2 * Copyright 2016 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.
23 #include <folly/io/IOBuf.h>
26 * Compression / decompression over IOBufs
29 namespace folly { namespace io {
31 enum class CodecType {
33 * This codec type is not defined; getCodec() will throw an exception
34 * if used. Useful if deriving your own classes from Codec without
35 * going through the getCodec() interface.
46 * Use LZ4 compression.
47 * Levels supported: 1 = fast, 2 = best; default = 1
52 * Use Snappy compression.
58 * Use zlib compression.
59 * Levels supported: 0 = no compression, 1 = fast, ..., 9 = best; default = 6
64 * Use LZ4 compression, prefixed with size (as Varint).
69 * Use LZMA2 compression.
70 * Levels supported: 0 = no compression, 1 = fast, ..., 9 = best; default = 6
73 LZMA2_VARINT_SIZE = 7,
76 * Use ZSTD_BETA compression.
77 * This format is not yet final; please do not rely on it for anything other
78 * than testing purposes yet.
83 * Use gzip compression. This is the same compression algorithm as ZLIB but
84 * gzip-compressed files tend to be easier to work with from the command line.
85 * Levels supported: 0 = no compression, 1 = fast, ..., 9 = best; default = 6
97 * Return the maximum length of data that may be compressed with this codec.
98 * NO_COMPRESSION and ZLIB support arbitrary lengths;
99 * LZ4 supports up to 1.9GiB; SNAPPY supports up to 4GiB.
100 * May return UNLIMITED_UNCOMPRESSED_LENGTH if unlimited.
102 uint64_t maxUncompressedLength() const;
105 * Return the codec's type.
107 CodecType type() const { return type_; }
110 * Does this codec need the exact uncompressed length on decompression?
112 bool needsUncompressedLength() const;
115 * Compress data, returning an IOBuf (which may share storage with data).
116 * Throws std::invalid_argument if data is larger than
117 * maxUncompressedLength().
119 * Regardless of the behavior of the underlying compressor, compressing
120 * an empty IOBuf chain will return an empty IOBuf chain.
122 std::unique_ptr<IOBuf> compress(const folly::IOBuf* data);
125 * Uncompress data. Throws std::runtime_error on decompression error.
127 * Some codecs (LZ4) require the exact uncompressed length; this is indicated
128 * by needsUncompressedLength().
130 * For other codes (zlib), knowing the exact uncompressed length ahead of
131 * time might be faster.
133 * Regardless of the behavior of the underlying compressor, uncompressing
134 * an empty IOBuf chain will return an empty IOBuf chain.
136 static constexpr uint64_t UNKNOWN_UNCOMPRESSED_LENGTH = uint64_t(-1);
137 static constexpr uint64_t UNLIMITED_UNCOMPRESSED_LENGTH = uint64_t(-2);
139 std::unique_ptr<IOBuf> uncompress(
141 uint64_t uncompressedLength = UNKNOWN_UNCOMPRESSED_LENGTH);
144 explicit Codec(CodecType type);
147 // default: no limits (save for special value UNKNOWN_UNCOMPRESSED_LENGTH)
148 virtual uint64_t doMaxUncompressedLength() const;
149 // default: doesn't need uncompressed length
150 virtual bool doNeedsUncompressedLength() const;
151 virtual std::unique_ptr<IOBuf> doCompress(const folly::IOBuf* data) = 0;
152 virtual std::unique_ptr<IOBuf> doUncompress(const folly::IOBuf* data,
153 uint64_t uncompressedLength) = 0;
158 constexpr int COMPRESSION_LEVEL_FASTEST = -1;
159 constexpr int COMPRESSION_LEVEL_DEFAULT = -2;
160 constexpr int COMPRESSION_LEVEL_BEST = -3;
163 * Return a codec for the given type. Throws on error. The level
164 * is a non-negative codec-dependent integer indicating the level of
165 * compression desired, or one of the following constants:
167 * COMPRESSION_LEVEL_FASTEST is fastest (uses least CPU / memory,
169 * COMPRESSION_LEVEL_DEFAULT is the default (likely a tradeoff between
171 * COMPRESSION_LEVEL_BEST is the best compression (uses most CPU / memory,
174 * When decompressing, the compression level is ignored. All codecs will
175 * decompress all data compressed with the a codec of the same type, regardless
176 * of compression level.
178 std::unique_ptr<Codec> getCodec(CodecType type,
179 int level = COMPRESSION_LEVEL_DEFAULT);