(fix): write out full PCA results when not run before neighbors (scve…

…rse#3079)

Co-authored-by: Philipp A <[email protected]>

docs/release-notes/1.10.2.md

@@ -17,6 +17,7 @@
 
 * Compatibility with `matplotlib` 3.9 {pr}`2999` {smaller}`I Virshup`
 * Add clear errors where `backed` mode-like matrices (i.e., from `sparse_dataset`) are not supported {pr}`3048` {smaller}`I gold`
+* Write out full pca results when `_choose_representation` is called i.e., `~sc.pp.neighbors` without `sc.pp.pca` {pr}`3078` {smaller}`I gold`
 * Fix deprecated use of `.A` with sparse matrices {pr}`3084` {smaller}`P Angerer`
 
 ```{rubric} Performance

scanpy/tests/test_neighbors_key_added.py

@@ -32,6 +32,16 @@ def test_neighbors_key_added(adata):
     )
 
 
+def test_neighbors_pca_keys_added_without_previous_pca_run(adata):
+    assert "pca" not in adata.uns and "X_pca" not in adata.obsm
+    with pytest.warns(
+        UserWarning,
+        match=r".*Falling back to preprocessing with `sc.pp.pca` and default params",
+    ):
+        sc.pp.neighbors(adata, n_neighbors=n_neighbors, random_state=0)
+    assert "pca" in adata.uns
+
+
 # test functions with neighbors_key and obsp
 @needs.igraph
 @needs.leidenalg

scanpy/tools/_utils.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import warnings
 from typing import TYPE_CHECKING
 
 import numpy as np
@@ -37,14 +38,14 @@ def _choose_representation(
                 X = adata.obsm["X_pca"][:, :n_pcs]
                 logg.info(f"    using 'X_pca' with n_pcs = {X.shape[1]}")
             else:
-                logg.warning(
+                warnings.warn(
                     f"You’re trying to run this on {adata.n_vars} dimensions of `.X`, "
                     "if you really want this, set `use_rep='X'`.\n         "
                     "Falling back to preprocessing with `sc.pp.pca` and default params."
                 )
                 n_pcs_pca = n_pcs if n_pcs is not None else settings.N_PCS
-                X = pca(adata.X, n_comps=n_pcs_pca)
-                adata.obsm["X_pca"] = X
+                pca(adata, n_comps=n_pcs_pca)
+                X = adata.obsm["X_pca"]
         else:
             logg.info("    using data matrix X directly")
             X = adata.X