xdiff / xdiffi.con commit xdl_change_compact(): introduce the concept of a change group (e8adf23)
   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_MAX_COST_MIN 256
  28#define XDL_HEUR_MIN_COST 256
  29#define XDL_LINE_MAX (long)((1UL << (CHAR_BIT * sizeof(long) - 1)) - 1)
  30#define XDL_SNAKE_CNT 20
  31#define XDL_K_HEUR 4
  32
  33
  34
  35typedef struct s_xdpsplit {
  36        long i1, i2;
  37        int min_lo, min_hi;
  38} xdpsplit_t;
  39
  40
  41
  42
  43static long xdl_split(unsigned long const *ha1, long off1, long lim1,
  44                      unsigned long const *ha2, long off2, long lim2,
  45                      long *kvdf, long *kvdb, int need_min, xdpsplit_t *spl,
  46                      xdalgoenv_t *xenv);
  47static xdchange_t *xdl_add_change(xdchange_t *xscr, long i1, long i2, long chg1, long chg2);
  48
  49
  50
  51
  52
  53/*
  54 * See "An O(ND) Difference Algorithm and its Variations", by Eugene Myers.
  55 * Basically considers a "box" (off1, off2, lim1, lim2) and scan from both
  56 * the forward diagonal starting from (off1, off2) and the backward diagonal
  57 * starting from (lim1, lim2). If the K values on the same diagonal crosses
  58 * returns the furthest point of reach. We might end up having to expensive
  59 * cases using this algorithm is full, so a little bit of heuristic is needed
  60 * to cut the search and to return a suboptimal point.
  61 */
  62static long xdl_split(unsigned long const *ha1, long off1, long lim1,
  63                      unsigned long const *ha2, long off2, long lim2,
  64                      long *kvdf, long *kvdb, int need_min, xdpsplit_t *spl,
  65                      xdalgoenv_t *xenv) {
  66        long dmin = off1 - lim2, dmax = lim1 - off2;
  67        long fmid = off1 - off2, bmid = lim1 - lim2;
  68        long odd = (fmid - bmid) & 1;
  69        long fmin = fmid, fmax = fmid;
  70        long bmin = bmid, bmax = bmid;
  71        long ec, d, i1, i2, prev1, best, dd, v, k;
  72
  73        /*
  74         * Set initial diagonal values for both forward and backward path.
  75         */
  76        kvdf[fmid] = off1;
  77        kvdb[bmid] = lim1;
  78
  79        for (ec = 1;; ec++) {
  80                int got_snake = 0;
  81
  82                /*
  83                 * We need to extent the diagonal "domain" by one. If the next
  84                 * values exits the box boundaries we need to change it in the
  85                 * opposite direction because (max - min) must be a power of two.
  86                 * Also we initialize the external K value to -1 so that we can
  87                 * avoid extra conditions check inside the core loop.
  88                 */
  89                if (fmin > dmin)
  90                        kvdf[--fmin - 1] = -1;
  91                else
  92                        ++fmin;
  93                if (fmax < dmax)
  94                        kvdf[++fmax + 1] = -1;
  95                else
  96                        --fmax;
  97
  98                for (d = fmax; d >= fmin; d -= 2) {
  99                        if (kvdf[d - 1] >= kvdf[d + 1])
 100                                i1 = kvdf[d - 1] + 1;
 101                        else
 102                                i1 = kvdf[d + 1];
 103                        prev1 = i1;
 104                        i2 = i1 - d;
 105                        for (; i1 < lim1 && i2 < lim2 && ha1[i1] == ha2[i2]; i1++, i2++);
 106                        if (i1 - prev1 > xenv->snake_cnt)
 107                                got_snake = 1;
 108                        kvdf[d] = i1;
 109                        if (odd && bmin <= d && d <= bmax && kvdb[d] <= i1) {
 110                                spl->i1 = i1;
 111                                spl->i2 = i2;
 112                                spl->min_lo = spl->min_hi = 1;
 113                                return ec;
 114                        }
 115                }
 116
 117                /*
 118                 * We need to extent the diagonal "domain" by one. If the next
 119                 * values exits the box boundaries we need to change it in the
 120                 * opposite direction because (max - min) must be a power of two.
 121                 * Also we initialize the external K value to -1 so that we can
 122                 * avoid extra conditions check inside the core loop.
 123                 */
 124                if (bmin > dmin)
 125                        kvdb[--bmin - 1] = XDL_LINE_MAX;
 126                else
 127                        ++bmin;
 128                if (bmax < dmax)
 129                        kvdb[++bmax + 1] = XDL_LINE_MAX;
 130                else
 131                        --bmax;
 132
 133                for (d = bmax; d >= bmin; d -= 2) {
 134                        if (kvdb[d - 1] < kvdb[d + 1])
 135                                i1 = kvdb[d - 1];
 136                        else
 137                                i1 = kvdb[d + 1] - 1;
 138                        prev1 = i1;
 139                        i2 = i1 - d;
 140                        for (; i1 > off1 && i2 > off2 && ha1[i1 - 1] == ha2[i2 - 1]; i1--, i2--);
 141                        if (prev1 - i1 > xenv->snake_cnt)
 142                                got_snake = 1;
 143                        kvdb[d] = i1;
 144                        if (!odd && fmin <= d && d <= fmax && i1 <= kvdf[d]) {
 145                                spl->i1 = i1;
 146                                spl->i2 = i2;
 147                                spl->min_lo = spl->min_hi = 1;
 148                                return ec;
 149                        }
 150                }
 151
 152                if (need_min)
 153                        continue;
 154
 155                /*
 156                 * If the edit cost is above the heuristic trigger and if
 157                 * we got a good snake, we sample current diagonals to see
 158                 * if some of the, have reached an "interesting" path. Our
 159                 * measure is a function of the distance from the diagonal
 160                 * corner (i1 + i2) penalized with the distance from the
 161                 * mid diagonal itself. If this value is above the current
 162                 * edit cost times a magic factor (XDL_K_HEUR) we consider
 163                 * it interesting.
 164                 */
 165                if (got_snake && ec > xenv->heur_min) {
 166                        for (best = 0, d = fmax; d >= fmin; d -= 2) {
 167                                dd = d > fmid ? d - fmid: fmid - d;
 168                                i1 = kvdf[d];
 169                                i2 = i1 - d;
 170                                v = (i1 - off1) + (i2 - off2) - dd;
 171
 172                                if (v > XDL_K_HEUR * ec && v > best &&
 173                                    off1 + xenv->snake_cnt <= i1 && i1 < lim1 &&
 174                                    off2 + xenv->snake_cnt <= i2 && i2 < lim2) {
 175                                        for (k = 1; ha1[i1 - k] == ha2[i2 - k]; k++)
 176                                                if (k == xenv->snake_cnt) {
 177                                                        best = v;
 178                                                        spl->i1 = i1;
 179                                                        spl->i2 = i2;
 180                                                        break;
 181                                                }
 182                                }
 183                        }
 184                        if (best > 0) {
 185                                spl->min_lo = 1;
 186                                spl->min_hi = 0;
 187                                return ec;
 188                        }
 189
 190                        for (best = 0, d = bmax; d >= bmin; d -= 2) {
 191                                dd = d > bmid ? d - bmid: bmid - d;
 192                                i1 = kvdb[d];
 193                                i2 = i1 - d;
 194                                v = (lim1 - i1) + (lim2 - i2) - dd;
 195
 196                                if (v > XDL_K_HEUR * ec && v > best &&
 197                                    off1 < i1 && i1 <= lim1 - xenv->snake_cnt &&
 198                                    off2 < i2 && i2 <= lim2 - xenv->snake_cnt) {
 199                                        for (k = 0; ha1[i1 + k] == ha2[i2 + k]; k++)
 200                                                if (k == xenv->snake_cnt - 1) {
 201                                                        best = v;
 202                                                        spl->i1 = i1;
 203                                                        spl->i2 = i2;
 204                                                        break;
 205                                                }
 206                                }
 207                        }
 208                        if (best > 0) {
 209                                spl->min_lo = 0;
 210                                spl->min_hi = 1;
 211                                return ec;
 212                        }
 213                }
 214
 215                /*
 216                 * Enough is enough. We spent too much time here and now we collect
 217                 * the furthest reaching path using the (i1 + i2) measure.
 218                 */
 219                if (ec >= xenv->mxcost) {
 220                        long fbest, fbest1, bbest, bbest1;
 221
 222                        fbest = fbest1 = -1;
 223                        for (d = fmax; d >= fmin; d -= 2) {
 224                                i1 = XDL_MIN(kvdf[d], lim1);
 225                                i2 = i1 - d;
 226                                if (lim2 < i2)
 227                                        i1 = lim2 + d, i2 = lim2;
 228                                if (fbest < i1 + i2) {
 229                                        fbest = i1 + i2;
 230                                        fbest1 = i1;
 231                                }
 232                        }
 233
 234                        bbest = bbest1 = XDL_LINE_MAX;
 235                        for (d = bmax; d >= bmin; d -= 2) {
 236                                i1 = XDL_MAX(off1, kvdb[d]);
 237                                i2 = i1 - d;
 238                                if (i2 < off2)
 239                                        i1 = off2 + d, i2 = off2;
 240                                if (i1 + i2 < bbest) {
 241                                        bbest = i1 + i2;
 242                                        bbest1 = i1;
 243                                }
 244                        }
 245
 246                        if ((lim1 + lim2) - bbest < fbest - (off1 + off2)) {
 247                                spl->i1 = fbest1;
 248                                spl->i2 = fbest - fbest1;
 249                                spl->min_lo = 1;
 250                                spl->min_hi = 0;
 251                        } else {
 252                                spl->i1 = bbest1;
 253                                spl->i2 = bbest - bbest1;
 254                                spl->min_lo = 0;
 255                                spl->min_hi = 1;
 256                        }
 257                        return ec;
 258                }
 259        }
 260}
 261
 262
 263/*
 264 * Rule: "Divide et Impera". Recursively split the box in sub-boxes by calling
 265 * the box splitting function. Note that the real job (marking changed lines)
 266 * is done in the two boundary reaching checks.
 267 */
 268int xdl_recs_cmp(diffdata_t *dd1, long off1, long lim1,
 269                 diffdata_t *dd2, long off2, long lim2,
 270                 long *kvdf, long *kvdb, int need_min, xdalgoenv_t *xenv) {
 271        unsigned long const *ha1 = dd1->ha, *ha2 = dd2->ha;
 272
 273        /*
 274         * Shrink the box by walking through each diagonal snake (SW and NE).
 275         */
 276        for (; off1 < lim1 && off2 < lim2 && ha1[off1] == ha2[off2]; off1++, off2++);
 277        for (; off1 < lim1 && off2 < lim2 && ha1[lim1 - 1] == ha2[lim2 - 1]; lim1--, lim2--);
 278
 279        /*
 280         * If one dimension is empty, then all records on the other one must
 281         * be obviously changed.
 282         */
 283        if (off1 == lim1) {
 284                char *rchg2 = dd2->rchg;
 285                long *rindex2 = dd2->rindex;
 286
 287                for (; off2 < lim2; off2++)
 288                        rchg2[rindex2[off2]] = 1;
 289        } else if (off2 == lim2) {
 290                char *rchg1 = dd1->rchg;
 291                long *rindex1 = dd1->rindex;
 292
 293                for (; off1 < lim1; off1++)
 294                        rchg1[rindex1[off1]] = 1;
 295        } else {
 296                xdpsplit_t spl;
 297                spl.i1 = spl.i2 = 0;
 298
 299                /*
 300                 * Divide ...
 301                 */
 302                if (xdl_split(ha1, off1, lim1, ha2, off2, lim2, kvdf, kvdb,
 303                              need_min, &spl, xenv) < 0) {
 304
 305                        return -1;
 306                }
 307
 308                /*
 309                 * ... et Impera.
 310                 */
 311                if (xdl_recs_cmp(dd1, off1, spl.i1, dd2, off2, spl.i2,
 312                                 kvdf, kvdb, spl.min_lo, xenv) < 0 ||
 313                    xdl_recs_cmp(dd1, spl.i1, lim1, dd2, spl.i2, lim2,
 314                                 kvdf, kvdb, spl.min_hi, xenv) < 0) {
 315
 316                        return -1;
 317                }
 318        }
 319
 320        return 0;
 321}
 322
 323
 324int xdl_do_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
 325                xdfenv_t *xe) {
 326        long ndiags;
 327        long *kvd, *kvdf, *kvdb;
 328        xdalgoenv_t xenv;
 329        diffdata_t dd1, dd2;
 330
 331        if (XDF_DIFF_ALG(xpp->flags) == XDF_PATIENCE_DIFF)
 332                return xdl_do_patience_diff(mf1, mf2, xpp, xe);
 333
 334        if (XDF_DIFF_ALG(xpp->flags) == XDF_HISTOGRAM_DIFF)
 335                return xdl_do_histogram_diff(mf1, mf2, xpp, xe);
 336
 337        if (xdl_prepare_env(mf1, mf2, xpp, xe) < 0) {
 338
 339                return -1;
 340        }
 341
 342        /*
 343         * Allocate and setup K vectors to be used by the differential algorithm.
 344         * One is to store the forward path and one to store the backward path.
 345         */
 346        ndiags = xe->xdf1.nreff + xe->xdf2.nreff + 3;
 347        if (!(kvd = (long *) xdl_malloc((2 * ndiags + 2) * sizeof(long)))) {
 348
 349                xdl_free_env(xe);
 350                return -1;
 351        }
 352        kvdf = kvd;
 353        kvdb = kvdf + ndiags;
 354        kvdf += xe->xdf2.nreff + 1;
 355        kvdb += xe->xdf2.nreff + 1;
 356
 357        xenv.mxcost = xdl_bogosqrt(ndiags);
 358        if (xenv.mxcost < XDL_MAX_COST_MIN)
 359                xenv.mxcost = XDL_MAX_COST_MIN;
 360        xenv.snake_cnt = XDL_SNAKE_CNT;
 361        xenv.heur_min = XDL_HEUR_MIN_COST;
 362
 363        dd1.nrec = xe->xdf1.nreff;
 364        dd1.ha = xe->xdf1.ha;
 365        dd1.rchg = xe->xdf1.rchg;
 366        dd1.rindex = xe->xdf1.rindex;
 367        dd2.nrec = xe->xdf2.nreff;
 368        dd2.ha = xe->xdf2.ha;
 369        dd2.rchg = xe->xdf2.rchg;
 370        dd2.rindex = xe->xdf2.rindex;
 371
 372        if (xdl_recs_cmp(&dd1, 0, dd1.nrec, &dd2, 0, dd2.nrec,
 373                         kvdf, kvdb, (xpp->flags & XDF_NEED_MINIMAL) != 0, &xenv) < 0) {
 374
 375                xdl_free(kvd);
 376                xdl_free_env(xe);
 377                return -1;
 378        }
 379
 380        xdl_free(kvd);
 381
 382        return 0;
 383}
 384
 385
 386static xdchange_t *xdl_add_change(xdchange_t *xscr, long i1, long i2, long chg1, long chg2) {
 387        xdchange_t *xch;
 388
 389        if (!(xch = (xdchange_t *) xdl_malloc(sizeof(xdchange_t))))
 390                return NULL;
 391
 392        xch->next = xscr;
 393        xch->i1 = i1;
 394        xch->i2 = i2;
 395        xch->chg1 = chg1;
 396        xch->chg2 = chg2;
 397        xch->ignore = 0;
 398
 399        return xch;
 400}
 401
 402
 403static int is_blank_line(xrecord_t *rec, long flags)
 404{
 405        return xdl_blankline(rec->ptr, rec->size, flags);
 406}
 407
 408static int recs_match(xrecord_t *rec1, xrecord_t *rec2, long flags)
 409{
 410        return (rec1->ha == rec2->ha &&
 411                xdl_recmatch(rec1->ptr, rec1->size,
 412                             rec2->ptr, rec2->size,
 413                             flags));
 414}
 415
 416/*
 417 * Represent a group of changed lines in an xdfile_t (i.e., a contiguous group
 418 * of lines that was inserted or deleted from the corresponding version of the
 419 * file). We consider there to be such a group at the beginning of the file, at
 420 * the end of the file, and between any two unchanged lines, though most such
 421 * groups will usually be empty.
 422 *
 423 * If the first line in a group is equal to the line following the group, then
 424 * the group can be slid down. Similarly, if the last line in a group is equal
 425 * to the line preceding the group, then the group can be slid up. See
 426 * group_slide_down() and group_slide_up().
 427 *
 428 * Note that loops that are testing for changed lines in xdf->rchg do not need
 429 * index bounding since the array is prepared with a zero at position -1 and N.
 430 */
 431struct group {
 432        /*
 433         * The index of the first changed line in the group, or the index of
 434         * the unchanged line above which the (empty) group is located.
 435         */
 436        long start;
 437
 438        /*
 439         * The index of the first unchanged line after the group. For an empty
 440         * group, end is equal to start.
 441         */
 442        long end;
 443};
 444
 445/*
 446 * Initialize g to point at the first group in xdf.
 447 */
 448static void group_init(xdfile_t *xdf, struct group *g)
 449{
 450        g->start = g->end = 0;
 451        while (xdf->rchg[g->end])
 452                g->end++;
 453}
 454
 455/*
 456 * Move g to describe the next (possibly empty) group in xdf and return 0. If g
 457 * is already at the end of the file, do nothing and return -1.
 458 */
 459static inline int group_next(xdfile_t *xdf, struct group *g)
 460{
 461        if (g->end == xdf->nrec)
 462                return -1;
 463
 464        g->start = g->end + 1;
 465        for (g->end = g->start; xdf->rchg[g->end]; g->end++)
 466                ;
 467
 468        return 0;
 469}
 470
 471/*
 472 * Move g to describe the previous (possibly empty) group in xdf and return 0.
 473 * If g is already at the beginning of the file, do nothing and return -1.
 474 */
 475static inline int group_previous(xdfile_t *xdf, struct group *g)
 476{
 477        if (g->start == 0)
 478                return -1;
 479
 480        g->end = g->start - 1;
 481        for (g->start = g->end; xdf->rchg[g->start - 1]; g->start--)
 482                ;
 483
 484        return 0;
 485}
 486
 487/*
 488 * If g can be slid toward the end of the file, do so, and if it bumps into a
 489 * following group, expand this group to include it. Return 0 on success or -1
 490 * if g cannot be slid down.
 491 */
 492static int group_slide_down(xdfile_t *xdf, struct group *g, long flags)
 493{
 494        if (g->end < xdf->nrec &&
 495            recs_match(xdf->recs[g->start], xdf->recs[g->end], flags)) {
 496                xdf->rchg[g->start++] = 0;
 497                xdf->rchg[g->end++] = 1;
 498
 499                while (xdf->rchg[g->end])
 500                        g->end++;
 501
 502                return 0;
 503        } else {
 504                return -1;
 505        }
 506}
 507
 508/*
 509 * If g can be slid toward the beginning of the file, do so, and if it bumps
 510 * into a previous group, expand this group to include it. Return 0 on success
 511 * or -1 if g cannot be slid up.
 512 */
 513static int group_slide_up(xdfile_t *xdf, struct group *g, long flags)
 514{
 515        if (g->start > 0 &&
 516            recs_match(xdf->recs[g->start - 1], xdf->recs[g->end - 1], flags)) {
 517                xdf->rchg[--g->start] = 1;
 518                xdf->rchg[--g->end] = 0;
 519
 520                while (xdf->rchg[g->start - 1])
 521                        g->start--;
 522
 523                return 0;
 524        } else {
 525                return -1;
 526        }
 527}
 528
 529static void xdl_bug(const char *msg)
 530{
 531        fprintf(stderr, "BUG: %s\n", msg);
 532        exit(1);
 533}
 534
 535/*
 536 * Move back and forward change groups for a consistent and pretty diff output.
 537 * This also helps in finding joinable change groups and reducing the diff
 538 * size.
 539 */
 540int xdl_change_compact(xdfile_t *xdf, xdfile_t *xdfo, long flags) {
 541        struct group g, go;
 542        long earliest_end, end_matching_other;
 543        long groupsize;
 544        unsigned int blank_lines;
 545
 546        group_init(xdf, &g);
 547        group_init(xdfo, &go);
 548
 549        while (1) {
 550                /* If the group is empty in the to-be-compacted file, skip it: */
 551                if (g.end == g.start)
 552                        goto next;
 553
 554                /*
 555                 * Now shift the change up and then down as far as possible in
 556                 * each direction. If it bumps into any other changes, merge them.
 557                 */
 558                do {
 559                        groupsize = g.end - g.start;
 560
 561                        /*
 562                         * Keep track of the last "end" index that causes this
 563                         * group to align with a group of changed lines in the
 564                         * other file. -1 indicates that we haven't found such
 565                         * a match yet:
 566                         */
 567                        end_matching_other = -1;
 568
 569                        /*
 570                         * Boolean value that records whether there are any blank
 571                         * lines that could be made to be the last line of this
 572                         * group.
 573                         */
 574                        blank_lines = 0;
 575
 576                        /* Shift the group backward as much as possible: */
 577                        while (!group_slide_up(xdf, &g, flags))
 578                                if (group_previous(xdfo, &go))
 579                                        xdl_bug("group sync broken sliding up");
 580
 581                        /*
 582                         * This is this highest that this group can be shifted.
 583                         * Record its end index:
 584                         */
 585                        earliest_end = g.end;
 586
 587                        if (go.end > go.start)
 588                                end_matching_other = g.end;
 589
 590                        /* Now shift the group forward as far as possible: */
 591                        while (1) {
 592                                if (!blank_lines)
 593                                        blank_lines = is_blank_line(
 594                                                        xdf->recs[g.end - 1],
 595                                                        flags);
 596
 597                                if (group_slide_down(xdf, &g, flags))
 598                                        break;
 599                                if (group_next(xdfo, &go))
 600                                        xdl_bug("group sync broken sliding down");
 601
 602                                if (go.end > go.start)
 603                                        end_matching_other = g.end;
 604                        }
 605                } while (groupsize != g.end - g.start);
 606
 607                if (g.end == earliest_end) {
 608                        /* no shifting was possible */
 609                } else if (end_matching_other != -1) {
 610                        /*
 611                         * Move the possibly merged group of changes back to line
 612                         * up with the last group of changes from the other file
 613                         * that it can align with.
 614                         */
 615                        while (go.end == go.start) {
 616                                if (group_slide_up(xdf, &g, flags))
 617                                        xdl_bug("match disappeared");
 618                                if (group_previous(xdfo, &go))
 619                                        xdl_bug("group sync broken sliding to match");
 620                        }
 621                } else if ((flags & XDF_COMPACTION_HEURISTIC) && blank_lines) {
 622                        /*
 623                         * Compaction heuristic: if it is possible to shift the
 624                         * group to make its bottom line a blank line, do so.
 625                         *
 626                         * As we already shifted the group forward as far as
 627                         * possible in the earlier loop, we only need to handle
 628                         * backward shifts, not forward ones.
 629                         */
 630                        while (!is_blank_line(xdf->recs[g.end - 1], flags)) {
 631                                if (group_slide_up(xdf, &g, flags))
 632                                        xdl_bug("blank line disappeared");
 633                                if (group_previous(xdfo, &go))
 634                                        xdl_bug("group sync broken sliding to blank line");
 635                        }
 636                }
 637
 638        next:
 639                /* Move past the just-processed group: */
 640                if (group_next(xdf, &g))
 641                        break;
 642                if (group_next(xdfo, &go))
 643                        xdl_bug("group sync broken moving to next group");
 644        }
 645
 646        if (!group_next(xdfo, &go))
 647                xdl_bug("group sync broken at end of file");
 648
 649        return 0;
 650}
 651
 652
 653int xdl_build_script(xdfenv_t *xe, xdchange_t **xscr) {
 654        xdchange_t *cscr = NULL, *xch;
 655        char *rchg1 = xe->xdf1.rchg, *rchg2 = xe->xdf2.rchg;
 656        long i1, i2, l1, l2;
 657
 658        /*
 659         * Trivial. Collects "groups" of changes and creates an edit script.
 660         */
 661        for (i1 = xe->xdf1.nrec, i2 = xe->xdf2.nrec; i1 >= 0 || i2 >= 0; i1--, i2--)
 662                if (rchg1[i1 - 1] || rchg2[i2 - 1]) {
 663                        for (l1 = i1; rchg1[i1 - 1]; i1--);
 664                        for (l2 = i2; rchg2[i2 - 1]; i2--);
 665
 666                        if (!(xch = xdl_add_change(cscr, i1, i2, l1 - i1, l2 - i2))) {
 667                                xdl_free_script(cscr);
 668                                return -1;
 669                        }
 670                        cscr = xch;
 671                }
 672
 673        *xscr = cscr;
 674
 675        return 0;
 676}
 677
 678
 679void xdl_free_script(xdchange_t *xscr) {
 680        xdchange_t *xch;
 681
 682        while ((xch = xscr) != NULL) {
 683                xscr = xscr->next;
 684                xdl_free(xch);
 685        }
 686}
 687
 688static int xdl_call_hunk_func(xdfenv_t *xe, xdchange_t *xscr, xdemitcb_t *ecb,
 689                              xdemitconf_t const *xecfg)
 690{
 691        xdchange_t *xch, *xche;
 692
 693        for (xch = xscr; xch; xch = xche->next) {
 694                xche = xdl_get_hunk(&xch, xecfg);
 695                if (!xch)
 696                        break;
 697                if (xecfg->hunk_func(xch->i1, xche->i1 + xche->chg1 - xch->i1,
 698                                     xch->i2, xche->i2 + xche->chg2 - xch->i2,
 699                                     ecb->priv) < 0)
 700                        return -1;
 701        }
 702        return 0;
 703}
 704
 705static void xdl_mark_ignorable(xdchange_t *xscr, xdfenv_t *xe, long flags)
 706{
 707        xdchange_t *xch;
 708
 709        for (xch = xscr; xch; xch = xch->next) {
 710                int ignore = 1;
 711                xrecord_t **rec;
 712                long i;
 713
 714                rec = &xe->xdf1.recs[xch->i1];
 715                for (i = 0; i < xch->chg1 && ignore; i++)
 716                        ignore = xdl_blankline(rec[i]->ptr, rec[i]->size, flags);
 717
 718                rec = &xe->xdf2.recs[xch->i2];
 719                for (i = 0; i < xch->chg2 && ignore; i++)
 720                        ignore = xdl_blankline(rec[i]->ptr, rec[i]->size, flags);
 721
 722                xch->ignore = ignore;
 723        }
 724}
 725
 726int xdl_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
 727             xdemitconf_t const *xecfg, xdemitcb_t *ecb) {
 728        xdchange_t *xscr;
 729        xdfenv_t xe;
 730        emit_func_t ef = xecfg->hunk_func ? xdl_call_hunk_func : xdl_emit_diff;
 731
 732        if (xdl_do_diff(mf1, mf2, xpp, &xe) < 0) {
 733
 734                return -1;
 735        }
 736        if (xdl_change_compact(&xe.xdf1, &xe.xdf2, xpp->flags) < 0 ||
 737            xdl_change_compact(&xe.xdf2, &xe.xdf1, xpp->flags) < 0 ||
 738            xdl_build_script(&xe, &xscr) < 0) {
 739
 740                xdl_free_env(&xe);
 741                return -1;
 742        }
 743        if (xscr) {
 744                if (xpp->flags & XDF_IGNORE_BLANK_LINES)
 745                        xdl_mark_ignorable(xscr, &xe, xpp->flags);
 746
 747                if (ef(&xe, xscr, ecb, xecfg) < 0) {
 748
 749                        xdl_free_script(xscr);
 750                        xdl_free_env(&xe);
 751                        return -1;
 752                }
 753                xdl_free_script(xscr);
 754        }
 755        xdl_free_env(&xe);
 756
 757        return 0;
 758}