srtdisk v0.3.2

Release Date: 2026-03-18

Changes

• Removed immature h5mu support
• Refactors and more comprehensive h5ad format support.

---

srtdisk v0.3.1

Release Date: 2026-02-25

Bug Fixes

• Fixed CSR sparse matrix loading failure in ReadSparseMatrix() that caused 'dims' must contain all (i,j) pairs errors when loading large h5ad-converted h5Seurat files. The dims argument to Matrix::sparseMatrix() was incorrectly swapped as (ncols, nrows) instead of (nrows, ncols), and the unnecessary transpose has been removed.
• Added direct h5ad-to-RDS conversion path in Convert() so that Convert("file.h5ad", dest = "file.rds") works without a manual intermediate h5Seurat step.

Validated on mouse atlas (356K cells, 16.8K genes) and human atlas (1.04M cells, 17.1K genes).

---

v0.2.1

Fixed gene set mapping

• Changed the mapping logic: Seurat scale.data layer is no longer mapped during conversion
• In previous version, the default behavior is to map scale.data to adata.X when present. This will cause non-highly-variable gene symbols being lost.
• Current behavior is data->X, counts->raw (or counts->X if data layer does not exist).
• Added a new column is_highly_variable to match scanpy behavior.

Enhanced robustness for column conversions

• Implemented column name cleaning steps to avoid metadata conversion failure.

Loom Support (Beta)

• New SaveLoom() / as.loom() updated for Seurat v5, with a new convert-loom
  vignette.

Bug Fixes

• Fixed spatial image dimension ordering (images were transposed incorrectly)
• Fixed data/counts layer mixup when loading h5Seurat files
• Fixed NA encoding in categorical h5ad columns
• Added spatial centroid fallback for native scanpy/squidpy h5ad files
• Spatial conversion promoted from beta to fully supported
• Minimum R version now 4.1.0

v0.2.0

Release of srtdisk v0.2.0

What's Changed

• Added support for Visium data, h5ad <-> Seurat obj
• Added support for multi-assay object, such as CITE-seq, with examples (1-assay at a time)
• Fixed issues with FOV and image misalignment in Visium data conversion