Recently, the python scripts under the importing directory were restructured to work with zc.buildout. I switched to Buildout because:

  • Buildout automatically installs (almost) all of the dependencies. With Jenkins running unit tests, the scripts are now running on three different computers. I didn’t have access the the server running Jenkins so I couldn’t install all the dependencies by hand. With build out, that’s not a problem.
  • Buildout is a well-known standard in the Python community. Many Python developers already have experience with Buildout due to its wide use, so the structure should be easier for them to understand.
  • Buildout easily generates jUnit xml test reports. In order to integrate the existing unit tests with Jenkins, jUnit xml formatted reports need to be generated. Buildout already has a recipe for this called collective.xmltestreport.
  • Buildout can use a specific version of Python for all executables. Incompatibility bugs were being introduced occasionally because I was using Python version 2.7 for development, but the production servers only have version 2.6 installed. With Buildout I can easily switch to Python 2.6 locally for development to avoid the problem.

The costs of switching to Buildout are:

  • The time taken to learn it and restructure the code. I had never used Buildout before, so this took about 8 hours all up. The 8 hours includes the Jenkins integration which would have taken time regardless of whether Buildout was used. The time spent switching will be be offset by the benefits like faster setup and debugging.
  • Added complexity. Buildout is light-weight, but it’s still a bit more complex than a bunch of .py files. Although, you could argue that using a non-standard structure for a bunch of .py files is more complicated than using a standard structure.

blog comments powered by Disqus