View benchmarks for latest deployment
A simple lightweight set of implementations and bindings for compression algorithms written in Go.
This project contains the source code for a summer mentorship about learning how to implement and create different compression algorithms. This includes common algorithms such as huffman, lempel-ziv, and arithmetic along with bindings for builtin Go compression algorithms.
To start using this package from the command line, install it with go get
and go install
$ go get -u github.com/go-compression/raisin
$ go install github.com/go-compression/raisin/cmd/...
Once done, you should be able to start using it
$ echo "Hello world!" > test.txt
$ raisin test.txt
Compressing...
Original bytes: 13
Compressed bytes: 14
Compression ratio: 107.69%
$ cat test.txt.rsn
��ӷ �?��KD+
�
$ rm test.txt
$ raisin -decompress test.txt.rsn
Decompressing...
$ cat test.txt
Hello world!
The possible commands include:
-compress
- Compress a given file and output the compressed contents to a file with ".rsn" at the end-decompress
- Decompress a given file and output the decompressed contents to a file without ".rsn" at the end-benchmark
- Benchmark a given file and measure the compression ratio, outputs a .rsn and a .decompressed file
The most important flag is the -algorithm
flag which allows you to specify which algorithm to use during compression, decompression, or benchmarking. By default for compress
and decompress
this is lzss,arithmetic
. The possible algorithms include:
- lzss
- dmc
- huffman
- mcc
- arithmetic
- flate
- gzip
- lzw
- zlib
Here's an example of usage:
$ raisin -algorithm=arithmetic test.txt
You can also combine algorithms together in "layers", this will essentially compress the file with the first algorithm, then the second, etc. This stacking of algorithms is what powers virtually all modern compression, gzip and zip is powered by the FLATE algorithm which is essentially lempel-ziv (similar to lzss) and huffman coding stacked on toip of each other.
$ raisin -algorithm=lzss,huffman test.txt
Compressing...
Compression ratio: 307.69%
$ raisin -decompress -algorithm=lzss,huffman test.txt.rsn
Decompressing...
On top of this, you can easily compress or decompress multiple files by chaining them together with commas.
$ raisin test1.txt,test2.txt
Compressing...
Compression ratio: 68.53%
$ ls
test1.txt test1.txt.rsn test2.txt test2.txt.rsn
When using compress
and decompress
a few more options become available to make it easy to use from the command line:
delete
- Delete original file after compression/decompressed (defaults to true for decompression)out
- File name to be outputted (defaults to original file + .rsn for compression and file - .rsn for decompression, only available with a single file being compressed/decompressed)outext
- File extension to be outputted when compressing multiple files (unavailable with a single file being compressed/decompressed)
Let's take at the usage of delete
, keep in mind that delete
is on by default for decompress
ing.
$ echo "Hello world!" > test.txt
$ raisin -delete test.txt
Compressing...
Compression ratio: 107.69%
$ ls
test.txt.rsn
$ raisin -decompress -delete test.txt.rsn
Decompressing...
$ ls
test.txt
$ raisin -delete=false test.txt
Compressing...
Compression ratio: 107.69%
$ ls
test.txt test.txt.rsn
The out
command simply lets you change what file is outputted when compressing a single file:
$ echo "Hello world!" > test.txt
$ raisin -out=compressed.txt test.txt
Compressing...
Compression ratio: 107.69%
$ ls
test.txt compressed.txt
$ raisin -decompress -out=decompressed.txt compressed.txt
Decompressing...
$ ls
test.txt decompressed.txt
outext
is similar to out
but exists for when we compress/decompress multiple files. If outext
is provided, it will be used as the output extension for the files. Note that the default for compression for outext is .rsn
and for decompression it's an empty string (outext=
) which tells the program to remove the last extension.
$ ls
test1.txt test2.txt test3.txt
$ raisin -delete -outext=.testing test1.txt,test2.txt,test3.txt
Compressing...
Compression ratio: 107.69%
$ ls
test1.txt.testing test2.txt.testing test3.txt.testing
$ raisin -decompress -outext=.decompressed test1.txt.testing,test2.txt.testing,test3.txt.testing
Decompressing...
$ ls
test1.txt test2.txt test3.txt
You can use the benchmark
command to generate benchmarked results for a set of algorithms, layers, and files. This is helpful for generating results in a table, website, or in bindings for other languages such as python (see the ai
folder).
Usage is relatively similar to the compress
and decompress
commands.
$ echo "Hello world!" > test.txt
$ echo "abcabcabcabcabcabcabcabc" > test2.txt
$ raisin -benchmark -algorithm=lzss,huffman,arithmetic,gzip,[lzss,arithmetic] test.txt,test2.txt
┌─────────────────┬────────────┬───────────────────┬────────────────┬─────────────────────┬──────────┐
│ ENGINE │ TIME TAKEN │ COMPRESSION RATIO │ ACTUAL ENTROPY │ THEORETICAL ENTROPY │ LOSSLESS │
├─────────────────┼────────────┼───────────────────┼────────────────┼─────────────────────┼──────────┤
│ lzss │ 350µs │ 100.00% │ 2.20 │ 2.20 │ true │
│ arithmetic │ 50µs │ 107.69% │ 2.12 │ 2.20 │ true │
│ lzss,arithmetic │ 210µs │ 107.69% │ 2.12 │ 2.20 │ true │
│ gzip │ 280µs │ 284.62% │ 1.14 │ 2.20 │ true │
│ huffman │ 190µs │ 307.69% │ 1.08 │ 2.20 │ true │
├─────────────────┼────────────┼───────────────────┼────────────────┼─────────────────────┼──────────┤
│ File │ test.txt │ Size │ 13 B │ │ │
└─────────────────┴────────────┴───────────────────┴────────────────┴─────────────────────┴──────────┘
┌─────────────────┬────────────┬───────────────────┬────────────────┬─────────────────────┬──────────┐
│ ENGINE │ TIME TAKEN │ COMPRESSION RATIO │ ACTUAL ENTROPY │ THEORETICAL ENTROPY │ LOSSLESS │
├─────────────────┼────────────┼───────────────────┼────────────────┼─────────────────────┼──────────┤
│ lzss,arithmetic │ 160µs │ 84.00% │ 1.25 │ 1.22 │ true │
│ lzss │ 310µs │ 84.00% │ 1.25 │ 1.22 │ true │
│ arithmetic │ 160µs │ 84.00% │ 1.25 │ 1.22 │ true │
│ huffman │ 170µs │ 92.00% │ 1.24 │ 1.22 │ true │
│ gzip │ 430µs │ 120.00% │ 1.17 │ 1.22 │ true │
├─────────────────┼────────────┼───────────────────┼────────────────┼─────────────────────┼──────────┤
│ File │ test2.txt │ Size │ 25 B │ │ │
└─────────────────┴────────────┴───────────────────┴────────────────┴─────────────────────┴──────────┘
A larger example, taken from the .travis.yml
file to generate the benchmark page. Notice the -generate
flag, this tells it to generate an html file and output it as index.html
, which is then used and uploaded to the GitHub Pages branch. Keep in mind the program expects a template file to be at templates/benchmark.html
relative to your working directory. The command is as follows:
$ raisin -benchmark -generate -algorithm=lzss,dmc,huffman,flate,gzip,lzw,zlib,arithmetic,[lzss,huffman],[lzss,arithmetic],[arithmetic,huffman] alice29.txt,asyoulik.txt,cp.html,fields.c,grammar.lsp,kennedy.xls,lcet10.txt,plrabn12.txt,ptt5,sum,xargs.1
Shout-out to jedib0t for his wonderful go-pretty module for generating these tables and the HTML tables used in the GitHub Pages site.
To build the binary from source, simply go get
the package:
$ go get -u github.com/go-compression/raisin
Install the dependencies:
$ go get
And build:
$ go build
To use this package as a module, simply import the engine package and use the io.Reader and io.Writer interfaces.
import (
"fmt"
"github.com/go-compression/raisin/engine"
)
func main() {
text := []byte("Hello world!")
file := engine.CompressedFile{}
file.CompressionEngine = "arithmetic"
file.Write(text)
fmt.Println("Compressed:", string(file.Compressed))
}
Documentation is available at godoc, please note that most of the code is currently undocumented as it is still a work in progress.