feat: base without X_

[flying-sheep]

• Closes Don't prepend "X_" #3188
• Tests included or not required because:

• Release notes not necessary because:

TODO:

• scanpy/src/scanpy/plotting/_tools/scatterplots.py

  Line 858 in e6e9647

  if f"X_{basis}" not in adata.obsm:

• scanpy/src/scanpy/plotting/_utils.py

  Line 1034 in e6e9647

  elif f"X_{basis}" in adata.obsm:

• …

[codecov]

Codecov Report

❌ Patch coverage is 84.32432% with 29 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.91%. Comparing base (ee7707b) to head (03c8a2a).
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/scanpy/plotting/_tools/scatterplots.py	42.85%	8 Missing ⚠️
src/scanpy/_keys.py	88.88%	6 Missing ⚠️
src/scanpy/tools/_ingest.py	61.53%	5 Missing ⚠️
src/scanpy/external/tl/_trimap.py	20.00%	4 Missing ⚠️
src/scanpy/external/exporting.py	25.00%	3 Missing ⚠️
src/scanpy/_utils/__init__.py	83.33%	1 Missing ⚠️
src/scanpy/plotting/_utils.py	66.66%	1 Missing ⚠️
src/scanpy/tools/_utils.py	88.88%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4076      +/-   ##
==========================================
+ Coverage   79.74%   79.91%   +0.16%     
==========================================
  Files         120      121       +1     
  Lines       12852    12928      +76     
==========================================
+ Hits        10249    10331      +82     
+ Misses       2603     2597       -6     

Flag	Coverage Δ
hatch-test.low-vers	78.96% <84.32%> (+0.14%)	⬆️
hatch-test.pre	79.76% <84.32%> (+0.15%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/scanpy/_settings/presets.py	90.84% <100.00%> (+7.89%)	⬆️
src/scanpy/experimental/pp/_normalization.py	94.56% <100.00%> (+0.24%)	⬆️
src/scanpy/experimental/pp/_recipes.py	100.00% <100.00%> (ø)
src/scanpy/neighbors/__init__.py	82.23% <100.00%> (+0.18%)	⬆️
src/scanpy/plotting/_tools/__init__.py	76.17% <100.00%> (ø)
src/scanpy/plotting/_tools/paga.py	64.26% <100.00%> (ø)
src/scanpy/preprocessing/_pca/__init__.py	91.92% <100.00%> (-0.05%)	⬇️
src/scanpy/tools/_diffmap.py	82.60% <100.00%> (+0.79%)	⬆️
src/scanpy/tools/_dpt.py	65.84% <100.00%> (+0.23%)	⬆️
src/scanpy/tools/_draw_graph.py	69.86% <100.00%> (+0.84%)	⬆️
... and 11 more

[ilan-gold]

Are these the only two places that this is the case anymore i.e., prepend _X? It seems to have been done for PCA, but not UMAP:

scanpy/src/scanpy/tools/_umap.py

Line 147 in d96d91d

key_obsm, key_uns = ("X_umap", "umap") if key_added is None else [key_added] * 2

or am I misreading this?

[ilan-gold]

My only real comment is both concrete and exploratory - I think stronger types make sense here regardless of what I say below (after all, it should be nicer than indexing into a tuple with an integer?).

Furthermore, if we have more strongly typed returns, we could also investigate differing key names more cleanly. So instead of the present controlling one central key i.e., adata.varm["pca"] and adata.obsm["pca"] being identical, it could control all the keys (maybe?). This is really high-level, but my motivating thought here is that the different keys currently have the nice property of being a bit more readable. For example, adata.varm["PCs"], while a little clunky, represents "principal components in gene space" meaningfully (compared to adata.varm["pca"]).

Thoughts?

[flying-sheep]

the different keys currently have the nice property of being a bit more readable

I agree, varm["PCs"] is more informative than varm["pca"]. That does apply only

• to the default (I think it’s cleaner to directly use key_added when specified)
• to PCA (the others don’t have a varm key)

So the question boils down to:

should we leave differing keys when calling sc.pp.pca() while ScanpyV2Preview is enabled?

Or did I miss something?

[ilan-gold]

to PCA (the others don’t have a varm key)

This applies to diffmap too, for example:

https://github.com/scverse/scanpy/pull/4076/changes#diff-f25e128747ab8dc77cf527f9e972df2e5ddb32aa14f037d6d4b2d9219a99f397R70-R76

i.e., evals really mean "eigenvalues" which is different than X_diffmap in obsm