17 hours ago
- Scientific Python libraries often use Jupyter notebooks for user guides and documentation.
- User guide notebooks can serve as effective end-to-end tests since they cover major use cases.
- Notebooks need to be parametrized (e.g., reading server URL from environment) to work both for users and CI pipelines.
- Two methods to run notebooks automatically: using nbclient (simple error detection) or nbconvert with jq (detailed output but harder error handling).
- The nbclient method works well for testing both documentation and code integrity on each merge request.