Use StringDType when reading from zarr string variables

doc/whats-new.rst

@@ -38,6 +38,8 @@ New Features
   By `Matthew Willson <https://github.com/mjwillson>`_.
 - Better ordering of coordinates when displaying Xarray objects. (:pull:`11098`).
   By `Ian Hunt-Isaak <https://github.com/ianhi>`_, `Julia Signell <https://github.com/jsignell>`_.
+- Use ``np.dtypes.StringDType`` when reading Zarr string variables  (:pull:`11097`).
+  By `Julia Signell <https://github.com/jsignell>`_.
 
 Breaking Changes
 ~~~~~~~~~~~~~~~~

xarray/backends/zarr.py

@@ -206,6 +206,10 @@ def __init__(self, zarr_array):
             not _zarr_v3()
             and self._array.filters is not None
             and any(filt.codec_id == "vlen-utf8" for filt in self._array.filters)
+        ) or (
+            _zarr_v3()
+            and self._array.serializer
+            and self._array.serializer.to_dict()["name"] == "vlen-utf8"
         ):
             dtype = coding.strings.create_vlen_dtype(str)
         else:

xarray/core/duck_array_ops.py

@@ -26,6 +26,7 @@
 
 from xarray.compat import dask_array_compat, dask_array_ops
 from xarray.compat.array_api_compat import get_array_namespace
+from xarray.compat.npcompat import HAS_STRING_DTYPE
 from xarray.core import dtypes, nputils
 from xarray.core.extension_array import (
     PandasExtensionArray,
@@ -175,9 +176,10 @@ def isnull(data):
         # note: must check timedelta64 before integers, because currently
         # timedelta64 inherits from np.integer
         return isnat(data)
+    elif HAS_STRING_DTYPE and isinstance(scalar_type, np.dtypes.StringDType):
+        return xp.isnan(data)
     elif dtypes.isdtype(scalar_type, ("real floating", "complex floating"), xp=xp):
         # float types use NaN for null
-        xp = get_array_namespace(data)
         return xp.isnan(data)
     elif dtypes.isdtype(scalar_type, ("bool", "integral"), xp=xp) or (
         isinstance(scalar_type, np.dtype)

xarray/tests/test_backends.py

@@ -58,6 +58,7 @@
 from xarray.coding.cftime_offsets import date_range
 from xarray.coding.strings import check_vlen_dtype, create_vlen_dtype
 from xarray.coding.variables import SerializationWarning
+from xarray.compat.npcompat import HAS_STRING_DTYPE
 from xarray.conventions import encode_dataset_coordinates
 from xarray.core import indexing
 from xarray.core.common import _contains_cftime_datetimes
@@ -1088,8 +1089,9 @@ def test_roundtrip_empty_vlen_string_array(self) -> None:
                 # eg. NETCDF3 based backends do not roundtrip metadata
                 if actual["a"].dtype.metadata is not None:
                     assert check_vlen_dtype(actual["a"].dtype) is str
+            elif HAS_STRING_DTYPE:
+                assert np.issubdtype(actual["a"].dtype, np.dtypes.StringDType())
             else:
-                # zarr v3 sends back "<U1"
                 assert np.issubdtype(actual["a"].dtype, np.dtype("=U1"))
 
     @pytest.mark.parametrize(