Clean up Round - cut down anndata extension functionality (#880)

* remove delete_from_obs

* move get to proper get module, minimal test

* remove generate_anndata, never used

* delete unused private _detect_binary_columns

* move anndata ext functions mostly relevant for testing to test

* remove set_numeric_vars, reduce use get_column_indices

* remove _assert_encoded completely idk why it would be needed even

* remove not_encoded_error

* remove get_numeric_vars

* updates

* fixes

* underscore _assert_numeric_vars

* update submodule

* remove get_...

* remove doc of delete_from_obs

* fix typo

* fix typo

* to_dense and add get to docs

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>

* polish

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>

* submodule

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>

* fix docs

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>

* fix docs

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>

* fix docs

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>

---------

Signed-off-by: Lukas Heumos <lukas.heumos@posteo.net>
Co-authored-by: Lukas Heumos <lukas.heumos@posteo.net>

Author: eroell

.gitignore

@@ -158,3 +158,4 @@ notebooks/*-requirements.txt
 
 *.DS_Store
 test.ipynb
+test2.ipynb

docs/api/anndata_index.md

@@ -16,9 +16,15 @@
     anndata.df_to_anndata
     anndata.anndata_to_df
     anndata.move_to_obs
-    anndata.delete_from_obs
     anndata.move_to_x
-    anndata.get_obs_df
-    anndata.get_var_df
-    anndata.get_rank_features_df
+```
+
+```{eval-rst}
+.. autosummary::
+    :toctree: get
+    :nosignatures:
+
+    get.obs_df
+    get.var_df
+    get.rank_features_groups_df
 ```

docs/tutorials/notebooks

@@ -19,7 +19,7 @@
 
 from ehrapy import anndata as ad
 from ehrapy import data as dt
-from ehrapy import io
+from ehrapy import get, io
 from ehrapy import plot as pl
 from ehrapy import preprocessing as pp
 from ehrapy import tools as tl

ehrapy/__init__.py

@@ -6,15 +6,9 @@
 )
 from ehrapy.anndata.anndata_ext import (
     anndata_to_df,
-    delete_from_obs,
     df_to_anndata,
-    generate_anndata,
-    get_obs_df,
-    get_rank_features_df,
-    get_var_df,
     move_to_obs,
     move_to_x,
-    rank_genes_groups_df,
 )
 
 __all__ = [
@@ -23,13 +17,7 @@
     "feature_type_overview",
     "infer_feature_types",
     "anndata_to_df",
-    "delete_from_obs",
     "df_to_anndata",
-    "generate_anndata",
-    "get_obs_df",
-    "get_rank_features_df",
-    "get_var_df",
     "move_to_obs",
     "move_to_x",
-    "rank_genes_groups_df",
 ]

ehrapy/anndata/__init__.py

@@ -130,6 +130,27 @@ def infer_feature_types(
         raise ValueError(f"Output format {output} not recognized. Choose between 'tree', 'dataframe', or None.")
 
 
+# TODO: this function is a different flavor of inferring feature types. We should decide on a single implementation in the future.
+def _infer_numerical_column_indices(
+    adata: AnnData, layer: str | None = None, column_indices: Iterable[int] | None = None
+) -> list[int]:
+    mtx = adata.X if layer is None else adata[layer]
+    indices = (
+        list(range(mtx.shape[1])) if column_indices is None else [i for i in column_indices if i < mtx.shape[1] - 1]
+    )
+    non_numerical_indices = []
+    for i in indices:
+        # The astype("float64") call will throw only if the feature’s data type cannot be cast to float64, meaning in
+        # practice it contains non-numeric values. Consequently, it won’t throw if the values are numeric but stored
+        # as an "object" dtype, as astype("float64") can successfully convert them to floats.
+        try:
+            mtx[::, i].astype("float64")
+        except ValueError:
+            non_numerical_indices.append(i)
+
+    return [idx for idx in indices if idx not in non_numerical_indices]
+
+
 def check_feature_types(func):
     @wraps(func)
     def wrapper(adata, *args, **kwargs):