Merge branch 'main' into no-positional-bools

.github/workflows/test-gpu.yml

@@ -85,3 +85,10 @@ jobs:
           token: ${{ secrets.CODECOV_TOKEN }}
           fail_ci_if_error: true
           verbose: true
+
+      - name: Remove 'run-gpu-ci' Label
+        if: always()
+        uses: actions-ecosystem/action-remove-labels@v1
+        with:
+          labels: "run-gpu-ci"
+          github_token: ${{ secrets.GITHUB_TOKEN }}

.pre-commit-config.yaml

@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.6.8
+    rev: v0.7.0
     hooks:
       - id: ruff
         types_or: [python, pyi, jupyter]
@@ -14,7 +14,7 @@ repos:
         exclude_types:
           - markdown
   - repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.6.0
+    rev: v5.0.0
     hooks:
       - id: trailing-whitespace
       - id: end-of-file-fixer

src/anndata/_io/specs/registry.py

@@ -395,10 +395,62 @@ def read_elem_as_dask(
     chunks, optional
        length `n`, the same `n` as the size of the underlying array.
        Note that the minor axis dimension must match the shape for sparse.
+       Defaults to `(1000, adata.shape[1])` for CSR sparse,
+       `(adata.shape[0], 1000)` for CSC sparse,
+       and the on-disk chunking otherwise for dense.
 
     Returns
     -------
         DaskArray
+
+    Examples
+    --------
+
+    Setting up our example:
+
+    >>> from scanpy.datasets import pbmc3k
+    >>> import tempfile
+    >>> import anndata as ad
+    >>> import zarr
+
+    >>> tmp_path = tempfile.gettempdir()
+    >>> zarr_path = tmp_path + "/adata.zarr"
+
+    >>> adata = pbmc3k()
+    >>> adata.layers["dense"] = adata.X.toarray()
+    >>> adata.write_zarr(zarr_path)
+
+    Reading a sparse matrix from a zarr store lazily, with custom chunk size and default:
+
+    >>> g = zarr.open(zarr_path)
+    >>> adata.X = ad.experimental.read_elem_as_dask(g["X"])
+    >>> adata.X
+    dask.array<make_dask_chunk, shape=(2700, 32738), dtype=float32, chunksize=(1000, 32738), chunktype=scipy.csr_matrix>
+    >>> adata.X = ad.experimental.read_elem_as_dask(
+    ...     g["X"], chunks=(500, adata.shape[1])
+    ... )
+    >>> adata.X
+    dask.array<make_dask_chunk, shape=(2700, 32738), dtype=float32, chunksize=(500, 32738), chunktype=scipy.csr_matrix>
+
+    Reading a dense matrix from a zarr store lazily:
+
+    >>> adata.layers["dense"] = ad.experimental.read_elem_as_dask(g["layers/dense"])
+    >>> adata.layers["dense"]
+    dask.array<from-zarr, shape=(2700, 32738), dtype=float32, chunksize=(169, 2047), chunktype=numpy.ndarray>
+
+    Making a new anndata object from on-disk, with custom chunks:
+
+    >>> adata = ad.AnnData(
+    ...     obs=ad.io.read_elem(g["obs"]),
+    ...     var=ad.io.read_elem(g["var"]),
+    ...     uns=ad.io.read_elem(g["uns"]),
+    ...     obsm=ad.io.read_elem(g["obsm"]),
+    ...     varm=ad.io.read_elem(g["varm"]),
+    ... )
+    >>> adata.X = ad.experimental.read_elem_as_dask(
+    ...     g["X"], chunks=(500, adata.shape[1])
+    ... )
+    >>> adata.layers["dense"] = ad.experimental.read_elem_as_dask(g["layers/dense"])
     """
     return DaskReader(_LAZY_REGISTRY).read_elem(elem, chunks=chunks)