Tcl Source Code

Check-in [497b93405b]
Login

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview
Comment:[1080042][8f245009b0] Big bundle of regexp engine fixes and improvements contributed from Tom Lane of the postgres project.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1: 497b93405b3435aa98b1a1bc243b0b70e5f89192
User & Date: dgp 2015-10-21 14:10:13
Context
2016-04-19
20:35
Fork of Tcl used in the "Little" project. http://www.mcvoy.com/lm/little/index.html check-in: 69b737f5a1 user: dgp tags: little
2015-10-23
08:13
Merge trunk check-in: d68de1f600 user: jan.nijtmans tags: androwish
08:12
Change "clock scan/format -format %x -locale current" output on msgcat locale change. Bug [4a0c163d2... check-in: 5855bdf4a8 user: oehhar tags: trunk
2015-10-21
23:30
Micro-optimization: remove double checked lock from TclGetAllocCache in favour of initialization in ... check-in: fdbf64dc50 user: kbk tags: drh-micro-optimization
20:30
Change "clock format -format %x -locale current" output on msgcat locale change [4a0c163d24] check-in: e5bf4e6084 user: oehhar tags: bug-4a0c163d24
17:02
Merge updates from trunk. check-in: 6e776b8e33 user: mistachkin tags: tip-435, bug-57945b574a
14:10
[1080042][8f245009b0] Big bundle of regexp engine fixes and improvements contributed from Tom Lane o... check-in: 497b93405b user: dgp tags: trunk
13:50
[1080042][8f245009b0] Big bundle of regexp engine fixes and improvements contributed from Tom Lane o... check-in: 01bd72a637 user: dgp tags: core-8-5-branch
2015-10-19
14:07
[154f0982f2] Document that Tcl_NewObjectInstance() really needs to make a namespace. check-in: 2327af6915 user: dkf tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to generic/regc_nfa.c.

30
31
32
33
34
35
36



37
38
39
40
41
42
43
 *
 * One or two things that technically ought to be in here are actually in
 * color.c, thanks to some incestuous relationships in the color chains.
 */

#define	NISERR()	VISERR(nfa->v)
#define	NERR(e)		VERR(nfa->v, (e))




/*
 - newnfa - set up an NFA
 ^ static struct nfa *newnfa(struct vars *, struct colormap *, struct nfa *);
 */
static struct nfa *		/* the NFA, or NULL */
newnfa(







>
>
>







30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
 *
 * One or two things that technically ought to be in here are actually in
 * color.c, thanks to some incestuous relationships in the color chains.
 */

#define	NISERR()	VISERR(nfa->v)
#define	NERR(e)		VERR(nfa->v, (e))
#define STACK_TOO_DEEP(x) (0)
#define CANCEL_REQUESTED(x) (0)
#define REG_CANCEL 777

/*
 - newnfa - set up an NFA
 ^ static struct nfa *newnfa(struct vars *, struct colormap *, struct nfa *);
 */
static struct nfa *		/* the NFA, or NULL */
newnfa(
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

    nfa->states = NULL;
    nfa->slast = NULL;
    nfa->free = NULL;
    nfa->nstates = 0;
    nfa->cm = cm;
    nfa->v = v;
    nfa->size = 0;
    nfa->bos[0] = nfa->bos[1] = COLORLESS;
    nfa->eos[0] = nfa->eos[1] = COLORLESS;
    nfa->parent = parent;	/* Precedes newfstate so parent is valid. */
    nfa->post = newfstate(nfa, '@');	/* number 0 */
    nfa->pre = newfstate(nfa, '>');	/* number 1 */

    nfa->init = newstate(nfa);	/* May become invalid later. */







<







58
59
60
61
62
63
64

65
66
67
68
69
70
71

    nfa->states = NULL;
    nfa->slast = NULL;
    nfa->free = NULL;
    nfa->nstates = 0;
    nfa->cm = cm;
    nfa->v = v;

    nfa->bos[0] = nfa->bos[1] = COLORLESS;
    nfa->eos[0] = nfa->eos[1] = COLORLESS;
    nfa->parent = parent;	/* Precedes newfstate so parent is valid. */
    nfa->post = newfstate(nfa, '@');	/* number 0 */
    nfa->pre = newfstate(nfa, '>');	/* number 1 */

    nfa->init = newstate(nfa);	/* May become invalid later. */
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149

    if (ISERR()) {
	freenfa(nfa);
	return NULL;
    }
    return nfa;
}

/*
 - TooManyStates - checks if the max states exceeds the compile-time value
 ^ static int TooManyStates(struct nfa *);
 */
static int
TooManyStates(
    struct nfa *nfa)
{
    struct nfa *parent = nfa->parent;
    size_t sz = nfa->size;

    while (parent != NULL) {
	sz = parent->size;
	parent = parent->parent;
    }
    if (sz > REG_MAX_STATES) {
	return 1;
    }
    return 0;
}

/*
 - IncrementSize - increases the tracked size of the NFA and its parents.
 ^ static void IncrementSize(struct nfa *);
 */
static void
IncrementSize(
    struct nfa *nfa)
{
    struct nfa *parent = nfa->parent;

    nfa->size++;
    while (parent != NULL) {
	parent->size++;
	parent = parent->parent;
    }
}

/*
 - DecrementSize - increases the tracked size of the NFA and its parents.
 ^ static void DecrementSize(struct nfa *);
 */
static void
DecrementSize(
    struct nfa *nfa)
{
    struct nfa *parent = nfa->parent;

    nfa->size--;
    while (parent != NULL) {
	parent->size--;
	parent = parent->parent;
    }
}

/*
 - freenfa - free an entire NFA
 ^ static void freenfa(struct nfa *);
 */
static void
freenfa(







<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<







83
84
85
86
87
88
89























































90
91
92
93
94
95
96

    if (ISERR()) {
	freenfa(nfa);
	return NULL;
    }
    return nfa;
}
























































/*
 - freenfa - free an entire NFA
 ^ static void freenfa(struct nfa *);
 */
static void
freenfa(
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188




189
190
191
192
193

194
195
196
197
198
199
200
 */
static struct state *		/* NULL on error */
newstate(
    struct nfa *nfa)
{
    struct state *s;

    if (TooManyStates(nfa)) {
	/* XXX: add specific error for this */
	NERR(REG_ETOOBIG);
	return NULL;
    }
    if (nfa->free != NULL) {
	s = nfa->free;
	nfa->free = s->next;
    } else {




	s = (struct state *) MALLOC(sizeof(struct state));
	if (s == NULL) {
	    NERR(REG_ESPACE);
	    return NULL;
	}

	s->oas.next = NULL;
	s->free = NULL;
	s->noas = 0;
    }

    assert(nfa->nstates >= 0);
    s->no = nfa->nstates++;







<
<
<
<
<




>
>
>
>





>







120
121
122
123
124
125
126





127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
 */
static struct state *		/* NULL on error */
newstate(
    struct nfa *nfa)
{
    struct state *s;






    if (nfa->free != NULL) {
	s = nfa->free;
	nfa->free = s->next;
    } else {
	if (nfa->v->spaceused >= REG_MAX_COMPILE_SPACE) {
	    NERR(REG_ETOOBIG);
	    return NULL;
	}
	s = (struct state *) MALLOC(sizeof(struct state));
	if (s == NULL) {
	    NERR(REG_ESPACE);
	    return NULL;
	}
	nfa->v->spaceused += sizeof(struct state);
	s->oas.next = NULL;
	s->free = NULL;
	s->noas = 0;
    }

    assert(nfa->nstates >= 0);
    s->no = nfa->nstates++;
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
    s->next = NULL;
    if (nfa->slast != NULL) {
	assert(nfa->slast->next == NULL);
	nfa->slast->next = s;
    }
    s->prev = nfa->slast;
    nfa->slast = s;

    /*
     * Track the current size and the parent size.
     */

    IncrementSize(nfa);
    return s;
}

/*
 - newfstate - allocate an NFA state with a specified flag value
 ^ static struct state *newfstate(struct nfa *, int flag);
 */







<
<
<
<
<
<







157
158
159
160
161
162
163






164
165
166
167
168
169
170
    s->next = NULL;
    if (nfa->slast != NULL) {
	assert(nfa->slast->next == NULL);
	nfa->slast->next = s;
    }
    s->prev = nfa->slast;
    nfa->slast = s;






    return s;
}

/*
 - newfstate - allocate an NFA state with a specified flag value
 ^ static struct state *newfstate(struct nfa *, int flag);
 */
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311

312
313
314
315
316

317
318
319
320
321
322
323




324
325
326
327
328
329
330
331
332
333
334
335
336
337
338




339


340
341
342
343
344
345






















346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364




365
366




367
368
369
370
371
372
373
    } else {
	assert(s == nfa->states);
	nfa->states = s->next;
    }
    s->prev = NULL;
    s->next = nfa->free;	/* don't delete it, put it on the free list */
    nfa->free = s;
    DecrementSize(nfa);
}

/*
 - destroystate - really get rid of an already-freed state
 ^ static void destroystate(struct nfa *, struct state *);
 */
static void
destroystate(
    struct nfa *nfa,
    struct state *s)
{
    struct arcbatch *ab;
    struct arcbatch *abnext;

    assert(s->no == FREESTATE);
    for (ab=s->oas.next ; ab!=NULL ; ab=abnext) {
	abnext = ab->next;
	FREE(ab);

    }
    s->ins = NULL;
    s->outs = NULL;
    s->next = NULL;
    FREE(s);

}

/*
 - newarc - set up a new arc within an NFA
 ^ static void newarc(struct nfa *, int, pcolor, struct state *,
 ^	struct state *);
 */




static void
newarc(
    struct nfa *nfa,
    int t,
    pcolor co,
    struct state *from,
    struct state *to)
{
    struct arc *a;

    assert(from != NULL && to != NULL);

    /*
     * Check for duplicates.
     */







    for (a=from->outs ; a!=NULL ; a=a->outchain) {
	if (a->to == to && a->co == co && a->type == t) {
	    return;
	}
    }























    a = allocarc(nfa, from);
    if (NISERR()) {
	return;
    }
    assert(a != NULL);

    a->type = t;
    a->co = (color) co;
    a->to = to;
    a->from = from;

    /*
     * Put the new arc on the beginning, not the end, of the chains. Not only
     * is this easier, it has the very useful side effect that deleting the
     * most-recently-added arc is the cheapest case rather than the most
     * expensive one.
     */

    a->inchain = to->ins;




    to->ins = a;
    a->outchain = from->outs;




    from->outs = a;

    from->nouts++;
    to->nins++;

    if (COLORED(a) && nfa->parent == NULL) {
	colorchain(nfa->cm, a);







<


















>





>







>
>
>
>












<
|
<
>
>
>
>
|
>
>
|
|
|
|
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>












|
|
|
<

<

>
>
>
>


>
>
>
>







227
228
229
230
231
232
233

234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281

282

283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332

333

334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
    } else {
	assert(s == nfa->states);
	nfa->states = s->next;
    }
    s->prev = NULL;
    s->next = nfa->free;	/* don't delete it, put it on the free list */
    nfa->free = s;

}

/*
 - destroystate - really get rid of an already-freed state
 ^ static void destroystate(struct nfa *, struct state *);
 */
static void
destroystate(
    struct nfa *nfa,
    struct state *s)
{
    struct arcbatch *ab;
    struct arcbatch *abnext;

    assert(s->no == FREESTATE);
    for (ab=s->oas.next ; ab!=NULL ; ab=abnext) {
	abnext = ab->next;
	FREE(ab);
	nfa->v->spaceused -= sizeof(struct arcbatch);
    }
    s->ins = NULL;
    s->outs = NULL;
    s->next = NULL;
    FREE(s);
    nfa->v->spaceused -= sizeof(struct state);
}

/*
 - newarc - set up a new arc within an NFA
 ^ static void newarc(struct nfa *, int, pcolor, struct state *,
 ^	struct state *);
 */
/*
 * This function checks to make sure that no duplicate arcs are created.
 * In general we never want duplicates.
 */
static void
newarc(
    struct nfa *nfa,
    int t,
    pcolor co,
    struct state *from,
    struct state *to)
{
    struct arc *a;

    assert(from != NULL && to != NULL);


    /* check for duplicate arc, using whichever chain is shorter */

    if (from->nouts <= to->nins) {
	for (a = from->outs; a != NULL; a = a->outchain) {
	    if (a->to == to && a->co == co && a->type == t) {
		return;
	    }
	}
    } else {
	for (a = to->ins; a != NULL; a = a->inchain) {
	    if (a->from == from && a->co == co && a->type == t) {
		return;
	    }
	}
    }
  
    /* no dup, so create the arc */
    createarc(nfa, t, co, from, to);
}

/*
 * createarc - create a new arc within an NFA
 *
 * This function must *only* be used after verifying that there is no existing
 * identical arc (same type/color/from/to).
 */
static void
createarc(
    struct nfa * nfa,
    int t,
    pcolor co,
    struct state * from,
    struct state * to)
{
    struct arc *a;

    /* the arc is physically allocated within its from-state */
    a = allocarc(nfa, from);
    if (NISERR()) {
	return;
    }
    assert(a != NULL);

    a->type = t;
    a->co = (color) co;
    a->to = to;
    a->from = from;

    /*
     * Put the new arc on the beginning, not the end, of the chains; it's
     * simpler here, and freearc() is the same cost either way.  See also the
     * logic in moveins() and its cohorts, as well as fixempties().

     */

    a->inchain = to->ins;
    a->inchainRev = NULL;
    if (to->ins) {
	to->ins->inchainRev = a;
    }
    to->ins = a;
    a->outchain = from->outs;
    a->outchainRev = NULL;
    if (from->outs) {
	from->outs->outchainRev = a;
    }
    from->outs = a;

    from->nouts++;
    to->nins++;

    if (COLORED(a) && nfa->parent == NULL) {
	colorchain(nfa->cm, a);
396
397
398
399
400
401
402
403
404
405
406





407
408
409
410

411
412
413
414
415
416
417
    }

    /*
     * if none at hand, get more
     */

    if (s->free == NULL) {
	struct arcbatch *newAb = (struct arcbatch *)
		MALLOC(sizeof(struct arcbatch));
	int i;






	if (newAb == NULL) {
	    NERR(REG_ESPACE);
	    return NULL;
	}

	newAb->next = s->oas.next;
	s->oas.next = newAb;

	for (i=0 ; i<ABSIZE ; i++) {
	    newAb->a[i].type = 0;
	    newAb->a[i].freechain = &newAb->a[i+1];
	}







|
<


>
>
>
>
>




>







374
375
376
377
378
379
380
381

382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
    }

    /*
     * if none at hand, get more
     */

    if (s->free == NULL) {
	struct arcbatch *newAb;

	int i;

	if (nfa->v->spaceused >= REG_MAX_COMPILE_SPACE) {
	    NERR(REG_ETOOBIG);
	    return NULL;
	}
	newAb = (struct arcbatch *) MALLOC(sizeof(struct arcbatch));
	if (newAb == NULL) {
	    NERR(REG_ESPACE);
	    return NULL;
	}
	nfa->v->spaceused += sizeof(struct arcbatch);
	newAb->next = s->oas.next;
	s->oas.next = newAb;

	for (i=0 ; i<ABSIZE ; i++) {
	    newAb->a[i].type = 0;
	    newAb->a[i].freechain = &newAb->a[i+1];
	}
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455


456
457
458
459
460
461
462
463
464

465
466
467
468
469
470
471
472
473


474
475
476
477
478
479
480
481
482
483


484
485
486
487
488
489
490
491
492
493
494

495

496
497
498












































499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
static void
freearc(
    struct nfa *nfa,
    struct arc *victim)
{
    struct state *from = victim->from;
    struct state *to = victim->to;
    struct arc *a;

    assert(victim->type != 0);

    /*
     * Take it off color chain if necessary.
     */

    if (COLORED(victim) && nfa->parent == NULL) {
	uncolorchain(nfa->cm, victim);
    }

    /*
     * Take it off source's out-chain.
     */

    assert(from != NULL);


    assert(from->outs != NULL);
    a = from->outs;
    if (a == victim) {		/* simple case: first in chain */
	from->outs = victim->outchain;
    } else {
	for (; a!=NULL && a->outchain!=victim ; a=a->outchain) {
	    continue;
	}
	assert(a != NULL);

	a->outchain = victim->outchain;
    }
    from->nouts--;

    /*
     * Take it off target's in-chain.
     */

    assert(to != NULL);


    assert(to->ins != NULL);
    a = to->ins;
    if (a == victim) {		/* simple case: first in chain */
	to->ins = victim->inchain;
    } else {
	for (; a->inchain!=victim ; a=a->inchain) {
	    assert(a->inchain != NULL);
	    continue;
	}
	a->inchain = victim->inchain;


    }
    to->nins--;

    /*
     * Clean up and place on free list.
     */

    victim->type = 0;
    victim->from = NULL;	/* precautions... */
    victim->to = NULL;
    victim->inchain = NULL;

    victim->outchain = NULL;

    victim->freechain = from->free;
    from->free = victim;
}













































/*
 - hasnonemptyout - Does state have a non-EMPTY out arc?
 ^ static int hasnonemptyout(struct state *);
 */
static int
hasnonemptyout(
    struct state *s)
{
    struct arc *a;

    for (a = s->outs; a != NULL; a = a->outchain) {
	if (a->type != EMPTY) {
	    return 1;
	}
    }
    return 0;
}

/*
 - nonemptyouts - count non-EMPTY out arcs of a state
 ^ static int nonemptyouts(struct state *);
 */
static int
nonemptyouts(
    struct state *s)
{
    int n = 0;
    struct arc *a;

    for (a = s->outs; a != NULL; a = a->outchain) {
	if (a->type != EMPTY) {
	    n++;
	}
    }
    return n;
}

/*
 - nonemptyins - count non-EMPTY in arcs of a state
 ^ static int nonemptyins(struct state *);
 */
static int
nonemptyins(
    struct state *s)
{
    int n = 0;
    struct arc *a;

    for (a = s->ins; a != NULL; a = a->inchain) {
	if (a->type != EMPTY) {
	    n++;
	}
    }
    return n;
}

/*
 - findarc - find arc, if any, from given source with given type and color
 * If there is more than one such arc, the result is random.
 ^ static struct arc *findarc(struct state *, int, pcolor);
 */
static struct arc *







|
















>
>
|
<
<


|
|
|
|
>
|








>
>
|
<
<


|
|
<
|
|
>
>




|






>

>



>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>


















<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<







415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441


442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460


461
462
463
464

465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547






































548
549
550
551
552
553
554
static void
freearc(
    struct nfa *nfa,
    struct arc *victim)
{
    struct state *from = victim->from;
    struct state *to = victim->to;
    struct arc *predecessor;

    assert(victim->type != 0);

    /*
     * Take it off color chain if necessary.
     */

    if (COLORED(victim) && nfa->parent == NULL) {
	uncolorchain(nfa->cm, victim);
    }

    /*
     * Take it off source's out-chain.
     */

    assert(from != NULL);
    predecessor = victim->outchainRev;
    if (predecessor == NULL) {
	assert(from->outs == victim);


	from->outs = victim->outchain;
    } else {
	assert(predecessor->outchain == victim);
	predecessor->outchain = victim->outchain;
    }
    if (victim->outchain != NULL) {
	assert(victim->outchain->outchainRev == victim);
	victim->outchain->outchainRev = predecessor;
    }
    from->nouts--;

    /*
     * Take it off target's in-chain.
     */

    assert(to != NULL);
    predecessor = victim->inchainRev;
    if (predecessor == NULL) {
	assert(to->ins == victim);


	to->ins = victim->inchain;
    } else {
	assert(predecessor->inchain == victim);
	predecessor->inchain = victim->inchain;

    }
    if (victim->inchain != NULL) {
	assert(victim->inchain->inchainRev == victim);
	victim->inchain->inchainRev = predecessor;
    }
    to->nins--;

    /*
     * Clean up and place on from-state's free list.
     */

    victim->type = 0;
    victim->from = NULL;	/* precautions... */
    victim->to = NULL;
    victim->inchain = NULL;
    victim->inchainRev = NULL;
    victim->outchain = NULL;
    victim->outchainRev = NULL;
    victim->freechain = from->free;
    from->free = victim;
}

/*
 * changearctarget - flip an arc to have a different to state
 *
 * Caller must have verified that there is no pre-existing duplicate arc.
 *
 * Note that because we store arcs in their from state, we can't easily have
 * a similar changearcsource function.
 */
static void
changearctarget(struct arc * a, struct state * newto)
{
    struct state *oldto = a->to;
    struct arc *predecessor;

    assert(oldto != newto);

    /* take it off old target's in-chain */
    assert(oldto != NULL);
    predecessor = a->inchainRev;
    if (predecessor == NULL) {
	assert(oldto->ins == a);
	oldto->ins = a->inchain;
    } else {
	assert(predecessor->inchain == a);
	predecessor->inchain = a->inchain;
    }
    if (a->inchain != NULL) {
	assert(a->inchain->inchainRev == a);
	a->inchain->inchainRev = predecessor;
    }
    oldto->nins--;

    a->to = newto;

    /* prepend it to new target's in-chain */
    a->inchain = newto->ins;
    a->inchainRev = NULL;
    if (newto->ins) {
	newto->ins->inchainRev = a;
    }
    newto->ins = a;
    newto->nins++;
}

/*
 - hasnonemptyout - Does state have a non-EMPTY out arc?
 ^ static int hasnonemptyout(struct state *);
 */
static int
hasnonemptyout(
    struct state *s)
{
    struct arc *a;

    for (a = s->outs; a != NULL; a = a->outchain) {
	if (a->type != EMPTY) {
	    return 1;
	}
    }
    return 0;
}







































/*
 - findarc - find arc, if any, from given source with given type and color
 * If there is more than one such arc, the result is random.
 ^ static struct arc *findarc(struct state *, int, pcolor);
 */
static struct arc *
584
585
586
587
588
589
590
591
592



































































































































































593
594
595
596
597





598
599
600
601
602
603
604
605




606
607























608


609
610













611






612







613










614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634


635

636
637























































































































































638
639
640
641
642
643
644
645
646
647
648
649
650




651
652























653


654
655









656






657







658













659
660
661
662
663
664
665
666
667
668
669
670


671



672









673

674










675


676
677



678




679












680








681
682
683
684
685
686
687
    struct nfa *nfa,
    struct arc *oa,
    struct state *from,
    struct state *to)
{
    newarc(nfa, oa->type, oa->co, from, to);
}

/*



































































































































































 - moveins - move all in arcs of a state to another state
 * You might think this could be done better by just updating the
 * existing arcs, and you would be right if it weren't for the desire
 * for duplicate suppression, which makes it easier to just make new
 * ones to exploit the suppression built into newarc.





 ^ static void moveins(struct nfa *, struct state *, struct state *);
 */
static void
moveins(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState)
{




    struct arc *a;
























    assert(oldState != newState);



    while ((a = oldState->ins) != NULL) {













	cparc(nfa, a, a->from, newState);






	freearc(nfa, a);







    }










    assert(oldState->nins == 0);
    assert(oldState->ins == NULL);
}

/*
 - copyins - copy in arcs of a state to another state
 * Either all arcs, or only non-empty ones as determined by all value.
 ^ static VOID copyins(struct nfa *, struct state *, struct state *, int);
 */
static void
copyins(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState,
    int all)
{
    struct arc *a;

    assert(oldState != newState);

    for (a=oldState->ins ; a!=NULL ; a=a->inchain) {


	if (all || a->type != EMPTY) {

	    cparc(nfa, a, a->from, newState);
	}























































































































































    }
}

/*
 - moveouts - move all out arcs of a state to another state
 ^ static void moveouts(struct nfa *, struct state *, struct state *);
 */
static void
moveouts(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState)
{




    struct arc *a;
























    assert(oldState != newState);



    while ((a = oldState->outs) != NULL) {









	cparc(nfa, a, newState, a->to);






	freearc(nfa, a);







    }













}

/*
 - copyouts - copy out arcs of a state to another state
 * Either all arcs, or only non-empty ones as determined by all value.
 ^ static VOID copyouts(struct nfa *, struct state *, struct state *, int);
 */
static void
copyouts(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState,


    int all)



{









    struct arc *a;












    assert(oldState != newState);



    for (a=oldState->outs ; a!=NULL ; a=a->outchain) {



	if (all || a->type != EMPTY) {




	    cparc(nfa, a, newState, a->to);












	}








    }
}

/*
 - cloneouts - copy out arcs of a state to another state pair, modifying type
 ^ static void cloneouts(struct nfa *, struct state *, struct state *,
 ^ 	struct state *, int);







|

>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>


|


>
>
>
>
>








>
>
>
>
|

>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
|
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>






<






|
<

<
<


|
>
>
|
>


>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>













>
>
>
>
|

>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
|
|
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
|
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>




<






|
>
>
|
>
>
>
|
>
>
>
>
>
>
>
>
>
|
>

>
>
>
>
>
>
>
>
>
>
|
>
>
|
|
>
>
>
|
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>







577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845

846
847
848
849
850
851
852

853


854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102

1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
    struct nfa *nfa,
    struct arc *oa,
    struct state *from,
    struct state *to)
{
    newarc(nfa, oa->type, oa->co, from, to);
}

/*
 * sortins - sort the in arcs of a state by from/color/type
 */
static void
sortins(
    struct nfa * nfa,
    struct state * s)
{
    struct arc **sortarray;
    struct arc *a;
    int n = s->nins;
    int i;

    if (n <= 1) {
	return;		/* nothing to do */
    }
    /* make an array of arc pointers ... */
    sortarray = (struct arc **) MALLOC(n * sizeof(struct arc *));
    if (sortarray == NULL) {
	NERR(REG_ESPACE);
	return;
    }
    i = 0;
    for (a = s->ins; a != NULL; a = a->inchain) {
	sortarray[i++] = a;
    }
    assert(i == n);
    /* ... sort the array */
    qsort(sortarray, n, sizeof(struct arc *), sortins_cmp);
    /* ... and rebuild arc list in order */
    /* it seems worth special-casing first and last items to simplify loop */
    a = sortarray[0];
    s->ins = a;
    a->inchain = sortarray[1];
    a->inchainRev = NULL;
    for (i = 1; i < n - 1; i++) {
	a = sortarray[i];
	a->inchain = sortarray[i + 1];
	a->inchainRev = sortarray[i - 1];
    }
    a = sortarray[i];
    a->inchain = NULL;
    a->inchainRev = sortarray[i - 1];
    FREE(sortarray);
}

static int
sortins_cmp(
    const void *a,
    const void *b)
{
    const struct arc *aa = *((const struct arc * const *) a);
    const struct arc *bb = *((const struct arc * const *) b);

    /* we check the fields in the order they are most likely to be different */
    if (aa->from->no < bb->from->no) {
	return -1;
    }
    if (aa->from->no > bb->from->no) {
 	return 1;
    }
    if (aa->co < bb->co) {
 	return -1;
    }
    if (aa->co > bb->co) {
 	return 1;
    }
    if (aa->type < bb->type) {
 	return -1;
    }
    if (aa->type > bb->type) {
 	return 1;
    }
    return 0;
}
 
/*
 * sortouts - sort the out arcs of a state by to/color/type
 */
static void
sortouts(
    struct nfa * nfa,
    struct state * s)
{
    struct arc **sortarray;
    struct arc *a;
    int	n = s->nouts;
    int	i;

    if (n <= 1) {
	return;					/* nothing to do */
    }
    /* make an array of arc pointers ... */
    sortarray = (struct arc **) MALLOC(n * sizeof(struct arc *));
    if (sortarray == NULL) {
	NERR(REG_ESPACE);
	return;
    }
    i = 0;
    for (a = s->outs; a != NULL; a = a->outchain) {
	sortarray[i++] = a;
    }
    assert(i == n);
    /* ... sort the array */
    qsort(sortarray, n, sizeof(struct arc *), sortouts_cmp);
    /* ... and rebuild arc list in order */
    /* it seems worth special-casing first and last items to simplify loop */
    a = sortarray[0];
    s->outs = a;
    a->outchain = sortarray[1];
    a->outchainRev = NULL;
    for (i = 1; i < n - 1; i++) {
	a = sortarray[i];
	a->outchain = sortarray[i + 1];
	a->outchainRev = sortarray[i - 1];
    }
    a = sortarray[i];
    a->outchain = NULL;
    a->outchainRev = sortarray[i - 1];
    FREE(sortarray);
}

static int
sortouts_cmp(
    const void *a,
    const void *b)
{
    const struct arc *aa = *((const struct arc * const *) a);
    const struct arc *bb = *((const struct arc * const *) b);

    /* we check the fields in the order they are most likely to be different */
    if (aa->to->no < bb->to->no) {
	return -1;
    }
    if (aa->to->no > bb->to->no) {
	return 1;
    }
    if (aa->co < bb->co) {
	return -1;
    }
    if (aa->co > bb->co) {
	return 1;
    }
    if (aa->type < bb->type) {
	return -1;
    }
    if (aa->type > bb->type) {
	return 1;
    }
    return 0;
}

/*
 * Common decision logic about whether to use arc-by-arc operations or
 * sort/merge.  If there's just a few source arcs we cannot recoup the
 * cost of sorting the destination arc list, no matter how large it is.
 * Otherwise, limit the number of arc-by-arc comparisons to about 1000
 * (a somewhat arbitrary choice, but the breakeven point would probably
 * be machine dependent anyway).
 */
#define BULK_ARC_OP_USE_SORT(nsrcarcs, ndestarcs) \
	((nsrcarcs) < 4 ? 0 : ((nsrcarcs) > 32 || (ndestarcs) > 32))

/*
 - moveins - move all in arcs of a state to another state
 * You might think this could be done better by just updating the
 * existing arcs, and you would be right if it weren't for the need
 * for duplicate suppression, which makes it easier to just make new
 * ones to exploit the suppression built into newarc.
 *
 * However, if we have a whole lot of arcs to deal with, retail duplicate
 * checks become too slow.  In that case we proceed by sorting and merging
 * the arc lists, and then we can indeed just update the arcs in-place.
 *
 ^ static void moveins(struct nfa *, struct state *, struct state *);
 */
static void
moveins(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState)
{
    assert(oldState != newState);

    if (!BULK_ARC_OP_USE_SORT(oldState->nins, newState->nins)) {
	/* With not too many arcs, just do them one at a time */
	struct arc *a;

	while ((a = oldState->ins) != NULL) {
	    cparc(nfa, a, a->from, newState);
	    freearc(nfa, a);
	}
    } else {
	/*
	 * With many arcs, use a sort-merge approach.  Note changearctarget()
	 * will put the arc onto the front of newState's chain, so it does not
	 * break our walk through the sorted part of the chain.
	 */
	struct arc *oa;
	struct arc *na;

	/*
	 * Because we bypass newarc() in this code path, we'd better include a
	 * cancel check.
	 */
	if (CANCEL_REQUESTED(nfa->v->re)) {
	    NERR(REG_CANCEL);
	    return;
	}

	sortins(nfa, oldState);
	sortins(nfa, newState);
	if (NISERR()) {
	    return;		/* might have failed to sort */
	}
	oa = oldState->ins;
	na = newState->ins;
	while (oa != NULL && na != NULL) {
	    struct arc *a = oa;

	    switch (sortins_cmp(&oa, &na)) {
		case -1:
		    /* newState does not have anything matching oa */
		    oa = oa->inchain;

		    /*
		     * Rather than doing createarc+freearc, we can just unlink
		     * and relink the existing arc struct.
		     */
		    changearctarget(a, newState);
		    break;
		case 0:
		    /* match, advance in both lists */
		    oa = oa->inchain;
		    na = na->inchain;
		    /* ... and drop duplicate arc from oldState */
		    freearc(nfa, a);
		    break;
		case +1:
		    /* advance only na; oa might have a match later */
		    na = na->inchain;
		    break;
		default:
		    assert(NOTREACHED);
	    }
	}
	while (oa != NULL) {
	    /* newState does not have anything matching oa */
	    struct arc *a = oa;

	    oa = oa->inchain;
	    changearctarget(a, newState);
	}
    }

    assert(oldState->nins == 0);
    assert(oldState->ins == NULL);
}

/*
 - copyins - copy in arcs of a state to another state

 ^ static VOID copyins(struct nfa *, struct state *, struct state *, int);
 */
static void
copyins(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState)

{


    assert(oldState != newState);

    if (!BULK_ARC_OP_USE_SORT(oldState->nins, newState->nins)) {
	/* With not too many arcs, just do them one at a time */
	struct arc *a;

	for (a = oldState->ins; a != NULL; a = a->inchain) {
	    cparc(nfa, a, a->from, newState);
	}
    } else {
	/*
	 * With many arcs, use a sort-merge approach.  Note that createarc()
	 * will put new arcs onto the front of newState's chain, so it does
	 * not break our walk through the sorted part of the chain.
	 */
	struct arc *oa;
	struct arc *na;

	/*
	 * Because we bypass newarc() in this code path, we'd better include a
	 * cancel check.
	 */
	if (CANCEL_REQUESTED(nfa->v->re)) {
	    NERR(REG_CANCEL);
	    return;
	}

	sortins(nfa, oldState);
	sortins(nfa, newState);
	if (NISERR()) {
	    return;		/* might have failed to sort */
	}
	oa = oldState->ins;
	na = newState->ins;
	while (oa != NULL && na != NULL) {
	    struct arc *a = oa;

	    switch (sortins_cmp(&oa, &na)) {
		case -1:
		    /* newState does not have anything matching oa */
		    oa = oa->inchain;
		    createarc(nfa, a->type, a->co, a->from, newState);
		    break;
		case 0:
		    /* match, advance in both lists */
		    oa = oa->inchain;
		    na = na->inchain;
		    break;
		case +1:
		    /* advance only na; oa might have a match later */
		    na = na->inchain;
		    break;
		default:
		    assert(NOTREACHED);
	    }
	}
	while (oa != NULL) {
	    /* newState does not have anything matching oa */
	    struct arc *a = oa;

	    oa = oa->inchain;
	    createarc(nfa, a->type, a->co, a->from, newState);
	}
    }
}

/*
 * mergeins - merge a list of inarcs into a state
 *
 * This is much like copyins, but the source arcs are listed in an array,
 * and are not guaranteed unique.  It's okay to clobber the array contents.
 */
static void
mergeins(
    struct nfa * nfa,
    struct state * s,
    struct arc ** arcarray,
    int arccount)
{
    struct arc *na;
    int	i;
    int	j;

    if (arccount <= 0) {
	return;
    }

    /*
     * Because we bypass newarc() in this code path, we'd better include a
     * cancel check.
     */
    if (CANCEL_REQUESTED(nfa->v->re)) {
	NERR(REG_CANCEL);
	return;
    }

    /* Sort existing inarcs as well as proposed new ones */
    sortins(nfa, s);
    if (NISERR()) {
	return;			/* might have failed to sort */
    }

    qsort(arcarray, arccount, sizeof(struct arc *), sortins_cmp);

    /*
     * arcarray very likely includes dups, so we must eliminate them.  (This
     * could be folded into the next loop, but it's not worth the trouble.)
     */
    j = 0;
    for (i = 1; i < arccount; i++) {
	switch (sortins_cmp(&arcarray[j], &arcarray[i])) {
	    case -1:
		/* non-dup */
		arcarray[++j] = arcarray[i];
		break;
	    case 0:
		/* dup */
		break;
	    default:
		/* trouble */
		assert(NOTREACHED);
	}
    }
    arccount = j + 1;

    /*
     * Now merge into s' inchain.  Note that createarc() will put new arcs
     * onto the front of s's chain, so it does not break our walk through the
     * sorted part of the chain.
     */
    i = 0;
    na = s->ins;
    while (i < arccount && na != NULL) {
	struct arc *a = arcarray[i];

	switch (sortins_cmp(&a, &na)) {
	    case -1:
		/* s does not have anything matching a */
		createarc(nfa, a->type, a->co, a->from, s);
		i++;
		break;
	    case 0:
		/* match, advance in both lists */
		i++;
		na = na->inchain;
		break;
	    case +1:
		/* advance only na; array might have a match later */
		na = na->inchain;
		break;
	    default:
		assert(NOTREACHED);
	}
    }
    while (i < arccount) {
	/* s does not have anything matching a */
	struct arc *a = arcarray[i];

	createarc(nfa, a->type, a->co, a->from, s);
	i++;
    }
}

/*
 - moveouts - move all out arcs of a state to another state
 ^ static void moveouts(struct nfa *, struct state *, struct state *);
 */
static void
moveouts(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState)
{
    assert(oldState != newState);

    if (!BULK_ARC_OP_USE_SORT(oldState->nouts, newState->nouts)) {
	/* With not too many arcs, just do them one at a time */
	struct arc *a;

	while ((a = oldState->outs) != NULL) {
	    cparc(nfa, a, newState, a->to);
	    freearc(nfa, a);
	}
    } else {
	/*
	 * With many arcs, use a sort-merge approach.  Note that createarc()
	 * will put new arcs onto the front of newState's chain, so it does
	 * not break our walk through the sorted part of the chain.
	 */
	struct arc *oa;
	struct arc *na;

	/*
	 * Because we bypass newarc() in this code path, we'd better include a
	 * cancel check.
	 */
	if (CANCEL_REQUESTED(nfa->v->re)) {
	    NERR(REG_CANCEL);
	    return;
	}

	sortouts(nfa, oldState);
	sortouts(nfa, newState);
	if (NISERR()) {
	    return;	/* might have failed to sort */
	}
	oa = oldState->outs;
	na = newState->outs;
	while (oa != NULL && na != NULL) {
	    struct arc *a = oa;

	    switch (sortouts_cmp(&oa, &na)) {
		case -1:
		    /* newState does not have anything matching oa */
		    oa = oa->outchain;
		    createarc(nfa, a->type, a->co, newState, a->to);
		    freearc(nfa, a);
		    break;
		case 0:
		    /* match, advance in both lists */
		    oa = oa->outchain;
		    na = na->outchain;
		    /* ... and drop duplicate arc from oldState */
		    freearc(nfa, a);
		    break;
		case +1:
		    /* advance only na; oa might have a match later */
		    na = na->outchain;
		    break;
		default:
		    assert(NOTREACHED);
	    }
	}
	while (oa != NULL) {
	    /* newState does not have anything matching oa */
	    struct arc *a = oa;

	    oa = oa->outchain;
	    createarc(nfa, a->type, a->co, newState, a->to);
	    freearc(nfa, a);
	}
    }

    assert(oldState->nouts == 0);
    assert(oldState->outs == NULL);
}

/*
 - copyouts - copy out arcs of a state to another state

 ^ static VOID copyouts(struct nfa *, struct state *, struct state *, int);
 */
static void
copyouts(
    struct nfa *nfa,
    struct state *oldState,
    struct state *newState)
{
    assert(oldState != newState);

    if (!BULK_ARC_OP_USE_SORT(oldState->nouts, newState->nouts)) {
	/* With not too many arcs, just do them one at a time */
	struct arc *a;

	for (a = oldState->outs; a != NULL; a = a->outchain) {
	    cparc(nfa, a, newState, a->to);
	}
    } else {
 	/*
	 * With many arcs, use a sort-merge approach.  Note that createarc()
	 * will put new arcs onto the front of newState's chain, so it does
	 * not break our walk through the sorted part of the chain.
	 */
	struct arc *oa;
	struct arc *na;

	/*
	 * Because we bypass newarc() in this code path, we'd better include a
	 * cancel check.
	 */
	if (CANCEL_REQUESTED(nfa->v->re)) {
	    NERR(REG_CANCEL);
	    return;
	}

	sortouts(nfa, oldState);
	sortouts(nfa, newState);
	if (NISERR()) {
	    return;		/* might have failed to sort */
	}
	oa = oldState->outs;
	na = newState->outs;
	while (oa != NULL && na != NULL) {
	    struct arc *a = oa;

	    switch (sortouts_cmp(&oa, &na)) {
		case -1:
		    /* newState does not have anything matching oa */
		    oa = oa->outchain;
		    createarc(nfa, a->type, a->co, newState, a->to);
		    break;
		case 0:
		    /* match, advance in both lists */
		    oa = oa->outchain;
		    na = na->outchain;
		    break;
		case +1:
		    /* advance only na; oa might have a match later */
		    na = na->outchain;
		    break;
		default:
		    assert(NOTREACHED);
	    }
	}
	while (oa != NULL) {
	    /* newState does not have anything matching oa */
	    struct arc *a = oa;

	    oa = oa->outchain;
	    createarc(nfa, a->type, a->co, newState, a->to);
	}
    }
}

/*
 - cloneouts - copy out arcs of a state to another state pair, modifying type
 ^ static void cloneouts(struct nfa *, struct state *, struct state *,
 ^ 	struct state *, int);
892
893
894
895
896
897
898














899
900
901
902
903
904
905
    }
}

/*
 - optimize - optimize an NFA
 ^ static long optimize(struct nfa *, FILE *);
 */














static long			/* re_info bits */
optimize(
    struct nfa *nfa,
    FILE *f)			/* for debug output; NULL none */
{
    int verbose = (f != NULL) ? 1 : 0;








>
>
>
>
>
>
>
>
>
>
>
>
>
>







1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
    }
}

/*
 - optimize - optimize an NFA
 ^ static long optimize(struct nfa *, FILE *);
 */

 /*
  * The main goal of this function is not so much "optimization" (though it
  * does try to get rid of useless NFA states) as reducing the NFA to a form
  * the regex executor can handle.  The executor, and indeed the cNFA format
  * that is its input, can only handle PLAIN and LACON arcs.  The output of
  * the regex parser also includes EMPTY (do-nothing) arcs, as well as
  * ^, $, AHEAD, and BEHIND constraint arcs, which we must get rid of here.
  * We first get rid of EMPTY arcs and then deal with the constraint arcs.
  * The hardest part of either job is to get rid of circular loops of the
  * target arc type.  We would have to do that in any case, though, as such a
  * loop would otherwise allow the executor to cycle through the loop endlessly
  * without making any progress in the input string.
  */
static long			/* re_info bits */
optimize(
    struct nfa *nfa,
    FILE *f)			/* for debug output; NULL none */
{
    int verbose = (f != NULL) ? 1 : 0;

913
914
915
916
917
918
919

920
921
922
923
924
925





926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941

942
943
944
945
946
947
948
949
950
951

952
953
954
955
956
957
958
959











960
961
962
963
964
965
966
967
968
969






970
971
972
973
974
975
976
977
978
979
980
981




982
983


984






985
986
987
988
989
990

991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030


1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041



1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059








1060
1061
1062
1063

1064
1065
1066
1067
1068


1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082

1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098

1099
1100
1101
1102
1103
1104
1105
1106
1107
1108

1109
1110
1111
1112
1113
1114
1115




1116






1117
1118
1119
1120
1121
1122
1123
1124
1125
1126






1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138




1139
1140


1141






1142
1143
1144
1145
1146
1147

1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190


1191
1192
1193
1194
1195
1196
1197
1198
1199
1200



1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218








1219
1220
1221
1222



1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241

1242
1243
1244
1245
1246
1247
1248
    if (verbose) {
	fprintf(f, "\nempties:\n");
    }
    fixempties(nfa, f);		/* get rid of EMPTY arcs */
    if (verbose) {
	fprintf(f, "\nconstraints:\n");
    }

    pullback(nfa, f);		/* pull back constraints backward */
    pushfwd(nfa, f);		/* push fwd constraints forward */
    if (verbose) {
	fprintf(f, "\nfinal cleanup:\n");
    }
    cleanup(nfa);		/* final tidying */





    return analyze(nfa);	/* and analysis */
}

/*
 - pullback - pull back constraints backward to (with luck) eliminate them
 ^ static void pullback(struct nfa *, FILE *);
 */
static void
pullback(
    struct nfa *nfa,
    FILE *f)			/* for debug output; NULL none */
{
    struct state *s;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;

    int progress;

    /*
     * Find and pull until there are no more.
     */

    do {
	progress = 0;
	for (s=nfa->states ; s!=NULL && !NISERR() ; s=nexts) {
	    nexts = s->next;

	    for (a=s->outs ; a!=NULL && !NISERR() ; a=nexta) {
		nexta = a->outchain;
		if (a->type == '^' || a->type == BEHIND) {
		    if (pull(nfa, a)) {
			progress = 1;
		    }
		}
		assert(nexta == NULL || s->no != FREESTATE);











	    }
	}
	if (progress && f != NULL) {
	    dumpnfa(nfa, f);
	}
    } while (progress && !NISERR());
    if (NISERR()) {
	return;
    }







    for (a=nfa->pre->outs ; a!=NULL ; a=nexta) {
	nexta = a->outchain;
	if (a->type == '^') {
	    assert(a->co == 0 || a->co == 1);
	    newarc(nfa, PLAIN, nfa->bos[a->co], a->from, a->to);
	    freearc(nfa, a);
	}
    }
}

/*
 - pull - pull a back constraint backward past its source state




 * A significant property of this function is that it deletes at most
 * one state -- the constraint's from state -- and only if the constraint


 * was that state's last outarc.






 ^ static int pull(struct nfa *, struct arc *);
 */
static int			/* 0 couldn't, 1 could */
pull(
    struct nfa *nfa,
    struct arc *con)

{
    struct state *from = con->from;
    struct state *to = con->to;
    struct arc *a;
    struct arc *nexta;
    struct state *s;

    if (from == to) {		/* circular constraint is pointless */
	freearc(nfa, con);
	return 1;
    }
    if (from->flag) {		/* can't pull back beyond start */
	return 0;
    }
    if (from->nins == 0) {	/* unreachable */
	freearc(nfa, con);
	return 1;
    }

    /*
     * DGP 2007-11-15: Cloning a state with a circular constraint on its list
     * of outs can lead to trouble [Bug 1810038], so get rid of them first.
     */

    for (a = from->outs; a != NULL; a = nexta) {
	nexta = a->outchain;
	switch (a->type) {
	case '^':
	case '$':
	case BEHIND:
	case AHEAD:
	    if (from == a->to) {
		freearc(nfa, a);
	    }
	    break;
	}
    }

    /*
     * First, clone from state if necessary to avoid other outarcs.


     */

    if (from->nouts > 1) {
	s = newstate(nfa);
	if (NISERR()) {
	    return 0;
	}
	assert(to != from);		/* con is not an inarc */
	copyins(nfa, from, s, 1);	/* duplicate inarcs */
	cparc(nfa, con, s, to);		/* move constraint arc */
	freearc(nfa, con);



	from = s;
	con = from->outs;
    }
    assert(from->nouts == 1);

    /*
     * Propagate the constraint into the from state's inarcs.
     */

    for (a=from->ins ; a!=NULL ; a=nexta) {
	nexta = a->inchain;
	switch (combine(con, a)) {
	case INCOMPATIBLE:	/* destroy the arc */
	    freearc(nfa, a);
	    break;
	case SATISFIED:		/* no action needed */
	    break;
	case COMPATIBLE:	/* swap the two arcs, more or less */








	    s = newstate(nfa);
	    if (NISERR()) {
		return 0;
	    }

	    cparc(nfa, a, s, to);	/* anticipate move */
	    cparc(nfa, con, a->from, s);
	    if (NISERR()) {
		return 0;
	    }


	    freearc(nfa, a);
	    break;
	default:
	    assert(NOTREACHED);
	    break;
	}
    }

    /*
     * Remaining inarcs, if any, incorporate the constraint.
     */

    moveins(nfa, from, to);
    dropstate(nfa, from);	/* will free the constraint */

    return 1;
}

/*
 - pushfwd - push forward constraints forward to (with luck) eliminate them
 ^ static void pushfwd(struct nfa *, FILE *);
 */
static void
pushfwd(
    struct nfa *nfa,
    FILE *f)			/* for debug output; NULL none */
{
    struct state *s;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;

    int progress;

    /*
     * Find and push until there are no more.
     */

    do {
	progress = 0;
	for (s=nfa->states ; s!=NULL && !NISERR() ; s=nexts) {
	    nexts = s->next;

	    for (a = s->ins; a != NULL && !NISERR(); a = nexta) {
		nexta = a->inchain;
		if (a->type == '$' || a->type == AHEAD) {
		    if (push(nfa, a)) {
			progress = 1;
		    }
		}




		assert(nexta == NULL || s->no != FREESTATE);






	    }
	}
	if (progress && f != NULL) {
	    dumpnfa(nfa, f);
	}
    } while (progress && !NISERR());
    if (NISERR()) {
	return;
    }







    for (a = nfa->post->ins; a != NULL; a = nexta) {
	nexta = a->inchain;
	if (a->type == '$') {
	    assert(a->co == 0 || a->co == 1);
	    newarc(nfa, PLAIN, nfa->eos[a->co], a->from, a->to);
	    freearc(nfa, a);
	}
    }
}

/*
 - push - push a forward constraint forward past its destination state




 * A significant property of this function is that it deletes at most
 * one state -- the constraint's to state -- and only if the constraint


 * was that state's last inarc.






 ^ static int push(struct nfa *, struct arc *);
 */
static int			/* 0 couldn't, 1 could */
push(
    struct nfa *nfa,
    struct arc *con)

{
    struct state *from = con->from;
    struct state *to = con->to;
    struct arc *a;
    struct arc *nexta;
    struct state *s;

    if (to == from) {		/* circular constraint is pointless */
	freearc(nfa, con);
	return 1;
    }
    if (to->flag) {		/* can't push forward beyond end */
	return 0;
    }
    if (to->nouts == 0) {	/* dead end */
	freearc(nfa, con);
	return 1;
    }

    /*
     * DGP 2007-11-15: Here we duplicate the same protections as appear
     * in pull() above to avoid troubles with cloning a state with a
     * circular constraint on its list of ins.  It is not clear whether
     * this is necessary, or is protecting against a "can't happen".
     * Any test case that actually leads to a freearc() call here would
     * be a welcome addition to the test suite.
     */

    for (a = to->ins; a != NULL; a = nexta) {
	nexta = a->inchain;
	switch (a->type) {
	case '^':
	case '$':
	case BEHIND:
	case AHEAD:
	    if (a->from == to) {
		freearc(nfa, a);
	    }
	    break;
	}
    }
    /*
     * First, clone to state if necessary to avoid other inarcs.


     */

    if (to->nins > 1) {
	s = newstate(nfa);
	if (NISERR()) {
	    return 0;
	}
	copyouts(nfa, to, s, 1);	/* duplicate outarcs */
	cparc(nfa, con, from, s);	/* move constraint */
	freearc(nfa, con);



	to = s;
	con = to->ins;
    }
    assert(to->nins == 1);

    /*
     * Propagate the constraint into the to state's outarcs.
     */

    for (a = to->outs; a != NULL; a = nexta) {
	nexta = a->outchain;
	switch (combine(con, a)) {
	case INCOMPATIBLE:	/* destroy the arc */
	    freearc(nfa, a);
	    break;
	case SATISFIED:		/* no action needed */
	    break;
	case COMPATIBLE:	/* swap the two arcs, more or less */








	    s = newstate(nfa);
	    if (NISERR()) {
		return 0;
	    }



	    cparc(nfa, con, s, a->to);	/* anticipate move */
	    cparc(nfa, a, from, s);
	    if (NISERR()) {
		return 0;
	    }
	    freearc(nfa, a);
	    break;
	default:
	    assert(NOTREACHED);
	    break;
	}
    }

    /*
     * Remaining outarcs, if any, incorporate the constraint.
     */

    moveouts(nfa, to, from);
    dropstate(nfa, to);		/* will free the constraint */

    return 1;
}

/*
 - combine - constraint lands on an arc, what happens?
 ^ #def	INCOMPATIBLE	1	// destroys arc
 ^ #def	SATISFIED	2	// constraint satisfied







>






>
>
>
>
>




|











>










>



|




>
>
>
>
>
>
>
>
>
>
>










>
>
>
>
>
>












>
>
>
>
|
|
>
>
|
>
>
>
>
>
>


|


|
>







|
<
<
<









<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
>
>







<
|


>
>
>









|








>
>
>
>
>
>
>
>
|
|
|
|
>
|
<
<
<

>
>
|
|











|
>




|











>










>



|



>
>
>
>
|
>
>
>
>
>
>










>
>
>
>
>
>












>
>
>
>
|
|
>
>
|
>
>
>
>
>
>


|


|
>







|
<
<
<









<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
>
>







|
|

>
>
>









|








>
>
>
>
>
>
>
>
|
|
|
|
>
>
>
|
|
<
<
<
|
|











|
>







1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543



1544
1545
1546
1547
1548
1549
1550
1551
1552



















1553
1554
1555
1556
1557
1558
1559
1560
1561
1562

1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600



1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722



1723
1724
1725
1726
1727
1728
1729
1730
1731






















1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782



1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
    if (verbose) {
	fprintf(f, "\nempties:\n");
    }
    fixempties(nfa, f);		/* get rid of EMPTY arcs */
    if (verbose) {
	fprintf(f, "\nconstraints:\n");
    }
    fixconstraintloops(nfa, f);	/* get rid of constraint loops */
    pullback(nfa, f);		/* pull back constraints backward */
    pushfwd(nfa, f);		/* push fwd constraints forward */
    if (verbose) {
	fprintf(f, "\nfinal cleanup:\n");
    }
    cleanup(nfa);		/* final tidying */
#ifdef REG_DEBUG
    if (verbose) {
	dumpnfa(nfa, f);
    }
#endif
    return analyze(nfa);	/* and analysis */
}

/*
 - pullback - pull back constraints backward to eliminate them
 ^ static void pullback(struct nfa *, FILE *);
 */
static void
pullback(
    struct nfa *nfa,
    FILE *f)			/* for debug output; NULL none */
{
    struct state *s;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;
    struct state *intermediates;
    int progress;

    /*
     * Find and pull until there are no more.
     */

    do {
	progress = 0;
	for (s=nfa->states ; s!=NULL && !NISERR() ; s=nexts) {
	    nexts = s->next;
	    intermediates = NULL;
	    for (a=s->outs ; a!=NULL && !NISERR() ; a=nexta) {
		nexta = a->outchain;
		if (a->type == '^' || a->type == BEHIND) {
		    if (pull(nfa, a, &intermediates)) {
			progress = 1;
		    }
		}
		assert(nexta == NULL || s->no != FREESTATE);
	    }
	    /* clear tmp fields of intermediate states created here */
	    while (intermediates != NULL) {
		struct state *ns = intermediates->tmp;

		intermediates->tmp = NULL;
		intermediates = ns;
	    }
	    /* if s is now useless, get rid of it */
	    if ((s->nins == 0 || s->nouts == 0) && !s->flag) {
		dropstate(nfa, s);
	    }
	}
	if (progress && f != NULL) {
	    dumpnfa(nfa, f);
	}
    } while (progress && !NISERR());
    if (NISERR()) {
	return;
    }

    /*
     * Any ^ constraints we were able to pull to the start state can now be
     * replaced by PLAIN arcs referencing the BOS or BOL colors.  There should
     * be no other ^ or BEHIND arcs left in the NFA, though we do not check
     * that here (compact() will fail if so).
     */
    for (a=nfa->pre->outs ; a!=NULL ; a=nexta) {
	nexta = a->outchain;
	if (a->type == '^') {
	    assert(a->co == 0 || a->co == 1);
	    newarc(nfa, PLAIN, nfa->bos[a->co], a->from, a->to);
	    freearc(nfa, a);
	}
    }
}

/*
 - pull - pull a back constraint backward past its source state
 *
 * Returns 1 if successful (which it always is unless the source is the
 * start state or we have an internal error), 0 if nothing happened.
 *
 * A significant property of this function is that it deletes no pre-existing
 * states, and no outarcs of the constraint's from state other than the given
 * constraint arc.  This makes the loops in pullback() safe, at the cost that
 * we may leave useless states behind.  Therefore, we leave it to pullback()
 * to delete such states.
 *
 * If the from state has multiple back-constraint outarcs, and/or multiple
 * compatible constraint inarcs, we only need to create one new intermediate
 * state per combination of predecessor and successor states.  *intermediates
 * points to a list of such intermediate states for this from state (chained
 * through their tmp fields).
 ^ static int pull(struct nfa *, struct arc *);
 */
static int
pull(
    struct nfa *nfa,
    struct arc *con,
    struct state **intermediates)
{
    struct state *from = con->from;
    struct state *to = con->to;
    struct arc *a;
    struct arc *nexta;
    struct state *s;

    assert(from != to);		/* should have gotten rid of this earlier */



    if (from->flag) {		/* can't pull back beyond start */
	return 0;
    }
    if (from->nins == 0) {	/* unreachable */
	freearc(nfa, con);
	return 1;
    }

    /*



















     * First, clone from state if necessary to avoid other outarcs.  This may
     * seem wasteful, but it simplifies the logic, and we'll get rid of the
     * clone state again at the bottom.
     */

    if (from->nouts > 1) {
	s = newstate(nfa);
	if (NISERR()) {
	    return 0;
	}

	copyins(nfa, from, s);	/* duplicate inarcs */
	cparc(nfa, con, s, to);		/* move constraint arc */
	freearc(nfa, con);
	if (NISERR()) {
	    return 0;
	}
	from = s;
	con = from->outs;
    }
    assert(from->nouts == 1);

    /*
     * Propagate the constraint into the from state's inarcs.
     */

    for (a=from->ins ; a!=NULL && !NISERR(); a=nexta) {
	nexta = a->inchain;
	switch (combine(con, a)) {
	case INCOMPATIBLE:	/* destroy the arc */
	    freearc(nfa, a);
	    break;
	case SATISFIED:		/* no action needed */
	    break;
	case COMPATIBLE:	/* swap the two arcs, more or less */
	    /* need an intermediate state, but might have one already */
	    for (s = *intermediates; s != NULL; s = s->tmp) {
		assert(s->nins > 0 && s->nouts > 0);
		if (s->ins->from == a->from && s->outs->to == to) {
		    break;
		}
	    }
	    if (s == NULL) {
		s = newstate(nfa);
		if (NISERR()) {
		    return 0;
		}
		s->tmp = *intermediates;
		*intermediates = s;



	    }
  	    cparc(nfa, con, a->from, s);
	    cparc(nfa, a, s, to);
 	    freearc(nfa, a);
  	    break;
	default:
	    assert(NOTREACHED);
	    break;
	}
    }

    /*
     * Remaining inarcs, if any, incorporate the constraint.
     */

    moveins(nfa, from, to);
    freearc(nfa, con);
    /* from state is now useless, but we leave it to pullback() to clean up */
    return 1;
}

/*
 - pushfwd - push forward constraints forward to eliminate them
 ^ static void pushfwd(struct nfa *, FILE *);
 */
static void
pushfwd(
    struct nfa *nfa,
    FILE *f)			/* for debug output; NULL none */
{
    struct state *s;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;
    struct state *intermediates;
    int progress;

    /*
     * Find and push until there are no more.
     */

    do {
	progress = 0;
	for (s=nfa->states ; s!=NULL && !NISERR() ; s=nexts) {
	    nexts = s->next;
	    intermediates = NULL;
	    for (a = s->ins; a != NULL && !NISERR(); a = nexta) {
		nexta = a->inchain;
		if (a->type == '$' || a->type == AHEAD) {
		    if (push(nfa, a, &intermediates)) {
			progress = 1;
		    }
		}
	    }
	    /* clear tmp fields of intermediate states created here */
	    while (intermediates != NULL) {
		struct state *ns = intermediates->tmp;

		intermediates->tmp = NULL;
		intermediates = ns;
	    }
	    /* if s is now useless, get rid of it */
	    if ((s->nins == 0 || s->nouts == 0) && !s->flag) {
		dropstate(nfa, s);
	    }
	}
	if (progress && f != NULL) {
	    dumpnfa(nfa, f);
	}
    } while (progress && !NISERR());
    if (NISERR()) {
	return;
    }

    /*
     * Any $ constraints we were able to push to the post state can now be
     * replaced by PLAIN arcs referencing the EOS or EOL colors.  There should
     * be no other $ or AHEAD arcs left in the NFA, though we do not check
     * that here (compact() will fail if so).
     */
    for (a = nfa->post->ins; a != NULL; a = nexta) {
	nexta = a->inchain;
	if (a->type == '$') {
	    assert(a->co == 0 || a->co == 1);
	    newarc(nfa, PLAIN, nfa->eos[a->co], a->from, a->to);
	    freearc(nfa, a);
	}
    }
}

/*
 - push - push a forward constraint forward past its destination state
 *
 * Returns 1 if successful (which it always is unless the destination is the
 * post state or we have an internal error), 0 if nothing happened.
 *
 * A significant property of this function is that it deletes no pre-existing
 * states, and no inarcs of the constraint's to state other than the given
 * constraint arc.  This makes the loops in pushfwd() safe, at the cost that
 * we may leave useless states behind.  Therefore, we leave it to pushfwd()
 * to delete such states.
 *
 * If the to state has multiple forward-constraint inarcs, and/or multiple
 * compatible constraint outarcs, we only need to create one new intermediate
 * state per combination of predecessor and successor states.  *intermediates
 * points to a list of such intermediate states for this to state (chained
 * through their tmp fields).
 ^ static int push(struct nfa *, struct arc *);
 */
static int
push(
    struct nfa *nfa,
    struct arc *con,
    struct state **intermediates)
{
    struct state *from = con->from;
    struct state *to = con->to;
    struct arc *a;
    struct arc *nexta;
    struct state *s;

    assert(to != from);		/* should have gotten rid of this earlier */



    if (to->flag) {		/* can't push forward beyond end */
	return 0;
    }
    if (to->nouts == 0) {	/* dead end */
	freearc(nfa, con);
	return 1;
    }

    /*






















     * First, clone to state if necessary to avoid other inarcs.  This may
     * seem wasteful, but it simplifies the logic, and we'll get rid of the
     * clone state again at the bottom.
     */

    if (to->nins > 1) {
	s = newstate(nfa);
	if (NISERR()) {
	    return 0;
	}
	copyouts(nfa, to, s);		/* duplicate outarcs */
	cparc(nfa, con, from, s);	/* move constraint arc */
	freearc(nfa, con);
	if (NISERR()) {
	    return 0;
	}
	to = s;
	con = to->ins;
    }
    assert(to->nins == 1);

    /*
     * Propagate the constraint into the to state's outarcs.
     */

    for (a = to->outs; a != NULL && !NISERR(); a = nexta) {
	nexta = a->outchain;
	switch (combine(con, a)) {
	case INCOMPATIBLE:	/* destroy the arc */
	    freearc(nfa, a);
	    break;
	case SATISFIED:		/* no action needed */
	    break;
	case COMPATIBLE:	/* swap the two arcs, more or less */
	    /* need an intermediate state, but might have one already */
	    for (s = *intermediates; s != NULL; s = s->tmp) {
		assert(s->nins > 0 && s->nouts > 0);
		if (s->ins->from == from && s->outs->to == a->to) {
		    break;
		}
	    }
	    if (s == NULL) {
		s = newstate(nfa);
		if (NISERR()) {
		    return 0;
		}
		s->tmp = *intermediates;
		*intermediates = s;
	    }
	    cparc(nfa, con, s, a->to);
  	    cparc(nfa, a, from, s);



  	    freearc(nfa, a);
  	    break;
	default:
	    assert(NOTREACHED);
	    break;
	}
    }

    /*
     * Remaining outarcs, if any, incorporate the constraint.
     */

    moveouts(nfa, to, from);
    freearc(nfa, con);
    /* to state is now useless, but we leave it to pushfwd() to clean up */
    return 1;
}

/*
 - combine - constraint lands on an arc, what happens?
 ^ #def	INCOMPATIBLE	1	// destroys arc
 ^ #def	SATISFIED	2	// constraint satisfied
1312
1313
1314
1315
1316
1317
1318






1319
1320
1321
1322
1323
1324
1325
    FILE *f)			/* for debug output; NULL none */
{
    struct state *s;
    struct state *s2;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;







    /*
     * First, get rid of any states whose sole out-arc is an EMPTY,
     * since they're basically just aliases for their successor.  The
     * parsing algorithm creates enough of these that it's worth
     * special-casing this.
     */







>
>
>
>
>
>







1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
    FILE *f)			/* for debug output; NULL none */
{
    struct state *s;
    struct state *s2;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;
    int totalinarcs;
    struct arc **inarcsorig;
    struct arc **arcarray;
    int arccount;
    int prevnins;
    int nskip;

    /*
     * First, get rid of any states whose sole out-arc is an EMPTY,
     * since they're basically just aliases for their successor.  The
     * parsing algorithm creates enough of these that it's worth
     * special-casing this.
     */
1357
1358
1359
1360
1361
1362
1363




1364
1365
1366
1367
1368

1369
1370






1371



1372






1373
1374










1375




1376



1377
1378


1379
1380









1381



1382
1383
1384
1385
1386
1387
1388




1389










1390


1391








1392
1393
1394
1395
1396
1397
1398

1399


















1400
1401
1402
1403
1404
1405
1406
	}
	if (s != a->from) {
	    moveouts(nfa, s, a->from);
	}
	dropstate(nfa, s);
    }





    /*
     * For each remaining NFA state, find all other states that are
     * reachable from it by a chain of one or more EMPTY arcs.  Then
     * generate new arcs that eliminate the need for each such chain.
     *

     * If we just do this straightforwardly, the algorithm gets slow in
     * complex graphs, because the same arcs get copied to all






     * intermediate states of an EMPTY chain, and then uselessly pushed



     * repeatedly to the chain's final state; we waste a lot of time in






     * newarc's duplicate checking.  To improve matters, we decree that
     * any state with only EMPTY out-arcs is "doomed" and will not be










     * part of the final NFA. That can be ensured by not adding any new




     * out-arcs to such a state. Having ensured that, we need not update



     * the state's in-arcs list either; all arcs that might have gotten
     * pushed forward to it will just get pushed directly to successor


     * states.  This eliminates most of the useless duplicate arcs.
     */









    for (s = nfa->states; s != NULL && !NISERR(); s = s->next) {



	for (s2 = emptyreachable(s, s); s2 != s && !NISERR();
		s2 = nexts) {
	    /*
	     * If s2 is doomed, we decide that (1) we will always push
	     * arcs forward to it, not pull them back to s; and (2) we
	     * can optimize away the push-forward, per comment above.
	     * So do nothing.




	     */










	    if (s2->flag || hasnonemptyout(s2)) {


		replaceempty(nfa, s, s2);








	    }

	    /* Reset the tmp fields as we walk back */
	    nexts = s2->tmp;
	    s2->tmp = NULL;
	}
	s->tmp = NULL;

    }


















    if (NISERR()) {
	return;
    }

    /*
     * Remove all the EMPTY arcs, since we don't need them anymore.
     */







>
>
>
>

|
|
|

>
|
<
>
>
>
>
>
>
|
>
>
>
|
>
>
>
>
>
>
|
<
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
|
>
>
>
|
|
>
>
|

>
>
>
>
>
>
>
>
>
|
>
>
>
|
<
|
<
<
<
<
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
|
>
>
|
>
>
>
>
>
>
>
>

|
|
|
|
|
|
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936

1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954

1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993

1994




1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
	}
	if (s != a->from) {
	    moveouts(nfa, s, a->from);
	}
	dropstate(nfa, s);
    }

    if (NISERR()) {
	return;
    }

    /*
     * For each remaining NFA state, find all other states from which it is
     * reachable by a chain of one or more EMPTY arcs.  Then generate new arcs
     * that eliminate the need for each such chain.
     *
     * We could replace a chain of EMPTY arcs that leads from a "from" state
     * to a "to" state either by pushing non-EMPTY arcs forward (linking

     * directly from "from"'s predecessors to "to") or by pulling them back
     * (linking directly from "from" to "to"'s successors).  We choose to
     * always do the former; this choice is somewhat arbitrary, but the
     * approach below requires that we uniformly do one or the other.
     *
     * Suppose we have a chain of N successive EMPTY arcs (where N can easily
     * approach the size of the NFA).  All of the intermediate states must
     * have additional inarcs and outarcs, else they'd have been removed by
     * the steps above.  Assuming their inarcs are mostly not empties, we will
     * add O(N^2) arcs to the NFA, since a non-EMPTY inarc leading to any one
     * state in the chain must be duplicated to lead to all its successor
     * states as well.  So there is no hope of doing less than O(N^2) work;
     * however, we should endeavor to keep the big-O cost from being even
     * worse than that, which it can easily become without care.  In
     * particular, suppose we were to copy all S1's inarcs forward to S2, and
     * then also to S3, and then later we consider pushing S2's inarcs forward
     * to S3.  If we include the arcs already copied from S1 in that, we'd be
     * doing O(N^3) work.  (The duplicate-arc elimination built into newarc()

     * and its cohorts would get rid of the extra arcs, but not without cost.)
     *
     * We can avoid this cost by treating only arcs that existed at the start
     * of this phase as candidates to be pushed forward.  To identify those,
     * we remember the first inarc each state had to start with.  We rely on
     * the fact that newarc() and friends put new arcs on the front of their
     * to-states' inchains, and that this phase never deletes arcs, so that
     * the original arcs must be the last arcs in their to-states' inchains.
     *
     * So the process here is that, for each state in the NFA, we gather up
     * all non-EMPTY inarcs of states that can reach the target state via
     * EMPTY arcs.  We then sort, de-duplicate, and merge these arcs into the
     * target state's inchain.  (We can safely use sort-merge for this as long
     * as we update each state's original-arcs pointer after we add arcs to
     * it; the sort step of mergeins probably changed the order of the old
     * arcs.)
     *
     * Another refinement worth making is that, because we only add non-EMPTY
     * arcs during this phase, and all added arcs have the same from-state as
     * the non-EMPTY arc they were cloned from, we know ahead of time that any
     * states having only EMPTY outarcs will be useless for lack of outarcs
     * after we drop the EMPTY arcs.  (They cannot gain non-EMPTY outarcs if
     * they had none to start with.)  So we need not bother to update the
     * inchains of such states at all.
     */

    /* Remember the states' first original inarcs */
    /* ... and while at it, count how many old inarcs there are altogether */
    inarcsorig = (struct arc **) MALLOC(nfa->nstates * sizeof(struct arc *));
    if (inarcsorig == NULL) {
	NERR(REG_ESPACE);
	return;
    }
    totalinarcs = 0;
    for (s = nfa->states; s != NULL; s = s->next) {
	inarcsorig[s->no] = s->ins;
	totalinarcs += s->nins;
    }


    /*




     * Create a workspace for accumulating the inarcs to be added to the
     * current target state.  totalinarcs is probably a considerable
     * overestimate of the space needed, but the NFA is unlikely to be large
     * enough at this point to make it worth being smarter.
     */
    arcarray = (struct arc **) MALLOC(totalinarcs * sizeof(struct arc *));
    if (arcarray == NULL) {
	NERR(REG_ESPACE);
	FREE(inarcsorig);
	return;
    }

    /* And iterate over the target states */
    for (s = nfa->states; s != NULL && !NISERR(); s = s->next) {
	/* Ignore target states without non-EMPTY outarcs, per note above */
	if (!s->flag && !hasnonemptyout(s)) {
	    continue;
	}

	/* Find predecessor states and accumulate their original inarcs */
	arccount = 0;
	for (s2 = emptyreachable(nfa, s, s, inarcsorig); s2 != s; s2 = nexts) {
	    /* Add s2's original inarcs to arcarray[], but ignore empties */
	    for (a = inarcsorig[s2->no]; a != NULL; a = a->inchain) {
		if (a->type != EMPTY) {
		    arcarray[arccount++] = a;
		}
	    }
  
  	    /* Reset the tmp fields as we walk back */
  	    nexts = s2->tmp;
  	    s2->tmp = NULL;
  	}
  	s->tmp = NULL;
	assert(arccount <= totalinarcs);

	/* Remember how many original inarcs this state has */
	prevnins = s->nins;

	/* Add non-duplicate inarcs to target state */
	mergeins(nfa, s, arcarray, arccount);

	/* Now we must update the state's inarcsorig pointer */
	nskip = s->nins - prevnins;
	a = s->ins;
	while (nskip-- > 0) {
	    a = a->inchain;
	}
	inarcsorig[s->no] = a;
    }
  
    FREE(arcarray);
    FREE(inarcsorig);

    if (NISERR()) {
	return;
    }

    /*
     * Remove all the EMPTY arcs, since we don't need them anymore.
     */
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438





1439
1440
1441
1442
1443
1444
1445

1446
1447

1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459








































































1460













1461
1462
1463
1464
1465





1466
1467


1468


1469



1470













1471


1472

1473


1474





















1475






















1476



1477












































































































1478







1479








1480


1481



1482













































1483

1484
1485
1486





1487
1488





















1489





1490








1491






1492








1493




1494











1495








1496











1497








1498









1499

1500










1501





1502
1503

1504


1505








1506









1507






1508


















1509
1510






1511




1512


1513
1514

1515
1516
1517



1518
1519
1520
1521
1522
1523
1524

    if (f != NULL) {
	dumpnfa(nfa, f);
    }
}

/*
 - emptyreachable - recursively find all states reachable from s by EMPTY arcs
 * The return value is the last such state found.  Its tmp field links back
 * to the next-to-last such state, and so on back to s, so that all these
 * states can be located without searching the whole NFA.





 * The maximum recursion depth here is equal to the length of the longest
 * loop-free chain of EMPTY arcs, which is surely no more than the size of
 * the NFA, and in practice will be a lot less than that.
 ^ static struct state *emptyreachable(struct state *, struct state *);
 */
static struct state *
emptyreachable(

    struct state *s,
    struct state *lastfound)

{
    struct arc *a;

    s->tmp = lastfound;
    lastfound = s;
    for (a = s->outs; a != NULL; a = a->outchain) {
	if (a->type == EMPTY && a->to->tmp == NULL) {
	    lastfound = emptyreachable(a->to, lastfound);
	}
    }
    return lastfound;
}






















































































/*
 - replaceempty - replace an EMPTY arc chain with some non-empty arcs
 * The EMPTY arc(s) should be deleted later, but we can't do it here because
 * they may still be needed to identify other arc chains during fixempties().
 ^ static void replaceempty(struct nfa *, struct state *, struct state *);





 */
static void


replaceempty(


    struct nfa *nfa,



    struct state *from,













    struct state *to)


{

    int fromouts;


    int toins;












































    assert(from != to);
















































































































    /*







     * Create replacement arcs that bypass the need for the EMPTY chain.  We








     * can do this either by pushing arcs forward (linking directly from


     * "from"'s predecessors to "to") or by pulling them back (linking



     * directly from "from" to "to"'s successors).  In general, we choose













































     * whichever way creates greater fan-out or fan-in, so as to improve the

     * odds of reducing the other state to zero in-arcs or out-arcs and
     * thereby being able to delete it.  However, if "from" is doomed (has no
     * non-EMPTY out-arcs), we must keep it so, so always push forward in that





     * case.
     *





















     * The fan-out/fan-in comparison should count only non-EMPTY arcs.  If





     * "from" is doomed, we can skip counting "to"'s arcs, since we want to








     * force taking the copynonemptyins path in that case.






     */








    fromouts = nonemptyouts(from);




    toins = (fromouts == 0) ? 1 : nonemptyins(to);




















    if (fromouts > toins) {











	copyouts(nfa, to, from, 0);








	return;









    }

    if (fromouts < toins) {










	copyins(nfa, from, to, 0);





	return;
    }




    /*








     * fromouts == toins.  Decide on secondary issue: copy fewest arcs.









     *






     * Doesn't seem to be worth the trouble to exclude empties from these


















     * comparisons; that takes extra time and doesn't seem to improve the
     * resulting graph much.






     */




    if (from->nins > to->nouts) {


	copyouts(nfa, to, from, 0);
	return;

    }

    copyins(nfa, from, to, 0);



}

/*
 - cleanup - clean up NFA after optimizations
 ^ static void cleanup(struct nfa *);
 */
static void







|



>
>
>
>
>


|




>

|
>





|
|
|




>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
|
<
<
<
<
>
>
>
>
>
|
|
>
>
|
>
>
|
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
|
>
|
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
|
>
>
|
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
|
<
|
>
>
>
>
>
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
|
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
|
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
|
>
|
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
|
|
>
|
>
>
|
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
<
>
>
>
>
>
>

>
>
>
>
|
>
>
|
<
>
|
|
|
>
>
>







2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202




2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471

2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659

2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674

2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688

    if (f != NULL) {
	dumpnfa(nfa, f);
    }
}

/*
 - emptyreachable - recursively find all states that can reach s by EMPTY arcs
 * The return value is the last such state found.  Its tmp field links back
 * to the next-to-last such state, and so on back to s, so that all these
 * states can be located without searching the whole NFA.
 *
 * Since this is only used in fixempties(), we pass in the inarcsorig[] array
 * maintained by that function.  This lets us skip over all new inarcs, which
 * are certainly not EMPTY arcs.
 *
 * The maximum recursion depth here is equal to the length of the longest
 * loop-free chain of EMPTY arcs, which is surely no more than the size of
 * the NFA, and in practice will be less than that.
 ^ static struct state *emptyreachable(struct state *, struct state *);
 */
static struct state *
emptyreachable(
    struct nfa *nfa,
    struct state *s,
    struct state *lastfound,
    struct arc **inarcsorig)
{
    struct arc *a;

    s->tmp = lastfound;
    lastfound = s;
    for (a = inarcsorig[s->no]; a != NULL; a = a->inchain) {
	if (a->type == EMPTY && a->from->tmp == NULL) {
	    lastfound = emptyreachable(nfa, a->from, lastfound, inarcsorig);
	}
    }
    return lastfound;
}

/*
 * isconstraintarc - detect whether an arc is of a constraint type
 */
static inline int
isconstraintarc(struct arc * a)
{
    switch (a->type)
    {
	case '^':
	case '$':
	case BEHIND:
	case AHEAD:
	case LACON:
	    return 1;
    }
    return 0;
}

/*
 * hasconstraintout - does state have a constraint out arc?
 */
static int
hasconstraintout(struct state * s)
{
    struct arc *a;

    for (a = s->outs; a != NULL; a = a->outchain) {
	if (isconstraintarc(a)) {
	    return 1;
	}
    }
    return 0;
}

/*
 * fixconstraintloops - get rid of loops containing only constraint arcs
 *
 * A loop of states that contains only constraint arcs is useless, since
 * passing around the loop represents no forward progress.  Moreover, it
 * would cause infinite looping in pullback/pushfwd, so we need to get rid
 * of such loops before doing that.
 */
static void
fixconstraintloops(
    struct nfa * nfa,
    FILE *f)		/* for debug output; NULL none */
{
    struct state *s;
    struct state *nexts;
    struct arc *a;
    struct arc *nexta;
    int hasconstraints;

    /*
     * In the trivial case of a state that loops to itself, we can just drop
     * the constraint arc altogether.  This is worth special-casing because
     * such loops are far more common than loops containing multiple states.
     * While we're at it, note whether any constraint arcs survive.
     */
    hasconstraints = 0;
    for (s = nfa->states; s != NULL && !NISERR(); s = nexts) {
	nexts = s->next;
	/* while we're at it, ensure tmp fields are clear for next step */
	assert(s->tmp == NULL);
	for (a = s->outs; a != NULL && !NISERR(); a = nexta) {
	    nexta = a->outchain;
	    if (isconstraintarc(a)) {
		if (a->to == s) {
		    freearc(nfa, a);
		} else {
		    hasconstraints = 1;
 		}
	    }
	}
 	/* If we removed all the outarcs, the state is useless. */
 	if (s->nouts == 0 && !s->flag) {
 	    dropstate(nfa, s);
	}
    }
 
    /* Nothing to do if no remaining constraint arcs */
    if (NISERR() || !hasconstraints) {
	return;
    }

    /*




     * Starting from each remaining NFA state, search outwards for a
     * constraint loop.  If we find a loop, break the loop, then start the
     * search over.  (We could possibly retain some state from the first scan,
     * but it would complicate things greatly, and multi-state constraint
     * loops are rare enough that it's not worth optimizing the case.)
     */
  restart:
    for (s = nfa->states; s != NULL && !NISERR(); s = s->next) {
	if (findconstraintloop(nfa, s)) {
	    goto restart;
	}
    }

    if (NISERR()) {
	return;
    }

    /*
     * Now remove any states that have become useless.  (This cleanup is not
     * very thorough, and would be even less so if we tried to combine it with
     * the previous step; but cleanup() will take care of anything we miss.)
     *
     * Because findconstraintloop intentionally doesn't reset all tmp fields,
     * we have to clear them after it's done.  This is a convenient place to
     * do that, too.
     */
    for (s = nfa->states; s != NULL; s = nexts) {
	nexts = s->next;
	s->tmp = NULL;
	if ((s->nins == 0 || s->nouts == 0) && !s->flag) {
	    dropstate(nfa, s);
	}
    }

    if (f != NULL) {
 	dumpnfa(nfa, f);
    }
}

/*
 * findconstraintloop - recursively find a loop of constraint arcs
 *
 * If we find a loop, break it by calling breakconstraintloop(), then
 * return 1; otherwise return 0.
 *
 * State tmp fields are guaranteed all NULL on a success return, because
 * breakconstraintloop does that.  After a failure return, any state that
 * is known not to be part of a loop is marked with s->tmp == s; this allows
 * us not to have to re-prove that fact on later calls.  (This convention is
 * workable because we already eliminated single-state loops.)
 *
 * Note that the found loop doesn't necessarily include the first state we
 * are called on.  Any loop reachable from that state will do.
 *
 * The maximum recursion depth here is one more than the length of the longest
 * loop-free chain of constraint arcs, which is surely no more than the size
 * of the NFA, and in practice will be a lot less than that.
 */
static int
findconstraintloop(struct nfa * nfa, struct state * s)
{
    struct arc *a;

    /* Since this is recursive, it could be driven to stack overflow */
    if (STACK_TOO_DEEP(nfa->v->re)) {
	NERR(REG_ETOOBIG);
	return 1;		/* to exit as quickly as possible */
    }

    if (s->tmp != NULL) {
	/* Already proven uninteresting? */
	if (s->tmp == s) {
	    return 0;
	}
	/* Found a loop involving s */
	breakconstraintloop(nfa, s);
	/* The tmp fields have been cleaned up by breakconstraintloop */
	return 1;
    }
    for (a = s->outs; a != NULL; a = a->outchain) {
	if (isconstraintarc(a)) {
	    struct state *sto = a->to;

	    assert(sto != s);
	    s->tmp = sto;
	    if (findconstraintloop(nfa, sto)) {
		return 1;
	    }
	}
    }

    /*
     * If we get here, no constraint loop exists leading out from s.  Mark it
     * with s->tmp == s so we need not rediscover that fact again later.
     */
    s->tmp = s;
    return 0;
}

/*
 * breakconstraintloop - break a loop of constraint arcs
 *
 * sinitial is any one member state of the loop.  Each loop member's tmp
 * field links to its successor within the loop.  (Note that this function
 * will reset all the tmp fields to NULL.)
 *
 * We can break the loop by, for any one state S1 in the loop, cloning its
 * loop successor state S2 (and possibly following states), and then moving
 * all S1->S2 constraint arcs to point to the cloned S2.  The cloned S2 should
 * copy any non-constraint outarcs of S2.  Constraint outarcs should be
 * dropped if they point back to S1, else they need to be copied as arcs to
 * similarly cloned states S3, S4, etc.  In general, each cloned state copies
 * non-constraint outarcs, drops constraint outarcs that would lead to itself
 * or any earlier cloned state, and sends other constraint outarcs to newly
 * cloned states.  No cloned state will have any inarcs that aren't constraint
 * arcs or do not lead from S1 or earlier-cloned states.  It's okay to drop
 * constraint back-arcs since they would not take us to any state we've not
 * already been in; therefore, no new constraint loop is created.  In this way
 * we generate a modified NFA that can still represent every useful state
 * sequence, but not sequences that represent state loops with no consumption
 * of input data.  Note that the set of cloned states will certainly include
 * all of the loop member states other than S1, and it may also include
 * non-loop states that are reachable from S2 via constraint arcs.  This is
 * important because there is no guarantee that findconstraintloop found a
 * maximal loop (and searching for one would be NP-hard, so don't try).
 * Frequently the "non-loop states" are actually part of a larger loop that
 * we didn't notice, and indeed there may be several overlapping loops.
 * This technique ensures convergence in such cases, while considering only
 * the originally-found loop does not.
 *
 * If there is only one S1->S2 constraint arc, then that constraint is
 * certainly satisfied when we enter any of the clone states.  This means that
 * in the common case where many of the constraint arcs are identically
 * labeled, we can merge together clone states linked by a similarly-labeled
 * constraint: if we can get to the first one we can certainly get to the
 * second, so there's no need to distinguish.  This greatly reduces the number
 * of new states needed, so we preferentially break the given loop at a state
 * pair where this is true.
 *
 * Furthermore, it's fairly common to find that a cloned successor state has
 * no outarcs, especially if we're a bit aggressive about removing unnecessary
 * outarcs.  If that happens, then there is simply not any interesting state
 * that can be reached through the predecessor's loop arcs, which means we can
 * break the loop just by removing those loop arcs, with no new states added.
 */
static void
breakconstraintloop(struct nfa * nfa, struct state * sinitial)
{
    struct state *s;
    struct state *shead;
    struct state *stail;
    struct state *sclone;
    struct state *nexts;
    struct arc *refarc;
    struct arc *a;
    struct arc *nexta;

    /*
     * Start by identifying which loop step we want to break at.
     * Preferentially this is one with only one constraint arc.  (XXX are
     * there any other secondary heuristics we want to use here?)  Set refarc
     * to point to the selected lone constraint arc, if there is one.
     */
    refarc = NULL;
    s = sinitial;
    do {
	nexts = s->tmp;
	assert(nexts != s);	/* should not see any one-element loops */
	if (refarc == NULL) {
	    int narcs = 0;

	    for (a = s->outs; a != NULL; a = a->outchain) {
		if (a->to == nexts && isconstraintarc(a)) {
		    refarc = a;
		    narcs++;
		}
	    }
	    assert(narcs > 0);
	    if (narcs > 1) {
		refarc = NULL;	/* multiple constraint arcs here, no good */
	    }
	}
	s = nexts;
    } while (s != sinitial);

    if (refarc) {
	/* break at the refarc */
	shead = refarc->from;
	stail = refarc->to;
	assert(stail == shead->tmp);
    } else {
	/* for lack of a better idea, break after sinitial */
	shead = sinitial;
	stail = sinitial->tmp;
    }

    /*
     * Reset the tmp fields so that we can use them for local storage in
     * clonesuccessorstates.  (findconstraintloop won't mind, since it's just
     * going to abandon its search anyway.)
     */
    for (s = nfa->states; s != NULL; s = s->next) {
	s->tmp = NULL;
    }

    /*
     * Recursively build clone state(s) as needed.
     */
    sclone = newstate(nfa);
    if (sclone == NULL) {
	assert(NISERR());
	return;
    }

    clonesuccessorstates(nfa, stail, sclone, shead, refarc,
	    NULL, NULL, nfa->nstates);

    if (NISERR()) {
	return;
    }

    /*
     * It's possible that sclone has no outarcs at all, in which case it's
     * useless.  (We don't try extremely hard to get rid of useless states
     * here, but this is an easy and fairly common case.)
     */
    if (sclone->nouts == 0) {
	freestate(nfa, sclone);
	sclone = NULL;
    }

    /*
     * Move shead's constraint-loop arcs to point to sclone, or just drop them
     * if we discovered we don't need sclone.
     */
    for (a = shead->outs; a != NULL; a = nexta) {
	nexta = a->outchain;
	if (a->to == stail && isconstraintarc(a)) {
	    if (sclone) {
		cparc(nfa, a, shead, sclone);
	    }
	    freearc(nfa, a);
	    if (NISERR()) {
		break;
	    }
	}
    }
}

/*
 * clonesuccessorstates - create a tree of constraint-arc successor states
 *
 * ssource is the state to be cloned, and sclone is the state to copy its
 * outarcs into.  sclone's inarcs, if any, should already be set up.
 *
 * spredecessor is the original predecessor state that we are trying to build
 * successors for (it may not be the immediate predecessor of ssource).
 * refarc, if not NULL, is the original constraint arc that is known to have
 * been traversed out of spredecessor to reach the successor(s).
 *
 * For each cloned successor state, we transiently create a "donemap" that is
 * a boolean array showing which source states we've already visited for this
 * clone state.  This prevents infinite recursion as well as useless repeat
 * visits to the same state subtree (which can add up fast, since typical NFAs
 * have multiple redundant arc pathways).  Each donemap is a char array
 * indexed by state number.  The donemaps are all of the same size "nstates",
 * which is nfa->nstates as of the start of the recursion.  This is enough to
 * have entries for all pre-existing states, but *not* entries for clone
 * states created during the recursion.  That's okay since we have no need to

 * mark those.
 *
 * curdonemap is NULL when recursing to a new sclone state, or sclone's
 * donemap when we are recursing without having created a new state (which we
 * do when we decide we can merge a successor state into the current clone
 * state).  outerdonemap is NULL at the top level and otherwise the parent
 * clone state's donemap.
 *
 * The successor states we create and fill here form a strict tree structure,
 * with each state having exactly one predecessor, except that the toplevel
 * state has no inarcs as yet (breakconstraintloop will add its inarcs from
 * spredecessor after we're done).  Thus, we can examine sclone's inarcs back
 * to the root, plus refarc if any, to identify the set of constraints already
 * known valid at the current point.  This allows us to avoid generating extra
 * successor states.
 */
static void
clonesuccessorstates(
    struct nfa * nfa,
    struct state * ssource,
    struct state * sclone,
    struct state * spredecessor,
    struct arc * refarc,
    char *curdonemap,
    char *outerdonemap,
    int nstates)
{
    char *donemap;
    struct arc *a;

    /* Since this is recursive, it could be driven to stack overflow */
    if (STACK_TOO_DEEP(nfa->v->re)) {
	NERR(REG_ETOOBIG);
	return;
    }

    /* If this state hasn't already got a donemap, create one */
    donemap = curdonemap;
    if (donemap == NULL) {
	donemap = (char *) MALLOC(nstates * sizeof(char));
	if (donemap == NULL) {
	    NERR(REG_ESPACE);
	    return;
	}

	if (outerdonemap != NULL) {
	    /*
	     * Not at outermost recursion level, so copy the outer level's
	     * donemap; this ensures that we see states in process of being
	     * visited at outer levels, or already merged into predecessor
	     * states, as ones we shouldn't traverse back to.
	     */
	    memcpy(donemap, outerdonemap, nstates * sizeof(char));
	} else {
	    /* At outermost level, only spredecessor is off-limits */
	    memset(donemap, 0, nstates * sizeof(char));
	    assert(spredecessor->no < nstates);
	    donemap[spredecessor->no] = 1;
	}
    }

    /* Mark ssource as visited in the donemap */
    assert(ssource->no < nstates);
    assert(donemap[ssource->no] == 0);
    donemap[ssource->no] = 1;

    /*
     * We proceed by first cloning all of ssource's outarcs, creating new
     * clone states as needed but not doing more with them than that.  Then in
     * a second pass, recurse to process the child clone states.  This allows
     * us to have only one child clone state per reachable source state, even
     * when there are multiple outarcs leading to the same state.  Also, when
     * we do visit a child state, its set of inarcs is known exactly, which
     * makes it safe to apply the constraint-is-already-checked optimization.
     * Also, this ensures that we've merged all the states we can into the
     * current clone before we recurse to any children, thus possibly saving
     * them from making extra images of those states.
     *
     * While this function runs, child clone states of the current state are
     * marked by setting their tmp fields to point to the original state they
     * were cloned from.  This makes it possible to detect multiple outarcs
     * leading to the same state, and also makes it easy to distinguish clone
     * states from original states (which will have tmp == NULL).
     */
    for (a = ssource->outs; a != NULL && !NISERR(); a = a->outchain) {
	struct state *sto = a->to;

	/*
	 * We do not consider cloning successor states that have no constraint
	 * outarcs; just link to them as-is.  They cannot be part of a
	 * constraint loop so there is no need to make copies.  In particular,
	 * this rule keeps us from trying to clone the post state, which would
	 * be a bad idea.
	 */
	if (isconstraintarc(a) && hasconstraintout(sto)) {
	    struct state *prevclone;
	    int canmerge;
	    struct arc *a2;

	    /*
	     * Back-link constraint arcs must not be followed.  Nor is there a
	     * need to revisit states previously merged into this clone.
	     */
	    assert(sto->no < nstates);
	    if (donemap[sto->no] != 0) {
		continue;
	    }

	    /*
	     * Check whether we already have a child clone state for this
	     * source state.
	     */
	    prevclone = NULL;
	    for (a2 = sclone->outs; a2 != NULL; a2 = a2->outchain) {
		if (a2->to->tmp == sto) {
		    prevclone = a2->to;
		    break;
		}
	    }

	    /*
	     * If this arc is labeled the same as refarc, or the same as any
	     * arc we must have traversed to get to sclone, then no additional
	     * constraints need to be met to get to sto, so we should just
	     * merge its outarcs into sclone.
	     */
	    if (refarc && a->type == refarc->type && a->co == refarc->co) {
		canmerge = 1;
	    } else {
		struct state *s;

		canmerge = 0;
		for (s = sclone; s->ins; s = s->ins->from) {
		    if (s->nins == 1 &&
			    a->type == s->ins->type && a->co == s->ins->co) {
			canmerge = 1;
			break;
		    }
		}
	    }

	    if (canmerge) {
		/*
		 * We can merge into sclone.  If we previously made a child
		 * clone state, drop it; there's no need to visit it.  (This
		 * can happen if ssource has multiple pathways to sto, and we
		 * only just now found one that is provably a no-op.)
		 */
		if (prevclone) {
		    dropstate(nfa, prevclone);	/* kills our outarc, too */
		}

		/* Recurse to merge sto's outarcs into sclone */
		clonesuccessorstates(nfa, sto, sclone, spredecessor, refarc,
			donemap, outerdonemap, nstates);
		/* sto should now be marked as previously visited */
		assert(NISERR() || donemap[sto->no] == 1);
	    } else if (prevclone) {
		/*
		 * We already have a clone state for this successor, so just
		 * make another arc to it.
		 */
		cparc(nfa, a, sclone, prevclone);
	    } else {
		/*
		 * We need to create a new successor clone state.
		 */
		struct state *stoclone;

		stoclone = newstate(nfa);
		if (stoclone == NULL) {
		    assert(NISERR());
		    break;
		}
		/* Mark it as to what it's a clone of */
		stoclone->tmp = sto;
		/* ... and add the outarc leading to it */
		cparc(nfa, a, sclone, stoclone);
	    }
	} else {
	    /*
	     * Non-constraint outarcs just get copied to sclone, as do outarcs
	     * leading to states with no constraint outarc.
	     */
	    cparc(nfa, a, sclone, sto);
	}
    }


    /*
     * If we are at outer level for this clone state, recurse to all its child
     * clone states, clearing their tmp fields as we go.  (If we're not
     * outermost for sclone, leave this to be done by the outer call level.)
     * Note that if we have multiple outarcs leading to the same clone state,
     * it will only be recursed-to once.
     */
    if (curdonemap == NULL) {
	for (a = sclone->outs; a != NULL && !NISERR(); a = a->outchain) {
	    struct state *stoclone = a->to;
	    struct state *sto = stoclone->tmp;

	    if (sto != NULL) {
		stoclone->tmp = NULL;
		clonesuccessorstates(nfa, sto, stoclone, spredecessor, refarc,

			NULL, donemap, nstates);
	    }
	}

	/* Don't forget to free sclone's donemap when done with it */
	FREE(donemap);
    }
}

/*
 - cleanup - clean up NFA after optimizations
 ^ static void cleanup(struct nfa *);
 */
static void
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
	    }
	}
    }
    return 0;
}

/*
 - compact - compact an NFA
 ^ static void compact(struct nfa *, struct cnfa *);
 */
static void
compact(
    struct nfa *nfa,
    struct cnfa *cnfa)
{







|







2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
	    }
	}
    }
    return 0;
}

/*
 - compact - construct the compact representation of an NFA
 ^ static void compact(struct nfa *, struct cnfa *);
 */
static void
compact(
    struct nfa *nfa,
    struct cnfa *cnfa)
{
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735



1736


1737

1738



1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753


1754


1755

1756
1757
1758
1759
1760
1761
1762
		assert(s->no != cnfa->pre);
		ca->co = (color) (cnfa->ncolors + a->co);
		ca->to = a->to->no;
		ca++;
		cnfa->flags |= HASLACONS;
		break;
	    default:
		assert(NOTREACHED);
		break;
	    }
	}
	carcsort(first, ca-1);
	ca->co = COLORLESS;
	ca->to = 0;
	ca++;
    }
    assert(ca == &cnfa->arcs[narcs]);
    assert(cnfa->nstates != 0);

    /*
     * Mark no-progress states.
     */

    for (a = nfa->pre->outs; a != NULL; a = a->outchain) {
	cnfa->stflags[a->to->no] = CNFA_NOPROGRESS;
    }
    cnfa->stflags[nfa->pre->no] = CNFA_NOPROGRESS;
}

/*
 - carcsort - sort compacted-NFA arcs by color
 * Really dumb algorithm, but if the list is long enough for that to matter,
 * you're in real trouble anyway.
 ^ static void carcsort(struct carc *, struct carc *);
 */
static void
carcsort(
    struct carc *first,



    struct carc *last)


{

    struct carc *p;



    struct carc *q;
    struct carc tmp;

    if (last - first <= 1) {
	return;
    }

    for (p = first; p <= last; p++) {
	for (q = p; q <= last; q++) {
	    if (p->co > q->co || (p->co == q->co && p->to > q->to)) {
		assert(p != q);
		tmp = *p;
		*p = *q;
		*q = tmp;
	    }


	}


    }

}

/*
 - freecnfa - free a compacted NFA
 ^ static void freecnfa(struct cnfa *);
 */
static void







|



|



















<
<





>
>
>
|
>
>
|
>
|
>
>
>
|
|
|
|
|

|
<
<
<
<
|
<
<
|
>
>
|
>
>

>







2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892


2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916




2917


2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
		assert(s->no != cnfa->pre);
		ca->co = (color) (cnfa->ncolors + a->co);
		ca->to = a->to->no;
		ca++;
		cnfa->flags |= HASLACONS;
		break;
	    default:
		NERR(REG_ASSERT);
		break;
	    }
	}
	carcsort(first, ca - first);
	ca->co = COLORLESS;
	ca->to = 0;
	ca++;
    }
    assert(ca == &cnfa->arcs[narcs]);
    assert(cnfa->nstates != 0);

    /*
     * Mark no-progress states.
     */

    for (a = nfa->pre->outs; a != NULL; a = a->outchain) {
	cnfa->stflags[a->to->no] = CNFA_NOPROGRESS;
    }
    cnfa->stflags[nfa->pre->no] = CNFA_NOPROGRESS;
}

/*
 - carcsort - sort compacted-NFA arcs by color


 ^ static void carcsort(struct carc *, struct carc *);
 */
static void
carcsort(
    struct carc *first,
    size_t n)
{
    if (n > 1) {
	qsort(first, n, sizeof(struct carc), carc_cmp);
    }
}

static int
carc_cmp(
    const void *a,
    const void *b)
{
    const struct carc *aa = (const struct carc *) a;
    const struct carc *bb = (const struct carc *) b;
  
    if (aa->co < bb->co) {
	return -1;
    }
    if (aa->co > bb->co) {




	return +1;


    }
    if (aa->to < bb->to) {
	return -1;
    }
    if (aa->to > bb->to) {
	return +1;
    }
    return 0;
}

/*
 - freecnfa - free a compacted NFA
 ^ static void freecnfa(struct cnfa *);
 */
static void
1777
1778
1779
1780
1781
1782
1783


1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800


1801

1802
1803
1804
1805
1806
1807
1808
static void
dumpnfa(
    struct nfa *nfa,
    FILE *f)
{
#ifdef REG_DEBUG
    struct state *s;



    fprintf(f, "pre %d, post %d", nfa->pre->no, nfa->post->no);
    if (nfa->bos[0] != COLORLESS) {
	fprintf(f, ", bos [%ld]", (long) nfa->bos[0]);
    }
    if (nfa->bos[1] != COLORLESS) {
	fprintf(f, ", bol [%ld]", (long) nfa->bos[1]);
    }
    if (nfa->eos[0] != COLORLESS) {
	fprintf(f, ", eos [%ld]", (long) nfa->eos[0]);
    }
    if (nfa->eos[1] != COLORLESS) {
	fprintf(f, ", eol [%ld]", (long) nfa->eos[1]);
    }
    fprintf(f, "\n");
    for (s = nfa->states; s != NULL; s = s->next) {
	dumpstate(s, f);


    }

    if (nfa->parent == NULL) {
	dumpcolors(nfa->cm, f);
    }
    fflush(f);
#endif
}








>
>

















>
>

>







2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
static void
dumpnfa(
    struct nfa *nfa,
    FILE *f)
{
#ifdef REG_DEBUG
    struct state *s;
    int nstates = 0;
    int narcs = 0;

    fprintf(f, "pre %d, post %d", nfa->pre->no, nfa->post->no);
    if (nfa->bos[0] != COLORLESS) {
	fprintf(f, ", bos [%ld]", (long) nfa->bos[0]);
    }
    if (nfa->bos[1] != COLORLESS) {
	fprintf(f, ", bol [%ld]", (long) nfa->bos[1]);
    }
    if (nfa->eos[0] != COLORLESS) {
	fprintf(f, ", eos [%ld]", (long) nfa->eos[0]);
    }
    if (nfa->eos[1] != COLORLESS) {
	fprintf(f, ", eol [%ld]", (long) nfa->eos[1]);
    }
    fprintf(f, "\n");
    for (s = nfa->states; s != NULL; s = s->next) {
	dumpstate(s, f);
	nstates++;
	narcs += s->nouts;
    }
    fprintf(f, "total of %d states, %d arcs\n", nstates, narcs);
    if (nfa->parent == NULL) {
	dumpcolors(nfa->cm, f);
    }
    fflush(f);
#endif
}

1847
1848
1849
1850
1851
1852
1853

1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876


1877
1878
1879
1880
1881
1882
1883


1884

1885

1886
1887
1888
1889
1890
1891
1892
 */
static void
dumparcs(
    struct state *s,
    FILE *f)
{
    int pos;


    assert(s->nouts > 0);
    /* printing arcs in reverse order is usually clearer */
    pos = dumprarcs(s->outs, s, f, 1);
    if (pos != 1) {
	fprintf(f, "\n");
    }
}

/*
 - dumprarcs - dump remaining outarcs, recursively, in reverse order
 ^ static int dumprarcs(struct arc *, struct state *, FILE *, int);
 */
static int			/* resulting print position */
dumprarcs(
    struct arc *a,
    struct state *s,
    FILE *f,
    int pos)			/* initial print position */
{
    if (a->outchain != NULL) {
	pos = dumprarcs(a->outchain, s, f, pos);
    }


    dumparc(a, s, f);
    if (pos == 5) {
	fprintf(f, "\n");
	pos = 1;
    } else {
	pos++;
    }


    return pos;

}


/*
 - dumparc - dump one outarc in readable form, including prefixing tab
 ^ static void dumparc(struct arc *, struct state *, FILE *);
 */
static void
dumparc(







>

<
|
|
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
|
|

>
>
|
|
|
|
|
|
|
>
>
|
>
|
>







3022
3023
3024
3025
3026
3027
3028
3029
3030

3031
3032




3033











3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
 */
static void
dumparcs(
    struct state *s,
    FILE *f)
{
    int pos;
    struct arc *a;


    /* printing oldest arcs first is usually clearer */
    a = s->outs;




    assert(a != NULL);











    while (a->outchain != NULL) {
	a = a->outchain;
    }
    pos = 1;
    do {
	dumparc(a, s, f);
	if (pos == 5) {
	    fprintf(f, "\n");
	    pos = 1;
	} else {
	    pos++;
	}
	a = a->outchainRev;
    } while (a != NULL);
    if (pos != 1) {
	fprintf(f, "\n");
    }
}

/*
 - dumparc - dump one outarc in readable form, including prefixing tab
 ^ static void dumparc(struct arc *, struct state *, FILE *);
 */
static void
dumparc(

Changes to generic/regcomp.c.

112
113
114
115
116
117
118

119
120

121
122
123
124
125




126
127

128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147






148

149
150
151
152
153
154

155
156
157
158
159
160
161
162
163
164
165
166
167
static void freenfa(struct nfa *);
static struct state *newstate(struct nfa *);
static struct state *newfstate(struct nfa *, int flag);
static void dropstate(struct nfa *, struct state *);
static void freestate(struct nfa *, struct state *);
static void destroystate(struct nfa *, struct state *);
static void newarc(struct nfa *, int, pcolor, struct state *, struct state *);

static struct arc *allocarc(struct nfa *, struct state *);
static void freearc(struct nfa *, struct arc *);

static int hasnonemptyout(struct state *);
static int nonemptyouts(struct state *);
static int nonemptyins(struct state *);
static struct arc *findarc(struct state *, int, pcolor);
static void cparc(struct nfa *, struct arc *, struct state *, struct state *);




static void moveins(struct nfa *, struct state *, struct state *);
static void copyins(struct nfa *, struct state *, struct state *, int);

static void moveouts(struct nfa *, struct state *, struct state *);
static void copyouts(struct nfa *, struct state *, struct state *, int);
static void cloneouts(struct nfa *, struct state *, struct state *, struct state *, int);
static void delsub(struct nfa *, struct state *, struct state *);
static void deltraverse(struct nfa *, struct state *, struct state *);
static void dupnfa(struct nfa *, struct state *, struct state *, struct state *, struct state *);
static void duptraverse(struct nfa *, struct state *, struct state *, int);
static void cleartraverse(struct nfa *, struct state *);
static void specialcolors(struct nfa *);
static long optimize(struct nfa *, FILE *);
static void pullback(struct nfa *, FILE *);
static int pull(struct nfa *, struct arc *);
static void pushfwd(struct nfa *, FILE *);
static int push(struct nfa *, struct arc *);
#define	INCOMPATIBLE	1	/* destroys arc */
#define	SATISFIED	2	/* constraint satisfied */
#define	COMPATIBLE	3	/* compatible but not satisfied yet */
static int combine(struct arc *, struct arc *);
static void fixempties(struct nfa *, FILE *);
static struct state *emptyreachable(struct state *, struct state *);






static void replaceempty(struct nfa *, struct state *, struct state *);

static void cleanup(struct nfa *);
static void markreachable(struct nfa *, struct state *, struct state *, struct state *);
static void markcanreach(struct nfa *, struct state *, struct state *, struct state *);
static long analyze(struct nfa *);
static void compact(struct nfa *, struct cnfa *);
static void carcsort(struct carc *, struct carc *);

static void freecnfa(struct cnfa *);
static void dumpnfa(struct nfa *, FILE *);
#ifdef REG_DEBUG
static void dumpstate(struct state *, FILE *);
static void dumparcs(struct state *, FILE *);
static int dumprarcs(struct arc *, struct state *, FILE *, int);
static void dumparc(struct arc *, struct state *, FILE *);
#endif
static void dumpcnfa(struct cnfa *, FILE *);
#ifdef REG_DEBUG
static void dumpcstate(int, struct cnfa *, FILE *);
#endif
/* === regc_cvec.c === */







>


>

<
<


>
>
>
>

|
>

|









|

|





|
>
>
>
>
>
>
|
>





|
>





<







112
113
114
115
116
117
118
119
120
121
122
123


124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172

173
174
175
176
177
178
179
static void freenfa(struct nfa *);
static struct state *newstate(struct nfa *);
static struct state *newfstate(struct nfa *, int flag);
static void dropstate(struct nfa *, struct state *);
static void freestate(struct nfa *, struct state *);
static void destroystate(struct nfa *, struct state *);
static void newarc(struct nfa *, int, pcolor, struct state *, struct state *);
static void createarc(struct nfa *, int, pcolor, struct state *, struct state *);
static struct arc *allocarc(struct nfa *, struct state *);
static void freearc(struct nfa *, struct arc *);
static void changearctarget(struct arc *, struct state *);
static int hasnonemptyout(struct state *);


static struct arc *findarc(struct state *, int, pcolor);
static void cparc(struct nfa *, struct arc *, struct state *, struct state *);
static void sortins(struct nfa *, struct state *);
static int sortins_cmp(const void *, const void *);
static void sortouts(struct nfa *, struct state *);
static int sortouts_cmp(const void *, const void *);
static void moveins(struct nfa *, struct state *, struct state *);
static void copyins(struct nfa *, struct state *, struct state *);
static void mergeins(struct nfa *, struct state *, struct arc **, int);
static void moveouts(struct nfa *, struct state *, struct state *);
static void copyouts(struct nfa *, struct state *, struct state *);
static void cloneouts(struct nfa *, struct state *, struct state *, struct state *, int);
static void delsub(struct nfa *, struct state *, struct state *);
static void deltraverse(struct nfa *, struct state *, struct state *);
static void dupnfa(struct nfa *, struct state *, struct state *, struct state *, struct state *);
static void duptraverse(struct nfa *, struct state *, struct state *, int);
static void cleartraverse(struct nfa *, struct state *);
static void specialcolors(struct nfa *);
static long optimize(struct nfa *, FILE *);
static void pullback(struct nfa *, FILE *);
static int pull(struct nfa *, struct arc *, struct state **);
static void pushfwd(struct nfa *, FILE *);
static int push(struct nfa *, struct arc *, struct state **);
#define	INCOMPATIBLE	1	/* destroys arc */
#define	SATISFIED	2	/* constraint satisfied */
#define	COMPATIBLE	3	/* compatible but not satisfied yet */
static int combine(struct arc *, struct arc *);
static void fixempties(struct nfa *, FILE *);
static struct state *emptyreachable(struct nfa *, struct state *,
			struct state *, struct arc **);
static int	isconstraintarc(struct arc *);
static int	hasconstraintout(struct state *);
static void fixconstraintloops(struct nfa *, FILE *);
static int	findconstraintloop(struct nfa *, struct state *);
static void breakconstraintloop(struct nfa *, struct state *);
static void clonesuccessorstates(struct nfa *, struct state *, struct state *,
		 struct state *, struct arc *, char *, char *, int);
static void cleanup(struct nfa *);
static void markreachable(struct nfa *, struct state *, struct state *, struct state *);
static void markcanreach(struct nfa *, struct state *, struct state *, struct state *);
static long analyze(struct nfa *);
static void compact(struct nfa *, struct cnfa *);
static void carcsort(struct carc *, size_t);
static int carc_cmp(const void *, const void *);
static void freecnfa(struct cnfa *);
static void dumpnfa(struct nfa *, FILE *);
#ifdef REG_DEBUG
static void dumpstate(struct state *, FILE *);
static void dumparcs(struct state *, FILE *);

static void dumparc(struct arc *, struct state *, FILE *);
#endif
static void dumpcnfa(struct cnfa *, FILE *);
#ifdef REG_DEBUG
static void dumpcstate(int, struct cnfa *, FILE *);
#endif
/* === regc_cvec.c === */
208
209
210
211
212
213
214

215
216
217
218
219
220
221
    struct subre *treechain;	/* all tree nodes allocated */
    struct subre *treefree;	/* any free tree nodes */
    int ntree;			/* number of tree nodes, plus one */
    struct cvec *cv;		/* interface cvec */
    struct cvec *cv2;		/* utility cvec */
    struct subre *lacons;	/* lookahead-constraint vector */
    int nlacons;		/* size of lacons */

};

/* parsing macros; most know that `v' is the struct vars pointer */
#define	NEXT()	(next(v))		/* advance by one token */
#define	SEE(t)	(v->nexttype == (t))	/* is next token this? */
#define	EAT(t)	(SEE(t) && next(v))	/* if next is this, swallow it */
#define	VISERR(vv)	((vv)->err != 0)/* have we seen an error yet? */







>







220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
    struct subre *treechain;	/* all tree nodes allocated */
    struct subre *treefree;	/* any free tree nodes */
    int ntree;			/* number of tree nodes, plus one */
    struct cvec *cv;		/* interface cvec */
    struct cvec *cv2;		/* utility cvec */
    struct subre *lacons;	/* lookahead-constraint vector */
    int nlacons;		/* size of lacons */
    size_t spaceused;		/* approx. space used for compilation */
};

/* parsing macros; most know that `v' is the struct vars pointer */
#define	NEXT()	(next(v))		/* advance by one token */
#define	SEE(t)	(v->nexttype == (t))	/* is next token this? */
#define	EAT(t)	(SEE(t) && next(v))	/* if next is this, swallow it */
#define	VISERR(vv)	((vv)->err != 0)/* have we seen an error yet? */
319
320
321
322
323
324
325

326
327
328
329
330
331
332
    v->tree = NULL;
    v->treechain = NULL;
    v->treefree = NULL;
    v->cv = NULL;
    v->cv2 = NULL;
    v->lacons = NULL;
    v->nlacons = 0;

    re->re_magic = REMAGIC;
    re->re_info = 0;		/* bits get set during parse */
    re->re_csize = sizeof(chr);
    re->re_guts = NULL;
    re->re_fns = VS(&functions);

    /*







>







332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
    v->tree = NULL;
    v->treechain = NULL;
    v->treefree = NULL;
    v->cv = NULL;
    v->cv2 = NULL;
    v->lacons = NULL;
    v->nlacons = 0;
    v->spaceused = 0;
    re->re_magic = REMAGIC;
    re->re_info = 0;		/* bits get set during parse */
    re->re_csize = sizeof(chr);
    re->re_guts = NULL;
    re->re_fns = VS(&functions);

    /*
607
608
609
610
611
612
613
614
615

616
617
618
619
620
621
622

    /*
     * Do the splits.
     */

    for (s=slist ; s!=NULL ; s=s2) {
	s2 = newstate(nfa);

	copyouts(nfa, s, s2, 1);

	for (a=s->ins ; a!=NULL ; a=b) {
	    b = a->inchain;

	    if (a->from != pre) {
		cparc(nfa, a, a->from, s2);
		freearc(nfa, a);
	    }







|
|
>







621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637

    /*
     * Do the splits.
     */

    for (s=slist ; s!=NULL ; s=s2) {
	s2 = newstate(nfa);
	NOERR();
	copyouts(nfa, s, s2);
	NOERR();
	for (a=s->ins ; a!=NULL ; a=b) {
	    b = a->inchain;

	    if (a->from != pre) {
		cparc(nfa, a, a->from, s2);
		freearc(nfa, a);
	    }
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086

    fprintf(f, "\n\n\n========= DUMP ==========\n");
    fprintf(f, "nsub %d, info 0%lo, csize %d, ntree %d\n",
	    (int) re->re_nsub, re->re_info, re->re_csize, g->ntree);

    dumpcolors(&g->cmap, f);
    if (!NULLCNFA(g->search)) {
	printf("\nsearch:\n");
	dumpcnfa(&g->search, f);
    }
    for (i = 1; i < g->nlacons; i++) {
	fprintf(f, "\nla%d (%s):\n", i,
		(g->lacons[i].subno) ? "positive" : "negative");
	dumpcnfa(&g->lacons[i].cnfa, f);
    }







|







2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101

    fprintf(f, "\n\n\n========= DUMP ==========\n");
    fprintf(f, "nsub %d, info 0%lo, csize %d, ntree %d\n",
	    (int) re->re_nsub, re->re_info, re->re_csize, g->ntree);

    dumpcolors(&g->cmap, f);
    if (!NULLCNFA(g->search)) {
	fprintf(f, "\nsearch:\n");
	dumpcnfa(&g->search, f);
    }
    for (i = 1; i < g->nlacons; i++) {
	fprintf(f, "\nla%d (%s):\n", i,
		(g->lacons[i].subno) ? "positive" : "negative");
	dumpcnfa(&g->lacons[i].cnfa, f);
    }

Changes to generic/regerrs.h.

12
13
14
15
16
17
18
19
20
{ REG_ERANGE,	"REG_ERANGE",	"invalid character range" },
{ REG_ESPACE,	"REG_ESPACE",	"out of memory" },
{ REG_BADRPT,	"REG_BADRPT",	"quantifier operand invalid" },
{ REG_ASSERT,	"REG_ASSERT",	"\"can't happen\" -- you found a bug" },
{ REG_INVARG,	"REG_INVARG",	"invalid argument to regex function" },
{ REG_MIXED,	"REG_MIXED",	"character widths of regex and string differ" },
{ REG_BADOPT,	"REG_BADOPT",	"invalid embedded option" },
{ REG_ETOOBIG,	"REG_ETOOBIG",	"nfa has too many states" },
{ REG_ECOLORS,	"REG_ECOLORS",	"too many colors" },







|

12
13
14
15
16
17
18
19
20
{ REG_ERANGE,	"REG_ERANGE",	"invalid character range" },
{ REG_ESPACE,	"REG_ESPACE",	"out of memory" },
{ REG_BADRPT,	"REG_BADRPT",	"quantifier operand invalid" },
{ REG_ASSERT,	"REG_ASSERT",	"\"can't happen\" -- you found a bug" },
{ REG_INVARG,	"REG_INVARG",	"invalid argument to regex function" },
{ REG_MIXED,	"REG_MIXED",	"character widths of regex and string differ" },
{ REG_BADOPT,	"REG_BADOPT",	"invalid embedded option" },
{ REG_ETOOBIG,	"REG_ETOOBIG",	"regular expression is too complex" },
{ REG_ECOLORS,	"REG_ECOLORS",	"too many colors" },

Changes to generic/regex.h.

276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
#define	REG_ERANGE	11	/* invalid character range */
#define	REG_ESPACE	12	/* out of memory */
#define	REG_BADRPT	13	/* quantifier operand invalid */
#define	REG_ASSERT	15	/* "can't happen" -- you found a bug */
#define	REG_INVARG	16	/* invalid argument to regex function */
#define	REG_MIXED	17	/* character widths of regex and string differ */
#define	REG_BADOPT	18	/* invalid embedded option */
#define	REG_ETOOBIG	19	/* nfa has too many states */
#define	REG_ECOLORS	20	/* too many colors */
/* two specials for debugging and testing */
#define	REG_ATOI	101	/* convert error-code name to number */
#define	REG_ITOA	102	/* convert error-code number to name */

/*
 * the prototypes, as possibly munched by regfwd







|







276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
#define	REG_ERANGE	11	/* invalid character range */
#define	REG_ESPACE	12	/* out of memory */
#define	REG_BADRPT	13	/* quantifier operand invalid */
#define	REG_ASSERT	15	/* "can't happen" -- you found a bug */
#define	REG_INVARG	16	/* invalid argument to regex function */
#define	REG_MIXED	17	/* character widths of regex and string differ */
#define	REG_BADOPT	18	/* invalid embedded option */
#define	REG_ETOOBIG	19	/* regular expression is too complex */
#define	REG_ECOLORS	20	/* too many colors */
/* two specials for debugging and testing */
#define	REG_ATOI	101	/* convert error-code name to number */
#define	REG_ITOA	102	/* convert error-code number to name */

/*
 * the prototypes, as possibly munched by regfwd

Changes to generic/regguts.h.

250
251
252
253
254
255
256

257
258

259
260
261
262
263
264
265
266

struct arc {
    int type;			/* 0 if free, else an NFA arc type code */
    color co;
    struct state *from;		/* where it's from (and contained within) */
    struct state *to;		/* where it's to */
    struct arc *outchain;	/* link in *from's outs chain or free chain */

#define	freechain	outchain
    struct arc *inchain;	/* link in *to's ins chain */

    struct arc *colorchain;	/* link in color's arc chain */
    struct arc *colorchainRev;	/* back-link in color's arc chain */
};

struct arcbatch {		/* for bulk allocation of arcs */
    struct arcbatch *next;
#define	ABSIZE	10
    struct arc a[ABSIZE];







>
|
|
>
|







250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268

struct arc {
    int type;			/* 0 if free, else an NFA arc type code */
    color co;
    struct state *from;		/* where it's from (and contained within) */
    struct state *to;		/* where it's to */
    struct arc *outchain;	/* link in *from's outs chain or free chain */
    struct arc *outchainRev;	/* back-link in *from's outs chain */
#define	freechain outchain	/* we do not maintain "freechainRev" */
    struct arc *inchain;	/* *to's ins chain */
    struct arc *inchainRev;	/* back-link in *to's ins chain */
    struct arc *colorchain;	/* color's arc chain */
    struct arc *colorchainRev;	/* back-link in color's arc chain */
};

struct arcbatch {		/* for bulk allocation of arcs */
    struct arcbatch *next;
#define	ABSIZE	10
    struct arc a[ABSIZE];
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
    int nstates;		/* for numbering states */
    struct state *states;	/* state-chain header */
    struct state *slast;	/* tail of the chain */
    struct state *free;		/* free list */
    struct colormap *cm;	/* the color map */
    color bos[2];		/* colors, if any, assigned to BOS and BOL */
    color eos[2];		/* colors, if any, assigned to EOS and EOL */
    size_t size;		/* Current NFA size; differs from nstates as
				 * it also counts the number of states created
				 * by children of this state. */
    struct vars *v;		/* simplifies compile error reporting */
    struct nfa *parent;		/* parent NFA, if any */
};

/*
 * definitions for compacted NFA
 *







<
<
<







292
293
294
295
296
297
298



299
300
301
302
303
304
305
    int nstates;		/* for numbering states */
    struct state *states;	/* state-chain header */
    struct state *slast;	/* tail of the chain */
    struct state *free;		/* free list */
    struct colormap *cm;	/* the color map */
    color bos[2];		/* colors, if any, assigned to BOS and BOL */
    color eos[2];		/* colors, if any, assigned to EOS and EOL */



    struct vars *v;		/* simplifies compile error reporting */
    struct nfa *parent;		/* parent NFA, if any */
};

/*
 * definitions for compacted NFA
 *
336
337
338
339
340
341
342

343




344
345
346
347

348
349
350
351
352
353
354
    /* states[n] are pointers into a single malloc'd array of arcs */
    struct carc *arcs;		/* the area for the lists */
};
#define	ZAPCNFA(cnfa)	((cnfa).nstates = 0)
#define	NULLCNFA(cnfa)	((cnfa).nstates == 0)

/*

 * Used to limit the maximum NFA size to something sane. [Bug 1810264]




 */

#ifndef REG_MAX_STATES
#   define REG_MAX_STATES	100000

#endif

/*
 * subexpression tree
 *
 * "op" is one of:
 *	'='  plain regex without interesting substructure (implemented as DFA)







>
|
>
>
>
>

<
|
|
>







335
336
337
338
339
340
341
342
343
344
345
346
347
348

349
350
351
352
353
354
355
356
357
358
    /* states[n] are pointers into a single malloc'd array of arcs */
    struct carc *arcs;		/* the area for the lists */
};
#define	ZAPCNFA(cnfa)	((cnfa).nstates = 0)
#define	NULLCNFA(cnfa)	((cnfa).nstates == 0)

/*
 * This symbol limits the transient heap space used by the regex compiler,
 * and thereby also the maximum complexity of NFAs that we'll deal with.
 * Currently we only count NFA states and arcs against this; the other
 * transient data is generally not large enough to notice compared to those.
 * Note that we do not charge anything for the final output data structures
 * (the compacted NFA and the colormap).
 */

#ifndef REG_MAX_COMPILE_SPACE
#define REG_MAX_COMPILE_SPACE  \
	(100000 * sizeof(struct state) + 100000 * sizeof(struct arcbatch))
#endif

/*
 * subexpression tree
 *
 * "op" is one of:
 *	'='  plain regex without interesting substructure (implemented as DFA)

Changes to tests/reg.test.

1084
1085
1086
1087
1088
1089
1090

1091
1092
1093
1094
1095
1096
1097
1098
test reg-33.13 {Bug 1810264 - infinite loop} {
    regexp {($|^)*} {x}
} 1
# Some environments have small default stack sizes. [Bug 1905562]
test reg-33.14 {Bug 1810264 - super-expensive expression} nonPortable {
    regexp {(x{200}){200}$y} {x}
} 0

test reg-33.15 {Bug 3603557 - an "in the wild" RE} {
    lindex [regexp -expanded -about {
	^TETRA_MODE_CMD				# Message Type
	([[:blank:]]+)				# Pad
	(ETS_1_1|ETS_1_2|ETS_2_2)		# SystemCode
	([[:blank:]]+)				# Pad
	(CONTINUOUS|CARRIER|MCCH|TRAFFIC)	# SharingMode
	([[:blank:]]+)				# Pad







>
|







1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
test reg-33.13 {Bug 1810264 - infinite loop} {
    regexp {($|^)*} {x}
} 1
# Some environments have small default stack sizes. [Bug 1905562]
test reg-33.14 {Bug 1810264 - super-expensive expression} nonPortable {
    regexp {(x{200}){200}$y} {x}
} 0

test reg-33.15.1 {Bug 3603557 - an "in the wild" RE} {
    lindex [regexp -expanded -about {
	^TETRA_MODE_CMD				# Message Type
	([[:blank:]]+)				# Pad
	(ETS_1_1|ETS_1_2|ETS_2_2)		# SystemCode
	([[:blank:]]+)				# Pad
	(CONTINUOUS|CARRIER|MCCH|TRAFFIC)	# SharingMode
	([[:blank:]]+)				# Pad
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169




















































1170
1171
1172
1173
1174
1175
1176
	([0-7])					# MinPriority
	([[:blank:]]+)				# Pad
	(PASS|TRUE|FAIL|FALSE)			# ExtdSrvcsEnabled
	([[:blank:]]+)				# Pad
	(.*)					# ConditionalFields
    }] 0
} 68
test reg-33.16 {Bug [8d2c0da36d]- another "in the wild" RE} {
    lindex [regexp -about "^MRK:client1: =1339 14HKelly Talisman 10011000 (\[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]*) \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 8 0 8 0 0 0 77 77 1 1 2 0 11 { 1 3 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 13HC6 My Creator 2 3 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 31HC7 Slightly offensive name, huh 3 8 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 23HE-mail:[email protected] 4 9 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 17Hcompface must die 5 10 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 0 3HAir 6 12 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 14HPGP public key 7 13 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 [email protected] 8 30 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 0 12H2 text/plain 9 30 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 0 13H2 x-kom/basic 10 33 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 1H0 11 14 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 1H3 }\r?"] 0
} 1





















































# cleanup
::tcltest::cleanupTests
return

# Local Variables:
# mode: tcl
# End:







|


|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
	([0-7])					# MinPriority
	([[:blank:]]+)				# Pad
	(PASS|TRUE|FAIL|FALSE)			# ExtdSrvcsEnabled
	([[:blank:]]+)				# Pad
	(.*)					# ConditionalFields
    }] 0
} 68
test reg-33.16.1 {Bug [8d2c0da36d]- another "in the wild" RE} {
    lindex [regexp -about "^MRK:client1: =1339 14HKelly Talisman 10011000 (\[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]*) \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 8 0 8 0 0 0 77 77 1 1 2 0 11 { 1 3 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 13HC6 My Creator 2 3 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 31HC7 Slightly offensive name, huh 3 8 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 23HE-mail:[email protected] 4 9 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 17Hcompface must die 5 10 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 0 3HAir 6 12 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 14HPGP public key 7 13 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 [email protected] 8 30 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 0 12H2 text/plain 9 30 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 0 13H2 x-kom/basic 10 33 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 1H0 11 14 8 \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* \[0-9\]* 00000000 1 1H3 }\r?"] 0
} 1

test reg-33.15 {constraint fixes} {
    regexp {(^)+^} x
} 1
test reg-33.16 {constraint fixes} {
    regexp {($^)+} x
} 0
test reg-33.17 {constraint fixes} {
    regexp {(^$)*} x
} 1
test reg-33.18 {constraint fixes} {
    regexp {(^(?!aa))+} {aa bb cc}
} 0
test reg-33.19 {constraint fixes} {
    regexp {(^(?!aa)(?!bb)(?!cc))+} {aa x}
} 0
test reg-33.20 {constraint fixes} {
    regexp {(^(?!aa)(?!bb)(?!cc))+} {bb x}
} 0
test reg-33.21 {constraint fixes} {
    regexp {(^(?!aa)(?!bb)(?!cc))+} {cc x}
} 0
test reg-33.22 {constraint fixes} {
    regexp {(^(?!aa)(?!bb)(?!cc))+} {dd x}
} 1

test reg-33.23 {} {
    regexp {abcd(\m)+xyz} x
} 0
test reg-33.24 {} {
    regexp {abcd(\m)+xyz} a
} 0
test reg-33.25 {} {
    regexp {^abcd*(((((^(a c(e?d)a+|)+|)+|)+|)+|a)+|)} x
} 0
test reg-33.26 {} {
    regexp {a^(^)bcd*xy(((((($a+|)+|)+|)+$|)+|)+|)^$} x
} 0
test reg-33.27 {} {
    regexp {xyz(\Y\Y)+} x
} 0
test reg-33.28 {} {
    regexp {x|(?:\M)+} x
} 1
test reg-33.29 {} {
    # This is near the limits of the RE engine
    regexp [string repeat x*y*z* 480] x
} 1

test reg-33.30 {Bug 1080042} {
    regexp {(\Y)+} foo
} 1

# cleanup
::tcltest::cleanupTests
return

# Local Variables:
# mode: tcl
# End:

Changes to tests/regexp.test.

860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
	[a 667]([a 55])[a 668]([a 55])[a 669]([a 55])[a 668]([a 55]) \
	[a 671]([a 55])[a 669]([a 55])[a 668]([a 55])[a 669]([a 55]) \
	[a 669]([a 55])[a 669]([a 55])[a 668]([a 55])[a 669]([a 55]) \
	[a 668]([a 55])[a 710]([a 55])[a 668]([a 55])[a 668]([a 55]) \
	[a 668]([a 55])[a 668]([a 55])[a 668]([a 55])[a 511]] {}] a
} -cleanup {
    rename a {}
} -returnCodes 1 -result {couldn't compile regular expression pattern: nfa has too many states}
test regexp-22.5 {Bug 3610026} -setup {
    set e {}
    set cp 99
    while {$cp < 32864} {
	append e [format %c [incr cp]]
    }
} -body {







|







860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
	[a 667]([a 55])[a 668]([a 55])[a 669]([a 55])[a 668]([a 55]) \
	[a 671]([a 55])[a 669]([a 55])[a 668]([a 55])[a 669]([a 55]) \
	[a 669]([a 55])[a 669]([a 55])[a 668]([a 55])[a 669]([a 55]) \
	[a 668]([a 55])[a 710]([a 55])[a 668]([a 55])[a 668]([a 55]) \
	[a 668]([a 55])[a 668]([a 55])[a 668]([a 55])[a 511]] {}] a
} -cleanup {
    rename a {}
} -returnCodes 1 -match glob -result {couldn't compile regular expression pattern: *}
test regexp-22.5 {Bug 3610026} -setup {
    set e {}
    set cp 99
    while {$cp < 32864} {
	append e [format %c [incr cp]]
    }
} -body {