On the structure and reproducibility of Python packages - data crunch
summary
Data-driven analysis compares 1,180 top PyPI packages with ~2,260 biomedical projects and uses FawltyDeps to detect undeclared imports. Key findings: 99.9% of top PyPI projects declare dependencies while ~53–60% biomedical projects do, many biomedical repos keep code in the root rather than a src/ or same-name package, and biomedical projects prefer requirements.txt vs pyproject.toml for dependency declaration. The analysis flags undeclared imports and plans to compare declared versus actually imported packages to quantify reproducibility.
AI-written summary · about
chapters (8)
- 0:00Python project structure
- 3:40Python project pitfalls
- 10:09Experiment design & data
- 13:41Key findings
- 15:30Dependency patterns
- 19:10Conclusions & recommendations
- 21:18Actionable steps
- 23:27Q&A
related talks
24mThe trends in choosing licenses in Python ecosystem
32mHow (not) to upgrade a platform
25m"Security considerations in Python Packaging"
32mPyPI in the face: running jokes that PyPI download stats can play on you
54mKey Challenges In the PyData Ecosystem and How we Can All Make a Difference
24mBuilding a Sustainable Python Package Index
28mWhyPy? Understanding Trends Driving the Usage of Python
35mSecuring Python projects Supply Chain