| 1 |
//! Streaming compression functionality. |
| 2 |
|
| 3 |
use alloc::boxed::Box; |
| 4 |
use core::convert::TryInto; |
| 5 |
use core::{cmp, mem}; |
| 6 |
|
| 7 |
use super::super::*; |
| 8 |
use super::deflate_flags::*; |
| 9 |
use super::CompressionLevel; |
| 10 |
use crate::deflate::buffer::{ |
| 11 |
update_hash, HashBuffers, LocalBuf, LZ_CODE_BUF_MASK, LZ_CODE_BUF_SIZE, LZ_DICT_FULL_SIZE, |
| 12 |
LZ_HASH_BITS, LZ_HASH_SHIFT, LZ_HASH_SIZE, OUT_BUF_SIZE, |
| 13 |
}; |
| 14 |
use crate::deflate::stored::compress_stored; |
| 15 |
use crate::deflate::zlib; |
| 16 |
use crate::shared::{update_adler32, HUFFMAN_LENGTH_ORDER, MZ_ADLER32_INIT}; |
| 17 |
use crate::DataFormat; |
| 18 |
|
| 19 |
// Currently not bubbled up outside this module, so can fill in with more |
| 20 |
// context eventually if needed. |
| 21 |
type Result<T, E = Error> = core::result::Result<T, E>; |
| 22 |
pub(crate) struct Error {} |
| 23 |
|
| 24 |
pub(crate) const MAX_PROBES_MASK: u32 = 0xFFF; |
| 25 |
|
| 26 |
const MAX_SUPPORTED_HUFF_CODESIZE: usize = 15; |
| 27 |
|
| 28 |
// Length code for length values - 256. |
| 29 |
// We use an offset to help with bound check avoidance as we can mask values to 32 |
| 30 |
// and it also saves some memory as we can use a u8 instead of a u16. |
| 31 |
// Conventiently our table is large enough that we can get away with using an |
| 32 |
// offset of 256 which results in very efficient code. |
| 33 |
const LEN_SYM: [u8; 256] = [ |
| 34 |
1, 2, 3, 4, 5, 6, 7, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 13, 13, 14, 14, 14, 14, 15, 15, |
| 35 |
15, 15, 16, 16, 16, 16, 17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 19, 19, |
| 36 |
19, 19, 19, 19, 19, 19, 20, 20, 20, 20, 20, 20, 20, 20, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, |
| 37 |
21, 21, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, |
| 38 |
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, |
| 39 |
24, 24, 24, 24, 24, 24, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, |
| 40 |
25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, |
| 41 |
26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 27, 27, |
| 42 |
27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, |
| 43 |
27, 27, 27, 27, 27, 27, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, |
| 44 |
28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 29, |
| 45 |
]; |
| 46 |
|
| 47 |
const LEN_SYM_OFFSET: usize = 256; |
| 48 |
|
| 49 |
/// Number of extra bits for length values. |
| 50 |
#[rustfmt::skip] |
| 51 |
const LEN_EXTRA: [u8; 256] = [ |
| 52 |
0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, |
| 53 |
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, |
| 54 |
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, |
| 55 |
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, |
| 56 |
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, |
| 57 |
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, |
| 58 |
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, |
| 59 |
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, |
| 60 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 61 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 62 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 63 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 64 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 65 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 66 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 67 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 0 |
| 68 |
]; |
| 69 |
|
| 70 |
/// Distance codes for distances smaller than 512. |
| 71 |
#[rustfmt::skip] |
| 72 |
const SMALL_DIST_SYM: [u8; 512] = [ |
| 73 |
0, 1, 2, 3, 4, 4, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, |
| 74 |
8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, |
| 75 |
10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, |
| 76 |
11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, |
| 77 |
12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, |
| 78 |
12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, |
| 79 |
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, |
| 80 |
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, |
| 81 |
14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, |
| 82 |
14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, |
| 83 |
14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, |
| 84 |
14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, |
| 85 |
15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, |
| 86 |
15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, |
| 87 |
15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, |
| 88 |
15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, |
| 89 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 90 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 91 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 92 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 93 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 94 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 95 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 96 |
16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, |
| 97 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 98 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 99 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 100 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 101 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 102 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 103 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, |
| 104 |
17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17 |
| 105 |
]; |
| 106 |
|
| 107 |
/// Number of extra bits for distances smaller than 512. |
| 108 |
#[rustfmt::skip] |
| 109 |
const SMALL_DIST_EXTRA: [u8; 512] = [ |
| 110 |
0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, |
| 111 |
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, |
| 112 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 113 |
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, |
| 114 |
6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, |
| 115 |
6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, |
| 116 |
6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, |
| 117 |
6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, |
| 118 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 119 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 120 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 121 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 122 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 123 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 124 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, |
| 125 |
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7 |
| 126 |
]; |
| 127 |
|
| 128 |
/// Base values to calculate distances above 512. |
| 129 |
#[rustfmt::skip] |
| 130 |
const LARGE_DIST_SYM: [u8; 128] = [ |
| 131 |
0, 0, 18, 19, 20, 20, 21, 21, 22, 22, 22, 22, 23, 23, 23, 23, |
| 132 |
24, 24, 24, 24, 24, 24, 24, 24, 25, 25, 25, 25, 25, 25, 25, 25, |
| 133 |
26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, |
| 134 |
27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, |
| 135 |
28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, |
| 136 |
28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, |
| 137 |
29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, |
| 138 |
29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29 |
| 139 |
]; |
| 140 |
|
| 141 |
/// Number of extra bits distances above 512. |
| 142 |
#[rustfmt::skip] |
| 143 |
const LARGE_DIST_EXTRA: [u8; 128] = [ |
| 144 |
0, 0, 8, 8, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 10, |
| 145 |
11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, |
| 146 |
12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, |
| 147 |
12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, |
| 148 |
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, |
| 149 |
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, |
| 150 |
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, |
| 151 |
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13 |
| 152 |
]; |
| 153 |
|
| 154 |
#[rustfmt::skip] |
| 155 |
const BITMASKS: [u32; 17] = [ |
| 156 |
0x0000, 0x0001, 0x0003, 0x0007, 0x000F, 0x001F, 0x003F, 0x007F, 0x00FF, |
| 157 |
0x01FF, 0x03FF, 0x07FF, 0x0FFF, 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF |
| 158 |
]; |
| 159 |
|
| 160 |
/// The maximum number of checks for matches in the hash table the compressor will make for each |
| 161 |
/// compression level. |
| 162 |
pub(crate) const NUM_PROBES: [u16; 11] = [0, 1, 6, 32, 16, 32, 128, 256, 512, 768, 1500]; |
| 163 |
|
| 164 |
#[derive(Copy, Clone)] |
| 165 |
struct SymFreq { |
| 166 |
key: u16, |
| 167 |
sym_index: u16, |
| 168 |
} |
| 169 |
|
| 170 |
pub mod deflate_flags { |
| 171 |
/// Whether to use a zlib wrapper. |
| 172 |
pub const TDEFL_WRITE_ZLIB_HEADER: u32 = 0x0000_1000; |
| 173 |
/// Should we compute the adler32 checksum. |
| 174 |
pub const TDEFL_COMPUTE_ADLER32: u32 = 0x0000_2000; |
| 175 |
/// Should we use greedy parsing (as opposed to lazy parsing where look ahead one or more |
| 176 |
/// bytes to check for better matches.) |
| 177 |
pub const TDEFL_GREEDY_PARSING_FLAG: u32 = 0x0000_4000; |
| 178 |
/// Used in miniz to skip zero-initializing hash and dict. We don't do this here, so |
| 179 |
/// this flag is ignored. |
| 180 |
pub const TDEFL_NONDETERMINISTIC_PARSING_FLAG: u32 = 0x0000_8000; |
| 181 |
/// Only look for matches with a distance of 0. |
| 182 |
pub const TDEFL_RLE_MATCHES: u32 = 0x0001_0000; |
| 183 |
/// Only use matches that are at least 6 bytes long. |
| 184 |
pub const TDEFL_FILTER_MATCHES: u32 = 0x0002_0000; |
| 185 |
/// Force the compressor to only output static blocks. (Blocks using the default huffman codes |
| 186 |
/// specified in the deflate specification.) |
| 187 |
pub const TDEFL_FORCE_ALL_STATIC_BLOCKS: u32 = 0x0004_0000; |
| 188 |
/// Force the compressor to only output raw/uncompressed blocks. |
| 189 |
pub const TDEFL_FORCE_ALL_RAW_BLOCKS: u32 = 0x0008_0000; |
| 190 |
} |
| 191 |
|
| 192 |
/// Strategy setting for compression. |
| 193 |
/// |
| 194 |
/// The non-default settings offer some special-case compression variants. |
| 195 |
#[repr(i32)] |
| 196 |
#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)] |
| 197 |
pub enum CompressionStrategy { |
| 198 |
/// Don't use any of the special strategies. |
| 199 |
Default = 0, |
| 200 |
/// Only use matches that are at least 5 bytes long. |
| 201 |
Filtered = 1, |
| 202 |
/// Don't look for matches, only huffman encode the literals. |
| 203 |
HuffmanOnly = 2, |
| 204 |
/// Only look for matches with a distance of 1, i.e do run-length encoding only. |
| 205 |
RLE = 3, |
| 206 |
/// Only use static/fixed blocks. (Blocks using the default huffman codes |
| 207 |
/// specified in the deflate specification.) |
| 208 |
Fixed = 4, |
| 209 |
} |
| 210 |
|
| 211 |
impl From<CompressionStrategy> for i32 { |
| 212 |
#[inline(always)] |
| 213 |
fn from(value: CompressionStrategy) -> Self { |
| 214 |
value as i32 |
| 215 |
} |
| 216 |
} |
| 217 |
|
| 218 |
/// A list of deflate flush types. |
| 219 |
#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)] |
| 220 |
pub enum TDEFLFlush { |
| 221 |
/// Normal operation. |
| 222 |
/// |
| 223 |
/// Compress as much as there is space for, and then return waiting for more input. |
| 224 |
None = 0, |
| 225 |
|
| 226 |
/// Try to flush all the current data and output an empty raw block. |
| 227 |
Sync = 2, |
| 228 |
|
| 229 |
/// Same as [`Sync`][Self::Sync], but reset the dictionary so that the following data does not |
| 230 |
/// depend on previous data. |
| 231 |
Full = 3, |
| 232 |
|
| 233 |
/// Try to flush everything and end the deflate stream. |
| 234 |
/// |
| 235 |
/// On success this will yield a [`TDEFLStatus::Done`] return status. |
| 236 |
Finish = 4, |
| 237 |
} |
| 238 |
|
| 239 |
impl From<MZFlush> for TDEFLFlush { |
| 240 |
fn from(flush: MZFlush) -> Self { |
| 241 |
match flush { |
| 242 |
MZFlush::None => TDEFLFlush::None, |
| 243 |
MZFlush::Sync => TDEFLFlush::Sync, |
| 244 |
MZFlush::Full => TDEFLFlush::Full, |
| 245 |
MZFlush::Finish => TDEFLFlush::Finish, |
| 246 |
_ => TDEFLFlush::None, // TODO: ??? What to do ??? |
| 247 |
} |
| 248 |
} |
| 249 |
} |
| 250 |
|
| 251 |
impl TDEFLFlush { |
| 252 |
pub const fn new(flush: i32) -> Result<Self, MZError> { |
| 253 |
match flush { |
| 254 |
0 => Ok(TDEFLFlush::None), |
| 255 |
2 => Ok(TDEFLFlush::Sync), |
| 256 |
3 => Ok(TDEFLFlush::Full), |
| 257 |
4 => Ok(TDEFLFlush::Finish), |
| 258 |
_ => Err(MZError::Param), |
| 259 |
} |
| 260 |
} |
| 261 |
} |
| 262 |
|
| 263 |
/// Return status of compression. |
| 264 |
#[repr(i32)] |
| 265 |
#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)] |
| 266 |
pub enum TDEFLStatus { |
| 267 |
/// Usage error. |
| 268 |
/// |
| 269 |
/// This indicates that either the [`CompressorOxide`] experienced a previous error, or the |
| 270 |
/// stream has already been [`TDEFLFlush::Finish`]'d. |
| 271 |
BadParam = -2, |
| 272 |
|
| 273 |
/// Error putting data into output buffer. |
| 274 |
/// |
| 275 |
/// This usually indicates a too-small buffer. |
| 276 |
PutBufFailed = -1, |
| 277 |
|
| 278 |
/// Compression succeeded normally. |
| 279 |
Okay = 0, |
| 280 |
|
| 281 |
/// Compression succeeded and the deflate stream was ended. |
| 282 |
/// |
| 283 |
/// This is the result of calling compression with [`TDEFLFlush::Finish`]. |
| 284 |
Done = 1, |
| 285 |
} |
| 286 |
|
| 287 |
const MAX_HUFF_SYMBOLS: usize = 288; |
| 288 |
/// Size of hash chain for fast compression mode. |
| 289 |
const LEVEL1_HASH_SIZE_MASK: u32 = 4095; |
| 290 |
/// The number of huffman tables used by the compressor. |
| 291 |
/// Literal/length, Distances and Length of the huffman codes for the other two tables. |
| 292 |
const MAX_HUFF_TABLES: usize = 3; |
| 293 |
/// Literal/length codes |
| 294 |
const MAX_HUFF_SYMBOLS_0: usize = 288; |
| 295 |
/// Distance codes. |
| 296 |
const MAX_HUFF_SYMBOLS_1: usize = 32; |
| 297 |
/// Huffman length values. |
| 298 |
const MAX_HUFF_SYMBOLS_2: usize = 19; |
| 299 |
/// Size of the chained hash table. |
| 300 |
pub(crate) const LZ_DICT_SIZE: usize = 32_768; |
| 301 |
/// Mask used when stepping through the hash chains. |
| 302 |
pub(crate) const LZ_DICT_SIZE_MASK: usize = (LZ_DICT_SIZE as u32 - 1) as usize; |
| 303 |
/// The minimum length of a match. |
| 304 |
pub(crate) const MIN_MATCH_LEN: u8 = 3; |
| 305 |
/// The maximum length of a match. |
| 306 |
pub(crate) const MAX_MATCH_LEN: usize = 258; |
| 307 |
|
| 308 |
pub(crate) const DEFAULT_FLAGS: u32 = NUM_PROBES[4] as u32 | TDEFL_WRITE_ZLIB_HEADER; |
| 309 |
|
| 310 |
#[cfg(test)] |
| 311 |
#[inline] |
| 312 |
fn write_u16_le(val: u16, slice: &mut [u8], pos: usize) { |
| 313 |
slice[pos] = val as u8; |
| 314 |
slice[pos + 1] = (val >> 8) as u8; |
| 315 |
} |
| 316 |
|
| 317 |
// Read the two bytes starting at pos and interpret them as an u16. |
| 318 |
#[inline(always)] |
| 319 |
const fn read_u16_le<const N: usize>(slice: &[u8; N], pos: usize) -> u16 { |
| 320 |
// The compiler is smart enough to optimize this into an unaligned load. |
| 321 |
slice[pos] as u16 | ((slice[pos + 1] as u16) << 8) |
| 322 |
} |
| 323 |
|
| 324 |
/// Main compression struct. |
| 325 |
pub struct CompressorOxide { |
| 326 |
pub(crate) lz: LZOxide, |
| 327 |
pub(crate) params: ParamsOxide, |
| 328 |
/// Put HuffmanOxide on the heap with default trick to avoid |
| 329 |
/// excessive stack copies. |
| 330 |
pub(crate) huff: Box<HuffmanOxide>, |
| 331 |
pub(crate) dict: DictOxide, |
| 332 |
} |
| 333 |
|
| 334 |
impl CompressorOxide { |
| 335 |
/// Create a new `CompressorOxide` with the given flags. |
| 336 |
/// |
| 337 |
/// # Notes |
| 338 |
/// This function may be changed to take different parameters in the future. |
| 339 |
pub fn new(flags: u32) -> Self { |
| 340 |
CompressorOxide { |
| 341 |
lz: LZOxide::new(), |
| 342 |
params: ParamsOxide::new(flags), |
| 343 |
huff: Box::default(), |
| 344 |
dict: DictOxide::new(flags), |
| 345 |
} |
| 346 |
} |
| 347 |
|
| 348 |
/// Get the adler32 checksum of the currently encoded data. |
| 349 |
pub const fn adler32(&self) -> u32 { |
| 350 |
self.params.adler32 |
| 351 |
} |
| 352 |
|
| 353 |
/// Get the return status of the previous [`compress`](fn.compress.html) |
| 354 |
/// call with this compressor. |
| 355 |
pub const fn prev_return_status(&self) -> TDEFLStatus { |
| 356 |
self.params.prev_return_status |
| 357 |
} |
| 358 |
|
| 359 |
/// Get the raw compressor flags. |
| 360 |
/// |
| 361 |
/// # Notes |
| 362 |
/// This function may be deprecated or changed in the future to use more rust-style flags. |
| 363 |
pub const fn flags(&self) -> i32 { |
| 364 |
self.params.flags as i32 |
| 365 |
} |
| 366 |
|
| 367 |
/// Returns whether the compressor is wrapping the data in a zlib format or not. |
| 368 |
pub const fn data_format(&self) -> DataFormat { |
| 369 |
if (self.params.flags & TDEFL_WRITE_ZLIB_HEADER) != 0 { |
| 370 |
DataFormat::Zlib |
| 371 |
} else { |
| 372 |
DataFormat::Raw |
| 373 |
} |
| 374 |
} |
| 375 |
|
| 376 |
/// Reset the state of the compressor, keeping the same parameters. |
| 377 |
/// |
| 378 |
/// This avoids re-allocating data. |
| 379 |
pub fn reset(&mut self) { |
| 380 |
// LZ buf and huffman has no settings or dynamic memory |
| 381 |
// that needs to be saved, so we simply replace them. |
| 382 |
self.lz = LZOxide::new(); |
| 383 |
self.params.reset(); |
| 384 |
*self.huff = HuffmanOxide::default(); |
| 385 |
self.dict.reset(); |
| 386 |
} |
| 387 |
|
| 388 |
/// Set the compression level of the compressor. |
| 389 |
/// |
| 390 |
/// Using this to change level after compression has started is supported. |
| 391 |
/// # Notes |
| 392 |
/// The compression strategy will be reset to the default one when this is called. |
| 393 |
pub fn set_compression_level(&mut self, level: CompressionLevel) { |
| 394 |
let format = self.data_format(); |
| 395 |
self.set_format_and_level(format, level as u8); |
| 396 |
} |
| 397 |
|
| 398 |
/// Set the compression level of the compressor using an integer value. |
| 399 |
/// |
| 400 |
/// Using this to change level after compression has started is supported. |
| 401 |
/// # Notes |
| 402 |
/// The compression strategy will be reset to the default one when this is called. |
| 403 |
pub fn set_compression_level_raw(&mut self, level: u8) { |
| 404 |
let format = self.data_format(); |
| 405 |
self.set_format_and_level(format, level); |
| 406 |
} |
| 407 |
|
| 408 |
/// Update the compression settings of the compressor. |
| 409 |
/// |
| 410 |
/// Changing the `DataFormat` after compression has started will result in |
| 411 |
/// a corrupted stream. |
| 412 |
/// |
| 413 |
/// # Notes |
| 414 |
/// This function mainly intended for setting the initial settings after e.g creating with |
| 415 |
/// `default` or after calling `CompressorOxide::reset()`, and behaviour may be changed |
| 416 |
/// to disallow calling it after starting compression in the future. |
| 417 |
pub fn set_format_and_level(&mut self, data_format: DataFormat, level: u8) { |
| 418 |
let flags = create_comp_flags_from_zip_params( |
| 419 |
level.into(), |
| 420 |
data_format.to_window_bits(), |
| 421 |
CompressionStrategy::Default as i32, |
| 422 |
); |
| 423 |
self.params.update_flags(flags); |
| 424 |
self.dict.update_flags(flags); |
| 425 |
} |
| 426 |
} |
| 427 |
|
| 428 |
impl Default for CompressorOxide { |
| 429 |
/// Initialize the compressor with a level of 4, zlib wrapper and |
| 430 |
/// the default strategy. |
| 431 |
fn default() -> Self { |
| 432 |
CompressorOxide { |
| 433 |
lz: LZOxide::new(), |
| 434 |
params: ParamsOxide::new(DEFAULT_FLAGS), |
| 435 |
huff: Box::default(), |
| 436 |
dict: DictOxide::new(DEFAULT_FLAGS), |
| 437 |
} |
| 438 |
} |
| 439 |
} |
| 440 |
|
| 441 |
/// Callback function and user used in `compress_to_output`. |
| 442 |
pub struct CallbackFunc<'a> { |
| 443 |
pub put_buf_func: &'a mut dyn FnMut(&[u8]) -> bool, |
| 444 |
} |
| 445 |
|
| 446 |
impl CallbackFunc<'_> { |
| 447 |
fn flush_output( |
| 448 |
&mut self, |
| 449 |
saved_output: SavedOutputBufferOxide, |
| 450 |
params: &mut ParamsOxide, |
| 451 |
) -> i32 { |
| 452 |
// TODO: As this could be unsafe since |
| 453 |
// we can't verify the function pointer |
| 454 |
// this whole function should maybe be unsafe as well. |
| 455 |
let call_success = (self.put_buf_func)(¶ms.local_buf.b[0..saved_output.pos]); |
| 456 |
|
| 457 |
if !call_success { |
| 458 |
params.prev_return_status = TDEFLStatus::PutBufFailed; |
| 459 |
return params.prev_return_status as i32; |
| 460 |
} |
| 461 |
|
| 462 |
params.flush_remaining as i32 |
| 463 |
} |
| 464 |
} |
| 465 |
|
| 466 |
struct CallbackBuf<'a> { |
| 467 |
pub out_buf: &'a mut [u8], |
| 468 |
} |
| 469 |
|
| 470 |
impl CallbackBuf<'_> { |
| 471 |
fn flush_output( |
| 472 |
&mut self, |
| 473 |
saved_output: SavedOutputBufferOxide, |
| 474 |
params: &mut ParamsOxide, |
| 475 |
) -> i32 { |
| 476 |
if saved_output.local { |
| 477 |
let n = cmp::min(saved_output.pos, self.out_buf.len() - params.out_buf_ofs); |
| 478 |
(self.out_buf[params.out_buf_ofs..params.out_buf_ofs + n]) |
| 479 |
.copy_from_slice(¶ms.local_buf.b[..n]); |
| 480 |
|
| 481 |
params.out_buf_ofs += n; |
| 482 |
if saved_output.pos != n { |
| 483 |
params.flush_ofs = n as u32; |
| 484 |
params.flush_remaining = (saved_output.pos - n) as u32; |
| 485 |
} |
| 486 |
} else { |
| 487 |
params.out_buf_ofs += saved_output.pos; |
| 488 |
} |
| 489 |
|
| 490 |
params.flush_remaining as i32 |
| 491 |
} |
| 492 |
} |
| 493 |
|
| 494 |
enum CallbackOut<'a> { |
| 495 |
Func(CallbackFunc<'a>), |
| 496 |
Buf(CallbackBuf<'a>), |
| 497 |
} |
| 498 |
|
| 499 |
impl CallbackOut<'_> { |
| 500 |
fn new_output_buffer<'b>( |
| 501 |
&'b mut self, |
| 502 |
local_buf: &'b mut [u8], |
| 503 |
out_buf_ofs: usize, |
| 504 |
) -> OutputBufferOxide<'b> { |
| 505 |
let is_local; |
| 506 |
let buf_len = OUT_BUF_SIZE - 16; |
| 507 |
let chosen_buffer = match *self { |
| 508 |
CallbackOut::Buf(ref mut cb) if cb.out_buf.len() - out_buf_ofs >= OUT_BUF_SIZE => { |
| 509 |
is_local = false; |
| 510 |
&mut cb.out_buf[out_buf_ofs..out_buf_ofs + buf_len] |
| 511 |
} |
| 512 |
_ => { |
| 513 |
is_local = true; |
| 514 |
&mut local_buf[..buf_len] |
| 515 |
} |
| 516 |
}; |
| 517 |
|
| 518 |
OutputBufferOxide { |
| 519 |
inner: chosen_buffer, |
| 520 |
inner_pos: 0, |
| 521 |
local: is_local, |
| 522 |
bit_buffer: 0, |
| 523 |
bits_in: 0, |
| 524 |
} |
| 525 |
} |
| 526 |
} |
| 527 |
|
| 528 |
pub(crate) struct CallbackOxide<'a> { |
| 529 |
in_buf: Option<&'a [u8]>, |
| 530 |
in_buf_size: Option<&'a mut usize>, |
| 531 |
out_buf_size: Option<&'a mut usize>, |
| 532 |
out: CallbackOut<'a>, |
| 533 |
} |
| 534 |
|
| 535 |
impl<'a> CallbackOxide<'a> { |
| 536 |
fn new_callback_buf(in_buf: &'a [u8], out_buf: &'a mut [u8]) -> Self { |
| 537 |
CallbackOxide { |
| 538 |
in_buf: Some(in_buf), |
| 539 |
in_buf_size: None, |
| 540 |
out_buf_size: None, |
| 541 |
out: CallbackOut::Buf(CallbackBuf { out_buf }), |
| 542 |
} |
| 543 |
} |
| 544 |
|
| 545 |
fn new_callback_func(in_buf: &'a [u8], callback_func: CallbackFunc<'a>) -> Self { |
| 546 |
CallbackOxide { |
| 547 |
in_buf: Some(in_buf), |
| 548 |
in_buf_size: None, |
| 549 |
out_buf_size: None, |
| 550 |
out: CallbackOut::Func(callback_func), |
| 551 |
} |
| 552 |
} |
| 553 |
|
| 554 |
fn update_size(&mut self, in_size: Option<usize>, out_size: Option<usize>) { |
| 555 |
if let (Some(in_size), Some(size)) = (in_size, self.in_buf_size.as_mut()) { |
| 556 |
**size = in_size; |
| 557 |
} |
| 558 |
|
| 559 |
if let (Some(out_size), Some(size)) = (out_size, self.out_buf_size.as_mut()) { |
| 560 |
**size = out_size |
| 561 |
} |
| 562 |
} |
| 563 |
|
| 564 |
fn flush_output( |
| 565 |
&mut self, |
| 566 |
saved_output: SavedOutputBufferOxide, |
| 567 |
params: &mut ParamsOxide, |
| 568 |
) -> i32 { |
| 569 |
if saved_output.pos == 0 { |
| 570 |
return params.flush_remaining as i32; |
| 571 |
} |
| 572 |
|
| 573 |
self.update_size(Some(params.src_pos), None); |
| 574 |
match self.out { |
| 575 |
CallbackOut::Func(ref mut cf) => cf.flush_output(saved_output, params), |
| 576 |
CallbackOut::Buf(ref mut cb) => cb.flush_output(saved_output, params), |
| 577 |
} |
| 578 |
} |
| 579 |
|
| 580 |
pub(crate) fn buf(&mut self) -> Option<&'a [u8]> { |
| 581 |
self.in_buf |
| 582 |
} |
| 583 |
} |
| 584 |
|
| 585 |
struct OutputBufferOxide<'a> { |
| 586 |
pub inner: &'a mut [u8], |
| 587 |
pub inner_pos: usize, |
| 588 |
pub local: bool, |
| 589 |
|
| 590 |
pub bit_buffer: u32, |
| 591 |
pub bits_in: u32, |
| 592 |
} |
| 593 |
|
| 594 |
impl OutputBufferOxide<'_> { |
| 595 |
/// Write bits to the bit buffer and flushes |
| 596 |
/// the bit buffer so any whole bytes are output |
| 597 |
/// to the underlying buffer. |
| 598 |
fn put_bits(&mut self, bits: u32, len: u32) { |
| 599 |
// TODO: Removing this assertion worsens performance |
| 600 |
// Need to figure out why |
| 601 |
assert!(bits <= ((1u32 << len) - 1u32)); |
| 602 |
self.bit_buffer |= bits << self.bits_in; |
| 603 |
self.bits_in += len; |
| 604 |
|
| 605 |
while self.bits_in >= 8 { |
| 606 |
self.inner[self.inner_pos] = self.bit_buffer as u8; |
| 607 |
self.inner_pos += 1; |
| 608 |
self.bit_buffer >>= 8; |
| 609 |
self.bits_in -= 8; |
| 610 |
} |
| 611 |
} |
| 612 |
|
| 613 |
#[inline] |
| 614 |
/// Write the provided bits to the bit buffer without flushing |
| 615 |
/// anything. Does not check if there is actually space for it. |
| 616 |
fn put_bits_no_flush(&mut self, bits: u32, len: u32) { |
| 617 |
self.bit_buffer |= bits << self.bits_in; |
| 618 |
self.bits_in += len; |
| 619 |
} |
| 620 |
|
| 621 |
const fn save(&self) -> SavedOutputBufferOxide { |
| 622 |
SavedOutputBufferOxide { |
| 623 |
pos: self.inner_pos, |
| 624 |
bit_buffer: self.bit_buffer, |
| 625 |
bits_in: self.bits_in, |
| 626 |
local: self.local, |
| 627 |
} |
| 628 |
} |
| 629 |
|
| 630 |
fn load(&mut self, saved: SavedOutputBufferOxide) { |
| 631 |
self.inner_pos = saved.pos; |
| 632 |
self.bit_buffer = saved.bit_buffer; |
| 633 |
self.bits_in = saved.bits_in; |
| 634 |
self.local = saved.local; |
| 635 |
} |
| 636 |
|
| 637 |
#[inline] |
| 638 |
/// Pad the bit buffer to a whole byte with |
| 639 |
/// zeroes and write that byte to the output buffer. |
| 640 |
fn pad_to_bytes(&mut self) { |
| 641 |
if self.bits_in != 0 { |
| 642 |
let len = 8 - self.bits_in; |
| 643 |
self.put_bits(0, len); |
| 644 |
} |
| 645 |
} |
| 646 |
|
| 647 |
#[inline] |
| 648 |
fn write_bytes(&mut self, bytes: &[u8]) { |
| 649 |
debug_assert_eq!(self.bits_in, 0); |
| 650 |
self.inner[self.inner_pos..self.inner_pos + bytes.len()].copy_from_slice(bytes); |
| 651 |
self.inner_pos += bytes.len(); |
| 652 |
} |
| 653 |
} |
| 654 |
|
| 655 |
struct SavedOutputBufferOxide { |
| 656 |
pub pos: usize, |
| 657 |
pub bit_buffer: u32, |
| 658 |
pub bits_in: u32, |
| 659 |
pub local: bool, |
| 660 |
} |
| 661 |
|
| 662 |
struct BitBuffer { |
| 663 |
pub bit_buffer: u64, |
| 664 |
pub bits_in: u32, |
| 665 |
} |
| 666 |
|
| 667 |
impl BitBuffer { |
| 668 |
fn put_fast(&mut self, bits: u64, len: u32) { |
| 669 |
self.bit_buffer |= bits << self.bits_in; |
| 670 |
self.bits_in += len; |
| 671 |
} |
| 672 |
|
| 673 |
fn flush(&mut self, output: &mut OutputBufferOxide) -> Result<()> { |
| 674 |
let pos = output.inner_pos; |
| 675 |
{ |
| 676 |
// isolation to please borrow checker |
| 677 |
let inner = &mut output.inner[pos..pos + 8]; |
| 678 |
let bytes = u64::to_le_bytes(self.bit_buffer); |
| 679 |
inner.copy_from_slice(&bytes); |
| 680 |
} |
| 681 |
match output.inner_pos.checked_add((self.bits_in >> 3) as usize) { |
| 682 |
Some(n) if n <= output.inner.len() => output.inner_pos = n, |
| 683 |
_ => return Err(Error {}), |
| 684 |
} |
| 685 |
self.bit_buffer >>= self.bits_in & !7; |
| 686 |
self.bits_in &= 7; |
| 687 |
Ok(()) |
| 688 |
} |
| 689 |
} |
| 690 |
|
| 691 |
/// A struct containing data about huffman codes and symbol frequencies. |
| 692 |
/// |
| 693 |
/// NOTE: Only the literal/lengths have enough symbols to actually use |
| 694 |
/// the full array. It's unclear why it's defined like this in miniz, |
| 695 |
/// it could be for cache/alignment reasons. |
| 696 |
pub(crate) struct HuffmanOxide { |
| 697 |
/// Number of occurrences of each symbol. |
| 698 |
pub count: [[u16; MAX_HUFF_SYMBOLS]; MAX_HUFF_TABLES], |
| 699 |
/// The bits of the huffman code assigned to the symbol |
| 700 |
pub codes: [[u16; MAX_HUFF_SYMBOLS]; MAX_HUFF_TABLES], |
| 701 |
/// The length of the huffman code assigned to the symbol. |
| 702 |
pub code_sizes: [[u8; MAX_HUFF_SYMBOLS]; MAX_HUFF_TABLES], |
| 703 |
} |
| 704 |
|
| 705 |
/// Tables used for literal/lengths in `HuffmanOxide`. |
| 706 |
const LITLEN_TABLE: usize = 0; |
| 707 |
/// Tables for distances. |
| 708 |
const DIST_TABLE: usize = 1; |
| 709 |
/// Tables for the run-length encoded huffman lengths for literals/lengths/distances. |
| 710 |
const HUFF_CODES_TABLE: usize = 2; |
| 711 |
|
| 712 |
/// Status of RLE encoding of huffman code lengths. |
| 713 |
struct Rle { |
| 714 |
pub z_count: u32, |
| 715 |
pub repeat_count: u16, |
| 716 |
pub prev_code_size: u8, |
| 717 |
} |
| 718 |
|
| 719 |
impl Rle { |
| 720 |
fn prev_code_size( |
| 721 |
&mut self, |
| 722 |
packed_code_sizes: &mut [u8], |
| 723 |
packed_pos: &mut usize, |
| 724 |
h: &mut HuffmanOxide, |
| 725 |
) -> Result<()> { |
| 726 |
let mut write = |buf| write(buf, packed_code_sizes, packed_pos); |
| 727 |
let counts = &mut h.count[HUFF_CODES_TABLE]; |
| 728 |
if self.repeat_count != 0 { |
| 729 |
if self.repeat_count < 3 { |
| 730 |
counts[self.prev_code_size as usize] = |
| 731 |
counts[self.prev_code_size as usize].wrapping_add(self.repeat_count); |
| 732 |
let code = self.prev_code_size; |
| 733 |
write(&[code, code, code][..self.repeat_count as usize])?; |
| 734 |
} else { |
| 735 |
counts[16] = counts[16].wrapping_add(1); |
| 736 |
write(&[16, (self.repeat_count - 3) as u8][..])?; |
| 737 |
} |
| 738 |
self.repeat_count = 0; |
| 739 |
} |
| 740 |
|
| 741 |
Ok(()) |
| 742 |
} |
| 743 |
|
| 744 |
fn zero_code_size( |
| 745 |
&mut self, |
| 746 |
packed_code_sizes: &mut [u8], |
| 747 |
packed_pos: &mut usize, |
| 748 |
h: &mut HuffmanOxide, |
| 749 |
) -> Result<()> { |
| 750 |
let mut write = |buf| write(buf, packed_code_sizes, packed_pos); |
| 751 |
let counts = &mut h.count[HUFF_CODES_TABLE]; |
| 752 |
if self.z_count != 0 { |
| 753 |
if self.z_count < 3 { |
| 754 |
counts[0] = counts[0].wrapping_add(self.z_count as u16); |
| 755 |
write(&[0, 0, 0][..self.z_count as usize])?; |
| 756 |
} else if self.z_count <= 10 { |
| 757 |
counts[17] = counts[17].wrapping_add(1); |
| 758 |
write(&[17, (self.z_count - 3) as u8][..])?; |
| 759 |
} else { |
| 760 |
counts[18] = counts[18].wrapping_add(1); |
| 761 |
write(&[18, (self.z_count - 11) as u8][..])?; |
| 762 |
} |
| 763 |
self.z_count = 0; |
| 764 |
} |
| 765 |
|
| 766 |
Ok(()) |
| 767 |
} |
| 768 |
} |
| 769 |
|
| 770 |
fn write(src: &[u8], dst: &mut [u8], dst_pos: &mut usize) -> Result<()> { |
| 771 |
match dst.get_mut(*dst_pos..*dst_pos + src.len()) { |
| 772 |
Some(s) => s.copy_from_slice(src), |
| 773 |
None => return Err(Error {}), |
| 774 |
} |
| 775 |
*dst_pos += src.len(); |
| 776 |
Ok(()) |
| 777 |
} |
| 778 |
|
| 779 |
impl Default for HuffmanOxide { |
| 780 |
fn default() -> Self { |
| 781 |
HuffmanOxide { |
| 782 |
count: [[0; MAX_HUFF_SYMBOLS]; MAX_HUFF_TABLES], |
| 783 |
codes: [[0; MAX_HUFF_SYMBOLS]; MAX_HUFF_TABLES], |
| 784 |
code_sizes: [[0; MAX_HUFF_SYMBOLS]; MAX_HUFF_TABLES], |
| 785 |
} |
| 786 |
} |
| 787 |
} |
| 788 |
|
| 789 |
impl HuffmanOxide { |
| 790 |
fn radix_sort_symbols<'a>( |
| 791 |
symbols0: &'a mut [SymFreq], |
| 792 |
symbols1: &'a mut [SymFreq], |
| 793 |
) -> &'a mut [SymFreq] { |
| 794 |
let mut hist = [[0; 256]; 2]; |
| 795 |
|
| 796 |
for freq in symbols0.iter() { |
| 797 |
hist[0][(freq.key & 0xFF) as usize] += 1; |
| 798 |
hist[1][((freq.key >> 8) & 0xFF) as usize] += 1; |
| 799 |
} |
| 800 |
|
| 801 |
let mut n_passes = 2; |
| 802 |
if symbols0.len() == hist[1][0] { |
| 803 |
n_passes -= 1; |
| 804 |
} |
| 805 |
|
| 806 |
let mut current_symbols = symbols0; |
| 807 |
let mut new_symbols = symbols1; |
| 808 |
|
| 809 |
for (pass, hist_item) in hist.iter().enumerate().take(n_passes) { |
| 810 |
let mut offsets = [0; 256]; |
| 811 |
let mut offset = 0; |
| 812 |
for i in 0..256 { |
| 813 |
offsets[i] = offset; |
| 814 |
offset += hist_item[i]; |
| 815 |
} |
| 816 |
|
| 817 |
for sym in current_symbols.iter() { |
| 818 |
let j = ((sym.key >> (pass * 8)) & 0xFF) as usize; |
| 819 |
new_symbols[offsets[j]] = *sym; |
| 820 |
offsets[j] += 1; |
| 821 |
} |
| 822 |
|
| 823 |
mem::swap(&mut current_symbols, &mut new_symbols); |
| 824 |
} |
| 825 |
|
| 826 |
current_symbols |
| 827 |
} |
| 828 |
|
| 829 |
fn calculate_minimum_redundancy(symbols: &mut [SymFreq]) { |
| 830 |
match symbols.len() { |
| 831 |
0 => (), |
| 832 |
1 => symbols[0].key = 1, |
| 833 |
n => { |
| 834 |
symbols[0].key += symbols[1].key; |
| 835 |
let mut root = 0; |
| 836 |
let mut leaf = 2; |
| 837 |
for next in 1..n - 1 { |
| 838 |
if (leaf >= n) || (symbols[root].key < symbols[leaf].key) { |
| 839 |
symbols[next].key = symbols[root].key; |
| 840 |
symbols[root].key = next as u16; |
| 841 |
root += 1; |
| 842 |
} else { |
| 843 |
symbols[next].key = symbols[leaf].key; |
| 844 |
leaf += 1; |
| 845 |
} |
| 846 |
|
| 847 |
if (leaf >= n) || (root < next && symbols[root].key < symbols[leaf].key) { |
| 848 |
symbols[next].key = symbols[next].key.wrapping_add(symbols[root].key); |
| 849 |
symbols[root].key = next as u16; |
| 850 |
root += 1; |
| 851 |
} else { |
| 852 |
symbols[next].key = symbols[next].key.wrapping_add(symbols[leaf].key); |
| 853 |
leaf += 1; |
| 854 |
} |
| 855 |
} |
| 856 |
|
| 857 |
symbols[n - 2].key = 0; |
| 858 |
for next in (0..n - 2).rev() { |
| 859 |
symbols[next].key = symbols[symbols[next].key as usize].key + 1; |
| 860 |
} |
| 861 |
|
| 862 |
let mut avbl = 1; |
| 863 |
let mut used = 0; |
| 864 |
let mut dpth = 0; |
| 865 |
let mut root = (n - 2) as i32; |
| 866 |
let mut next = (n - 1) as i32; |
| 867 |
while avbl > 0 { |
| 868 |
while (root >= 0) && (symbols[root as usize].key == dpth) { |
| 869 |
used += 1; |
| 870 |
root -= 1; |
| 871 |
} |
| 872 |
while avbl > used { |
| 873 |
symbols[next as usize].key = dpth; |
| 874 |
next -= 1; |
| 875 |
avbl -= 1; |
| 876 |
} |
| 877 |
avbl = 2 * used; |
| 878 |
dpth += 1; |
| 879 |
used = 0; |
| 880 |
} |
| 881 |
} |
| 882 |
} |
| 883 |
} |
| 884 |
|
| 885 |
fn enforce_max_code_size(num_codes: &mut [i32], code_list_len: usize, max_code_size: usize) { |
| 886 |
if code_list_len <= 1 { |
| 887 |
return; |
| 888 |
} |
| 889 |
|
| 890 |
num_codes[max_code_size] += num_codes[max_code_size + 1..].iter().sum::<i32>(); |
| 891 |
let total = num_codes[1..=max_code_size] |
| 892 |
.iter() |
| 893 |
.rev() |
| 894 |
.enumerate() |
| 895 |
.fold(0u32, |total, (i, &x)| total + ((x as u32) << i)); |
| 896 |
|
| 897 |
for _ in (1 << max_code_size)..total { |
| 898 |
num_codes[max_code_size] -= 1; |
| 899 |
for i in (1..max_code_size).rev() { |
| 900 |
if num_codes[i] != 0 { |
| 901 |
num_codes[i] -= 1; |
| 902 |
num_codes[i + 1] += 2; |
| 903 |
break; |
| 904 |
} |
| 905 |
} |
| 906 |
} |
| 907 |
} |
| 908 |
|
| 909 |
fn optimize_table( |
| 910 |
&mut self, |
| 911 |
table_num: usize, |
| 912 |
table_len: usize, |
| 913 |
code_size_limit: usize, |
| 914 |
static_table: bool, |
| 915 |
) { |
| 916 |
let mut num_codes = [0i32; 32 + 1]; |
| 917 |
let mut next_code = [0u32; MAX_SUPPORTED_HUFF_CODESIZE + 1]; |
| 918 |
|
| 919 |
if static_table { |
| 920 |
for &code_size in &self.code_sizes[table_num][..table_len] { |
| 921 |
num_codes[code_size as usize] += 1; |
| 922 |
} |
| 923 |
} else { |
| 924 |
let mut symbols0 = [SymFreq { |
| 925 |
key: 0, |
| 926 |
sym_index: 0, |
| 927 |
}; MAX_HUFF_SYMBOLS]; |
| 928 |
let mut symbols1 = [SymFreq { |
| 929 |
key: 0, |
| 930 |
sym_index: 0, |
| 931 |
}; MAX_HUFF_SYMBOLS]; |
| 932 |
|
| 933 |
let mut num_used_symbols = 0; |
| 934 |
for i in 0..table_len { |
| 935 |
if self.count[table_num][i] != 0 { |
| 936 |
symbols0[num_used_symbols] = SymFreq { |
| 937 |
key: self.count[table_num][i], |
| 938 |
sym_index: i as u16, |
| 939 |
}; |
| 940 |
num_used_symbols += 1; |
| 941 |
} |
| 942 |
} |
| 943 |
|
| 944 |
let symbols = Self::radix_sort_symbols( |
| 945 |
&mut symbols0[..num_used_symbols], |
| 946 |
&mut symbols1[..num_used_symbols], |
| 947 |
); |
| 948 |
Self::calculate_minimum_redundancy(symbols); |
| 949 |
|
| 950 |
for symbol in symbols.iter() { |
| 951 |
num_codes[symbol.key as usize] += 1; |
| 952 |
} |
| 953 |
|
| 954 |
Self::enforce_max_code_size(&mut num_codes, num_used_symbols, code_size_limit); |
| 955 |
|
| 956 |
self.code_sizes[table_num].fill(0); |
| 957 |
self.codes[table_num].fill(0); |
| 958 |
|
| 959 |
let mut last = num_used_symbols; |
| 960 |
for (i, &num_item) in num_codes |
| 961 |
.iter() |
| 962 |
.enumerate() |
| 963 |
.take(code_size_limit + 1) |
| 964 |
.skip(1) |
| 965 |
{ |
| 966 |
let first = last - num_item as usize; |
| 967 |
for symbol in &symbols[first..last] { |
| 968 |
self.code_sizes[table_num][symbol.sym_index as usize] = i as u8; |
| 969 |
} |
| 970 |
last = first; |
| 971 |
} |
| 972 |
} |
| 973 |
|
| 974 |
let mut j = 0; |
| 975 |
next_code[1] = 0; |
| 976 |
for i in 2..=code_size_limit { |
| 977 |
j = (j + num_codes[i - 1]) << 1; |
| 978 |
next_code[i] = j as u32; |
| 979 |
} |
| 980 |
|
| 981 |
for (&code_size, huff_code) in self.code_sizes[table_num] |
| 982 |
.iter() |
| 983 |
.take(table_len) |
| 984 |
.zip(self.codes[table_num].iter_mut().take(table_len)) |
| 985 |
{ |
| 986 |
if code_size == 0 { |
| 987 |
continue; |
| 988 |
} |
| 989 |
|
| 990 |
let code = next_code[code_size as usize]; |
| 991 |
|
| 992 |
next_code[code_size as usize] += 1; |
| 993 |
|
| 994 |
let rev_code = (code as u16).reverse_bits() >> (16 - code_size); |
| 995 |
|
| 996 |
*huff_code = rev_code; |
| 997 |
} |
| 998 |
} |
| 999 |
|
| 1000 |
fn start_static_block(&mut self, output: &mut OutputBufferOxide) { |
| 1001 |
self.code_sizes[LITLEN_TABLE][0..144].fill(8); |
| 1002 |
self.code_sizes[LITLEN_TABLE][144..256].fill(9); |
| 1003 |
self.code_sizes[LITLEN_TABLE][256..280].fill(7); |
| 1004 |
self.code_sizes[LITLEN_TABLE][280..288].fill(8); |
| 1005 |
|
| 1006 |
self.code_sizes[DIST_TABLE][..32].fill(5); |
| 1007 |
|
| 1008 |
self.optimize_table(LITLEN_TABLE, 288, 15, true); |
| 1009 |
self.optimize_table(DIST_TABLE, 32, 15, true); |
| 1010 |
|
| 1011 |
output.put_bits(0b01, 2) |
| 1012 |
} |
| 1013 |
|
| 1014 |
fn start_dynamic_block(&mut self, output: &mut OutputBufferOxide) -> Result<()> { |
| 1015 |
// There will always be one, and only one end of block code. |
| 1016 |
self.count[0][256] = 1; |
| 1017 |
|
| 1018 |
self.optimize_table(0, MAX_HUFF_SYMBOLS_0, 15, false); |
| 1019 |
self.optimize_table(1, MAX_HUFF_SYMBOLS_1, 15, false); |
| 1020 |
|
| 1021 |
let num_lit_codes = 286 |
| 1022 |
- &self.code_sizes[0][257..286] |
| 1023 |
.iter() |
| 1024 |
.rev() |
| 1025 |
.take_while(|&x| *x == 0) |
| 1026 |
.count(); |
| 1027 |
|
| 1028 |
let num_dist_codes = 30 |
| 1029 |
- &self.code_sizes[1][1..30] |
| 1030 |
.iter() |
| 1031 |
.rev() |
| 1032 |
.take_while(|&x| *x == 0) |
| 1033 |
.count(); |
| 1034 |
|
| 1035 |
let mut code_sizes_to_pack = [0u8; MAX_HUFF_SYMBOLS_0 + MAX_HUFF_SYMBOLS_1]; |
| 1036 |
let mut packed_code_sizes = [0u8; MAX_HUFF_SYMBOLS_0 + MAX_HUFF_SYMBOLS_1]; |
| 1037 |
|
| 1038 |
let total_code_sizes_to_pack = num_lit_codes + num_dist_codes; |
| 1039 |
|
| 1040 |
code_sizes_to_pack[..num_lit_codes].copy_from_slice(&self.code_sizes[0][..num_lit_codes]); |
| 1041 |
|
| 1042 |
code_sizes_to_pack[num_lit_codes..total_code_sizes_to_pack] |
| 1043 |
.copy_from_slice(&self.code_sizes[1][..num_dist_codes]); |
| 1044 |
|
| 1045 |
let mut rle = Rle { |
| 1046 |
z_count: 0, |
| 1047 |
repeat_count: 0, |
| 1048 |
prev_code_size: 0xFF, |
| 1049 |
}; |
| 1050 |
|
| 1051 |
self.count[HUFF_CODES_TABLE][..MAX_HUFF_SYMBOLS_2].fill(0); |
| 1052 |
|
| 1053 |
let mut packed_pos = 0; |
| 1054 |
for &code_size in &code_sizes_to_pack[..total_code_sizes_to_pack] { |
| 1055 |
if code_size == 0 { |
| 1056 |
rle.prev_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1057 |
rle.z_count += 1; |
| 1058 |
if rle.z_count == 138 { |
| 1059 |
rle.zero_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1060 |
} |
| 1061 |
} else { |
| 1062 |
rle.zero_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1063 |
if code_size != rle.prev_code_size { |
| 1064 |
rle.prev_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1065 |
self.count[HUFF_CODES_TABLE][code_size as usize] = |
| 1066 |
self.count[HUFF_CODES_TABLE][code_size as usize].wrapping_add(1); |
| 1067 |
write(&[code_size], &mut packed_code_sizes, &mut packed_pos)?; |
| 1068 |
} else { |
| 1069 |
rle.repeat_count += 1; |
| 1070 |
if rle.repeat_count == 6 { |
| 1071 |
rle.prev_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1072 |
} |
| 1073 |
} |
| 1074 |
} |
| 1075 |
rle.prev_code_size = code_size; |
| 1076 |
} |
| 1077 |
|
| 1078 |
if rle.repeat_count != 0 { |
| 1079 |
rle.prev_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1080 |
} else { |
| 1081 |
rle.zero_code_size(&mut packed_code_sizes, &mut packed_pos, self)?; |
| 1082 |
} |
| 1083 |
|
| 1084 |
self.optimize_table(2, MAX_HUFF_SYMBOLS_2, 7, false); |
| 1085 |
|
| 1086 |
output.put_bits_no_flush(2, 2); |
| 1087 |
|
| 1088 |
output.put_bits_no_flush((num_lit_codes - 257) as u32, 5); |
| 1089 |
output.put_bits_no_flush((num_dist_codes - 1) as u32, 5); |
| 1090 |
|
| 1091 |
let mut num_bit_lengths = 18 |
| 1092 |
- HUFFMAN_LENGTH_ORDER |
| 1093 |
.iter() |
| 1094 |
.rev() |
| 1095 |
.take_while(|&swizzle| self.code_sizes[HUFF_CODES_TABLE][*swizzle as usize] == 0) |
| 1096 |
.count(); |
| 1097 |
|
| 1098 |
num_bit_lengths = cmp::max(4, num_bit_lengths + 1); |
| 1099 |
output.put_bits(num_bit_lengths as u32 - 4, 4); |
| 1100 |
for &swizzle in &HUFFMAN_LENGTH_ORDER[..num_bit_lengths] { |
| 1101 |
output.put_bits( |
| 1102 |
u32::from(self.code_sizes[HUFF_CODES_TABLE][swizzle as usize]), |
| 1103 |
3, |
| 1104 |
); |
| 1105 |
} |
| 1106 |
|
| 1107 |
let mut packed_code_size_index = 0; |
| 1108 |
while packed_code_size_index < packed_pos { |
| 1109 |
let code = packed_code_sizes[packed_code_size_index] as usize; |
| 1110 |
packed_code_size_index += 1; |
| 1111 |
assert!(code < MAX_HUFF_SYMBOLS_2); |
| 1112 |
output.put_bits( |
| 1113 |
u32::from(self.codes[HUFF_CODES_TABLE][code]), |
| 1114 |
u32::from(self.code_sizes[HUFF_CODES_TABLE][code]), |
| 1115 |
); |
| 1116 |
if code >= 16 { |
| 1117 |
output.put_bits( |
| 1118 |
u32::from(packed_code_sizes[packed_code_size_index]), |
| 1119 |
[2, 3, 7][code - 16], |
| 1120 |
); |
| 1121 |
packed_code_size_index += 1; |
| 1122 |
} |
| 1123 |
} |
| 1124 |
|
| 1125 |
Ok(()) |
| 1126 |
} |
| 1127 |
} |
| 1128 |
|
| 1129 |
pub(crate) struct DictOxide { |
| 1130 |
/// The maximum number of checks in the hash chain, for the initial, |
| 1131 |
/// and the lazy match respectively. |
| 1132 |
pub max_probes: [u32; 2], |
| 1133 |
/// Buffer of input data. |
| 1134 |
/// Padded with 1 byte to simplify matching code in `compress_fast`. |
| 1135 |
pub b: HashBuffers, |
| 1136 |
|
| 1137 |
pub code_buf_dict_pos: usize, |
| 1138 |
pub lookahead_size: usize, |
| 1139 |
pub lookahead_pos: usize, |
| 1140 |
pub size: usize, |
| 1141 |
loop_len: u8, |
| 1142 |
} |
| 1143 |
|
| 1144 |
const fn probes_from_flags(flags: u32) -> [u32; 2] { |
| 1145 |
[ |
| 1146 |
1 + ((flags & 0xFFF) + 2) / 3, |
| 1147 |
1 + (((flags & 0xFFF) >> 2) + 2) / 3, |
| 1148 |
] |
| 1149 |
} |
| 1150 |
|
| 1151 |
impl DictOxide { |
| 1152 |
fn new(flags: u32) -> Self { |
| 1153 |
DictOxide { |
| 1154 |
max_probes: probes_from_flags(flags), |
| 1155 |
b: HashBuffers::default(), |
| 1156 |
code_buf_dict_pos: 0, |
| 1157 |
lookahead_size: 0, |
| 1158 |
lookahead_pos: 0, |
| 1159 |
size: 0, |
| 1160 |
loop_len: 32, |
| 1161 |
} |
| 1162 |
} |
| 1163 |
|
| 1164 |
fn update_flags(&mut self, flags: u32) { |
| 1165 |
self.max_probes = probes_from_flags(flags); |
| 1166 |
} |
| 1167 |
|
| 1168 |
fn reset(&mut self) { |
| 1169 |
self.b.reset(); |
| 1170 |
self.code_buf_dict_pos = 0; |
| 1171 |
self.lookahead_size = 0; |
| 1172 |
self.lookahead_pos = 0; |
| 1173 |
self.size = 0; |
| 1174 |
} |
| 1175 |
|
| 1176 |
/// Do an unaligned read of the data at `pos` in the dictionary and treat it as if it was of |
| 1177 |
/// type T. |
| 1178 |
#[inline] |
| 1179 |
fn read_unaligned_u32(&self, pos: usize) -> u32 { |
| 1180 |
// Masking the value here helps avoid bounds checks. |
| 1181 |
let pos = pos & LZ_DICT_SIZE_MASK; |
| 1182 |
let end = pos + 4; |
| 1183 |
// Somehow this assertion makes things faster. |
| 1184 |
// TODO: as of may 2024 this does not seem to make any difference |
| 1185 |
// so consider removing. |
| 1186 |
assert!(end < LZ_DICT_FULL_SIZE); |
| 1187 |
|
| 1188 |
let bytes: [u8; 4] = self.b.dict[pos..end].try_into().unwrap(); |
| 1189 |
u32::from_le_bytes(bytes) |
| 1190 |
} |
| 1191 |
|
| 1192 |
/// Do an unaligned read of the data at `pos` in the dictionary and treat it as if it was of |
| 1193 |
/// type T. |
| 1194 |
#[inline] |
| 1195 |
fn read_unaligned_u64(&self, pos: usize) -> u64 { |
| 1196 |
// Help evade bounds/panic code check by masking the position value |
| 1197 |
// This provides a small speedup at the cost of an instruction or two instead of |
| 1198 |
// having to use unsafe. |
| 1199 |
let pos = pos & LZ_DICT_SIZE_MASK; |
| 1200 |
let bytes: [u8; 8] = self.b.dict[pos..pos + 8].try_into().unwrap(); |
| 1201 |
u64::from_le_bytes(bytes) |
| 1202 |
} |
| 1203 |
|
| 1204 |
/// Try to find a match for the data at lookahead_pos in the dictionary that is |
| 1205 |
/// longer than `match_len`. |
| 1206 |
/// Returns a tuple containing (match_distance, match_length). Will be equal to the input |
| 1207 |
/// values if no better matches were found. |
| 1208 |
fn find_match( |
| 1209 |
&self, |
| 1210 |
lookahead_pos: usize, |
| 1211 |
max_dist: usize, |
| 1212 |
max_match_len: u32, |
| 1213 |
mut match_dist: u32, |
| 1214 |
mut match_len: u32, |
| 1215 |
) -> (u32, u32) { |
| 1216 |
// Clamp the match len and max_match_len to be valid. (It should be when this is called, but |
| 1217 |
// do it for now just in case for safety reasons.) |
| 1218 |
// This should normally end up as at worst conditional moves, |
| 1219 |
// so it shouldn't slow us down much. |
| 1220 |
// TODO: Statically verify these so we don't need to do this. |
| 1221 |
let max_match_len = cmp::min(MAX_MATCH_LEN as u32, max_match_len); |
| 1222 |
match_len = cmp::max(match_len, 1); |
| 1223 |
|
| 1224 |
// If we already have a match of the full length don't bother searching for another one. |
| 1225 |
if max_match_len <= match_len { |
| 1226 |
return (match_dist, match_len); |
| 1227 |
} |
| 1228 |
|
| 1229 |
let pos = lookahead_pos & LZ_DICT_SIZE_MASK; |
| 1230 |
let mut probe_pos = pos; |
| 1231 |
// Number of probes into the hash chains. |
| 1232 |
let mut num_probes_left = if match_len < 32 { |
| 1233 |
self.max_probes[0] |
| 1234 |
} else { |
| 1235 |
self.max_probes[1] |
| 1236 |
}; |
| 1237 |
|
| 1238 |
// Read the last byte of the current match, and the next one, used to compare matches. |
| 1239 |
let mut c01: u16 = read_u16_le(&self.b.dict, pos + match_len as usize - 1); |
| 1240 |
// Read the two bytes at the end position of the current match. |
| 1241 |
let s01: u16 = read_u16_le(&self.b.dict, pos); |
| 1242 |
|
| 1243 |
'outer: loop { |
| 1244 |
let mut dist; |
| 1245 |
'found: loop { |
| 1246 |
num_probes_left -= 1; |
| 1247 |
if num_probes_left == 0 { |
| 1248 |
// We have done as many probes in the hash chain as the current compression |
| 1249 |
// settings allow, so return the best match we found, if any. |
| 1250 |
return (match_dist, match_len); |
| 1251 |
} |
| 1252 |
|
| 1253 |
for _ in 0..3 { |
| 1254 |
let next_probe_pos = self.b.next[probe_pos] as usize; |
| 1255 |
|
| 1256 |
dist = (lookahead_pos - next_probe_pos) & 0xFFFF; |
| 1257 |
// Optimization: The last condition should never be hit but helps the compiler by avoiding |
| 1258 |
// doing the bounds check in the read_u16_le call and adding the extra instructions |
| 1259 |
// for branching to a panic after that and instead just adds the extra instruction here |
| 1260 |
// instead saving some instructions and thus improving performance a bit. |
| 1261 |
// May want to investigate whether we can avoid it entirely but as of now the compiler |
| 1262 |
// isn't able to deduce that match_len - 1 is bounded to [1-257] |
| 1263 |
// Disable clippy lint as it needs to be written in this specific way |
| 1264 |
// rather than MAX_MATCH_LEN to work |
| 1265 |
// because the compiler isn't super smart.... |
| 1266 |
#[allow(clippy::int_plus_one)] |
| 1267 |
if next_probe_pos == 0 |
| 1268 |
|| dist > max_dist |
| 1269 |
|| match_len as usize - 1 >= MAX_MATCH_LEN |
| 1270 |
{ |
| 1271 |
// We reached the end of the hash chain, or the next value is further away |
| 1272 |
// than the maximum allowed distance, so return the best match we found, if |
| 1273 |
// any. |
| 1274 |
return (match_dist, match_len); |
| 1275 |
} |
| 1276 |
|
| 1277 |
// Mask the position value to get the position in the hash chain of the next |
| 1278 |
// position to match against. |
| 1279 |
probe_pos = next_probe_pos & LZ_DICT_SIZE_MASK; |
| 1280 |
|
| 1281 |
if read_u16_le(&self.b.dict, probe_pos + match_len as usize - 1) == c01 { |
| 1282 |
break 'found; |
| 1283 |
} |
| 1284 |
} |
| 1285 |
} |
| 1286 |
|
| 1287 |
if dist == 0 { |
| 1288 |
// We've looked through the whole match range, so return the best match we |
| 1289 |
// found. |
| 1290 |
return (match_dist, match_len); |
| 1291 |
} |
| 1292 |
|
| 1293 |
// Check if the two first bytes match. |
| 1294 |
if read_u16_le(&self.b.dict, probe_pos) != s01 { |
| 1295 |
continue; |
| 1296 |
} |
| 1297 |
|
| 1298 |
let mut p = pos + 2; |
| 1299 |
let mut q = probe_pos + 2; |
| 1300 |
// The first two bytes matched, so check the full length of the match. |
| 1301 |
// TODO: This is a workaround for an upstream issue introduced after a LLVM upgrade in rust 1.82. |
| 1302 |
// the compiler is too smart and ends up unrolling the loop which causes the performance to get worse |
| 1303 |
// Using a variable instead of a constant here to prevent it seems to at least get back some of the performance loss. |
| 1304 |
for _ in 0..self.loop_len as i32 { |
| 1305 |
let p_data: u64 = self.read_unaligned_u64(p); |
| 1306 |
let q_data: u64 = self.read_unaligned_u64(q); |
| 1307 |
// Compare of 8 bytes at a time by using unaligned loads of 64-bit integers. |
| 1308 |
let xor_data = p_data ^ q_data; |
| 1309 |
if xor_data == 0 { |
| 1310 |
p += 8; |
| 1311 |
q += 8; |
| 1312 |
} else { |
| 1313 |
// If not all of the last 8 bytes matched, check how may of them did. |
| 1314 |
let trailing = xor_data.trailing_zeros(); |
| 1315 |
|
| 1316 |
let probe_len = p - pos + (trailing as usize >> 3); |
| 1317 |
if probe_len > match_len as usize { |
| 1318 |
match_dist = dist as u32; |
| 1319 |
match_len = cmp::min(max_match_len, probe_len as u32); |
| 1320 |
if match_len >= max_match_len { |
| 1321 |
// We found a match that had the maximum allowed length, |
| 1322 |
// so there is now point searching further. |
| 1323 |
return (match_dist, match_len); |
| 1324 |
} |
| 1325 |
// We found a better match, so save the last two bytes for further match |
| 1326 |
// comparisons. |
| 1327 |
// Optimization: use saturating_sub makes the compiler able to evade the bounds check |
| 1328 |
// at the cost of some extra instructions since it avoids any possibility of wraparound. |
| 1329 |
// need to see if we can find a better way to do this since this is still a bit costly. |
| 1330 |
c01 = |
| 1331 |
read_u16_le(&self.b.dict, (pos + match_len as usize).saturating_sub(1)); |
| 1332 |
} |
| 1333 |
continue 'outer; |
| 1334 |
} |
| 1335 |
} |
| 1336 |
|
| 1337 |
return (dist as u32, cmp::min(max_match_len, MAX_MATCH_LEN as u32)); |
| 1338 |
} |
| 1339 |
} |
| 1340 |
} |
| 1341 |
|
| 1342 |
pub(crate) struct ParamsOxide { |
| 1343 |
pub flags: u32, |
| 1344 |
pub greedy_parsing: bool, |
| 1345 |
pub block_index: u32, |
| 1346 |
|
| 1347 |
pub saved_match_dist: u32, |
| 1348 |
pub saved_match_len: u32, |
| 1349 |
pub saved_lit: u8, |
| 1350 |
|
| 1351 |
pub flush: TDEFLFlush, |
| 1352 |
pub flush_ofs: u32, |
| 1353 |
pub flush_remaining: u32, |
| 1354 |
pub finished: bool, |
| 1355 |
|
| 1356 |
pub adler32: u32, |
| 1357 |
|
| 1358 |
pub src_pos: usize, |
| 1359 |
|
| 1360 |
pub out_buf_ofs: usize, |
| 1361 |
pub prev_return_status: TDEFLStatus, |
| 1362 |
|
| 1363 |
pub saved_bit_buffer: u32, |
| 1364 |
pub saved_bits_in: u32, |
| 1365 |
|
| 1366 |
pub local_buf: Box<LocalBuf>, |
| 1367 |
} |
| 1368 |
|
| 1369 |
impl ParamsOxide { |
| 1370 |
fn new(flags: u32) -> Self { |
| 1371 |
ParamsOxide { |
| 1372 |
flags, |
| 1373 |
greedy_parsing: flags & TDEFL_GREEDY_PARSING_FLAG != 0, |
| 1374 |
block_index: 0, |
| 1375 |
saved_match_dist: 0, |
| 1376 |
saved_match_len: 0, |
| 1377 |
saved_lit: 0, |
| 1378 |
flush: TDEFLFlush::None, |
| 1379 |
flush_ofs: 0, |
| 1380 |
flush_remaining: 0, |
| 1381 |
finished: false, |
| 1382 |
adler32: MZ_ADLER32_INIT, |
| 1383 |
src_pos: 0, |
| 1384 |
out_buf_ofs: 0, |
| 1385 |
prev_return_status: TDEFLStatus::Okay, |
| 1386 |
saved_bit_buffer: 0, |
| 1387 |
saved_bits_in: 0, |
| 1388 |
local_buf: Box::default(), |
| 1389 |
} |
| 1390 |
} |
| 1391 |
|
| 1392 |
fn update_flags(&mut self, flags: u32) { |
| 1393 |
self.flags = flags; |
| 1394 |
self.greedy_parsing = self.flags & TDEFL_GREEDY_PARSING_FLAG != 0; |
| 1395 |
} |
| 1396 |
|
| 1397 |
/// Reset state, saving settings. |
| 1398 |
fn reset(&mut self) { |
| 1399 |
self.block_index = 0; |
| 1400 |
self.saved_match_len = 0; |
| 1401 |
self.saved_match_dist = 0; |
| 1402 |
self.saved_lit = 0; |
| 1403 |
self.flush = TDEFLFlush::None; |
| 1404 |
self.flush_ofs = 0; |
| 1405 |
self.flush_remaining = 0; |
| 1406 |
self.finished = false; |
| 1407 |
self.adler32 = MZ_ADLER32_INIT; |
| 1408 |
self.src_pos = 0; |
| 1409 |
self.out_buf_ofs = 0; |
| 1410 |
self.prev_return_status = TDEFLStatus::Okay; |
| 1411 |
self.saved_bit_buffer = 0; |
| 1412 |
self.saved_bits_in = 0; |
| 1413 |
self.local_buf.b = [0; OUT_BUF_SIZE]; |
| 1414 |
} |
| 1415 |
} |
| 1416 |
|
| 1417 |
pub(crate) struct LZOxide { |
| 1418 |
pub codes: [u8; LZ_CODE_BUF_SIZE], |
| 1419 |
pub code_position: usize, |
| 1420 |
pub flag_position: usize, |
| 1421 |
|
| 1422 |
// The total number of bytes in the current block. |
| 1423 |
pub total_bytes: u32, |
| 1424 |
pub num_flags_left: u32, |
| 1425 |
} |
| 1426 |
|
| 1427 |
impl LZOxide { |
| 1428 |
const fn new() -> Self { |
| 1429 |
LZOxide { |
| 1430 |
codes: [0; LZ_CODE_BUF_SIZE], |
| 1431 |
code_position: 1, |
| 1432 |
flag_position: 0, |
| 1433 |
total_bytes: 0, |
| 1434 |
num_flags_left: 8, |
| 1435 |
} |
| 1436 |
} |
| 1437 |
|
| 1438 |
fn write_code(&mut self, val: u8) { |
| 1439 |
// Perf - go via u16 to help evade bounds check |
| 1440 |
// TODO: see if we can use u16 for flag_position in general. |
| 1441 |
self.codes[usize::from(self.code_position as u16)] = val; |
| 1442 |
self.code_position += 1; |
| 1443 |
} |
| 1444 |
|
| 1445 |
fn init_flag(&mut self) { |
| 1446 |
if self.num_flags_left == 8 { |
| 1447 |
*self.get_flag() = 0; |
| 1448 |
self.code_position -= 1; |
| 1449 |
} else { |
| 1450 |
*self.get_flag() >>= self.num_flags_left; |
| 1451 |
} |
| 1452 |
} |
| 1453 |
|
| 1454 |
fn get_flag(&mut self) -> &mut u8 { |
| 1455 |
// Perf - go via u16 to help evade bounds check |
| 1456 |
// TODO: see if we can use u16 for flag_position in general. |
| 1457 |
&mut self.codes[usize::from(self.flag_position as u16)] |
| 1458 |
} |
| 1459 |
|
| 1460 |
fn plant_flag(&mut self) { |
| 1461 |
self.flag_position = self.code_position; |
| 1462 |
self.code_position += 1; |
| 1463 |
} |
| 1464 |
|
| 1465 |
fn consume_flag(&mut self) { |
| 1466 |
self.num_flags_left -= 1; |
| 1467 |
if self.num_flags_left == 0 { |
| 1468 |
self.num_flags_left = 8; |
| 1469 |
self.plant_flag(); |
| 1470 |
} |
| 1471 |
} |
| 1472 |
} |
| 1473 |
|
| 1474 |
fn compress_lz_codes( |
| 1475 |
huff: &HuffmanOxide, |
| 1476 |
output: &mut OutputBufferOxide, |
| 1477 |
lz_code_buf: &[u8; LZ_CODE_BUF_SIZE], |
| 1478 |
lz_code_buf_used_len: usize, |
| 1479 |
) -> Result<bool> { |
| 1480 |
let mut flags = 1; |
| 1481 |
let mut bb = BitBuffer { |
| 1482 |
bit_buffer: u64::from(output.bit_buffer), |
| 1483 |
bits_in: output.bits_in, |
| 1484 |
}; |
| 1485 |
|
| 1486 |
// Help out the compiler know this variable won't be larger than |
| 1487 |
// the buffer length since the constants won't propagate through the function call. |
| 1488 |
let lz_code_buf_used_len = cmp::min(lz_code_buf.len(), lz_code_buf_used_len); |
| 1489 |
|
| 1490 |
let mut i: usize = 0; |
| 1491 |
while i < lz_code_buf_used_len { |
| 1492 |
if flags == 1 { |
| 1493 |
flags = u32::from(lz_code_buf[i]) | 0x100; |
| 1494 |
i += 1; |
| 1495 |
} |
| 1496 |
|
| 1497 |
// The lz code was a length code |
| 1498 |
if flags & 1 == 1 { |
| 1499 |
flags >>= 1; |
| 1500 |
|
| 1501 |
let sym; |
| 1502 |
let num_extra_bits; |
| 1503 |
|
| 1504 |
let match_len = lz_code_buf[i & LZ_CODE_BUF_MASK] as usize; |
| 1505 |
|
| 1506 |
let match_dist = lz_code_buf[(i + 1) & LZ_CODE_BUF_MASK] as u16 |
| 1507 |
| ((lz_code_buf[(i + 2) & LZ_CODE_BUF_MASK] as u16) << 8); |
| 1508 |
|
| 1509 |
i += 3; |
| 1510 |
|
| 1511 |
debug_assert!(huff.code_sizes[0][LEN_SYM[match_len] as usize + LEN_SYM_OFFSET] != 0); |
| 1512 |
let len_sym = (LEN_SYM[match_len] & 31) as usize + LEN_SYM_OFFSET; |
| 1513 |
|
| 1514 |
bb.put_fast( |
| 1515 |
u64::from(huff.codes[0][len_sym]), |
| 1516 |
u32::from(huff.code_sizes[0][len_sym]), |
| 1517 |
); |
| 1518 |
bb.put_fast( |
| 1519 |
match_len as u64 & u64::from(BITMASKS[(LEN_EXTRA[match_len] & 7) as usize]), |
| 1520 |
u32::from(LEN_EXTRA[match_len]), |
| 1521 |
); |
| 1522 |
|
| 1523 |
if match_dist < 512 { |
| 1524 |
sym = SMALL_DIST_SYM[match_dist as usize] as usize; |
| 1525 |
num_extra_bits = SMALL_DIST_EXTRA[match_dist as usize] as usize; |
| 1526 |
} else { |
| 1527 |
sym = LARGE_DIST_SYM[(match_dist >> 8) as usize] as usize; |
| 1528 |
num_extra_bits = LARGE_DIST_EXTRA[(match_dist >> 8) as usize] as usize; |
| 1529 |
} |
| 1530 |
|
| 1531 |
debug_assert!(huff.code_sizes[1][sym] != 0); |
| 1532 |
bb.put_fast( |
| 1533 |
u64::from(huff.codes[1][sym]), |
| 1534 |
u32::from(huff.code_sizes[1][sym]), |
| 1535 |
); |
| 1536 |
bb.put_fast( |
| 1537 |
u64::from(match_dist) & u64::from(BITMASKS[num_extra_bits & 15]), |
| 1538 |
num_extra_bits as u32, |
| 1539 |
); |
| 1540 |
} else { |
| 1541 |
// The lz code was a literal |
| 1542 |
for _ in 0..3 { |
| 1543 |
flags >>= 1; |
| 1544 |
let lit = lz_code_buf[i & LZ_CODE_BUF_MASK]; |
| 1545 |
i += 1; |
| 1546 |
|
| 1547 |
debug_assert!(huff.code_sizes[0][lit as usize] != 0); |
| 1548 |
bb.put_fast( |
| 1549 |
u64::from(huff.codes[0][lit as usize]), |
| 1550 |
u32::from(huff.code_sizes[0][lit as usize]), |
| 1551 |
); |
| 1552 |
|
| 1553 |
if flags & 1 == 1 || i >= lz_code_buf_used_len { |
| 1554 |
break; |
| 1555 |
} |
| 1556 |
} |
| 1557 |
} |
| 1558 |
|
| 1559 |
bb.flush(output)?; |
| 1560 |
} |
| 1561 |
|
| 1562 |
output.bits_in = 0; |
| 1563 |
output.bit_buffer = 0; |
| 1564 |
while bb.bits_in != 0 { |
| 1565 |
let n = cmp::min(bb.bits_in, 16); |
| 1566 |
output.put_bits(bb.bit_buffer as u32 & BITMASKS[n as usize], n); |
| 1567 |
bb.bit_buffer >>= n; |
| 1568 |
bb.bits_in -= n; |
| 1569 |
} |
| 1570 |
|
| 1571 |
// Output the end of block symbol. |
| 1572 |
output.put_bits( |
| 1573 |
u32::from(huff.codes[0][256]), |
| 1574 |
u32::from(huff.code_sizes[0][256]), |
| 1575 |
); |
| 1576 |
|
| 1577 |
Ok(true) |
| 1578 |
} |
| 1579 |
|
| 1580 |
fn compress_block( |
| 1581 |
huff: &mut HuffmanOxide, |
| 1582 |
output: &mut OutputBufferOxide, |
| 1583 |
lz: &LZOxide, |
| 1584 |
static_block: bool, |
| 1585 |
) -> Result<bool> { |
| 1586 |
if static_block { |
| 1587 |
huff.start_static_block(output); |
| 1588 |
} else { |
| 1589 |
huff.start_dynamic_block(output)?; |
| 1590 |
} |
| 1591 |
|
| 1592 |
compress_lz_codes(huff, output, &lz.codes, lz.code_position) |
| 1593 |
} |
| 1594 |
|
| 1595 |
pub(crate) fn flush_block( |
| 1596 |
d: &mut CompressorOxide, |
| 1597 |
callback: &mut CallbackOxide, |
| 1598 |
flush: TDEFLFlush, |
| 1599 |
) -> Result<i32> { |
| 1600 |
let mut saved_buffer; |
| 1601 |
{ |
| 1602 |
let mut output = callback |
| 1603 |
.out |
| 1604 |
.new_output_buffer(&mut d.params.local_buf.b, d.params.out_buf_ofs); |
| 1605 |
output.bit_buffer = d.params.saved_bit_buffer; |
| 1606 |
output.bits_in = d.params.saved_bits_in; |
| 1607 |
|
| 1608 |
// TODO: Don't think this second condition should be here but need to verify. |
| 1609 |
let use_raw_block = (d.params.flags & TDEFL_FORCE_ALL_RAW_BLOCKS != 0) |
| 1610 |
&& (d.dict.lookahead_pos - d.dict.code_buf_dict_pos) <= d.dict.size; |
| 1611 |
debug_assert_eq!( |
| 1612 |
use_raw_block, |
| 1613 |
d.params.flags & TDEFL_FORCE_ALL_RAW_BLOCKS != 0 |
| 1614 |
); |
| 1615 |
|
| 1616 |
assert!(d.params.flush_remaining == 0); |
| 1617 |
d.params.flush_ofs = 0; |
| 1618 |
d.params.flush_remaining = 0; |
| 1619 |
|
| 1620 |
d.lz.init_flag(); |
| 1621 |
|
| 1622 |
// If we are at the start of the stream, write the zlib header if requested. |
| 1623 |
if d.params.flags & TDEFL_WRITE_ZLIB_HEADER != 0 && d.params.block_index == 0 { |
| 1624 |
let header = zlib::header_from_flags(d.params.flags); |
| 1625 |
output.put_bits_no_flush(header[0].into(), 8); |
| 1626 |
output.put_bits(header[1].into(), 8); |
| 1627 |
} |
| 1628 |
|
| 1629 |
// Output the block header. |
| 1630 |
output.put_bits((flush == TDEFLFlush::Finish) as u32, 1); |
| 1631 |
|
| 1632 |
saved_buffer = output.save(); |
| 1633 |
|
| 1634 |
let comp_success = if !use_raw_block { |
| 1635 |
let use_static = |
| 1636 |
(d.params.flags & TDEFL_FORCE_ALL_STATIC_BLOCKS != 0) || (d.lz.total_bytes < 48); |
| 1637 |
compress_block(&mut d.huff, &mut output, &d.lz, use_static)? |
| 1638 |
} else { |
| 1639 |
false |
| 1640 |
}; |
| 1641 |
|
| 1642 |
// If we failed to compress anything and the output would take up more space than the output |
| 1643 |
// data, output a stored block instead, which has at most 5 bytes of overhead. |
| 1644 |
// We only use some simple heuristics for now. |
| 1645 |
// A stored block will have an overhead of at least 4 bytes containing the block length |
| 1646 |
// but usually more due to the length parameters having to start at a byte boundary and thus |
| 1647 |
// requiring up to 5 bytes of padding. |
| 1648 |
// As a static block will have an overhead of at most 1 bit per byte |
| 1649 |
// (as literals are either 8 or 9 bytes), a raw block will |
| 1650 |
// never take up less space if the number of input bytes are less than 32. |
| 1651 |
let expanded = (d.lz.total_bytes > 32) |
| 1652 |
&& (output.inner_pos - saved_buffer.pos + 1 >= (d.lz.total_bytes as usize)) |
| 1653 |
&& (d.dict.lookahead_pos - d.dict.code_buf_dict_pos <= d.dict.size); |
| 1654 |
|
| 1655 |
if use_raw_block || expanded { |
| 1656 |
output.load(saved_buffer); |
| 1657 |
|
| 1658 |
// Block header. |
| 1659 |
output.put_bits(0, 2); |
| 1660 |
|
| 1661 |
// Block length has to start on a byte boundary, so pad. |
| 1662 |
output.pad_to_bytes(); |
| 1663 |
|
| 1664 |
// Block length and ones complement of block length. |
| 1665 |
output.put_bits(d.lz.total_bytes & 0xFFFF, 16); |
| 1666 |
output.put_bits(!d.lz.total_bytes & 0xFFFF, 16); |
| 1667 |
|
| 1668 |
// Write the actual bytes. |
| 1669 |
let start = d.dict.code_buf_dict_pos & LZ_DICT_SIZE_MASK; |
| 1670 |
let end = (d.dict.code_buf_dict_pos + d.lz.total_bytes as usize) & LZ_DICT_SIZE_MASK; |
| 1671 |
let dict = &mut d.dict.b.dict; |
| 1672 |
if start < end { |
| 1673 |
// The data does not wrap around. |
| 1674 |
output.write_bytes(&dict[start..end]); |
| 1675 |
} else if d.lz.total_bytes > 0 { |
| 1676 |
// The data wraps around and the input was not 0 bytes. |
| 1677 |
output.write_bytes(&dict[start..LZ_DICT_SIZE]); |
| 1678 |
output.write_bytes(&dict[..end]); |
| 1679 |
} |
| 1680 |
} else if !comp_success { |
| 1681 |
output.load(saved_buffer); |
| 1682 |
compress_block(&mut d.huff, &mut output, &d.lz, true)?; |
| 1683 |
} |
| 1684 |
|
| 1685 |
if flush != TDEFLFlush::None { |
| 1686 |
if flush == TDEFLFlush::Finish { |
| 1687 |
output.pad_to_bytes(); |
| 1688 |
if d.params.flags & TDEFL_WRITE_ZLIB_HEADER != 0 { |
| 1689 |
let mut adler = d.params.adler32; |
| 1690 |
for _ in 0..4 { |
| 1691 |
output.put_bits((adler >> 24) & 0xFF, 8); |
| 1692 |
adler <<= 8; |
| 1693 |
} |
| 1694 |
} |
| 1695 |
} else { |
| 1696 |
// Sync or Full flush. |
| 1697 |
// Output an empty raw block. |
| 1698 |
output.put_bits(0, 3); |
| 1699 |
output.pad_to_bytes(); |
| 1700 |
output.put_bits(0, 16); |
| 1701 |
output.put_bits(0xFFFF, 16); |
| 1702 |
} |
| 1703 |
} |
| 1704 |
|
| 1705 |
d.huff.count[0][..MAX_HUFF_SYMBOLS_0].fill(0); |
| 1706 |
d.huff.count[1][..MAX_HUFF_SYMBOLS_1].fill(0); |
| 1707 |
|
| 1708 |
// Clear LZ buffer for the next block. |
| 1709 |
d.lz.code_position = 1; |
| 1710 |
d.lz.flag_position = 0; |
| 1711 |
d.lz.num_flags_left = 8; |
| 1712 |
d.dict.code_buf_dict_pos += d.lz.total_bytes as usize; |
| 1713 |
d.lz.total_bytes = 0; |
| 1714 |
d.params.block_index += 1; |
| 1715 |
|
| 1716 |
saved_buffer = output.save(); |
| 1717 |
|
| 1718 |
d.params.saved_bit_buffer = saved_buffer.bit_buffer; |
| 1719 |
d.params.saved_bits_in = saved_buffer.bits_in; |
| 1720 |
} |
| 1721 |
|
| 1722 |
Ok(callback.flush_output(saved_buffer, &mut d.params)) |
| 1723 |
} |
| 1724 |
|
| 1725 |
pub(crate) fn record_literal(h: &mut HuffmanOxide, lz: &mut LZOxide, lit: u8) { |
| 1726 |
lz.total_bytes += 1; |
| 1727 |
lz.write_code(lit); |
| 1728 |
|
| 1729 |
*lz.get_flag() >>= 1; |
| 1730 |
lz.consume_flag(); |
| 1731 |
|
| 1732 |
h.count[0][lit as usize] += 1; |
| 1733 |
} |
| 1734 |
|
| 1735 |
fn record_match(h: &mut HuffmanOxide, lz: &mut LZOxide, match_len: u32, mut match_dist: u32) { |
| 1736 |
debug_assert!(match_len >= MIN_MATCH_LEN.into()); |
| 1737 |
debug_assert!(match_dist >= 1); |
| 1738 |
debug_assert!(match_dist as usize <= LZ_DICT_SIZE); |
| 1739 |
|
| 1740 |
lz.total_bytes += match_len; |
| 1741 |
match_dist -= 1; |
| 1742 |
let match_len = (match_len - u32::from(MIN_MATCH_LEN)) as u8; |
| 1743 |
lz.write_code(match_len); |
| 1744 |
lz.write_code(match_dist as u8); |
| 1745 |
lz.write_code((match_dist >> 8) as u8); |
| 1746 |
|
| 1747 |
*lz.get_flag() >>= 1; |
| 1748 |
*lz.get_flag() |= 0x80; |
| 1749 |
lz.consume_flag(); |
| 1750 |
|
| 1751 |
let symbol = if match_dist < 512 { |
| 1752 |
SMALL_DIST_SYM[match_dist as usize] |
| 1753 |
} else { |
| 1754 |
LARGE_DIST_SYM[((match_dist >> 8) & 127) as usize] |
| 1755 |
} as usize; |
| 1756 |
h.count[1][symbol] += 1; |
| 1757 |
// Mask the values from LEN_SYM here as the compiler isn't quite smart enough to infer |
| 1758 |
// that it only contains values smaller than 32. |
| 1759 |
h.count[0][(LEN_SYM[match_len as usize] as usize & 31) + LEN_SYM_OFFSET] += 1; |
| 1760 |
} |
| 1761 |
|
| 1762 |
fn compress_normal(d: &mut CompressorOxide, callback: &mut CallbackOxide) -> bool { |
| 1763 |
let in_buf = match callback.in_buf { |
| 1764 |
None => return true, |
| 1765 |
Some(in_buf) => in_buf, |
| 1766 |
}; |
| 1767 |
|
| 1768 |
let mut src_pos = d.params.src_pos; |
| 1769 |
let mut lookahead_size = d.dict.lookahead_size; |
| 1770 |
let mut lookahead_pos = d.dict.lookahead_pos; |
| 1771 |
let mut saved_lit = d.params.saved_lit; |
| 1772 |
let mut saved_match_dist = d.params.saved_match_dist; |
| 1773 |
let mut saved_match_len = d.params.saved_match_len; |
| 1774 |
|
| 1775 |
while src_pos < in_buf.len() || (d.params.flush != TDEFLFlush::None && lookahead_size != 0) { |
| 1776 |
let src_buf_left = in_buf.len() - src_pos; |
| 1777 |
let num_bytes_to_process = cmp::min(src_buf_left, MAX_MATCH_LEN - lookahead_size); |
| 1778 |
|
| 1779 |
if lookahead_size + d.dict.size >= usize::from(MIN_MATCH_LEN) - 1 |
| 1780 |
&& num_bytes_to_process > 0 |
| 1781 |
{ |
| 1782 |
let dictb = &mut d.dict.b; |
| 1783 |
|
| 1784 |
let mut dst_pos = (lookahead_pos + lookahead_size) & LZ_DICT_SIZE_MASK; |
| 1785 |
let mut ins_pos = lookahead_pos + lookahead_size - 2; |
| 1786 |
// Start the hash value from the first two bytes |
| 1787 |
let mut hash = update_hash( |
| 1788 |
u16::from(dictb.dict[ins_pos & LZ_DICT_SIZE_MASK]), |
| 1789 |
dictb.dict[(ins_pos + 1) & LZ_DICT_SIZE_MASK], |
| 1790 |
); |
| 1791 |
|
| 1792 |
lookahead_size += num_bytes_to_process; |
| 1793 |
|
| 1794 |
for &c in &in_buf[src_pos..src_pos + num_bytes_to_process] { |
| 1795 |
// Add byte to input buffer. |
| 1796 |
dictb.dict[dst_pos] = c; |
| 1797 |
if dst_pos < MAX_MATCH_LEN - 1 { |
| 1798 |
dictb.dict[LZ_DICT_SIZE + dst_pos] = c; |
| 1799 |
} |
| 1800 |
|
| 1801 |
// Generate hash from the current byte, |
| 1802 |
hash = update_hash(hash, c); |
| 1803 |
dictb.next[ins_pos & LZ_DICT_SIZE_MASK] = dictb.hash[hash as usize]; |
| 1804 |
// and insert it into the hash chain. |
| 1805 |
dictb.hash[hash as usize] = ins_pos as u16; |
| 1806 |
dst_pos = (dst_pos + 1) & LZ_DICT_SIZE_MASK; |
| 1807 |
ins_pos += 1; |
| 1808 |
} |
| 1809 |
|
| 1810 |
src_pos += num_bytes_to_process; |
| 1811 |
} else { |
| 1812 |
let dictb = &mut d.dict.b; |
| 1813 |
for &c in &in_buf[src_pos..src_pos + num_bytes_to_process] { |
| 1814 |
let dst_pos = (lookahead_pos + lookahead_size) & LZ_DICT_SIZE_MASK; |
| 1815 |
dictb.dict[dst_pos] = c; |
| 1816 |
if dst_pos < MAX_MATCH_LEN - 1 { |
| 1817 |
dictb.dict[LZ_DICT_SIZE + dst_pos] = c; |
| 1818 |
} |
| 1819 |
|
| 1820 |
lookahead_size += 1; |
| 1821 |
if lookahead_size + d.dict.size >= MIN_MATCH_LEN.into() { |
| 1822 |
let ins_pos = lookahead_pos + lookahead_size - 3; |
| 1823 |
let hash = ((u32::from(dictb.dict[ins_pos & LZ_DICT_SIZE_MASK]) |
| 1824 |
<< (LZ_HASH_SHIFT * 2)) |
| 1825 |
^ ((u32::from(dictb.dict[(ins_pos + 1) & LZ_DICT_SIZE_MASK]) |
| 1826 |
<< LZ_HASH_SHIFT) |
| 1827 |
^ u32::from(c))) |
| 1828 |
& (LZ_HASH_SIZE as u32 - 1); |
| 1829 |
|
| 1830 |
dictb.next[ins_pos & LZ_DICT_SIZE_MASK] = dictb.hash[hash as usize]; |
| 1831 |
dictb.hash[hash as usize] = ins_pos as u16; |
| 1832 |
} |
| 1833 |
} |
| 1834 |
|
| 1835 |
src_pos += num_bytes_to_process; |
| 1836 |
} |
| 1837 |
|
| 1838 |
d.dict.size = cmp::min(LZ_DICT_SIZE - lookahead_size, d.dict.size); |
| 1839 |
if d.params.flush == TDEFLFlush::None && lookahead_size < MAX_MATCH_LEN { |
| 1840 |
break; |
| 1841 |
} |
| 1842 |
|
| 1843 |
let mut len_to_move = 1; |
| 1844 |
let mut cur_match_dist = 0; |
| 1845 |
let mut cur_match_len = if saved_match_len != 0 { |
| 1846 |
saved_match_len |
| 1847 |
} else { |
| 1848 |
u32::from(MIN_MATCH_LEN) - 1 |
| 1849 |
}; |
| 1850 |
let cur_pos = lookahead_pos & LZ_DICT_SIZE_MASK; |
| 1851 |
if d.params.flags & TDEFL_RLE_MATCHES != 0 { |
| 1852 |
// If TDEFL_RLE_MATCHES is set, we only look for repeating sequences of the current byte. |
| 1853 |
if d.dict.size != 0 { |
| 1854 |
let c = d.dict.b.dict[(cur_pos.wrapping_sub(1)) & LZ_DICT_SIZE_MASK]; |
| 1855 |
cur_match_len = d.dict.b.dict[cur_pos..(cur_pos + lookahead_size)] |
| 1856 |
.iter() |
| 1857 |
.take_while(|&x| *x == c) |
| 1858 |
.count() as u32; |
| 1859 |
if cur_match_len < MIN_MATCH_LEN.into() { |
| 1860 |
cur_match_len = 0 |
| 1861 |
} else { |
| 1862 |
cur_match_dist = 1 |
| 1863 |
} |
| 1864 |
} |
| 1865 |
} else { |
| 1866 |
// Try to find a match for the bytes at the current position. |
| 1867 |
let dist_len = d.dict.find_match( |
| 1868 |
lookahead_pos, |
| 1869 |
d.dict.size, |
| 1870 |
lookahead_size as u32, |
| 1871 |
cur_match_dist, |
| 1872 |
cur_match_len, |
| 1873 |
); |
| 1874 |
cur_match_dist = dist_len.0; |
| 1875 |
cur_match_len = dist_len.1; |
| 1876 |
} |
| 1877 |
|
| 1878 |
let far_and_small = cur_match_len == MIN_MATCH_LEN.into() && cur_match_dist >= 8 * 1024; |
| 1879 |
let filter_small = d.params.flags & TDEFL_FILTER_MATCHES != 0 && cur_match_len <= 5; |
| 1880 |
if far_and_small || filter_small || cur_pos == cur_match_dist as usize { |
| 1881 |
cur_match_dist = 0; |
| 1882 |
cur_match_len = 0; |
| 1883 |
} |
| 1884 |
|
| 1885 |
if saved_match_len != 0 { |
| 1886 |
if cur_match_len > saved_match_len { |
| 1887 |
record_literal(&mut d.huff, &mut d.lz, saved_lit); |
| 1888 |
if cur_match_len >= 128 { |
| 1889 |
record_match(&mut d.huff, &mut d.lz, cur_match_len, cur_match_dist); |
| 1890 |
saved_match_len = 0; |
| 1891 |
len_to_move = cur_match_len as usize; |
| 1892 |
} else { |
| 1893 |
saved_lit = d.dict.b.dict[cur_pos]; |
| 1894 |
saved_match_dist = cur_match_dist; |
| 1895 |
saved_match_len = cur_match_len; |
| 1896 |
} |
| 1897 |
} else { |
| 1898 |
record_match(&mut d.huff, &mut d.lz, saved_match_len, saved_match_dist); |
| 1899 |
len_to_move = (saved_match_len - 1) as usize; |
| 1900 |
saved_match_len = 0; |
| 1901 |
} |
| 1902 |
} else if cur_match_dist == 0 { |
| 1903 |
record_literal( |
| 1904 |
&mut d.huff, |
| 1905 |
&mut d.lz, |
| 1906 |
d.dict.b.dict[cmp::min(cur_pos, d.dict.b.dict.len() - 1)], |
| 1907 |
); |
| 1908 |
} else if d.params.greedy_parsing |
| 1909 |
|| (d.params.flags & TDEFL_RLE_MATCHES != 0) |
| 1910 |
|| cur_match_len >= 128 |
| 1911 |
{ |
| 1912 |
// If we are using lazy matching, check for matches at the next byte if the current |
| 1913 |
// match was shorter than 128 bytes. |
| 1914 |
record_match(&mut d.huff, &mut d.lz, cur_match_len, cur_match_dist); |
| 1915 |
len_to_move = cur_match_len as usize; |
| 1916 |
} else { |
| 1917 |
saved_lit = d.dict.b.dict[cmp::min(cur_pos, d.dict.b.dict.len() - 1)]; |
| 1918 |
saved_match_dist = cur_match_dist; |
| 1919 |
saved_match_len = cur_match_len; |
| 1920 |
} |
| 1921 |
|
| 1922 |
lookahead_pos += len_to_move; |
| 1923 |
assert!(lookahead_size >= len_to_move); |
| 1924 |
lookahead_size -= len_to_move; |
| 1925 |
d.dict.size = cmp::min(d.dict.size + len_to_move, LZ_DICT_SIZE); |
| 1926 |
|
| 1927 |
let lz_buf_tight = d.lz.code_position > LZ_CODE_BUF_SIZE - 8; |
| 1928 |
let fat = ((d.lz.code_position * 115) >> 7) >= d.lz.total_bytes as usize; |
| 1929 |
let buf_fat = (d.lz.total_bytes > 31 * 1024) && fat; |
| 1930 |
|
| 1931 |
if lz_buf_tight || buf_fat { |
| 1932 |
d.params.src_pos = src_pos; |
| 1933 |
// These values are used in flush_block, so we need to write them back here. |
| 1934 |
d.dict.lookahead_size = lookahead_size; |
| 1935 |
d.dict.lookahead_pos = lookahead_pos; |
| 1936 |
|
| 1937 |
let n = flush_block(d, callback, TDEFLFlush::None) |
| 1938 |
.unwrap_or(TDEFLStatus::PutBufFailed as i32); |
| 1939 |
if n != 0 { |
| 1940 |
d.params.saved_lit = saved_lit; |
| 1941 |
d.params.saved_match_dist = saved_match_dist; |
| 1942 |
d.params.saved_match_len = saved_match_len; |
| 1943 |
return n > 0; |
| 1944 |
} |
| 1945 |
} |
| 1946 |
} |
| 1947 |
|
| 1948 |
d.params.src_pos = src_pos; |
| 1949 |
d.dict.lookahead_size = lookahead_size; |
| 1950 |
d.dict.lookahead_pos = lookahead_pos; |
| 1951 |
d.params.saved_lit = saved_lit; |
| 1952 |
d.params.saved_match_dist = saved_match_dist; |
| 1953 |
d.params.saved_match_len = saved_match_len; |
| 1954 |
true |
| 1955 |
} |
| 1956 |
|
| 1957 |
const COMP_FAST_LOOKAHEAD_SIZE: usize = 4096; |
| 1958 |
|
| 1959 |
fn compress_fast(d: &mut CompressorOxide, callback: &mut CallbackOxide) -> bool { |
| 1960 |
let mut src_pos = d.params.src_pos; |
| 1961 |
let mut lookahead_size = d.dict.lookahead_size; |
| 1962 |
let mut lookahead_pos = d.dict.lookahead_pos; |
| 1963 |
|
| 1964 |
let mut cur_pos = lookahead_pos & LZ_DICT_SIZE_MASK; |
| 1965 |
let in_buf = match callback.in_buf { |
| 1966 |
None => return true, |
| 1967 |
Some(in_buf) => in_buf, |
| 1968 |
}; |
| 1969 |
|
| 1970 |
debug_assert!(d.lz.code_position < LZ_CODE_BUF_SIZE - 2); |
| 1971 |
|
| 1972 |
while src_pos < in_buf.len() || (d.params.flush != TDEFLFlush::None && lookahead_size > 0) { |
| 1973 |
let mut dst_pos = (lookahead_pos + lookahead_size) & LZ_DICT_SIZE_MASK; |
| 1974 |
let mut num_bytes_to_process = cmp::min( |
| 1975 |
in_buf.len() - src_pos, |
| 1976 |
COMP_FAST_LOOKAHEAD_SIZE - lookahead_size, |
| 1977 |
); |
| 1978 |
lookahead_size += num_bytes_to_process; |
| 1979 |
|
| 1980 |
while num_bytes_to_process != 0 { |
| 1981 |
let n = cmp::min(LZ_DICT_SIZE - dst_pos, num_bytes_to_process); |
| 1982 |
d.dict.b.dict[dst_pos..dst_pos + n].copy_from_slice(&in_buf[src_pos..src_pos + n]); |
| 1983 |
|
| 1984 |
if dst_pos < MAX_MATCH_LEN - 1 { |
| 1985 |
let m = cmp::min(n, MAX_MATCH_LEN - 1 - dst_pos); |
| 1986 |
d.dict.b.dict[dst_pos + LZ_DICT_SIZE..dst_pos + LZ_DICT_SIZE + m] |
| 1987 |
.copy_from_slice(&in_buf[src_pos..src_pos + m]); |
| 1988 |
} |
| 1989 |
|
| 1990 |
src_pos += n; |
| 1991 |
dst_pos = (dst_pos + n) & LZ_DICT_SIZE_MASK; |
| 1992 |
num_bytes_to_process -= n; |
| 1993 |
} |
| 1994 |
|
| 1995 |
d.dict.size = cmp::min(LZ_DICT_SIZE - lookahead_size, d.dict.size); |
| 1996 |
if d.params.flush == TDEFLFlush::None && lookahead_size < COMP_FAST_LOOKAHEAD_SIZE { |
| 1997 |
break; |
| 1998 |
} |
| 1999 |
|
| 2000 |
while lookahead_size >= 4 { |
| 2001 |
let mut cur_match_len = 1; |
| 2002 |
|
| 2003 |
let first_trigram = d.dict.read_unaligned_u32(cur_pos) & 0xFF_FFFF; |
| 2004 |
|
| 2005 |
let hash = (first_trigram ^ (first_trigram >> (24 - (LZ_HASH_BITS - 8)))) |
| 2006 |
& LEVEL1_HASH_SIZE_MASK; |
| 2007 |
|
| 2008 |
let mut probe_pos = usize::from(d.dict.b.hash[hash as usize]); |
| 2009 |
d.dict.b.hash[hash as usize] = lookahead_pos as u16; |
| 2010 |
|
| 2011 |
let mut cur_match_dist = (lookahead_pos - probe_pos) as u16; |
| 2012 |
if cur_match_dist as usize <= d.dict.size { |
| 2013 |
probe_pos &= LZ_DICT_SIZE_MASK; |
| 2014 |
|
| 2015 |
let trigram = d.dict.read_unaligned_u32(probe_pos) & 0xFF_FFFF; |
| 2016 |
|
| 2017 |
if first_trigram == trigram { |
| 2018 |
// Trigram was tested, so we can start with "+ 3" displacement. |
| 2019 |
let mut p = cur_pos + 3; |
| 2020 |
let mut q = probe_pos + 3; |
| 2021 |
cur_match_len = (|| { |
| 2022 |
for _ in 0..32 { |
| 2023 |
let p_data: u64 = d.dict.read_unaligned_u64(p); |
| 2024 |
let q_data: u64 = d.dict.read_unaligned_u64(q); |
| 2025 |
let xor_data = p_data ^ q_data; |
| 2026 |
if xor_data == 0 { |
| 2027 |
p += 8; |
| 2028 |
q += 8; |
| 2029 |
} else { |
| 2030 |
let trailing = xor_data.trailing_zeros(); |
| 2031 |
return p as u32 - cur_pos as u32 + (trailing >> 3); |
| 2032 |
} |
| 2033 |
} |
| 2034 |
|
| 2035 |
if cur_match_dist == 0 { |
| 2036 |
0 |
| 2037 |
} else { |
| 2038 |
MAX_MATCH_LEN as u32 |
| 2039 |
} |
| 2040 |
})(); |
| 2041 |
|
| 2042 |
if cur_match_len < MIN_MATCH_LEN.into() |
| 2043 |
|| (cur_match_len == MIN_MATCH_LEN.into() && cur_match_dist >= 8 * 1024) |
| 2044 |
{ |
| 2045 |
let lit = first_trigram as u8; |
| 2046 |
cur_match_len = 1; |
| 2047 |
d.lz.write_code(lit); |
| 2048 |
*d.lz.get_flag() >>= 1; |
| 2049 |
d.huff.count[0][lit as usize] += 1; |
| 2050 |
} else { |
| 2051 |
// Limit the match to the length of the lookahead so we don't create a match |
| 2052 |
// that ends after the end of the input data. |
| 2053 |
cur_match_len = cmp::min(cur_match_len, lookahead_size as u32); |
| 2054 |
debug_assert!(cur_match_len >= MIN_MATCH_LEN.into()); |
| 2055 |
debug_assert!(cur_match_len <= MAX_MATCH_LEN as u32); |
| 2056 |
debug_assert!(cur_match_dist >= 1); |
| 2057 |
debug_assert!(cur_match_dist as usize <= LZ_DICT_SIZE); |
| 2058 |
cur_match_dist -= 1; |
| 2059 |
|
| 2060 |
d.lz.write_code((cur_match_len - u32::from(MIN_MATCH_LEN)) as u8); |
| 2061 |
d.lz.write_code(cur_match_dist as u8); |
| 2062 |
d.lz.write_code((cur_match_dist >> 8) as u8); |
| 2063 |
|
| 2064 |
*d.lz.get_flag() >>= 1; |
| 2065 |
*d.lz.get_flag() |= 0x80; |
| 2066 |
if cur_match_dist < 512 { |
| 2067 |
d.huff.count[1][SMALL_DIST_SYM[cur_match_dist as usize] as usize] += 1; |
| 2068 |
} else { |
| 2069 |
d.huff.count[1] |
| 2070 |
[LARGE_DIST_SYM[(cur_match_dist >> 8) as usize] as usize] += 1; |
| 2071 |
} |
| 2072 |
|
| 2073 |
d.huff.count[0][(LEN_SYM |
| 2074 |
[(cur_match_len - u32::from(MIN_MATCH_LEN)) as usize & 255] |
| 2075 |
as usize |
| 2076 |
& 31) |
| 2077 |
+ LEN_SYM_OFFSET] += 1; |
| 2078 |
} |
| 2079 |
} else { |
| 2080 |
d.lz.write_code(first_trigram as u8); |
| 2081 |
*d.lz.get_flag() >>= 1; |
| 2082 |
d.huff.count[0][first_trigram as u8 as usize] += 1; |
| 2083 |
} |
| 2084 |
|
| 2085 |
d.lz.consume_flag(); |
| 2086 |
d.lz.total_bytes += cur_match_len; |
| 2087 |
lookahead_pos += cur_match_len as usize; |
| 2088 |
d.dict.size = cmp::min(d.dict.size + cur_match_len as usize, LZ_DICT_SIZE); |
| 2089 |
cur_pos = (cur_pos + cur_match_len as usize) & LZ_DICT_SIZE_MASK; |
| 2090 |
lookahead_size -= cur_match_len as usize; |
| 2091 |
|
| 2092 |
if d.lz.code_position > LZ_CODE_BUF_SIZE - 8 { |
| 2093 |
// These values are used in flush_block, so we need to write them back here. |
| 2094 |
d.dict.lookahead_size = lookahead_size; |
| 2095 |
d.dict.lookahead_pos = lookahead_pos; |
| 2096 |
|
| 2097 |
let n = match flush_block(d, callback, TDEFLFlush::None) { |
| 2098 |
Err(_) => { |
| 2099 |
d.params.src_pos = src_pos; |
| 2100 |
d.params.prev_return_status = TDEFLStatus::PutBufFailed; |
| 2101 |
return false; |
| 2102 |
} |
| 2103 |
Ok(status) => status, |
| 2104 |
}; |
| 2105 |
if n != 0 { |
| 2106 |
d.params.src_pos = src_pos; |
| 2107 |
return n > 0; |
| 2108 |
} |
| 2109 |
debug_assert!(d.lz.code_position < LZ_CODE_BUF_SIZE - 2); |
| 2110 |
|
| 2111 |
lookahead_size = d.dict.lookahead_size; |
| 2112 |
lookahead_pos = d.dict.lookahead_pos; |
| 2113 |
} |
| 2114 |
} |
| 2115 |
} |
| 2116 |
|
| 2117 |
while lookahead_size != 0 { |
| 2118 |
let lit = d.dict.b.dict[cur_pos]; |
| 2119 |
d.lz.total_bytes += 1; |
| 2120 |
d.lz.write_code(lit); |
| 2121 |
*d.lz.get_flag() >>= 1; |
| 2122 |
d.lz.consume_flag(); |
| 2123 |
|
| 2124 |
d.huff.count[0][lit as usize] += 1; |
| 2125 |
lookahead_pos += 1; |
| 2126 |
d.dict.size = cmp::min(d.dict.size + 1, LZ_DICT_SIZE); |
| 2127 |
cur_pos = (cur_pos + 1) & LZ_DICT_SIZE_MASK; |
| 2128 |
lookahead_size -= 1; |
| 2129 |
|
| 2130 |
if d.lz.code_position > LZ_CODE_BUF_SIZE - 8 { |
| 2131 |
// These values are used in flush_block, so we need to write them back here. |
| 2132 |
d.dict.lookahead_size = lookahead_size; |
| 2133 |
d.dict.lookahead_pos = lookahead_pos; |
| 2134 |
|
| 2135 |
let n = match flush_block(d, callback, TDEFLFlush::None) { |
| 2136 |
Err(_) => { |
| 2137 |
d.params.prev_return_status = TDEFLStatus::PutBufFailed; |
| 2138 |
d.params.src_pos = src_pos; |
| 2139 |
return false; |
| 2140 |
} |
| 2141 |
Ok(status) => status, |
| 2142 |
}; |
| 2143 |
if n != 0 { |
| 2144 |
d.params.src_pos = src_pos; |
| 2145 |
return n > 0; |
| 2146 |
} |
| 2147 |
|
| 2148 |
lookahead_size = d.dict.lookahead_size; |
| 2149 |
lookahead_pos = d.dict.lookahead_pos; |
| 2150 |
} |
| 2151 |
} |
| 2152 |
} |
| 2153 |
|
| 2154 |
d.params.src_pos = src_pos; |
| 2155 |
d.dict.lookahead_size = lookahead_size; |
| 2156 |
d.dict.lookahead_pos = lookahead_pos; |
| 2157 |
true |
| 2158 |
} |
| 2159 |
|
| 2160 |
fn flush_output_buffer(c: &mut CallbackOxide, p: &mut ParamsOxide) -> (TDEFLStatus, usize, usize) { |
| 2161 |
let mut res = (TDEFLStatus::Okay, p.src_pos, 0); |
| 2162 |
if let CallbackOut::Buf(ref mut cb) = c.out { |
| 2163 |
let n = cmp::min(cb.out_buf.len() - p.out_buf_ofs, p.flush_remaining as usize); |
| 2164 |
if n != 0 { |
| 2165 |
cb.out_buf[p.out_buf_ofs..p.out_buf_ofs + n] |
| 2166 |
.copy_from_slice(&p.local_buf.b[p.flush_ofs as usize..p.flush_ofs as usize + n]); |
| 2167 |
} |
| 2168 |
p.flush_ofs += n as u32; |
| 2169 |
p.flush_remaining -= n as u32; |
| 2170 |
p.out_buf_ofs += n; |
| 2171 |
res.2 = p.out_buf_ofs; |
| 2172 |
} |
| 2173 |
|
| 2174 |
if p.finished && p.flush_remaining == 0 { |
| 2175 |
res.0 = TDEFLStatus::Done |
| 2176 |
} |
| 2177 |
res |
| 2178 |
} |
| 2179 |
|
| 2180 |
/// Main compression function. Tries to compress as much as possible from `in_buf` and |
| 2181 |
/// puts compressed output into `out_buf`. |
| 2182 |
/// |
| 2183 |
/// The value of `flush` determines if the compressor should attempt to flush all output |
| 2184 |
/// and alternatively try to finish the stream. |
| 2185 |
/// |
| 2186 |
/// Use [`TDEFLFlush::Finish`] on the final call to signal that the stream is finishing. |
| 2187 |
/// |
| 2188 |
/// Note that this function does not keep track of whether a flush marker has been output, so |
| 2189 |
/// if called using [`TDEFLFlush::Sync`], the caller needs to ensure there is enough space in the |
| 2190 |
/// output buffer if they want to avoid repeated flush markers. |
| 2191 |
/// See #105 for details. |
| 2192 |
/// |
| 2193 |
/// # Returns |
| 2194 |
/// Returns a tuple containing the current status of the compressor, the current position |
| 2195 |
/// in the input buffer and the current position in the output buffer. |
| 2196 |
pub fn compress( |
| 2197 |
d: &mut CompressorOxide, |
| 2198 |
in_buf: &[u8], |
| 2199 |
out_buf: &mut [u8], |
| 2200 |
flush: TDEFLFlush, |
| 2201 |
) -> (TDEFLStatus, usize, usize) { |
| 2202 |
compress_inner( |
| 2203 |
d, |
| 2204 |
&mut CallbackOxide::new_callback_buf(in_buf, out_buf), |
| 2205 |
flush, |
| 2206 |
) |
| 2207 |
} |
| 2208 |
|
| 2209 |
/// Main compression function. Callbacks output. |
| 2210 |
/// |
| 2211 |
/// # Returns |
| 2212 |
/// Returns a tuple containing the current status of the compressor, the current position |
| 2213 |
/// in the input buffer. |
| 2214 |
/// |
| 2215 |
/// The caller is responsible for ensuring the `CallbackFunc` struct will not cause undefined |
| 2216 |
/// behaviour. |
| 2217 |
pub fn compress_to_output( |
| 2218 |
d: &mut CompressorOxide, |
| 2219 |
in_buf: &[u8], |
| 2220 |
flush: TDEFLFlush, |
| 2221 |
mut callback_func: impl FnMut(&[u8]) -> bool, |
| 2222 |
) -> (TDEFLStatus, usize) { |
| 2223 |
let res = compress_inner( |
| 2224 |
d, |
| 2225 |
&mut CallbackOxide::new_callback_func( |
| 2226 |
in_buf, |
| 2227 |
CallbackFunc { |
| 2228 |
put_buf_func: &mut callback_func, |
| 2229 |
}, |
| 2230 |
), |
| 2231 |
flush, |
| 2232 |
); |
| 2233 |
|
| 2234 |
(res.0, res.1) |
| 2235 |
} |
| 2236 |
|
| 2237 |
fn compress_inner( |
| 2238 |
d: &mut CompressorOxide, |
| 2239 |
callback: &mut CallbackOxide, |
| 2240 |
flush: TDEFLFlush, |
| 2241 |
) -> (TDEFLStatus, usize, usize) { |
| 2242 |
d.params.out_buf_ofs = 0; |
| 2243 |
d.params.src_pos = 0; |
| 2244 |
|
| 2245 |
let prev_ok = d.params.prev_return_status == TDEFLStatus::Okay; |
| 2246 |
let flush_finish_once = d.params.flush != TDEFLFlush::Finish || flush == TDEFLFlush::Finish; |
| 2247 |
|
| 2248 |
d.params.flush = flush; |
| 2249 |
if !prev_ok || !flush_finish_once { |
| 2250 |
d.params.prev_return_status = TDEFLStatus::BadParam; |
| 2251 |
return (d.params.prev_return_status, 0, 0); |
| 2252 |
} |
| 2253 |
|
| 2254 |
if d.params.flush_remaining != 0 || d.params.finished { |
| 2255 |
let res = flush_output_buffer(callback, &mut d.params); |
| 2256 |
d.params.prev_return_status = res.0; |
| 2257 |
return res; |
| 2258 |
} |
| 2259 |
|
| 2260 |
let one_probe = d.params.flags & MAX_PROBES_MASK == 1; |
| 2261 |
let greedy = d.params.flags & TDEFL_GREEDY_PARSING_FLAG != 0; |
| 2262 |
let filter_or_rle = d.params.flags & (TDEFL_FILTER_MATCHES | TDEFL_FORCE_ALL_RAW_BLOCKS) != 0; |
| 2263 |
|
| 2264 |
let raw = d.params.flags & TDEFL_FORCE_ALL_RAW_BLOCKS != 0; |
| 2265 |
|
| 2266 |
let compress_success = if raw { |
| 2267 |
compress_stored(d, callback) |
| 2268 |
} else if one_probe && greedy && !filter_or_rle { |
| 2269 |
compress_fast(d, callback) |
| 2270 |
} else { |
| 2271 |
compress_normal(d, callback) |
| 2272 |
}; |
| 2273 |
|
| 2274 |
if !compress_success { |
| 2275 |
return ( |
| 2276 |
d.params.prev_return_status, |
| 2277 |
d.params.src_pos, |
| 2278 |
d.params.out_buf_ofs, |
| 2279 |
); |
| 2280 |
} |
| 2281 |
|
| 2282 |
if let Some(in_buf) = callback.in_buf { |
| 2283 |
if d.params.flags & (TDEFL_WRITE_ZLIB_HEADER | TDEFL_COMPUTE_ADLER32) != 0 { |
| 2284 |
d.params.adler32 = update_adler32(d.params.adler32, &in_buf[..d.params.src_pos]); |
| 2285 |
} |
| 2286 |
} |
| 2287 |
|
| 2288 |
let flush_none = d.params.flush == TDEFLFlush::None; |
| 2289 |
let in_left = callback.in_buf.map_or(0, |buf| buf.len()) - d.params.src_pos; |
| 2290 |
let remaining = in_left != 0 || d.params.flush_remaining != 0; |
| 2291 |
if !flush_none && d.dict.lookahead_size == 0 && !remaining { |
| 2292 |
let flush = d.params.flush; |
| 2293 |
match flush_block(d, callback, flush) { |
| 2294 |
Err(_) => { |
| 2295 |
d.params.prev_return_status = TDEFLStatus::PutBufFailed; |
| 2296 |
return ( |
| 2297 |
d.params.prev_return_status, |
| 2298 |
d.params.src_pos, |
| 2299 |
d.params.out_buf_ofs, |
| 2300 |
); |
| 2301 |
} |
| 2302 |
Ok(x) if x < 0 => { |
| 2303 |
return ( |
| 2304 |
d.params.prev_return_status, |
| 2305 |
d.params.src_pos, |
| 2306 |
d.params.out_buf_ofs, |
| 2307 |
) |
| 2308 |
} |
| 2309 |
_ => { |
| 2310 |
d.params.finished = d.params.flush == TDEFLFlush::Finish; |
| 2311 |
if d.params.flush == TDEFLFlush::Full { |
| 2312 |
d.dict.b.hash.fill(0); |
| 2313 |
d.dict.b.next.fill(0); |
| 2314 |
d.dict.size = 0; |
| 2315 |
} |
| 2316 |
} |
| 2317 |
} |
| 2318 |
} |
| 2319 |
|
| 2320 |
let res = flush_output_buffer(callback, &mut d.params); |
| 2321 |
d.params.prev_return_status = res.0; |
| 2322 |
|
| 2323 |
res |
| 2324 |
} |
| 2325 |
|
| 2326 |
/// Create a set of compression flags using parameters used by zlib and other compressors. |
| 2327 |
/// Mainly intended for use with transition from c libraries as it deals with raw integers. |
| 2328 |
/// |
| 2329 |
/// # Parameters |
| 2330 |
/// `level` determines compression level. Clamped to maximum of 10. Negative values result in |
| 2331 |
/// `CompressionLevel::DefaultLevel`. |
| 2332 |
/// `window_bits`: Above 0, wraps the stream in a zlib wrapper, 0 or negative for a raw deflate |
| 2333 |
/// stream. |
| 2334 |
/// `strategy`: Sets the strategy if this conforms to any of the values in `CompressionStrategy`. |
| 2335 |
/// |
| 2336 |
/// # Notes |
| 2337 |
/// This function may be removed or moved to the `miniz_oxide_c_api` in the future. |
| 2338 |
pub fn create_comp_flags_from_zip_params(level: i32, window_bits: i32, strategy: i32) -> u32 { |
| 2339 |
let num_probes = (if level >= 0 { |
| 2340 |
cmp::min(10, level) |
| 2341 |
} else { |
| 2342 |
CompressionLevel::DefaultLevel as i32 |
| 2343 |
}) as usize; |
| 2344 |
let greedy = if level <= 3 { |
| 2345 |
TDEFL_GREEDY_PARSING_FLAG |
| 2346 |
} else { |
| 2347 |
0 |
| 2348 |
}; |
| 2349 |
let mut comp_flags = u32::from(NUM_PROBES[num_probes]) | greedy; |
| 2350 |
|
| 2351 |
if window_bits > 0 { |
| 2352 |
comp_flags |= TDEFL_WRITE_ZLIB_HEADER; |
| 2353 |
} |
| 2354 |
|
| 2355 |
if level == 0 { |
| 2356 |
comp_flags |= TDEFL_FORCE_ALL_RAW_BLOCKS; |
| 2357 |
} else if strategy == CompressionStrategy::Filtered as i32 { |
| 2358 |
comp_flags |= TDEFL_FILTER_MATCHES; |
| 2359 |
} else if strategy == CompressionStrategy::HuffmanOnly as i32 { |
| 2360 |
comp_flags &= !MAX_PROBES_MASK; |
| 2361 |
} else if strategy == CompressionStrategy::Fixed as i32 { |
| 2362 |
comp_flags |= TDEFL_FORCE_ALL_STATIC_BLOCKS; |
| 2363 |
} else if strategy == CompressionStrategy::RLE as i32 { |
| 2364 |
comp_flags |= TDEFL_RLE_MATCHES; |
| 2365 |
} |
| 2366 |
|
| 2367 |
comp_flags |
| 2368 |
} |
| 2369 |
|
| 2370 |
#[cfg(test)] |
| 2371 |
mod test { |
| 2372 |
use super::{ |
| 2373 |
compress_to_output, create_comp_flags_from_zip_params, read_u16_le, write_u16_le, |
| 2374 |
CompressionStrategy, CompressorOxide, TDEFLFlush, TDEFLStatus, DEFAULT_FLAGS, |
| 2375 |
MZ_DEFAULT_WINDOW_BITS, |
| 2376 |
}; |
| 2377 |
use crate::inflate::decompress_to_vec; |
| 2378 |
use alloc::vec; |
| 2379 |
|
| 2380 |
#[test] |
| 2381 |
fn u16_to_slice() { |
| 2382 |
let mut slice = [0, 0]; |
| 2383 |
write_u16_le(2000, &mut slice, 0); |
| 2384 |
assert_eq!(slice, [208, 7]); |
| 2385 |
} |
| 2386 |
|
| 2387 |
#[test] |
| 2388 |
fn u16_from_slice() { |
| 2389 |
let slice = [208, 7]; |
| 2390 |
assert_eq!(read_u16_le(&slice, 0), 2000); |
| 2391 |
} |
| 2392 |
|
| 2393 |
#[test] |
| 2394 |
fn compress_output() { |
| 2395 |
assert_eq!( |
| 2396 |
DEFAULT_FLAGS, |
| 2397 |
create_comp_flags_from_zip_params( |
| 2398 |
4, |
| 2399 |
MZ_DEFAULT_WINDOW_BITS, |
| 2400 |
CompressionStrategy::Default as i32 |
| 2401 |
) |
| 2402 |
); |
| 2403 |
|
| 2404 |
let slice = [ |
| 2405 |
1, 2, 3, 4, 1, 2, 3, 1, 2, 3, 1, 2, 6, 1, 2, 3, 1, 2, 3, 2, 3, 1, 2, 3, |
| 2406 |
]; |
| 2407 |
let mut encoded = vec![]; |
| 2408 |
let flags = create_comp_flags_from_zip_params(6, 0, 0); |
| 2409 |
let mut d = CompressorOxide::new(flags); |
| 2410 |
let (status, in_consumed) = |
| 2411 |
compress_to_output(&mut d, &slice, TDEFLFlush::Finish, |out: &[u8]| { |
| 2412 |
encoded.extend_from_slice(out); |
| 2413 |
true |
| 2414 |
}); |
| 2415 |
|
| 2416 |
assert_eq!(status, TDEFLStatus::Done); |
| 2417 |
assert_eq!(in_consumed, slice.len()); |
| 2418 |
|
| 2419 |
let decoded = decompress_to_vec(&encoded[..]).unwrap(); |
| 2420 |
assert_eq!(&decoded[..], &slice[..]); |
| 2421 |
} |
| 2422 |
|
| 2423 |
#[test] |
| 2424 |
/// Check fast compress mode |
| 2425 |
fn compress_fast() { |
| 2426 |
let slice = [ |
| 2427 |
1, 2, 3, 4, 1, 2, 3, 1, 2, 3, 1, 2, 6, 1, 2, 3, 1, 2, 3, 2, 3, 1, 2, 3, |
| 2428 |
]; |
| 2429 |
let mut encoded = vec![]; |
| 2430 |
let flags = create_comp_flags_from_zip_params(1, 0, 0); |
| 2431 |
let mut d = CompressorOxide::new(flags); |
| 2432 |
let (status, in_consumed) = |
| 2433 |
compress_to_output(&mut d, &slice, TDEFLFlush::Finish, |out: &[u8]| { |
| 2434 |
encoded.extend_from_slice(out); |
| 2435 |
true |
| 2436 |
}); |
| 2437 |
|
| 2438 |
assert_eq!(status, TDEFLStatus::Done); |
| 2439 |
assert_eq!(in_consumed, slice.len()); |
| 2440 |
|
| 2441 |
// Needs to be altered if algorithm improves. |
| 2442 |
assert_eq!( |
| 2443 |
&encoded[..], |
| 2444 |
[99, 100, 98, 102, 1, 98, 48, 98, 3, 147, 204, 76, 204, 140, 76, 204, 0] |
| 2445 |
); |
| 2446 |
|
| 2447 |
let decoded = decompress_to_vec(&encoded[..]).unwrap(); |
| 2448 |
assert_eq!(&decoded[..], &slice[..]); |
| 2449 |
} |
| 2450 |
|
| 2451 |
#[test] |
| 2452 |
fn zlib_window_bits() { |
| 2453 |
use crate::inflate::stream::{inflate, InflateState}; |
| 2454 |
use crate::DataFormat; |
| 2455 |
use alloc::boxed::Box; |
| 2456 |
let slice = [ |
| 2457 |
1, 2, 3, 4, 1, 2, 3, 1, 2, 3, 1, 2, 6, 1, 2, 3, 1, 2, 3, 2, 3, 1, 2, 3, 35, 22, 22, 2, |
| 2458 |
6, 2, 6, |
| 2459 |
]; |
| 2460 |
let mut encoded = vec![]; |
| 2461 |
let flags = create_comp_flags_from_zip_params(2, 1, CompressionStrategy::RLE.into()); |
| 2462 |
let mut d = CompressorOxide::new(flags); |
| 2463 |
let (status, in_consumed) = |
| 2464 |
compress_to_output(&mut d, &slice, TDEFLFlush::Finish, |out: &[u8]| { |
| 2465 |
encoded.extend_from_slice(out); |
| 2466 |
true |
| 2467 |
}); |
| 2468 |
|
| 2469 |
assert_eq!(status, TDEFLStatus::Done); |
| 2470 |
assert_eq!(in_consumed, slice.len()); |
| 2471 |
|
| 2472 |
let mut output = vec![0; slice.len()]; |
| 2473 |
|
| 2474 |
let mut decompressor = Box::new(InflateState::new(DataFormat::Zlib)); |
| 2475 |
|
| 2476 |
let mut out_slice = output.as_mut_slice(); |
| 2477 |
// Feed 1 byte at a time and no back buffer to test that RLE encoding has been used. |
| 2478 |
for i in 0..encoded.len() { |
| 2479 |
let result = inflate( |
| 2480 |
&mut decompressor, |
| 2481 |
&encoded[i..i + 1], |
| 2482 |
out_slice, |
| 2483 |
crate::MZFlush::None, |
| 2484 |
); |
| 2485 |
out_slice = &mut out_slice[result.bytes_written..]; |
| 2486 |
} |
| 2487 |
let cmf = decompressor.decompressor().zlib_header().0; |
| 2488 |
assert_eq!(cmf, 8); |
| 2489 |
assert_eq!(output, slice) |
| 2490 |
} |
| 2491 |
} |