Version 0.1 - May 2014
This project skeleton provides some basic capabilities to support collaborative data analysis across a number of machines. The most general model assumes that multiple team members will be operating on the code base simultaneously from their local environments. The project skeleton simplifies the processes of capturing the Python environment specification supporting the project and recreating that environment as needed on any machine that has conda and pip installed.
To ensure that team members are always operating on the same data, we will employ a model where all derived data is defined by codified transformations of the original source data. At no time will the source data itself be directly modified as the source data resides outside of version control. By forcing data transformations to happen through code that is under version control, we can establish a tight coupling between analytic results and the supporting data transformations needed to produce them.
By enforcing this workflow, we simplify the processes of synchronizing the team during the project and delivering reproducible analytic results to our clients at the conclusion of the project.
This workflow is undoubtedly a work in progress. It came to life after particular challenges arose during a client engagement. Please pass along suggestions or make contributions through pull requests to improve it!
Chris Diehl
[email protected]