[WIP] adding an option to read file from s3

anndata/_core/file_backing.py

@@ -6,6 +6,7 @@
 from typing import TYPE_CHECKING, Literal
 
 import h5py
+import remfile
 
 from ..compat import AwkArray, DaskArray, ZarrArray, ZarrGroup
 from .sparse_dataset import BaseCompressedSparseDataset
@@ -65,7 +66,13 @@ def filename(self) -> Path:
 
     @filename.setter
     def filename(self, filename: PathLike | None):
-        self._filename = None if filename is None else Path(filename)
+        self._filename = (
+            None
+            if filename is None
+            else filename
+            if isinstance(filename, remfile.RemFile.RemFile)
+            else Path(filename)
+        )
 
     def open(
         self,

anndata/tests/test_readwrite.py

@@ -797,3 +797,19 @@ def test_io_dtype(tmp_path, diskfmt, dtype):
     curr = read(pth)
 
     assert curr.X.dtype == dtype
+
+
+needs_remfile = pytest.mark.skipif(
+    not find_spec("remfile"), reason="remfile is not installed"
+)
+
+
+@needs_remfile
+def test_read_s3_remfile():
+    import remfile
+
+    # TODO: change to a different datafile?
+    url_had = "https://allen-brain-cell-atlas.s3.us-west-2.amazonaws.com/expression_matrices/WMB-10Xv2/20230630/WMB-10Xv2-TH-log2.h5ad"
+    file_had = remfile.File(url_had)
+    data = ad.read_h5ad(file_had, backed="r")
+    assert data.X.shape == (131212, 32285)

pyproject.toml

@@ -43,6 +43,7 @@ dependencies = [
     "numpy>=1.16.5",                         # required by pandas 1.x
     "scipy>1.4",
     "h5py>=3",
+    "remfile",
     "exceptiongroup; python_version<'3.11'",
     "natsort",
     "packaging>=20",