Merge branch 'main' into ig/cache_indptr

anndata/_core/sparse_dataset.py

@@ -17,6 +17,7 @@
 from abc import ABC
 from functools import cached_property
 from itertools import accumulate, chain
+from pathlib import Path
 from typing import TYPE_CHECKING, Literal, NamedTuple
 
 import h5py
@@ -25,7 +26,7 @@
 from scipy.sparse import _sparsetools
 
 from anndata._core.index import _fix_slice_bounds
-from anndata.compat import H5Group, ZarrGroup
+from anndata.compat import H5Group, ZarrArray, ZarrGroup
 
 from ..compat import _read_attr
 
@@ -58,8 +59,17 @@ class BackedSparseMatrix(_cs_matrix):
     def copy(self) -> ss.spmatrix:
         if isinstance(self.data, h5py.Dataset):
             return sparse_dataset(self.data.parent).to_memory()
-        else:
-            return super().copy()
+        if isinstance(self.data, ZarrArray):
+            import zarr
+
+            return sparse_dataset(
+                zarr.open(
+                    store=self.data.store,
+                    mode="r",
+                    chunk_store=self.data.chunk_store,  # chunk_store is needed, not clear why
+                )[Path(self.data.path).parent]
+            ).to_memory()
+        return super().copy()
 
     def _set_many(self, i: Iterable[int], j: Iterable[int], x):
         """\

anndata/tests/test_backed_sparse.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-from contextlib import contextmanager
 from typing import TYPE_CHECKING, Callable, Literal
 
 import h5py
@@ -64,7 +63,7 @@ def callback(func, elem_name, elem, iospec):
                         **{k: read_dispatched(v, callback) for k, v in elem.items()}
                     )
                 if iospec.encoding_type in {"csc_matrix", "csr_matrix"}:
-                    return sparse_dataset(elem)._to_backed()
+                    return sparse_dataset(elem)
                 return func(elem)
 
             adata = read_dispatched(f, callback=callback)
@@ -90,6 +89,7 @@ def test_backed_indexing(
 
     assert_equal(csr_mem[obs_idx, var_idx].X, csr_disk[obs_idx, var_idx].X)
     assert_equal(csr_mem[obs_idx, var_idx].X, csc_disk[obs_idx, var_idx].X)
+    assert_equal(csr_mem.X[...], csc_disk.X[...])
     assert_equal(csr_mem[obs_idx, :].X, dense_disk[obs_idx, :].X)
     assert_equal(csr_mem[obs_idx].X, csr_disk[obs_idx].X)
     assert_equal(csr_mem[:, var_idx].X, dense_disk[:, var_idx].X)
@@ -288,17 +288,6 @@ def test_anndata_sparse_compat(tmp_path: Path, diskfmt: Literal["h5ad", "zarr"])
     assert_equal(adata.X, base)
 
 
-@contextmanager
-def xfail_if_zarr(diskfmt: Literal["h5ad", "zarr"]):
-    if diskfmt == "zarr":
-        with pytest.raises(AssertionError):
-            yield
-        # TODO: Zarr backed mode https://github.com/scverse/anndata/issues/219
-        pytest.xfail("Backed zarr not really supported yet")
-    else:
-        yield
-
-
 def test_backed_sizeof(
     ondisk_equivalent_adata: tuple[AnnData, AnnData, AnnData, AnnData],
     diskfmt: Literal["h5ad", "zarr"],
@@ -308,6 +297,5 @@ def test_backed_sizeof(
     assert csr_mem.__sizeof__() == csr_disk.__sizeof__(with_disk=True)
     assert csr_mem.__sizeof__() == csc_disk.__sizeof__(with_disk=True)
     assert csr_disk.__sizeof__(with_disk=True) == csc_disk.__sizeof__(with_disk=True)
-    with xfail_if_zarr(diskfmt):
-        assert csr_mem.__sizeof__() > csr_disk.__sizeof__()
-        assert csr_mem.__sizeof__() > csc_disk.__sizeof__()
+    assert csr_mem.__sizeof__() > csr_disk.__sizeof__()
+    assert csr_mem.__sizeof__() > csc_disk.__sizeof__()

docs/release-notes/0.10.4.md

@@ -7,6 +7,7 @@
 * `adata[:, []]` now returns an `AnnData` object empty on the appropriate dimensions instead of erroring {pr}`1243` {user}`ilan-gold`
 * `adata.X[mask]` works in newer `numpy` versions when `X` is `backed` {pr}`1255` {user}`ilan-gold`
 * `BaseCompressedSparseDataset`'s `indptr` is cached {pr}`1266` {user}`ilan-gold`
+* `adata.X[...]` fixed for `X` as a `BaseCompressedSparseDataset` with `zarr` backend {pr}`1265` {user}`ilan-gold`
 
 ```{rubric} Documentation
 ```