2 * LibXDiff by Davide Libenzi ( File Differential Library )
3 * Copyright (C) 2003 Davide Libenzi
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2.1 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, see
17 * <http://www.gnu.org/licenses/>.
19 * Davide Libenzi <davidel@xmailserver.org>
26 #define XDL_KPDIS_RUN 4
27 #define XDL_MAX_EQLIMIT 1024
28 #define XDL_SIMSCAN_WINDOW 100
29 #define XDL_GUESS_NLINES1 256
30 #define XDL_GUESS_NLINES2 20
33 typedef struct s_xdlclass
{
34 struct s_xdlclass
*next
;
42 typedef struct s_xdlclassifier
{
56 static int xdl_init_classifier(xdlclassifier_t
*cf
, long size
, long flags
);
57 static void xdl_free_classifier(xdlclassifier_t
*cf
);
58 static int xdl_classify_record(unsigned int pass
, xdlclassifier_t
*cf
, xrecord_t
**rhash
,
59 unsigned int hbits
, xrecord_t
*rec
);
60 static int xdl_prepare_ctx(unsigned int pass
, mmfile_t
*mf
, long narec
, xpparam_t
const *xpp
,
61 xdlclassifier_t
*cf
, xdfile_t
*xdf
);
62 static void xdl_free_ctx(xdfile_t
*xdf
);
63 static int xdl_clean_mmatch(char const *dis
, long i
, long s
, long e
);
64 static int xdl_cleanup_records(xdlclassifier_t
*cf
, xdfile_t
*xdf1
, xdfile_t
*xdf2
);
65 static int xdl_trim_ends(xdfile_t
*xdf1
, xdfile_t
*xdf2
);
66 static int xdl_optimize_ctxs(xdlclassifier_t
*cf
, xdfile_t
*xdf1
, xdfile_t
*xdf2
);
71 static int xdl_init_classifier(xdlclassifier_t
*cf
, long size
, long flags
) {
74 cf
->hbits
= xdl_hashbits((unsigned int) size
);
75 cf
->hsize
= 1 << cf
->hbits
;
77 if (xdl_cha_init(&cf
->ncha
, sizeof(xdlclass_t
), size
/ 4 + 1) < 0) {
81 if (!XDL_CALLOC_ARRAY(cf
->rchash
, cf
->hsize
)) {
83 xdl_cha_free(&cf
->ncha
);
88 if (!XDL_ALLOC_ARRAY(cf
->rcrecs
, cf
->alloc
)) {
91 xdl_cha_free(&cf
->ncha
);
101 static void xdl_free_classifier(xdlclassifier_t
*cf
) {
103 xdl_free(cf
->rcrecs
);
104 xdl_free(cf
->rchash
);
105 xdl_cha_free(&cf
->ncha
);
109 static int xdl_classify_record(unsigned int pass
, xdlclassifier_t
*cf
, xrecord_t
**rhash
,
110 unsigned int hbits
, xrecord_t
*rec
) {
116 hi
= (long) XDL_HASHLONG(rec
->ha
, cf
->hbits
);
117 for (rcrec
= cf
->rchash
[hi
]; rcrec
; rcrec
= rcrec
->next
)
118 if (rcrec
->ha
== rec
->ha
&&
119 xdl_recmatch(rcrec
->line
, rcrec
->size
,
120 rec
->ptr
, rec
->size
, cf
->flags
))
124 if (!(rcrec
= xdl_cha_alloc(&cf
->ncha
))) {
128 rcrec
->idx
= cf
->count
++;
129 if (XDL_ALLOC_GROW(cf
->rcrecs
, cf
->count
, cf
->alloc
))
131 cf
->rcrecs
[rcrec
->idx
] = rcrec
;
133 rcrec
->size
= rec
->size
;
135 rcrec
->len1
= rcrec
->len2
= 0;
136 rcrec
->next
= cf
->rchash
[hi
];
137 cf
->rchash
[hi
] = rcrec
;
140 (pass
== 1) ? rcrec
->len1
++ : rcrec
->len2
++;
142 rec
->ha
= (unsigned long) rcrec
->idx
;
144 hi
= (long) XDL_HASHLONG(rec
->ha
, hbits
);
145 rec
->next
= rhash
[hi
];
152 static int xdl_prepare_ctx(unsigned int pass
, mmfile_t
*mf
, long narec
, xpparam_t
const *xpp
,
153 xdlclassifier_t
*cf
, xdfile_t
*xdf
) {
155 long nrec
, hsize
, bsize
;
157 char const *blk
, *cur
, *top
, *prev
;
171 if (xdl_cha_init(&xdf
->rcha
, sizeof(xrecord_t
), narec
/ 4 + 1) < 0)
173 if (!XDL_ALLOC_ARRAY(recs
, narec
))
176 hbits
= xdl_hashbits((unsigned int) narec
);
178 if (!XDL_CALLOC_ARRAY(rhash
, hsize
))
182 if ((cur
= blk
= xdl_mmfile_first(mf
, &bsize
))) {
183 for (top
= blk
+ bsize
; cur
< top
; ) {
185 hav
= xdl_hash_record(&cur
, top
, xpp
->flags
);
186 if (XDL_ALLOC_GROW(recs
, nrec
+ 1, narec
))
188 if (!(crec
= xdl_cha_alloc(&xdf
->rcha
)))
191 crec
->size
= (long) (cur
- prev
);
194 if (xdl_classify_record(pass
, cf
, rhash
, hbits
, crec
) < 0)
199 if (!XDL_CALLOC_ARRAY(rchg
, nrec
+ 2))
202 if ((XDF_DIFF_ALG(xpp
->flags
) != XDF_PATIENCE_DIFF
) &&
203 (XDF_DIFF_ALG(xpp
->flags
) != XDF_HISTOGRAM_DIFF
)) {
204 if (!XDL_ALLOC_ARRAY(rindex
, nrec
+ 1))
206 if (!XDL_ALLOC_ARRAY(ha
, nrec
+ 1))
214 xdf
->rchg
= rchg
+ 1;
215 xdf
->rindex
= rindex
;
219 xdf
->dend
= nrec
- 1;
229 xdl_cha_free(&xdf
->rcha
);
234 static void xdl_free_ctx(xdfile_t
*xdf
) {
236 xdl_free(xdf
->rhash
);
237 xdl_free(xdf
->rindex
);
238 xdl_free(xdf
->rchg
- 1);
241 xdl_cha_free(&xdf
->rcha
);
245 int xdl_prepare_env(mmfile_t
*mf1
, mmfile_t
*mf2
, xpparam_t
const *xpp
,
247 long enl1
, enl2
, sample
;
250 memset(&cf
, 0, sizeof(cf
));
253 * For histogram diff, we can afford a smaller sample size and
254 * thus a poorer estimate of the number of lines, as the hash
255 * table (rhash) won't be filled up/grown. The number of lines
256 * (nrecs) will be updated correctly anyway by
259 sample
= (XDF_DIFF_ALG(xpp
->flags
) == XDF_HISTOGRAM_DIFF
260 ? XDL_GUESS_NLINES2
: XDL_GUESS_NLINES1
);
262 enl1
= xdl_guess_lines(mf1
, sample
) + 1;
263 enl2
= xdl_guess_lines(mf2
, sample
) + 1;
265 if (xdl_init_classifier(&cf
, enl1
+ enl2
+ 1, xpp
->flags
) < 0)
268 if (xdl_prepare_ctx(1, mf1
, enl1
, xpp
, &cf
, &xe
->xdf1
) < 0) {
270 xdl_free_classifier(&cf
);
273 if (xdl_prepare_ctx(2, mf2
, enl2
, xpp
, &cf
, &xe
->xdf2
) < 0) {
275 xdl_free_ctx(&xe
->xdf1
);
276 xdl_free_classifier(&cf
);
280 if ((XDF_DIFF_ALG(xpp
->flags
) != XDF_PATIENCE_DIFF
) &&
281 (XDF_DIFF_ALG(xpp
->flags
) != XDF_HISTOGRAM_DIFF
) &&
282 xdl_optimize_ctxs(&cf
, &xe
->xdf1
, &xe
->xdf2
) < 0) {
284 xdl_free_ctx(&xe
->xdf2
);
285 xdl_free_ctx(&xe
->xdf1
);
286 xdl_free_classifier(&cf
);
290 xdl_free_classifier(&cf
);
296 void xdl_free_env(xdfenv_t
*xe
) {
298 xdl_free_ctx(&xe
->xdf2
);
299 xdl_free_ctx(&xe
->xdf1
);
303 static int xdl_clean_mmatch(char const *dis
, long i
, long s
, long e
) {
304 long r
, rdis0
, rpdis0
, rdis1
, rpdis1
;
307 * Limits the window the is examined during the similar-lines
308 * scan. The loops below stops when dis[i - r] == 1 (line that
309 * has no match), but there are corner cases where the loop
310 * proceed all the way to the extremities by causing huge
311 * performance penalties in case of big files.
313 if (i
- s
> XDL_SIMSCAN_WINDOW
)
314 s
= i
- XDL_SIMSCAN_WINDOW
;
315 if (e
- i
> XDL_SIMSCAN_WINDOW
)
316 e
= i
+ XDL_SIMSCAN_WINDOW
;
319 * Scans the lines before 'i' to find a run of lines that either
320 * have no match (dis[j] == 0) or have multiple matches (dis[j] > 1).
321 * Note that we always call this function with dis[i] > 1, so the
322 * current line (i) is already a multimatch line.
324 for (r
= 1, rdis0
= 0, rpdis0
= 1; (i
- r
) >= s
; r
++) {
327 else if (dis
[i
- r
] == 2)
333 * If the run before the line 'i' found only multimatch lines, we
334 * return 0 and hence we don't make the current line (i) discarded.
335 * We want to discard multimatch lines only when they appear in the
336 * middle of runs with nomatch lines (dis[j] == 0).
340 for (r
= 1, rdis1
= 0, rpdis1
= 1; (i
+ r
) <= e
; r
++) {
343 else if (dis
[i
+ r
] == 2)
349 * If the run after the line 'i' found only multimatch lines, we
350 * return 0 and hence we don't make the current line (i) discarded.
357 return rpdis1
* XDL_KPDIS_RUN
< (rpdis1
+ rdis1
);
362 * Try to reduce the problem complexity, discard records that have no
363 * matches on the other file. Also, lines that have multiple matches
364 * might be potentially discarded if they happear in a run of discardable.
366 static int xdl_cleanup_records(xdlclassifier_t
*cf
, xdfile_t
*xdf1
, xdfile_t
*xdf2
) {
367 long i
, nm
, nreff
, mlim
;
370 char *dis
, *dis1
, *dis2
;
372 if (!XDL_CALLOC_ARRAY(dis
, xdf1
->nrec
+ xdf2
->nrec
+ 2))
375 dis2
= dis1
+ xdf1
->nrec
+ 1;
377 if ((mlim
= xdl_bogosqrt(xdf1
->nrec
)) > XDL_MAX_EQLIMIT
)
378 mlim
= XDL_MAX_EQLIMIT
;
379 for (i
= xdf1
->dstart
, recs
= &xdf1
->recs
[xdf1
->dstart
]; i
<= xdf1
->dend
; i
++, recs
++) {
380 rcrec
= cf
->rcrecs
[(*recs
)->ha
];
381 nm
= rcrec
? rcrec
->len2
: 0;
382 dis1
[i
] = (nm
== 0) ? 0: (nm
>= mlim
) ? 2: 1;
385 if ((mlim
= xdl_bogosqrt(xdf2
->nrec
)) > XDL_MAX_EQLIMIT
)
386 mlim
= XDL_MAX_EQLIMIT
;
387 for (i
= xdf2
->dstart
, recs
= &xdf2
->recs
[xdf2
->dstart
]; i
<= xdf2
->dend
; i
++, recs
++) {
388 rcrec
= cf
->rcrecs
[(*recs
)->ha
];
389 nm
= rcrec
? rcrec
->len1
: 0;
390 dis2
[i
] = (nm
== 0) ? 0: (nm
>= mlim
) ? 2: 1;
393 for (nreff
= 0, i
= xdf1
->dstart
, recs
= &xdf1
->recs
[xdf1
->dstart
];
394 i
<= xdf1
->dend
; i
++, recs
++) {
396 (dis1
[i
] == 2 && !xdl_clean_mmatch(dis1
, i
, xdf1
->dstart
, xdf1
->dend
))) {
397 xdf1
->rindex
[nreff
] = i
;
398 xdf1
->ha
[nreff
] = (*recs
)->ha
;
405 for (nreff
= 0, i
= xdf2
->dstart
, recs
= &xdf2
->recs
[xdf2
->dstart
];
406 i
<= xdf2
->dend
; i
++, recs
++) {
408 (dis2
[i
] == 2 && !xdl_clean_mmatch(dis2
, i
, xdf2
->dstart
, xdf2
->dend
))) {
409 xdf2
->rindex
[nreff
] = i
;
410 xdf2
->ha
[nreff
] = (*recs
)->ha
;
424 * Early trim initial and terminal matching records.
426 static int xdl_trim_ends(xdfile_t
*xdf1
, xdfile_t
*xdf2
) {
428 xrecord_t
**recs1
, **recs2
;
432 for (i
= 0, lim
= XDL_MIN(xdf1
->nrec
, xdf2
->nrec
); i
< lim
;
433 i
++, recs1
++, recs2
++)
434 if ((*recs1
)->ha
!= (*recs2
)->ha
)
437 xdf1
->dstart
= xdf2
->dstart
= i
;
439 recs1
= xdf1
->recs
+ xdf1
->nrec
- 1;
440 recs2
= xdf2
->recs
+ xdf2
->nrec
- 1;
441 for (lim
-= i
, i
= 0; i
< lim
; i
++, recs1
--, recs2
--)
442 if ((*recs1
)->ha
!= (*recs2
)->ha
)
445 xdf1
->dend
= xdf1
->nrec
- i
- 1;
446 xdf2
->dend
= xdf2
->nrec
- i
- 1;
452 static int xdl_optimize_ctxs(xdlclassifier_t
*cf
, xdfile_t
*xdf1
, xdfile_t
*xdf2
) {
454 if (xdl_trim_ends(xdf1
, xdf2
) < 0 ||
455 xdl_cleanup_records(cf
, xdf1
, xdf2
) < 0) {