• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
3  * All rights reserved.
4  *
5  * This source code is licensed under both the BSD-style license (found in the
6  * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7  * in the COPYING file in the root directory of this source tree).
8  * You may select, at your option, one of the above-listed licenses.
9  */
10 
11 /* This match finder leverages techniques used in file comparison algorithms
12  * to find matches between a dictionary and a source file.
13  *
14  * The original motivation for studying this approach was to try and optimize
15  * Zstandard for the use case of patching: the most common scenario being
16  * updating an existing software package with the next version. When patching,
17  * the difference between the old version of the package and the new version
18  * is generally tiny (most of the new file will be identical to
19  * the old one). In more technical terms, the edit distance (the minimal number
20  * of changes required to take one sequence of bytes to another) between the
21  * files would be small relative to the size of the file.
22  *
23  * Various 'diffing' algorithms utilize this notion of edit distance and
24  * the corrensponding concept of a minimal edit script between two
25  * sequences to identify the regions within two files where they differ.
26  * The core algorithm used in this match finder is described in:
27  *
28  * "An O(ND) Difference Algorithm and its Variations", Eugene W. Myers,
29  *    Algorithmica Vol. 1, 1986, pp. 251-266,
30  *    <https://doi.org/10.1007/BF01840446>.
31  *
32  * Additional algorithmic heuristics for speed improvement have also been included.
33  * These we inspired from implementations of various regular and binary diffing
34  * algorithms such as GNU diff, bsdiff, and Xdelta.
35  *
36  * Note: after some experimentation, this approach proved to not provide enough
37  * utility to justify the additional CPU used in finding matches. The one area
38  * where this approach consistenly outperforms Zstandard even on level 19 is
39  * when compressing small files (<10 KB) using a equally small dictionary that
40  * is very similar to the source file. For the use case that this was intended,
41  * (large similar files) this approach by itself took 5-10X longer than zstd-19 and
42  * generally resulted in 2-3X larger files. The core advantage that zstd-19 has
43  * over this appraoch for match finding is the overlapping matches. This approach
44  * cannot find any.
45  *
46  * I'm leaving this in the contrib section in case this ever becomes interesting
47  * to explore again.
48  * */
49 
50 #ifndef ZSTD_EDIST_H
51 #define ZSTD_EDIST_H
52 
53 /*-*************************************
54 *  Dependencies
55 ***************************************/
56 
57 #include <stddef.h>
58 #include "zstd_internal.h" /* ZSTD_Sequence */
59 
60 /*! ZSTD_eDist_genSequences() :
61  * Will populate the provided ZSTD_Sequence buffer with sequences
62  * based on the optimal or near-optimal (depending on 'useHeuristics')
63  * edit script between 'dict' and 'src.'
64  * @return : the number of sequences found */
65 size_t ZSTD_eDist_genSequences(ZSTD_Sequence* sequences,
66                         const void* dict, size_t dictSize,
67                         const void* src, size_t srcSize,
68                         int useHeuristics);
69 
70 #endif
71