v0.4.0

New features! This doesn't change anything in the preprocessing/singularity workflow.

1. new write_10x function with Dandelion class so that it can now format AIRR-formatted tables to 10x's input files (e.g. all_contig_annotations.csv and all_contig.fasta)
2. new saving mode - makes no difference to users but the new format is different in the backend as i have dropped the dependency on pandas - now it directly saves with h5py, which means that the way the files can be accessed is different. files saves previously should still be possible to be read in - if resaved, it will overwrite with the new format.
3. downsampling in generate_network now can also downsample orphan VJ chains (previously it relied on VDJ chains primarily), and also works on alternative clone_key
4. dandelion is now compatible with python 3.12. Just in time before python 3.13 😩
5. compatibility with AnnData is fixed. @grst, you should be able to unpin dandelion in scirpy!
6. revamped the testing suite with micromamba with new caching options.

What's Changed

• Fix anndata compat by @zktuong in #408
• fix clone id key by @zktuong in #414
• new saving mode by @zktuong in #420
• fix downsampling logic in generate network by @zktuong in #424
• fix clone_key not respected properly if non-default clone_id by @zktuong in #429
• prepare for python 3.12 by @zktuong in #363

dependency updates

• pip prod(deps): update pandas requirement from <=2.2.2,>=1.0.3 to >=1.0.3,<=2.2.3 by @dependabot in #431
• pip dev(deps-dev): update sphinx-autodoc-typehints requirement from <=2.2.3 to <=2.4.4 by @dependabot in #430
• pip dev(deps-dev): update scirpy requirement from <=0.17.2 to <=0.18.0 by @dependabot in #426

Full Changelog: v0.3.8...v0.4.0