CI madness with Ibis: testing 20 query engines on every commit
summary
Project runs tests against 20 backends (about 17 SQL, 3 non‑SQL) on every commit and claims CI completes quickly by making individual jobs fast. Tools actually used: poetry for locked deps, Docker/docker‑compose for identical dev and CI environments, GitHub Actions with a paid team plan ($624/year) to increase concurrency, plus the task runner just. Measured metrics—average job duration, queue time, workflow duration—drop variability after switching to poetry; CI metrics are pulled daily via the GitHub API and loaded into BigQuery.
AI-written summary · about
chapters (13)
- 0:00Testing multiple databases
- 1:53IBIS project overview
- 3:38Challenges of 20 backends
- 4:48Fast CI testing principles
- 6:33CI pipeline tools
- 9:27CI performance metrics
- 13:45Optimizing CI metrics
- 14:52Best practices summary
- 15:13IBIS with Polars backend
- 16:20Collecting CI metrics
- 16:41Plotting graph tools
- 17:47Favorite backend and filetype
- 18:31Analyzing workflow spikes
related talks
24mSupercharging CI: Test Suite Optimization for Speed, Stability, and Savings
35mManaging CI Across Multiple Farms for GPU Testing
32mBuilding Continuous Integration and Feedback
35mMassively Continuous Integration: From 3 days to 30 minutes
41mContinuous Integration Tips & Tricks
28mUsing CI to Test the Frontend: the TeamCity Example
32mBuilding Confidence Together: Continuous Integration & Testing
58mSpeed up your test suite by throwing computers at it