This package provides various compression algorithms.
- zstandard compression and decompression in pure Go.
- S2 is a high performance replacement for Snappy.
- Optimized deflate packages which can be used as a dropin replacement for gzip, zip and zlib.
- snappy is a drop-in replacement for
github.com/golang/snappyoffering better compression and concurrent streams. - huff0 and FSE implementations for raw entropy encoding.
- gzhttp Provides client and server wrappers for handling gzipped requests efficiently.
- pgzip is a separate package that provides a very fast parallel gzip implementation.
- fuzz package for fuzz testing all compressors/decompressors here.
-
July 21, 2022 (v1.15.9)
- zstd: Fix decoder crash on amd64 (no BMI) on invalid input klauspost#645
- zstd: Disable decoder extended memory copies (amd64) due to possible crashes klauspost#644
- zstd: Allow single segments up to "max decoded size" by @klauspost in klauspost#643
-
July 13, 2022 (v1.15.8)
- gzip: fix stack exhaustion bug in Reader.Read klauspost#641
- s2: Add Index header trim/restore klauspost#638
- zstd: Optimize seqdeq amd64 asm by @greatroar in klauspost#636
- zstd: Improve decoder memcopy klauspost#637
- huff0: Pass a single bitReader pointer to asm by @greatroar in klauspost#634
- zstd: Branchless getBits for amd64 w/o BMI2 by @greatroar in klauspost#640
- gzhttp: Remove header before writing klauspost#639
-
June 29, 2022 (v1.15.7)
- s2: Fix absolute forward seeks klauspost#633
- zip: Merge upstream klauspost#631
- zip: Re-add zip64 fix klauspost#624
- zstd: translate fseDecoder.buildDtable into asm by @WojciechMula in klauspost#598
- flate: Faster histograms klauspost#620
- deflate: Use compound hcode klauspost#622
-
June 3, 2022 (v1.15.6)
- s2: Improve coding for long, close matches klauspost#613
- s2c: Add Snappy/S2 stream recompression klauspost#611
- zstd: Always use configured block size klauspost#605
- zstd: Fix incorrect hash table placement for dict encoding in default klauspost#606
- zstd: Apply default config to ZipDecompressor without options klauspost#608
- gzhttp: Exclude more common archive formats klauspost#612
- s2: Add ReaderIgnoreCRC klauspost#609
- s2: Remove sanity load on index creation klauspost#607
- snappy: Use dedicated function for scoring klauspost#614
- s2c+s2d: Use official snappy framed extension klauspost#610
-
May 25, 2022 (v1.15.5)
- s2: Add concurrent stream decompression klauspost#602
- s2: Fix final emit oob read crash on amd64 klauspost#601
- huff0: asm implementation of Decompress1X by @WojciechMula klauspost#596
- zstd: Use 1 less goroutine for stream decoding klauspost#588
- zstd: Copy literal in 16 byte blocks when possible klauspost#592
- zstd: Speed up when WithDecoderLowmem(false) klauspost#599
- zstd: faster next state update in BMI2 version of decode by @WojciechMula in klauspost#593
- huff0: Do not check max size when reading table. klauspost#586
- flate: Inplace hashing for level 7-9 by @klauspost in klauspost#590
-
May 11, 2022 (v1.15.4)
-
May 5, 2022 (v1.15.3)
-
Apr 26, 2022 (v1.15.2)
-
Mar 11, 2022 (v1.15.1)
-
Mar 3, 2022 (v1.15.0)
- zstd: Refactor decoder by @klauspost in #498
- zstd: Add stream encoding without goroutines by @klauspost in #505
- huff0: Prevent single blocks exceeding 16 bits by @klauspost in#507
- flate: Inline literal emission by @klauspost in #509
- gzhttp: Add zstd to transport by @klauspost in #400
- gzhttp: Make content-type optional by @klauspost in #510
See Details
Both compression and decompression now supports "synchronous" stream operations. This means that whenever "concurrency" is set to 1, they will operate without spawning goroutines.Stream decompression is now faster on asynchronous, since the goroutine allocation much more effectively splits the workload. On typical streams this will typically use 2 cores fully for decompression. When a stream has finished decoding no goroutines will be left over, so decoders can now safely be pooled and still be garbage collected.
While the release has been extensively tested, it is recommended to testing when upgrading.
-
Feb 22, 2022 (v1.14.4)
-
Feb 17, 2022 (v1.14.3)
-
Jan 25, 2022 (v1.14.2)
-
Jan 11, 2022 (v1.14.1)
See changes to v1.13.x
-
Aug 30, 2021 (v1.13.5)
-
Aug 12, 2021 (v1.13.4)
- Add snappy replacement package.
- zstd: Fix incorrect encoding in "best" mode #415
-
Aug 3, 2021 (v1.13.3)
-
Jun 14, 2021 (v1.13.1)
- s2: Add full Snappy output support #396
- zstd: Add configurable Decoder window size #394
- gzhttp: Add header to skip compression #389
- s2: Improve speed with bigger output margin #395
-
Jun 3, 2021 (v1.13.0)
See changes to v1.12.x
-
May 25, 2021 (v1.12.3)
-
Apr 27, 2021 (v1.12.2)
-
Apr 14, 2021 (v1.12.1)
- snappy package removed. Upstream added as dependency.
- s2: Better compression in "best" mode #353
- s2sx: Add stdin input and detect pre-compressed from signature #352
- s2c/s2d: Add http as possible input #348
- s2c/s2d/s2sx: Always truncate when writing files #352
- zstd: Reduce memory usage further when using WithLowerEncoderMem #346
- s2: Fix potential problem with amd64 assembly and profilers #349
See changes to v1.11.x
-
Mar 26, 2021 (v1.11.13)
-
Mar 5, 2021 (v1.11.12)
- s2: Add
s2sxbinary that creates self extracting archives. - s2: Speed up decompression on non-assembly platforms #328
- s2: Add
-
Mar 1, 2021 (v1.11.9)
-
Feb 25, 2021 (v1.11.8)
-
Jan 14, 2021 (v1.11.7)
-
Jan 7, 2021 (v1.11.6)
-
Dec 20, 2020 (v1.11.4)
-
Nov 15, 2020 (v1.11.3)
-
Oct 11, 2020 (v1.11.2)
- s2: Fix out of bounds read in "better" block compression #291
-
Oct 1, 2020 (v1.11.1)
- zstd: Set allLitEntropy true in default configuration #286
-
Sept 8, 2020 (v1.11.0)
- zstd: Add experimental compression dictionaries #281
- zstd: Fix mixed Write and ReadFrom calls #282
- inflate/gz: Limit variable shifts, ~5% faster decompression #274
See changes to v1.10.x
-
July 8, 2020 (v1.10.11)
-
June 23, 2020 (v1.10.10)
- zstd: Skip entropy compression in fastest mode when no matches. #270
-
June 16, 2020 (v1.10.9):
-
June 5, 2020 (v1.10.8):
- 1.15x faster zstd block decompression. #265
-
June 1, 2020 (v1.10.7):
- Added zstd decompression dictionary support
- Increase zstd decompression speed up to 1.19x. #259
- Remove internal reset call in zstd compression and reduce allocations. #263
-
May 21, 2020: (v1.10.6)
-
April 12, 2020: (v1.10.5)
- s2-commands: Flush output when receiving SIGINT. #239
-
Apr 8, 2020: (v1.10.4)
-
Mar 11, 2020: (v1.10.3)
-
Feb 27, 2020: (v1.10.2)
-
Feb 18, 2020: (v1.10.1)
-
Feb 4, 2020: (v1.10.0)
- Add optional dictionary to stateless deflate. Breaking change, send
nilfor previous behaviour. #216 - Fix buffer overflow on repeated small block deflate. #218
- Allow copying content from an existing ZIP file without decompressing+compressing. #214
- Added S2 AMD64 assembler and various optimizations. Stream speed >10GB/s. #186
- Add optional dictionary to stateless deflate. Breaking change, send
See changes prior to v1.10.0
- Jan 20,2020 (v1.9.8) Optimize gzip/deflate with better size estimates and faster table generation. #207 by luyu6056, #206.
- Jan 11, 2020: S2 Encode/Decode will use provided buffer if capacity is big enough. #204
- Jan 5, 2020: (v1.9.7) Fix another zstd regression in v1.9.5 - v1.9.6 removed.
- Jan 4, 2020: (v1.9.6) Regression in v1.9.5 fixed causing corrupt zstd encodes in rare cases.
- Jan 4, 2020: Faster IO in s2c + s2d commandline tools compression/decompression. #192
- Dec 29, 2019: Removed v1.9.5 since fuzz tests showed a compatibility problem with the reference zstandard decoder.
- Dec 29, 2019: (v1.9.5) zstd: 10-20% faster block compression. #199
- Dec 29, 2019: zip package updated with latest Go features
- Dec 29, 2019: zstd: Single segment flag condintions tweaked. #197
- Dec 18, 2019: s2: Faster compression when ReadFrom is used. #198
- Dec 10, 2019: s2: Fix repeat length output when just above at 16MB limit.
- Dec 10, 2019: zstd: Add function to get decoder as io.ReadCloser. #191
- Dec 3, 2019: (v1.9.4) S2: limit max repeat length. #188
- Dec 3, 2019: Add WithNoEntropyCompression to zstd #187
- Dec 3, 2019: Reduce memory use for tests. Check for leaked goroutines.
- Nov 28, 2019 (v1.9.3) Less allocations in stateless deflate.
- Nov 28, 2019: 5-20% Faster huff0 decode. Impacts zstd as well. #184
- Nov 12, 2019 (v1.9.2) Added Stateless Compression for gzip/deflate.
- Nov 12, 2019: Fixed zstd decompression of large single blocks. #180
- Nov 11, 2019: Set default s2c block size to 4MB.
- Nov 11, 2019: Reduce inflate memory use by 1KB.
- Nov 10, 2019: Less allocations in deflate bit writer.
- Nov 10, 2019: Fix inconsistent error returned by zstd decoder.
- Oct 28, 2019 (v1.9.1) ztsd: Fix crash when compressing blocks. #174
- Oct 24, 2019 (v1.9.0) zstd: Fix rare data corruption #173
- Oct 24, 2019 zstd: Fix huff0 out of buffer write #171 and always return errors #172
- Oct 10, 2019: Big deflate rewrite, 30-40% faster with better compression #105
See changes prior to v1.9.0
- Oct 10, 2019: (v1.8.6) zstd: Allow partial reads to get flushed data. #169
- Oct 3, 2019: Fix inconsistent results on broken zstd streams.
- Sep 25, 2019: Added
-rm(remove source files) and-q(no output except errors) tos2cands2dcommands - Sep 16, 2019: (v1.8.4) Add
s2cands2dcommandline tools. - Sep 10, 2019: (v1.8.3) Fix s2 decoder Skip.
- Sep 7, 2019: zstd: Added WithWindowSize, contributed by ianwilkes.
- Sep 5, 2019: (v1.8.2) Add WithZeroFrames which adds full zero payload block encoding option.
- Sep 5, 2019: Lazy initialization of zstandard predefined en/decoder tables.
- Aug 26, 2019: (v1.8.1) S2: 1-2% compression increase in "better" compression mode.
- Aug 26, 2019: zstd: Check maximum size of Huffman 1X compressed literals while decoding.
- Aug 24, 2019: (v1.8.0) Added S2 compression, a high performance replacement for Snappy.
- Aug 21, 2019: (v1.7.6) Fixed minor issues found by fuzzer. One could lead to zstd not decompressing.
- Aug 18, 2019: Add fuzzit continuous fuzzing.
- Aug 14, 2019: zstd: Skip incompressible data 2x faster. #147
- Aug 4, 2019 (v1.7.5): Better literal compression. #146
- Aug 4, 2019: Faster zstd compression. #143 #144
- Aug 4, 2019: Faster zstd decompression. #145 #143 #142
- July 15, 2019 (v1.7.4): Fix double EOF block in rare cases on zstd encoder.
- July 15, 2019 (v1.7.3): Minor speedup/compression increase in default zstd encoder.
- July 14, 2019: zstd decoder: Fix decompression error on multiple uses with mixed content.
- July 7, 2019 (v1.7.2): Snappy update, zstd decoder potential race fix.
- June 17, 2019: zstd decompression bugfix.
- June 17, 2019: fix 32 bit builds.
- June 17, 2019: Easier use in modules (less dependencies).
- June 9, 2019: New stronger "default" zstd compression mode. Matches zstd default compression ratio.
- June 5, 2019: 20-40% throughput in zstandard compression and better compression.
- June 5, 2019: deflate/gzip compression: Reduce memory usage of lower compression levels.
- June 2, 2019: Added zstandard compression!
- May 25, 2019: deflate/gzip: 10% faster bit writer, mostly visible in lower levels.
- Apr 22, 2019: zstd decompression added.
- Aug 1, 2018: Added huff0 README.
- Jul 8, 2018: Added Performance Update 2018 below.
- Jun 23, 2018: Merged Go 1.11 inflate optimizations. Go 1.9 is now required. Backwards compatible version tagged with v1.3.0.
- Apr 2, 2018: Added huff0 en/decoder. Experimental for now, API may change.
- Mar 4, 2018: Added FSE Entropy en/decoder. Experimental for now, API may change.
- Nov 3, 2017: Add compression Estimate function.
- May 28, 2017: Reduce allocations when resetting decoder.
- Apr 02, 2017: Change back to official crc32, since changes were merged in Go 1.7.
- Jan 14, 2017: Reduce stack pressure due to array copies. See Issue #18625.
- Oct 25, 2016: Level 2-4 have been rewritten and now offers significantly better performance than before.
- Oct 20, 2016: Port zlib changes from Go 1.7 to fix zlib writer issue. Please update.
- Oct 16, 2016: Go 1.7 changes merged. Apples to apples this package is a few percent faster, but has a significantly better balance between speed and compression per level.
- Mar 24, 2016: Always attempt Huffman encoding on level 4-7. This improves base 64 encoded data compression.
- Mar 24, 2016: Small speedup for level 1-3.
- Feb 19, 2016: Faster bit writer, level -2 is 15% faster, level 1 is 4% faster.
- Feb 19, 2016: Handle small payloads faster in level 1-3.
- Feb 19, 2016: Added faster level 2 + 3 compression modes.
- Feb 19, 2016: Rebalanced compression levels, so there is a more even progresssion in terms of compression. New default level is 5.
- Feb 14, 2016: Snappy: Merge upstream changes.
- Feb 14, 2016: Snappy: Fix aggressive skipping.
- Feb 14, 2016: Snappy: Update benchmark.
- Feb 13, 2016: Deflate: Fixed assembler problem that could lead to sub-optimal compression.
- Feb 12, 2016: Snappy: Added AMD64 SSE 4.2 optimizations to matching, which makes easy to compress material run faster. Typical speedup is around 25%.
- Feb 9, 2016: Added Snappy package fork. This version is 5-7% faster, much more on hard to compress content.
- Jan 30, 2016: Optimize level 1 to 3 by not considering static dictionary or storing uncompressed. ~4-5% speedup.
- Jan 16, 2016: Optimization on deflate level 1,2,3 compression.
- Jan 8 2016: Merge CL 18317: fix reading, writing of zip64 archives.
- Dec 8 2015: Make level 1 and -2 deterministic even if write size differs.
- Dec 8 2015: Split encoding functions, so hashing and matching can potentially be inlined. 1-3% faster on AMD64. 5% faster on other platforms.
- Dec 8 2015: Fixed rare one byte out-of bounds read. Please update!
- Nov 23 2015: Optimization on token writer. ~2-4% faster. Contributed by @dsnet.
- Nov 20 2015: Small optimization to bit writer on 64 bit systems.
- Nov 17 2015: Fixed out-of-bound errors if the underlying Writer returned an error. See #15.
- Nov 12 2015: Added io.WriterTo support to gzip/inflate.
- Nov 11 2015: Merged CL 16669: archive/zip: enable overriding (de)compressors per file
- Oct 15 2015: Added skipping on uncompressible data. Random data speed up >5x.
The packages are drop-in replacements for standard libraries. Simply replace the import path to use them:
| old import | new import | Documentation |
|---|---|---|
compress/gzip |
github.com/klauspost/compress/gzip |
gzip |
compress/zlib |
github.com/klauspost/compress/zlib |
zlib |
archive/zip |
github.com/klauspost/compress/zip |
zip |
compress/flate |
github.com/klauspost/compress/flate |
flate |
You may also be interested in pgzip, which is a drop in replacement for gzip, which support multithreaded compression on big files and the optimized crc32 package used by these packages.
The packages contains the same as the standard library, so you can use the godoc for that: gzip, zip, zlib, flate.
Currently there is only minor speedup on decompression (mostly CRC32 calculation).
Memory usage is typically 1MB for a Writer. stdlib is in the same range. If you expect to have a lot of concurrently allocated Writers consider using the stateless compress described below.
For compression performance, see: this spreadsheet.
This package offers stateless compression as a special option for gzip/deflate. It will do compression but without maintaining any state between Write calls.
This means there will be no memory kept between Write calls, but compression and speed will be suboptimal.
This is only relevant in cases where you expect to run many thousands of compressors concurrently, but with very little activity. This is not intended for regular web servers serving individual requests.
Because of this, the size of actual Write calls will affect output size.
In gzip, specify level -3 / gzip.StatelessCompression to enable.
For direct deflate use, NewStatelessWriter and StatelessDeflate are available. See documentation
A bufio.Writer can of course be used to control write sizes. For example, to use a 4KB buffer:
// replace 'ioutil.Discard' with your output.
gzw, err := gzip.NewWriterLevel(ioutil.Discard, gzip.StatelessCompression)
if err != nil {
return err
}
defer gzw.Close()
w := bufio.NewWriterSize(gzw, 4096)
defer w.Flush()
// Write to 'w'
This will only use up to 4KB in memory when the writer is idle.
Compression is almost always worse than the fastest compression level and each write will allocate (a little) memory.
It has been a while since we have been looking at the speed of this package compared to the standard library, so I thought I would re-do my tests and give some overall recommendations based on the current state. All benchmarks have been performed with Go 1.10 on my Desktop Intel(R) Core(TM) i7-2600 CPU @3.40GHz. Since I last ran the tests, I have gotten more RAM, which means tests with big files are no longer limited by my SSD.
The raw results are in my updated spreadsheet. Due to cgo changes and upstream updates i could not get the cgo version of gzip to compile. Instead I included the zstd cgo implementation. If I get cgo gzip to work again, I might replace the results in the sheet.
The columns to take note of are: MB/s - the throughput. Reduction - the data size reduction in percent of the original. Rel Speed relative speed compared to the standard library at the same level. Smaller - how many percent smaller is the compressed output compared to stdlib. Negative means the output was bigger. Loss means the loss (or gain) in compression as a percentage difference of the input.
The gzstd (standard library gzip) and gzkp (this package gzip) only uses one CPU core. pgzip, bgzf uses all 4 cores. zstd uses one core, and is a beast (but not Go, yet).
There appears to be a roughly 5-10% speed advantage over the standard library when comparing at similar compression levels.
The biggest difference you will see is the result of re-balancing the compression levels. I wanted by library to give a smoother transition between the compression levels than the standard library.
This package attempts to provide a more smooth transition, where "1" is taking a lot of shortcuts, "5" is the reasonable trade-off and "9" is the "give me the best compression", and the values in between gives something reasonable in between. The standard library has big differences in levels 1-4, but levels 5-9 having no significant gains - often spending a lot more time than can be justified by the achieved compression.
There are links to all the test data in the spreadsheet in the top left field on each tab.
This test set aims to emulate typical use in a web server. The test-set is 4GB data in 53k files, and is a mixture of (mostly) HTML, JS, CSS.
Since level 1 and 9 are close to being the same code, they are quite close. But looking at the levels in-between the differences are quite big.
Looking at level 6, this package is 88% faster, but will output about 6% more data. For a web server, this means you can serve 88% more data, but have to pay for 6% more bandwidth. You can draw your own conclusions on what would be the most expensive for your case.
This test is for typical data files stored on a server. In this case it is a collection of Go precompiled objects. They are very compressible.
The picture is similar to the web content, but with small differences since this is very compressible. Levels 2-3 offer good speed, but is sacrificing quite a bit of compression.
The standard library seems suboptimal on level 3 and 4 - offering both worse compression and speed than level 6 & 7 of this package respectively.
This is a JSON file with very high redundancy. The reduction starts at 95% on level 1, so in real life terms we are dealing with something like a highly redundant stream of data, etc.
It is definitely visible that we are dealing with specialized content here, so the results are very scattered. This package does not do very well at levels 1-4, but picks up significantly at level 5 and levels 7 and 8 offering great speed for the achieved compression.
So if you know you content is extremely compressible you might want to go slightly higher than the defaults. The standard library has a huge gap between levels 3 and 4 in terms of speed (2.75x slowdown), so it offers little "middle ground".
This is a pretty common test corpus: enwik9. It contains the first 10^9 bytes of the English Wikipedia dump on Mar. 3, 2006. This is a very good test of typical text based compression and more data heavy streams.
We see a similar picture here as in "Web Content". On equal levels some compression is sacrificed for more speed. Level 5 seems to be the best trade-off between speed and size, beating stdlib level 3 in both.
I will combine two test sets, one 10GB file set and a VM disk image (~8GB). Both contain different data types and represent a typical backup scenario.
The most notable thing is how quickly the standard library drops to very low compression speeds around level 5-6 without any big gains in compression. Since this type of data is fairly common, this does not seem like good behavior.
This is mainly a test of how good the algorithms are at detecting un-compressible input. The standard library only offers this feature with very conservative settings at level 1. Obviously there is no reason for the algorithms to try to compress input that cannot be compressed. The only downside is that it might skip some compressible data on false detections.
This compression library adds a special compression level, named HuffmanOnly, which allows near linear time compression. This is done by completely disabling matching of previous data, and only reduce the number of bits to represent each character.
This means that often used characters, like 'e' and ' ' (space) in text use the fewest bits to represent, and rare characters like '¤' takes more bits to represent. For more information see wikipedia or this nice video.
Since this type of compression has much less variance, the compression speed is mostly unaffected by the input data, and is usually more than 180MB/s for a single core.
The downside is that the compression ratio is usually considerably worse than even the fastest conventional compression. The compression ratio can never be better than 8:1 (12.5%).
The linear time compression can be used as a "better than nothing" mode, where you cannot risk the encoder to slow down on some content. For comparison, the size of the "Twain" text is 233460 bytes (+29% vs. level 1) and encode speed is 144MB/s (4.5x level 1). So in this case you trade a 30% size increase for a 4 times speedup.
For more information see my blog post on Fast Linear Time Compression.
This is implemented on Go 1.7 as "Huffman Only" mode, though not exposed for gzip.
Here are other packages of good quality and pure Go (no cgo wrappers or autoconverted code):
- github.com/pierrec/lz4 - strong multithreaded LZ4 compression.
- github.com/cosnicolaou/pbzip2 - multithreaded bzip2 decompression.
- github.com/dsnet/compress - brotli decompression, bzip2 writer.
This code is licensed under the same conditions as the original Go code. See LICENSE file.