Automated tests

Automated tests

Adding new dependencies

If you need to add new dependencies (e.g. pip/conda packages or system-level packages) then edit the Dockerfile to add these. An example is shown in this commit.

Note: only add dependencies after the line that says Add new dependencies after here. This is just because the contents above that line are cached, and editing anything above there will cause the Docker image to be rebuilt, making the tests take much longer.

Running the tests locally

todo

Limitations/what the tests won't catch

• won't detect bugs associated with parallelism in the heatmap code (parallelism/concurrency is set to 1 on the automated test environment - see this commit)
• won't detect bugs in folder watching functionality (because we only test the pipeline on a single recording at a time)
• won't detect bugs associated with full length recordings (we use a chopped recording because the test server has limited RAM and storage space)
• won't detect VR specific bugs (because we currently only test against an open-field test dataset). This can be improved by adding other datasets.
• less likely to detect bugs that occur outside the golden path (e.g. bugs in error handling code). The test recording is a working recording that shouldn't produce errors, so we won't know how the code handles dodgy recordings (e.g. that have incorrect parameters). This can be improved by adding dodgy recordings and improving the unit tests