Blaze extends the usability of NumPy and Pandas to distributed and out-of-core computing. Blaze provides an interface similar to that of the NumPy ND-Array or Pandas DataFrame. Blaze maps inputs from these familiar interfaces onto a variety of other computational engines like Postgres or Spark. Blaze connects users to big computation from the comfort of an interactive object.
Blaze includes a rich set of computational and data primitives useful in building and communicating between computational systems. Blaze can help with consistent and robust data migration, as well as remote execution.
Blaze aims to be a foundational project allowing many different users of other PyData projects (Pandas, Theano, Numba, SciPy, Scikit-Learn) to interoperate at the application level and at the library level with the goal of being able to to lift their existing functionality into a distributed context.
Blaze is in development. Take a look at the release notes.
Documentation is available at blaze.pydata.org/
The easiest way to try out Blaze is through the Anaconda distribution.
To make sure you're running the latest released version of Blaze, use the conda package manager to update.
$ conda update blaze
Source code for the latest development version of blaze can be obtained from Github.
The Blaze project itself is spread out over multiple projects,
in addition to the main blaze
repo. Other Blaze projects include the
following:
Additionally, Blaze hooks into and depends on several excellent projects
written by the general community. See requirements.txt
for a full list of
dependencies.
Install all the pre-requisites using conda or another mechanism, then run:
$ python setup.py install
If you're on a Linux or Mac OS-X platform, you can install a development version of Blaze (hosted on Binstar) by typing the following:
$ conda install -c mwiebe -c mrocklin blaze
Anyone wishing to discuss on Blaze should join the blaze-dev mailing list. To get started contributing, read through the Developer Workflow documentation.
Blaze development is sponsored by Continuum Analytics.
Released under BSD license. See LICENSE.txt for details.