2 * Copyright 2015 Facebook, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 #ifndef FOLLY_IO_COMPRESSION_H_
18 #define FOLLY_IO_COMPRESSION_H_
24 #include <folly/io/IOBuf.h>
27 * Compression / decompression over IOBufs
30 namespace folly { namespace io {
32 enum class CodecType {
34 * This codec type is not defined; getCodec() will throw an exception
35 * if used. Useful if deriving your own classes from Codec without
36 * going through the getCodec() interface.
47 * Use LZ4 compression.
48 * Levels supported: 1 = fast, 2 = best; default = 1
53 * Use Snappy compression.
59 * Use zlib compression.
60 * Levels supported: 0 = no compression, 1 = fast, ..., 9 = best; default = 6
65 * Use LZ4 compression, prefixed with size (as Varint).
70 * Use LZMA2 compression.
71 * Levels supported: 0 = no compression, 1 = fast, ..., 9 = best; default = 6
74 LZMA2_VARINT_SIZE = 7,
77 * Use ZSTD_BETA compression.
78 * This format is not yet final; please do not rely on it for anything other
79 * than testing purposes yet.
91 * Return the maximum length of data that may be compressed with this codec.
92 * NO_COMPRESSION and ZLIB support arbitrary lengths;
93 * LZ4 supports up to 1.9GiB; SNAPPY supports up to 4GiB.
94 * May return UNLIMITED_UNCOMPRESSED_LENGTH if unlimited.
96 uint64_t maxUncompressedLength() const;
99 * Return the codec's type.
101 CodecType type() const { return type_; }
104 * Does this codec need the exact uncompressed length on decompression?
106 bool needsUncompressedLength() const;
109 * Compress data, returning an IOBuf (which may share storage with data).
110 * Throws std::invalid_argument if data is larger than
111 * maxUncompressedLength().
113 * Regardless of the behavior of the underlying compressor, compressing
114 * an empty IOBuf chain will return an empty IOBuf chain.
116 std::unique_ptr<IOBuf> compress(const folly::IOBuf* data);
119 * Uncompress data. Throws std::runtime_error on decompression error.
121 * Some codecs (LZ4) require the exact uncompressed length; this is indicated
122 * by needsUncompressedLength().
124 * For other codes (zlib), knowing the exact uncompressed length ahead of
125 * time might be faster.
127 * Regardless of the behavior of the underlying compressor, uncompressing
128 * an empty IOBuf chain will return an empty IOBuf chain.
130 static constexpr uint64_t UNKNOWN_UNCOMPRESSED_LENGTH = uint64_t(-1);
131 static constexpr uint64_t UNLIMITED_UNCOMPRESSED_LENGTH = uint64_t(-2);
133 std::unique_ptr<IOBuf> uncompress(
135 uint64_t uncompressedLength = UNKNOWN_UNCOMPRESSED_LENGTH);
138 explicit Codec(CodecType type);
141 // default: no limits (save for special value UNKNOWN_UNCOMPRESSED_LENGTH)
142 virtual uint64_t doMaxUncompressedLength() const;
143 // default: doesn't need uncompressed length
144 virtual bool doNeedsUncompressedLength() const;
145 virtual std::unique_ptr<IOBuf> doCompress(const folly::IOBuf* data) = 0;
146 virtual std::unique_ptr<IOBuf> doUncompress(const folly::IOBuf* data,
147 uint64_t uncompressedLength) = 0;
152 constexpr int COMPRESSION_LEVEL_FASTEST = -1;
153 constexpr int COMPRESSION_LEVEL_DEFAULT = -2;
154 constexpr int COMPRESSION_LEVEL_BEST = -3;
157 * Return a codec for the given type. Throws on error. The level
158 * is a non-negative codec-dependent integer indicating the level of
159 * compression desired, or one of the following constants:
161 * COMPRESSION_LEVEL_FASTEST is fastest (uses least CPU / memory,
163 * COMPRESSION_LEVEL_DEFAULT is the default (likely a tradeoff between
165 * COMPRESSION_LEVEL_BEST is the best compression (uses most CPU / memory,
168 * When decompressing, the compression level is ignored. All codecs will
169 * decompress all data compressed with the a codec of the same type, regardless
170 * of compression level.
172 std::unique_ptr<Codec> getCodec(CodecType type,
173 int level = COMPRESSION_LEVEL_DEFAULT);
177 #endif /* FOLLY_IO_COMPRESSION_H_ */