Ref paths

anndata/__init__.py

@@ -1,5 +1,6 @@
 from ._core.anndata import AnnData, ImplicitModificationWarning
 from ._core.raw import Raw
+from ._core.ref_path import RefPath
 from ._io import (
     read_h5ad,
     read_loom,
@@ -71,6 +72,13 @@
    AnnData.write_loom
    AnnData.write_zarr
 
+Helper classes
+--------------
+
+.. autosummary::
+   :toctree: .
+
+   RefPath
 
 Errors and warnings
 -------------------

anndata/_core/aligned_mapping.py

@@ -230,6 +230,11 @@ def __init__(
         if vals is not None:
             self.update(vals)
 
+    def __setitem__(self, key: str, value: V):
+        if hasattr(value, "index") and isinstance(value.index, pd.RangeIndex):
+            value.index = self.dim_names
+        super().__setitem__(key, value)
+
 
 class AxisArraysView(AlignedViewMixin, AxisArraysBase):
     def __init__(

anndata/_core/anndata.py

@@ -42,6 +42,7 @@
     _resolve_idxs,
 )
 from .sparse_dataset import SparseDataset
+from . import ref_path
 from .. import utils
 from ..utils import convert_to_dict, ensure_df_homogeneous
 from ..logging import anndata_logger as logger
@@ -275,8 +276,18 @@ def __init__(
         obs: Optional[Union[pd.DataFrame, Mapping[str, Iterable[Any]]]] = None,
         var: Optional[Union[pd.DataFrame, Mapping[str, Iterable[Any]]]] = None,
         uns: Optional[Mapping[str, Any]] = None,
-        obsm: Optional[Union[np.ndarray, Mapping[str, Sequence[Any]]]] = None,
-        varm: Optional[Union[np.ndarray, Mapping[str, Sequence[Any]]]] = None,
+        obsm: Optional[
+            Union[
+                np.ndarray,
+                Mapping[str, Union[np.ndarray, sparse.spmatrix, pd.DataFrame]],
+            ]
+        ] = None,
+        varm: Optional[
+            Union[
+                np.ndarray,
+                Mapping[str, Union[np.ndarray, sparse.spmatrix, pd.DataFrame]],
+            ]
+        ] = None,
         layers: Optional[Mapping[str, Union[np.ndarray, sparse.spmatrix]]] = None,
         raw: Optional[Mapping[str, Any]] = None,
         dtype: Union[np.dtype, str] = "float32",
@@ -285,8 +296,12 @@ def __init__(
         filemode: Optional[Literal["r", "r+"]] = None,
         asview: bool = False,
         *,
-        obsp: Optional[Union[np.ndarray, Mapping[str, Sequence[Any]]]] = None,
-        varp: Optional[Union[np.ndarray, Mapping[str, Sequence[Any]]]] = None,
+        obsp: Optional[
+            Union[np.ndarray, Mapping[str, Union[np.ndarray, sparse.spmatrix]]]
+        ] = None,
+        varp: Optional[
+            Union[np.ndarray, Mapping[str, Union[np.ndarray, sparse.spmatrix]]]
+        ] = None,
         oidx: Index1D = None,
         vidx: Index1D = None,
     ):
@@ -1334,6 +1349,10 @@ def _get_X(self, use_raw=False, layer=None):
         else:
             return self.X
 
+    resolve_path = ref_path.resolve_path
+    get_df = ref_path.get_df
+    get_vector = ref_path.get_vector
+
     def obs_vector(self, k: str, *, layer: Optional[str] = None) -> np.ndarray:
         """\
         Convenience function for returning a 1 dimensional ndarray of values

anndata/_core/index.py

@@ -7,6 +7,9 @@
 import pandas as pd
 from scipy.sparse import spmatrix, issparse
 
+from ..compat import Literal
+from . import anndata
+
 
 Index1D = Union[slice, int, str, np.int64, np.ndarray]
 Index = Union[Index1D, Tuple[Index1D, Index1D], spmatrix]
@@ -138,9 +141,14 @@ def make_slice(idx, dimidx, n=2):
     return tuple(mut)
 
 
-def get_vector(adata, k, coldim, idxdim, layer=None):
+def get_vector(
+    adata: "anndata.AnnData",
+    k: str,
+    coldim: Literal["obs", "var"],
+    idxdim: Literal["obs", "var"],
+    layer: Optional[str] = None,
+):
     # adata could be self if Raw and AnnData shared a parent
-    dims = ("obs", "var")
     col = getattr(adata, coldim).columns
     idx = getattr(adata, f"{idxdim}_names")
 
@@ -158,9 +166,11 @@ def get_vector(adata, k, coldim, idxdim, layer=None):
     elif in_col:
         return getattr(adata, coldim)[k].values
     elif in_idx:
-        selected_dim = dims.index(idxdim)
-        idx = adata._normalize_indices(make_slice(k, selected_dim))
-        a = adata._get_X(layer=layer)[idx]
-    if issparse(a):
-        a = a.toarray()
-    return np.ravel(a)
+        return get_x_vector(adata, idxdim, k, layer)
+
+
+def get_x_vector(adata, idxdim: Literal["obs", "var"], k: str, layer: str = None):
+    selected_dim = ("obs", "var").index(idxdim)
+    idx = adata._normalize_indices(make_slice(k, selected_dim))
+    a = adata._get_X(layer=layer)[idx]
+    return np.ravel(a.toarray() if issparse(a) else a)

anndata/_core/raw.py

@@ -97,6 +97,10 @@ def var_names(self):
     def obs_names(self):
         return self._adata.obs_names
 
+    @property
+    def layers(self):
+        return {None: self.X}
+
     def __getitem__(self, index):
         oidx, vidx = self._normalize_indices(index)