]>
Commit | Line | Data |
---|---|---|
3443546f LT |
1 | /* |
2 | * LibXDiff by Davide Libenzi ( File Differential Library ) | |
3 | * Copyright (C) 2003 Davide Libenzi | |
4 | * | |
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. | |
9 | * | |
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. | |
14 | * | |
15 | * You should have received a copy of the GNU Lesser General Public | |
16 | * License along with this library; if not, write to the Free Software | |
17 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
18 | * | |
19 | * Davide Libenzi <davidel@xmailserver.org> | |
20 | * | |
21 | */ | |
22 | ||
23 | #include "xinclude.h" | |
24 | ||
25 | ||
3443546f | 26 | #define XDL_KPDIS_RUN 4 |
ca557aff | 27 | #define XDL_MAX_EQLIMIT 1024 |
9b28d554 | 28 | #define XDL_SIMSCAN_WINDOW 100 |
86abba80 TRC |
29 | #define XDL_GUESS_NLINES1 256 |
30 | #define XDL_GUESS_NLINES2 20 | |
3443546f LT |
31 | |
32 | ||
33 | typedef struct s_xdlclass { | |
34 | struct s_xdlclass *next; | |
35 | unsigned long ha; | |
36 | char const *line; | |
37 | long size; | |
38 | long idx; | |
27af01d5 | 39 | long len1, len2; |
3443546f LT |
40 | } xdlclass_t; |
41 | ||
42 | typedef struct s_xdlclassifier { | |
43 | unsigned int hbits; | |
44 | long hsize; | |
45 | xdlclass_t **rchash; | |
46 | chastore_t ncha; | |
27af01d5 TRC |
47 | xdlclass_t **rcrecs; |
48 | long alloc; | |
3443546f | 49 | long count; |
0d21efa5 | 50 | long flags; |
3443546f LT |
51 | } xdlclassifier_t; |
52 | ||
53 | ||
54 | ||
55 | ||
0d21efa5 | 56 | static int xdl_init_classifier(xdlclassifier_t *cf, long size, long flags); |
3443546f | 57 | static void xdl_free_classifier(xdlclassifier_t *cf); |
27af01d5 TRC |
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, | |
3443546f LT |
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); | |
27af01d5 | 64 | static int xdl_cleanup_records(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2); |
3443546f | 65 | static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2); |
27af01d5 | 66 | static int xdl_optimize_ctxs(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2); |
3443546f LT |
67 | |
68 | ||
69 | ||
70 | ||
0d21efa5 | 71 | static int xdl_init_classifier(xdlclassifier_t *cf, long size, long flags) { |
0d21efa5 JS |
72 | cf->flags = flags; |
73 | ||
3443546f LT |
74 | cf->hbits = xdl_hashbits((unsigned int) size); |
75 | cf->hsize = 1 << cf->hbits; | |
76 | ||
77 | if (xdl_cha_init(&cf->ncha, sizeof(xdlclass_t), size / 4 + 1) < 0) { | |
78 | ||
79 | return -1; | |
80 | } | |
81 | if (!(cf->rchash = (xdlclass_t **) xdl_malloc(cf->hsize * sizeof(xdlclass_t *)))) { | |
82 | ||
83 | xdl_cha_free(&cf->ncha); | |
84 | return -1; | |
85 | } | |
452f4fa5 | 86 | memset(cf->rchash, 0, cf->hsize * sizeof(xdlclass_t *)); |
3443546f | 87 | |
27af01d5 TRC |
88 | cf->alloc = size; |
89 | if (!(cf->rcrecs = (xdlclass_t **) xdl_malloc(cf->alloc * sizeof(xdlclass_t *)))) { | |
90 | ||
91 | xdl_free(cf->rchash); | |
92 | xdl_cha_free(&cf->ncha); | |
93 | return -1; | |
94 | } | |
95 | ||
3443546f LT |
96 | cf->count = 0; |
97 | ||
98 | return 0; | |
99 | } | |
100 | ||
101 | ||
102 | static void xdl_free_classifier(xdlclassifier_t *cf) { | |
103 | ||
27af01d5 | 104 | xdl_free(cf->rcrecs); |
3443546f LT |
105 | xdl_free(cf->rchash); |
106 | xdl_cha_free(&cf->ncha); | |
107 | } | |
108 | ||
109 | ||
27af01d5 TRC |
110 | static int xdl_classify_record(unsigned int pass, xdlclassifier_t *cf, xrecord_t **rhash, |
111 | unsigned int hbits, xrecord_t *rec) { | |
3443546f LT |
112 | long hi; |
113 | char const *line; | |
114 | xdlclass_t *rcrec; | |
27af01d5 | 115 | xdlclass_t **rcrecs; |
3443546f LT |
116 | |
117 | line = rec->ptr; | |
118 | hi = (long) XDL_HASHLONG(rec->ha, cf->hbits); | |
119 | for (rcrec = cf->rchash[hi]; rcrec; rcrec = rcrec->next) | |
0d21efa5 JS |
120 | if (rcrec->ha == rec->ha && |
121 | xdl_recmatch(rcrec->line, rcrec->size, | |
122 | rec->ptr, rec->size, cf->flags)) | |
3443546f LT |
123 | break; |
124 | ||
125 | if (!rcrec) { | |
126 | if (!(rcrec = xdl_cha_alloc(&cf->ncha))) { | |
127 | ||
128 | return -1; | |
129 | } | |
130 | rcrec->idx = cf->count++; | |
27af01d5 TRC |
131 | if (cf->count > cf->alloc) { |
132 | cf->alloc *= 2; | |
133 | if (!(rcrecs = (xdlclass_t **) xdl_realloc(cf->rcrecs, cf->alloc * sizeof(xdlclass_t *)))) { | |
134 | ||
135 | return -1; | |
136 | } | |
137 | cf->rcrecs = rcrecs; | |
138 | } | |
139 | cf->rcrecs[rcrec->idx] = rcrec; | |
3443546f LT |
140 | rcrec->line = line; |
141 | rcrec->size = rec->size; | |
142 | rcrec->ha = rec->ha; | |
27af01d5 | 143 | rcrec->len1 = rcrec->len2 = 0; |
3443546f LT |
144 | rcrec->next = cf->rchash[hi]; |
145 | cf->rchash[hi] = rcrec; | |
146 | } | |
147 | ||
27af01d5 TRC |
148 | (pass == 1) ? rcrec->len1++ : rcrec->len2++; |
149 | ||
3443546f LT |
150 | rec->ha = (unsigned long) rcrec->idx; |
151 | ||
152 | hi = (long) XDL_HASHLONG(rec->ha, hbits); | |
153 | rec->next = rhash[hi]; | |
154 | rhash[hi] = rec; | |
155 | ||
156 | return 0; | |
157 | } | |
158 | ||
159 | ||
27af01d5 | 160 | static int xdl_prepare_ctx(unsigned int pass, mmfile_t *mf, long narec, xpparam_t const *xpp, |
3443546f LT |
161 | xdlclassifier_t *cf, xdfile_t *xdf) { |
162 | unsigned int hbits; | |
452f4fa5 | 163 | long nrec, hsize, bsize; |
3443546f LT |
164 | unsigned long hav; |
165 | char const *blk, *cur, *top, *prev; | |
166 | xrecord_t *crec; | |
167 | xrecord_t **recs, **rrecs; | |
168 | xrecord_t **rhash; | |
169 | unsigned long *ha; | |
170 | char *rchg; | |
171 | long *rindex; | |
172 | ||
159607a8 TRC |
173 | ha = NULL; |
174 | rindex = NULL; | |
175 | rchg = NULL; | |
176 | rhash = NULL; | |
177 | recs = NULL; | |
3443546f | 178 | |
159607a8 TRC |
179 | if (xdl_cha_init(&xdf->rcha, sizeof(xrecord_t), narec / 4 + 1) < 0) |
180 | goto abort; | |
181 | if (!(recs = (xrecord_t **) xdl_malloc(narec * sizeof(xrecord_t *)))) | |
182 | goto abort; | |
3443546f | 183 | |
9f37c275 TRC |
184 | if (xpp->flags & XDF_HISTOGRAM_DIFF) |
185 | hbits = hsize = 0; | |
186 | else { | |
187 | hbits = xdl_hashbits((unsigned int) narec); | |
188 | hsize = 1 << hbits; | |
189 | if (!(rhash = (xrecord_t **) xdl_malloc(hsize * sizeof(xrecord_t *)))) | |
190 | goto abort; | |
191 | memset(rhash, 0, hsize * sizeof(xrecord_t *)); | |
3443546f | 192 | } |
3443546f LT |
193 | |
194 | nrec = 0; | |
195 | if ((cur = blk = xdl_mmfile_first(mf, &bsize)) != NULL) { | |
739864b1 | 196 | for (top = blk + bsize; cur < top; ) { |
3443546f | 197 | prev = cur; |
0d21efa5 | 198 | hav = xdl_hash_record(&cur, top, xpp->flags); |
3443546f LT |
199 | if (nrec >= narec) { |
200 | narec *= 2; | |
159607a8 TRC |
201 | if (!(rrecs = (xrecord_t **) xdl_realloc(recs, narec * sizeof(xrecord_t *)))) |
202 | goto abort; | |
3443546f LT |
203 | recs = rrecs; |
204 | } | |
159607a8 TRC |
205 | if (!(crec = xdl_cha_alloc(&xdf->rcha))) |
206 | goto abort; | |
3443546f LT |
207 | crec->ptr = prev; |
208 | crec->size = (long) (cur - prev); | |
209 | crec->ha = hav; | |
210 | recs[nrec++] = crec; | |
211 | ||
9f37c275 | 212 | if (!(xpp->flags & XDF_HISTOGRAM_DIFF) && |
b14b969a | 213 | xdl_classify_record(pass, cf, rhash, hbits, crec) < 0) |
159607a8 | 214 | goto abort; |
3443546f LT |
215 | } |
216 | } | |
217 | ||
159607a8 TRC |
218 | if (!(rchg = (char *) xdl_malloc((nrec + 2) * sizeof(char)))) |
219 | goto abort; | |
3443546f LT |
220 | memset(rchg, 0, (nrec + 2) * sizeof(char)); |
221 | ||
159607a8 TRC |
222 | if (!(rindex = (long *) xdl_malloc((nrec + 1) * sizeof(long)))) |
223 | goto abort; | |
224 | if (!(ha = (unsigned long *) xdl_malloc((nrec + 1) * sizeof(unsigned long)))) | |
225 | goto abort; | |
3443546f LT |
226 | |
227 | xdf->nrec = nrec; | |
228 | xdf->recs = recs; | |
229 | xdf->hbits = hbits; | |
230 | xdf->rhash = rhash; | |
231 | xdf->rchg = rchg + 1; | |
232 | xdf->rindex = rindex; | |
233 | xdf->nreff = 0; | |
234 | xdf->ha = ha; | |
235 | xdf->dstart = 0; | |
236 | xdf->dend = nrec - 1; | |
237 | ||
238 | return 0; | |
159607a8 TRC |
239 | |
240 | abort: | |
241 | xdl_free(ha); | |
242 | xdl_free(rindex); | |
243 | xdl_free(rchg); | |
244 | xdl_free(rhash); | |
245 | xdl_free(recs); | |
246 | xdl_cha_free(&xdf->rcha); | |
247 | return -1; | |
3443546f LT |
248 | } |
249 | ||
250 | ||
251 | static void xdl_free_ctx(xdfile_t *xdf) { | |
252 | ||
253 | xdl_free(xdf->rhash); | |
254 | xdl_free(xdf->rindex); | |
255 | xdl_free(xdf->rchg - 1); | |
256 | xdl_free(xdf->ha); | |
257 | xdl_free(xdf->recs); | |
258 | xdl_cha_free(&xdf->rcha); | |
259 | } | |
260 | ||
261 | ||
262 | int xdl_prepare_env(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp, | |
263 | xdfenv_t *xe) { | |
86abba80 | 264 | long enl1, enl2, sample; |
3443546f LT |
265 | xdlclassifier_t cf; |
266 | ||
86abba80 TRC |
267 | /* |
268 | * For histogram diff, we can afford a smaller sample size and | |
269 | * thus a poorer estimate of the number of lines, as the hash | |
270 | * table (rhash) won't be filled up/grown. The number of lines | |
271 | * (nrecs) will be updated correctly anyway by | |
272 | * xdl_prepare_ctx(). | |
273 | */ | |
274 | sample = xpp->flags & XDF_HISTOGRAM_DIFF ? XDL_GUESS_NLINES2 : XDL_GUESS_NLINES1; | |
275 | ||
276 | enl1 = xdl_guess_lines(mf1, sample) + 1; | |
277 | enl2 = xdl_guess_lines(mf2, sample) + 1; | |
3443546f | 278 | |
9f37c275 TRC |
279 | if (!(xpp->flags & XDF_HISTOGRAM_DIFF) && |
280 | xdl_init_classifier(&cf, enl1 + enl2 + 1, xpp->flags) < 0) { | |
3443546f LT |
281 | |
282 | return -1; | |
283 | } | |
284 | ||
27af01d5 | 285 | if (xdl_prepare_ctx(1, mf1, enl1, xpp, &cf, &xe->xdf1) < 0) { |
3443546f LT |
286 | |
287 | xdl_free_classifier(&cf); | |
288 | return -1; | |
289 | } | |
27af01d5 | 290 | if (xdl_prepare_ctx(2, mf2, enl2, xpp, &cf, &xe->xdf2) < 0) { |
3443546f LT |
291 | |
292 | xdl_free_ctx(&xe->xdf1); | |
293 | xdl_free_classifier(&cf); | |
294 | return -1; | |
295 | } | |
296 | ||
92b7de93 | 297 | if (!(xpp->flags & XDF_PATIENCE_DIFF) && |
9f37c275 | 298 | !(xpp->flags & XDF_HISTOGRAM_DIFF) && |
27af01d5 | 299 | xdl_optimize_ctxs(&cf, &xe->xdf1, &xe->xdf2) < 0) { |
3443546f LT |
300 | |
301 | xdl_free_ctx(&xe->xdf2); | |
302 | xdl_free_ctx(&xe->xdf1); | |
303 | return -1; | |
304 | } | |
305 | ||
b14b969a JH |
306 | if (!(xpp->flags & XDF_HISTOGRAM_DIFF)) |
307 | xdl_free_classifier(&cf); | |
27af01d5 | 308 | |
3443546f LT |
309 | return 0; |
310 | } | |
311 | ||
312 | ||
313 | void xdl_free_env(xdfenv_t *xe) { | |
314 | ||
315 | xdl_free_ctx(&xe->xdf2); | |
316 | xdl_free_ctx(&xe->xdf1); | |
317 | } | |
318 | ||
319 | ||
320 | static int xdl_clean_mmatch(char const *dis, long i, long s, long e) { | |
ca557aff DL |
321 | long r, rdis0, rpdis0, rdis1, rpdis1; |
322 | ||
9b28d554 DL |
323 | /* |
324 | * Limits the window the is examined during the similar-lines | |
325 | * scan. The loops below stops when dis[i - r] == 1 (line that | |
326 | * has no match), but there are corner cases where the loop | |
327 | * proceed all the way to the extremities by causing huge | |
328 | * performance penalties in case of big files. | |
329 | */ | |
330 | if (i - s > XDL_SIMSCAN_WINDOW) | |
331 | s = i - XDL_SIMSCAN_WINDOW; | |
332 | if (e - i > XDL_SIMSCAN_WINDOW) | |
333 | e = i + XDL_SIMSCAN_WINDOW; | |
334 | ||
ca557aff DL |
335 | /* |
336 | * Scans the lines before 'i' to find a run of lines that either | |
337 | * have no match (dis[j] == 0) or have multiple matches (dis[j] > 1). | |
338 | * Note that we always call this function with dis[i] > 1, so the | |
339 | * current line (i) is already a multimatch line. | |
340 | */ | |
341 | for (r = 1, rdis0 = 0, rpdis0 = 1; (i - r) >= s; r++) { | |
3443546f | 342 | if (!dis[i - r]) |
ca557aff | 343 | rdis0++; |
3443546f | 344 | else if (dis[i - r] == 2) |
ca557aff | 345 | rpdis0++; |
3443546f LT |
346 | else |
347 | break; | |
348 | } | |
ca557aff DL |
349 | /* |
350 | * If the run before the line 'i' found only multimatch lines, we | |
351 | * return 0 and hence we don't make the current line (i) discarded. | |
352 | * We want to discard multimatch lines only when they appear in the | |
353 | * middle of runs with nomatch lines (dis[j] == 0). | |
354 | */ | |
355 | if (rdis0 == 0) | |
356 | return 0; | |
357 | for (r = 1, rdis1 = 0, rpdis1 = 1; (i + r) <= e; r++) { | |
3443546f | 358 | if (!dis[i + r]) |
ca557aff | 359 | rdis1++; |
3443546f | 360 | else if (dis[i + r] == 2) |
ca557aff | 361 | rpdis1++; |
3443546f LT |
362 | else |
363 | break; | |
364 | } | |
ca557aff DL |
365 | /* |
366 | * If the run after the line 'i' found only multimatch lines, we | |
367 | * return 0 and hence we don't make the current line (i) discarded. | |
368 | */ | |
369 | if (rdis1 == 0) | |
370 | return 0; | |
371 | rdis1 += rdis0; | |
372 | rpdis1 += rpdis0; | |
373 | ||
374 | return rpdis1 * XDL_KPDIS_RUN < (rpdis1 + rdis1); | |
3443546f LT |
375 | } |
376 | ||
377 | ||
378 | /* | |
379 | * Try to reduce the problem complexity, discard records that have no | |
380 | * matches on the other file. Also, lines that have multiple matches | |
381 | * might be potentially discarded if they happear in a run of discardable. | |
382 | */ | |
27af01d5 TRC |
383 | static int xdl_cleanup_records(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2) { |
384 | long i, nm, nreff; | |
3443546f | 385 | xrecord_t **recs; |
27af01d5 | 386 | xdlclass_t *rcrec; |
3443546f LT |
387 | char *dis, *dis1, *dis2; |
388 | ||
ca557aff | 389 | if (!(dis = (char *) xdl_malloc(xdf1->nrec + xdf2->nrec + 2))) { |
3443546f LT |
390 | |
391 | return -1; | |
392 | } | |
ca557aff | 393 | memset(dis, 0, xdf1->nrec + xdf2->nrec + 2); |
3443546f LT |
394 | dis1 = dis; |
395 | dis2 = dis1 + xdf1->nrec + 1; | |
396 | ||
397 | for (i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart]; i <= xdf1->dend; i++, recs++) { | |
27af01d5 TRC |
398 | rcrec = cf->rcrecs[(*recs)->ha]; |
399 | nm = rcrec ? rcrec->len2 : 0; | |
400 | dis1[i] = (nm == 0) ? 0: 1; | |
3443546f LT |
401 | } |
402 | ||
403 | for (i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart]; i <= xdf2->dend; i++, recs++) { | |
27af01d5 TRC |
404 | rcrec = cf->rcrecs[(*recs)->ha]; |
405 | nm = rcrec ? rcrec->len1 : 0; | |
406 | dis2[i] = (nm == 0) ? 0: 1; | |
3443546f LT |
407 | } |
408 | ||
409 | for (nreff = 0, i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart]; | |
410 | i <= xdf1->dend; i++, recs++) { | |
411 | if (dis1[i] == 1 || | |
412 | (dis1[i] == 2 && !xdl_clean_mmatch(dis1, i, xdf1->dstart, xdf1->dend))) { | |
413 | xdf1->rindex[nreff] = i; | |
414 | xdf1->ha[nreff] = (*recs)->ha; | |
415 | nreff++; | |
416 | } else | |
417 | xdf1->rchg[i] = 1; | |
418 | } | |
419 | xdf1->nreff = nreff; | |
420 | ||
421 | for (nreff = 0, i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart]; | |
422 | i <= xdf2->dend; i++, recs++) { | |
423 | if (dis2[i] == 1 || | |
424 | (dis2[i] == 2 && !xdl_clean_mmatch(dis2, i, xdf2->dstart, xdf2->dend))) { | |
425 | xdf2->rindex[nreff] = i; | |
426 | xdf2->ha[nreff] = (*recs)->ha; | |
427 | nreff++; | |
428 | } else | |
429 | xdf2->rchg[i] = 1; | |
430 | } | |
431 | xdf2->nreff = nreff; | |
432 | ||
433 | xdl_free(dis); | |
434 | ||
435 | return 0; | |
436 | } | |
437 | ||
438 | ||
439 | /* | |
440 | * Early trim initial and terminal matching records. | |
441 | */ | |
442 | static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2) { | |
443 | long i, lim; | |
444 | xrecord_t **recs1, **recs2; | |
445 | ||
446 | recs1 = xdf1->recs; | |
447 | recs2 = xdf2->recs; | |
448 | for (i = 0, lim = XDL_MIN(xdf1->nrec, xdf2->nrec); i < lim; | |
449 | i++, recs1++, recs2++) | |
450 | if ((*recs1)->ha != (*recs2)->ha) | |
451 | break; | |
452 | ||
453 | xdf1->dstart = xdf2->dstart = i; | |
454 | ||
455 | recs1 = xdf1->recs + xdf1->nrec - 1; | |
456 | recs2 = xdf2->recs + xdf2->nrec - 1; | |
457 | for (lim -= i, i = 0; i < lim; i++, recs1--, recs2--) | |
458 | if ((*recs1)->ha != (*recs2)->ha) | |
459 | break; | |
460 | ||
461 | xdf1->dend = xdf1->nrec - i - 1; | |
462 | xdf2->dend = xdf2->nrec - i - 1; | |
463 | ||
464 | return 0; | |
465 | } | |
466 | ||
467 | ||
27af01d5 | 468 | static int xdl_optimize_ctxs(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2) { |
3443546f LT |
469 | |
470 | if (xdl_trim_ends(xdf1, xdf2) < 0 || | |
27af01d5 | 471 | xdl_cleanup_records(cf, xdf1, xdf2) < 0) { |
3443546f LT |
472 | |
473 | return -1; | |
474 | } | |
475 | ||
476 | return 0; | |
477 | } |