- Implemented a fully functional ASV continuous benchmarking GitHub Actions workflow for DIPY
- Replaced isolated benchmark execution with
asv continuous master HEAD for automated performance regression detection
- Integrated automated PR benchmark comparison reporting with sticky pull request comments
- Pinned GitHub Actions to stable versions consistent with the rest of DIPY CI infrastructure
- Added artifact upload support and improved workflow maintainability following patterns used by NumPy, SciPy, and scikit-image
- Addressed maintainer review feedback, rebased conflicts, and validated workflow compatibility across CI environments
View Pull Request ↗
- Identified simulation tutorials missing explicit
plt.figure() calls before plotting in sphinx-gallery builds
- Replaced bare
plt.plot() calls with fig, ax = plt.subplots() (object-oriented API) in simulate_dki.py and simulate_multi_tensor.py
- Built documentation locally on Windows using meson-python and MSVC, verifying both tutorials executed successfully
- Confirmed clean, separate figure generation with no canvas overlap in generated documentation
- Documentation-only fix with no functional behavior changes
View Pull Request ↗
- Standardized docstrings in
dipy/data/fetcher.py to follow numpydoc conventions
- Corrected
Returns section formatting and removed duplicate entries
- Updated type annotations from
string to Path where appropriate
- Improved docstring clarity and ensured proper Sphinx rendering
- Maintained full backward compatibility (documentation-only change)
View Pull Request ↗