Over a couple of months in 2018 I developed a script for refactoring GDAL’s ancient and arcane test suite to use Pytest. The final pull request was over 118,000 lines. This talk covers how I did it and what I learned.
At Koordinates we’ve been using Pytest for all our testing. It’s so much better than the other options that I cringe whenever I have to test a project that uses the stdlib.
We also make heavy use of GDAL, a broad and ancient data-munging library and tool. Unfortunately the test suite was 20 years old and homegrown. The test runner predates the unittest tools in the standard library, and sprouts weird quirky behaviour all over the place.
At Pycon AU 2018 I happened across John Reese’s excellent talk on Refactoring Using the Standard Library and realised I could use Bowler to do an automated rewrite of the GDAL test suite, and hopefully convert it to use Pytest.
This talk describes my experience, including:
- when to use automated refactoring tools
- the basics of automated refactoring
- how to avoid git conflicts when working on a huge pull request on a very active project
- how to actually get your >100,000 line pull request merged upstream (!)
Watch 'How I auto-refactored a huge OSS project to use pytest' on PyCon AU's YouTube account
Craig de Stigter
I’ve been a platform engineer for Koordinates for ages. I’m a Christian, a husband and father of two small people.
Back when I used to have spare time, I spent most of it snowboarding, drinking craft beer and doing open source programming. I do a bit of brewing now and then.
I like Python. I like upgrading stuff and refactoring code. Sometimes I think I’d be a good janitor, or possibly a demolition person.
Platinum Sponsors
Gold Sponsors
Sponsors
Auspice
