xdiff/xpatience.c

   1 /*
   2  *  LibXDiff by Davide Libenzi ( File Differential Library )
   3  *  Copyright (C) 2003-2016 Davide Libenzi, Johannes E. Schindelin
   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, see
  17  *  <http://www.gnu.org/licenses/>.
  18  *
  19  *  Davide Libenzi <davidel@xmailserver.org>
  20  *
  21  */
  22 #include "xinclude.h"
  23 #include "xtypes.h"
  24 #include "xdiff.h"
  25
  26 /*
  27  * The basic idea of patience diff is to find lines that are unique in
  28  * both files.  These are intuitively the ones that we want to see as
  29  * common lines.
  30  *
  31  * The maximal ordered sequence of such line pairs (where ordered means
  32  * that the order in the sequence agrees with the order of the lines in
  33  * both files) naturally defines an initial set of common lines.
  34  *
  35  * Now, the algorithm tries to extend the set of common lines by growing
  36  * the line ranges where the files have identical lines.
  37  *
  38  * Between those common lines, the patience diff algorithm is applied
  39  * recursively, until no unique line pairs can be found; these line ranges
  40  * are handled by the well-known Myers algorithm.
  41  */
  42
  43 #define NON_UNIQUE ULONG_MAX
  44
  45 /*
  46  * This is a hash mapping from line hash to line numbers in the first and
  47  * second file.
  48  */
  49 struct hashmap {
  50         int nr, alloc;
  51         struct entry {
  52                 unsigned long hash;
  53                 /*
  54                  * 0 = unused entry, 1 = first line, 2 = second, etc.
  55                  * line2 is NON_UNIQUE if the line is not unique
  56                  * in either the first or the second file.
  57                  */
  58                 unsigned long line1, line2;
  59                 /*
  60                  * "next" & "previous" are used for the longest common
  61                  * sequence;
  62                  * initially, "next" reflects only the order in file1.
  63                  */
  64                 struct entry *next, *previous;
  65
  66                 /*
  67                  * If 1, this entry can serve as an anchor. See
  68                  * Documentation/diff-options.txt for more information.
  69                  */
  70                 unsigned anchor : 1;
  71         } *entries, *first, *last;
  72         /* were common records found? */
  73         unsigned long has_matches;
  74         mmfile_t *file1, *file2;
  75         xdfenv_t *env;
  76         xpparam_t const *xpp;
  77 };
  78
  79 static int is_anchor(xpparam_t const *xpp, const char *line)
  80 {
  81         int i;
  82         for (i = 0; i < xpp->anchors_nr; i++) {
  83                 if (!strncmp(line, xpp->anchors[i], strlen(xpp->anchors[i])))
  84                         return 1;
  85         }
  86         return 0;
  87 }
  88
  89 /* The argument "pass" is 1 for the first file, 2 for the second. */
  90 static void insert_record(xpparam_t const *xpp, int line, struct hashmap *map,
  91                           int pass)
  92 {
  93         xrecord_t **records = pass == 1 ?
  94                 map->env->xdf1.recs : map->env->xdf2.recs;
  95         xrecord_t *record = records[line - 1], *other;
  96         /*
  97          * After xdl_prepare_env() (or more precisely, due to
  98          * xdl_classify_record()), the "ha" member of the records (AKA lines)
  99          * is _not_ the hash anymore, but a linearized version of it.  In
 100          * other words, the "ha" member is guaranteed to start with 0 and
 101          * the second record's ha can only be 0 or 1, etc.
 102          *
 103          * So we multiply ha by 2 in the hope that the hashing was
 104          * "unique enough".
 105          */
 106         int index = (int)((record->ha << 1) % map->alloc);
 107
 108         while (map->entries[index].line1) {
 109                 other = map->env->xdf1.recs[map->entries[index].line1 - 1];
 110                 if (map->entries[index].hash != record->ha ||
 111                                 !xdl_recmatch(record->ptr, record->size,
 112                                         other->ptr, other->size,
 113                                         map->xpp->flags)) {
 114                         if (++index >= map->alloc)
 115                                 index = 0;
 116                         continue;
 117                 }
 118                 if (pass == 2)
 119                         map->has_matches = 1;
 120                 if (pass == 1 || map->entries[index].line2)
 121                         map->entries[index].line2 = NON_UNIQUE;
 122                 else
 123                         map->entries[index].line2 = line;
 124                 return;
 125         }
 126         if (pass == 2)
 127                 return;
 128         map->entries[index].line1 = line;
 129         map->entries[index].hash = record->ha;
 130         map->entries[index].anchor = is_anchor(xpp, map->env->xdf1.recs[line - 1]->ptr);
 131         if (!map->first)
 132                 map->first = map->entries + index;
 133         if (map->last) {
 134                 map->last->next = map->entries + index;
 135                 map->entries[index].previous = map->last;
 136         }
 137         map->last = map->entries + index;
 138         map->nr++;
 139 }
 140
 141 /*
 142  * This function has to be called for each recursion into the inter-hunk
 143  * parts, as previously non-unique lines can become unique when being
 144  * restricted to a smaller part of the files.
 145  *
 146  * It is assumed that env has been prepared using xdl_prepare().
 147  */
 148 static int fill_hashmap(mmfile_t *file1, mmfile_t *file2,
 149                 xpparam_t const *xpp, xdfenv_t *env,
 150                 struct hashmap *result,
 151                 int line1, int count1, int line2, int count2)
 152 {
 153         result->file1 = file1;
 154         result->file2 = file2;
 155         result->xpp = xpp;
 156         result->env = env;
 157
 158         /* We know exactly how large we want the hash map */
 159         result->alloc = count1 * 2;
 160         result->entries = (struct entry *)
 161                 xdl_malloc(result->alloc * sizeof(struct entry));
 162         if (!result->entries)
 163                 return -1;
 164         memset(result->entries, 0, result->alloc * sizeof(struct entry));
 165
 166         /* First, fill with entries from the first file */
 167         while (count1--)
 168                 insert_record(xpp, line1++, result, 1);
 169
 170         /* Then search for matches in the second file */
 171         while (count2--)
 172                 insert_record(xpp, line2++, result, 2);
 173
 174         return 0;
 175 }
 176
 177 /*
 178  * Find the longest sequence with a smaller last element (meaning a smaller
 179  * line2, as we construct the sequence with entries ordered by line1).
 180  */
 181 static int binary_search(struct entry **sequence, int longest,
 182                 struct entry *entry)
 183 {
 184         int left = -1, right = longest;
 185
 186         while (left + 1 < right) {
 187                 int middle = left + (right - left) / 2;
 188                 /* by construction, no two entries can be equal */
 189                 if (sequence[middle]->line2 > entry->line2)
 190                         right = middle;
 191                 else
 192                         left = middle;
 193         }
 194         /* return the index in "sequence", _not_ the sequence length */
 195         return left;
 196 }
 197
 198 /*
 199  * The idea is to start with the list of common unique lines sorted by
 200  * the order in file1.  For each of these pairs, the longest (partial)
 201  * sequence whose last element's line2 is smaller is determined.
 202  *
 203  * For efficiency, the sequences are kept in a list containing exactly one
 204  * item per sequence length: the sequence with the smallest last
 205  * element (in terms of line2).
 206  */
 207 static struct entry *find_longest_common_sequence(struct hashmap *map)
 208 {
 209         struct entry **sequence = xdl_malloc(map->nr * sizeof(struct entry *));
 210         int longest = 0, i;
 211         struct entry *entry;
 212
 213         /*
 214          * If not -1, this entry in sequence must never be overridden.
 215          * Therefore, overriding entries before this has no effect, so
 216          * do not do that either.
 217          */
 218         int anchor_i = -1;
 219
 220         for (entry = map->first; entry; entry = entry->next) {
 221                 if (!entry->line2 || entry->line2 == NON_UNIQUE)
 222                         continue;
 223                 i = binary_search(sequence, longest, entry);
 224                 entry->previous = i < 0 ? NULL : sequence[i];
 225                 ++i;
 226                 if (i <= anchor_i)
 227                         continue;
 228                 sequence[i] = entry;
 229                 if (entry->anchor) {
 230                         anchor_i = i;
 231                         longest = anchor_i + 1;
 232                 } else if (i == longest) {
 233                         longest++;
 234                 }
 235         }
 236
 237         /* No common unique lines were found */
 238         if (!longest) {
 239                 xdl_free(sequence);
 240                 return NULL;
 241         }
 242
 243         /* Iterate starting at the last element, adjusting the "next" members */
 244         entry = sequence[longest - 1];
 245         entry->next = NULL;
 246         while (entry->previous) {
 247                 entry->previous->next = entry;
 248                 entry = entry->previous;
 249         }
 250         xdl_free(sequence);
 251         return entry;
 252 }
 253
 254 static int match(struct hashmap *map, int line1, int line2)
 255 {
 256         xrecord_t *record1 = map->env->xdf1.recs[line1 - 1];
 257         xrecord_t *record2 = map->env->xdf2.recs[line2 - 1];
 258         return xdl_recmatch(record1->ptr, record1->size,
 259                 record2->ptr, record2->size, map->xpp->flags);
 260 }
 261
 262 static int patience_diff(mmfile_t *file1, mmfile_t *file2,
 263                 xpparam_t const *xpp, xdfenv_t *env,
 264                 int line1, int count1, int line2, int count2);
 265
 266 static int walk_common_sequence(struct hashmap *map, struct entry *first,
 267                 int line1, int count1, int line2, int count2)
 268 {
 269         int end1 = line1 + count1, end2 = line2 + count2;
 270         int next1, next2;
 271
 272         for (;;) {
 273                 /* Try to grow the line ranges of common lines */
 274                 if (first) {
 275                         next1 = first->line1;
 276                         next2 = first->line2;
 277                         while (next1 > line1 && next2 > line2 &&
 278                                         match(map, next1 - 1, next2 - 1)) {
 279                                 next1--;
 280                                 next2--;
 281                         }
 282                 } else {
 283                         next1 = end1;
 284                         next2 = end2;
 285                 }
 286                 while (line1 < next1 && line2 < next2 &&
 287                                 match(map, line1, line2)) {
 288                         line1++;
 289                         line2++;
 290                 }
 291
 292                 /* Recurse */
 293                 if (next1 > line1 || next2 > line2) {
 294                         struct hashmap submap;
 295
 296                         memset(&submap, 0, sizeof(submap));
 297                         if (patience_diff(map->file1, map->file2,
 298                                         map->xpp, map->env,
 299                                         line1, next1 - line1,
 300                                         line2, next2 - line2))
 301                                 return -1;
 302                 }
 303
 304                 if (!first)
 305                         return 0;
 306
 307                 while (first->next &&
 308                                 first->next->line1 == first->line1 + 1 &&
 309                                 first->next->line2 == first->line2 + 1)
 310                         first = first->next;
 311
 312                 line1 = first->line1 + 1;
 313                 line2 = first->line2 + 1;
 314
 315                 first = first->next;
 316         }
 317 }
 318
 319 static int fall_back_to_classic_diff(struct hashmap *map,
 320                 int line1, int count1, int line2, int count2)
 321 {
 322         xpparam_t xpp;
 323         xpp.flags = map->xpp->flags & ~XDF_DIFF_ALGORITHM_MASK;
 324
 325         return xdl_fall_back_diff(map->env, &xpp,
 326                                   line1, count1, line2, count2);
 327 }
 328
 329 /*
 330  * Recursively find the longest common sequence of unique lines,
 331  * and if none was found, ask xdl_do_diff() to do the job.
 332  *
 333  * This function assumes that env was prepared with xdl_prepare_env().
 334  */
 335 static int patience_diff(mmfile_t *file1, mmfile_t *file2,
 336                 xpparam_t const *xpp, xdfenv_t *env,
 337                 int line1, int count1, int line2, int count2)
 338 {
 339         struct hashmap map;
 340         struct entry *first;
 341         int result = 0;
 342
 343         /* trivial case: one side is empty */
 344         if (!count1) {
 345                 while(count2--)
 346                         env->xdf2.rchg[line2++ - 1] = 1;
 347                 return 0;
 348         } else if (!count2) {
 349                 while(count1--)
 350                         env->xdf1.rchg[line1++ - 1] = 1;
 351                 return 0;
 352         }
 353
 354         memset(&map, 0, sizeof(map));
 355         if (fill_hashmap(file1, file2, xpp, env, &map,
 356                         line1, count1, line2, count2))
 357                 return -1;
 358
 359         /* are there any matching lines at all? */
 360         if (!map.has_matches) {
 361                 while(count1--)
 362                         env->xdf1.rchg[line1++ - 1] = 1;
 363                 while(count2--)
 364                         env->xdf2.rchg[line2++ - 1] = 1;
 365                 xdl_free(map.entries);
 366                 return 0;
 367         }
 368
 369         first = find_longest_common_sequence(&map);
 370         if (first)
 371                 result = walk_common_sequence(&map, first,
 372                         line1, count1, line2, count2);
 373         else
 374                 result = fall_back_to_classic_diff(&map,
 375                         line1, count1, line2, count2);
 376
 377         xdl_free(map.entries);
 378         return result;
 379 }
 380
 381 int xdl_do_patience_diff(mmfile_t *file1, mmfile_t *file2,
 382                 xpparam_t const *xpp, xdfenv_t *env)
 383 {
 384         if (xdl_prepare_env(file1, file2, xpp, env) < 0)
 385                 return -1;
 386
 387         /* environment is cleaned up in xdl_diff() */
 388         return patience_diff(file1, file2, xpp, env,
 389                         1, env->xdf1.nrec, 1, env->xdf2.nrec);
 390 }