Skip to content

TODO.md

Latest commit

 

History

History
63 lines (61 loc) · 3.49 KB

TODO.md

File metadata and controls

63 lines (61 loc) · 3.49 KB
Raw

Primary

  • Consolidate plotting paths with render backend (lets_plot or matplotlib)
  • NaN and Inf support for _anndata_io
  • Recreate LazyTransform from uns
  • Documentation audit
  • Numerical precision inconsistency
    • Mix of float32 and float64 (mainly in operator, sometimes in-mem) causes drift within and across modes
    • C++ uses 64-bit, R/Rcpp does also — always consistent
  • Fix R/Python result parity
  • Reorganize and consolidate code
    • Audit for dead functions
  • Interactive cell selector
  • UMAP points have no stroke
  • compute_archetype_feature_specificity(): key_added > key_prefix
  • Make archetype specificity and network centrality optional in run_actionet()
  • Split _backed_persist.py
  • [Deferred] Simplify anndata 0.13 backed compatibility patch once anndata>=0.13 is the floor
    • Currently src/actionet/io/persist.py, src/actionet/io/subset.py, and src/actionet/io/checkpoint.py filter out the None key from adata.layers.keys() via a _real_layer_keys helper to avoid writing spurious layers/None HDF5 groups (anndata 0.13 aliases .X as layers[None]).
    • init_from_reopened in src/actionet/io/backed_adapter.py also unpacks the reopened AnnData into explicit kwargs (and drives the "init from file" branch via filename=) to sidestep the X is layers[None] identity check that fails when backed _CSRDataset wrappers are recreated per attribute access.
    • When we drop anndata<0.13 support, revisit both workarounds: the _real_layer_keys helper can likely be inlined or removed entirely, and init_from_reopened can be simplified now that layers[None] is a stable, documented alias for .X.

Secondary

  • Consolidate normalization code-paths
  • Add network centrality to run_actionet?
  • Lazy transform in-memory?
  • Allow changes to lazy transform params
    • Force recompute of attributes
  • compute_transparency() use scale()
  • Explore more accurate and faster log approximations for JSD and lazy transform
  • Combine plot_umap* paths with raster=bool
  • Python compute_feature_specificity() in-mem is still garbage (maybe not?)

Done

  • ARMA_DONT_USE_WRAPPER multiple redefinition
  • Parallel specificity bug (fixed???)
  • Standardize key args
  • Test impute features
  • Implement plotFeatures
  • Test backed SVD
  • MatrixSource supports layers, but layers can't be backed.
    • Added validation logic
  • Add pseudocount to normalize_anndata()
  • compute_feature_specificity backed/in-mem parity
  • Fix memory usage in compute_network_diffusion() (in-memory)
  • Speed up plot/impute features
    • I/O bound. Probably as good as it's gonna get
  • Add run_lpa/propagate_labels
  • Decouple archetype specificity from run_actionet
    • Compute specificity on the fly for impute_from_archetypes()
  • Decouple compute_network_diffusion from anndata
  • Add 3D to plotly
  • Optionally omit C_* and specificity matrices to reduce object size
  • Document everything (OpenMP hard requirement, README overhaul, context files updated)
  • annotate_cells enrichment → DataFrame
  • Fixed violin/boxplot alignment
  • Delete PRIMME sources, headers, vendored tree, CMake wiring, ALG_PRIMME, and remaining test/wrapper references (see context/DECISIONS.md "SVD algorithm strategy").
  • Delete Feng SVD C++ sources (svd_feng.{cpp,hpp}), the ALG_FENG enum, and Feng switch cases in runSVD/runSVD_Operator. The libactionet/wrappers_r/ copies here are reference-only and were intentionally left untouched; a reminder to patch the standalone actionet-r package is tracked in src/libactionet/TODO.md.