| 1 |
/* |
| 2 |
* Copyright (c) Meta Platforms, Inc. and affiliates. |
| 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 corresponding 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 consistently outperforms Zstandard even on level 19 is |
| 39 |
* when compressing small files (<10 KB) using an 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 approach 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 |