xdiff/xprepare.c

   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
  26
  27 #define XDL_KPDIS_RUN 4
  28
  29
  30
  31 typedef struct s_xdlclass {
  32         struct s_xdlclass *next;
  33         unsigned long ha;
  34         char const *line;
  35         long size;
  36         long idx;
  37 } xdlclass_t;
  38
  39 typedef struct s_xdlclassifier {
  40         unsigned int hbits;
  41         long hsize;
  42         xdlclass_t **rchash;
  43         chastore_t ncha;
  44         long count;
  45 } xdlclassifier_t;
  46
  47
  48
  49
  50 static int xdl_init_classifier(xdlclassifier_t *cf, long size);
  51 static void xdl_free_classifier(xdlclassifier_t *cf);
  52 static int xdl_classify_record(xdlclassifier_t *cf, xrecord_t **rhash, unsigned int hbits,
  53                                xrecord_t *rec);
  54 static int xdl_prepare_ctx(mmfile_t *mf, long narec, xpparam_t const *xpp,
  55                            xdlclassifier_t *cf, xdfile_t *xdf);
  56 static void xdl_free_ctx(xdfile_t *xdf);
  57 static int xdl_clean_mmatch(char const *dis, long i, long s, long e);
  58 static int xdl_cleanup_records(xdfile_t *xdf1, xdfile_t *xdf2);
  59 static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2);
  60 static int xdl_optimize_ctxs(xdfile_t *xdf1, xdfile_t *xdf2);
  61
  62
  63
  64
  65 static int xdl_init_classifier(xdlclassifier_t *cf, long size) {
  66         long i;
  67
  68         cf->hbits = xdl_hashbits((unsigned int) size);
  69         cf->hsize = 1 << cf->hbits;
  70
  71         if (xdl_cha_init(&cf->ncha, sizeof(xdlclass_t), size / 4 + 1) < 0) {
  72
  73                 return -1;
  74         }
  75         if (!(cf->rchash = (xdlclass_t **) xdl_malloc(cf->hsize * sizeof(xdlclass_t *)))) {
  76
  77                 xdl_cha_free(&cf->ncha);
  78                 return -1;
  79         }
  80         for (i = 0; i < cf->hsize; i++)
  81                 cf->rchash[i] = NULL;
  82
  83         cf->count = 0;
  84
  85         return 0;
  86 }
  87
  88
  89 static void xdl_free_classifier(xdlclassifier_t *cf) {
  90
  91         xdl_free(cf->rchash);
  92         xdl_cha_free(&cf->ncha);
  93 }
  94
  95
  96 static int xdl_classify_record(xdlclassifier_t *cf, xrecord_t **rhash, unsigned int hbits,
  97                                xrecord_t *rec) {
  98         long hi;
  99         char const *line;
 100         xdlclass_t *rcrec;
 101
 102         line = rec->ptr;
 103         hi = (long) XDL_HASHLONG(rec->ha, cf->hbits);
 104         for (rcrec = cf->rchash[hi]; rcrec; rcrec = rcrec->next)
 105                 if (rcrec->ha == rec->ha && rcrec->size == rec->size &&
 106                     !memcmp(line, rcrec->line, rec->size))
 107                         break;
 108
 109         if (!rcrec) {
 110                 if (!(rcrec = xdl_cha_alloc(&cf->ncha))) {
 111
 112                         return -1;
 113                 }
 114                 rcrec->idx = cf->count++;
 115                 rcrec->line = line;
 116                 rcrec->size = rec->size;
 117                 rcrec->ha = rec->ha;
 118                 rcrec->next = cf->rchash[hi];
 119                 cf->rchash[hi] = rcrec;
 120         }
 121
 122         rec->ha = (unsigned long) rcrec->idx;
 123
 124         hi = (long) XDL_HASHLONG(rec->ha, hbits);
 125         rec->next = rhash[hi];
 126         rhash[hi] = rec;
 127
 128         return 0;
 129 }
 130
 131
 132 static int xdl_prepare_ctx(mmfile_t *mf, long narec, xpparam_t const *xpp,
 133                            xdlclassifier_t *cf, xdfile_t *xdf) {
 134         unsigned int hbits;
 135         long i, nrec, hsize, bsize;
 136         unsigned long hav;
 137         char const *blk, *cur, *top, *prev;
 138         xrecord_t *crec;
 139         xrecord_t **recs, **rrecs;
 140         xrecord_t **rhash;
 141         unsigned long *ha;
 142         char *rchg;
 143         long *rindex;
 144
 145         if (xdl_cha_init(&xdf->rcha, sizeof(xrecord_t), narec / 4 + 1) < 0) {
 146
 147                 return -1;
 148         }
 149         if (!(recs = (xrecord_t **) xdl_malloc(narec * sizeof(xrecord_t *)))) {
 150
 151                 xdl_cha_free(&xdf->rcha);
 152                 return -1;
 153         }
 154
 155         hbits = xdl_hashbits((unsigned int) narec);
 156         hsize = 1 << hbits;
 157         if (!(rhash = (xrecord_t **) xdl_malloc(hsize * sizeof(xrecord_t *)))) {
 158
 159                 xdl_free(recs);
 160                 xdl_cha_free(&xdf->rcha);
 161                 return -1;
 162         }
 163         for (i = 0; i < hsize; i++)
 164                 rhash[i] = NULL;
 165
 166         nrec = 0;
 167         if ((cur = blk = xdl_mmfile_first(mf, &bsize)) != NULL) {
 168                 for (top = blk + bsize;;) {
 169                         if (cur >= top) {
 170                                 if (!(cur = blk = xdl_mmfile_next(mf, &bsize)))
 171                                         break;
 172                                 top = blk + bsize;
 173                         }
 174                         prev = cur;
 175                         hav = xdl_hash_record(&cur, top);
 176                         if (nrec >= narec) {
 177                                 narec *= 2;
 178                                 if (!(rrecs = (xrecord_t **) xdl_realloc(recs, narec * sizeof(xrecord_t *)))) {
 179
 180                                         xdl_free(rhash);
 181                                         xdl_free(recs);
 182                                         xdl_cha_free(&xdf->rcha);
 183                                         return -1;
 184                                 }
 185                                 recs = rrecs;
 186                         }
 187                         if (!(crec = xdl_cha_alloc(&xdf->rcha))) {
 188
 189                                 xdl_free(rhash);
 190                                 xdl_free(recs);
 191                                 xdl_cha_free(&xdf->rcha);
 192                                 return -1;
 193                         }
 194                         crec->ptr = prev;
 195                         crec->size = (long) (cur - prev);
 196                         crec->ha = hav;
 197                         recs[nrec++] = crec;
 198
 199                         if (xdl_classify_record(cf, rhash, hbits, crec) < 0) {
 200
 201                                 xdl_free(rhash);
 202                                 xdl_free(recs);
 203                                 xdl_cha_free(&xdf->rcha);
 204                                 return -1;
 205                         }
 206                 }
 207         }
 208
 209         if (!(rchg = (char *) xdl_malloc((nrec + 2) * sizeof(char)))) {
 210
 211                 xdl_free(rhash);
 212                 xdl_free(recs);
 213                 xdl_cha_free(&xdf->rcha);
 214                 return -1;
 215         }
 216         memset(rchg, 0, (nrec + 2) * sizeof(char));
 217
 218         if (!(rindex = (long *) xdl_malloc((nrec + 1) * sizeof(long)))) {
 219
 220                 xdl_free(rchg);
 221                 xdl_free(rhash);
 222                 xdl_free(recs);
 223                 xdl_cha_free(&xdf->rcha);
 224                 return -1;
 225         }
 226         if (!(ha = (unsigned long *) xdl_malloc((nrec + 1) * sizeof(unsigned long)))) {
 227
 228                 xdl_free(rindex);
 229                 xdl_free(rchg);
 230                 xdl_free(rhash);
 231                 xdl_free(recs);
 232                 xdl_cha_free(&xdf->rcha);
 233                 return -1;
 234         }
 235
 236         xdf->nrec = nrec;
 237         xdf->recs = recs;
 238         xdf->hbits = hbits;
 239         xdf->rhash = rhash;
 240         xdf->rchg = rchg + 1;
 241         xdf->rindex = rindex;
 242         xdf->nreff = 0;
 243         xdf->ha = ha;
 244         xdf->dstart = 0;
 245         xdf->dend = nrec - 1;
 246
 247         return 0;
 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) {
 264         long enl1, enl2;
 265         xdlclassifier_t cf;
 266
 267         enl1 = xdl_guess_lines(mf1) + 1;
 268         enl2 = xdl_guess_lines(mf2) + 1;
 269
 270         if (xdl_init_classifier(&cf, enl1 + enl2 + 1) < 0) {
 271
 272                 return -1;
 273         }
 274
 275         if (xdl_prepare_ctx(mf1, enl1, xpp, &cf, &xe->xdf1) < 0) {
 276
 277                 xdl_free_classifier(&cf);
 278                 return -1;
 279         }
 280         if (xdl_prepare_ctx(mf2, enl2, xpp, &cf, &xe->xdf2) < 0) {
 281
 282                 xdl_free_ctx(&xe->xdf1);
 283                 xdl_free_classifier(&cf);
 284                 return -1;
 285         }
 286
 287         xdl_free_classifier(&cf);
 288
 289         if (xdl_optimize_ctxs(&xe->xdf1, &xe->xdf2) < 0) {
 290
 291                 xdl_free_ctx(&xe->xdf2);
 292                 xdl_free_ctx(&xe->xdf1);
 293                 return -1;
 294         }
 295
 296         return 0;
 297 }
 298
 299
 300 void xdl_free_env(xdfenv_t *xe) {
 301
 302         xdl_free_ctx(&xe->xdf2);
 303         xdl_free_ctx(&xe->xdf1);
 304 }
 305
 306
 307 static int xdl_clean_mmatch(char const *dis, long i, long s, long e) {
 308         long r, rdis, rpdis;
 309
 310         for (r = 1, rdis = 0, rpdis = 1; (i - r) >= s; r++) {
 311                 if (!dis[i - r])
 312                         rdis++;
 313                 else if (dis[i - r] == 2)
 314                         rpdis++;
 315                 else
 316                         break;
 317         }
 318         for (r = 1; (i + r) <= e; r++) {
 319                 if (!dis[i + r])
 320                         rdis++;
 321                 else if (dis[i + r] == 2)
 322                         rpdis++;
 323                 else
 324                         break;
 325         }
 326
 327         return rpdis * XDL_KPDIS_RUN < (rpdis + rdis);
 328 }
 329
 330
 331 /*
 332  * Try to reduce the problem complexity, discard records that have no
 333  * matches on the other file. Also, lines that have multiple matches
 334  * might be potentially discarded if they happear in a run of discardable.
 335  */
 336 static int xdl_cleanup_records(xdfile_t *xdf1, xdfile_t *xdf2) {
 337         long i, rhi, nreff;
 338         unsigned long hav;
 339         xrecord_t **recs;
 340         xrecord_t *rec;
 341         char *dis, *dis1, *dis2;
 342
 343         if (!(dis = (char *) xdl_malloc((xdf1->nrec + xdf2->nrec + 2) * sizeof(char)))) {
 344
 345                 return -1;
 346         }
 347         memset(dis, 0, (xdf1->nrec + xdf2->nrec + 2) * sizeof(char));
 348         dis1 = dis;
 349         dis2 = dis1 + xdf1->nrec + 1;
 350
 351         for (i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart]; i <= xdf1->dend; i++, recs++) {
 352                 hav = (*recs)->ha;
 353                 rhi = (long) XDL_HASHLONG(hav, xdf2->hbits);
 354                 for (rec = xdf2->rhash[rhi]; rec; rec = rec->next)
 355                         if (rec->ha == hav && ++dis1[i] == 2)
 356                                 break;
 357         }
 358
 359         for (i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart]; i <= xdf2->dend; i++, recs++) {
 360                 hav = (*recs)->ha;
 361                 rhi = (long) XDL_HASHLONG(hav, xdf1->hbits);
 362                 for (rec = xdf1->rhash[rhi]; rec; rec = rec->next)
 363                         if (rec->ha == hav && ++dis2[i] == 2)
 364                                 break;
 365         }
 366
 367         for (nreff = 0, i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart];
 368              i <= xdf1->dend; i++, recs++) {
 369                 if (dis1[i] == 1 ||
 370                     (dis1[i] == 2 && !xdl_clean_mmatch(dis1, i, xdf1->dstart, xdf1->dend))) {
 371                         xdf1->rindex[nreff] = i;
 372                         xdf1->ha[nreff] = (*recs)->ha;
 373                         nreff++;
 374                 } else
 375                         xdf1->rchg[i] = 1;
 376         }
 377         xdf1->nreff = nreff;
 378
 379         for (nreff = 0, i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart];
 380              i <= xdf2->dend; i++, recs++) {
 381                 if (dis2[i] == 1 ||
 382                     (dis2[i] == 2 && !xdl_clean_mmatch(dis2, i, xdf2->dstart, xdf2->dend))) {
 383                         xdf2->rindex[nreff] = i;
 384                         xdf2->ha[nreff] = (*recs)->ha;
 385                         nreff++;
 386                 } else
 387                         xdf2->rchg[i] = 1;
 388         }
 389         xdf2->nreff = nreff;
 390
 391         xdl_free(dis);
 392
 393         return 0;
 394 }
 395
 396
 397 /*
 398  * Early trim initial and terminal matching records.
 399  */
 400 static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2) {
 401         long i, lim;
 402         xrecord_t **recs1, **recs2;
 403
 404         recs1 = xdf1->recs;
 405         recs2 = xdf2->recs;
 406         for (i = 0, lim = XDL_MIN(xdf1->nrec, xdf2->nrec); i < lim;
 407              i++, recs1++, recs2++)
 408                 if ((*recs1)->ha != (*recs2)->ha)
 409                         break;
 410
 411         xdf1->dstart = xdf2->dstart = i;
 412
 413         recs1 = xdf1->recs + xdf1->nrec - 1;
 414         recs2 = xdf2->recs + xdf2->nrec - 1;
 415         for (lim -= i, i = 0; i < lim; i++, recs1--, recs2--)
 416                 if ((*recs1)->ha != (*recs2)->ha)
 417                         break;
 418
 419         xdf1->dend = xdf1->nrec - i - 1;
 420         xdf2->dend = xdf2->nrec - i - 1;
 421
 422         return 0;
 423 }
 424
 425
 426 static int xdl_optimize_ctxs(xdfile_t *xdf1, xdfile_t *xdf2) {
 427
 428         if (xdl_trim_ends(xdf1, xdf2) < 0 ||
 429             xdl_cleanup_records(xdf1, xdf2) < 0) {
 430
 431                 return -1;
 432         }
 433
 434         return 0;
 435 }
 436