Ruby bindings for zstd library¶ ↑

See zstd library.

Other bindings: lzw, brotli, bzip2.

Installation¶ ↑

Operating systems: GNU/Linux, FreeBSD, OSX.

Dependencies: zstd 1.4.0+ version.

gem install ruby-zstds

You can build it from source.

rake gem
gem install pkg/ruby-zstds-*.gem

You can also use overlay for gentoo.

Installation in macOS on Apple Silicon¶ ↑

On M1 Macs, Homebrew installs to /opt/homebrew, so you’ll need to specify its include and lib paths when building the native extension for zstd.

brew install zstd
gem install ruby-zstds -- --with-opt-include=/opt/homebrew/include --with-opt-lib=/opt/homebrew/lib

You can also configure Bundler to use those options when installing:

bundle config set build.ruby-zstds "--with-opt-include=/opt/homebrew/include --with-opt-lib=/opt/homebrew/lib"

Usage¶ ↑

There are simple APIs: String and File. Also you can use generic streaming API: Stream::Writer and Stream::Reader.

require "zstds"

data = ZSTDS::String.compress "sample string"
puts ZSTDS::String.decompress(data)

ZSTDS::File.compress "file.txt", "file.txt.zst"
ZSTDS::File.decompress "file.txt.zst", "file.txt"

ZSTDS::Stream::Writer.open("file.txt.zst") { |writer| writer << "sample string" }
puts ZSTDS::Stream::Reader.open("file.txt.zst") { |reader| reader.read }

writer = ZSTDS::Stream::Writer.new output_socket
begin
  bytes_written = writer.write_nonblock "sample string"
  # handle "bytes_written"
rescue IO::WaitWritable
  # handle wait
ensure
  writer.close
end

reader = ZSTDS::Stream::Reader.new input_socket
begin
  puts reader.read_nonblock(512)
rescue IO::WaitReadable
  # handle wait
rescue ::EOFError
  # handle eof
ensure
  reader.close
end

You can create dictionary using ZSTDS::Dictionary.

require "securerandom"
require "zstds"

samples = (Array.new(8) { ::SecureRandom.random_bytes(1 << 8) } + ["sample string"]).shuffle

dictionary = ZSTDS::Dictionary.train samples
File.write "dictionary.bin", dictionary.buffer, :mode => "wb"

dictionary_buffer = File.read "dictionary.bin", :mode => "rb"
dictionary        = ZSTDS::Dictionary.new dictionary_buffer

data = ZSTDS::String.compress "sample string", :dictionary => dictionary
puts ZSTDS::String.decompress(data, :dictionary => dictionary)

You can create and read tar.zst archives with minitar.

require "zstds"
require "minitar"

ZSTDS::Stream::Writer.open "file.tar.zst" do |writer|
  Minitar::Writer.open writer do |tar|
    tar.add_file_simple "file", :data => "sample string"
  end
end

ZSTDS::Stream::Reader.open "file.tar.zst" do |reader|
  Minitar::Reader.open reader do |tar|
    tar.each_entry do |entry|
      puts entry.name
      puts entry.read
    end
  end
end

You can also use Content-Encoding: zstd with sinatra:

require "zstds"
require "sinatra"

get "/" do
  headers["Content-Encoding"] = "zstd"
  ZSTDS::String.compress "sample string"
end

All functionality (including streaming) can be used inside multiple threads with parallel. This code will provide heavy load for your CPU.

require "zstds"
require "parallel"

Parallel.each large_datas do |large_data|
  ZSTDS::String.compress large_data
end

Docs¶ ↑

Please review rdoc generated docs.

Options¶ ↑

There are internal buffers for compressed and decompressed data. For example you want to use 1 KB as source_buffer_length for compressor - please use 256 B as destination_buffer_length. You want to use 256 B as source_buffer_length for decompressor - please use 1 KB as destination_buffer_length.

gvl is disabled by default, this mode allows running multiple compressors/decompressors in different threads simultaneously. Please consider enabling gvl if you don’t want to launch processors in separate threads. If gvl is enabled ruby won’t waste time on acquiring/releasing VM lock.

String and File will set :pledged_size automaticaly.

You can also read zstd docs for more info about options.

Possible compressor options:

:source_buffer_length
:destination_buffer_length
:gvl
:compression_level
:window_log
:hash_log
:chain_log
:search_log
:min_match
:target_length
:strategy
:enable_long_distance_matching
:ldm_hash_log
:ldm_min_match
:ldm_bucket_size_log
:ldm_hash_rate_log
:content_size_flag
:checksum_flag
:dict_id_flag
:nb_workers
:job_size
:overlap_log
:dictionary
:pledged_size

Possible decompressor options:

:source_buffer_length
:destination_buffer_length
:gvl
:window_log_max
:dictionary

Example:

require "zstds"

data = ZSTDS::String.compress "sample string", :compression_level => 5
puts ZSTDS::String.decompress(data, :window_log_max => 11)

String¶ ↑

String maintains destination buffer only, so it accepts destination_buffer_length option only.

::compress(source, options = {})
::decompress(source, options = {})

source is a source string.

File¶ ↑

File maintains both source and destination buffers, it accepts both source_buffer_length and destination_buffer_length options.

::compress(source, destination, options = {})
::decompress(source, destination, options = {})

source and destination are file pathes.

Stream::Writer¶ ↑

Its behaviour is similar to builtin {Zlib::GzipWriter}.

Writer maintains destination buffer only, so it accepts destination_buffer_length option only.

::open(file_path, options = {}, :external_encoding => nil, :transcode_options => {}, &block)

Open file path and create stream writer associated with opened file. Data will be transcoded to :external_encoding using :transcode_options before compressing.

It may be tricky to use both :pledged_size and :transcode_options. You have to provide size of transcoded input.

::new(destination_io, options = {}, :external_encoding => nil, :transcode_options => {})

Create stream writer associated with destination io. Data will be transcoded to :external_encoding using :transcode_options before compressing.

It may be tricky to use both :pledged_size and :transcode_options. You have to provide size of transcoded input.

#set_encoding(external_encoding, nil, transcode_options)

Set another encodings, nil is just for compatibility with IO.

#io
#to_io
#stat
#external_encoding
#transcode_options
#pos
#tell

See {IO} docs.

#write(*objects)
#flush
#rewind
#close
#closed?

See {Zlib::GzipWriter} docs.

#write_nonblock(object, *options)
#flush_nonblock(*options)
#rewind_nonblock(*options)
#close_nonblock(*options)

Special asynchronous methods missing in Zlib::GzipWriter. rewind wants to close, close wants to write something and flush, flush want to write something. So it is possible to have asynchronous variants for these synchronous methods. Behaviour is the same as IO#write_nonblock method.

#<<(object)
#print(*objects)
#printf(*args)
#putc(object, :encoding => 'ASCII-8BIT')
#puts(*objects)

Typical helpers, see {Zlib::GzipWriter} docs.

Stream::Reader¶ ↑

Its behaviour is similar to builtin {Zlib::GzipReader}.

Reader maintains both source and destination buffers, it accepts both source_buffer_length and destination_buffer_length options.

::open(file_path, options = {}, :external_encoding => nil, :internal_encoding => nil, :transcode_options => {}, &block)

Open file path and create stream reader associated with opened file. Data will be force encoded to :external_encoding and transcoded to :internal_encoding using :transcode_options after decompressing.

::new(source_io, options = {}, :external_encoding => nil, :internal_encoding => nil, :transcode_options => {})

Create stream reader associated with source io. Data will be force encoded to :external_encoding and transcoded to :internal_encoding using :transcode_options after decompressing.

#set_encoding(external_encoding, internal_encoding, transcode_options)

Set another encodings.

#io
#to_io
#stat
#external_encoding
#internal_encoding
#transcode_options
#pos
#tell

See {IO} docs.

#read(bytes_to_read = nil, out_buffer = nil)
#eof?
#rewind
#close
#closed?

See {Zlib::GzipReader} docs.

#readpartial(bytes_to_read = nil, out_buffer = nil)
#read_nonblock(bytes_to_read, out_buffer = nil, *options)

See {IO} docs.

#getbyte
#each_byte(&block)
#readbyte
#ungetbyte(byte)

#getc
#readchar
#each_char(&block)
#ungetc(char)

#lineno
#lineno=
#gets(separator = $OUTPUT_RECORD_SEPARATOR, limit = nil)
#readline
#readlines
#each(&block)
#each_line(&block)
#ungetline(line)

Typical helpers, see {Zlib::GzipReader} docs.

Dictionary¶ ↑

You can train dictionary from samples using train class method.

::train(samples, :capacity => 0)

Please review zstd code before using it. There are many validation requirements and it changes between versions.

#buffer

There is an attribute reader for buffer. You can use it to store dictionary somewhere.

::new(buffer)

Please use regular constructor to create dictionary from buffer.

#id

Read dictionary id from buffer.

Thread safety¶ ↑

:gvl option is disabled by default, you can use bindings effectively in multiple threads. Please be careful: bindings are not thread safe. You should lock all shared data between threads.

For example: you should not use same compressor/decompressor inside multiple threads. Please verify that you are using each processor inside single thread at the same time.

CI¶ ↑

Please visit scripts/test-images. See universal test script scripts/ci_test.sh for CI.

License¶ ↑

MIT license, see LICENSE and AUTHORS.