritten */
Symbol_insert(struct symbol * data,const char * key)5478 int Symbol_insert(struct symbol *data, const char *key)
5479 {
5480   x2node *np;
5481   unsigned h;
5482   unsigned ph;
5483 
5484   if( x2a==0 ) return 0;
5485   ph = strhash(key);
5486   h = ph & (x2a->size-1);
5487   np = x2a->ht[h];
5488   while( np ){
5489     if( strcmp(np->key,key)==0 ){
5490       /* An existing entry with the same key is found. */
5491       /* Fail because overwrite is not allows. */
5492       return 0;
5493     }
5494     np = np->next;
5495   }
5496   if( x2a->count>=x2a->size ){
5497     /* Need to make the hash table bigger */
5498     int i,arrSize;
5499     struct s_x2 array;
5500     array.size = arrSize = x2a->size*2;
5501     array.count = x2a->count;
5502     array.tbl = (x2node*)calloc(arrSize, sizeof(x2node) + sizeof(x2node*));
5503     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
5504     array.ht = (x2node**)&(array.tbl[arrSize]);
5505     for(i=0; i<arrSize; i++) array.ht[i] = 0;
5506     for(i=0; i<x2a->count; i++){
5507       x2node *oldnp, *newnp;
5508       oldnp = &(x2a->tbl[i]);
5509       h = strhash(oldnp->key) & (arrSize-1);
5510       newnp = &(array.tbl[i]);
5511       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5512       newnp->next = array.ht[h];
5513       newnp->key = oldnp->key;
5514       newnp->data = oldnp->data;
5515       newnp->from = &(array.ht[h]);
5516       array.ht[h] = newnp;
5517     }
5518     free(x2a->tbl);
5519     memcpy(x2a, &array, sizeof(array)); /* *x2a = array; */
5520   }
5521   /* Insert the new data */
5522   h = ph & (x2a->size-1);
5523   np = &(x2a->tbl[x2a->count++]);
5524   np->key = key;
5525   np->data = data;
5526   if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
5527   np->next = x2a->ht[h];
5528   x2a->ht[h] = np;
5529   np->from = &(x2a->ht[h]);
5530   return 1;
5531 }
5532 
5533 /* Return a pointer to data assigned to the given key.  Return NULL
5534 ** if no such key. */
Symbol_find(const char * key)5535 struct symbol *Symbol_find(const char *key)
5536 {
5537   unsigned h;
5538   x2node *np;
5539 
5540   if( x2a==0 ) return 0;
5541   h = strhash(key) & (x2a->size-1);
5542   np = x2a->ht[h];
5543   while( np ){
5544     if( strcmp(np->key,key)==0 ) break;
5545     np = np->next;
5546   }
5547   return np ? np->data : 0;
5548 }
5549 
5550 /* Return the n-th data.  Return NULL if n is out of range. */
Symbol_Nth(int n)5551 struct symbol *Symbol_Nth(int n)
5552 {
5553   struct symbol *data;
5554   if( x2a && n>0 && n<=x2a->count ){
5555     data = x2a->tbl[n-1].data;
5556   }else{
5557     data = 0;
5558   }
5559   return data;
5560 }
5561 
5562 /* Return the size of the array */
Symbol_count()5563 int Symbol_count()
5564 {
5565   return x2a ? x2a->count : 0;
5566 }
5567 
5568 /* Return an array of pointers to all data in the table.
5569 ** The array is obtained from malloc.  Return NULL if memory allocation
5570 ** problems, or if the array is empty. */
Symbol_arrayof()5571 struct symbol **Symbol_arrayof()
5572 {
5573   struct symbol **array;
5574   int i,arrSize;
5575   if( x2a==0 ) return 0;
5576   arrSize = x2a->count;
5577   array = (struct symbol **)calloc(arrSize, sizeof(struct symbol *));
5578   if( array ){
5579     for(i=0; i<arrSize; i++) array[i] = x2a->tbl[i].data;
5580   }
5581   return array;
5582 }
5583 
5584 /* Compare two configurations */
Configcmp(const char * _a,const char * _b)5585 int Configcmp(const char *_a,const char *_b)
5586 {
5587   const struct config *a = (struct config *) _a;
5588   const struct config *b = (struct config *) _b;
5589   int x;
5590   x = a->rp->index - b->rp->index;
5591   if( x==0 ) x = a->dot - b->dot;
5592   return x;
5593 }
5594 
5595 /* Compare two states */
statecmp(struct config * a,struct config * b)5596 PRIVATE int statecmp(struct config *a, struct config *b)
5597 {
5598   int rc;
5599   for(rc=0; rc==0 && a && b;  a=a->bp, b=b->bp){
5600     rc = a->rp->index - b->rp->index;
5601     if( rc==0 ) rc = a->dot - b->dot;
5602   }
5603   if( rc==0 ){
5604     if( a ) rc = 1;
5605     if( b ) rc = -1;
5606   }
5607   return rc;
5608 }
5609 
5610 /* Hash a state */
statehash(struct config * a)5611 PRIVATE unsigned statehash(struct config *a)
5612 {
5613   unsigned h=0;
5614   while( a ){
5615     h = h*571 + a->rp->index*37 + a->dot;
5616     a = a->bp;
5617   }
5618   return h;
5619 }
5620 
5621 /* Allocate a new state structure */
State_new()5622 struct state *State_new()
5623 {
5624   struct state *newstate;
5625   newstate = (struct state *)calloc(1, sizeof(struct state) );
5626   MemoryCheck(newstate);
5627   return newstate;
5628 }
5629 
5630 /* There is one instance of the following structure for each
5631 ** associative array of type "x3".
5632 */
5633 struct s_x3 {
5634   int size;               /* The number of available slots. */
5635                           /*   Must be a power of 2 greater than or */
5636                           /*   equal to 1 */
5637   int count;              /* Number of currently slots filled */
5638   struct s_x3node *tbl;  /* The data stored here */
5639   struct s_x3node **ht;  /* Hash table for lookups */
5640 };
5641 
5642 /* There is one instance of this structure for every data element
5643 ** in an associative array of type "x3".
5644 */
5645 typedef struct s_x3node {
5646   struct state *data;                  /* The data */
5647   struct config *key;                   /* The key */
5648   struct s_x3node *next;   /* Next entry with the same hash */
5649   struct s_x3node **from;  /* Previous link */
5650 } x3node;
5651 
5652 /* There is only one instance of the array, which is the following */
5653 static struct s_x3 *x3a;
5654 
5655 /* Allocate a new associative array */
State_init(void)5656 void State_init(void){
5657   if( x3a ) return;
5658   x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
5659   if( x3a ){
5660     x3a->size = 128;
5661     x3a->count = 0;
5662     x3a->tbl = (x3node*)calloc(128, sizeof(x3node) + sizeof(x3node*));
5663     if( x3a->tbl==0 ){
5664       free(x3a);
5665       x3a = 0;
5666     }else{
5667       int i;
5668       x3a->ht = (x3node**)&(x3a->tbl[128]);
5669       for(i=0; i<128; i++) x3a->ht[i] = 0;
5670     }
5671   }
5672 }
5673 /* Insert a new record into the array.  Return TRUE if successful.
5674 ** Prior data with the same key is NOT overwritten */
State_insert(struct state * data,struct config * key)5675 int State_insert(struct state *data, struct config *key)
5676 {
5677   x3node *np;
5678   unsigned h;
5679   unsigned ph;
5680 
5681   if( x3a==0 ) return 0;
5682   ph = statehash(key);
5683   h = ph & (x3a->size-1);
5684   np = x3a->ht[h];
5685   while( np ){
5686     if( statecmp(np->key,key)==0 ){
5687       /* An existing entry with the same key is found. */
5688       /* Fail because overwrite is not allows. */
5689       return 0;
5690     }
5691     np = np->next;
5692   }
5693   if( x3a->count>=x3a->size ){
5694     /* Need to make the hash table bigger */
5695     int i,arrSize;
5696     struct s_x3 array;
5697     array.size = arrSize = x3a->size*2;
5698     array.count = x3a->count;
5699     array.tbl = (x3node*)calloc(arrSize, sizeof(x3node) + sizeof(x3node*));
5700     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
5701     array.ht = (x3node**)&(array.tbl[arrSize]);
5702     for(i=0; i<arrSize; i++) array.ht[i] = 0;
5703     for(i=0; i<x3a->count; i++){
5704       x3node *oldnp, *newnp;
5705       oldnp = &(x3a->tbl[i]);
5706       h = statehash(oldnp->key) & (arrSize-1);
5707       newnp = &(array.tbl[i]);
5708       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5709       newnp->next = array.ht[h];
5710       newnp->key = oldnp->key;
5711       newnp->data = oldnp->data;
5712       newnp->from = &(array.ht[h]);
5713       array.ht[h] = newnp;
5714     }
5715     free(x3a->tbl);
5716     memcpy(x3a, &array, sizeof(array)); /* *x3a = array; */
5717   }
5718   /* Insert the new data */
5719   h = ph & (x3a->size-1);
5720   np = &(x3a->tbl[x3a->count++]);
5721   np->key = key;
5722   np->data = data;
5723   if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
5724   np->next = x3a->ht[h];
5725   x3a->ht[h] = np;
5726   np->from = &(x3a->ht[h]);
5727   return 1;
5728 }
5729 
5730 /* Return a pointer to data assigned to the given key.  Return NULL
5731 ** if no such key. */
State_find(struct config * key)5732 struct state *State_find(struct config *key)
5733 {
5734   unsigned h;
5735   x3node *np;
5736 
5737   if( x3a==0 ) return 0;
5738   h = statehash(key) & (x3a->size-1);
5739   np = x3a->ht[h];
5740   while( np ){
5741     if( statecmp(np->key,key)==0 ) break;
5742     np = np->next;
5743   }
5744   return np ? np->data : 0;
5745 }
5746 
5747 /* Return an array of pointers to all data in the table.
5748 ** The array is obtained from malloc.  Return NULL if memory allocation
5749 ** problems, or if the array is empty. */
State_arrayof(void)5750 struct state **State_arrayof(void)
5751 {
5752   struct state **array;
5753   int i,arrSize;
5754   if( x3a==0 ) return 0;
5755   arrSize = x3a->count;
5756   array = (struct state **)calloc(arrSize, sizeof(struct state *));
5757   if( array ){
5758     for(i=0; i<arrSize; i++) array[i] = x3a->tbl[i].data;
5759   }
5760   return array;
5761 }
5762 
5763 /* Hash a configuration */
confighash(struct config * a)5764 PRIVATE unsigned confighash(struct config *a)
5765 {
5766   unsigned h=0;
5767   h = h*571 + a->rp->index*37 + a->dot;
5768   return h;
5769 }
5770 
5771 /* There is one instance of the following structure for each
5772 ** associative array of type "x4".
5773 */
5774 struct s_x4 {
5775   int size;               /* The number of available slots. */
5776                           /*   Must be a power of 2 greater than or */
5777                           /*   equal to 1 */
5778   int count;              /* Number of currently slots filled */
5779   struct s_x4node *tbl;  /* The data stored here */
5780   struct s_x4node **ht;  /* Hash table for lookups */
5781 };
5782 
5783 /* There is one instance of this structure for every data element
5784 ** in an associative array of type "x4".
5785 */
5786 typedef struct s_x4node {
5787   struct config *data;                  /* The data */
5788   struct s_x4node *next;   /* Next entry with the same hash */
5789   struct s_x4node **from;  /* Previous link */
5790 } x4node;
5791 
5792 /* There is only one instance of the array, which is the following */
5793 static struct s_x4 *x4a;
5794 
5795 /* Allocate a new associative array */
Configtable_init(void)5796 void Configtable_init(void){
5797   if( x4a ) return;
5798   x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
5799   if( x4a ){
5800     x4a->size = 64;
5801     x4a->count = 0;
5802     x4a->tbl = (x4node*)calloc(64, sizeof(x4node) + sizeof(x4node*));
5803     if( x4a->tbl==0 ){
5804       free(x4a);
5805       x4a = 0;
5806     }else{
5807       int i;
5808       x4a->ht = (x4node**)&(x4a->tbl[64]);
5809       for(i=0; i<64; i++) x4a->ht[i] = 0;
5810     }
5811   }
5812 }
5813 /* Insert a new record into the array.  Return TRUE if successful.
5814 ** Prior data with the same key is NOT overwritten */
Configtable_insert(struct config * data)5815 int Configtable_insert(struct config *data)
5816 {
5817   x4node *np;
5818   unsigned h;
5819   unsigned ph;
5820 
5821   if( x4a==0 ) return 0;
5822   ph = confighash(data);
5823   h = ph & (x4a->size-1);
5824   np = x4a->ht[h];
5825   while( np ){
5826     if( Configcmp((const char *) np->data,(const char *) data)==0 ){
5827       /* An existing entry with the same key is found. */
5828       /* Fail because overwrite is not allows. */
5829       return 0;
5830     }
5831     np = np->next;
5832   }
5833   if( x4a->count>=x4a->size ){
5834     /* Need to make the hash table bigger */
5835     int i,arrSize;
5836     struct s_x4 array;
5837     array.size = arrSize = x4a->size*2;
5838     array.count = x4a->count;
5839     array.tbl = (x4node*)calloc(arrSize, sizeof(x4node) + sizeof(x4node*));
5840     if( array.tbl==0 ) return 0;  /* Fail due to malloc failure */
5841     array.ht = (x4node**)&(array.tbl[arrSize]);
5842     for(i=0; i<arrSize; i++) array.ht[i] = 0;
5843     for(i=0; i<x4a->count; i++){
5844       x4node *oldnp, *newnp;
5845       oldnp = &(x4a->tbl[i]);
5846       h = confighash(oldnp->data) & (arrSize-1);
5847       newnp = &(array.tbl[i]);
5848       if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
5849       newnp->next = array.ht[h];
5850       newnp->data = oldnp->data;
5851       newnp->from = &(array.ht[h]);
5852       array.ht[h] = newnp;
5853     }
5854     free(x4a->tbl);
5855     memcpy(x4a, &array, sizeof(array)); /* *x4a = array; */
5856   }
5857   /* Insert the new data */
5858   h = ph & (x4a->size-1);
5859   np = &(x4a->tbl[x4a->count++]);
5860   np->data = data;
5861   if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
5862   np->next = x4a->ht[h];
5863   x4a->ht[h] = np;
5864   np->from = &(x4a->ht[h]);
5865   return 1;
5866 }
5867 
5868 /* Return a pointer to data assigned to the given key.  Return NULL
5869 ** if no such key. */
Configtable_find(struct config * key)5870 struct config *Configtable_find(struct config *key)
5871 {
5872   int h;
5873   x4node *np;
5874 
5875   if( x4a==0 ) return 0;
5876   h = confighash(key) & (x4a->size-1);
5877   np = x4a->ht[h];
5878   while( np ){
5879     if( Configcmp((const char *) np->data,(const char *) key)==0 ) break;
5880     np = np->next;
5881   }
5882   return np ? np->data : 0;
5883 }
5884 
5885 /* Remove all data from the table.  Pass each data to the function "f"
5886 ** as it is removed.  ("f" may be null to avoid this step.) */
Configtable_clear(int (* f)(struct config *))5887 void Configtable_clear(int(*f)(struct config *))
5888 {
5889   int i;
5890   if( x4a==0 || x4a->count==0 ) return;
5891   if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
5892   for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;
5893   x4a->count = 0;
5894   return;
5895 }
5896