lir 1.5.0

v1.5.0

on Python PyPI

What's Changed

New Features

Score-LLR plot / Score histogram — scores are now captured at intermediate stages, enabling richer diagnostic plots directly from pipeline output.

ValidateFeatureData transformer — enforces expected feature data types at .apply time, catching type mismatches early and making pipelines more robust.

Documentation

• Improved doc linting and cleaned up documentation pages
• Updated get_apidocs_uri, fixed the registry overview, and converted several Markdown links to RST
• Extended Pipeline docstring
• Removed __init__ docstrings; replaced with proper Attributes sections in class docstrings
• Added an explanation of PAV interpretation
• Reorganised unit tests to mirror the package structure

Testing

• Added unit tests for conversion functions
• Unit test suite now runs cleanly without warnings

Fixes & Improvements

• SourcePairing.ratio_limit now accepts float arguments
• Removed the requests dependency (now covered implicitly by requests-cache)
• Adopted YAML anchors to reduce config duplication
• Replaced elif chains with match/case statements in McMc
• Moved devpav from lir.metrics to lir.algorithms
• mcmc no longer requires labels to be passed during .apply
• Refactored LLR/score histogram classes to share a common base class
• Removed all assert statements from production code
• Replaced several custom methods with their NumPy equivalents

Contributors

Thanks to [@PimMeulensteen](https://github.com/PimMeulensteen) and [@wowtor](https://github.com/wowtor).