This is a green test light (?):
-
we want to have a fully tested pipeline for empirical projects
- read raw data
- clean
- analysis
- produce output
each step should be tested. ideally the entire pipeline rebuilds if we make any kind of change anywhere in the code base. This requires an automatique setup (you don't want to to have to do that manually yourself). for example in this line we run our pipeline, in the next line we check it's output.
-
the badge above links to buildkite.com which is like a private CI platform. I have 2 dedicated machines in cumulus just waiting till I push a commit to repo x on github which is linked up to buildkite which triggers a job on a machine in cumulus. Can be installed on any machine. You can run anything you want on cumulus of course (even stata ;-) ) and have a big a dataset as well. It’s useful for build setups with license restrictions or other stuff that is impossible on Travis. One has to maintain the build bot which is a bit of a pain, but it’s a good solution still.
-
communication between RAs: that’s one of my main motivations. Also we could @tag people within the org an assign jobs and share solutions across projects. We could even share RAs.