aboutsummaryrefslogtreecommitdiffstatshomepage
path: root/libraries
diff options
context:
space:
mode:
authorDavid Walter Seikel2012-01-24 00:17:38 +1000
committerDavid Walter Seikel2012-01-24 00:17:38 +1000
commit6eb12c56c62201b4426fc817493e3a290388033e (patch)
tree80eb9871d8d8a4854559c2a9b72609f54d91acc2 /libraries
parentUpdate .gitignore for the new libraries. (diff)
downloadSledjHamr-6eb12c56c62201b4426fc817493e3a290388033e.zip
SledjHamr-6eb12c56c62201b4426fc817493e3a290388033e.tar.gz
SledjHamr-6eb12c56c62201b4426fc817493e3a290388033e.tar.bz2
SledjHamr-6eb12c56c62201b4426fc817493e3a290388033e.tar.xz
Move the lemon source to the libraries directory.
Diffstat (limited to 'libraries')
-rw-r--r--libraries/lemon/lemon.c4898
-rw-r--r--libraries/lemon/lempar.c850
2 files changed, 5748 insertions, 0 deletions
diff --git a/libraries/lemon/lemon.c b/libraries/lemon/lemon.c
new file mode 100644
index 0000000..a089bc7
--- /dev/null
+++ b/libraries/lemon/lemon.c
@@ -0,0 +1,4898 @@
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#ifndef __WIN32__
17# if defined(_WIN32) || defined(WIN32)
18# define __WIN32__
19# endif
20#endif
21
22#ifdef __WIN32__
23#ifdef __cplusplus
24extern "C" {
25#endif
26extern int access(const char *path, int mode);
27#ifdef __cplusplus
28}
29#endif
30#else
31#include <unistd.h>
32#endif
33
34/* #define PRIVATE static */
35#define PRIVATE
36
37#ifdef TEST
38#define MAXRHS 5 /* Set low to exercise exception code */
39#else
40#define MAXRHS 1000
41#endif
42
43static int showPrecedenceConflict = 0;
44static char *msort(char*,char**,int(*)(const char*,const char*));
45
46/*
47** Compilers are getting increasingly pedantic about type conversions
48** as C evolves ever closer to Ada.... To work around the latest problems
49** we have to define the following variant of strlen().
50*/
51#define lemonStrlen(X) ((int)strlen(X))
52
53/* a few forward declarations... */
54struct rule;
55struct lemon;
56struct action;
57
58static struct action *Action_new(void);
59static struct action *Action_sort(struct action *);
60
61/********** From the file "build.h" ************************************/
62void FindRulePrecedences();
63void FindFirstSets();
64void FindStates();
65void FindLinks();
66void FindFollowSets();
67void FindActions();
68
69/********* From the file "configlist.h" *********************************/
70void Configlist_init(void);
71struct config *Configlist_add(struct rule *, int);
72struct config *Configlist_addbasis(struct rule *, int);
73void Configlist_closure(struct lemon *);
74void Configlist_sort(void);
75void Configlist_sortbasis(void);
76struct config *Configlist_return(void);
77struct config *Configlist_basis(void);
78void Configlist_eat(struct config *);
79void Configlist_reset(void);
80
81/********* From the file "error.h" ***************************************/
82void ErrorMsg(const char *, int,const char *, ...);
83
84/****** From the file "option.h" ******************************************/
85enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
86 OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR};
87struct s_options {
88 enum option_type type;
89 const char *label;
90 char *arg;
91 const char *message;
92};
93int OptInit(char**,struct s_options*,FILE*);
94int OptNArgs(void);
95char *OptArg(int);
96void OptErr(int);
97void OptPrint(void);
98
99/******** From the file "parse.h" *****************************************/
100void Parse(struct lemon *lemp);
101
102/********* From the file "plink.h" ***************************************/
103struct plink *Plink_new(void);
104void Plink_add(struct plink **, struct config *);
105void Plink_copy(struct plink **, struct plink *);
106void Plink_delete(struct plink *);
107
108/********** From the file "report.h" *************************************/
109void Reprint(struct lemon *);
110void ReportOutput(struct lemon *);
111void ReportTable(struct lemon *, int);
112void ReportHeader(struct lemon *);
113void CompressTables(struct lemon *);
114void ResortStates(struct lemon *);
115
116/********** From the file "set.h" ****************************************/
117void SetSize(int); /* All sets will be of size N */
118char *SetNew(void); /* A new set for element 0..N */
119void SetFree(char*); /* Deallocate a set */
120
121char *SetNew(void); /* A new set for element 0..N */
122int SetAdd(char*,int); /* Add element to a set */
123int SetUnion(char *,char *); /* A <- A U B, thru element N */
124#define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
125
126/********** From the file "struct.h" *************************************/
127/*
128** Principal data structures for the LEMON parser generator.
129*/
130
131typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
132
133/* Symbols (terminals and nonterminals) of the grammar are stored
134** in the following: */
135enum symbol_type {
136 TERMINAL,
137 NONTERMINAL,
138 MULTITERMINAL
139};
140enum e_assoc {
141 LEFT,
142 RIGHT,
143 NONE,
144 UNK
145};
146struct symbol {
147 const char *name; /* Name of the symbol */
148 int index; /* Index number for this symbol */
149 enum symbol_type type; /* Symbols are all either TERMINALS or NTs */
150 struct rule *rule; /* Linked list of rules of this (if an NT) */
151 struct symbol *fallback; /* fallback token in case this token doesn't parse */
152 int prec; /* Precedence if defined (-1 otherwise) */
153 enum e_assoc assoc; /* Associativity if precedence is defined */
154 char *firstset; /* First-set for all rules of this symbol */
155 Boolean lambda; /* True if NT and can generate an empty string */
156 int useCnt; /* Number of times used */
157 char *destructor; /* Code which executes whenever this symbol is
158 ** popped from the stack during error processing */
159 int destLineno; /* Line number for start of destructor */
160 char *datatype; /* The data type of information held by this
161 ** object. Only used if type==NONTERMINAL */
162 int dtnum; /* The data type number. In the parser, the value
163 ** stack is a union. The .yy%d element of this
164 ** union is the correct data type for this object */
165 /* The following fields are used by MULTITERMINALs only */
166 int nsubsym; /* Number of constituent symbols in the MULTI */
167 struct symbol **subsym; /* Array of constituent symbols */
168};
169
170/* Each production rule in the grammar is stored in the following
171** structure. */
172struct rule {
173 struct symbol *lhs; /* Left-hand side of the rule */
174 const char *lhsalias; /* Alias for the LHS (NULL if none) */
175 int lhsStart; /* True if left-hand side is the start symbol */
176 int ruleline; /* Line number for the rule */
177 int nrhs; /* Number of RHS symbols */
178 struct symbol **rhs; /* The RHS symbols */
179 const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
180 int line; /* Line number at which code begins */
181 const char *code; /* The code executed when this rule is reduced */
182 struct symbol *precsym; /* Precedence symbol for this rule */
183 int index; /* An index number for this rule */
184 Boolean canReduce; /* True if this rule is ever reduced */
185 struct rule *nextlhs; /* Next rule with the same LHS */
186 struct rule *next; /* Next rule in the global list */
187};
188
189/* A configuration is a production rule of the grammar together with
190** a mark (dot) showing how much of that rule has been processed so far.
191** Configurations also contain a follow-set which is a list of terminal
192** symbols which are allowed to immediately follow the end of the rule.
193** Every configuration is recorded as an instance of the following: */
194enum cfgstatus {
195 COMPLETE,
196 INCOMPLETE
197};
198struct config {
199 struct rule *rp; /* The rule upon which the configuration is based */
200 int dot; /* The parse point */
201 char *fws; /* Follow-set for this configuration only */
202 struct plink *fplp; /* Follow-set forward propagation links */
203 struct plink *bplp; /* Follow-set backwards propagation links */
204 struct state *stp; /* Pointer to state which contains this */
205 enum cfgstatus status; /* used during followset and shift computations */
206 struct config *next; /* Next configuration in the state */
207 struct config *bp; /* The next basis configuration */
208};
209
210enum e_action {
211 SHIFT,
212 ACCEPT,
213 REDUCE,
214 ERROR,
215 SSCONFLICT, /* A shift/shift conflict */
216 SRCONFLICT, /* Was a reduce, but part of a conflict */
217 RRCONFLICT, /* Was a reduce, but part of a conflict */
218 SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
219 RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
220 NOT_USED /* Deleted by compression */
221};
222
223/* Every shift or reduce operation is stored as one of the following */
224struct action {
225 struct symbol *sp; /* The look-ahead symbol */
226 enum e_action type;
227 union {
228 struct state *stp; /* The new state, if a shift */
229 struct rule *rp; /* The rule, if a reduce */
230 } x;
231 struct action *next; /* Next action for this state */
232 struct action *collide; /* Next action with the same hash */
233};
234
235/* Each state of the generated parser's finite state machine
236** is encoded as an instance of the following structure. */
237struct state {
238 struct config *bp; /* The basis configurations for this state */
239 struct config *cfp; /* All configurations in this set */
240 int statenum; /* Sequential number for this state */
241 struct action *ap; /* Array of actions for this state */
242 int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
243 int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
244 int iDflt; /* Default action */
245};
246#define NO_OFFSET (-2147483647)
247
248/* A followset propagation link indicates that the contents of one
249** configuration followset should be propagated to another whenever
250** the first changes. */
251struct plink {
252 struct config *cfp; /* The configuration to which linked */
253 struct plink *next; /* The next propagate link */
254};
255
256/* The state vector for the entire parser generator is recorded as
257** follows. (LEMON uses no global variables and makes little use of
258** static variables. Fields in the following structure can be thought
259** of as begin global variables in the program.) */
260struct lemon {
261 struct state **sorted; /* Table of states sorted by state number */
262 struct rule *rule; /* List of all rules */
263 int nstate; /* Number of states */
264 int nrule; /* Number of rules */
265 int nsymbol; /* Number of terminal and nonterminal symbols */
266 int nterminal; /* Number of terminal symbols */
267 struct symbol **symbols; /* Sorted array of pointers to symbols */
268 int errorcnt; /* Number of errors */
269 struct symbol *errsym; /* The error symbol */
270 struct symbol *wildcard; /* Token that matches anything */
271 char *name; /* Name of the generated parser */
272 char *arg; /* Declaration of the 3th argument to parser */
273 char *tokentype; /* Type of terminal symbols in the parser stack */
274 char *vartype; /* The default type of non-terminal symbols */
275 char *start; /* Name of the start symbol for the grammar */
276 char *stacksize; /* Size of the parser stack */
277 char *include; /* Code to put at the start of the C file */
278 char *error; /* Code to execute when an error is seen */
279 char *overflow; /* Code to execute on a stack overflow */
280 char *failure; /* Code to execute on parser failure */
281 char *accept; /* Code to execute when the parser excepts */
282 char *extracode; /* Code appended to the generated file */
283 char *tokendest; /* Code to execute to destroy token data */
284 char *vardest; /* Code for the default non-terminal destructor */
285 char *filename; /* Name of the input file */
286 char *outname; /* Name of the current output file */
287 char *tokenprefix; /* A prefix added to token names in the .h file */
288 int nconflict; /* Number of parsing conflicts */
289 int tablesize; /* Size of the parse tables */
290 int basisflag; /* Print only basis configurations */
291 int has_fallback; /* True if any %fallback is seen in the grammar */
292 int nolinenosflag; /* True if #line statements should not be printed */
293 char *argv0; /* Name of the program */
294};
295
296#define MemoryCheck(X) if((X)==0){ \
297 extern void memory_error(); \
298 memory_error(); \
299}
300
301/**************** From the file "table.h" *********************************/
302/*
303** All code in this file has been automatically generated
304** from a specification in the file
305** "table.q"
306** by the associative array code building program "aagen".
307** Do not edit this file! Instead, edit the specification
308** file, then rerun aagen.
309*/
310/*
311** Code for processing tables in the LEMON parser generator.
312*/
313/* Routines for handling a strings */
314
315const char *Strsafe(const char *);
316
317void Strsafe_init(void);
318int Strsafe_insert(const char *);
319const char *Strsafe_find(const char *);
320
321/* Routines for handling symbols of the grammar */
322
323struct symbol *Symbol_new(const char *);
324int Symbolcmpp(const void *, const void *);
325void Symbol_init(void);
326int Symbol_insert(struct symbol *, const char *);
327struct symbol *Symbol_find(const char *);
328struct symbol *Symbol_Nth(int);
329int Symbol_count(void);
330struct symbol **Symbol_arrayof(void);
331
332/* Routines to manage the state table */
333
334int Configcmp(const char *, const char *);
335struct state *State_new(void);
336void State_init(void);
337int State_insert(struct state *, struct config *);
338struct state *State_find(struct config *);
339struct state **State_arrayof(/* */);
340
341/* Routines used for efficiency in Configlist_add */
342
343void Configtable_init(void);
344int Configtable_insert(struct config *);
345struct config *Configtable_find(struct config *);
346void Configtable_clear(int(*)(struct config *));
347
348/****************** From the file "action.c" *******************************/
349/*
350** Routines processing parser actions in the LEMON parser generator.
351*/
352
353/* Allocate a new parser action */
354static struct action *Action_new(void){
355 static struct action *freelist = 0;
356 struct action *newaction;
357
358 if( freelist==0 ){
359 int i;
360 int amt = 100;
361 freelist = (struct action *)calloc(amt, sizeof(struct action));
362 if( freelist==0 ){
363 fprintf(stderr,"Unable to allocate memory for a new parser action.");
364 exit(1);
365 }
366 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
367 freelist[amt-1].next = 0;
368 }
369 newaction = freelist;
370 freelist = freelist->next;
371 return newaction;
372}
373
374/* Compare two actions for sorting purposes. Return negative, zero, or
375** positive if the first action is less than, equal to, or greater than
376** the first
377*/
378static int actioncmp(
379 struct action *ap1,
380 struct action *ap2
381){
382 int rc;
383 rc = ap1->sp->index - ap2->sp->index;
384 if( rc==0 ){
385 rc = (int)ap1->type - (int)ap2->type;
386 }
387 if( rc==0 && ap1->type==REDUCE ){
388 rc = ap1->x.rp->index - ap2->x.rp->index;
389 }
390 if( rc==0 ){
391 rc = (int) (ap2 - ap1);
392 }
393 return rc;
394}
395
396/* Sort parser actions */
397static struct action *Action_sort(
398 struct action *ap
399){
400 ap = (struct action *)msort((char *)ap,(char **)&ap->next,
401 (int(*)(const char*,const char*))actioncmp);
402 return ap;
403}
404
405void Action_add(
406 struct action **app,
407 enum e_action type,
408 struct symbol *sp,
409 char *arg
410){
411 struct action *newaction;
412 newaction = Action_new();
413 newaction->next = *app;
414 *app = newaction;
415 newaction->type = type;
416 newaction->sp = sp;
417 if( type==SHIFT ){
418 newaction->x.stp = (struct state *)arg;
419 }else{
420 newaction->x.rp = (struct rule *)arg;
421 }
422}
423/********************** New code to implement the "acttab" module ***********/
424/*
425** This module implements routines use to construct the yy_action[] table.
426*/
427
428/*
429** The state of the yy_action table under construction is an instance of
430** the following structure.
431**
432** The yy_action table maps the pair (state_number, lookahead) into an
433** action_number. The table is an array of integers pairs. The state_number
434** determines an initial offset into the yy_action array. The lookahead
435** value is then added to this initial offset to get an index X into the
436** yy_action array. If the aAction[X].lookahead equals the value of the
437** of the lookahead input, then the value of the action_number output is
438** aAction[X].action. If the lookaheads do not match then the
439** default action for the state_number is returned.
440**
441** All actions associated with a single state_number are first entered
442** into aLookahead[] using multiple calls to acttab_action(). Then the
443** actions for that single state_number are placed into the aAction[]
444** array with a single call to acttab_insert(). The acttab_insert() call
445** also resets the aLookahead[] array in preparation for the next
446** state number.
447*/
448struct lookahead_action {
449 int lookahead; /* Value of the lookahead token */
450 int action; /* Action to take on the given lookahead */
451};
452typedef struct acttab acttab;
453struct acttab {
454 int nAction; /* Number of used slots in aAction[] */
455 int nActionAlloc; /* Slots allocated for aAction[] */
456 struct lookahead_action
457 *aAction, /* The yy_action[] table under construction */
458 *aLookahead; /* A single new transaction set */
459 int mnLookahead; /* Minimum aLookahead[].lookahead */
460 int mnAction; /* Action associated with mnLookahead */
461 int mxLookahead; /* Maximum aLookahead[].lookahead */
462 int nLookahead; /* Used slots in aLookahead[] */
463 int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
464};
465
466/* Return the number of entries in the yy_action table */
467#define acttab_size(X) ((X)->nAction)
468
469/* The value for the N-th entry in yy_action */
470#define acttab_yyaction(X,N) ((X)->aAction[N].action)
471
472/* The value for the N-th entry in yy_lookahead */
473#define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
474
475/* Free all memory associated with the given acttab */
476void acttab_free(acttab *p){
477 free( p->aAction );
478 free( p->aLookahead );
479 free( p );
480}
481
482/* Allocate a new acttab structure */
483acttab *acttab_alloc(void){
484 acttab *p = (acttab *) calloc( 1, sizeof(*p) );
485 if( p==0 ){
486 fprintf(stderr,"Unable to allocate memory for a new acttab.");
487 exit(1);
488 }
489 memset(p, 0, sizeof(*p));
490 return p;
491}
492
493/* Add a new action to the current transaction set.
494**
495** This routine is called once for each lookahead for a particular
496** state.
497*/
498void acttab_action(acttab *p, int lookahead, int action){
499 if( p->nLookahead>=p->nLookaheadAlloc ){
500 p->nLookaheadAlloc += 25;
501 p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead,
502 sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
503 if( p->aLookahead==0 ){
504 fprintf(stderr,"malloc failed\n");
505 exit(1);
506 }
507 }
508 if( p->nLookahead==0 ){
509 p->mxLookahead = lookahead;
510 p->mnLookahead = lookahead;
511 p->mnAction = action;
512 }else{
513 if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
514 if( p->mnLookahead>lookahead ){
515 p->mnLookahead = lookahead;
516 p->mnAction = action;
517 }
518 }
519 p->aLookahead[p->nLookahead].lookahead = lookahead;
520 p->aLookahead[p->nLookahead].action = action;
521 p->nLookahead++;
522}
523
524/*
525** Add the transaction set built up with prior calls to acttab_action()
526** into the current action table. Then reset the transaction set back
527** to an empty set in preparation for a new round of acttab_action() calls.
528**
529** Return the offset into the action table of the new transaction.
530*/
531int acttab_insert(acttab *p){
532 int i, j, k, n;
533 assert( p->nLookahead>0 );
534
535 /* Make sure we have enough space to hold the expanded action table
536 ** in the worst case. The worst case occurs if the transaction set
537 ** must be appended to the current action table
538 */
539 n = p->mxLookahead + 1;
540 if( p->nAction + n >= p->nActionAlloc ){
541 int oldAlloc = p->nActionAlloc;
542 p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
543 p->aAction = (struct lookahead_action *) realloc( p->aAction,
544 sizeof(p->aAction[0])*p->nActionAlloc);
545 if( p->aAction==0 ){
546 fprintf(stderr,"malloc failed\n");
547 exit(1);
548 }
549 for(i=oldAlloc; i<p->nActionAlloc; i++){
550 p->aAction[i].lookahead = -1;
551 p->aAction[i].action = -1;
552 }
553 }
554
555 /* Scan the existing action table looking for an offset that is a
556 ** duplicate of the current transaction set. Fall out of the loop
557 ** if and when the duplicate is found.
558 **
559 ** i is the index in p->aAction[] where p->mnLookahead is inserted.
560 */
561 for(i=p->nAction-1; i>=0; i--){
562 if( p->aAction[i].lookahead==p->mnLookahead ){
563 /* All lookaheads and actions in the aLookahead[] transaction
564 ** must match against the candidate aAction[i] entry. */
565 if( p->aAction[i].action!=p->mnAction ) continue;
566 for(j=0; j<p->nLookahead; j++){
567 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
568 if( k<0 || k>=p->nAction ) break;
569 if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
570 if( p->aLookahead[j].action!=p->aAction[k].action ) break;
571 }
572 if( j<p->nLookahead ) continue;
573
574 /* No possible lookahead value that is not in the aLookahead[]
575 ** transaction is allowed to match aAction[i] */
576 n = 0;
577 for(j=0; j<p->nAction; j++){
578 if( p->aAction[j].lookahead<0 ) continue;
579 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
580 }
581 if( n==p->nLookahead ){
582 break; /* An exact match is found at offset i */
583 }
584 }
585 }
586
587 /* If no existing offsets exactly match the current transaction, find an
588 ** an empty offset in the aAction[] table in which we can add the
589 ** aLookahead[] transaction.
590 */
591 if( i<0 ){
592 /* Look for holes in the aAction[] table that fit the current
593 ** aLookahead[] transaction. Leave i set to the offset of the hole.
594 ** If no holes are found, i is left at p->nAction, which means the
595 ** transaction will be appended. */
596 for(i=0; i<p->nActionAlloc - p->mxLookahead; i++){
597 if( p->aAction[i].lookahead<0 ){
598 for(j=0; j<p->nLookahead; j++){
599 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
600 if( k<0 ) break;
601 if( p->aAction[k].lookahead>=0 ) break;
602 }
603 if( j<p->nLookahead ) continue;
604 for(j=0; j<p->nAction; j++){
605 if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
606 }
607 if( j==p->nAction ){
608 break; /* Fits in empty slots */
609 }
610 }
611 }
612 }
613 /* Insert transaction set at index i. */
614 for(j=0; j<p->nLookahead; j++){
615 k = p->aLookahead[j].lookahead - p->mnLookahead + i;
616 p->aAction[k] = p->aLookahead[j];
617 if( k>=p->nAction ) p->nAction = k+1;
618 }
619 p->nLookahead = 0;
620
621 /* Return the offset that is added to the lookahead in order to get the
622 ** index into yy_action of the action */
623 return i - p->mnLookahead;
624}
625
626/********************** From the file "build.c" *****************************/
627/*
628** Routines to construction the finite state machine for the LEMON
629** parser generator.
630*/
631
632/* Find a precedence symbol of every rule in the grammar.
633**
634** Those rules which have a precedence symbol coded in the input
635** grammar using the "[symbol]" construct will already have the
636** rp->precsym field filled. Other rules take as their precedence
637** symbol the first RHS symbol with a defined precedence. If there
638** are not RHS symbols with a defined precedence, the precedence
639** symbol field is left blank.
640*/
641void FindRulePrecedences(struct lemon *xp)
642{
643 struct rule *rp;
644 for(rp=xp->rule; rp; rp=rp->next){
645 if( rp->precsym==0 ){
646 int i, j;
647 for(i=0; i<rp->nrhs && rp->precsym==0; i++){
648 struct symbol *sp = rp->rhs[i];
649 if( sp->type==MULTITERMINAL ){
650 for(j=0; j<sp->nsubsym; j++){
651 if( sp->subsym[j]->prec>=0 ){
652 rp->precsym = sp->subsym[j];
653 break;
654 }
655 }
656 }else if( sp->prec>=0 ){
657 rp->precsym = rp->rhs[i];
658 }
659 }
660 }
661 }
662 return;
663}
664
665/* Find all nonterminals which will generate the empty string.
666** Then go back and compute the first sets of every nonterminal.
667** The first set is the set of all terminal symbols which can begin
668** a string generated by that nonterminal.
669*/
670void FindFirstSets(struct lemon *lemp)
671{
672 int i, j;
673 struct rule *rp;
674 int progress;
675
676 for(i=0; i<lemp->nsymbol; i++){
677 lemp->symbols[i]->lambda = LEMON_FALSE;
678 }
679 for(i=lemp->nterminal; i<lemp->nsymbol; i++){
680 lemp->symbols[i]->firstset = SetNew();
681 }
682
683 /* First compute all lambdas */
684 do{
685 progress = 0;
686 for(rp=lemp->rule; rp; rp=rp->next){
687 if( rp->lhs->lambda ) continue;
688 for(i=0; i<rp->nrhs; i++){
689 struct symbol *sp = rp->rhs[i];
690 assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE );
691 if( sp->lambda==LEMON_FALSE ) break;
692 }
693 if( i==rp->nrhs ){
694 rp->lhs->lambda = LEMON_TRUE;
695 progress = 1;
696 }
697 }
698 }while( progress );
699
700 /* Now compute all first sets */
701 do{
702 struct symbol *s1, *s2;
703 progress = 0;
704 for(rp=lemp->rule; rp; rp=rp->next){
705 s1 = rp->lhs;
706 for(i=0; i<rp->nrhs; i++){
707 s2 = rp->rhs[i];
708 if( s2->type==TERMINAL ){
709 progress += SetAdd(s1->firstset,s2->index);
710 break;
711 }else if( s2->type==MULTITERMINAL ){
712 for(j=0; j<s2->nsubsym; j++){
713 progress += SetAdd(s1->firstset,s2->subsym[j]->index);
714 }
715 break;
716 }else if( s1==s2 ){
717 if( s1->lambda==LEMON_FALSE ) break;
718 }else{
719 progress += SetUnion(s1->firstset,s2->firstset);
720 if( s2->lambda==LEMON_FALSE ) break;
721 }
722 }
723 }
724 }while( progress );
725 return;
726}
727
728/* Compute all LR(0) states for the grammar. Links
729** are added to between some states so that the LR(1) follow sets
730** can be computed later.
731*/
732PRIVATE struct state *getstate(struct lemon *); /* forward reference */
733void FindStates(struct lemon *lemp)
734{
735 struct symbol *sp;
736 struct rule *rp;
737
738 Configlist_init();
739
740 /* Find the start symbol */
741 if( lemp->start ){
742 sp = Symbol_find(lemp->start);
743 if( sp==0 ){
744 ErrorMsg(lemp->filename,0,
745"The specified start symbol \"%s\" is not \
746in a nonterminal of the grammar. \"%s\" will be used as the start \
747symbol instead.",lemp->start,lemp->rule->lhs->name);
748 lemp->errorcnt++;
749 sp = lemp->rule->lhs;
750 }
751 }else{
752 sp = lemp->rule->lhs;
753 }
754
755 /* Make sure the start symbol doesn't occur on the right-hand side of
756 ** any rule. Report an error if it does. (YACC would generate a new
757 ** start symbol in this case.) */
758 for(rp=lemp->rule; rp; rp=rp->next){
759 int i;
760 for(i=0; i<rp->nrhs; i++){
761 if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
762 ErrorMsg(lemp->filename,0,
763"The start symbol \"%s\" occurs on the \
764right-hand side of a rule. This will result in a parser which \
765does not work properly.",sp->name);
766 lemp->errorcnt++;
767 }
768 }
769 }
770
771 /* The basis configuration set for the first state
772 ** is all rules which have the start symbol as their
773 ** left-hand side */
774 for(rp=sp->rule; rp; rp=rp->nextlhs){
775 struct config *newcfp;
776 rp->lhsStart = 1;
777 newcfp = Configlist_addbasis(rp,0);
778 SetAdd(newcfp->fws,0);
779 }
780
781 /* Compute the first state. All other states will be
782 ** computed automatically during the computation of the first one.
783 ** The returned pointer to the first state is not used. */
784 (void)getstate(lemp);
785 return;
786}
787
788/* Return a pointer to a state which is described by the configuration
789** list which has been built from calls to Configlist_add.
790*/
791PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */
792PRIVATE struct state *getstate(struct lemon *lemp)
793{
794 struct config *cfp, *bp;
795 struct state *stp;
796
797 /* Extract the sorted basis of the new state. The basis was constructed
798 ** by prior calls to "Configlist_addbasis()". */
799 Configlist_sortbasis();
800 bp = Configlist_basis();
801
802 /* Get a state with the same basis */
803 stp = State_find(bp);
804 if( stp ){
805 /* A state with the same basis already exists! Copy all the follow-set
806 ** propagation links from the state under construction into the
807 ** preexisting state, then return a pointer to the preexisting state */
808 struct config *x, *y;
809 for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
810 Plink_copy(&y->bplp,x->bplp);
811 Plink_delete(x->fplp);
812 x->fplp = x->bplp = 0;
813 }
814 cfp = Configlist_return();
815 Configlist_eat(cfp);
816 }else{
817 /* This really is a new state. Construct all the details */
818 Configlist_closure(lemp); /* Compute the configuration closure */
819 Configlist_sort(); /* Sort the configuration closure */
820 cfp = Configlist_return(); /* Get a pointer to the config list */
821 stp = State_new(); /* A new state structure */
822 MemoryCheck(stp);
823 stp->bp = bp; /* Remember the configuration basis */
824 stp->cfp = cfp; /* Remember the configuration closure */
825 stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
826 stp->ap = 0; /* No actions, yet. */
827 State_insert(stp,stp->bp); /* Add to the state table */
828 buildshifts(lemp,stp); /* Recursively compute successor states */
829 }
830 return stp;
831}
832
833/*
834** Return true if two symbols are the same.
835*/
836int same_symbol(struct symbol *a, struct symbol *b)
837{
838 int i;
839 if( a==b ) return 1;
840 if( a->type!=MULTITERMINAL ) return 0;
841 if( b->type!=MULTITERMINAL ) return 0;
842 if( a->nsubsym!=b->nsubsym ) return 0;
843 for(i=0; i<a->nsubsym; i++){
844 if( a->subsym[i]!=b->subsym[i] ) return 0;
845 }
846 return 1;
847}
848
849/* Construct all successor states to the given state. A "successor"
850** state is any state which can be reached by a shift action.
851*/
852PRIVATE void buildshifts(struct lemon *lemp, struct state *stp)
853{
854 struct config *cfp; /* For looping thru the config closure of "stp" */
855 struct config *bcfp; /* For the inner loop on config closure of "stp" */
856 struct config *newcfg; /* */
857 struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
858 struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
859 struct state *newstp; /* A pointer to a successor state */
860
861 /* Each configuration becomes complete after it contibutes to a successor
862 ** state. Initially, all configurations are incomplete */
863 for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
864
865 /* Loop through all configurations of the state "stp" */
866 for(cfp=stp->cfp; cfp; cfp=cfp->next){
867 if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
868 if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
869 Configlist_reset(); /* Reset the new config set */
870 sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
871
872 /* For every configuration in the state "stp" which has the symbol "sp"
873 ** following its dot, add the same configuration to the basis set under
874 ** construction but with the dot shifted one symbol to the right. */
875 for(bcfp=cfp; bcfp; bcfp=bcfp->next){
876 if( bcfp->status==COMPLETE ) continue; /* Already used */
877 if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
878 bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
879 if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
880 bcfp->status = COMPLETE; /* Mark this config as used */
881 newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
882 Plink_add(&newcfg->bplp,bcfp);
883 }
884
885 /* Get a pointer to the state described by the basis configuration set
886 ** constructed in the preceding loop */
887 newstp = getstate(lemp);
888
889 /* The state "newstp" is reached from the state "stp" by a shift action
890 ** on the symbol "sp" */
891 if( sp->type==MULTITERMINAL ){
892 int i;
893 for(i=0; i<sp->nsubsym; i++){
894 Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
895 }
896 }else{
897 Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
898 }
899 }
900}
901
902/*
903** Construct the propagation links
904*/
905void FindLinks(struct lemon *lemp)
906{
907 int i;
908 struct config *cfp, *other;
909 struct state *stp;
910 struct plink *plp;
911
912 /* Housekeeping detail:
913 ** Add to every propagate link a pointer back to the state to
914 ** which the link is attached. */
915 for(i=0; i<lemp->nstate; i++){
916 stp = lemp->sorted[i];
917 for(cfp=stp->cfp; cfp; cfp=cfp->next){
918 cfp->stp = stp;
919 }
920 }
921
922 /* Convert all backlinks into forward links. Only the forward
923 ** links are used in the follow-set computation. */
924 for(i=0; i<lemp->nstate; i++){
925 stp = lemp->sorted[i];
926 for(cfp=stp->cfp; cfp; cfp=cfp->next){
927 for(plp=cfp->bplp; plp; plp=plp->next){
928 other = plp->cfp;
929 Plink_add(&other->fplp,cfp);
930 }
931 }
932 }
933}
934
935/* Compute all followsets.
936**
937** A followset is the set of all symbols which can come immediately
938** after a configuration.
939*/
940void FindFollowSets(struct lemon *lemp)
941{
942 int i;
943 struct config *cfp;
944 struct plink *plp;
945 int progress;
946 int change;
947
948 for(i=0; i<lemp->nstate; i++){
949 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
950 cfp->status = INCOMPLETE;
951 }
952 }
953
954 do{
955 progress = 0;
956 for(i=0; i<lemp->nstate; i++){
957 for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
958 if( cfp->status==COMPLETE ) continue;
959 for(plp=cfp->fplp; plp; plp=plp->next){
960 change = SetUnion(plp->cfp->fws,cfp->fws);
961 if( change ){
962 plp->cfp->status = INCOMPLETE;
963 progress = 1;
964 }
965 }
966 cfp->status = COMPLETE;
967 }
968 }
969 }while( progress );
970}
971
972static int resolve_conflict(struct action *,struct action *, struct symbol *);
973
974/* Compute the reduce actions, and resolve conflicts.
975*/
976void FindActions(struct lemon *lemp)
977{
978 int i,j;
979 struct config *cfp;
980 struct state *stp;
981 struct symbol *sp;
982 struct rule *rp;
983
984 /* Add all of the reduce actions
985 ** A reduce action is added for each element of the followset of
986 ** a configuration which has its dot at the extreme right.
987 */
988 for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
989 stp = lemp->sorted[i];
990 for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
991 if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
992 for(j=0; j<lemp->nterminal; j++){
993 if( SetFind(cfp->fws,j) ){
994 /* Add a reduce action to the state "stp" which will reduce by the
995 ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
996 Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
997 }
998 }
999 }
1000 }
1001 }
1002
1003 /* Add the accepting token */
1004 if( lemp->start ){
1005 sp = Symbol_find(lemp->start);
1006 if( sp==0 ) sp = lemp->rule->lhs;
1007 }else{
1008 sp = lemp->rule->lhs;
1009 }
1010 /* Add to the first state (which is always the starting state of the
1011 ** finite state machine) an action to ACCEPT if the lookahead is the
1012 ** start nonterminal. */
1013 Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
1014
1015 /* Resolve conflicts */
1016 for(i=0; i<lemp->nstate; i++){
1017 struct action *ap, *nap;
1018 struct state *stp;
1019 stp = lemp->sorted[i];
1020 /* assert( stp->ap ); */
1021 stp->ap = Action_sort(stp->ap);
1022 for(ap=stp->ap; ap && ap->next; ap=ap->next){
1023 for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
1024 /* The two actions "ap" and "nap" have the same lookahead.
1025 ** Figure out which one should be used */
1026 lemp->nconflict += resolve_conflict(ap,nap,lemp->errsym);
1027 }
1028 }
1029 }
1030
1031 /* Report an error for each rule that can never be reduced. */
1032 for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
1033 for(i=0; i<lemp->nstate; i++){
1034 struct action *ap;
1035 for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
1036 if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
1037 }
1038 }
1039 for(rp=lemp->rule; rp; rp=rp->next){
1040 if( rp->canReduce ) continue;
1041 ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
1042 lemp->errorcnt++;
1043 }
1044}
1045
1046/* Resolve a conflict between the two given actions. If the
1047** conflict can't be resolved, return non-zero.
1048**
1049** NO LONGER TRUE:
1050** To resolve a conflict, first look to see if either action
1051** is on an error rule. In that case, take the action which
1052** is not associated with the error rule. If neither or both
1053** actions are associated with an error rule, then try to
1054** use precedence to resolve the conflict.
1055**
1056** If either action is a SHIFT, then it must be apx. This
1057** function won't work if apx->type==REDUCE and apy->type==SHIFT.
1058*/
1059static int resolve_conflict(
1060 struct action *apx,
1061 struct action *apy,
1062 struct symbol *errsym /* The error symbol (if defined. NULL otherwise) */
1063){
1064 struct symbol *spx, *spy;
1065 int errcnt = 0;
1066 assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
1067 if( apx->type==SHIFT && apy->type==SHIFT ){
1068 apy->type = SSCONFLICT;
1069 errcnt++;
1070 }
1071 if( apx->type==SHIFT && apy->type==REDUCE ){
1072 spx = apx->sp;
1073 spy = apy->x.rp->precsym;
1074 if( spy==0 || spx->prec<0 || spy->prec<0 ){
1075 /* Not enough precedence information. */
1076 apy->type = SRCONFLICT;
1077 errcnt++;
1078 }else if( spx->prec>spy->prec ){ /* higher precedence wins */
1079 apy->type = RD_RESOLVED;
1080 }else if( spx->prec<spy->prec ){
1081 apx->type = SH_RESOLVED;
1082 }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
1083 apy->type = RD_RESOLVED; /* associativity */
1084 }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
1085 apx->type = SH_RESOLVED;
1086 }else{
1087 assert( spx->prec==spy->prec && spx->assoc==NONE );
1088 apy->type = SRCONFLICT;
1089 errcnt++;
1090 }
1091 }else if( apx->type==REDUCE && apy->type==REDUCE ){
1092 spx = apx->x.rp->precsym;
1093 spy = apy->x.rp->precsym;
1094 if( spx==0 || spy==0 || spx->prec<0 ||
1095 spy->prec<0 || spx->prec==spy->prec ){
1096 apy->type = RRCONFLICT;
1097 errcnt++;
1098 }else if( spx->prec>spy->prec ){
1099 apy->type = RD_RESOLVED;
1100 }else if( spx->prec<spy->prec ){
1101 apx->type = RD_RESOLVED;
1102 }
1103 }else{
1104 assert(
1105 apx->type==SH_RESOLVED ||
1106 apx->type==RD_RESOLVED ||
1107 apx->type==SSCONFLICT ||
1108 apx->type==SRCONFLICT ||
1109 apx->type==RRCONFLICT ||
1110 apy->type==SH_RESOLVED ||
1111 apy->type==RD_RESOLVED ||
1112 apy->type==SSCONFLICT ||
1113 apy->type==SRCONFLICT ||
1114 apy->type==RRCONFLICT
1115 );
1116 /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
1117 ** REDUCEs on the list. If we reach this point it must be because
1118 ** the parser conflict had already been resolved. */
1119 }
1120 return errcnt;
1121}
1122/********************* From the file "configlist.c" *************************/
1123/*
1124** Routines to processing a configuration list and building a state
1125** in the LEMON parser generator.
1126*/
1127
1128static struct config *freelist = 0; /* List of free configurations */
1129static struct config *current = 0; /* Top of list of configurations */
1130static struct config **currentend = 0; /* Last on list of configs */
1131static struct config *basis = 0; /* Top of list of basis configs */
1132static struct config **basisend = 0; /* End of list of basis configs */
1133
1134/* Return a pointer to a new configuration */
1135PRIVATE struct config *newconfig(){
1136 struct config *newcfg;
1137 if( freelist==0 ){
1138 int i;
1139 int amt = 3;
1140 freelist = (struct config *)calloc( amt, sizeof(struct config) );
1141 if( freelist==0 ){
1142 fprintf(stderr,"Unable to allocate memory for a new configuration.");
1143 exit(1);
1144 }
1145 for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
1146 freelist[amt-1].next = 0;
1147 }
1148 newcfg = freelist;
1149 freelist = freelist->next;
1150 return newcfg;
1151}
1152
1153/* The configuration "old" is no longer used */
1154PRIVATE void deleteconfig(struct config *old)
1155{
1156 old->next = freelist;
1157 freelist = old;
1158}
1159
1160/* Initialized the configuration list builder */
1161void Configlist_init(){
1162 current = 0;
1163 currentend = &current;
1164 basis = 0;
1165 basisend = &basis;
1166 Configtable_init();
1167 return;
1168}
1169
1170/* Initialized the configuration list builder */
1171void Configlist_reset(){
1172 current = 0;
1173 currentend = &current;
1174 basis = 0;
1175 basisend = &basis;
1176 Configtable_clear(0);
1177 return;
1178}
1179
1180/* Add another configuration to the configuration list */
1181struct config *Configlist_add(
1182 struct rule *rp, /* The rule */
1183 int dot /* Index into the RHS of the rule where the dot goes */
1184){
1185 struct config *cfp, model;
1186
1187 assert( currentend!=0 );
1188 model.rp = rp;
1189 model.dot = dot;
1190 cfp = Configtable_find(&model);
1191 if( cfp==0 ){
1192 cfp = newconfig();
1193 cfp->rp = rp;
1194 cfp->dot = dot;
1195 cfp->fws = SetNew();
1196 cfp->stp = 0;
1197 cfp->fplp = cfp->bplp = 0;
1198 cfp->next = 0;
1199 cfp->bp = 0;
1200 *currentend = cfp;
1201 currentend = &cfp->next;
1202 Configtable_insert(cfp);
1203 }
1204 return cfp;
1205}
1206
1207/* Add a basis configuration to the configuration list */
1208struct config *Configlist_addbasis(struct rule *rp, int dot)
1209{
1210 struct config *cfp, model;
1211
1212 assert( basisend!=0 );
1213 assert( currentend!=0 );
1214 model.rp = rp;
1215 model.dot = dot;
1216 cfp = Configtable_find(&model);
1217 if( cfp==0 ){
1218 cfp = newconfig();
1219 cfp->rp = rp;
1220 cfp->dot = dot;
1221 cfp->fws = SetNew();
1222 cfp->stp = 0;
1223 cfp->fplp = cfp->bplp = 0;
1224 cfp->next = 0;
1225 cfp->bp = 0;
1226 *currentend = cfp;
1227 currentend = &cfp->next;
1228 *basisend = cfp;
1229 basisend = &cfp->bp;
1230 Configtable_insert(cfp);
1231 }
1232 return cfp;
1233}
1234
1235/* Compute the closure of the configuration list */
1236void Configlist_closure(struct lemon *lemp)
1237{
1238 struct config *cfp, *newcfp;
1239 struct rule *rp, *newrp;
1240 struct symbol *sp, *xsp;
1241 int i, dot;
1242
1243 assert( currentend!=0 );
1244 for(cfp=current; cfp; cfp=cfp->next){
1245 rp = cfp->rp;
1246 dot = cfp->dot;
1247 if( dot>=rp->nrhs ) continue;
1248 sp = rp->rhs[dot];
1249 if( sp->type==NONTERMINAL ){
1250 if( sp->rule==0 && sp!=lemp->errsym ){
1251 ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
1252 sp->name);
1253 lemp->errorcnt++;
1254 }
1255 for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
1256 newcfp = Configlist_add(newrp,0);
1257 for(i=dot+1; i<rp->nrhs; i++){
1258 xsp = rp->rhs[i];
1259 if( xsp->type==TERMINAL ){
1260 SetAdd(newcfp->fws,xsp->index);
1261 break;
1262 }else if( xsp->type==MULTITERMINAL ){
1263 int k;
1264 for(k=0; k<xsp->nsubsym; k++){
1265 SetAdd(newcfp->fws, xsp->subsym[k]->index);
1266 }
1267 break;
1268 }else{
1269 SetUnion(newcfp->fws,xsp->firstset);
1270 if( xsp->lambda==LEMON_FALSE ) break;
1271 }
1272 }
1273 if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
1274 }
1275 }
1276 }
1277 return;
1278}
1279
1280/* Sort the configuration list */
1281void Configlist_sort(){
1282 current = (struct config *)msort((char *)current,(char **)&(current->next),Configcmp);
1283 currentend = 0;
1284 return;
1285}
1286
1287/* Sort the basis configuration list */
1288void Configlist_sortbasis(){
1289 basis = (struct config *)msort((char *)current,(char **)&(current->bp),Configcmp);
1290 basisend = 0;
1291 return;
1292}
1293
1294/* Return a pointer to the head of the configuration list and
1295** reset the list */
1296struct config *Configlist_return(){
1297 struct config *old;
1298 old = current;
1299 current = 0;
1300 currentend = 0;
1301 return old;
1302}
1303
1304/* Return a pointer to the head of the configuration list and
1305** reset the list */
1306struct config *Configlist_basis(){
1307 struct config *old;
1308 old = basis;
1309 basis = 0;
1310 basisend = 0;
1311 return old;
1312}
1313
1314/* Free all elements of the given configuration list */
1315void Configlist_eat(struct config *cfp)
1316{
1317 struct config *nextcfp;
1318 for(; cfp; cfp=nextcfp){
1319 nextcfp = cfp->next;
1320 assert( cfp->fplp==0 );
1321 assert( cfp->bplp==0 );
1322 if( cfp->fws ) SetFree(cfp->fws);
1323 deleteconfig(cfp);
1324 }
1325 return;
1326}
1327/***************** From the file "error.c" *********************************/
1328/*
1329** Code for printing error message.
1330*/
1331
1332void ErrorMsg(const char *filename, int lineno, const char *format, ...){
1333 va_list ap;
1334 fprintf(stderr, "%s:%d: ", filename, lineno);
1335 va_start(ap, format);
1336 vfprintf(stderr,format,ap);
1337 va_end(ap);
1338 fprintf(stderr, "\n");
1339}
1340/**************** From the file "main.c" ************************************/
1341/*
1342** Main program file for the LEMON parser generator.
1343*/
1344
1345/* Report an out-of-memory condition and abort. This function
1346** is used mostly by the "MemoryCheck" macro in struct.h
1347*/
1348void memory_error(){
1349 fprintf(stderr,"Out of memory. Aborting...\n");
1350 exit(1);
1351}
1352
1353static int nDefine = 0; /* Number of -D options on the command line */
1354static char **azDefine = 0; /* Name of the -D macros */
1355
1356/* This routine is called with the argument to each -D command-line option.
1357** Add the macro defined to the azDefine array.
1358*/
1359static void handle_D_option(char *z){
1360 char **paz;
1361 nDefine++;
1362 azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine);
1363 if( azDefine==0 ){
1364 fprintf(stderr,"out of memory\n");
1365 exit(1);
1366 }
1367 paz = &azDefine[nDefine-1];
1368 *paz = (char *) malloc( lemonStrlen(z)+1 );
1369 if( *paz==0 ){
1370 fprintf(stderr,"out of memory\n");
1371 exit(1);
1372 }
1373 strcpy(*paz, z);
1374 for(z=*paz; *z && *z!='='; z++){}
1375 *z = 0;
1376}
1377
1378static char *user_templatename = NULL;
1379static void handle_T_option(char *z){
1380 user_templatename = (char *) malloc( lemonStrlen(z)+1 );
1381 if( user_templatename==0 ){
1382 memory_error();
1383 }
1384 strcpy(user_templatename, z);
1385}
1386
1387/* The main program. Parse the command line and do it... */
1388int main(int argc, char **argv)
1389{
1390 static int version = 0;
1391 static int rpflag = 0;
1392 static int basisflag = 0;
1393 static int compress = 0;
1394 static int quiet = 0;
1395 static int statistics = 0;
1396 static int mhflag = 0;
1397 static int nolinenosflag = 0;
1398 static int noResort = 0;
1399 static struct s_options options[] = {
1400 {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
1401 {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
1402 {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
1403 {OPT_FSTR, "T", (char*)handle_T_option, "Specify a template file."},
1404 {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
1405 {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."},
1406 {OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."},
1407 {OPT_FLAG, "p", (char*)&showPrecedenceConflict,
1408 "Show conflicts resolved by precedence rules"},
1409 {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
1410 {OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"},
1411 {OPT_FLAG, "s", (char*)&statistics,
1412 "Print parser stats to standard output."},
1413 {OPT_FLAG, "x", (char*)&version, "Print the version number."},
1414 {OPT_FLAG,0,0,0}
1415 };
1416 int i;
1417 int exitcode;
1418 struct lemon lem;
1419
1420 OptInit(argv,options,stderr);
1421 if( version ){
1422 printf("Lemon version 1.0\n");
1423 exit(0);
1424 }
1425 if( OptNArgs()!=1 ){
1426 fprintf(stderr,"Exactly one filename argument is required.\n");
1427 exit(1);
1428 }
1429 memset(&lem, 0, sizeof(lem));
1430 lem.errorcnt = 0;
1431
1432 /* Initialize the machine */
1433 Strsafe_init();
1434 Symbol_init();
1435 State_init();
1436 lem.argv0 = argv[0];
1437 lem.filename = OptArg(0);
1438 lem.basisflag = basisflag;
1439 lem.nolinenosflag = nolinenosflag;
1440 Symbol_new("$");
1441 lem.errsym = Symbol_new("error");
1442 lem.errsym->useCnt = 0;
1443
1444 /* Parse the input file */
1445 Parse(&lem);
1446 if( lem.errorcnt ) exit(lem.errorcnt);
1447 if( lem.nrule==0 ){
1448 fprintf(stderr,"Empty grammar.\n");
1449 exit(1);
1450 }
1451
1452 /* Count and index the symbols of the grammar */
1453 lem.nsymbol = Symbol_count();
1454 Symbol_new("{default}");
1455 lem.symbols = Symbol_arrayof();
1456 for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
1457 qsort(lem.symbols,lem.nsymbol+1,sizeof(struct symbol*), Symbolcmpp);
1458 for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
1459 for(i=1; isupper(lem.symbols[i]->name[0]); i++);
1460 lem.nterminal = i;
1461
1462 /* Generate a reprint of the grammar, if requested on the command line */
1463 if( rpflag ){
1464 Reprint(&lem);
1465 }else{
1466 /* Initialize the size for all follow and first sets */
1467 SetSize(lem.nterminal+1);
1468
1469 /* Find the precedence for every production rule (that has one) */
1470 FindRulePrecedences(&lem);
1471
1472 /* Compute the lambda-nonterminals and the first-sets for every
1473 ** nonterminal */
1474 FindFirstSets(&lem);
1475
1476 /* Compute all LR(0) states. Also record follow-set propagation
1477 ** links so that the follow-set can be computed later */
1478 lem.nstate = 0;
1479 FindStates(&lem);
1480 lem.sorted = State_arrayof();
1481
1482 /* Tie up loose ends on the propagation links */
1483 FindLinks(&lem);
1484
1485 /* Compute the follow set of every reducible configuration */
1486 FindFollowSets(&lem);
1487
1488 /* Compute the action tables */
1489 FindActions(&lem);
1490
1491 /* Compress the action tables */
1492 if( compress==0 ) CompressTables(&lem);
1493
1494 /* Reorder and renumber the states so that states with fewer choices
1495 ** occur at the end. This is an optimization that helps make the
1496 ** generated parser tables smaller. */
1497 if( noResort==0 ) ResortStates(&lem);
1498
1499 /* Generate a report of the parser generated. (the "y.output" file) */
1500 if( !quiet ) ReportOutput(&lem);
1501
1502 /* Generate the source code for the parser */
1503 ReportTable(&lem, mhflag);
1504
1505 /* Produce a header file for use by the scanner. (This step is
1506 ** omitted if the "-m" option is used because makeheaders will
1507 ** generate the file for us.) */
1508 if( !mhflag ) ReportHeader(&lem);
1509 }
1510 if( statistics ){
1511 printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
1512 lem.nterminal, lem.nsymbol - lem.nterminal, lem.nrule);
1513 printf(" %d states, %d parser table entries, %d conflicts\n",
1514 lem.nstate, lem.tablesize, lem.nconflict);
1515 }
1516 if( lem.nconflict > 0 ){
1517 fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
1518 }
1519
1520 /* return 0 on success, 1 on failure. */
1521 exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0;
1522 exit(exitcode);
1523 return (exitcode);
1524}
1525/******************** From the file "msort.c" *******************************/
1526/*
1527** A generic merge-sort program.
1528**
1529** USAGE:
1530** Let "ptr" be a pointer to some structure which is at the head of
1531** a null-terminated list. Then to sort the list call:
1532**
1533** ptr = msort(ptr,&(ptr->next),cmpfnc);
1534**
1535** In the above, "cmpfnc" is a pointer to a function which compares
1536** two instances of the structure and returns an integer, as in
1537** strcmp. The second argument is a pointer to the pointer to the
1538** second element of the linked list. This address is used to compute
1539** the offset to the "next" field within the structure. The offset to
1540** the "next" field must be constant for all structures in the list.
1541**
1542** The function returns a new pointer which is the head of the list
1543** after sorting.
1544**
1545** ALGORITHM:
1546** Merge-sort.
1547*/
1548
1549/*
1550** Return a pointer to the next structure in the linked list.
1551*/
1552#define NEXT(A) (*(char**)(((unsigned long)A)+offset))
1553
1554/*
1555** Inputs:
1556** a: A sorted, null-terminated linked list. (May be null).
1557** b: A sorted, null-terminated linked list. (May be null).
1558** cmp: A pointer to the comparison function.
1559** offset: Offset in the structure to the "next" field.
1560**
1561** Return Value:
1562** A pointer to the head of a sorted list containing the elements
1563** of both a and b.
1564**
1565** Side effects:
1566** The "next" pointers for elements in the lists a and b are
1567** changed.
1568*/
1569static char *merge(
1570 char *a,
1571 char *b,
1572 int (*cmp)(const char*,const char*),
1573 int offset
1574){
1575 char *ptr, *head;
1576
1577 if( a==0 ){
1578 head = b;
1579 }else if( b==0 ){
1580 head = a;
1581 }else{
1582 if( (*cmp)(a,b)<=0 ){
1583 ptr = a;
1584 a = NEXT(a);
1585 }else{
1586 ptr = b;
1587 b = NEXT(b);
1588 }
1589 head = ptr;
1590 while( a && b ){
1591 if( (*cmp)(a,b)<=0 ){
1592 NEXT(ptr) = a;
1593 ptr = a;
1594 a = NEXT(a);
1595 }else{
1596 NEXT(ptr) = b;
1597 ptr = b;
1598 b = NEXT(b);
1599 }
1600 }
1601 if( a ) NEXT(ptr) = a;
1602 else NEXT(ptr) = b;
1603 }
1604 return head;
1605}
1606
1607/*
1608** Inputs:
1609** list: Pointer to a singly-linked list of structures.
1610** next: Pointer to pointer to the second element of the list.
1611** cmp: A comparison function.
1612**
1613** Return Value:
1614** A pointer to the head of a sorted list containing the elements
1615** orginally in list.
1616**
1617** Side effects:
1618** The "next" pointers for elements in list are changed.
1619*/
1620#define LISTSIZE 30
1621static char *msort(
1622 char *list,
1623 char **next,
1624 int (*cmp)(const char*,const char*)
1625){
1626 unsigned long offset;
1627 char *ep;
1628 char *set[LISTSIZE];
1629 int i;
1630 offset = (unsigned long)next - (unsigned long)list;
1631 for(i=0; i<LISTSIZE; i++) set[i] = 0;
1632 while( list ){
1633 ep = list;
1634 list = NEXT(list);
1635 NEXT(ep) = 0;
1636 for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
1637 ep = merge(ep,set[i],cmp,offset);
1638 set[i] = 0;
1639 }
1640 set[i] = ep;
1641 }
1642 ep = 0;
1643 for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset);
1644 return ep;
1645}
1646/************************ From the file "option.c" **************************/
1647static char **argv;
1648static struct s_options *op;
1649static FILE *errstream;
1650
1651#define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
1652
1653/*
1654** Print the command line with a carrot pointing to the k-th character
1655** of the n-th field.
1656*/
1657static void errline(int n, int k, FILE *err)
1658{
1659 int spcnt, i;
1660 if( argv[0] ) fprintf(err,"%s",argv[0]);
1661 spcnt = lemonStrlen(argv[0]) + 1;
1662 for(i=1; i<n && argv[i]; i++){
1663 fprintf(err," %s",argv[i]);
1664 spcnt += lemonStrlen(argv[i])+1;
1665 }
1666 spcnt += k;
1667 for(; argv[i]; i++) fprintf(err," %s",argv[i]);
1668 if( spcnt<20 ){
1669 fprintf(err,"\n%*s^-- here\n",spcnt,"");
1670 }else{
1671 fprintf(err,"\n%*shere --^\n",spcnt-7,"");
1672 }
1673}
1674
1675/*
1676** Return the index of the N-th non-switch argument. Return -1
1677** if N is out of range.
1678*/
1679static int argindex(int n)
1680{
1681 int i;
1682 int dashdash = 0;
1683 if( argv!=0 && *argv!=0 ){
1684 for(i=1; argv[i]; i++){
1685 if( dashdash || !ISOPT(argv[i]) ){
1686 if( n==0 ) return i;
1687 n--;
1688 }
1689 if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1690 }
1691 }
1692 return -1;
1693}
1694
1695static char emsg[] = "Command line syntax error: ";
1696
1697/*
1698** Process a flag command line argument.
1699*/
1700static int handleflags(int i, FILE *err)
1701{
1702 int v;
1703 int errcnt = 0;
1704 int j;
1705 for(j=0; op[j].label; j++){
1706 if( strncmp(&argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
1707 }
1708 v = argv[i][0]=='-' ? 1 : 0;
1709 if( op[j].label==0 ){
1710 if( err ){
1711 fprintf(err,"%sundefined option.\n",emsg);
1712 errline(i,1,err);
1713 }
1714 errcnt++;
1715 }else if( op[j].type==OPT_FLAG ){
1716 *((int*)op[j].arg) = v;
1717 }else if( op[j].type==OPT_FFLAG ){
1718 (*(void(*)(int))(op[j].arg))(v);
1719 }else if( op[j].type==OPT_FSTR ){
1720 (*(void(*)(char *))(op[j].arg))(&argv[i][2]);
1721 }else{
1722 if( err ){
1723 fprintf(err,"%smissing argument on switch.\n",emsg);
1724 errline(i,1,err);
1725 }
1726 errcnt++;
1727 }
1728 return errcnt;
1729}
1730
1731/*
1732** Process a command line switch which has an argument.
1733*/
1734static int handleswitch(int i, FILE *err)
1735{
1736 int lv = 0;
1737 double dv = 0.0;
1738 char *sv = 0, *end;
1739 char *cp;
1740 int j;
1741 int errcnt = 0;
1742 cp = strchr(argv[i],'=');
1743 assert( cp!=0 );
1744 *cp = 0;
1745 for(j=0; op[j].label; j++){
1746 if( strcmp(argv[i],op[j].label)==0 ) break;
1747 }
1748 *cp = '=';
1749 if( op[j].label==0 ){
1750 if( err ){
1751 fprintf(err,"%sundefined option.\n",emsg);
1752 errline(i,0,err);
1753 }
1754 errcnt++;
1755 }else{
1756 cp++;
1757 switch( op[j].type ){
1758 case OPT_FLAG:
1759 case OPT_FFLAG:
1760 if( err ){
1761 fprintf(err,"%soption requires an argument.\n",emsg);
1762 errline(i,0,err);
1763 }
1764 errcnt++;
1765 break;
1766 case OPT_DBL:
1767 case OPT_FDBL:
1768 dv = strtod(cp,&end);
1769 if( *end ){
1770 if( err ){
1771 fprintf(err,"%sillegal character in floating-point argument.\n",emsg);
1772 errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1773 }
1774 errcnt++;
1775 }
1776 break;
1777 case OPT_INT:
1778 case OPT_FINT:
1779 lv = strtol(cp,&end,0);
1780 if( *end ){
1781 if( err ){
1782 fprintf(err,"%sillegal character in integer argument.\n",emsg);
1783 errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
1784 }
1785 errcnt++;
1786 }
1787 break;
1788 case OPT_STR:
1789 case OPT_FSTR:
1790 sv = cp;
1791 break;
1792 }
1793 switch( op[j].type ){
1794 case OPT_FLAG:
1795 case OPT_FFLAG:
1796 break;
1797 case OPT_DBL:
1798 *(double*)(op[j].arg) = dv;
1799 break;
1800 case OPT_FDBL:
1801 (*(void(*)(double))(op[j].arg))(dv);
1802 break;
1803 case OPT_INT:
1804 *(int*)(op[j].arg) = lv;
1805 break;
1806 case OPT_FINT:
1807 (*(void(*)(int))(op[j].arg))((int)lv);
1808 break;
1809 case OPT_STR:
1810 *(char**)(op[j].arg) = sv;
1811 break;
1812 case OPT_FSTR:
1813 (*(void(*)(char *))(op[j].arg))(sv);
1814 break;
1815 }
1816 }
1817 return errcnt;
1818}
1819
1820int OptInit(char **a, struct s_options *o, FILE *err)
1821{
1822 int errcnt = 0;
1823 argv = a;
1824 op = o;
1825 errstream = err;
1826 if( argv && *argv && op ){
1827 int i;
1828 for(i=1; argv[i]; i++){
1829 if( argv[i][0]=='+' || argv[i][0]=='-' ){
1830 errcnt += handleflags(i,err);
1831 }else if( strchr(argv[i],'=') ){
1832 errcnt += handleswitch(i,err);
1833 }
1834 }
1835 }
1836 if( errcnt>0 ){
1837 fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
1838 OptPrint();
1839 exit(1);
1840 }
1841 return 0;
1842}
1843
1844int OptNArgs(){
1845 int cnt = 0;
1846 int dashdash = 0;
1847 int i;
1848 if( argv!=0 && argv[0]!=0 ){
1849 for(i=1; argv[i]; i++){
1850 if( dashdash || !ISOPT(argv[i]) ) cnt++;
1851 if( strcmp(argv[i],"--")==0 ) dashdash = 1;
1852 }
1853 }
1854 return cnt;
1855}
1856
1857char *OptArg(int n)
1858{
1859 int i;
1860 i = argindex(n);
1861 return i>=0 ? argv[i] : 0;
1862}
1863
1864void OptErr(int n)
1865{
1866 int i;
1867 i = argindex(n);
1868 if( i>=0 ) errline(i,0,errstream);
1869}
1870
1871void OptPrint(){
1872 int i;
1873 int max, len;
1874 max = 0;
1875 for(i=0; op[i].label; i++){
1876 len = lemonStrlen(op[i].label) + 1;
1877 switch( op[i].type ){
1878 case OPT_FLAG:
1879 case OPT_FFLAG:
1880 break;
1881 case OPT_INT:
1882 case OPT_FINT:
1883 len += 9; /* length of "<integer>" */
1884 break;
1885 case OPT_DBL:
1886 case OPT_FDBL:
1887 len += 6; /* length of "<real>" */
1888 break;
1889 case OPT_STR:
1890 case OPT_FSTR:
1891 len += 8; /* length of "<string>" */
1892 break;
1893 }
1894 if( len>max ) max = len;
1895 }
1896 for(i=0; op[i].label; i++){
1897 switch( op[i].type ){
1898 case OPT_FLAG:
1899 case OPT_FFLAG:
1900 fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
1901 break;
1902 case OPT_INT:
1903 case OPT_FINT:
1904 fprintf(errstream," %s=<integer>%*s %s\n",op[i].label,
1905 (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
1906 break;
1907 case OPT_DBL:
1908 case OPT_FDBL:
1909 fprintf(errstream," %s=<real>%*s %s\n",op[i].label,
1910 (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
1911 break;
1912 case OPT_STR:
1913 case OPT_FSTR:
1914 fprintf(errstream," %s=<string>%*s %s\n",op[i].label,
1915 (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
1916 break;
1917 }
1918 }
1919}
1920/*********************** From the file "parse.c" ****************************/
1921/*
1922** Input file parser for the LEMON parser generator.
1923*/
1924
1925/* The state of the parser */
1926enum e_state {
1927 INITIALIZE,
1928 WAITING_FOR_DECL_OR_RULE,
1929 WAITING_FOR_DECL_KEYWORD,
1930 WAITING_FOR_DECL_ARG,
1931 WAITING_FOR_PRECEDENCE_SYMBOL,
1932 WAITING_FOR_ARROW,
1933 IN_RHS,
1934 LHS_ALIAS_1,
1935 LHS_ALIAS_2,
1936 LHS_ALIAS_3,
1937 RHS_ALIAS_1,
1938 RHS_ALIAS_2,
1939 PRECEDENCE_MARK_1,
1940 PRECEDENCE_MARK_2,
1941 RESYNC_AFTER_RULE_ERROR,
1942 RESYNC_AFTER_DECL_ERROR,
1943 WAITING_FOR_DESTRUCTOR_SYMBOL,
1944 WAITING_FOR_DATATYPE_SYMBOL,
1945 WAITING_FOR_FALLBACK_ID,
1946 WAITING_FOR_WILDCARD_ID
1947};
1948struct pstate {
1949 char *filename; /* Name of the input file */
1950 int tokenlineno; /* Linenumber at which current token starts */
1951 int errorcnt; /* Number of errors so far */
1952 char *tokenstart; /* Text of current token */
1953 struct lemon *gp; /* Global state vector */
1954 enum e_state state; /* The state of the parser */
1955 struct symbol *fallback; /* The fallback token */
1956 struct symbol *lhs; /* Left-hand side of current rule */
1957 const char *lhsalias; /* Alias for the LHS */
1958 int nrhs; /* Number of right-hand side symbols seen */
1959 struct symbol *rhs[MAXRHS]; /* RHS symbols */
1960 const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
1961 struct rule *prevrule; /* Previous rule parsed */
1962 const char *declkeyword; /* Keyword of a declaration */
1963 char **declargslot; /* Where the declaration argument should be put */
1964 int insertLineMacro; /* Add #line before declaration insert */
1965 int *decllinenoslot; /* Where to write declaration line number */
1966 enum e_assoc declassoc; /* Assign this association to decl arguments */
1967 int preccounter; /* Assign this precedence to decl arguments */
1968 struct rule *firstrule; /* Pointer to first rule in the grammar */
1969 struct rule *lastrule; /* Pointer to the most recently parsed rule */
1970};
1971
1972/* Parse a single token */
1973static void parseonetoken(struct pstate *psp)
1974{
1975 const char *x;
1976 x = Strsafe(psp->tokenstart); /* Save the token permanently */
1977#if 0
1978 printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
1979 x,psp->state);
1980#endif
1981 switch( psp->state ){
1982 case INITIALIZE:
1983 psp->prevrule = 0;
1984 psp->preccounter = 0;
1985 psp->firstrule = psp->lastrule = 0;
1986 psp->gp->nrule = 0;
1987 /* Fall thru to next case */
1988 case WAITING_FOR_DECL_OR_RULE:
1989 if( x[0]=='%' ){
1990 psp->state = WAITING_FOR_DECL_KEYWORD;
1991 }else if( islower(x[0]) ){
1992 psp->lhs = Symbol_new(x);
1993 psp->nrhs = 0;
1994 psp->lhsalias = 0;
1995 psp->state = WAITING_FOR_ARROW;
1996 }else if( x[0]=='{' ){
1997 if( psp->prevrule==0 ){
1998 ErrorMsg(psp->filename,psp->tokenlineno,
1999"There is no prior rule opon which to attach the code \
2000fragment which begins on this line.");
2001 psp->errorcnt++;
2002 }else if( psp->prevrule->code!=0 ){
2003 ErrorMsg(psp->filename,psp->tokenlineno,
2004"Code fragment beginning on this line is not the first \
2005to follow the previous rule.");
2006 psp->errorcnt++;
2007 }else{
2008 psp->prevrule->line = psp->tokenlineno;
2009 psp->prevrule->code = &x[1];
2010 }
2011 }else if( x[0]=='[' ){
2012 psp->state = PRECEDENCE_MARK_1;
2013 }else{
2014 ErrorMsg(psp->filename,psp->tokenlineno,
2015 "Token \"%s\" should be either \"%%\" or a nonterminal name.",
2016 x);
2017 psp->errorcnt++;
2018 }
2019 break;
2020 case PRECEDENCE_MARK_1:
2021 if( !isupper(x[0]) ){
2022 ErrorMsg(psp->filename,psp->tokenlineno,
2023 "The precedence symbol must be a terminal.");
2024 psp->errorcnt++;
2025 }else if( psp->prevrule==0 ){
2026 ErrorMsg(psp->filename,psp->tokenlineno,
2027 "There is no prior rule to assign precedence \"[%s]\".",x);
2028 psp->errorcnt++;
2029 }else if( psp->prevrule->precsym!=0 ){
2030 ErrorMsg(psp->filename,psp->tokenlineno,
2031"Precedence mark on this line is not the first \
2032to follow the previous rule.");
2033 psp->errorcnt++;
2034 }else{
2035 psp->prevrule->precsym = Symbol_new(x);
2036 }
2037 psp->state = PRECEDENCE_MARK_2;
2038 break;
2039 case PRECEDENCE_MARK_2:
2040 if( x[0]!=']' ){
2041 ErrorMsg(psp->filename,psp->tokenlineno,
2042 "Missing \"]\" on precedence mark.");
2043 psp->errorcnt++;
2044 }
2045 psp->state = WAITING_FOR_DECL_OR_RULE;
2046 break;
2047 case WAITING_FOR_ARROW:
2048 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2049 psp->state = IN_RHS;
2050 }else if( x[0]=='(' ){
2051 psp->state = LHS_ALIAS_1;
2052 }else{
2053 ErrorMsg(psp->filename,psp->tokenlineno,
2054 "Expected to see a \":\" following the LHS symbol \"%s\".",
2055 psp->lhs->name);
2056 psp->errorcnt++;
2057 psp->state = RESYNC_AFTER_RULE_ERROR;
2058 }
2059 break;
2060 case LHS_ALIAS_1:
2061 if( isalpha(x[0]) ){
2062 psp->lhsalias = x;
2063 psp->state = LHS_ALIAS_2;
2064 }else{
2065 ErrorMsg(psp->filename,psp->tokenlineno,
2066 "\"%s\" is not a valid alias for the LHS \"%s\"\n",
2067 x,psp->lhs->name);
2068 psp->errorcnt++;
2069 psp->state = RESYNC_AFTER_RULE_ERROR;
2070 }
2071 break;
2072 case LHS_ALIAS_2:
2073 if( x[0]==')' ){
2074 psp->state = LHS_ALIAS_3;
2075 }else{
2076 ErrorMsg(psp->filename,psp->tokenlineno,
2077 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2078 psp->errorcnt++;
2079 psp->state = RESYNC_AFTER_RULE_ERROR;
2080 }
2081 break;
2082 case LHS_ALIAS_3:
2083 if( x[0]==':' && x[1]==':' && x[2]=='=' ){
2084 psp->state = IN_RHS;
2085 }else{
2086 ErrorMsg(psp->filename,psp->tokenlineno,
2087 "Missing \"->\" following: \"%s(%s)\".",
2088 psp->lhs->name,psp->lhsalias);
2089 psp->errorcnt++;
2090 psp->state = RESYNC_AFTER_RULE_ERROR;
2091 }
2092 break;
2093 case IN_RHS:
2094 if( x[0]=='.' ){
2095 struct rule *rp;
2096 rp = (struct rule *)calloc( sizeof(struct rule) +
2097 sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
2098 if( rp==0 ){
2099 ErrorMsg(psp->filename,psp->tokenlineno,
2100 "Can't allocate enough memory for this rule.");
2101 psp->errorcnt++;
2102 psp->prevrule = 0;
2103 }else{
2104 int i;
2105 rp->ruleline = psp->tokenlineno;
2106 rp->rhs = (struct symbol**)&rp[1];
2107 rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]);
2108 for(i=0; i<psp->nrhs; i++){
2109 rp->rhs[i] = psp->rhs[i];
2110 rp->rhsalias[i] = psp->alias[i];
2111 }
2112 rp->lhs = psp->lhs;
2113 rp->lhsalias = psp->lhsalias;
2114 rp->nrhs = psp->nrhs;
2115 rp->code = 0;
2116 rp->precsym = 0;
2117 rp->index = psp->gp->nrule++;
2118 rp->nextlhs = rp->lhs->rule;
2119 rp->lhs->rule = rp;
2120 rp->next = 0;
2121 if( psp->firstrule==0 ){
2122 psp->firstrule = psp->lastrule = rp;
2123 }else{
2124 psp->lastrule->next = rp;
2125 psp->lastrule = rp;
2126 }
2127 psp->prevrule = rp;
2128 }
2129 psp->state = WAITING_FOR_DECL_OR_RULE;
2130 }else if( isalpha(x[0]) ){
2131 if( psp->nrhs>=MAXRHS ){
2132 ErrorMsg(psp->filename,psp->tokenlineno,
2133 "Too many symbols on RHS of rule beginning at \"%s\".",
2134 x);
2135 psp->errorcnt++;
2136 psp->state = RESYNC_AFTER_RULE_ERROR;
2137 }else{
2138 psp->rhs[psp->nrhs] = Symbol_new(x);
2139 psp->alias[psp->nrhs] = 0;
2140 psp->nrhs++;
2141 }
2142 }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 ){
2143 struct symbol *msp = psp->rhs[psp->nrhs-1];
2144 if( msp->type!=MULTITERMINAL ){
2145 struct symbol *origsp = msp;
2146 msp = (struct symbol *) calloc(1,sizeof(*msp));
2147 memset(msp, 0, sizeof(*msp));
2148 msp->type = MULTITERMINAL;
2149 msp->nsubsym = 1;
2150 msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*));
2151 msp->subsym[0] = origsp;
2152 msp->name = origsp->name;
2153 psp->rhs[psp->nrhs-1] = msp;
2154 }
2155 msp->nsubsym++;
2156 msp->subsym = (struct symbol **) realloc(msp->subsym,
2157 sizeof(struct symbol*)*msp->nsubsym);
2158 msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
2159 if( islower(x[1]) || islower(msp->subsym[0]->name[0]) ){
2160 ErrorMsg(psp->filename,psp->tokenlineno,
2161 "Cannot form a compound containing a non-terminal");
2162 psp->errorcnt++;
2163 }
2164 }else if( x[0]=='(' && psp->nrhs>0 ){
2165 psp->state = RHS_ALIAS_1;
2166 }else{
2167 ErrorMsg(psp->filename,psp->tokenlineno,
2168 "Illegal character on RHS of rule: \"%s\".",x);
2169 psp->errorcnt++;
2170 psp->state = RESYNC_AFTER_RULE_ERROR;
2171 }
2172 break;
2173 case RHS_ALIAS_1:
2174 if( isalpha(x[0]) ){
2175 psp->alias[psp->nrhs-1] = x;
2176 psp->state = RHS_ALIAS_2;
2177 }else{
2178 ErrorMsg(psp->filename,psp->tokenlineno,
2179 "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
2180 x,psp->rhs[psp->nrhs-1]->name);
2181 psp->errorcnt++;
2182 psp->state = RESYNC_AFTER_RULE_ERROR;
2183 }
2184 break;
2185 case RHS_ALIAS_2:
2186 if( x[0]==')' ){
2187 psp->state = IN_RHS;
2188 }else{
2189 ErrorMsg(psp->filename,psp->tokenlineno,
2190 "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
2191 psp->errorcnt++;
2192 psp->state = RESYNC_AFTER_RULE_ERROR;
2193 }
2194 break;
2195 case WAITING_FOR_DECL_KEYWORD:
2196 if( isalpha(x[0]) ){
2197 psp->declkeyword = x;
2198 psp->declargslot = 0;
2199 psp->decllinenoslot = 0;
2200 psp->insertLineMacro = 1;
2201 psp->state = WAITING_FOR_DECL_ARG;
2202 if( strcmp(x,"name")==0 ){
2203 psp->declargslot = &(psp->gp->name);
2204 psp->insertLineMacro = 0;
2205 }else if( strcmp(x,"include")==0 ){
2206 psp->declargslot = &(psp->gp->include);
2207 }else if( strcmp(x,"code")==0 ){
2208 psp->declargslot = &(psp->gp->extracode);
2209 }else if( strcmp(x,"token_destructor")==0 ){
2210 psp->declargslot = &psp->gp->tokendest;
2211 }else if( strcmp(x,"default_destructor")==0 ){
2212 psp->declargslot = &psp->gp->vardest;
2213 }else if( strcmp(x,"token_prefix")==0 ){
2214 psp->declargslot = &psp->gp->tokenprefix;
2215 psp->insertLineMacro = 0;
2216 }else if( strcmp(x,"syntax_error")==0 ){
2217 psp->declargslot = &(psp->gp->error);
2218 }else if( strcmp(x,"parse_accept")==0 ){
2219 psp->declargslot = &(psp->gp->accept);
2220 }else if( strcmp(x,"parse_failure")==0 ){
2221 psp->declargslot = &(psp->gp->failure);
2222 }else if( strcmp(x,"stack_overflow")==0 ){
2223 psp->declargslot = &(psp->gp->overflow);
2224 }else if( strcmp(x,"extra_argument")==0 ){
2225 psp->declargslot = &(psp->gp->arg);
2226 psp->insertLineMacro = 0;
2227 }else if( strcmp(x,"token_type")==0 ){
2228 psp->declargslot = &(psp->gp->tokentype);
2229 psp->insertLineMacro = 0;
2230 }else if( strcmp(x,"default_type")==0 ){
2231 psp->declargslot = &(psp->gp->vartype);
2232 psp->insertLineMacro = 0;
2233 }else if( strcmp(x,"stack_size")==0 ){
2234 psp->declargslot = &(psp->gp->stacksize);
2235 psp->insertLineMacro = 0;
2236 }else if( strcmp(x,"start_symbol")==0 ){
2237 psp->declargslot = &(psp->gp->start);
2238 psp->insertLineMacro = 0;
2239 }else if( strcmp(x,"left")==0 ){
2240 psp->preccounter++;
2241 psp->declassoc = LEFT;
2242 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2243 }else if( strcmp(x,"right")==0 ){
2244 psp->preccounter++;
2245 psp->declassoc = RIGHT;
2246 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2247 }else if( strcmp(x,"nonassoc")==0 ){
2248 psp->preccounter++;
2249 psp->declassoc = NONE;
2250 psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
2251 }else if( strcmp(x,"destructor")==0 ){
2252 psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
2253 }else if( strcmp(x,"type")==0 ){
2254 psp->state = WAITING_FOR_DATATYPE_SYMBOL;
2255 }else if( strcmp(x,"fallback")==0 ){
2256 psp->fallback = 0;
2257 psp->state = WAITING_FOR_FALLBACK_ID;
2258 }else if( strcmp(x,"wildcard")==0 ){
2259 psp->state = WAITING_FOR_WILDCARD_ID;
2260 }else{
2261 ErrorMsg(psp->filename,psp->tokenlineno,
2262 "Unknown declaration keyword: \"%%%s\".",x);
2263 psp->errorcnt++;
2264 psp->state = RESYNC_AFTER_DECL_ERROR;
2265 }
2266 }else{
2267 ErrorMsg(psp->filename,psp->tokenlineno,
2268 "Illegal declaration keyword: \"%s\".",x);
2269 psp->errorcnt++;
2270 psp->state = RESYNC_AFTER_DECL_ERROR;
2271 }
2272 break;
2273 case WAITING_FOR_DESTRUCTOR_SYMBOL:
2274 if( !isalpha(x[0]) ){
2275 ErrorMsg(psp->filename,psp->tokenlineno,
2276 "Symbol name missing after %%destructor keyword");
2277 psp->errorcnt++;
2278 psp->state = RESYNC_AFTER_DECL_ERROR;
2279 }else{
2280 struct symbol *sp = Symbol_new(x);
2281 psp->declargslot = &sp->destructor;
2282 psp->decllinenoslot = &sp->destLineno;
2283 psp->insertLineMacro = 1;
2284 psp->state = WAITING_FOR_DECL_ARG;
2285 }
2286 break;
2287 case WAITING_FOR_DATATYPE_SYMBOL:
2288 if( !isalpha(x[0]) ){
2289 ErrorMsg(psp->filename,psp->tokenlineno,
2290 "Symbol name missing after %%type keyword");
2291 psp->errorcnt++;
2292 psp->state = RESYNC_AFTER_DECL_ERROR;
2293 }else{
2294 struct symbol *sp = Symbol_find(x);
2295 if((sp) && (sp->datatype)){
2296 ErrorMsg(psp->filename,psp->tokenlineno,
2297 "Symbol %%type \"%s\" already defined", x);
2298 psp->errorcnt++;
2299 psp->state = RESYNC_AFTER_DECL_ERROR;
2300 }else{
2301 if (!sp){
2302 sp = Symbol_new(x);
2303 }
2304 psp->declargslot = &sp->datatype;
2305 psp->insertLineMacro = 0;
2306 psp->state = WAITING_FOR_DECL_ARG;
2307 }
2308 }
2309 break;
2310 case WAITING_FOR_PRECEDENCE_SYMBOL:
2311 if( x[0]=='.' ){
2312 psp->state = WAITING_FOR_DECL_OR_RULE;
2313 }else if( isupper(x[0]) ){
2314 struct symbol *sp;
2315 sp = Symbol_new(x);
2316 if( sp->prec>=0 ){
2317 ErrorMsg(psp->filename,psp->tokenlineno,
2318 "Symbol \"%s\" has already be given a precedence.",x);
2319 psp->errorcnt++;
2320 }else{
2321 sp->prec = psp->preccounter;
2322 sp->assoc = psp->declassoc;
2323 }
2324 }else{
2325 ErrorMsg(psp->filename,psp->tokenlineno,
2326 "Can't assign a precedence to \"%s\".",x);
2327 psp->errorcnt++;
2328 }
2329 break;
2330 case WAITING_FOR_DECL_ARG:
2331 if( x[0]=='{' || x[0]=='\"' || isalnum(x[0]) ){
2332 const char *zOld, *zNew;
2333 char *zBuf, *z;
2334 int nOld, n, nLine, nNew, nBack;
2335 int addLineMacro;
2336 char zLine[50];
2337 zNew = x;
2338 if( zNew[0]=='"' || zNew[0]=='{' ) zNew++;
2339 nNew = lemonStrlen(zNew);
2340 if( *psp->declargslot ){
2341 zOld = *psp->declargslot;
2342 }else{
2343 zOld = "";
2344 }
2345 nOld = lemonStrlen(zOld);
2346 n = nOld + nNew + 20;
2347 addLineMacro = !psp->gp->nolinenosflag && psp->insertLineMacro &&
2348 (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0);
2349 if( addLineMacro ){
2350 for(z=psp->filename, nBack=0; *z; z++){
2351 if( *z=='\\' ) nBack++;
2352 }
2353 sprintf(zLine, "#line %d ", psp->tokenlineno);
2354 nLine = lemonStrlen(zLine);
2355 n += nLine + lemonStrlen(psp->filename) + nBack;
2356 }
2357 *psp->declargslot = (char *) realloc(*psp->declargslot, n);
2358 zBuf = *psp->declargslot + nOld;
2359 if( addLineMacro ){
2360 if( nOld && zBuf[-1]!='\n' ){
2361 *(zBuf++) = '\n';
2362 }
2363 memcpy(zBuf, zLine, nLine);
2364 zBuf += nLine;
2365 *(zBuf++) = '"';
2366 for(z=psp->filename; *z; z++){
2367 if( *z=='\\' ){
2368 *(zBuf++) = '\\';
2369 }
2370 *(zBuf++) = *z;
2371 }
2372 *(zBuf++) = '"';
2373 *(zBuf++) = '\n';
2374 }
2375 if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){
2376 psp->decllinenoslot[0] = psp->tokenlineno;
2377 }
2378 memcpy(zBuf, zNew, nNew);
2379 zBuf += nNew;
2380 *zBuf = 0;
2381 psp->state = WAITING_FOR_DECL_OR_RULE;
2382 }else{
2383 ErrorMsg(psp->filename,psp->tokenlineno,
2384 "Illegal argument to %%%s: %s",psp->declkeyword,x);
2385 psp->errorcnt++;
2386 psp->state = RESYNC_AFTER_DECL_ERROR;
2387 }
2388 break;
2389 case WAITING_FOR_FALLBACK_ID:
2390 if( x[0]=='.' ){
2391 psp->state = WAITING_FOR_DECL_OR_RULE;
2392 }else if( !isupper(x[0]) ){
2393 ErrorMsg(psp->filename, psp->tokenlineno,
2394 "%%fallback argument \"%s\" should be a token", x);
2395 psp->errorcnt++;
2396 }else{
2397 struct symbol *sp = Symbol_new(x);
2398 if( psp->fallback==0 ){
2399 psp->fallback = sp;
2400 }else if( sp->fallback ){
2401 ErrorMsg(psp->filename, psp->tokenlineno,
2402 "More than one fallback assigned to token %s", x);
2403 psp->errorcnt++;
2404 }else{
2405 sp->fallback = psp->fallback;
2406 psp->gp->has_fallback = 1;
2407 }
2408 }
2409 break;
2410 case WAITING_FOR_WILDCARD_ID:
2411 if( x[0]=='.' ){
2412 psp->state = WAITING_FOR_DECL_OR_RULE;
2413 }else if( !isupper(x[0]) ){
2414 ErrorMsg(psp->filename, psp->tokenlineno,
2415 "%%wildcard argument \"%s\" should be a token", x);
2416 psp->errorcnt++;
2417 }else{
2418 struct symbol *sp = Symbol_new(x);
2419 if( psp->gp->wildcard==0 ){
2420 psp->gp->wildcard = sp;
2421 }else{
2422 ErrorMsg(psp->filename, psp->tokenlineno,
2423 "Extra wildcard to token: %s", x);
2424 psp->errorcnt++;
2425 }
2426 }
2427 break;
2428 case RESYNC_AFTER_RULE_ERROR:
2429/* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2430** break; */
2431 case RESYNC_AFTER_DECL_ERROR:
2432 if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
2433 if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
2434 break;
2435 }
2436}
2437
2438/* Run the preprocessor over the input file text. The global variables
2439** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
2440** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
2441** comments them out. Text in between is also commented out as appropriate.
2442*/
2443static void preprocess_input(char *z){
2444 int i, j, k, n;
2445 int exclude = 0;
2446 int start = 0;
2447 int lineno = 1;
2448 int start_lineno = 1;
2449 for(i=0; z[i]; i++){
2450 if( z[i]=='\n' ) lineno++;
2451 if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue;
2452 if( strncmp(&z[i],"%endif",6)==0 && isspace(z[i+6]) ){
2453 if( exclude ){
2454 exclude--;
2455 if( exclude==0 ){
2456 for(j=start; j<i; j++) if( z[j]!='\n' ) z[j] = ' ';
2457 }
2458 }
2459 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2460 }else if( (strncmp(&z[i],"%ifdef",6)==0 && isspace(z[i+6]))
2461 || (strncmp(&z[i],"%ifndef",7)==0 && isspace(z[i+7])) ){
2462 if( exclude ){
2463 exclude++;
2464 }else{
2465 for(j=i+7; isspace(z[j]); j++){}
2466 for(n=0; z[j+n] && !isspace(z[j+n]); n++){}
2467 exclude = 1;
2468 for(k=0; k<nDefine; k++){
2469 if( strncmp(azDefine[k],&z[j],n)==0 && lemonStrlen(azDefine[k])==n ){
2470 exclude = 0;
2471 break;
2472 }
2473 }
2474 if( z[i+3]=='n' ) exclude = !exclude;
2475 if( exclude ){
2476 start = i;
2477 start_lineno = lineno;
2478 }
2479 }
2480 for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' ';
2481 }
2482 }
2483 if( exclude ){
2484 fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno);
2485 exit(1);
2486 }
2487}
2488
2489/* In spite of its name, this function is really a scanner. It read
2490** in the entire input file (all at once) then tokenizes it. Each
2491** token is passed to the function "parseonetoken" which builds all
2492** the appropriate data structures in the global state vector "gp".
2493*/
2494void Parse(struct lemon *gp)
2495{
2496 struct pstate ps;
2497 FILE *fp;
2498 char *filebuf;
2499 int filesize;
2500 int lineno;
2501 int c;
2502 char *cp, *nextcp;
2503 int startline = 0;
2504
2505 memset(&ps, '\0', sizeof(ps));
2506 ps.gp = gp;
2507 ps.filename = gp->filename;
2508 ps.errorcnt = 0;
2509 ps.state = INITIALIZE;
2510
2511 /* Begin by reading the input file */
2512 fp = fopen(ps.filename,"rb");
2513 if( fp==0 ){
2514 ErrorMsg(ps.filename,0,"Can't open this file for reading.");
2515 gp->errorcnt++;
2516 return;
2517 }
2518 fseek(fp,0,2);
2519 filesize = ftell(fp);
2520 rewind(fp);
2521 filebuf = (char *)malloc( filesize+1 );
2522 if( filebuf==0 ){
2523 ErrorMsg(ps.filename,0,"Can't allocate %d of memory to hold this file.",
2524 filesize+1);
2525 gp->errorcnt++;
2526 fclose(fp);
2527 return;
2528 }
2529 if( fread(filebuf,1,filesize,fp)!=filesize ){
2530 ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.",
2531 filesize);
2532 free(filebuf);
2533 gp->errorcnt++;
2534 fclose(fp);
2535 return;
2536 }
2537 fclose(fp);
2538 filebuf[filesize] = 0;
2539
2540 /* Make an initial pass through the file to handle %ifdef and %ifndef */
2541 preprocess_input(filebuf);
2542
2543 /* Now scan the text of the input file */
2544 lineno = 1;
2545 for(cp=filebuf; (c= *cp)!=0; ){
2546 if( c=='\n' ) lineno++; /* Keep track of the line number */
2547 if( isspace(c) ){ cp++; continue; } /* Skip all white space */
2548 if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */
2549 cp+=2;
2550 while( (c= *cp)!=0 && c!='\n' ) cp++;
2551 continue;
2552 }
2553 if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */
2554 cp+=2;
2555 while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){
2556 if( c=='\n' ) lineno++;
2557 cp++;
2558 }
2559 if( c ) cp++;
2560 continue;
2561 }
2562 ps.tokenstart = cp; /* Mark the beginning of the token */
2563 ps.tokenlineno = lineno; /* Linenumber on which token begins */
2564 if( c=='\"' ){ /* String literals */
2565 cp++;
2566 while( (c= *cp)!=0 && c!='\"' ){
2567 if( c=='\n' ) lineno++;
2568 cp++;
2569 }
2570 if( c==0 ){
2571 ErrorMsg(ps.filename,startline,
2572"String starting on this line is not terminated before the end of the file.");
2573 ps.errorcnt++;
2574 nextcp = cp;
2575 }else{
2576 nextcp = cp+1;
2577 }
2578 }else if( c=='{' ){ /* A block of C code */
2579 int level;
2580 cp++;
2581 for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){
2582 if( c=='\n' ) lineno++;
2583 else if( c=='{' ) level++;
2584 else if( c=='}' ) level--;
2585 else if( c=='/' && cp[1]=='*' ){ /* Skip comments */
2586 int prevc;
2587 cp = &cp[2];
2588 prevc = 0;
2589 while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){
2590 if( c=='\n' ) lineno++;
2591 prevc = c;
2592 cp++;
2593 }
2594 }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */
2595 cp = &cp[2];
2596 while( (c= *cp)!=0 && c!='\n' ) cp++;
2597 if( c ) lineno++;
2598 }else if( c=='\'' || c=='\"' ){ /* String a character literals */
2599 int startchar, prevc;
2600 startchar = c;
2601 prevc = 0;
2602 for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){
2603 if( c=='\n' ) lineno++;
2604 if( prevc=='\\' ) prevc = 0;
2605 else prevc = c;
2606 }
2607 }
2608 }
2609 if( c==0 ){
2610 ErrorMsg(ps.filename,ps.tokenlineno,
2611"C code starting on this line is not terminated before the end of the file.");
2612 ps.errorcnt++;
2613 nextcp = cp;
2614 }else{
2615 nextcp = cp+1;
2616 }
2617 }else if( isalnum(c) ){ /* Identifiers */
2618 while( (c= *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
2619 nextcp = cp;
2620 }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
2621 cp += 3;
2622 nextcp = cp;
2623 }else if( (c=='/' || c=='|') && isalpha(cp[1]) ){
2624 cp += 2;
2625 while( (c = *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
2626 nextcp = cp;
2627 }else{ /* All other (one character) operators */
2628 cp++;
2629 nextcp = cp;
2630 }
2631 c = *cp;
2632 *cp = 0; /* Null terminate the token */
2633 parseonetoken(&ps); /* Parse the token */
2634 *cp = c; /* Restore the buffer */
2635 cp = nextcp;
2636 }
2637 free(filebuf); /* Release the buffer after parsing */
2638 gp->rule = ps.firstrule;
2639 gp->errorcnt = ps.errorcnt;
2640}
2641/*************************** From the file "plink.c" *********************/
2642/*
2643** Routines processing configuration follow-set propagation links
2644** in the LEMON parser generator.
2645*/
2646static struct plink *plink_freelist = 0;
2647
2648/* Allocate a new plink */
2649struct plink *Plink_new(){
2650 struct plink *newlink;
2651
2652 if( plink_freelist==0 ){
2653 int i;
2654 int amt = 100;
2655 plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) );
2656 if( plink_freelist==0 ){
2657 fprintf(stderr,
2658 "Unable to allocate memory for a new follow-set propagation link.\n");
2659 exit(1);
2660 }
2661 for(i=0; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1];
2662 plink_freelist[amt-1].next = 0;
2663 }
2664 newlink = plink_freelist;
2665 plink_freelist = plink_freelist->next;
2666 return newlink;
2667}
2668
2669/* Add a plink to a plink list */
2670void Plink_add(struct plink **plpp, struct config *cfp)
2671{
2672 struct plink *newlink;
2673 newlink = Plink_new();
2674 newlink->next = *plpp;
2675 *plpp = newlink;
2676 newlink->cfp = cfp;
2677}
2678
2679/* Transfer every plink on the list "from" to the list "to" */
2680void Plink_copy(struct plink **to, struct plink *from)
2681{
2682 struct plink *nextpl;
2683 while( from ){
2684 nextpl = from->next;
2685 from->next = *to;
2686 *to = from;
2687 from = nextpl;
2688 }
2689}
2690
2691/* Delete every plink on the list */
2692void Plink_delete(struct plink *plp)
2693{
2694 struct plink *nextpl;
2695
2696 while( plp ){
2697 nextpl = plp->next;
2698 plp->next = plink_freelist;
2699 plink_freelist = plp;
2700 plp = nextpl;
2701 }
2702}
2703/*********************** From the file "report.c" **************************/
2704/*
2705** Procedures for generating reports and tables in the LEMON parser generator.
2706*/
2707
2708/* Generate a filename with the given suffix. Space to hold the
2709** name comes from malloc() and must be freed by the calling
2710** function.
2711*/
2712PRIVATE char *file_makename(struct lemon *lemp, const char *suffix)
2713{
2714 char *name;
2715 char *cp;
2716
2717 name = (char*)malloc( lemonStrlen(lemp->filename) + lemonStrlen(suffix) + 5 );
2718 if( name==0 ){
2719 fprintf(stderr,"Can't allocate space for a filename.\n");
2720 exit(1);
2721 }
2722 strcpy(name,lemp->filename);
2723 cp = strrchr(name,'.');
2724 if( cp ) *cp = 0;
2725 strcat(name,suffix);
2726 return name;
2727}
2728
2729/* Open a file with a name based on the name of the input file,
2730** but with a different (specified) suffix, and return a pointer
2731** to the stream */
2732PRIVATE FILE *file_open(
2733 struct lemon *lemp,
2734 const char *suffix,
2735 const char *mode
2736){
2737 FILE *fp;
2738
2739 if( lemp->outname ) free(lemp->outname);
2740 lemp->outname = file_makename(lemp, suffix);
2741 fp = fopen(lemp->outname,mode);
2742 if( fp==0 && *mode=='w' ){
2743 fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname);
2744 lemp->errorcnt++;
2745 return 0;
2746 }
2747 return fp;
2748}
2749
2750/* Duplicate the input file without comments and without actions
2751** on rules */
2752void Reprint(struct lemon *lemp)
2753{
2754 struct rule *rp;
2755 struct symbol *sp;
2756 int i, j, maxlen, len, ncolumns, skip;
2757 printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename);
2758 maxlen = 10;
2759 for(i=0; i<lemp->nsymbol; i++){
2760 sp = lemp->symbols[i];
2761 len = lemonStrlen(sp->name);
2762 if( len>maxlen ) maxlen = len;
2763 }
2764 ncolumns = 76/(maxlen+5);
2765 if( ncolumns<1 ) ncolumns = 1;
2766 skip = (lemp->nsymbol + ncolumns - 1)/ncolumns;
2767 for(i=0; i<skip; i++){
2768 printf("//");
2769 for(j=i; j<lemp->nsymbol; j+=skip){
2770 sp = lemp->symbols[j];
2771 assert( sp->index==j );
2772 printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
2773 }
2774 printf("\n");
2775 }
2776 for(rp=lemp->rule; rp; rp=rp->next){
2777 printf("%s",rp->lhs->name);
2778 /* if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
2779 printf(" ::=");
2780 for(i=0; i<rp->nrhs; i++){
2781 sp = rp->rhs[i];
2782 printf(" %s", sp->name);
2783 if( sp->type==MULTITERMINAL ){
2784 for(j=1; j<sp->nsubsym; j++){
2785 printf("|%s", sp->subsym[j]->name);
2786 }
2787 }
2788 /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
2789 }
2790 printf(".");
2791 if( rp->precsym ) printf(" [%s]",rp->precsym->name);
2792 /* if( rp->code ) printf("\n %s",rp->code); */
2793 printf("\n");
2794 }
2795}
2796
2797void ConfigPrint(FILE *fp, struct config *cfp)
2798{
2799 struct rule *rp;
2800 struct symbol *sp;
2801 int i, j;
2802 rp = cfp->rp;
2803 fprintf(fp,"%s ::=",rp->lhs->name);
2804 for(i=0; i<=rp->nrhs; i++){
2805 if( i==cfp->dot ) fprintf(fp," *");
2806 if( i==rp->nrhs ) break;
2807 sp = rp->rhs[i];
2808 fprintf(fp," %s", sp->name);
2809 if( sp->type==MULTITERMINAL ){
2810 for(j=1; j<sp->nsubsym; j++){
2811 fprintf(fp,"|%s",sp->subsym[j]->name);
2812 }
2813 }
2814 }
2815}
2816
2817/* #define TEST */
2818#if 0
2819/* Print a set */
2820PRIVATE void SetPrint(out,set,lemp)
2821FILE *out;
2822char *set;
2823struct lemon *lemp;
2824{
2825 int i;
2826 char *spacer;
2827 spacer = "";
2828 fprintf(out,"%12s[","");
2829 for(i=0; i<lemp->nterminal; i++){
2830 if( SetFind(set,i) ){
2831 fprintf(out,"%s%s",spacer,lemp->symbols[i]->name);
2832 spacer = " ";
2833 }
2834 }
2835 fprintf(out,"]\n");
2836}
2837
2838/* Print a plink chain */
2839PRIVATE void PlinkPrint(out,plp,tag)
2840FILE *out;
2841struct plink *plp;
2842char *tag;
2843{
2844 while( plp ){
2845 fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum);
2846 ConfigPrint(out,plp->cfp);
2847 fprintf(out,"\n");
2848 plp = plp->next;
2849 }
2850}
2851#endif
2852
2853/* Print an action to the given file descriptor. Return FALSE if
2854** nothing was actually printed.
2855*/
2856int PrintAction(struct action *ap, FILE *fp, int indent){
2857 int result = 1;
2858 switch( ap->type ){
2859 case SHIFT:
2860 fprintf(fp,"%*s shift %d",indent,ap->sp->name,ap->x.stp->statenum);
2861 break;
2862 case REDUCE:
2863 fprintf(fp,"%*s reduce %d",indent,ap->sp->name,ap->x.rp->index);
2864 break;
2865 case ACCEPT:
2866 fprintf(fp,"%*s accept",indent,ap->sp->name);
2867 break;
2868 case ERROR:
2869 fprintf(fp,"%*s error",indent,ap->sp->name);
2870 break;
2871 case SRCONFLICT:
2872 case RRCONFLICT:
2873 fprintf(fp,"%*s reduce %-3d ** Parsing conflict **",
2874 indent,ap->sp->name,ap->x.rp->index);
2875 break;
2876 case SSCONFLICT:
2877 fprintf(fp,"%*s shift %-3d ** Parsing conflict **",
2878 indent,ap->sp->name,ap->x.stp->statenum);
2879 break;
2880 case SH_RESOLVED:
2881 if( showPrecedenceConflict ){
2882 fprintf(fp,"%*s shift %-3d -- dropped by precedence",
2883 indent,ap->sp->name,ap->x.stp->statenum);
2884 }else{
2885 result = 0;
2886 }
2887 break;
2888 case RD_RESOLVED:
2889 if( showPrecedenceConflict ){
2890 fprintf(fp,"%*s reduce %-3d -- dropped by precedence",
2891 indent,ap->sp->name,ap->x.rp->index);
2892 }else{
2893 result = 0;
2894 }
2895 break;
2896 case NOT_USED:
2897 result = 0;
2898 break;
2899 }
2900 return result;
2901}
2902
2903/* Generate the "y.output" log file */
2904void ReportOutput(struct lemon *lemp)
2905{
2906 int i;
2907 struct state *stp;
2908 struct config *cfp;
2909 struct action *ap;
2910 FILE *fp;
2911
2912 fp = file_open(lemp,".out","wb");
2913 if( fp==0 ) return;
2914 for(i=0; i<lemp->nstate; i++){
2915 stp = lemp->sorted[i];
2916 fprintf(fp,"State %d:\n",stp->statenum);
2917 if( lemp->basisflag ) cfp=stp->bp;
2918 else cfp=stp->cfp;
2919 while( cfp ){
2920 char buf[20];
2921 if( cfp->dot==cfp->rp->nrhs ){
2922 sprintf(buf,"(%d)",cfp->rp->index);
2923 fprintf(fp," %5s ",buf);
2924 }else{
2925 fprintf(fp," ");
2926 }
2927 ConfigPrint(fp,cfp);
2928 fprintf(fp,"\n");
2929#if 0
2930 SetPrint(fp,cfp->fws,lemp);
2931 PlinkPrint(fp,cfp->fplp,"To ");
2932 PlinkPrint(fp,cfp->bplp,"From");
2933#endif
2934 if( lemp->basisflag ) cfp=cfp->bp;
2935 else cfp=cfp->next;
2936 }
2937 fprintf(fp,"\n");
2938 for(ap=stp->ap; ap; ap=ap->next){
2939 if( PrintAction(ap,fp,30) ) fprintf(fp,"\n");
2940 }
2941 fprintf(fp,"\n");
2942 }
2943 fprintf(fp, "----------------------------------------------------\n");
2944 fprintf(fp, "Symbols:\n");
2945 for(i=0; i<lemp->nsymbol; i++){
2946 int j;
2947 struct symbol *sp;
2948
2949 sp = lemp->symbols[i];
2950 fprintf(fp, " %3d: %s", i, sp->name);
2951 if( sp->type==NONTERMINAL ){
2952 fprintf(fp, ":");
2953 if( sp->lambda ){
2954 fprintf(fp, " <lambda>");
2955 }
2956 for(j=0; j<lemp->nterminal; j++){
2957 if( sp->firstset && SetFind(sp->firstset, j) ){
2958 fprintf(fp, " %s", lemp->symbols[j]->name);
2959 }
2960 }
2961 }
2962 fprintf(fp, "\n");
2963 }
2964 fclose(fp);
2965 return;
2966}
2967
2968/* Search for the file "name" which is in the same directory as
2969** the exacutable */
2970PRIVATE char *pathsearch(char *argv0, char *name, int modemask)
2971{
2972 const char *pathlist;
2973 char *pathbufptr;
2974 char *pathbuf;
2975 char *path,*cp;
2976 char c;
2977
2978#ifdef __WIN32__
2979 cp = strrchr(argv0,'\\');
2980#else
2981 cp = strrchr(argv0,'/');
2982#endif
2983 if( cp ){
2984 c = *cp;
2985 *cp = 0;
2986 path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 );
2987 if( path ) sprintf(path,"%s/%s",argv0,name);
2988 *cp = c;
2989 }else{
2990 pathlist = getenv("PATH");
2991 if( pathlist==0 ) pathlist = ".:/bin:/usr/bin";
2992 pathbuf = (char *) malloc( lemonStrlen(pathlist) + 1 );
2993 path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 );
2994 if( (pathbuf != 0) && (path!=0) ){
2995 pathbufptr = pathbuf;
2996 strcpy(pathbuf, pathlist);
2997 while( *pathbuf ){
2998 cp = strchr(pathbuf,':');
2999 if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)];
3000 c = *cp;
3001 *cp = 0;
3002 sprintf(path,"%s/%s",pathbuf,name);
3003 *cp = c;
3004 if( c==0 ) pathbuf[0] = 0;
3005 else pathbuf = &cp[1];
3006 if( access(path,modemask)==0 ) break;
3007 }
3008 free(pathbufptr);
3009 }
3010 }
3011 return path;
3012}
3013
3014/* Given an action, compute the integer value for that action
3015** which is to be put in the action table of the generated machine.
3016** Return negative if no action should be generated.
3017*/
3018PRIVATE int compute_action(struct lemon *lemp, struct action *ap)
3019{
3020 int act;
3021 switch( ap->type ){
3022 case SHIFT: act = ap->x.stp->statenum; break;
3023 case REDUCE: act = ap->x.rp->index + lemp->nstate; break;
3024 case ERROR: act = lemp->nstate + lemp->nrule; break;
3025 case ACCEPT: act = lemp->nstate + lemp->nrule + 1; break;
3026 default: act = -1; break;
3027 }
3028 return act;
3029}
3030
3031#define LINESIZE 1000
3032/* The next cluster of routines are for reading the template file
3033** and writing the results to the generated parser */
3034/* The first function transfers data from "in" to "out" until
3035** a line is seen which begins with "%%". The line number is
3036** tracked.
3037**
3038** if name!=0, then any word that begin with "Parse" is changed to
3039** begin with *name instead.
3040*/
3041PRIVATE void tplt_xfer(char *name, FILE *in, FILE *out, int *lineno)
3042{
3043 int i, iStart;
3044 char line[LINESIZE];
3045 while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){
3046 (*lineno)++;
3047 iStart = 0;
3048 if( name ){
3049 for(i=0; line[i]; i++){
3050 if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0
3051 && (i==0 || !isalpha(line[i-1]))
3052 ){
3053 if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]);
3054 fprintf(out,"%s",name);
3055 i += 4;
3056 iStart = i+1;
3057 }
3058 }
3059 }
3060 fprintf(out,"%s",&line[iStart]);
3061 }
3062}
3063
3064/* The next function finds the template file and opens it, returning
3065** a pointer to the opened file. */
3066PRIVATE FILE *tplt_open(struct lemon *lemp)
3067{
3068 static char templatename[] = "lempar.c";
3069 char buf[1000];
3070 FILE *in;
3071 char *tpltname;
3072 char *cp;
3073
3074 /* first, see if user specified a template filename on the command line. */
3075 if (user_templatename != 0) {
3076 if( access(user_templatename,004)==-1 ){
3077 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3078 user_templatename);
3079 lemp->errorcnt++;
3080 return 0;
3081 }
3082 in = fopen(user_templatename,"rb");
3083 if( in==0 ){
3084 fprintf(stderr,"Can't open the template file \"%s\".\n",user_templatename);
3085 lemp->errorcnt++;
3086 return 0;
3087 }
3088 return in;
3089 }
3090
3091 cp = strrchr(lemp->filename,'.');
3092 if( cp ){
3093 sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename);
3094 }else{
3095 sprintf(buf,"%s.lt",lemp->filename);
3096 }
3097 if( access(buf,004)==0 ){
3098 tpltname = buf;
3099 }else if( access(templatename,004)==0 ){
3100 tpltname = templatename;
3101 }else{
3102 tpltname = pathsearch(lemp->argv0,templatename,0);
3103 }
3104 if( tpltname==0 ){
3105 fprintf(stderr,"Can't find the parser driver template file \"%s\".\n",
3106 templatename);
3107 lemp->errorcnt++;
3108 return 0;
3109 }
3110 in = fopen(tpltname,"rb");
3111 if( in==0 ){
3112 fprintf(stderr,"Can't open the template file \"%s\".\n",templatename);
3113 lemp->errorcnt++;
3114 return 0;
3115 }
3116 return in;
3117}
3118
3119/* Print a #line directive line to the output file. */
3120PRIVATE void tplt_linedir(FILE *out, int lineno, char *filename)
3121{
3122 fprintf(out,"#line %d \"",lineno);
3123 while( *filename ){
3124 if( *filename == '\\' ) putc('\\',out);
3125 putc(*filename,out);
3126 filename++;
3127 }
3128 fprintf(out,"\"\n");
3129}
3130
3131/* Print a string to the file and keep the linenumber up to date */
3132PRIVATE void tplt_print(FILE *out, struct lemon *lemp, char *str, int *lineno)
3133{
3134 if( str==0 ) return;
3135 while( *str ){
3136 putc(*str,out);
3137 if( *str=='\n' ) (*lineno)++;
3138 str++;
3139 }
3140 if( str[-1]!='\n' ){
3141 putc('\n',out);
3142 (*lineno)++;
3143 }
3144 if (!lemp->nolinenosflag) {
3145 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3146 }
3147 return;
3148}
3149
3150/*
3151** The following routine emits code for the destructor for the
3152** symbol sp
3153*/
3154void emit_destructor_code(
3155 FILE *out,
3156 struct symbol *sp,
3157 struct lemon *lemp,
3158 int *lineno
3159){
3160 char *cp = 0;
3161
3162 if( sp->type==TERMINAL ){
3163 cp = lemp->tokendest;
3164 if( cp==0 ) return;
3165 fprintf(out,"{\n"); (*lineno)++;
3166 }else if( sp->destructor ){
3167 cp = sp->destructor;
3168 fprintf(out,"{\n"); (*lineno)++;
3169 if (!lemp->nolinenosflag) { (*lineno)++; tplt_linedir(out,sp->destLineno,lemp->filename); }
3170 }else if( lemp->vardest ){
3171 cp = lemp->vardest;
3172 if( cp==0 ) return;
3173 fprintf(out,"{\n"); (*lineno)++;
3174 }else{
3175 assert( 0 ); /* Cannot happen */
3176 }
3177 for(; *cp; cp++){
3178 if( *cp=='$' && cp[1]=='$' ){
3179 fprintf(out,"(yypminor->yy%d)",sp->dtnum);
3180 cp++;
3181 continue;
3182 }
3183 if( *cp=='\n' ) (*lineno)++;
3184 fputc(*cp,out);
3185 }
3186 fprintf(out,"\n"); (*lineno)++;
3187 if (!lemp->nolinenosflag) {
3188 (*lineno)++; tplt_linedir(out,*lineno,lemp->outname);
3189 }
3190 fprintf(out,"}\n"); (*lineno)++;
3191 return;
3192}
3193
3194/*
3195** Return TRUE (non-zero) if the given symbol has a destructor.
3196*/
3197int has_destructor(struct symbol *sp, struct lemon *lemp)
3198{
3199 int ret;
3200 if( sp->type==TERMINAL ){
3201 ret = lemp->tokendest!=0;
3202 }else{
3203 ret = lemp->vardest!=0 || sp->destructor!=0;
3204 }
3205 return ret;
3206}
3207
3208/*
3209** Append text to a dynamically allocated string. If zText is 0 then
3210** reset the string to be empty again. Always return the complete text
3211** of the string (which is overwritten with each call).
3212**
3213** n bytes of zText are stored. If n==0 then all of zText up to the first
3214** \000 terminator is stored. zText can contain up to two instances of
3215** %d. The values of p1 and p2 are written into the first and second
3216** %d.
3217**
3218** If n==-1, then the previous character is overwritten.
3219*/
3220PRIVATE char *append_str(const char *zText, int n, int p1, int p2){
3221 static char empty[1] = { 0 };
3222 static char *z = 0;
3223 static int alloced = 0;
3224 static int used = 0;
3225 int c;
3226 char zInt[40];
3227 if( zText==0 ){
3228 used = 0;
3229 return z;
3230 }
3231 if( n<=0 ){
3232 if( n<0 ){
3233 used += n;
3234 assert( used>=0 );
3235 }
3236 n = lemonStrlen(zText);
3237 }
3238 if( (int) (n+sizeof(zInt)*2+used) >= alloced ){
3239 alloced = n + sizeof(zInt)*2 + used + 200;
3240 z = (char *) realloc(z, alloced);
3241 }
3242 if( z==0 ) return empty;
3243 while( n-- > 0 ){
3244 c = *(zText++);
3245 if( c=='%' && n>0 && zText[0]=='d' ){
3246 sprintf(zInt, "%d", p1);
3247 p1 = p2;
3248 strcpy(&z[used], zInt);
3249 used += lemonStrlen(&z[used]);
3250 zText++;
3251 n--;
3252 }else{
3253 z[used++] = c;
3254 }
3255 }
3256 z[used] = 0;
3257 return z;
3258}
3259
3260/*
3261** zCode is a string that is the action associated with a rule. Expand
3262** the symbols in this string so that the refer to elements of the parser
3263** stack.
3264*/
3265PRIVATE void translate_code(struct lemon *lemp, struct rule *rp){
3266 char *cp, *xp;
3267 int i;
3268 char lhsused = 0; /* True if the LHS element has been used */
3269 char used[MAXRHS]; /* True for each RHS element which is used */
3270
3271 for(i=0; i<rp->nrhs; i++) used[i] = 0;
3272 lhsused = 0;
3273
3274 if( rp->code==0 ){
3275 static char newlinestr[2] = { '\n', '\0' };
3276 rp->code = newlinestr;
3277 rp->line = rp->ruleline;
3278 }
3279
3280 append_str(0,0,0,0);
3281
3282 /* This const cast is wrong but harmless, if we're careful. */
3283 for(cp=(char *)rp->code; *cp; cp++){
3284 if( isalpha(*cp) && (cp==rp->code || (!isalnum(cp[-1]) && cp[-1]!='_')) ){
3285 char saved;
3286 for(xp= &cp[1]; isalnum(*xp) || *xp=='_'; xp++);
3287 saved = *xp;
3288 *xp = 0;
3289 if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){
3290 append_str("yygotominor.yy%d",0,rp->lhs->dtnum,0);
3291 cp = xp;
3292 lhsused = 1;
3293 }else{
3294 for(i=0; i<rp->nrhs; i++){
3295 if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
3296 if( cp!=rp->code && cp[-1]=='@' ){
3297 /* If the argument is of the form @X then substituted
3298 ** the token number of X, not the value of X */
3299 append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
3300 }else{
3301 struct symbol *sp = rp->rhs[i];
3302 int dtnum;
3303 if( sp->type==MULTITERMINAL ){
3304 dtnum = sp->subsym[0]->dtnum;
3305 }else{
3306 dtnum = sp->dtnum;
3307 }
3308 append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);
3309 }
3310 cp = xp;
3311 used[i] = 1;
3312 break;
3313 }
3314 }
3315 }
3316 *xp = saved;
3317 }
3318 append_str(cp, 1, 0, 0);
3319 } /* End loop */
3320
3321 /* Check to make sure the LHS has been used */
3322 if( rp->lhsalias && !lhsused ){
3323 ErrorMsg(lemp->filename,rp->ruleline,
3324 "Label \"%s\" for \"%s(%s)\" is never used.",
3325 rp->lhsalias,rp->lhs->name,rp->lhsalias);
3326 lemp->errorcnt++;
3327 }
3328
3329 /* Generate destructor code for RHS symbols which are not used in the
3330 ** reduce code */
3331 for(i=0; i<rp->nrhs; i++){
3332 if( rp->rhsalias[i] && !used[i] ){
3333 ErrorMsg(lemp->filename,rp->ruleline,
3334 "Label %s for \"%s(%s)\" is never used.",
3335 rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]);
3336 lemp->errorcnt++;
3337 }else if( rp->rhsalias[i]==0 ){
3338 if( has_destructor(rp->rhs[i],lemp) ){
3339 append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0,
3340 rp->rhs[i]->index,i-rp->nrhs+1);
3341 }else{
3342 /* No destructor defined for this term */
3343 }
3344 }
3345 }
3346 if( rp->code ){
3347 cp = append_str(0,0,0,0);
3348 rp->code = Strsafe(cp?cp:"");
3349 }
3350}
3351
3352/*
3353** Generate code which executes when the rule "rp" is reduced. Write
3354** the code to "out". Make sure lineno stays up-to-date.
3355*/
3356PRIVATE void emit_code(
3357 FILE *out,
3358 struct rule *rp,
3359 struct lemon *lemp,
3360 int *lineno
3361){
3362 const char *cp;
3363
3364 /* Generate code to do the reduce action */
3365 if( rp->code ){
3366 if (!lemp->nolinenosflag) { (*lineno)++; tplt_linedir(out,rp->line,lemp->filename); }
3367 fprintf(out,"{%s",rp->code);
3368 for(cp=rp->code; *cp; cp++){
3369 if( *cp=='\n' ) (*lineno)++;
3370 } /* End loop */
3371 fprintf(out,"}\n"); (*lineno)++;
3372 if (!lemp->nolinenosflag) { (*lineno)++; tplt_linedir(out,*lineno,lemp->outname); }
3373 } /* End if( rp->code ) */
3374
3375 return;
3376}
3377
3378/*
3379** Print the definition of the union used for the parser's data stack.
3380** This union contains fields for every possible data type for tokens
3381** and nonterminals. In the process of computing and printing this
3382** union, also set the ".dtnum" field of every terminal and nonterminal
3383** symbol.
3384*/
3385void print_stack_union(
3386 FILE *out, /* The output stream */
3387 struct lemon *lemp, /* The main info structure for this parser */
3388 int *plineno, /* Pointer to the line number */
3389 int mhflag /* True if generating makeheaders output */
3390){
3391 int lineno = *plineno; /* The line number of the output */
3392 char **types; /* A hash table of datatypes */
3393 int arraysize; /* Size of the "types" array */
3394 int maxdtlength; /* Maximum length of any ".datatype" field. */
3395 char *stddt; /* Standardized name for a datatype */
3396 int i,j; /* Loop counters */
3397 int hash; /* For hashing the name of a type */
3398 const char *name; /* Name of the parser */
3399
3400 /* Allocate and initialize types[] and allocate stddt[] */
3401 arraysize = lemp->nsymbol * 2;
3402 types = (char**)calloc( arraysize, sizeof(char*) );
3403 if( types==0 ){
3404 fprintf(stderr,"Out of memory.\n");
3405 exit(1);
3406 }
3407 for(i=0; i<arraysize; i++) types[i] = 0;
3408 maxdtlength = 0;
3409 if( lemp->vartype ){
3410 maxdtlength = lemonStrlen(lemp->vartype);
3411 }
3412 for(i=0; i<lemp->nsymbol; i++){
3413 int len;
3414 struct symbol *sp = lemp->symbols[i];
3415 if( sp->datatype==0 ) continue;
3416 len = lemonStrlen(sp->datatype);
3417 if( len>maxdtlength ) maxdtlength = len;
3418 }
3419 stddt = (char*)malloc( maxdtlength*2 + 1 );
3420 if( stddt==0 ){
3421 fprintf(stderr,"Out of memory.\n");
3422 exit(1);
3423 }
3424
3425 /* Build a hash table of datatypes. The ".dtnum" field of each symbol
3426 ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is
3427 ** used for terminal symbols. If there is no %default_type defined then
3428 ** 0 is also used as the .dtnum value for nonterminals which do not specify
3429 ** a datatype using the %type directive.
3430 */
3431 for(i=0; i<lemp->nsymbol; i++){
3432 struct symbol *sp = lemp->symbols[i];
3433 char *cp;
3434 if( sp==lemp->errsym ){
3435 sp->dtnum = arraysize+1;
3436 continue;
3437 }
3438 if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){
3439 sp->dtnum = 0;
3440 continue;
3441 }
3442 cp = sp->datatype;
3443 if( cp==0 ) cp = lemp->vartype;
3444 j = 0;
3445 while( isspace(*cp) ) cp++;
3446 while( *cp ) stddt[j++] = *cp++;
3447 while( j>0 && isspace(stddt[j-1]) ) j--;
3448 stddt[j] = 0;
3449 if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){
3450 sp->dtnum = 0;
3451 continue;
3452 }
3453 hash = 0;
3454 for(j=0; stddt[j]; j++){
3455 hash = hash*53 + stddt[j];
3456 }
3457 hash = (hash & 0x7fffffff)%arraysize;
3458 while( types[hash] ){
3459 if( strcmp(types[hash],stddt)==0 ){
3460 sp->dtnum = hash + 1;
3461 break;
3462 }
3463 hash++;
3464 if( hash>=arraysize ) hash = 0;
3465 }
3466 if( types[hash]==0 ){
3467 sp->dtnum = hash + 1;
3468 types[hash] = (char*)malloc( lemonStrlen(stddt)+1 );
3469 if( types[hash]==0 ){
3470 fprintf(stderr,"Out of memory.\n");
3471 exit(1);
3472 }
3473 strcpy(types[hash],stddt);
3474 }
3475 }
3476
3477 /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */
3478 name = lemp->name ? lemp->name : "Parse";
3479 lineno = *plineno;
3480 if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; }
3481 fprintf(out,"#define %sTOKENTYPE %s\n",name,
3482 lemp->tokentype?lemp->tokentype:"void*"); lineno++;
3483 if( mhflag ){ fprintf(out,"#endif\n"); lineno++; }
3484 fprintf(out,"typedef union {\n"); lineno++;
3485 fprintf(out," int yyinit;\n"); lineno++;
3486 fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++;
3487 for(i=0; i<arraysize; i++){
3488 if( types[i]==0 ) continue;
3489 fprintf(out," %s yy%d;\n",types[i],i+1); lineno++;
3490 free(types[i]);
3491 }
3492 if( lemp->errsym->useCnt ){
3493 fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++;
3494 }
3495 free(stddt);
3496 free(types);
3497 fprintf(out,"} YYMINORTYPE;\n"); lineno++;
3498 *plineno = lineno;
3499}
3500
3501/*
3502** Return the name of a C datatype able to represent values between
3503** lwr and upr, inclusive.
3504*/
3505static const char *minimum_size_type(int lwr, int upr){
3506 if( lwr>=0 ){
3507 if( upr<=255 ){
3508 return "unsigned char";
3509 }else if( upr<65535 ){
3510 return "unsigned short int";
3511 }else{
3512 return "unsigned int";
3513 }
3514 }else if( lwr>=-127 && upr<=127 ){
3515 return "signed char";
3516 }else if( lwr>=-32767 && upr<32767 ){
3517 return "short";
3518 }else{
3519 return "int";
3520 }
3521}
3522
3523/*
3524** Each state contains a set of token transaction and a set of
3525** nonterminal transactions. Each of these sets makes an instance
3526** of the following structure. An array of these structures is used
3527** to order the creation of entries in the yy_action[] table.
3528*/
3529struct axset {
3530 struct state *stp; /* A pointer to a state */
3531 int isTkn; /* True to use tokens. False for non-terminals */
3532 int nAction; /* Number of actions */
3533 int iOrder; /* Original order of action sets */
3534};
3535
3536/*
3537** Compare to axset structures for sorting purposes
3538*/
3539static int axset_compare(const void *a, const void *b){
3540 struct axset *p1 = (struct axset*)a;
3541 struct axset *p2 = (struct axset*)b;
3542 int c;
3543 c = p2->nAction - p1->nAction;
3544 if( c==0 ){
3545 c = p2->iOrder - p1->iOrder;
3546 }
3547 assert( c!=0 || p1==p2 );
3548 return c;
3549}
3550
3551/*
3552** Write text on "out" that describes the rule "rp".
3553*/
3554static void writeRuleText(FILE *out, struct rule *rp){
3555 int j;
3556 fprintf(out,"%s ::=", rp->lhs->name);
3557 for(j=0; j<rp->nrhs; j++){
3558 struct symbol *sp = rp->rhs[j];
3559 fprintf(out," %s", sp->name);
3560 if( sp->type==MULTITERMINAL ){
3561 int k;
3562 for(k=1; k<sp->nsubsym; k++){
3563 fprintf(out,"|%s",sp->subsym[k]->name);
3564 }
3565 }
3566 }
3567}
3568
3569
3570/* Generate C source code for the parser */
3571void ReportTable(
3572 struct lemon *lemp,
3573 int mhflag /* Output in makeheaders format if true */
3574){
3575 FILE *out, *in;
3576 char line[LINESIZE];
3577 int lineno;
3578 struct state *stp;
3579 struct action *ap;
3580 struct rule *rp;
3581 struct acttab *pActtab;
3582 int i, j, n;
3583 const char *name;
3584 int mnTknOfst, mxTknOfst;
3585 int mnNtOfst, mxNtOfst;
3586 struct axset *ax;
3587
3588 in = tplt_open(lemp);
3589 if( in==0 ) return;
3590 out = file_open(lemp,".c","wb");
3591 if( out==0 ){
3592 fclose(in);
3593 return;
3594 }
3595 lineno = 1;
3596 tplt_xfer(lemp->name,in,out,&lineno);
3597
3598 /* Generate the include code, if any */
3599 tplt_print(out,lemp,lemp->include,&lineno);
3600 if( mhflag ){
3601 char *name = file_makename(lemp, ".h");
3602 fprintf(out,"#include \"%s\"\n", name); lineno++;
3603 free(name);
3604 }
3605 tplt_xfer(lemp->name,in,out,&lineno);
3606
3607 /* Generate #defines for all tokens */
3608 if( mhflag ){
3609 const char *prefix;
3610 fprintf(out,"#if INTERFACE\n"); lineno++;
3611 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
3612 else prefix = "";
3613 for(i=1; i<lemp->nterminal; i++){
3614 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
3615 lineno++;
3616 }
3617 fprintf(out,"#endif\n"); lineno++;
3618 }
3619 tplt_xfer(lemp->name,in,out,&lineno);
3620
3621 /* Generate the defines */
3622 fprintf(out,"#define YYCODETYPE %s\n",
3623 minimum_size_type(0, lemp->nsymbol+1)); lineno++;
3624 fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol+1); lineno++;
3625 fprintf(out,"#define YYACTIONTYPE %s\n",
3626 minimum_size_type(0, lemp->nstate+lemp->nrule+5)); lineno++;
3627 if( lemp->wildcard ){
3628 fprintf(out,"#define YYWILDCARD %d\n",
3629 lemp->wildcard->index); lineno++;
3630 }
3631 print_stack_union(out,lemp,&lineno,mhflag);
3632 fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++;
3633 if( lemp->stacksize ){
3634 fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++;
3635 }else{
3636 fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++;
3637 }
3638 fprintf(out, "#endif\n"); lineno++;
3639 if( mhflag ){
3640 fprintf(out,"#if INTERFACE\n"); lineno++;
3641 }
3642 name = lemp->name ? lemp->name : "Parse";
3643 if( lemp->arg && lemp->arg[0] ){
3644 int i;
3645 i = lemonStrlen(lemp->arg);
3646 while( i>=1 && isspace(lemp->arg[i-1]) ) i--;
3647 while( i>=1 && (isalnum(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--;
3648 fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++;
3649 fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++;
3650 fprintf(out,"#define %sARG_FETCH %s = yypParser->%s\n",
3651 name,lemp->arg,&lemp->arg[i]); lineno++;
3652 fprintf(out,"#define %sARG_STORE yypParser->%s = %s\n",
3653 name,&lemp->arg[i],&lemp->arg[i]); lineno++;
3654 }else{
3655 fprintf(out,"#define %sARG_SDECL\n",name); lineno++;
3656 fprintf(out,"#define %sARG_PDECL\n",name); lineno++;
3657 fprintf(out,"#define %sARG_FETCH\n",name); lineno++;
3658 fprintf(out,"#define %sARG_STORE\n",name); lineno++;
3659 }
3660 if( mhflag ){
3661 fprintf(out,"#endif\n"); lineno++;
3662 }
3663 fprintf(out,"#define YYNSTATE %d\n",lemp->nstate); lineno++;
3664 fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++;
3665 if( lemp->errsym->useCnt ){
3666 fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++;
3667 fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++;
3668 }
3669 if( lemp->has_fallback ){
3670 fprintf(out,"#define YYFALLBACK 1\n"); lineno++;
3671 }
3672 tplt_xfer(lemp->name,in,out,&lineno);
3673
3674 /* Generate the action table and its associates:
3675 **
3676 ** yy_action[] A single table containing all actions.
3677 ** yy_lookahead[] A table containing the lookahead for each entry in
3678 ** yy_action. Used to detect hash collisions.
3679 ** yy_shift_ofst[] For each state, the offset into yy_action for
3680 ** shifting terminals.
3681 ** yy_reduce_ofst[] For each state, the offset into yy_action for
3682 ** shifting non-terminals after a reduce.
3683 ** yy_default[] Default action for each state.
3684 */
3685
3686 /* Compute the actions on all states and count them up */
3687 ax = (struct axset *) calloc(lemp->nstate*2, sizeof(ax[0]));
3688 if( ax==0 ){
3689 fprintf(stderr,"malloc failed\n");
3690 exit(1);
3691 }
3692 for(i=0; i<lemp->nstate; i++){
3693 stp = lemp->sorted[i];
3694 ax[i*2].stp = stp;
3695 ax[i*2].isTkn = 1;
3696 ax[i*2].nAction = stp->nTknAct;
3697 ax[i*2+1].stp = stp;
3698 ax[i*2+1].isTkn = 0;
3699 ax[i*2+1].nAction = stp->nNtAct;
3700 }
3701 mxTknOfst = mnTknOfst = 0;
3702 mxNtOfst = mnNtOfst = 0;
3703
3704 /* Compute the action table. In order to try to keep the size of the
3705 ** action table to a minimum, the heuristic of placing the largest action
3706 ** sets first is used.
3707 */
3708 for(i=0; i<lemp->nstate*2; i++) ax[i].iOrder = i;
3709 qsort(ax, lemp->nstate*2, sizeof(ax[0]), axset_compare);
3710 pActtab = acttab_alloc();
3711 for(i=0; i<lemp->nstate*2 && ax[i].nAction>0; i++){
3712 stp = ax[i].stp;
3713 if( ax[i].isTkn ){
3714 for(ap=stp->ap; ap; ap=ap->next){
3715 int action;
3716 if( ap->sp->index>=lemp->nterminal ) continue;
3717 action = compute_action(lemp, ap);
3718 if( action<0 ) continue;
3719 acttab_action(pActtab, ap->sp->index, action);
3720 }
3721 stp->iTknOfst = acttab_insert(pActtab);
3722 if( stp->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst;
3723 if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst;
3724 }else{
3725 for(ap=stp->ap; ap; ap=ap->next){
3726 int action;
3727 if( ap->sp->index<lemp->nterminal ) continue;
3728 if( ap->sp->index==lemp->nsymbol ) continue;
3729 action = compute_action(lemp, ap);
3730 if( action<0 ) continue;
3731 acttab_action(pActtab, ap->sp->index, action);
3732 }
3733 stp->iNtOfst = acttab_insert(pActtab);
3734 if( stp->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst;
3735 if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst;
3736 }
3737 }
3738 free(ax);
3739
3740 /* Output the yy_action table */
3741 n = acttab_size(pActtab);
3742 fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++;
3743 fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++;
3744 for(i=j=0; i<n; i++){
3745 int action = acttab_yyaction(pActtab, i);
3746 if( action<0 ) action = lemp->nstate + lemp->nrule + 2;
3747 if( j==0 ) fprintf(out," /* %5d */ ", i);
3748 fprintf(out, " %4d,", action);
3749 if( j==9 || i==n-1 ){
3750 fprintf(out, "\n"); lineno++;
3751 j = 0;
3752 }else{
3753 j++;
3754 }
3755 }
3756 fprintf(out, "};\n"); lineno++;
3757
3758 /* Output the yy_lookahead table */
3759 fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++;
3760 for(i=j=0; i<n; i++){
3761 int la = acttab_yylookahead(pActtab, i);
3762 if( la<0 ) la = lemp->nsymbol;
3763 if( j==0 ) fprintf(out," /* %5d */ ", i);
3764 fprintf(out, " %4d,", la);
3765 if( j==9 || i==n-1 ){
3766 fprintf(out, "\n"); lineno++;
3767 j = 0;
3768 }else{
3769 j++;
3770 }
3771 }
3772 fprintf(out, "};\n"); lineno++;
3773
3774 /* Output the yy_shift_ofst[] table */
3775 fprintf(out, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst-1); lineno++;
3776 n = lemp->nstate;
3777 while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--;
3778 fprintf(out, "#define YY_SHIFT_COUNT (%d)\n", n-1); lineno++;
3779 fprintf(out, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst); lineno++;
3780 fprintf(out, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst); lineno++;
3781 fprintf(out, "static const %s yy_shift_ofst[] = {\n",
3782 minimum_size_type(mnTknOfst-1, mxTknOfst)); lineno++;
3783 for(i=j=0; i<n; i++){
3784 int ofst;
3785 stp = lemp->sorted[i];
3786 ofst = stp->iTknOfst;
3787 if( ofst==NO_OFFSET ) ofst = mnTknOfst - 1;
3788 if( j==0 ) fprintf(out," /* %5d */ ", i);
3789 fprintf(out, " %4d,", ofst);
3790 if( j==9 || i==n-1 ){
3791 fprintf(out, "\n"); lineno++;
3792 j = 0;
3793 }else{
3794 j++;
3795 }
3796 }
3797 fprintf(out, "};\n"); lineno++;
3798
3799 /* Output the yy_reduce_ofst[] table */
3800 fprintf(out, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst-1); lineno++;
3801 n = lemp->nstate;
3802 while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--;
3803 fprintf(out, "#define YY_REDUCE_COUNT (%d)\n", n-1); lineno++;
3804 fprintf(out, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst); lineno++;
3805 fprintf(out, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst); lineno++;
3806 fprintf(out, "static const %s yy_reduce_ofst[] = {\n",
3807 minimum_size_type(mnNtOfst-1, mxNtOfst)); lineno++;
3808 for(i=j=0; i<n; i++){
3809 int ofst;
3810 stp = lemp->sorted[i];
3811 ofst = stp->iNtOfst;
3812 if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1;
3813 if( j==0 ) fprintf(out," /* %5d */ ", i);
3814 fprintf(out, " %4d,", ofst);
3815 if( j==9 || i==n-1 ){
3816 fprintf(out, "\n"); lineno++;
3817 j = 0;
3818 }else{
3819 j++;
3820 }
3821 }
3822 fprintf(out, "};\n"); lineno++;
3823
3824 /* Output the default action table */
3825 fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++;
3826 n = lemp->nstate;
3827 for(i=j=0; i<n; i++){
3828 stp = lemp->sorted[i];
3829 if( j==0 ) fprintf(out," /* %5d */ ", i);
3830 fprintf(out, " %4d,", stp->iDflt);
3831 if( j==9 || i==n-1 ){
3832 fprintf(out, "\n"); lineno++;
3833 j = 0;
3834 }else{
3835 j++;
3836 }
3837 }
3838 fprintf(out, "};\n"); lineno++;
3839 tplt_xfer(lemp->name,in,out,&lineno);
3840
3841 /* Generate the table of fallback tokens.
3842 */
3843 if( lemp->has_fallback ){
3844 int mx = lemp->nterminal - 1;
3845 while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; }
3846 for(i=0; i<=mx; i++){
3847 struct symbol *p = lemp->symbols[i];
3848 if( p->fallback==0 ){
3849 fprintf(out, " 0, /* %10s => nothing */\n", p->name);
3850 }else{
3851 fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index,
3852 p->name, p->fallback->name);
3853 }
3854 lineno++;
3855 }
3856 }
3857 tplt_xfer(lemp->name, in, out, &lineno);
3858
3859 /* Generate a table containing the symbolic name of every symbol
3860 */
3861 for(i=0; i<lemp->nsymbol; i++){
3862 sprintf(line,"\"%s\",",lemp->symbols[i]->name);
3863 fprintf(out," %-15s",line);
3864 if( (i&3)==3 ){ fprintf(out,"\n"); lineno++; }
3865 }
3866 if( (i&3)!=0 ){ fprintf(out,"\n"); lineno++; }
3867 tplt_xfer(lemp->name,in,out,&lineno);
3868
3869 /* Generate a table containing a text string that describes every
3870 ** rule in the rule set of the grammar. This information is used
3871 ** when tracing REDUCE actions.
3872 */
3873 for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
3874 assert( rp->index==i );
3875 fprintf(out," /* %3d */ \"", i);
3876 writeRuleText(out, rp);
3877 fprintf(out,"\",\n"); lineno++;
3878 }
3879 tplt_xfer(lemp->name,in,out,&lineno);
3880
3881 /* Generate code which executes every time a symbol is popped from
3882 ** the stack while processing errors or while destroying the parser.
3883 ** (In other words, generate the %destructor actions)
3884 */
3885 if( lemp->tokendest ){
3886 int once = 1;
3887 for(i=0; i<lemp->nsymbol; i++){
3888 struct symbol *sp = lemp->symbols[i];
3889 if( sp==0 || sp->type!=TERMINAL ) continue;
3890 if( once ){
3891 fprintf(out, " /* TERMINAL Destructor */\n"); lineno++;
3892 once = 0;
3893 }
3894 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
3895 }
3896 for(i=0; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++);
3897 if( i<lemp->nsymbol ){
3898 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
3899 fprintf(out," break;\n"); lineno++;
3900 }
3901 }
3902 if( lemp->vardest ){
3903 struct symbol *dflt_sp = 0;
3904 int once = 1;
3905 for(i=0; i<lemp->nsymbol; i++){
3906 struct symbol *sp = lemp->symbols[i];
3907 if( sp==0 || sp->type==TERMINAL ||
3908 sp->index<=0 || sp->destructor!=0 ) continue;
3909 if( once ){
3910 fprintf(out, " /* Default NON-TERMINAL Destructor */\n"); lineno++;
3911 once = 0;
3912 }
3913 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
3914 dflt_sp = sp;
3915 }
3916 if( dflt_sp!=0 ){
3917 emit_destructor_code(out,dflt_sp,lemp,&lineno);
3918 }
3919 fprintf(out," break;\n"); lineno++;
3920 }
3921 for(i=0; i<lemp->nsymbol; i++){
3922 struct symbol *sp = lemp->symbols[i];
3923 if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue;
3924 fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++;
3925
3926 /* Combine duplicate destructors into a single case */
3927 for(j=i+1; j<lemp->nsymbol; j++){
3928 struct symbol *sp2 = lemp->symbols[j];
3929 if( sp2 && sp2->type!=TERMINAL && sp2->destructor
3930 && sp2->dtnum==sp->dtnum
3931 && strcmp(sp->destructor,sp2->destructor)==0 ){
3932 fprintf(out," case %d: /* %s */\n",
3933 sp2->index, sp2->name); lineno++;
3934 sp2->destructor = 0;
3935 }
3936 }
3937
3938 emit_destructor_code(out,lemp->symbols[i],lemp,&lineno);
3939 fprintf(out," break;\n"); lineno++;
3940 }
3941 tplt_xfer(lemp->name,in,out,&lineno);
3942
3943 /* Generate code which executes whenever the parser stack overflows */
3944 tplt_print(out,lemp,lemp->overflow,&lineno);
3945 tplt_xfer(lemp->name,in,out,&lineno);
3946
3947 /* Generate the table of rule information
3948 **
3949 ** Note: This code depends on the fact that rules are number
3950 ** sequentually beginning with 0.
3951 */
3952 for(rp=lemp->rule; rp; rp=rp->next){
3953 fprintf(out," { %d, %d },\n",rp->lhs->index,rp->nrhs); lineno++;
3954 }
3955 tplt_xfer(lemp->name,in,out,&lineno);
3956
3957 /* Generate code which execution during each REDUCE action */
3958 for(rp=lemp->rule; rp; rp=rp->next){
3959 translate_code(lemp, rp);
3960 }
3961 /* First output rules other than the default: rule */
3962 for(rp=lemp->rule; rp; rp=rp->next){
3963 struct rule *rp2; /* Other rules with the same action */
3964 if( rp->code==0 ) continue;
3965 if( rp->code[0]=='\n' && rp->code[1]==0 ) continue; /* Will be default: */
3966 fprintf(out," case %d: /* ", rp->index);
3967 writeRuleText(out, rp);
3968 fprintf(out, " */\n"); lineno++;
3969 for(rp2=rp->next; rp2; rp2=rp2->next){
3970 if( rp2->code==rp->code ){
3971 fprintf(out," case %d: /* ", rp2->index);
3972 writeRuleText(out, rp2);
3973 fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->index); lineno++;
3974 rp2->code = 0;
3975 }
3976 }
3977 emit_code(out,rp,lemp,&lineno);
3978 fprintf(out," break;\n"); lineno++;
3979 rp->code = 0;
3980 }
3981 /* Finally, output the default: rule. We choose as the default: all
3982 ** empty actions. */
3983 fprintf(out," default:\n"); lineno++;
3984 for(rp=lemp->rule; rp; rp=rp->next){
3985 if( rp->code==0 ) continue;
3986 assert( rp->code[0]=='\n' && rp->code[1]==0 );
3987 fprintf(out," /* (%d) ", rp->index);
3988 writeRuleText(out, rp);
3989 fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->index); lineno++;
3990 }
3991 fprintf(out," break;\n"); lineno++;
3992 tplt_xfer(lemp->name,in,out,&lineno);
3993
3994 /* Generate code which executes if a parse fails */
3995 tplt_print(out,lemp,lemp->failure,&lineno);
3996 tplt_xfer(lemp->name,in,out,&lineno);
3997
3998 /* Generate code which executes when a syntax error occurs */
3999 tplt_print(out,lemp,lemp->error,&lineno);
4000 tplt_xfer(lemp->name,in,out,&lineno);
4001
4002 /* Generate code which executes when the parser accepts its input */
4003 tplt_print(out,lemp,lemp->accept,&lineno);
4004 tplt_xfer(lemp->name,in,out,&lineno);
4005
4006 /* Append any addition code the user desires */
4007 tplt_print(out,lemp,lemp->extracode,&lineno);
4008
4009 fclose(in);
4010 fclose(out);
4011 return;
4012}
4013
4014/* Generate a header file for the parser */
4015void ReportHeader(struct lemon *lemp)
4016{
4017 FILE *out, *in;
4018 const char *prefix;
4019 char line[LINESIZE];
4020 char pattern[LINESIZE];
4021 int i;
4022
4023 if( lemp->tokenprefix ) prefix = lemp->tokenprefix;
4024 else prefix = "";
4025 in = file_open(lemp,".h","rb");
4026 if( in ){
4027 for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){
4028 sprintf(pattern,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
4029 if( strcmp(line,pattern) ) break;
4030 }
4031 fclose(in);
4032 if( i==lemp->nterminal ){
4033 /* No change in the file. Don't rewrite it. */
4034 return;
4035 }
4036 }
4037 out = file_open(lemp,".h","wb");
4038 if( out ){
4039 for(i=1; i<lemp->nterminal; i++){
4040 fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i);
4041 }
4042 fclose(out);
4043 }
4044 return;
4045}
4046
4047/* Reduce the size of the action tables, if possible, by making use
4048** of defaults.
4049**
4050** In this version, we take the most frequent REDUCE action and make
4051** it the default. Except, there is no default if the wildcard token
4052** is a possible look-ahead.
4053*/
4054void CompressTables(struct lemon *lemp)
4055{
4056 struct state *stp;
4057 struct action *ap, *ap2;
4058 struct rule *rp, *rp2, *rbest;
4059 int nbest, n;
4060 int i;
4061 int usesWildcard;
4062
4063 for(i=0; i<lemp->nstate; i++){
4064 stp = lemp->sorted[i];
4065 nbest = 0;
4066 rbest = 0;
4067 usesWildcard = 0;
4068
4069 for(ap=stp->ap; ap; ap=ap->next){
4070 if( ap->type==SHIFT && ap->sp==lemp->wildcard ){
4071 usesWildcard = 1;
4072 }
4073 if( ap->type!=REDUCE ) continue;
4074 rp = ap->x.rp;
4075 if( rp->lhsStart ) continue;
4076 if( rp==rbest ) continue;
4077 n = 1;
4078 for(ap2=ap->next; ap2; ap2=ap2->next){
4079 if( ap2->type!=REDUCE ) continue;
4080 rp2 = ap2->x.rp;
4081 if( rp2==rbest ) continue;
4082 if( rp2==rp ) n++;
4083 }
4084 if( n>nbest ){
4085 nbest = n;
4086 rbest = rp;
4087 }
4088 }
4089
4090 /* Do not make a default if the number of rules to default
4091 ** is not at least 1 or if the wildcard token is a possible
4092 ** lookahead.
4093 */
4094 if( nbest<1 || usesWildcard ) continue;
4095
4096
4097 /* Combine matching REDUCE actions into a single default */
4098 for(ap=stp->ap; ap; ap=ap->next){
4099 if( ap->type==REDUCE && ap->x.rp==rbest ) break;
4100 }
4101 assert( ap );
4102 ap->sp = Symbol_new("{default}");
4103 for(ap=ap->next; ap; ap=ap->next){
4104 if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED;
4105 }
4106 stp->ap = Action_sort(stp->ap);
4107 }
4108}
4109
4110
4111/*
4112** Compare two states for sorting purposes. The smaller state is the
4113** one with the most non-terminal actions. If they have the same number
4114** of non-terminal actions, then the smaller is the one with the most
4115** token actions.
4116*/
4117static int stateResortCompare(const void *a, const void *b){
4118 const struct state *pA = *(const struct state**)a;
4119 const struct state *pB = *(const struct state**)b;
4120 int n;
4121
4122 n = pB->nNtAct - pA->nNtAct;
4123 if( n==0 ){
4124 n = pB->nTknAct - pA->nTknAct;
4125 if( n==0 ){
4126 n = pB->statenum - pA->statenum;
4127 }
4128 }
4129 assert( n!=0 );
4130 return n;
4131}
4132
4133
4134/*
4135** Renumber and resort states so that states with fewer choices
4136** occur at the end. Except, keep state 0 as the first state.
4137*/
4138void ResortStates(struct lemon *lemp)
4139{
4140 int i;
4141 struct state *stp;
4142 struct action *ap;
4143
4144 for(i=0; i<lemp->nstate; i++){
4145 stp = lemp->sorted[i];
4146 stp->nTknAct = stp->nNtAct = 0;
4147 stp->iDflt = lemp->nstate + lemp->nrule;
4148 stp->iTknOfst = NO_OFFSET;
4149 stp->iNtOfst = NO_OFFSET;
4150 for(ap=stp->ap; ap; ap=ap->next){
4151 if( compute_action(lemp,ap)>=0 ){
4152 if( ap->sp->index<lemp->nterminal ){
4153 stp->nTknAct++;
4154 }else if( ap->sp->index<lemp->nsymbol ){
4155 stp->nNtAct++;
4156 }else{
4157 stp->iDflt = compute_action(lemp, ap);
4158 }
4159 }
4160 }
4161 }
4162 qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]),
4163 stateResortCompare);
4164 for(i=0; i<lemp->nstate; i++){
4165 lemp->sorted[i]->statenum = i;
4166 }
4167}
4168
4169
4170/***************** From the file "set.c" ************************************/
4171/*
4172** Set manipulation routines for the LEMON parser generator.
4173*/
4174
4175static int size = 0;
4176
4177/* Set the set size */
4178void SetSize(int n)
4179{
4180 size = n+1;
4181}
4182
4183/* Allocate a new set */
4184char *SetNew(){
4185 char *s;
4186 s = (char*)calloc( size, 1);
4187 if( s==0 ){
4188 extern void memory_error();
4189 memory_error();
4190 }
4191 return s;
4192}
4193
4194/* Deallocate a set */
4195void SetFree(char *s)
4196{
4197 free(s);
4198}
4199
4200/* Add a new element to the set. Return TRUE if the element was added
4201** and FALSE if it was already there. */
4202int SetAdd(char *s, int e)
4203{
4204 int rv;
4205 assert( e>=0 && e<size );
4206 rv = s[e];
4207 s[e] = 1;
4208 return !rv;
4209}
4210
4211/* Add every element of s2 to s1. Return TRUE if s1 changes. */
4212int SetUnion(char *s1, char *s2)
4213{
4214 int i, progress;
4215 progress = 0;
4216 for(i=0; i<size; i++){
4217 if( s2[i]==0 ) continue;
4218 if( s1[i]==0 ){
4219 progress = 1;
4220 s1[i] = 1;
4221 }
4222 }
4223 return progress;
4224}
4225/********************** From the file "table.c" ****************************/
4226/*
4227** All code in this file has been automatically generated
4228** from a specification in the file
4229** "table.q"
4230** by the associative array code building program "aagen".
4231** Do not edit this file! Instead, edit the specification
4232** file, then rerun aagen.
4233*/
4234/*
4235** Code for processing tables in the LEMON parser generator.
4236*/
4237
4238PRIVATE int strhash(const char *x)
4239{
4240 int h = 0;
4241 while( *x) h = h*13 + *(x++);
4242 return h;
4243}
4244
4245/* Works like strdup, sort of. Save a string in malloced memory, but
4246** keep strings in a table so that the same string is not in more
4247** than one place.
4248*/
4249const char *Strsafe(const char *y)
4250{
4251 const char *z;
4252 char *cpy;
4253
4254 if( y==0 ) return 0;
4255 z = Strsafe_find(y);
4256 if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){
4257 strcpy(cpy,y);
4258 z = cpy;
4259 Strsafe_insert(z);
4260 }
4261 MemoryCheck(z);
4262 return z;
4263}
4264
4265/* There is one instance of the following structure for each
4266** associative array of type "x1".
4267*/
4268struct s_x1 {
4269 int size; /* The number of available slots. */
4270 /* Must be a power of 2 greater than or */
4271 /* equal to 1 */
4272 int count; /* Number of currently slots filled */
4273 struct s_x1node *tbl; /* The data stored here */
4274 struct s_x1node **ht; /* Hash table for lookups */
4275};
4276
4277/* There is one instance of this structure for every data element
4278** in an associative array of type "x1".
4279*/
4280typedef struct s_x1node {
4281 const char *data; /* The data */
4282 struct s_x1node *next; /* Next entry with the same hash */
4283 struct s_x1node **from; /* Previous link */
4284} x1node;
4285
4286/* There is only one instance of the array, which is the following */
4287static struct s_x1 *x1a;
4288
4289/* Allocate a new associative array */
4290void Strsafe_init(){
4291 if( x1a ) return;
4292 x1a = (struct s_x1*)malloc( sizeof(struct s_x1) );
4293 if( x1a ){
4294 x1a->size = 1024;
4295 x1a->count = 0;
4296 x1a->tbl = (x1node*)malloc(
4297 (sizeof(x1node) + sizeof(x1node*))*1024 );
4298 if( x1a->tbl==0 ){
4299 free(x1a);
4300 x1a = 0;
4301 }else{
4302 int i;
4303 x1a->ht = (x1node**)&(x1a->tbl[1024]);
4304 for(i=0; i<1024; i++) x1a->ht[i] = 0;
4305 }
4306 }
4307}
4308/* Insert a new record into the array. Return TRUE if successful.
4309** Prior data with the same key is NOT overwritten */
4310int Strsafe_insert(const char *data)
4311{
4312 x1node *np;
4313 int h;
4314 int ph;
4315
4316 if( x1a==0 ) return 0;
4317 ph = strhash(data);
4318 h = ph & (x1a->size-1);
4319 np = x1a->ht[h];
4320 while( np ){
4321 if( strcmp(np->data,data)==0 ){
4322 /* An existing entry with the same key is found. */
4323 /* Fail because overwrite is not allows. */
4324 return 0;
4325 }
4326 np = np->next;
4327 }
4328 if( x1a->count>=x1a->size ){
4329 /* Need to make the hash table bigger */
4330 int i,size;
4331 struct s_x1 array;
4332 array.size = size = x1a->size*2;
4333 array.count = x1a->count;
4334 array.tbl = (x1node*)malloc(
4335 (sizeof(x1node) + sizeof(x1node*))*size );
4336 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4337 array.ht = (x1node**)&(array.tbl[size]);
4338 for(i=0; i<size; i++) array.ht[i] = 0;
4339 for(i=0; i<x1a->count; i++){
4340 x1node *oldnp, *newnp;
4341 oldnp = &(x1a->tbl[i]);
4342 h = strhash(oldnp->data) & (size-1);
4343 newnp = &(array.tbl[i]);
4344 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4345 newnp->next = array.ht[h];
4346 newnp->data = oldnp->data;
4347 newnp->from = &(array.ht[h]);
4348 array.ht[h] = newnp;
4349 }
4350 free(x1a->tbl);
4351 *x1a = array;
4352 }
4353 /* Insert the new data */
4354 h = ph & (x1a->size-1);
4355 np = &(x1a->tbl[x1a->count++]);
4356 np->data = data;
4357 if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next);
4358 np->next = x1a->ht[h];
4359 x1a->ht[h] = np;
4360 np->from = &(x1a->ht[h]);
4361 return 1;
4362}
4363
4364/* Return a pointer to data assigned to the given key. Return NULL
4365** if no such key. */
4366const char *Strsafe_find(const char *key)
4367{
4368 int h;
4369 x1node *np;
4370
4371 if( x1a==0 ) return 0;
4372 h = strhash(key) & (x1a->size-1);
4373 np = x1a->ht[h];
4374 while( np ){
4375 if( strcmp(np->data,key)==0 ) break;
4376 np = np->next;
4377 }
4378 return np ? np->data : 0;
4379}
4380
4381/* Return a pointer to the (terminal or nonterminal) symbol "x".
4382** Create a new symbol if this is the first time "x" has been seen.
4383*/
4384struct symbol *Symbol_new(const char *x)
4385{
4386 struct symbol *sp;
4387
4388 sp = Symbol_find(x);
4389 if( sp==0 ){
4390 sp = (struct symbol *)calloc(1, sizeof(struct symbol) );
4391 MemoryCheck(sp);
4392 sp->name = Strsafe(x);
4393 sp->type = isupper(*x) ? TERMINAL : NONTERMINAL;
4394 sp->rule = 0;
4395 sp->fallback = 0;
4396 sp->prec = -1;
4397 sp->assoc = UNK;
4398 sp->firstset = 0;
4399 sp->lambda = LEMON_FALSE;
4400 sp->destructor = 0;
4401 sp->destLineno = 0;
4402 sp->datatype = 0;
4403 sp->useCnt = 0;
4404 Symbol_insert(sp,sp->name);
4405 }
4406 sp->useCnt++;
4407 return sp;
4408}
4409
4410/* Compare two symbols for working purposes
4411**
4412** Symbols that begin with upper case letters (terminals or tokens)
4413** must sort before symbols that begin with lower case letters
4414** (non-terminals). Other than that, the order does not matter.
4415**
4416** We find experimentally that leaving the symbols in their original
4417** order (the order they appeared in the grammar file) gives the
4418** smallest parser tables in SQLite.
4419*/
4420int Symbolcmpp(const void *_a, const void *_b)
4421{
4422 const struct symbol **a = (const struct symbol **) _a;
4423 const struct symbol **b = (const struct symbol **) _b;
4424 int i1 = (**a).index + 10000000*((**a).name[0]>'Z');
4425 int i2 = (**b).index + 10000000*((**b).name[0]>'Z');
4426 assert( i1!=i2 || strcmp((**a).name,(**b).name)==0 );
4427 return i1-i2;
4428}
4429
4430/* There is one instance of the following structure for each
4431** associative array of type "x2".
4432*/
4433struct s_x2 {
4434 int size; /* The number of available slots. */
4435 /* Must be a power of 2 greater than or */
4436 /* equal to 1 */
4437 int count; /* Number of currently slots filled */
4438 struct s_x2node *tbl; /* The data stored here */
4439 struct s_x2node **ht; /* Hash table for lookups */
4440};
4441
4442/* There is one instance of this structure for every data element
4443** in an associative array of type "x2".
4444*/
4445typedef struct s_x2node {
4446 struct symbol *data; /* The data */
4447 const char *key; /* The key */
4448 struct s_x2node *next; /* Next entry with the same hash */
4449 struct s_x2node **from; /* Previous link */
4450} x2node;
4451
4452/* There is only one instance of the array, which is the following */
4453static struct s_x2 *x2a;
4454
4455/* Allocate a new associative array */
4456void Symbol_init(){
4457 if( x2a ) return;
4458 x2a = (struct s_x2*)malloc( sizeof(struct s_x2) );
4459 if( x2a ){
4460 x2a->size = 128;
4461 x2a->count = 0;
4462 x2a->tbl = (x2node*)malloc(
4463 (sizeof(x2node) + sizeof(x2node*))*128 );
4464 if( x2a->tbl==0 ){
4465 free(x2a);
4466 x2a = 0;
4467 }else{
4468 int i;
4469 x2a->ht = (x2node**)&(x2a->tbl[128]);
4470 for(i=0; i<128; i++) x2a->ht[i] = 0;
4471 }
4472 }
4473}
4474/* Insert a new record into the array. Return TRUE if successful.
4475** Prior data with the same key is NOT overwritten */
4476int Symbol_insert(struct symbol *data, const char *key)
4477{
4478 x2node *np;
4479 int h;
4480 int ph;
4481
4482 if( x2a==0 ) return 0;
4483 ph = strhash(key);
4484 h = ph & (x2a->size-1);
4485 np = x2a->ht[h];
4486 while( np ){
4487 if( strcmp(np->key,key)==0 ){
4488 /* An existing entry with the same key is found. */
4489 /* Fail because overwrite is not allows. */
4490 return 0;
4491 }
4492 np = np->next;
4493 }
4494 if( x2a->count>=x2a->size ){
4495 /* Need to make the hash table bigger */
4496 int i,size;
4497 struct s_x2 array;
4498 array.size = size = x2a->size*2;
4499 array.count = x2a->count;
4500 array.tbl = (x2node*)malloc(
4501 (sizeof(x2node) + sizeof(x2node*))*size );
4502 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4503 array.ht = (x2node**)&(array.tbl[size]);
4504 for(i=0; i<size; i++) array.ht[i] = 0;
4505 for(i=0; i<x2a->count; i++){
4506 x2node *oldnp, *newnp;
4507 oldnp = &(x2a->tbl[i]);
4508 h = strhash(oldnp->key) & (size-1);
4509 newnp = &(array.tbl[i]);
4510 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4511 newnp->next = array.ht[h];
4512 newnp->key = oldnp->key;
4513 newnp->data = oldnp->data;
4514 newnp->from = &(array.ht[h]);
4515 array.ht[h] = newnp;
4516 }
4517 free(x2a->tbl);
4518 *x2a = array;
4519 }
4520 /* Insert the new data */
4521 h = ph & (x2a->size-1);
4522 np = &(x2a->tbl[x2a->count++]);
4523 np->key = key;
4524 np->data = data;
4525 if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next);
4526 np->next = x2a->ht[h];
4527 x2a->ht[h] = np;
4528 np->from = &(x2a->ht[h]);
4529 return 1;
4530}
4531
4532/* Return a pointer to data assigned to the given key. Return NULL
4533** if no such key. */
4534struct symbol *Symbol_find(const char *key)
4535{
4536 int h;
4537 x2node *np;
4538
4539 if( x2a==0 ) return 0;
4540 h = strhash(key) & (x2a->size-1);
4541 np = x2a->ht[h];
4542 while( np ){
4543 if( strcmp(np->key,key)==0 ) break;
4544 np = np->next;
4545 }
4546 return np ? np->data : 0;
4547}
4548
4549/* Return the n-th data. Return NULL if n is out of range. */
4550struct symbol *Symbol_Nth(int n)
4551{
4552 struct symbol *data;
4553 if( x2a && n>0 && n<=x2a->count ){
4554 data = x2a->tbl[n-1].data;
4555 }else{
4556 data = 0;
4557 }
4558 return data;
4559}
4560
4561/* Return the size of the array */
4562int Symbol_count()
4563{
4564 return x2a ? x2a->count : 0;
4565}
4566
4567/* Return an array of pointers to all data in the table.
4568** The array is obtained from malloc. Return NULL if memory allocation
4569** problems, or if the array is empty. */
4570struct symbol **Symbol_arrayof()
4571{
4572 struct symbol **array;
4573 int i,size;
4574 if( x2a==0 ) return 0;
4575 size = x2a->count;
4576 array = (struct symbol **)calloc(size, sizeof(struct symbol *));
4577 if( array ){
4578 for(i=0; i<size; i++) array[i] = x2a->tbl[i].data;
4579 }
4580 return array;
4581}
4582
4583/* Compare two configurations */
4584int Configcmp(const char *_a,const char *_b)
4585{
4586 const struct config *a = (struct config *) _a;
4587 const struct config *b = (struct config *) _b;
4588 int x;
4589 x = a->rp->index - b->rp->index;
4590 if( x==0 ) x = a->dot - b->dot;
4591 return x;
4592}
4593
4594/* Compare two states */
4595PRIVATE int statecmp(struct config *a, struct config *b)
4596{
4597 int rc;
4598 for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){
4599 rc = a->rp->index - b->rp->index;
4600 if( rc==0 ) rc = a->dot - b->dot;
4601 }
4602 if( rc==0 ){
4603 if( a ) rc = 1;
4604 if( b ) rc = -1;
4605 }
4606 return rc;
4607}
4608
4609/* Hash a state */
4610PRIVATE int statehash(struct config *a)
4611{
4612 int h=0;
4613 while( a ){
4614 h = h*571 + a->rp->index*37 + a->dot;
4615 a = a->bp;
4616 }
4617 return h;
4618}
4619
4620/* Allocate a new state structure */
4621struct state *State_new()
4622{
4623 struct state *newstate;
4624 newstate = (struct state *)calloc(1, sizeof(struct state) );
4625 MemoryCheck(newstate);
4626 return newstate;
4627}
4628
4629/* There is one instance of the following structure for each
4630** associative array of type "x3".
4631*/
4632struct s_x3 {
4633 int size; /* The number of available slots. */
4634 /* Must be a power of 2 greater than or */
4635 /* equal to 1 */
4636 int count; /* Number of currently slots filled */
4637 struct s_x3node *tbl; /* The data stored here */
4638 struct s_x3node **ht; /* Hash table for lookups */
4639};
4640
4641/* There is one instance of this structure for every data element
4642** in an associative array of type "x3".
4643*/
4644typedef struct s_x3node {
4645 struct state *data; /* The data */
4646 struct config *key; /* The key */
4647 struct s_x3node *next; /* Next entry with the same hash */
4648 struct s_x3node **from; /* Previous link */
4649} x3node;
4650
4651/* There is only one instance of the array, which is the following */
4652static struct s_x3 *x3a;
4653
4654/* Allocate a new associative array */
4655void State_init(){
4656 if( x3a ) return;
4657 x3a = (struct s_x3*)malloc( sizeof(struct s_x3) );
4658 if( x3a ){
4659 x3a->size = 128;
4660 x3a->count = 0;
4661 x3a->tbl = (x3node*)malloc(
4662 (sizeof(x3node) + sizeof(x3node*))*128 );
4663 if( x3a->tbl==0 ){
4664 free(x3a);
4665 x3a = 0;
4666 }else{
4667 int i;
4668 x3a->ht = (x3node**)&(x3a->tbl[128]);
4669 for(i=0; i<128; i++) x3a->ht[i] = 0;
4670 }
4671 }
4672}
4673/* Insert a new record into the array. Return TRUE if successful.
4674** Prior data with the same key is NOT overwritten */
4675int State_insert(struct state *data, struct config *key)
4676{
4677 x3node *np;
4678 int h;
4679 int ph;
4680
4681 if( x3a==0 ) return 0;
4682 ph = statehash(key);
4683 h = ph & (x3a->size-1);
4684 np = x3a->ht[h];
4685 while( np ){
4686 if( statecmp(np->key,key)==0 ){
4687 /* An existing entry with the same key is found. */
4688 /* Fail because overwrite is not allows. */
4689 return 0;
4690 }
4691 np = np->next;
4692 }
4693 if( x3a->count>=x3a->size ){
4694 /* Need to make the hash table bigger */
4695 int i,size;
4696 struct s_x3 array;
4697 array.size = size = x3a->size*2;
4698 array.count = x3a->count;
4699 array.tbl = (x3node*)malloc(
4700 (sizeof(x3node) + sizeof(x3node*))*size );
4701 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4702 array.ht = (x3node**)&(array.tbl[size]);
4703 for(i=0; i<size; i++) array.ht[i] = 0;
4704 for(i=0; i<x3a->count; i++){
4705 x3node *oldnp, *newnp;
4706 oldnp = &(x3a->tbl[i]);
4707 h = statehash(oldnp->key) & (size-1);
4708 newnp = &(array.tbl[i]);
4709 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4710 newnp->next = array.ht[h];
4711 newnp->key = oldnp->key;
4712 newnp->data = oldnp->data;
4713 newnp->from = &(array.ht[h]);
4714 array.ht[h] = newnp;
4715 }
4716 free(x3a->tbl);
4717 *x3a = array;
4718 }
4719 /* Insert the new data */
4720 h = ph & (x3a->size-1);
4721 np = &(x3a->tbl[x3a->count++]);
4722 np->key = key;
4723 np->data = data;
4724 if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next);
4725 np->next = x3a->ht[h];
4726 x3a->ht[h] = np;
4727 np->from = &(x3a->ht[h]);
4728 return 1;
4729}
4730
4731/* Return a pointer to data assigned to the given key. Return NULL
4732** if no such key. */
4733struct state *State_find(struct config *key)
4734{
4735 int h;
4736 x3node *np;
4737
4738 if( x3a==0 ) return 0;
4739 h = statehash(key) & (x3a->size-1);
4740 np = x3a->ht[h];
4741 while( np ){
4742 if( statecmp(np->key,key)==0 ) break;
4743 np = np->next;
4744 }
4745 return np ? np->data : 0;
4746}
4747
4748/* Return an array of pointers to all data in the table.
4749** The array is obtained from malloc. Return NULL if memory allocation
4750** problems, or if the array is empty. */
4751struct state **State_arrayof()
4752{
4753 struct state **array;
4754 int i,size;
4755 if( x3a==0 ) return 0;
4756 size = x3a->count;
4757 array = (struct state **)malloc( sizeof(struct state *)*size );
4758 if( array ){
4759 for(i=0; i<size; i++) array[i] = x3a->tbl[i].data;
4760 }
4761 return array;
4762}
4763
4764/* Hash a configuration */
4765PRIVATE int confighash(struct config *a)
4766{
4767 int h=0;
4768 h = h*571 + a->rp->index*37 + a->dot;
4769 return h;
4770}
4771
4772/* There is one instance of the following structure for each
4773** associative array of type "x4".
4774*/
4775struct s_x4 {
4776 int size; /* The number of available slots. */
4777 /* Must be a power of 2 greater than or */
4778 /* equal to 1 */
4779 int count; /* Number of currently slots filled */
4780 struct s_x4node *tbl; /* The data stored here */
4781 struct s_x4node **ht; /* Hash table for lookups */
4782};
4783
4784/* There is one instance of this structure for every data element
4785** in an associative array of type "x4".
4786*/
4787typedef struct s_x4node {
4788 struct config *data; /* The data */
4789 struct s_x4node *next; /* Next entry with the same hash */
4790 struct s_x4node **from; /* Previous link */
4791} x4node;
4792
4793/* There is only one instance of the array, which is the following */
4794static struct s_x4 *x4a;
4795
4796/* Allocate a new associative array */
4797void Configtable_init(){
4798 if( x4a ) return;
4799 x4a = (struct s_x4*)malloc( sizeof(struct s_x4) );
4800 if( x4a ){
4801 x4a->size = 64;
4802 x4a->count = 0;
4803 x4a->tbl = (x4node*)malloc(
4804 (sizeof(x4node) + sizeof(x4node*))*64 );
4805 if( x4a->tbl==0 ){
4806 free(x4a);
4807 x4a = 0;
4808 }else{
4809 int i;
4810 x4a->ht = (x4node**)&(x4a->tbl[64]);
4811 for(i=0; i<64; i++) x4a->ht[i] = 0;
4812 }
4813 }
4814}
4815/* Insert a new record into the array. Return TRUE if successful.
4816** Prior data with the same key is NOT overwritten */
4817int Configtable_insert(struct config *data)
4818{
4819 x4node *np;
4820 int h;
4821 int ph;
4822
4823 if( x4a==0 ) return 0;
4824 ph = confighash(data);
4825 h = ph & (x4a->size-1);
4826 np = x4a->ht[h];
4827 while( np ){
4828 if( Configcmp((const char *) np->data,(const char *) data)==0 ){
4829 /* An existing entry with the same key is found. */
4830 /* Fail because overwrite is not allows. */
4831 return 0;
4832 }
4833 np = np->next;
4834 }
4835 if( x4a->count>=x4a->size ){
4836 /* Need to make the hash table bigger */
4837 int i,size;
4838 struct s_x4 array;
4839 array.size = size = x4a->size*2;
4840 array.count = x4a->count;
4841 array.tbl = (x4node*)malloc(
4842 (sizeof(x4node) + sizeof(x4node*))*size );
4843 if( array.tbl==0 ) return 0; /* Fail due to malloc failure */
4844 array.ht = (x4node**)&(array.tbl[size]);
4845 for(i=0; i<size; i++) array.ht[i] = 0;
4846 for(i=0; i<x4a->count; i++){
4847 x4node *oldnp, *newnp;
4848 oldnp = &(x4a->tbl[i]);
4849 h = confighash(oldnp->data) & (size-1);
4850 newnp = &(array.tbl[i]);
4851 if( array.ht[h] ) array.ht[h]->from = &(newnp->next);
4852 newnp->next = array.ht[h];
4853 newnp->data = oldnp->data;
4854 newnp->from = &(array.ht[h]);
4855 array.ht[h] = newnp;
4856 }
4857 free(x4a->tbl);
4858 *x4a = array;
4859 }
4860 /* Insert the new data */
4861 h = ph & (x4a->size-1);
4862 np = &(x4a->tbl[x4a->count++]);
4863 np->data = data;
4864 if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next);
4865 np->next = x4a->ht[h];
4866 x4a->ht[h] = np;
4867 np->from = &(x4a->ht[h]);
4868 return 1;
4869}
4870
4871/* Return a pointer to data assigned to the given key. Return NULL
4872** if no such key. */
4873struct config *Configtable_find(struct config *key)
4874{
4875 int h;
4876 x4node *np;
4877
4878 if( x4a==0 ) return 0;
4879 h = confighash(key) & (x4a->size-1);
4880 np = x4a->ht[h];
4881 while( np ){
4882 if( Configcmp((const char *) np->data,(const char *) key)==0 ) break;
4883 np = np->next;
4884 }
4885 return np ? np->data : 0;
4886}
4887
4888/* Remove all data from the table. Pass each data to the function "f"
4889** as it is removed. ("f" may be null to avoid this step.) */
4890void Configtable_clear(int(*f)(struct config *))
4891{
4892 int i;
4893 if( x4a==0 || x4a->count==0 ) return;
4894 if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data);
4895 for(i=0; i<x4a->size; i++) x4a->ht[i] = 0;
4896 x4a->count = 0;
4897 return;
4898}
diff --git a/libraries/lemon/lempar.c b/libraries/lemon/lempar.c
new file mode 100644
index 0000000..fe56d2d
--- /dev/null
+++ b/libraries/lemon/lempar.c
@@ -0,0 +1,850 @@
1/* Driver template for the LEMON parser generator.
2** The author disclaims copyright to this source code.
3*/
4/* First off, code is included that follows the "include" declaration
5** in the input grammar file. */
6#include <stdio.h>
7%%
8/* Next is all token values, in a form suitable for use by makeheaders.
9** This section will be null unless lemon is run with the -m switch.
10*/
11/*
12** These constants (all generated automatically by the parser generator)
13** specify the various kinds of tokens (terminals) that the parser
14** understands.
15**
16** Each symbol here is a terminal symbol in the grammar.
17*/
18%%
19/* Make sure the INTERFACE macro is defined.
20*/
21#ifndef INTERFACE
22# define INTERFACE 1
23#endif
24/* The next thing included is series of defines which control
25** various aspects of the generated parser.
26** YYCODETYPE is the data type used for storing terminal
27** and nonterminal numbers. "unsigned char" is
28** used if there are fewer than 250 terminals
29** and nonterminals. "int" is used otherwise.
30** YYNOCODE is a number of type YYCODETYPE which corresponds
31** to no legal terminal or nonterminal number. This
32** number is used to fill in empty slots of the hash
33** table.
34** YYFALLBACK If defined, this indicates that one or more tokens
35** have fall-back values which should be used if the
36** original value of the token will not parse.
37** YYACTIONTYPE is the data type used for storing terminal
38** and nonterminal numbers. "unsigned char" is
39** used if there are fewer than 250 rules and
40** states combined. "int" is used otherwise.
41** ParseTOKENTYPE is the data type used for minor tokens given
42** directly to the parser from the tokenizer.
43** YYMINORTYPE is the data type used for all minor tokens.
44** This is typically a union of many types, one of
45** which is ParseTOKENTYPE. The entry in the union
46** for base tokens is called "yy0".
47** YYSTACKDEPTH is the maximum depth of the parser's stack. If
48** zero the stack is dynamically sized using realloc()
49** ParseARG_SDECL A static variable declaration for the %extra_argument
50** ParseARG_PDECL A parameter declaration for the %extra_argument
51** ParseARG_STORE Code to store %extra_argument into yypParser
52** ParseARG_FETCH Code to extract %extra_argument from yypParser
53** YYNSTATE the combined number of states.
54** YYNRULE the number of rules in the grammar
55** YYERRORSYMBOL is the code number of the error symbol. If not
56** defined, then do no error processing.
57*/
58%%
59#define YY_NO_ACTION (YYNSTATE+YYNRULE+2)
60#define YY_ACCEPT_ACTION (YYNSTATE+YYNRULE+1)
61#define YY_ERROR_ACTION (YYNSTATE+YYNRULE)
62
63/* The yyzerominor constant is used to initialize instances of
64** YYMINORTYPE objects to zero. */
65static const YYMINORTYPE yyzerominor = { 0 };
66
67/* Define the yytestcase() macro to be a no-op if is not already defined
68** otherwise.
69**
70** Applications can choose to define yytestcase() in the %include section
71** to a macro that can assist in verifying code coverage. For production
72** code the yytestcase() macro should be turned off. But it is useful
73** for testing.
74*/
75#ifndef yytestcase
76# define yytestcase(X)
77#endif
78
79
80/* Next are the tables used to determine what action to take based on the
81** current state and lookahead token. These tables are used to implement
82** functions that take a state number and lookahead value and return an
83** action integer.
84**
85** Suppose the action integer is N. Then the action is determined as
86** follows
87**
88** 0 <= N < YYNSTATE Shift N. That is, push the lookahead
89** token onto the stack and goto state N.
90**
91** YYNSTATE <= N < YYNSTATE+YYNRULE Reduce by rule N-YYNSTATE.
92**
93** N == YYNSTATE+YYNRULE A syntax error has occurred.
94**
95** N == YYNSTATE+YYNRULE+1 The parser accepts its input.
96**
97** N == YYNSTATE+YYNRULE+2 No such action. Denotes unused
98** slots in the yy_action[] table.
99**
100** The action table is constructed as a single large table named yy_action[].
101** Given state S and lookahead X, the action is computed as
102**
103** yy_action[ yy_shift_ofst[S] + X ]
104**
105** If the index value yy_shift_ofst[S]+X is out of range or if the value
106** yy_lookahead[yy_shift_ofst[S]+X] is not equal to X or if yy_shift_ofst[S]
107** is equal to YY_SHIFT_USE_DFLT, it means that the action is not in the table
108** and that yy_default[S] should be used instead.
109**
110** The formula above is for computing the action when the lookahead is
111** a terminal symbol. If the lookahead is a non-terminal (as occurs after
112** a reduce action) then the yy_reduce_ofst[] array is used in place of
113** the yy_shift_ofst[] array and YY_REDUCE_USE_DFLT is used in place of
114** YY_SHIFT_USE_DFLT.
115**
116** The following are the tables generated in this section:
117**
118** yy_action[] A single table containing all actions.
119** yy_lookahead[] A table containing the lookahead for each entry in
120** yy_action. Used to detect hash collisions.
121** yy_shift_ofst[] For each state, the offset into yy_action for
122** shifting terminals.
123** yy_reduce_ofst[] For each state, the offset into yy_action for
124** shifting non-terminals after a reduce.
125** yy_default[] Default action for each state.
126*/
127%%
128
129/* The next table maps tokens into fallback tokens. If a construct
130** like the following:
131**
132** %fallback ID X Y Z.
133**
134** appears in the grammar, then ID becomes a fallback token for X, Y,
135** and Z. Whenever one of the tokens X, Y, or Z is input to the parser
136** but it does not parse, the type of the token is changed to ID and
137** the parse is retried before an error is thrown.
138*/
139#ifdef YYFALLBACK
140static const YYCODETYPE yyFallback[] = {
141%%
142};
143#endif /* YYFALLBACK */
144
145/* The following structure represents a single element of the
146** parser's stack. Information stored includes:
147**
148** + The state number for the parser at this level of the stack.
149**
150** + The value of the token stored at this level of the stack.
151** (In other words, the "major" token.)
152**
153** + The semantic value stored at this level of the stack. This is
154** the information used by the action routines in the grammar.
155** It is sometimes called the "minor" token.
156*/
157struct yyStackEntry {
158 YYACTIONTYPE stateno; /* The state-number */
159 YYCODETYPE major; /* The major token value. This is the code
160 ** number for the token at this stack level */
161 YYMINORTYPE minor; /* The user-supplied minor token value. This
162 ** is the value of the token */
163};
164typedef struct yyStackEntry yyStackEntry;
165
166/* The state of the parser is completely contained in an instance of
167** the following structure */
168struct yyParser {
169 int yyidx; /* Index of top element in stack */
170#ifdef YYTRACKMAXSTACKDEPTH
171 int yyidxMax; /* Maximum value of yyidx */
172#endif
173 int yyerrcnt; /* Shifts left before out of the error */
174 ParseARG_SDECL /* A place to hold %extra_argument */
175#if YYSTACKDEPTH<=0
176 int yystksz; /* Current side of the stack */
177 yyStackEntry *yystack; /* The parser's stack */
178#else
179 yyStackEntry yystack[YYSTACKDEPTH]; /* The parser's stack */
180#endif
181};
182typedef struct yyParser yyParser;
183
184#ifndef NDEBUG
185#include <stdio.h>
186static FILE *yyTraceFILE = 0;
187static char *yyTracePrompt = 0;
188#endif /* NDEBUG */
189
190#ifndef NDEBUG
191/*
192** Turn parser tracing on by giving a stream to which to write the trace
193** and a prompt to preface each trace message. Tracing is turned off
194** by making either argument NULL
195**
196** Inputs:
197** <ul>
198** <li> A FILE* to which trace output should be written.
199** If NULL, then tracing is turned off.
200** <li> A prefix string written at the beginning of every
201** line of trace output. If NULL, then tracing is
202** turned off.
203** </ul>
204**
205** Outputs:
206** None.
207*/
208void ParseTrace(FILE *TraceFILE, char *zTracePrompt){
209 yyTraceFILE = TraceFILE;
210 yyTracePrompt = zTracePrompt;
211 if( yyTraceFILE==0 ) yyTracePrompt = 0;
212 else if( yyTracePrompt==0 ) yyTraceFILE = 0;
213}
214#endif /* NDEBUG */
215
216#ifndef NDEBUG
217/* For tracing shifts, the names of all terminals and nonterminals
218** are required. The following table supplies these names */
219static const char *const yyTokenName[] = {
220%%
221};
222#endif /* NDEBUG */
223
224#ifndef NDEBUG
225/* For tracing reduce actions, the names of all rules are required.
226*/
227static const char *const yyRuleName[] = {
228%%
229};
230#endif /* NDEBUG */
231
232
233#if YYSTACKDEPTH<=0
234/*
235** Try to increase the size of the parser stack.
236*/
237static void yyGrowStack(yyParser *p){
238 int newSize;
239 yyStackEntry *pNew;
240
241 newSize = p->yystksz*2 + 100;
242 pNew = realloc(p->yystack, newSize*sizeof(pNew[0]));
243 if( pNew ){
244 p->yystack = pNew;
245 p->yystksz = newSize;
246#ifndef NDEBUG
247 if( yyTraceFILE ){
248 fprintf(yyTraceFILE,"%sStack grows to %d entries!\n",
249 yyTracePrompt, p->yystksz);
250 }
251#endif
252 }
253}
254#endif
255
256/*
257** This function allocates a new parser.
258** The only argument is a pointer to a function which works like
259** malloc.
260**
261** Inputs:
262** A pointer to the function used to allocate memory.
263**
264** Outputs:
265** A pointer to a parser. This pointer is used in subsequent calls
266** to Parse and ParseFree.
267*/
268void *ParseAlloc(void *(*mallocProc)(size_t)){
269 yyParser *pParser;
270 pParser = (yyParser*)(*mallocProc)( (size_t)sizeof(yyParser) );
271 if( pParser ){
272 pParser->yyidx = -1;
273#ifdef YYTRACKMAXSTACKDEPTH
274 pParser->yyidxMax = 0;
275#endif
276#if YYSTACKDEPTH<=0
277 pParser->yystack = NULL;
278 pParser->yystksz = 0;
279 yyGrowStack(pParser);
280#endif
281 }
282 return pParser;
283}
284
285/* The following function deletes the value associated with a
286** symbol. The symbol can be either a terminal or nonterminal.
287** "yymajor" is the symbol code, and "yypminor" is a pointer to
288** the value.
289*/
290static void yy_destructor(
291 yyParser *yypParser, /* The parser */
292 YYCODETYPE yymajor, /* Type code for object to destroy */
293 YYMINORTYPE *yypminor /* The object to be destroyed */
294){
295 ParseARG_FETCH;
296 switch( yymajor ){
297 /* Here is inserted the actions which take place when a
298 ** terminal or non-terminal is destroyed. This can happen
299 ** when the symbol is popped from the stack during a
300 ** reduce or during error processing or when a parser is
301 ** being destroyed before it is finished parsing.
302 **
303 ** Note: during a reduce, the only symbols destroyed are those
304 ** which appear on the RHS of the rule, but which are not used
305 ** inside the C code.
306 */
307%%
308 default: break; /* If no destructor action specified: do nothing */
309 }
310}
311
312/*
313** Pop the parser's stack once.
314**
315** If there is a destructor routine associated with the token which
316** is popped from the stack, then call it.
317**
318** Return the major token number for the symbol popped.
319*/
320static int yy_pop_parser_stack(yyParser *pParser){
321 YYCODETYPE yymajor;
322 yyStackEntry *yytos = &pParser->yystack[pParser->yyidx];
323
324 if( pParser->yyidx<0 ) return 0;
325#ifndef NDEBUG
326 if( yyTraceFILE && pParser->yyidx>=0 ){
327 fprintf(yyTraceFILE,"%sPopping %s\n",
328 yyTracePrompt,
329 yyTokenName[yytos->major]);
330 }
331#endif
332 yymajor = yytos->major;
333 yy_destructor(pParser, yymajor, &yytos->minor);
334 pParser->yyidx--;
335 return yymajor;
336}
337
338/*
339** Deallocate and destroy a parser. Destructors are all called for
340** all stack elements before shutting the parser down.
341**
342** Inputs:
343** <ul>
344** <li> A pointer to the parser. This should be a pointer
345** obtained from ParseAlloc.
346** <li> A pointer to a function used to reclaim memory obtained
347** from malloc.
348** </ul>
349*/
350void ParseFree(
351 void *p, /* The parser to be deleted */
352 void (*freeProc)(void*) /* Function used to reclaim memory */
353){
354 yyParser *pParser = (yyParser*)p;
355 if( pParser==0 ) return;
356 while( pParser->yyidx>=0 ) yy_pop_parser_stack(pParser);
357#if YYSTACKDEPTH<=0
358 free(pParser->yystack);
359#endif
360 (*freeProc)((void*)pParser);
361}
362
363/*
364** Return the peak depth of the stack for a parser.
365*/
366#ifdef YYTRACKMAXSTACKDEPTH
367int ParseStackPeak(void *p){
368 yyParser *pParser = (yyParser*)p;
369 return pParser->yyidxMax;
370}
371#endif
372
373/*
374** Find the appropriate action for a parser given the terminal
375** look-ahead token iLookAhead.
376**
377** If the look-ahead token is YYNOCODE, then check to see if the action is
378** independent of the look-ahead. If it is, return the action, otherwise
379** return YY_NO_ACTION.
380*/
381static int yy_find_shift_action(
382 yyParser *pParser, /* The parser */
383 YYCODETYPE iLookAhead /* The look-ahead token */
384){
385 int i;
386 int stateno = pParser->yystack[pParser->yyidx].stateno;
387
388 if( stateno>YY_SHIFT_COUNT
389 || (i = yy_shift_ofst[stateno])==YY_SHIFT_USE_DFLT ){
390 return yy_default[stateno];
391 }
392 assert( iLookAhead!=YYNOCODE );
393 i += iLookAhead;
394 if( i<0 || i>=YY_ACTTAB_COUNT || yy_lookahead[i]!=iLookAhead ){
395 if( iLookAhead>0 ){
396#ifdef YYFALLBACK
397 YYCODETYPE iFallback; /* Fallback token */
398 if( iLookAhead<sizeof(yyFallback)/sizeof(yyFallback[0])
399 && (iFallback = yyFallback[iLookAhead])!=0 ){
400#ifndef NDEBUG
401 if( yyTraceFILE ){
402 fprintf(yyTraceFILE, "%sFALLBACK %s => %s\n",
403 yyTracePrompt, yyTokenName[iLookAhead], yyTokenName[iFallback]);
404 }
405#endif
406 return yy_find_shift_action(pParser, iFallback);
407 }
408#endif
409#ifdef YYWILDCARD
410 {
411 int j = i - iLookAhead + YYWILDCARD;
412 if(
413#if YY_SHIFT_MIN+YYWILDCARD<0
414 j>=0 &&
415#endif
416#if YY_SHIFT_MAX+YYWILDCARD>=YY_ACTTAB_COUNT
417 j<YY_ACTTAB_COUNT &&
418#endif
419 yy_lookahead[j]==YYWILDCARD
420 ){
421#ifndef NDEBUG
422 if( yyTraceFILE ){
423 fprintf(yyTraceFILE, "%sWILDCARD %s => %s\n",
424 yyTracePrompt, yyTokenName[iLookAhead], yyTokenName[YYWILDCARD]);
425 }
426#endif /* NDEBUG */
427 return yy_action[j];
428 }
429 }
430#endif /* YYWILDCARD */
431 }
432 return yy_default[stateno];
433 }else{
434 return yy_action[i];
435 }
436}
437
438/*
439** Find the appropriate action for a parser given the non-terminal
440** look-ahead token iLookAhead.
441**
442** If the look-ahead token is YYNOCODE, then check to see if the action is
443** independent of the look-ahead. If it is, return the action, otherwise
444** return YY_NO_ACTION.
445*/
446static int yy_find_reduce_action(
447 int stateno, /* Current state number */
448 YYCODETYPE iLookAhead /* The look-ahead token */
449){
450 int i;
451#ifdef YYERRORSYMBOL
452 if( stateno>YY_REDUCE_COUNT ){
453 return yy_default[stateno];
454 }
455#else
456 assert( stateno<=YY_REDUCE_COUNT );
457#endif
458 i = yy_reduce_ofst[stateno];
459 assert( i!=YY_REDUCE_USE_DFLT );
460 assert( iLookAhead!=YYNOCODE );
461 i += iLookAhead;
462#ifdef YYERRORSYMBOL
463 if( i<0 || i>=YY_ACTTAB_COUNT || yy_lookahead[i]!=iLookAhead ){
464 return yy_default[stateno];
465 }
466#else
467 assert( i>=0 && i<YY_ACTTAB_COUNT );
468 assert( yy_lookahead[i]==iLookAhead );
469#endif
470 return yy_action[i];
471}
472
473/*
474** The following routine is called if the stack overflows.
475*/
476static void yyStackOverflow(yyParser *yypParser, YYMINORTYPE *yypMinor){
477 ParseARG_FETCH;
478 yypParser->yyidx--;
479#ifndef NDEBUG
480 if( yyTraceFILE ){
481 fprintf(yyTraceFILE,"%sStack Overflow!\n",yyTracePrompt);
482 }
483#endif
484 while( yypParser->yyidx>=0 ) yy_pop_parser_stack(yypParser);
485 /* Here code is inserted which will execute if the parser
486 ** stack every overflows */
487%%
488 ParseARG_STORE; /* Suppress warning about unused %extra_argument var */
489}
490
491/*
492** Perform a shift action.
493*/
494static void yy_shift(
495 yyParser *yypParser, /* The parser to be shifted */
496 int yyNewState, /* The new state to shift in */
497 int yyMajor, /* The major token to shift in */
498 YYMINORTYPE *yypMinor /* Pointer to the minor token to shift in */
499){
500 yyStackEntry *yytos;
501 yypParser->yyidx++;
502#ifdef YYTRACKMAXSTACKDEPTH
503 if( yypParser->yyidx>yypParser->yyidxMax ){
504 yypParser->yyidxMax = yypParser->yyidx;
505 }
506#endif
507#if YYSTACKDEPTH>0
508 if( yypParser->yyidx>=YYSTACKDEPTH ){
509 yyStackOverflow(yypParser, yypMinor);
510 return;
511 }
512#else
513 if( yypParser->yyidx>=yypParser->yystksz ){
514 yyGrowStack(yypParser);
515 if( yypParser->yyidx>=yypParser->yystksz ){
516 yyStackOverflow(yypParser, yypMinor);
517 return;
518 }
519 }
520#endif
521 yytos = &yypParser->yystack[yypParser->yyidx];
522 yytos->stateno = (YYACTIONTYPE)yyNewState;
523 yytos->major = (YYCODETYPE)yyMajor;
524 yytos->minor = *yypMinor;
525#ifndef NDEBUG
526 if( yyTraceFILE && yypParser->yyidx>0 ){
527 int i;
528 fprintf(yyTraceFILE,"%sShift %d\n",yyTracePrompt,yyNewState);
529 fprintf(yyTraceFILE,"%sStack:",yyTracePrompt);
530 for(i=1; i<=yypParser->yyidx; i++)
531 fprintf(yyTraceFILE," %s",yyTokenName[yypParser->yystack[i].major]);
532 fprintf(yyTraceFILE,"\n");
533 }
534#endif
535}
536
537/* The following table contains information about every rule that
538** is used during the reduce.
539*/
540static const struct {
541 YYCODETYPE lhs; /* Symbol on the left-hand side of the rule */
542 unsigned char nrhs; /* Number of right-hand side symbols in the rule */
543} yyRuleInfo[] = {
544%%
545};
546
547static void yy_accept(yyParser*); /* Forward Declaration */
548
549/*
550** Perform a reduce action and the shift that must immediately
551** follow the reduce.
552*/
553static void yy_reduce(
554 yyParser *yypParser, /* The parser */
555 int yyruleno /* Number of the rule by which to reduce */
556){
557 int yygoto; /* The next state */
558 int yyact; /* The next action */
559 YYMINORTYPE yygotominor; /* The LHS of the rule reduced */
560 yyStackEntry *yymsp; /* The top of the parser's stack */
561 int yysize; /* Amount to pop the stack */
562 ParseARG_FETCH;
563 yymsp = &yypParser->yystack[yypParser->yyidx];
564#ifndef NDEBUG
565 if( yyTraceFILE && yyruleno>=0
566 && yyruleno<(int)(sizeof(yyRuleName)/sizeof(yyRuleName[0])) ){
567 fprintf(yyTraceFILE, "%sReduce [%s].\n", yyTracePrompt,
568 yyRuleName[yyruleno]);
569 }
570#endif /* NDEBUG */
571
572 /* Silence complaints from purify about yygotominor being uninitialized
573 ** in some cases when it is copied into the stack after the following
574 ** switch. yygotominor is uninitialized when a rule reduces that does
575 ** not set the value of its left-hand side nonterminal. Leaving the
576 ** value of the nonterminal uninitialized is utterly harmless as long
577 ** as the value is never used. So really the only thing this code
578 ** accomplishes is to quieten purify.
579 **
580 ** 2007-01-16: The wireshark project (www.wireshark.org) reports that
581 ** without this code, their parser segfaults. I'm not sure what there
582 ** parser is doing to make this happen. This is the second bug report
583 ** from wireshark this week. Clearly they are stressing Lemon in ways
584 ** that it has not been previously stressed... (SQLite ticket #2172)
585 */
586 /*memset(&yygotominor, 0, sizeof(yygotominor));*/
587 yygotominor = yyzerominor;
588
589
590 switch( yyruleno ){
591 /* Beginning here are the reduction cases. A typical example
592 ** follows:
593 ** case 0:
594 ** #line <lineno> <grammarfile>
595 ** { ... } // User supplied code
596 ** #line <lineno> <thisfile>
597 ** break;
598 */
599%%
600 };
601 yygoto = yyRuleInfo[yyruleno].lhs;
602 yysize = yyRuleInfo[yyruleno].nrhs;
603 yypParser->yyidx -= yysize;
604 yyact = yy_find_reduce_action(yymsp[-yysize].stateno,(YYCODETYPE)yygoto);
605 if( yyact < YYNSTATE ){
606#ifdef NDEBUG
607 /* If we are not debugging and the reduce action popped at least
608 ** one element off the stack, then we can push the new element back
609 ** onto the stack here, and skip the stack overflow test in yy_shift().
610 ** That gives a significant speed improvement. */
611 if( yysize ){
612 yypParser->yyidx++;
613 yymsp -= yysize-1;
614 yymsp->stateno = (YYACTIONTYPE)yyact;
615 yymsp->major = (YYCODETYPE)yygoto;
616 yymsp->minor = yygotominor;
617 }else
618#endif
619 {
620 yy_shift(yypParser,yyact,yygoto,&yygotominor);
621 }
622 }else{
623 assert( yyact == YYNSTATE + YYNRULE + 1 );
624 yy_accept(yypParser);
625 }
626}
627
628/*
629** The following code executes when the parse fails
630*/
631#ifndef YYNOERRORRECOVERY
632static void yy_parse_failed(
633 yyParser *yypParser /* The parser */
634){
635 ParseARG_FETCH;
636#ifndef NDEBUG
637 if( yyTraceFILE ){
638 fprintf(yyTraceFILE,"%sFail!\n",yyTracePrompt);
639 }
640#endif
641 while( yypParser->yyidx>=0 ) yy_pop_parser_stack(yypParser);
642 /* Here code is inserted which will be executed whenever the
643 ** parser fails */
644%%
645 ParseARG_STORE; /* Suppress warning about unused %extra_argument variable */
646}
647#endif /* YYNOERRORRECOVERY */
648
649/*
650** The following code executes when a syntax error first occurs.
651*/
652static void yy_syntax_error(
653 yyParser *yypParser, /* The parser */
654 int yymajor, /* The major type of the error token */
655 YYMINORTYPE yyminor /* The minor type of the error token */
656){
657 ParseARG_FETCH;
658#define TOKEN (yyminor.yy0)
659%%
660 ParseARG_STORE; /* Suppress warning about unused %extra_argument variable */
661}
662
663/*
664** The following is executed when the parser accepts
665*/
666static void yy_accept(
667 yyParser *yypParser /* The parser */
668){
669 ParseARG_FETCH;
670#ifndef NDEBUG
671 if( yyTraceFILE ){
672 fprintf(yyTraceFILE,"%sAccept!\n",yyTracePrompt);
673 }
674#endif
675 while( yypParser->yyidx>=0 ) yy_pop_parser_stack(yypParser);
676 /* Here code is inserted which will be executed whenever the
677 ** parser accepts */
678%%
679 ParseARG_STORE; /* Suppress warning about unused %extra_argument variable */
680}
681
682/* The main parser program.
683** The first argument is a pointer to a structure obtained from
684** "ParseAlloc" which describes the current state of the parser.
685** The second argument is the major token number. The third is
686** the minor token. The fourth optional argument is whatever the
687** user wants (and specified in the grammar) and is available for
688** use by the action routines.
689**
690** Inputs:
691** <ul>
692** <li> A pointer to the parser (an opaque structure.)
693** <li> The major token number.
694** <li> The minor token number.
695** <li> An option argument of a grammar-specified type.
696** </ul>
697**
698** Outputs:
699** None.
700*/
701void Parse(
702 void *yyp, /* The parser */
703 int yymajor, /* The major token code number */
704 ParseTOKENTYPE yyminor /* The value for the token */
705 ParseARG_PDECL /* Optional %extra_argument parameter */
706){
707 YYMINORTYPE yyminorunion;
708 int yyact; /* The parser action. */
709 int yyendofinput; /* True if we are at the end of input */
710#ifdef YYERRORSYMBOL
711 int yyerrorhit = 0; /* True if yymajor has invoked an error */
712#endif
713 yyParser *yypParser; /* The parser */
714
715 /* (re)initialize the parser, if necessary */
716 yypParser = (yyParser*)yyp;
717 if( yypParser->yyidx<0 ){
718#if YYSTACKDEPTH<=0
719 if( yypParser->yystksz <=0 ){
720 /*memset(&yyminorunion, 0, sizeof(yyminorunion));*/
721 yyminorunion = yyzerominor;
722 yyStackOverflow(yypParser, &yyminorunion);
723 return;
724 }
725#endif
726 yypParser->yyidx = 0;
727 yypParser->yyerrcnt = -1;
728 yypParser->yystack[0].stateno = 0;
729 yypParser->yystack[0].major = 0;
730 }
731 yyminorunion.yy0 = yyminor;
732 yyendofinput = (yymajor==0);
733 ParseARG_STORE;
734
735#ifndef NDEBUG
736 if( yyTraceFILE ){
737 fprintf(yyTraceFILE,"%sInput %s\n",yyTracePrompt,yyTokenName[yymajor]);
738 }
739#endif
740
741 do{
742 yyact = yy_find_shift_action(yypParser,(YYCODETYPE)yymajor);
743 if( yyact<YYNSTATE ){
744 assert( !yyendofinput ); /* Impossible to shift the $ token */
745 yy_shift(yypParser,yyact,yymajor,&yyminorunion);
746 yypParser->yyerrcnt--;
747 yymajor = YYNOCODE;
748 }else if( yyact < YYNSTATE + YYNRULE ){
749 yy_reduce(yypParser,yyact-YYNSTATE);
750 }else{
751 assert( yyact == YY_ERROR_ACTION );
752#ifdef YYERRORSYMBOL
753 int yymx;
754#endif
755#ifndef NDEBUG
756 if( yyTraceFILE ){
757 fprintf(yyTraceFILE,"%sSyntax Error!\n",yyTracePrompt);
758 }
759#endif
760#ifdef YYERRORSYMBOL
761 /* A syntax error has occurred.
762 ** The response to an error depends upon whether or not the
763 ** grammar defines an error token "ERROR".
764 **
765 ** This is what we do if the grammar does define ERROR:
766 **
767 ** * Call the %syntax_error function.
768 **
769 ** * Begin popping the stack until we enter a state where
770 ** it is legal to shift the error symbol, then shift
771 ** the error symbol.
772 **
773 ** * Set the error count to three.
774 **
775 ** * Begin accepting and shifting new tokens. No new error
776 ** processing will occur until three tokens have been
777 ** shifted successfully.
778 **
779 */
780 if( yypParser->yyerrcnt<0 ){
781 yy_syntax_error(yypParser,yymajor,yyminorunion);
782 }
783 yymx = yypParser->yystack[yypParser->yyidx].major;
784 if( yymx==YYERRORSYMBOL || yyerrorhit ){
785#ifndef NDEBUG
786 if( yyTraceFILE ){
787 fprintf(yyTraceFILE,"%sDiscard input token %s\n",
788 yyTracePrompt,yyTokenName[yymajor]);
789 }
790#endif
791 yy_destructor(yypParser, (YYCODETYPE)yymajor,&yyminorunion);
792 yymajor = YYNOCODE;
793 }else{
794 while(
795 yypParser->yyidx >= 0 &&
796 yymx != YYERRORSYMBOL &&
797 (yyact = yy_find_reduce_action(
798 yypParser->yystack[yypParser->yyidx].stateno,
799 YYERRORSYMBOL)) >= YYNSTATE
800 ){
801 yy_pop_parser_stack(yypParser);
802 }
803 if( yypParser->yyidx < 0 || yymajor==0 ){
804 yy_destructor(yypParser,(YYCODETYPE)yymajor,&yyminorunion);
805 yy_parse_failed(yypParser);
806 yymajor = YYNOCODE;
807 }else if( yymx!=YYERRORSYMBOL ){
808 YYMINORTYPE u2;
809 u2.YYERRSYMDT = 0;
810 yy_shift(yypParser,yyact,YYERRORSYMBOL,&u2);
811 }
812 }
813 yypParser->yyerrcnt = 3;
814 yyerrorhit = 1;
815#elif defined(YYNOERRORRECOVERY)
816 /* If the YYNOERRORRECOVERY macro is defined, then do not attempt to
817 ** do any kind of error recovery. Instead, simply invoke the syntax
818 ** error routine and continue going as if nothing had happened.
819 **
820 ** Applications can set this macro (for example inside %include) if
821 ** they intend to abandon the parse upon the first syntax error seen.
822 */
823 yy_syntax_error(yypParser,yymajor,yyminorunion);
824 yy_destructor(yypParser,(YYCODETYPE)yymajor,&yyminorunion);
825 yymajor = YYNOCODE;
826
827#else /* YYERRORSYMBOL is not defined */
828 /* This is what we do if the grammar does not define ERROR:
829 **
830 ** * Report an error message, and throw away the input token.
831 **
832 ** * If the input token is $, then fail the parse.
833 **
834 ** As before, subsequent error messages are suppressed until
835 ** three input tokens have been successfully shifted.
836 */
837 if( yypParser->yyerrcnt<=0 ){
838 yy_syntax_error(yypParser,yymajor,yyminorunion);
839 }
840 yypParser->yyerrcnt = 3;
841 yy_destructor(yypParser,(YYCODETYPE)yymajor,&yyminorunion);
842 if( yyendofinput ){
843 yy_parse_failed(yypParser);
844 }
845 yymajor = YYNOCODE;
846#endif
847 }
848 }while( yymajor!=YYNOCODE && yypParser->yyidx>=0 );
849 return;
850}