Changelog¶

0.2.1 (2026-09-01)¶

Rename the PyPI distribution to acidgenomics-acidgenomes. The import name is unchanged: import acidgenomes as ag still works.
Publish to PyPI instead of python.acidgenomics.com.
Pin [tool.uv.build-backend] module-name = "acidgenomes" explicitly, since the build backend’s default module name now derives from the distribution name (acidgenomics_acidgenomes), not the import name.

Work around a bug in genomicranges 0.8.4, where GenomicRanges.to_pandas() silently doubles the row count. It sets the range DataFrame’s index to gr.names before it concatenates gr.mcols.to_pandas(), whose index is still a positional RangeIndex. pandas.concat(..., axis=1) then outer-joins two disjoint indices instead of binding columns positionally, so every row loses either its coordinate columns or its metadata columns.
Add _genomicranges_to_pandas_safe(), and use it in place of GenomicRanges.to_pandas() at both call sites: the _GRangesWrapper.data property, and make_ensembl_genes_from_gtf(). Every provider (Ensembl, GENCODE, RefSeq, UCSC, FlyBase, WormBase) routes through the shared dataframe_to_granges / make_granges_from_gff pipeline, so this fixes all of them, not Ensembl alone.
The 0.2.0 package index artifact was rebuilt in place on 2026-08-21 to carry this fix. The v0.2.0 git tag was moved to match, so git checkout v0.2.0 reproduces the artifact currently served from the index.

Add tests/test_genomicranges_to_pandas_safe.py. It covers the upstream bug and the workaround through the real dataframe_to_granges construction path.

Handle None/NaN biotype values in _apply_broad_class that caused AttributeError: 'float' object has no attribute 'lower' with Ensembl release 116 GTF files.

Initial release.