Contribution activity.
Upstream pull requests to projects I do not own. Status is synced from GitHub.
perf: avoid dense BasePlot aggregation
Keeps sparse and backed inputs intact in dotplot, matrixplot, and stacked violin by aggregating with Scanpy kernels instead of building a dense cell-level dataframe.
Impact: Targets memory blow-ups when plotting large single-cell datasets; closes a long-standing performance issue.
Fix PersistenceLength normalization for sliced run
Corrects PersistenceLength normalization so sliced trajectory runs use the number of analyzed frames rather than the full trajectory length.
Impact: Prevents silently rescaled polymer persistence-length results when users analyze trajectory subsets.
Promote bool columns to nullable boolean when reindexing alt-axis annotations under outer join
Fixes outer-join reindexing that can promote boolean annotation columns to object dtype and make merged AnnData files unwritable.
Impact: Targets shared merge behavior used by both in-memory and on-disk concatenation paths.
Skip files_unchanged lint when no author names are found in the manifest
Fixes an unhandled validation error in the files_unchanged lint when manifest contributors have no names, falling back to the manifest author or skipping the test with a message.
Impact: Stops pipeline linting from crashing on pipelines such as pairgenomealign whose contributors list has no names.