pyf.splitter 3.1

Dataflow splitting system for PyF framework

Introduction

pyf.splitter is a fully independent module that can be used with pyf or in any other projet. It does not have dependencies on pyf.

Purpose

The splitter purpose is simple and will stay so. It gives you an abstraction above a data flow (or any python iterable) and gives the illusion of manipulating in memory iterables when in fact everything is serialized on disk to avoid memory consumption.

The second and last purpose is to split (hence the name) your data flow according to some simple rules. Splitting is at the very least important to be able to store huge data chunks on disk without hitting file systems limitations (ever tried to store 600Gb files on a fat file system?)

It is important to note that we do not encapsulate (ie: hide) the bucket files. The splitter gives you the bucket file names it produced, you then use another function to read the files into another stream.

Running tests

To run tests you need to install tox:

pip install tox

and then just launch tox if you want the whole test suite, ie python2.7, python3.4 and pep8.

If you want to only run only kind of test (ie: python2.7 only) you can specify it like so:

tox -e py27

all defined test envs are defined in the tox.ini file

Changes

Oct 9 2015, version 3.1

• Version 3.0 introduced python3 support and version 3.1 is a bugfix release to add more sane defaults in the separator. (The datetime objects were not serializable using the python3 version with the default separator)