]>
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 | |
48425792 TZ |
16 | * License along with this library; if not, see |
17 | * <http://www.gnu.org/licenses/>. | |
3443546f LT |
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 | } | |
18aae7e2 | 81 | if (!(cf->rchash = xdl_calloc(cf->hsize, sizeof(*cf->rchash)))) { |
3443546f LT |
82 | |
83 | xdl_cha_free(&cf->ncha); | |
84 | return -1; | |
85 | } | |
3443546f | 86 | |
27af01d5 | 87 | cf->alloc = size; |
abf04bda | 88 | if (!XDL_ALLOC_ARRAY(cf->rcrecs, cf->alloc)) { |
27af01d5 TRC |
89 | |
90 | xdl_free(cf->rchash); | |
91 | xdl_cha_free(&cf->ncha); | |
92 | return -1; | |
93 | } | |
94 | ||
3443546f LT |
95 | cf->count = 0; |
96 | ||
97 | return 0; | |
98 | } | |
99 | ||
100 | ||
101 | static void xdl_free_classifier(xdlclassifier_t *cf) { | |
102 | ||
27af01d5 | 103 | xdl_free(cf->rcrecs); |
3443546f LT |
104 | xdl_free(cf->rchash); |
105 | xdl_cha_free(&cf->ncha); | |
106 | } | |
107 | ||
108 | ||
27af01d5 TRC |
109 | static int xdl_classify_record(unsigned int pass, xdlclassifier_t *cf, xrecord_t **rhash, |
110 | unsigned int hbits, xrecord_t *rec) { | |
3443546f LT |
111 | long hi; |
112 | char const *line; | |
113 | xdlclass_t *rcrec; | |
27af01d5 | 114 | xdlclass_t **rcrecs; |
3443546f LT |
115 | |
116 | line = rec->ptr; | |
117 | hi = (long) XDL_HASHLONG(rec->ha, cf->hbits); | |
118 | for (rcrec = cf->rchash[hi]; rcrec; rcrec = rcrec->next) | |
0d21efa5 JS |
119 | if (rcrec->ha == rec->ha && |
120 | xdl_recmatch(rcrec->line, rcrec->size, | |
121 | rec->ptr, rec->size, cf->flags)) | |
3443546f LT |
122 | break; |
123 | ||
124 | if (!rcrec) { | |
125 | if (!(rcrec = xdl_cha_alloc(&cf->ncha))) { | |
126 | ||
127 | return -1; | |
128 | } | |
129 | rcrec->idx = cf->count++; | |
27af01d5 TRC |
130 | if (cf->count > cf->alloc) { |
131 | cf->alloc *= 2; | |
132 | if (!(rcrecs = (xdlclass_t **) xdl_realloc(cf->rcrecs, cf->alloc * sizeof(xdlclass_t *)))) { | |
133 | ||
134 | return -1; | |
135 | } | |
136 | cf->rcrecs = rcrecs; | |
137 | } | |
138 | cf->rcrecs[rcrec->idx] = rcrec; | |
3443546f LT |
139 | rcrec->line = line; |
140 | rcrec->size = rec->size; | |
141 | rcrec->ha = rec->ha; | |
27af01d5 | 142 | rcrec->len1 = rcrec->len2 = 0; |
3443546f LT |
143 | rcrec->next = cf->rchash[hi]; |
144 | cf->rchash[hi] = rcrec; | |
145 | } | |
146 | ||
27af01d5 TRC |
147 | (pass == 1) ? rcrec->len1++ : rcrec->len2++; |
148 | ||
3443546f LT |
149 | rec->ha = (unsigned long) rcrec->idx; |
150 | ||
151 | hi = (long) XDL_HASHLONG(rec->ha, hbits); | |
152 | rec->next = rhash[hi]; | |
153 | rhash[hi] = rec; | |
154 | ||
155 | return 0; | |
156 | } | |
157 | ||
158 | ||
27af01d5 | 159 | static int xdl_prepare_ctx(unsigned int pass, mmfile_t *mf, long narec, xpparam_t const *xpp, |
3443546f LT |
160 | xdlclassifier_t *cf, xdfile_t *xdf) { |
161 | unsigned int hbits; | |
452f4fa5 | 162 | long nrec, hsize, bsize; |
3443546f LT |
163 | unsigned long hav; |
164 | char const *blk, *cur, *top, *prev; | |
165 | xrecord_t *crec; | |
166 | xrecord_t **recs, **rrecs; | |
167 | xrecord_t **rhash; | |
168 | unsigned long *ha; | |
169 | char *rchg; | |
170 | long *rindex; | |
171 | ||
159607a8 TRC |
172 | ha = NULL; |
173 | rindex = NULL; | |
174 | rchg = NULL; | |
175 | rhash = NULL; | |
176 | recs = NULL; | |
3443546f | 177 | |
159607a8 TRC |
178 | if (xdl_cha_init(&xdf->rcha, sizeof(xrecord_t), narec / 4 + 1) < 0) |
179 | goto abort; | |
abf04bda | 180 | if (!XDL_ALLOC_ARRAY(recs, narec)) |
159607a8 | 181 | goto abort; |
3443546f | 182 | |
663c5ad0 PW |
183 | hbits = xdl_hashbits((unsigned int) narec); |
184 | hsize = 1 << hbits; | |
18aae7e2 | 185 | if (!(rhash = xdl_calloc(hsize, sizeof(*rhash)))) |
663c5ad0 | 186 | goto abort; |
3443546f LT |
187 | |
188 | nrec = 0; | |
afe8a907 | 189 | if ((cur = blk = xdl_mmfile_first(mf, &bsize))) { |
739864b1 | 190 | for (top = blk + bsize; cur < top; ) { |
3443546f | 191 | prev = cur; |
0d21efa5 | 192 | hav = xdl_hash_record(&cur, top, xpp->flags); |
3443546f LT |
193 | if (nrec >= narec) { |
194 | narec *= 2; | |
159607a8 TRC |
195 | if (!(rrecs = (xrecord_t **) xdl_realloc(recs, narec * sizeof(xrecord_t *)))) |
196 | goto abort; | |
3443546f LT |
197 | recs = rrecs; |
198 | } | |
159607a8 TRC |
199 | if (!(crec = xdl_cha_alloc(&xdf->rcha))) |
200 | goto abort; | |
3443546f LT |
201 | crec->ptr = prev; |
202 | crec->size = (long) (cur - prev); | |
203 | crec->ha = hav; | |
204 | recs[nrec++] = crec; | |
663c5ad0 | 205 | if (xdl_classify_record(pass, cf, rhash, hbits, crec) < 0) |
159607a8 | 206 | goto abort; |
3443546f LT |
207 | } |
208 | } | |
209 | ||
18aae7e2 | 210 | if (!(rchg = xdl_calloc((nrec + 2), sizeof(*rchg)))) |
159607a8 | 211 | goto abort; |
3443546f | 212 | |
b82dd3f7 PW |
213 | if ((XDF_DIFF_ALG(xpp->flags) != XDF_PATIENCE_DIFF) && |
214 | (XDF_DIFF_ALG(xpp->flags) != XDF_HISTOGRAM_DIFF)) { | |
abf04bda | 215 | if (!XDL_ALLOC_ARRAY(rindex, nrec + 1)) |
b82dd3f7 | 216 | goto abort; |
abf04bda | 217 | if (!XDL_ALLOC_ARRAY(ha, nrec + 1)) |
b82dd3f7 PW |
218 | goto abort; |
219 | } | |
3443546f LT |
220 | |
221 | xdf->nrec = nrec; | |
222 | xdf->recs = recs; | |
223 | xdf->hbits = hbits; | |
224 | xdf->rhash = rhash; | |
225 | xdf->rchg = rchg + 1; | |
226 | xdf->rindex = rindex; | |
227 | xdf->nreff = 0; | |
228 | xdf->ha = ha; | |
229 | xdf->dstart = 0; | |
230 | xdf->dend = nrec - 1; | |
231 | ||
232 | return 0; | |
159607a8 TRC |
233 | |
234 | abort: | |
235 | xdl_free(ha); | |
236 | xdl_free(rindex); | |
237 | xdl_free(rchg); | |
238 | xdl_free(rhash); | |
239 | xdl_free(recs); | |
240 | xdl_cha_free(&xdf->rcha); | |
241 | return -1; | |
3443546f LT |
242 | } |
243 | ||
244 | ||
245 | static void xdl_free_ctx(xdfile_t *xdf) { | |
246 | ||
247 | xdl_free(xdf->rhash); | |
248 | xdl_free(xdf->rindex); | |
249 | xdl_free(xdf->rchg - 1); | |
250 | xdl_free(xdf->ha); | |
251 | xdl_free(xdf->recs); | |
252 | xdl_cha_free(&xdf->rcha); | |
253 | } | |
254 | ||
255 | ||
256 | int xdl_prepare_env(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp, | |
257 | xdfenv_t *xe) { | |
86abba80 | 258 | long enl1, enl2, sample; |
3443546f LT |
259 | xdlclassifier_t cf; |
260 | ||
2738bc3f TRC |
261 | memset(&cf, 0, sizeof(cf)); |
262 | ||
86abba80 TRC |
263 | /* |
264 | * For histogram diff, we can afford a smaller sample size and | |
265 | * thus a poorer estimate of the number of lines, as the hash | |
266 | * table (rhash) won't be filled up/grown. The number of lines | |
267 | * (nrecs) will be updated correctly anyway by | |
268 | * xdl_prepare_ctx(). | |
269 | */ | |
307ab20b JH |
270 | sample = (XDF_DIFF_ALG(xpp->flags) == XDF_HISTOGRAM_DIFF |
271 | ? XDL_GUESS_NLINES2 : XDL_GUESS_NLINES1); | |
86abba80 TRC |
272 | |
273 | enl1 = xdl_guess_lines(mf1, sample) + 1; | |
274 | enl2 = xdl_guess_lines(mf2, sample) + 1; | |
3443546f | 275 | |
663c5ad0 | 276 | if (xdl_init_classifier(&cf, enl1 + enl2 + 1, xpp->flags) < 0) |
3443546f | 277 | return -1; |
3443546f | 278 | |
27af01d5 | 279 | if (xdl_prepare_ctx(1, mf1, enl1, xpp, &cf, &xe->xdf1) < 0) { |
3443546f LT |
280 | |
281 | xdl_free_classifier(&cf); | |
282 | return -1; | |
283 | } | |
27af01d5 | 284 | if (xdl_prepare_ctx(2, mf2, enl2, xpp, &cf, &xe->xdf2) < 0) { |
3443546f LT |
285 | |
286 | xdl_free_ctx(&xe->xdf1); | |
287 | xdl_free_classifier(&cf); | |
288 | return -1; | |
289 | } | |
290 | ||
307ab20b JH |
291 | if ((XDF_DIFF_ALG(xpp->flags) != XDF_PATIENCE_DIFF) && |
292 | (XDF_DIFF_ALG(xpp->flags) != XDF_HISTOGRAM_DIFF) && | |
293 | xdl_optimize_ctxs(&cf, &xe->xdf1, &xe->xdf2) < 0) { | |
3443546f LT |
294 | |
295 | xdl_free_ctx(&xe->xdf2); | |
296 | xdl_free_ctx(&xe->xdf1); | |
87f16258 | 297 | xdl_free_classifier(&cf); |
3443546f LT |
298 | return -1; |
299 | } | |
300 | ||
663c5ad0 | 301 | xdl_free_classifier(&cf); |
27af01d5 | 302 | |
3443546f LT |
303 | return 0; |
304 | } | |
305 | ||
306 | ||
307 | void xdl_free_env(xdfenv_t *xe) { | |
308 | ||
309 | xdl_free_ctx(&xe->xdf2); | |
310 | xdl_free_ctx(&xe->xdf1); | |
311 | } | |
312 | ||
313 | ||
314 | static int xdl_clean_mmatch(char const *dis, long i, long s, long e) { | |
ca557aff DL |
315 | long r, rdis0, rpdis0, rdis1, rpdis1; |
316 | ||
9b28d554 DL |
317 | /* |
318 | * Limits the window the is examined during the similar-lines | |
319 | * scan. The loops below stops when dis[i - r] == 1 (line that | |
320 | * has no match), but there are corner cases where the loop | |
321 | * proceed all the way to the extremities by causing huge | |
322 | * performance penalties in case of big files. | |
323 | */ | |
324 | if (i - s > XDL_SIMSCAN_WINDOW) | |
325 | s = i - XDL_SIMSCAN_WINDOW; | |
326 | if (e - i > XDL_SIMSCAN_WINDOW) | |
327 | e = i + XDL_SIMSCAN_WINDOW; | |
328 | ||
ca557aff DL |
329 | /* |
330 | * Scans the lines before 'i' to find a run of lines that either | |
331 | * have no match (dis[j] == 0) or have multiple matches (dis[j] > 1). | |
332 | * Note that we always call this function with dis[i] > 1, so the | |
333 | * current line (i) is already a multimatch line. | |
334 | */ | |
335 | for (r = 1, rdis0 = 0, rpdis0 = 1; (i - r) >= s; r++) { | |
3443546f | 336 | if (!dis[i - r]) |
ca557aff | 337 | rdis0++; |
3443546f | 338 | else if (dis[i - r] == 2) |
ca557aff | 339 | rpdis0++; |
3443546f LT |
340 | else |
341 | break; | |
342 | } | |
ca557aff DL |
343 | /* |
344 | * If the run before the line 'i' found only multimatch lines, we | |
345 | * return 0 and hence we don't make the current line (i) discarded. | |
346 | * We want to discard multimatch lines only when they appear in the | |
347 | * middle of runs with nomatch lines (dis[j] == 0). | |
348 | */ | |
349 | if (rdis0 == 0) | |
350 | return 0; | |
351 | for (r = 1, rdis1 = 0, rpdis1 = 1; (i + r) <= e; r++) { | |
3443546f | 352 | if (!dis[i + r]) |
ca557aff | 353 | rdis1++; |
3443546f | 354 | else if (dis[i + r] == 2) |
ca557aff | 355 | rpdis1++; |
3443546f LT |
356 | else |
357 | break; | |
358 | } | |
ca557aff DL |
359 | /* |
360 | * If the run after the line 'i' found only multimatch lines, we | |
361 | * return 0 and hence we don't make the current line (i) discarded. | |
362 | */ | |
363 | if (rdis1 == 0) | |
364 | return 0; | |
365 | rdis1 += rdis0; | |
366 | rpdis1 += rpdis0; | |
367 | ||
368 | return rpdis1 * XDL_KPDIS_RUN < (rpdis1 + rdis1); | |
3443546f LT |
369 | } |
370 | ||
371 | ||
372 | /* | |
373 | * Try to reduce the problem complexity, discard records that have no | |
374 | * matches on the other file. Also, lines that have multiple matches | |
375 | * might be potentially discarded if they happear in a run of discardable. | |
376 | */ | |
27af01d5 | 377 | static int xdl_cleanup_records(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2) { |
c5aa9068 | 378 | long i, nm, nreff, mlim; |
3443546f | 379 | xrecord_t **recs; |
27af01d5 | 380 | xdlclass_t *rcrec; |
3443546f LT |
381 | char *dis, *dis1, *dis2; |
382 | ||
18aae7e2 | 383 | if (!(dis = xdl_calloc(xdf1->nrec + xdf2->nrec + 2, sizeof(*dis)))) { |
3443546f LT |
384 | |
385 | return -1; | |
386 | } | |
3443546f LT |
387 | dis1 = dis; |
388 | dis2 = dis1 + xdf1->nrec + 1; | |
389 | ||
c5aa9068 RS |
390 | if ((mlim = xdl_bogosqrt(xdf1->nrec)) > XDL_MAX_EQLIMIT) |
391 | mlim = XDL_MAX_EQLIMIT; | |
3443546f | 392 | for (i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart]; i <= xdf1->dend; i++, recs++) { |
27af01d5 TRC |
393 | rcrec = cf->rcrecs[(*recs)->ha]; |
394 | nm = rcrec ? rcrec->len2 : 0; | |
c5aa9068 | 395 | dis1[i] = (nm == 0) ? 0: (nm >= mlim) ? 2: 1; |
3443546f LT |
396 | } |
397 | ||
c5aa9068 RS |
398 | if ((mlim = xdl_bogosqrt(xdf2->nrec)) > XDL_MAX_EQLIMIT) |
399 | mlim = XDL_MAX_EQLIMIT; | |
3443546f | 400 | for (i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart]; i <= xdf2->dend; i++, recs++) { |
27af01d5 TRC |
401 | rcrec = cf->rcrecs[(*recs)->ha]; |
402 | nm = rcrec ? rcrec->len1 : 0; | |
c5aa9068 | 403 | dis2[i] = (nm == 0) ? 0: (nm >= mlim) ? 2: 1; |
3443546f LT |
404 | } |
405 | ||
406 | for (nreff = 0, i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart]; | |
407 | i <= xdf1->dend; i++, recs++) { | |
408 | if (dis1[i] == 1 || | |
409 | (dis1[i] == 2 && !xdl_clean_mmatch(dis1, i, xdf1->dstart, xdf1->dend))) { | |
410 | xdf1->rindex[nreff] = i; | |
411 | xdf1->ha[nreff] = (*recs)->ha; | |
412 | nreff++; | |
413 | } else | |
414 | xdf1->rchg[i] = 1; | |
415 | } | |
416 | xdf1->nreff = nreff; | |
417 | ||
418 | for (nreff = 0, i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart]; | |
419 | i <= xdf2->dend; i++, recs++) { | |
420 | if (dis2[i] == 1 || | |
421 | (dis2[i] == 2 && !xdl_clean_mmatch(dis2, i, xdf2->dstart, xdf2->dend))) { | |
422 | xdf2->rindex[nreff] = i; | |
423 | xdf2->ha[nreff] = (*recs)->ha; | |
424 | nreff++; | |
425 | } else | |
426 | xdf2->rchg[i] = 1; | |
427 | } | |
428 | xdf2->nreff = nreff; | |
429 | ||
430 | xdl_free(dis); | |
431 | ||
432 | return 0; | |
433 | } | |
434 | ||
435 | ||
436 | /* | |
437 | * Early trim initial and terminal matching records. | |
438 | */ | |
439 | static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2) { | |
440 | long i, lim; | |
441 | xrecord_t **recs1, **recs2; | |
442 | ||
443 | recs1 = xdf1->recs; | |
444 | recs2 = xdf2->recs; | |
445 | for (i = 0, lim = XDL_MIN(xdf1->nrec, xdf2->nrec); i < lim; | |
446 | i++, recs1++, recs2++) | |
447 | if ((*recs1)->ha != (*recs2)->ha) | |
448 | break; | |
449 | ||
450 | xdf1->dstart = xdf2->dstart = i; | |
451 | ||
452 | recs1 = xdf1->recs + xdf1->nrec - 1; | |
453 | recs2 = xdf2->recs + xdf2->nrec - 1; | |
454 | for (lim -= i, i = 0; i < lim; i++, recs1--, recs2--) | |
455 | if ((*recs1)->ha != (*recs2)->ha) | |
456 | break; | |
457 | ||
458 | xdf1->dend = xdf1->nrec - i - 1; | |
459 | xdf2->dend = xdf2->nrec - i - 1; | |
460 | ||
461 | return 0; | |
462 | } | |
463 | ||
464 | ||
27af01d5 | 465 | static int xdl_optimize_ctxs(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2) { |
3443546f LT |
466 | |
467 | if (xdl_trim_ends(xdf1, xdf2) < 0 || | |
27af01d5 | 468 | xdl_cleanup_records(cf, xdf1, xdf2) < 0) { |
3443546f LT |
469 | |
470 | return -1; | |
471 | } | |
472 | ||
473 | return 0; | |
474 | } |