27 Stars
Updated Last
1 Year Ago
Started In
November 2014

The Blosc Module for Julia

Build Status Build status

This module provides fast lossless compression for the Julia language by interfacing the Blosc Library.

Note that Blosc is currently limited to 32-bit buffer sizes. Blosc does run just fine on 64-bit systems; it just can't compress arrays bigger than 2GB. Note also that this limitation does not affect the use of Blosc compression for HDF5, since HDF5 breaks up large arrays into small chunks before compressing them. So, don't worry about this if you are just using Blosc within the HDF5 package.


To install, simply run Pkg.add("Blosc") in Julia. Precompiled binaries are provided for Mac and Windows systems, while on other systems the Blosc library will be downloaded and compiled.


The functions provided are:

  • compress(src::Array{T}; level=5, shuffle=true, itemsize=sizeof(T)): returns a Vector{UInt8} consisting of src in compressed form. level is the compression level (between 0=no compression and 9=max), shuffle indicates whether to use Blosc's shuffling preconditioner, which is optimized for arrays of binary items of size itemsize.

  • compress!(dest::Vector{UInt8}, src; ...): as compress, but uses a pre-allocated destination buffer dest. Returns the size (in bytes) of the compressed data, or 0 if the buffer was too small.

  • decompress(T::Type, src::Vector{UInt8}): return the compressed buffer src as an array of element type T.

  • decompress!(dest::Vector{T}, src::Vector{UInt8}): like decompress, but uses a pre-allocated destination buffer, which is resized as needed to store the decompressed data.

  • Blosc.set_num_threads(n=CPU_CORES): tells Blosc to use n threads (initially 1).

  • Blosc.compressors(): returns an array of strings for the available compression algorithms. (The blosclz compressor is the default.)

  • Blosc.set_compressor(s::AbstractString): set the current compression algorithm


This module was written by Steven G. Johnson and Jake Bolewski (who had independently written his own Blosc.jl package which is now merged with this one).

Required Packages

No packages found.