Repository navigation
The Horse Who Steadied the Snake's Garden
Summary
- Made pytest the canonical test runner, establishing a single authoritative entry point for the test suite
- Fixed deduplication of pytest runtime profile records to ensure clean profiling data
- Broadened and refined warning filters for sklearn, dataframe interchange, and general dataframe warnings in tests
- Merged test authority gates (PR #629) to solidify testing infrastructure reliability
Details
Release v4.0.1 is a quiet but purposeful refinement — the kind of work that rarely makes headlines yet profoundly shapes the day-to-day experience of every contributor to Limen. At its core, this release is about bringing order and clarity to the project's testing foundations.
The most consequential change is the elevation of pytest as the canonical test runner. While pytest has long been the tool of choice in practice, this release formalizes that relationship, removing ambiguity about how tests should be invoked and ensuring that CI pipelines, contributor workflows, and documentation all speak the same language. It is a small declaration with far-reaching consequences: fewer surprises for newcomers, fewer edge cases in automation, and a single source of truth for how quality is validated.
Alongside this, a subtle but important fix addresses the deduplication of pytest runtime profile records. Previously, under certain conditions, profiling data could contain duplicate entries, clouding performance analysis and making it harder to identify genuine bottlenecks. With this fix, profile records are now clean and trustworthy — a prerequisite for meaningful performance work in future releases.
The remaining changes form a cohesive effort to tame the noise in test output. Warning filters have been broadened and refined across several fronts: sklearn's liblinear solver now gracefully handles n_jobs warnings that are irrelevant to Limen's usage patterns, dataframe interchange deprecation notices are properly suppressed in the test environment, and a more general dataframe warning filter ensures that upstream library chatter does not obscure genuine issues. These are the kinds of housekeeping tasks that, left unattended, slowly erode developer confidence in the test suite — when every run is peppered with spurious warnings, real problems become harder to spot.
Taken together, these changes reflect a philosophy of stewardship: the belief that a well-maintained foundation is not a luxury but a necessity. In the spirit of the steadfast Horse tempering the Snake's restless ingenuity, this release prioritizes stability, clarity, and trust in the tools we rely on every day.