zstd

README

zstd

Zstandard is a real-time compression algorithm, providing high compression ratios. It offers a very wide range of compression / speed trade-off, while being backed by a very fast decoder. A high performance compression algorithm is implemented. For now focused on speed.

This package provides compression to and decompression of Zstandard content.

This package is pure Go. Use noasm and nounsafe to disable relevant features.

The zstd package is provided as open source software using a Go standard license.

Currently the package is heavily optimized for 64 bit processors and will be significantly slower on 32 bit processors.

For seekable zstd streams, see this excellent package.

Installation

Install using go get -u github.com/klauspost/compress. The package is located in github.com/klauspost/compress/zstd.

Compressor

Status:

STABLE - there may always be subtle bugs, a wide variety of content has been tested and the library is actively used by several projects. This library is being fuzz-tested for all updates.

There may still be specific combinations of data types/size/settings that could lead to edge cases, so as always, testing is recommended.

For now, a high speed (fastest) and medium-fast (default) compressor has been implemented.

• The "Fastest" compression ratio is roughly equivalent to zstd level 1.
• The "Default" compression ratio is roughly equivalent to zstd level 3 (default).
• The "Better" compression ratio is roughly equivalent to zstd level 7.
• The "Best" compression ratio is roughly equivalent to zstd level 11.

In terms of speed, it is typically 2x as fast as the stdlib deflate/gzip in its fastest mode. The compression ratio compared to stdlib is around level 3, but usually 3x as fast.

Usage

An Encoder can be used for either compressing a stream via the io.WriteCloser interface supported by the Encoder or as multiple independent tasks via the EncodeAll function. Smaller encodes are encouraged to use the EncodeAll function. Use NewWriter to create a new instance that can be used for both.

To create a writer with default options, do like this:

// Compress input to output.
func Compress(in io.Reader, out io.Writer) error {
    enc, err := zstd.NewWriter(out)
    if err != nil {
        return err
    }
    _, err = io.Copy(enc, in)
    if err != nil {
        enc.Close()
        return err
    }
    return enc.Close()
}

Now you can encode by writing data to enc. The output will be finished writing when Close() is called. Even if your encode fails, you should still call Close() to release any resources that may be held up.

The above is fine for big encodes. However, whenever possible try to reuse the writer.

To reuse the encoder, you can use the Reset(io.Writer) function to change to another output. This will allow the encoder to reuse all resources and avoid wasteful allocations.

Currently stream encoding has 'light' concurrency, meaning up to 2 goroutines can be working on part of a stream. This is independent of the WithEncoderConcurrency(n), but that is likely to change in the future. So if you want to limit concurrency for future updates, specify the concurrency you would like.

If you would like stream encoding to be done without spawning async goroutines, use WithEncoderConcurrency(1) which will compress input as each block is completed, blocking on writes until each has completed.

You can specify your desired compression level using WithEncoderLevel() option. Currently only pre-defined compression settings can be specified.

Future Compatibility Guarantees

This will be an evolving project. When using this package it is important to note that both the compression efficiency and speed may change.

The goal will be to keep the default efficiency at the default zstd (level 3). However the encoding should never be assumed to remain the same, and you should not use hashes of compressed output for similarity checks.

The Encoder can be assumed to produce the same output from the exact same code version. However, the may be modes in the future that break this, although they will not be enabled without an explicit option.

This encoder is not designed to (and will probably never) output the exact same bitstream as the reference encoder.

Also note, that the cgo decompressor currently does not report all errors on invalid input, omits error checks, ignores checksums and seems to ignore concatenated streams, even though it is part of the spec.

Blocks

For compressing small blocks, the returned encoder has a function called EncodeAll(src, dst []byte) []byte.

EncodeAll will encode all input in src and append it to dst. This function can be called concurrently. Each call will only run on a same goroutine as the caller.

Encoded blocks can be concatenated and the result will be the combined input stream. Data compressed with EncodeAll can be decoded with the Decoder, using either a stream or DecodeAll.

Especially when encoding blocks you should take special care to reuse the encoder. This will effectively make it run without allocations after a warmup period. To make it run completely without allocations, supply a destination buffer with space for all content.

import "github.com/klauspost/compress/zstd"

// Create a writer that caches compressors.
// For this operation type we supply a nil Reader.
var encoder, _ = zstd.NewWriter(nil)

// Compress a buffer. 
// If you have a destination buffer, the allocation in the call can also be eliminated.
func Compress(src []byte) []byte {
    return encoder.EncodeAll(src, make([]byte, 0, len(src)))
} 

You can control the maximum number of concurrent encodes using the WithEncoderConcurrency(n) option when creating the writer.

Using the Encoder for both a stream and individual blocks concurrently is safe.

Performance

I have collected some speed examples to compare speed and compression against other compressors.

• file is the input file.
• out is the compressor used. zskp is this package. zstd is the Datadog cgo library. gzstd/gzkp is gzip standard and this library.
• level is the compression level used. For zskp level 1 is "fastest", level 2 is "default"; 3 is "better", 4 is "best".
• insize/outsize is the input/output size.
• millis is the number of milliseconds used for compression.
• mb/s is megabytes (2^20 bytes) per second.

Silesia Corpus:
http://sun.aei.polsl.pl/~sdeor/corpus/silesia.zip

This package:
file    out     level   insize      outsize     millis  mb/s
silesia.tar zskp    1   211947520   73821326    634     318.47
silesia.tar zskp    2   211947520   67655404    1508    133.96
silesia.tar zskp    3   211947520   64746933    3000    67.37
silesia.tar zskp    4   211947520   60073508    16926   11.94

cgo zstd:
silesia.tar zstd    1   211947520   73605392    543     371.56
silesia.tar zstd    3   211947520   66793289    864     233.68
silesia.tar zstd    6   211947520   62916450    1913    105.66
silesia.tar zstd    9   211947520   60212393    5063    39.92

gzip, stdlib/this package:
silesia.tar gzstd   1   211947520   80007735    1498    134.87
silesia.tar gzkp    1   211947520   80088272    1009    200.31

GOB stream of binary data. Highly compressible.
https://files.klauspost.com/compress/gob-stream.7z

file        out     level   insize  outsize     millis  mb/s
gob-stream  zskp    1   1911399616  233948096   3230    564.34
gob-stream  zskp    2   1911399616  203997694   4997    364.73
gob-stream  zskp    3   1911399616  173526523   13435   135.68
gob-stream  zskp    4   1911399616  162195235   47559   38.33

gob-stream  zstd    1   1911399616  249810424   2637    691.26
gob-stream  zstd    3   1911399616  208192146   3490    522.31
gob-stream  zstd    6   1911399616  193632038   6687    272.56
gob-stream  zstd    9   1911399616  177620386   16175   112.70

gob-stream  gzstd   1   1911399616  357382013   9046    201.49
gob-stream  gzkp    1   1911399616  359136669   4885    373.08

The test data for the Large Text Compression Benchmark is the first
10^9 bytes of the English Wikipedia dump on Mar. 3, 2006.
http://mattmahoney.net/dc/textdata.html

file    out level   insize      outsize     millis  mb/s
enwik9  zskp    1   1000000000  343833605   3687    258.64
enwik9  zskp    2   1000000000  317001237   7672    124.29
enwik9  zskp    3   1000000000  291915823   15923   59.89
enwik9  zskp    4   1000000000  261710291   77697   12.27

enwik9  zstd    1   1000000000  358072021   3110    306.65
enwik9  zstd    3   1000000000  313734672   4784    199.35
enwik9  zstd    6   1000000000  295138875   10290   92.68
enwik9  zstd    9   1000000000  278348700   28549   33.40

enwik9  gzstd   1   1000000000  382578136   8608    110.78
enwik9  gzkp    1   1000000000  382781160   5628    169.45

Highly compressible JSON file.
https://files.klauspost.com/compress/github-june-2days-2019.json.zst

file                        out level   insize      outsize     millis  mb/s
github-june-2days-2019.json zskp    1   6273951764  697439532   9789    611.17
github-june-2days-2019.json zskp    2   6273951764  610876538   18553   322.49
github-june-2days-2019.json zskp    3   6273951764  517662858   44186   135.41
github-june-2days-2019.json zskp    4   6273951764  464617114   165373  36.18

github-june-2days-2019.json zstd    1   6273951764  766284037   8450    708.00
github-june-2days-2019.json zstd    3   6273951764  661889476   10927   547.57
github-june-2days-2019.json zstd    6   6273951764  642756859   22996   260.18
github-june-2days-2019.json zstd    9   6273951764  601974523   52413   114.16

github-june-2days-2019.json gzstd   1   6273951764  1164397768  26793   223.32
github-june-2days-2019.json gzkp    1   6273951764  1120631856  17693   338.16

VM Image, Linux mint with a few installed applications:
https://files.klauspost.com/compress/rawstudio-mint14.7z

file                    out level   insize      outsize     millis  mb/s
rawstudio-mint14.tar    zskp    1   8558382592  3718400221  18206   448.29
rawstudio-mint14.tar    zskp    2   8558382592  3326118337  37074   220.15
rawstudio-mint14.tar    zskp    3   8558382592  3163842361  87306   93.49
rawstudio-mint14.tar    zskp    4   8558382592  2970480650  783862  10.41

rawstudio-mint14.tar    zstd    1   8558382592  3609250104  17136   476.27
rawstudio-mint14.tar    zstd    3   8558382592  3341679997  29262   278.92
rawstudio-mint14.tar    zstd    6   8558382592  3235846406  77904   104.77
rawstudio-mint14.tar    zstd    9   8558382592  3160778861  140946  57.91

rawstudio-mint14.tar    gzstd   1   8558382592  3926234992  51345   158.96
rawstudio-mint14.tar    gzkp    1   8558382592  3960117298  36722   222.26

CSV data:
https://files.klauspost.com/compress/nyc-taxi-data-10M.csv.zst

file                    out level   insize      outsize     millis  mb/s
nyc-taxi-data-10M.csv   zskp    1   3325605752  641319332   9462    335.17
nyc-taxi-data-10M.csv   zskp    2   3325605752  588976126   17570   180.50
nyc-taxi-data-10M.csv   zskp    3   3325605752  529329260   32432   97.79
nyc-taxi-data-10M.csv   zskp    4   3325605752  474949772   138025  22.98

nyc-taxi-data-10M.csv   zstd    1   3325605752  687399637   8233    385.18
nyc-taxi-data-10M.csv   zstd    3   3325605752  598514411   10065   315.07
nyc-taxi-data-10M.csv   zstd    6   3325605752  570522953   20038   158.27
nyc-taxi-data-10M.csv   zstd    9   3325605752  517554797   64565   49.12

nyc-taxi-data-10M.csv   gzstd   1   3325605752  928654908   21270   149.11
nyc-taxi-data-10M.csv   gzkp    1   3325605752  922273214   13929   227.68

Decompressor

Status: STABLE - there may still be subtle bugs, but a wide variety of content has been tested.

This library is being continuously fuzz-tested, kindly supplied by fuzzit.dev. The main purpose of the fuzz testing is to ensure that it is not possible to crash the decoder, or run it past its limits with ANY input provided.

Usage

The package has been designed for two main usages, big streams of data and smaller in-memory buffers. There are two main usages of the package for these. Both of them are accessed by creating a Decoder.

For streaming use a simple setup could look like this:

import "github.com/klauspost/compress/zstd"

func Decompress(in io.Reader, out io.Writer) error {
    d, err := zstd.NewReader(in)
    if err != nil {
        return err
    }
    defer d.Close()
    
    // Copy content...
    _, err = io.Copy(out, d)
    return err
}

It is important to use the "Close" function when you no longer need the Reader to stop running goroutines, when running with default settings. Goroutines will exit once an error has been returned, including io.EOF at the end of a stream.

Streams are decoded concurrently in 4 asynchronous stages to give the best possible throughput. However, if you prefer synchronous decompression, use WithDecoderConcurrency(1) which will decompress data as it is being requested only.

For decoding buffers, it could look something like this:

import "github.com/klauspost/compress/zstd"

// Create a reader that caches decompressors.
// For this operation type we supply a nil Reader.
var decoder, _ = zstd.NewReader(nil, zstd.WithDecoderConcurrency(0))

// Decompress a buffer. We don't supply a destination buffer,
// so it will be allocated by the decoder.
func Decompress(src []byte) ([]byte, error) {
    return decoder.DecodeAll(src, nil)
} 

Both of these cases should provide the functionality needed. The decoder can be used for concurrent decompression of multiple buffers. By default 4 decompressors will be created.

It will only allow a certain number of concurrent operations to run. To tweak that yourself use the WithDecoderConcurrency(n) option when creating the decoder. It is possible to use WithDecoderConcurrency(0) to create GOMAXPROCS decoders.

Dictionaries

Data compressed with dictionaries can be decompressed.

Dictionaries are added individually to Decoders. Dictionaries are generated by the zstd --train command and contains an initial state for the decoder. To add a dictionary use the WithDecoderDicts(dicts ...[]byte) option with the dictionary data. Several dictionaries can be added at once.

The dictionary will be used automatically for the data that specifies them. A re-used Decoder will still contain the dictionaries registered.

When registering multiple dictionaries with the same ID, the last one will be used.

It is possible to use dictionaries when compressing data.

To enable a dictionary use WithEncoderDict(dict []byte). Here only one dictionary will be used and it will likely be used even if it doesn't improve compression.

The used dictionary must be used to decompress the content.

For any real gains, the dictionary should be built with similar data. If an unsuitable dictionary is used the output may be slightly larger than using no dictionary. Use the zstd commandline tool to build a dictionary from sample data. For information see zstd dictionary information.

For now there is a fixed startup performance penalty for compressing content with dictionaries. This will likely be improved over time. Just be aware to test performance when implementing.

Allocation-less operation

The decoder has been designed to operate without allocations after a warmup.

This means that you should store the decoder for best performance. To re-use a stream decoder, use the Reset(r io.Reader) error to switch to another stream. A decoder can safely be re-used even if the previous stream failed.

To release the resources, you must call the Close() function on a decoder. After this it can no longer be reused, but all running goroutines will be stopped. So you must use this if you will no longer need the Reader.

For decompressing smaller buffers a single decoder can be used. When decoding buffers, you can supply a destination slice with length 0 and your expected capacity. In this case no unneeded allocations should be made.

Concurrency

The buffer decoder does everything on the same goroutine and does nothing concurrently. It can however decode several buffers concurrently. Use WithDecoderConcurrency(n) to limit that.

The stream decoder will create goroutines that:

1. Reads input and splits the input into blocks.
2. Decompression of literals.
3. Decompression of sequences.
4. Reconstruction of output stream.

So effectively this also means the decoder will "read ahead" and prepare data to always be available for output.

The concurrency level will, for streams, determine how many blocks ahead the compression will start.

Since "blocks" are quite dependent on the output of the previous block stream decoding will only have limited concurrency.

In practice this means that concurrency is often limited to utilizing about 3 cores effectively.

Benchmarks

The first two are streaming decodes and the last are smaller inputs.

Running on AMD Ryzen 9 3950X 16-Core Processor. AMD64 assembly used.

BenchmarkDecoderSilesia-32    	                   5	 206878840 ns/op	1024.50 MB/s	   49808 B/op	      43 allocs/op
BenchmarkDecoderEnwik9-32                          1	1271809000 ns/op	 786.28 MB/s	   72048 B/op	      52 allocs/op

Concurrent blocks, performance:

BenchmarkDecoder_DecodeAllParallel/kppkn.gtb.zst-32         	   67356	     17857 ns/op	10321.96 MB/s	        22.48 pct	     102 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/geo.protodata.zst-32     	  266656	      4421 ns/op	26823.21 MB/s	        11.89 pct	      19 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/plrabn12.txt.zst-32      	   20992	     56842 ns/op	8477.17 MB/s	        39.90 pct	     754 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/lcet10.txt.zst-32        	   27456	     43932 ns/op	9714.01 MB/s	        33.27 pct	     524 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/asyoulik.txt.zst-32      	   78432	     15047 ns/op	8319.15 MB/s	        40.34 pct	      66 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/alice29.txt.zst-32       	   65800	     18436 ns/op	8249.63 MB/s	        37.75 pct	      88 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/html_x_4.zst-32          	  102993	     11523 ns/op	35546.09 MB/s	         3.637 pct	     143 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/paper-100k.pdf.zst-32    	 1000000	      1070 ns/op	95720.98 MB/s	        80.53 pct	       3 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/fireworks.jpeg.zst-32    	  749802	      1752 ns/op	70272.35 MB/s	       100.0 pct	       5 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/urls.10K.zst-32          	   22640	     52934 ns/op	13263.37 MB/s	        26.25 pct	    1014 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/html.zst-32              	  226412	      5232 ns/op	19572.27 MB/s	        14.49 pct	      20 B/op	       0 allocs/op
BenchmarkDecoder_DecodeAllParallel/comp-data.bin.zst-32     	  923041	      1276 ns/op	3194.71 MB/s	        31.26 pct	       0 B/op	       0 allocs/op

This reflects the performance around May 2022, but this may be out of date.

Zstd inside ZIP files

It is possible to use zstandard to compress individual files inside zip archives. While this isn't widely supported it can be useful for internal files.

To support the compression and decompression of these files you must register a compressor and decompressor.

It is highly recommended registering the (de)compressors on individual zip Reader/Writer and NOT use the global registration functions. The main reason for this is that 2 registrations from different packages will result in a panic.

It is a good idea to only have a single compressor and decompressor, since they can be used for multiple zip files concurrently, and using a single instance will allow reusing some resources.

See this example for how to compress and decompress files inside zip archives.

Contributions

Contributions are always welcome. For new features/fixes, remember to add tests and for performance enhancements include benchmarks.

For general feedback and experience reports, feel free to open an issue or write me on Twitter.

This package includes the excellent github.com/cespare/xxhash package Copyright (c) 2016 Caleb Spare.

Documentation

Overview

Package zstd provides decompression of zstandard files.

For advanced usage and examples, go to the README: https://github.com/klauspost/compress/tree/master/zstd#zstd

Index

• Constants
• Variables
• func BuildDict(o BuildDictOptions) ([]byte, error)
• func InspectDictionary(b []byte) (interface{ ... }, error)
• func ZipCompressor(opts ...EOption) func(w io.Writer) (io.WriteCloser, error)
• func ZipDecompressor(opts ...DOption) func(r io.Reader) io.ReadCloser
• type BuildDictOptions
• type DOption
  • func IgnoreChecksum(b bool) DOption
  • func WithDecodeAllCapLimit(b bool) DOption
  • func WithDecodeBuffersBelow(size int) DOption
  • func WithDecoderConcurrency(n int) DOption
  • func WithDecoderDictRaw(id uint32, content []byte) DOption
  • func WithDecoderDicts(dicts ...[]byte) DOption
  • func WithDecoderLowmem(b bool) DOption
  • func WithDecoderMaxMemory(n uint64) DOption
  • func WithDecoderMaxWindow(size uint64) DOption
• type Decoder
  • func NewReader(r io.Reader, opts ...DOption) (*Decoder, error)
  • func (d *Decoder) Close()
  • func (d *Decoder) DecodeAll(input, dst []byte) ([]byte, error)
  • func (d *Decoder) IOReadCloser() io.ReadCloser
  • func (d *Decoder) Read(p []byte) (int, error)
  • func (d *Decoder) Reset(r io.Reader) error
  • func (d *Decoder) WriteTo(w io.Writer) (int64, error)
• type EOption
  • func WithAllLitEntropyCompression(b bool) EOption
  • func WithEncoderCRC(b bool) EOption
  • func WithEncoderConcurrency(n int) EOption
  • func WithEncoderDict(dict []byte) EOption
  • func WithEncoderDictRaw(id uint32, content []byte) EOption
  • func WithEncoderLevel(l EncoderLevel) EOption
  • func WithEncoderPadding(n int) EOption
  • func WithLowerEncoderMem(b bool) EOption
  • func WithNoEntropyCompression(b bool) EOption
  • func WithSingleSegment(b bool) EOption
  • func WithWindowSize(n int) EOption
  • func WithZeroFrames(b bool) EOption
• type Encoder
  • func NewWriter(w io.Writer, opts ...EOption) (*Encoder, error)
  • func (e *Encoder) Close() error
  • func (e *Encoder) EncodeAll(src, dst []byte) []byte
  • func (e *Encoder) Flush() error
  • func (e *Encoder) MaxEncodedSize(size int) int
  • func (e *Encoder) ReadFrom(r io.Reader) (n int64, err error)
  • func (e *Encoder) Reset(w io.Writer)
  • func (e *Encoder) ResetContentSize(w io.Writer, size int64)
  • func (e *Encoder) Write(p []byte) (n int, err error)
• type EncoderLevel
  • func EncoderLevelFromString(s string) (bool, EncoderLevel)
  • func EncoderLevelFromZstd(level int) EncoderLevel
  • func (e EncoderLevel) String() string
• type Header
  • func (h *Header) AppendTo(dst []byte) ([]byte, error)
  • func (h *Header) Decode(in []byte) error
  • func (h *Header) DecodeAndStrip(in []byte) (remain []byte, err error)
• type SnappyConverter
  • func (r *SnappyConverter) Convert(in io.Reader, w io.Writer) (int64, error)

Examples

• WithEncoderDictRaw
• ZipCompressor

Constants

const (
	// MinWindowSize is the minimum Window Size, which is 1 KB.
	MinWindowSize = 1 << 10

	// MaxWindowSize is the maximum encoder window size
	// and the default decoder maximum window size.
	MaxWindowSize = 1 << 29
)

const HeaderMaxSize = 14 + 3

HeaderMaxSize is the maximum size of a Frame and Block Header. If less is sent to Header.Decode it *may* still contain enough information.

const ZipMethodPKWare = 20

ZipMethodPKWare is the original method number used by PKWARE to indicate Zstandard compression. Deprecated: This has been deprecated by PKWARE, use ZipMethodWinZip instead for compression. See https://pkware.cachefly.net/webdocs/APPNOTE/APPNOTE-6.3.9.TXT

const ZipMethodWinZip = 93

ZipMethodWinZip is the method for Zstandard compressed data inside Zip files for WinZip. See https://www.winzip.com/win/en/comp_info.html

Variables

var (
	// ErrSnappyCorrupt reports that the input is invalid.
	ErrSnappyCorrupt = errors.New("snappy: corrupt input")
	// ErrSnappyTooLarge reports that the uncompressed length is too large.
	ErrSnappyTooLarge = errors.New("snappy: decoded block is too large")
	// ErrSnappyUnsupported reports that the input isn't supported.
	ErrSnappyUnsupported = errors.New("snappy: unsupported input")
)

var (
	// ErrReservedBlockType is returned when a reserved block type is found.
	// Typically this indicates wrong or corrupted input.
	ErrReservedBlockType = errors.New("invalid input: reserved block type encountered")

	// ErrCompressedSizeTooBig is returned when a block is bigger than allowed.
	// Typically this indicates wrong or corrupted input.
	ErrCompressedSizeTooBig = errors.New("invalid input: compressed size too big")

	// ErrBlockTooSmall is returned when a block is too small to be decoded.
	// Typically returned on invalid input.
	ErrBlockTooSmall = errors.New("block too small")

	// ErrUnexpectedBlockSize is returned when a block has unexpected size.
	// Typically returned on invalid input.
	ErrUnexpectedBlockSize = errors.New("unexpected block size")

	// ErrMagicMismatch is returned when a "magic" number isn't what is expected.
	// Typically this indicates wrong or corrupted input.
	ErrMagicMismatch = errors.New("invalid input: magic number mismatch")

	// ErrWindowSizeExceeded is returned when a reference exceeds the valid window size.
	// Typically this indicates wrong or corrupted input.
	ErrWindowSizeExceeded = errors.New("window size exceeded")

	// ErrWindowSizeTooSmall is returned when no window size is specified.
	// Typically this indicates wrong or corrupted input.
	ErrWindowSizeTooSmall = errors.New("invalid input: window size was too small")

	// ErrDecoderSizeExceeded is returned if decompressed size exceeds the configured limit.
	ErrDecoderSizeExceeded = errors.New("decompressed size exceeds configured limit")

	// ErrUnknownDictionary is returned if the dictionary ID is unknown.
	ErrUnknownDictionary = errors.New("unknown dictionary")

	// ErrFrameSizeExceeded is returned if the stated frame size is exceeded.
	// This is only returned if SingleSegment is specified on the frame.
	ErrFrameSizeExceeded = errors.New("frame size exceeded")

	// ErrFrameSizeMismatch is returned if the stated frame size does not match the expected size.
	// This is only returned if SingleSegment is specified on the frame.
	ErrFrameSizeMismatch = errors.New("frame size does not match size on stream")

	// ErrCRCMismatch is returned if CRC mismatches.
	ErrCRCMismatch = errors.New("CRC check failed")

	// ErrDecoderClosed will be returned if the Decoder was used after
	// Close has been called.
	ErrDecoderClosed = errors.New("decoder used after Close")

	// ErrEncoderClosed will be returned if the Encoder was used after
	// Close has been called.
	ErrEncoderClosed = errors.New("encoder used after Close")

	// ErrDecoderNilInput is returned when a nil Reader was provided
	// and an operation other than Reset/DecodeAll/Close was attempted.
	ErrDecoderNilInput = errors.New("nil input provided as reader")
)

Functions

func BuildDict

func BuildDict(o BuildDictOptions) ([]byte, error)

func InspectDictionary

func InspectDictionary(b []byte) (interface {
	ID() uint32
	ContentSize() int
	Content() []byte
	Offsets() [3]int
	LitEncoder() *huff0.Scratch
}, error)

InspectDictionary loads a zstd dictionary and provides functions to inspect the content.

func ZipCompressor

func ZipCompressor(opts ...EOption) func(w io.Writer) (io.WriteCloser, error)

ZipCompressor returns a compressor that can be registered with zip libraries. The provided encoder options will be used on all encodes.

Example

package main

import (
	"archive/zip"
	"bytes"
	"fmt"
	"io"

	"github.com/klauspost/compress/zstd"
)

func main() {
	// Get zstandard de/compressors for zip.
	// These can be used by multiple readers and writers.
	compr := zstd.ZipCompressor(zstd.WithWindowSize(1<<20), zstd.WithEncoderCRC(false))
	decomp := zstd.ZipDecompressor()

	// Try it out...
	var buf bytes.Buffer
	zw := zip.NewWriter(&buf)
	zw.RegisterCompressor(zstd.ZipMethodWinZip, compr)
	zw.RegisterCompressor(zstd.ZipMethodPKWare, compr)

	// Create 1MB data
	tmp := make([]byte, 1<<20)
	for i := range tmp {
		tmp[i] = byte(i)
	}
	w, err := zw.CreateHeader(&zip.FileHeader{
		Name:   "file1.txt",
		Method: zstd.ZipMethodWinZip,
	})
	if err != nil {
		panic(err)
	}
	w.Write(tmp)

	// Another...
	w, err = zw.CreateHeader(&zip.FileHeader{
		Name:   "file2.txt",
		Method: zstd.ZipMethodPKWare,
	})
	w.Write(tmp)
	zw.Close()

	zr, err := zip.NewReader(bytes.NewReader(buf.Bytes()), int64(buf.Len()))
	if err != nil {
		panic(err)
	}
	zr.RegisterDecompressor(zstd.ZipMethodWinZip, decomp)
	zr.RegisterDecompressor(zstd.ZipMethodPKWare, decomp)
	for _, file := range zr.File {
		rc, err := file.Open()
		if err != nil {
			panic(err)
		}
		b, err := io.ReadAll(rc)
		rc.Close()
		if bytes.Equal(b, tmp) {
			fmt.Println(file.Name, "ok")
		} else {
			fmt.Println(file.Name, "mismatch")
		}
	}
}

Output:

file1.txt ok
file2.txt ok

func ZipDecompressor

func ZipDecompressor(opts ...DOption) func(r io.Reader) io.ReadCloser

ZipDecompressor returns a decompressor that can be registered with zip libraries. See ZipCompressor for example. Options can be specified. WithDecoderConcurrency(1) is forced, and by default a 128MB maximum decompression window is specified. The window size can be overridden if required.

Types

type BuildDictOptions

type BuildDictOptions struct {
	// Dictionary ID.
	ID uint32

	// Content to use to create dictionary tables.
	Contents [][]byte

	// History to use for all blocks.
	History []byte

	// Offsets to use.
	Offsets [3]int

	// CompatV155 will make the dictionary compatible with Zstd v1.5.5 and earlier.
	// See https://github.com/facebook/zstd/issues/3724
	CompatV155 bool

	// Use the specified encoder level.
	// The dictionary will be built using the specified encoder level,
	// which will reflect speed and make the dictionary tailored for that level.
	// If not set SpeedBestCompression will be used.
	Level EncoderLevel

	// DebugOut will write stats and other details here if set.
	DebugOut io.Writer
}

type DOption

type DOption func(*decoderOptions) error

DOption is an option for creating a decoder.

func IgnoreChecksum

func IgnoreChecksum(b bool) DOption

IgnoreChecksum allows to forcibly ignore checksum checking.

func WithDecodeAllCapLimit

func WithDecodeAllCapLimit(b bool) DOption

WithDecodeAllCapLimit will limit DecodeAll to decoding cap(dst)-len(dst) bytes, or any size set in WithDecoderMaxMemory. This can be used to limit decoding to a specific maximum output size. Disabled by default.

func WithDecodeBuffersBelow

func WithDecodeBuffersBelow(size int) DOption

WithDecodeBuffersBelow will fully decode readers that have a `Bytes() []byte` and `Len() int` interface similar to bytes.Buffer. This typically uses less allocations but will have the full decompressed object in memory. Note that DecodeAllCapLimit will disable this, as well as giving a size of 0 or less. Default is 128KiB.

func WithDecoderConcurrency

func WithDecoderConcurrency(n int) DOption

WithDecoderConcurrency sets the number of created decoders. When decoding block with DecodeAll, this will limit the number of possible concurrently running decodes. When decoding streams, this will limit the number of inflight blocks. When decoding streams and setting maximum to 1, no async decoding will be done. When a value of 0 is provided GOMAXPROCS will be used. By default this will be set to 4 or GOMAXPROCS, whatever is lower.

func WithDecoderDictRaw

func WithDecoderDictRaw(id uint32, content []byte) DOption

WithDecoderDictRaw registers a dictionary that may be used by the decoder. The slice content can be arbitrary data.

func WithDecoderDicts

func WithDecoderDicts(dicts ...[]byte) DOption

WithDecoderDicts allows to register one or more dictionaries for the decoder.

Each slice in dict must be in the dictionary format produced by "zstd --train" from the Zstandard reference implementation.

If several dictionaries with the same ID are provided, the last one will be used.

func WithDecoderLowmem

func WithDecoderLowmem(b bool) DOption

WithDecoderLowmem will set whether to use a lower amount of memory, but possibly have to allocate more while running.

func WithDecoderMaxMemory

func WithDecoderMaxMemory(n uint64) DOption

WithDecoderMaxMemory allows to set a maximum decoded size for in-memory non-streaming operations or maximum window size for streaming operations. This can be used to control memory usage of potentially hostile content. Maximum is 1 << 63 bytes. Default is 64GiB.

func WithDecoderMaxWindow

func WithDecoderMaxWindow(size uint64) DOption

WithDecoderMaxWindow allows to set a maximum window size for decodes. This allows rejecting packets that will cause big memory usage. The Decoder will likely allocate more memory based on the WithDecoderLowmem setting. If WithDecoderMaxMemory is set to a lower value, that will be used. Default is 512MB, Maximum is ~3.75 TB as per zstandard spec.

type Decoder

type Decoder struct {
	// contains filtered or unexported fields
}

Decoder provides decoding of zstandard streams. The decoder has been designed to operate without allocations after a warmup. This means that you should store the decoder for best performance. To re-use a stream decoder, use the Reset(r io.Reader) error to switch to another stream. A decoder can safely be re-used even if the previous stream failed. To release the resources, you must call the Close() function on a decoder.

func NewReader

func NewReader(r io.Reader, opts ...DOption) (*Decoder, error)

NewReader creates a new decoder. A nil Reader can be provided in which case Reset can be used to start a decode.

A Decoder can be used in two modes:

1) As a stream, or 2) For stateless decoding using DecodeAll.

Only a single stream can be decoded concurrently, but the same decoder can run multiple concurrent stateless decodes. It is even possible to use stateless decodes while a stream is being decoded.

The Reset function can be used to initiate a new stream, which will considerably reduce the allocations normally caused by NewReader.

func (*Decoder) Close

func (d *Decoder) Close()

Close will release all resources. It is NOT possible to reuse the decoder after this.

func (*Decoder) DecodeAll

func (d *Decoder) DecodeAll(input, dst []byte) ([]byte, error)

DecodeAll allows stateless decoding of a blob of bytes. Output will be appended to dst, so if the destination size is known you can pre-allocate the destination slice to avoid allocations. DecodeAll can be used concurrently. The Decoder concurrency limits will be respected.

func (*Decoder) IOReadCloser

func (d *Decoder) IOReadCloser() io.ReadCloser

IOReadCloser returns the decoder as an io.ReadCloser for convenience. Any changes to the decoder will be reflected, so the returned ReadCloser can be reused along with the decoder. io.WriterTo is also supported by the returned ReadCloser.

func (*Decoder) Read

func (d *Decoder) Read(p []byte) (int, error)

Read bytes from the decompressed stream into p. Returns the number of bytes read and any error that occurred. When the stream is done, io.EOF will be returned.

func (*Decoder) Reset

func (d *Decoder) Reset(r io.Reader) error

Reset will reset the decoder the supplied stream after the current has finished processing. Note that this functionality cannot be used after Close has been called. Reset can be called with a nil reader to release references to the previous reader. After being called with a nil reader, no other operations than Reset or DecodeAll or Close should be used.

func (*Decoder) WriteTo

func (d *Decoder) WriteTo(w io.Writer) (int64, error)

WriteTo writes data to w until there's no more data to write or when an error occurs. The return value n is the number of bytes written. Any error encountered during the write is also returned.

type EOption

type EOption func(*encoderOptions) error

EOption is an option for creating a encoder.

func WithAllLitEntropyCompression

func WithAllLitEntropyCompression(b bool) EOption

WithAllLitEntropyCompression will apply entropy compression if no matches are found. Disabling this will skip incompressible data faster, but in cases with no matches but skewed character distribution compression is lost. Default value depends on the compression level selected.

func WithEncoderCRC

func WithEncoderCRC(b bool) EOption

WithEncoderCRC will add CRC value to output. Output will be 4 bytes larger.

func WithEncoderConcurrency

func WithEncoderConcurrency(n int) EOption

WithEncoderConcurrency will set the concurrency, meaning the maximum number of encoders to run concurrently. The value supplied must be at least 1. For streams, setting a value of 1 will disable async compression. By default this will be set to GOMAXPROCS.

func WithEncoderDict

func WithEncoderDict(dict []byte) EOption

WithEncoderDict allows to register a dictionary that will be used for the encode.

The slice dict must be in the dictionary format produced by "zstd --train" from the Zstandard reference implementation.

The encoder *may* choose to use no dictionary instead for certain payloads.

func WithEncoderDictRaw

func WithEncoderDictRaw(id uint32, content []byte) EOption

WithEncoderDictRaw registers a dictionary that may be used by the encoder.

The slice content may contain arbitrary data. It will be used as an initial history.

Example

package main

import (
	"bytes"
	"fmt"

	"github.com/klauspost/compress/zstd"
)

func main() {
	// "Raw" dictionaries can be used for compressed delta encoding.

	source := []byte(`
		This is the source file. Compression of the target file with
		the source file as the dictionary will produce a compressed
		delta encoding of the target file.`)
	target := []byte(`
		This is the target file. Decompression of the delta encoding with
		the source file as the dictionary will produce this file.`)

	// The dictionary id is arbitrary. We use zero for compatibility
	// with zstd --patch-from, but applications can use any id
	// not in the range [32768, 1<<31).
	const id = 0

	bestLevel := zstd.WithEncoderLevel(zstd.SpeedBestCompression)

	w, _ := zstd.NewWriter(nil, bestLevel,
		zstd.WithEncoderDictRaw(id, source))
	delta := w.EncodeAll(target, nil)

	r, _ := zstd.NewReader(nil, zstd.WithDecoderDictRaw(id, source))
	out, err := r.DecodeAll(delta, nil)
	if err != nil || !bytes.Equal(out, target) {
		panic("decoding error")
	}

	// Ordinary compression, for reference.
	w, _ = zstd.NewWriter(nil, bestLevel)
	compressed := w.EncodeAll(target, nil)

	// Check that the delta is at most half as big as the compressed file.
	fmt.Println(len(delta) < len(compressed)/2)
}

Output:

true

func WithEncoderLevel

func WithEncoderLevel(l EncoderLevel) EOption

WithEncoderLevel specifies a predefined compression level.

func WithEncoderPadding

func WithEncoderPadding(n int) EOption

WithEncoderPadding will add padding to all output so the size will be a multiple of n. This can be used to obfuscate the exact output size or make blocks of a certain size. The contents will be a skippable frame, so it will be invisible by the decoder. n must be > 0 and <= 1GB, 1<<30 bytes. The padded area will be filled with data from crypto/rand.Reader. If `EncodeAll` is used with data already in the destination, the total size will be multiple of this.

func WithLowerEncoderMem

func WithLowerEncoderMem(b bool) EOption

WithLowerEncoderMem will trade in some memory cases trade less memory usage for slower encoding speed. This will not change the window size which is the primary function for reducing memory usage. See WithWindowSize.

func WithNoEntropyCompression

func WithNoEntropyCompression(b bool) EOption

WithNoEntropyCompression will always skip entropy compression of literals. This can be useful if content has matches, but unlikely to benefit from entropy compression. Usually the slight speed improvement is not worth enabling this.

func WithSingleSegment

func WithSingleSegment(b bool) EOption

WithSingleSegment will set the "single segment" flag when EncodeAll is used. If this flag is set, data must be regenerated within a single continuous memory segment. In this case, Window_Descriptor byte is skipped, but Frame_Content_Size is necessarily present. As a consequence, the decoder must allocate a memory segment of size equal or larger than size of your content. In order to preserve the decoder from unreasonable memory requirements, a decoder is allowed to reject a compressed frame which requests a memory size beyond decoder's authorized range. For broader compatibility, decoders are recommended to support memory sizes of at least 8 MB. This is only a recommendation, each decoder is free to support higher or lower limits, depending on local limitations. If this is not specified, block encodes will automatically choose this based on the input size and the window size. This setting has no effect on streamed encodes.

func WithWindowSize

func WithWindowSize(n int) EOption

WithWindowSize will set the maximum allowed back-reference distance. The value must be a power of two between MinWindowSize and MaxWindowSize. A larger value will enable better compression but allocate more memory and, for above-default values, take considerably longer. The default value is determined by the compression level and max 8MB.

func WithZeroFrames

func WithZeroFrames(b bool) EOption

WithZeroFrames will encode 0 length input as full frames. This can be needed for compatibility with zstandard usage, but is not needed for this package.

type Encoder

type Encoder struct {
	// contains filtered or unexported fields
}

Encoder provides encoding to Zstandard. An Encoder can be used for either compressing a stream via the io.WriteCloser interface supported by the Encoder or as multiple independent tasks via the EncodeAll function. Smaller encodes are encouraged to use the EncodeAll function. Use NewWriter to create a new instance.

func NewWriter

func NewWriter(w io.Writer, opts ...EOption) (*Encoder, error)

NewWriter will create a new Zstandard encoder. If the encoder will be used for encoding blocks a nil writer can be used.

func (*Encoder) Close

func (e *Encoder) Close() error

Close will flush the final output and close the stream. The function will block until everything has been written. The Encoder can still be re-used after calling this.

func (*Encoder) EncodeAll

func (e *Encoder) EncodeAll(src, dst []byte) []byte

EncodeAll will encode all input in src and append it to dst. This function can be called concurrently, but each call will only run on a single goroutine. If empty input is given, nothing is returned, unless WithZeroFrames is specified. Encoded blocks can be concatenated and the result will be the combined input stream. Data compressed with EncodeAll can be decoded with the Decoder, using either a stream or DecodeAll.

func (*Encoder) Flush

func (e *Encoder) Flush() error

Flush will send the currently written data to output and block until everything has been written. This should only be used on rare occasions where pushing the currently queued data is critical.

func (*Encoder) MaxEncodedSize

func (e *Encoder) MaxEncodedSize(size int) int

MaxEncodedSize returns the expected maximum size of an encoded block or stream.

func (*Encoder) ReadFrom

func (e *Encoder) ReadFrom(r io.Reader) (n int64, err error)

ReadFrom reads data from r until EOF or error. The return value n is the number of bytes read. Any error except io.EOF encountered during the read is also returned.

The Copy function uses ReaderFrom if available.

func (*Encoder) Reset

func (e *Encoder) Reset(w io.Writer)

Reset will re-initialize the writer and new writes will encode to the supplied writer as a new, independent stream.

func (*Encoder) ResetContentSize

func (e *Encoder) ResetContentSize(w io.Writer, size int64)

ResetContentSize will reset and set a content size for the next stream. If the bytes written does not match the size given an error will be returned when calling Close(). This is removed when Reset is called. Sizes <= 0 results in no content size set.

func (*Encoder) Write

func (e *Encoder) Write(p []byte) (n int, err error)

Write data to the encoder. Input data will be buffered and as the buffer fills up content will be compressed and written to the output. When done writing, use Close to flush the remaining output and write CRC if requested.

type EncoderLevel

type EncoderLevel int

EncoderLevel predefines encoder compression levels. Only use the constants made available, since the actual mapping of these values are very likely to change and your compression could change unpredictably when upgrading the library.

const (

	// SpeedFastest will choose the fastest reasonable compression.
	// This is roughly equivalent to the fastest Zstandard mode.
	SpeedFastest EncoderLevel

	// SpeedDefault is the default "pretty fast" compression option.
	// This is roughly equivalent to the default Zstandard mode (level 3).
	SpeedDefault

	// SpeedBetterCompression will yield better compression than the default.
	// Currently it is about zstd level 7-8 with ~ 2x-3x the default CPU usage.
	// By using this, notice that CPU usage may go up in the future.
	SpeedBetterCompression

	// SpeedBestCompression will choose the best available compression option.
	// This will offer the best compression no matter the CPU cost.
	SpeedBestCompression
)

func EncoderLevelFromString

func EncoderLevelFromString(s string) (bool, EncoderLevel)

EncoderLevelFromString will convert a string representation of an encoding level back to a compression level. The compare is not case sensitive. If the string wasn't recognized, (false, SpeedDefault) will be returned.

func EncoderLevelFromZstd

func EncoderLevelFromZstd(level int) EncoderLevel

EncoderLevelFromZstd will return an encoder level that closest matches the compression ratio of a specific zstd compression level. Many input values will provide the same compression level.

func (EncoderLevel) String

func (e EncoderLevel) String() string

String provides a string representation of the compression level.

type Header

type Header struct {
	// SingleSegment specifies whether the data is to be decompressed into a
	// single contiguous memory segment.
	// It implies that WindowSize is invalid and that FrameContentSize is valid.
	SingleSegment bool

	// WindowSize is the window of data to keep while decoding.
	// Will only be set if SingleSegment is false.
	WindowSize uint64

	// Dictionary ID.
	// If 0, no dictionary.
	DictionaryID uint32

	// HasFCS specifies whether FrameContentSize has a valid value.
	HasFCS bool

	// FrameContentSize is the expected uncompressed size of the entire frame.
	FrameContentSize uint64

	// Skippable will be true if the frame is meant to be skipped.
	// This implies that FirstBlock.OK is false.
	Skippable bool

	// SkippableID is the user-specific ID for the skippable frame.
	// Valid values are between 0 to 15, inclusive.
	SkippableID int

	// SkippableSize is the length of the user data to skip following
	// the header.
	SkippableSize uint32

	// HeaderSize is the raw size of the frame header.
	//
	// For normal frames, it includes the size of the magic number and
	// the size of the header (per section 3.1.1.1).
	// It does not include the size for any data blocks (section 3.1.1.2) nor
	// the size for the trailing content checksum.
	//
	// For skippable frames, this counts the size of the magic number
	// along with the size of the size field of the payload.
	// It does not include the size of the skippable payload itself.
	// The total frame size is the HeaderSize plus the SkippableSize.
	HeaderSize int

	// First block information.
	FirstBlock struct {
		// OK will be set if first block could be decoded.
		OK bool

		// Is this the last block of a frame?
		Last bool

		// Is the data compressed?
		// If true CompressedSize will be populated.
		// Unfortunately DecompressedSize cannot be determined
		// without decoding the blocks.
		Compressed bool

		// DecompressedSize is the expected decompressed size of the block.
		// Will be 0 if it cannot be determined.
		DecompressedSize int

		// CompressedSize of the data in the block.
		// Does not include the block header.
		// Will be equal to DecompressedSize if not Compressed.
		CompressedSize int
	}

	// If set there is a checksum present for the block content.
	// The checksum field at the end is always 4 bytes long.
	HasCheckSum bool
}

Header contains information about the first frame and block within that.

func (*Header) AppendTo

func (h *Header) AppendTo(dst []byte) ([]byte, error)

AppendTo will append the encoded header to the dst slice. There is no error checking performed on the header values.

func (*Header) Decode

func (h *Header) Decode(in []byte) error

Decode the header from the beginning of the stream. This will decode the frame header and the first block header if enough bytes are provided. It is recommended to provide at least HeaderMaxSize bytes. If the frame header cannot be read an error will be returned. If there isn't enough input, io.ErrUnexpectedEOF is returned. The FirstBlock.OK will indicate if enough information was available to decode the first block header.

func (*Header) DecodeAndStrip

func (h *Header) DecodeAndStrip(in []byte) (remain []byte, err error)

DecodeAndStrip will decode the header from the beginning of the stream and on success return the remaining bytes. This will decode the frame header and the first block header if enough bytes are provided. It is recommended to provide at least HeaderMaxSize bytes. If the frame header cannot be read an error will be returned. If there isn't enough input, io.ErrUnexpectedEOF is returned. The FirstBlock.OK will indicate if enough information was available to decode the first block header.

type SnappyConverter

type SnappyConverter struct {
	// contains filtered or unexported fields
}

SnappyConverter can read SnappyConverter-compressed streams and convert them to zstd. Conversion is done by converting the stream directly from Snappy without intermediate full decoding. Therefore the compression ratio is much less than what can be done by a full decompression and compression, and a faulty Snappy stream may lead to a faulty Zstandard stream without any errors being generated. No CRC value is being generated and not all CRC values of the Snappy stream are checked. However, it provides really fast recompression of Snappy streams. The converter can be reused to avoid allocations, even after errors.

func (*SnappyConverter) Convert

func (r *SnappyConverter) Convert(in io.Reader, w io.Writer) (int64, error)

Convert the Snappy stream supplied in 'in' and write the zStandard stream to 'w'. If any error is detected on the Snappy stream it is returned. The number of bytes written is returned.