adding conversion to tensorstore

iohub/ngff/nodes.py

@@ -9,6 +9,7 @@
 import math
 import os
 from copy import deepcopy
+from pathlib import Path
 from typing import TYPE_CHECKING, Generator, Literal, Sequence, Type
 
 import numpy as np
@@ -344,7 +345,27 @@ def downscale(self):
         raise NotImplementedError
 
     def tensorstore(self):
-        raise NotImplementedError
+        import tensorstore as ts
+
+        metadata = {
+            "dtype": self.dtype.str,
+            "shape": self.shape,
+            "chunks": self.chunks,
+        }
+        ts_spec = {
+            "driver": "zarr",
+            "kvstore": {
+                "driver": "file",
+                "path": str((Path(self._store.path) / self.path).resolve()),
+            },
+            "metadata": metadata,
+        }
+        try:
+            zarr_dataset = ts.open(ts_spec, open=True).result()
+        except ValueError as e:
+            print(f"Error opening Zarr store: {e}")
+            raise
+        return zarr_dataset
 
 
 class TiledImageArray(ImageArray):

setup.cfg

@@ -45,6 +45,9 @@ install_requires =
     xarray>=2024.1.1
 
 [options.extras_require]
+
+tensorstore=
+    tensorstore>=0.1.64
 dev =
     black
     flake8
@@ -53,7 +56,8 @@ dev =
     hypothesis>=6.61.0
     requests>=2.22.0
     wget>=3.2
-    ome-zarr>=0.9.0
+    ome-zarr>=0.9.
+
 doc =
     matplotlib
     numpydoc>=1.1.0

tests/ngff/test_ngff.py

@@ -250,6 +250,28 @@ def test_position_data(channels_and_random_5d, arr_name):
             _ = dataset.data
 
 
+@given(
+    channels_and_random_5d=_channels_and_random_5d(),
+    arr_name=short_alpha_numeric,
+)
+@settings(
+    max_examples=16,
+    deadline=2000,
+    suppress_health_check=[HealthCheck.data_too_large],
+)
+def test_ome_zarr_to_tensorstore(channels_and_random_5d, arr_name):
+    """Test `iohub.ngff.Position.data` to tensortore"""
+    pytest.importorskip("tensorstore")
+    channel_names, random_5d = channels_and_random_5d
+    with _temp_ome_zarr(random_5d, channel_names, "0") as dataset:
+        assert_array_almost_equal(dataset.data.numpy(), random_5d)
+    with pytest.raises(KeyError):
+        with _temp_ome_zarr(random_5d, channel_names, arr_name) as dataset:
+            t = dataset.data.tensorstore()
+            t.read().result()
+            del t
+
+
 @given(
     channels_and_random_5d=_channels_and_random_5d(),
     arr_name=short_alpha_numeric,