o Try to get it to work with fallback internal API and use default netcdf engine first for xarray loading - caused due to pydata/xarray#10755

Author: rajeeja

test/core/test_api.py

@@ -2,6 +2,10 @@
 import uxarray as ux
 import numpy as np
 import pytest
+import tempfile
+import xarray as xr
+from unittest.mock import patch
+from uxarray.core.utils import _open_dataset_with_fallback
 
 def test_open_geoflow_dataset(gridpath, datasetpath):
     """Loads a single dataset with its grid topology file using uxarray's
@@ -111,4 +115,39 @@ def test_open_dataset_grid_kwargs(gridpath, datasetpath):
             gridpath("ugrid", "outCSne30", "outCSne30.ug"),
             datasetpath("ugrid", "outCSne30", "outCSne30_var2.nc"),
             grid_kwargs={"drop_variables": "Mesh2_face_nodes"}
-        )
+                )
+
+
+def test_open_dataset_with_fallback():
+    """Test that the fallback mechanism works when the default engine fails."""
+
+    # Create a simple test dataset
+    with tempfile.NamedTemporaryFile(suffix='.nc', delete=False) as tmp:
+        data = xr.Dataset({'temp': (['x', 'y'], np.random.rand(5, 5))})
+        data.to_netcdf(tmp.name)
+        tmp_path = tmp.name
+
+    try:
+        # Test normal case
+        ds = _open_dataset_with_fallback(tmp_path)
+        assert isinstance(ds, xr.Dataset)
+        assert 'temp' in ds.data_vars
+
+        # Test fallback mechanism with mocked failure
+        original_open = xr.open_dataset
+        call_count = 0
+        def mock_open_dataset(*args, **kwargs):
+            nonlocal call_count
+            call_count += 1
+            if call_count == 1 and 'engine' not in kwargs:
+                raise Exception("Simulated engine failure")
+            return original_open(*args, **kwargs)
+
+        with patch('uxarray.core.utils.xr.open_dataset', side_effect=mock_open_dataset):
+            ds_fallback = _open_dataset_with_fallback(tmp_path)
+            assert isinstance(ds_fallback, xr.Dataset)
+            assert call_count == 2  # First failed, second succeeded
+
+    finally:
+        import os
+        os.unlink(tmp_path)

uxarray/core/api.py

@@ -7,7 +7,11 @@
 import numpy as np
 
 from uxarray.core.dataset import UxDataset
-from uxarray.core.utils import _map_dims_to_ugrid, match_chunks_to_ugrid
+from uxarray.core.utils import (
+    _map_dims_to_ugrid,
+    _open_dataset_with_fallback,
+    match_chunks_to_ugrid,
+)
 from uxarray.grid import Grid
 
 if TYPE_CHECKING:
@@ -178,8 +182,6 @@ def open_dataset(
     >>> import uxarray as ux
     >>> ux_ds = ux.open_dataset("grid_file.nc", "data_file.nc")
     """
-    import xarray as xr
-
     if grid_kwargs is None:
         grid_kwargs = {}
 
@@ -189,7 +191,7 @@ def open_dataset(
     )
 
     # Load the data as a Xarray Dataset
-    ds = xr.open_dataset(filename_or_obj, chunks=corrected_chunks, **kwargs)
+    ds = _open_dataset_with_fallback(filename_or_obj, chunks=corrected_chunks, **kwargs)
 
     # Map original dimensions to the UGRID conventions
     ds = _map_dims_to_ugrid(ds, uxgrid._source_dims_dict, uxgrid)

uxarray/core/dataset.py

@@ -14,7 +14,7 @@
 
 import uxarray
 from uxarray.core.dataarray import UxDataArray
-from uxarray.core.utils import _map_dims_to_ugrid
+from uxarray.core.utils import _map_dims_to_ugrid, _open_dataset_with_fallback
 from uxarray.formatting_html import dataset_repr
 from uxarray.grid import Grid
 from uxarray.grid.dual import construct_dual
@@ -326,7 +326,7 @@ def from_healpix(
         """
 
         if not isinstance(ds, xr.Dataset):
-            ds = xr.open_dataset(ds, **kwargs)
+            ds = _open_dataset_with_fallback(ds, **kwargs)
 
         if face_dim not in ds.dims:
             raise ValueError(

uxarray/core/utils.py

@@ -3,6 +3,41 @@
 from uxarray.io.utils import _get_source_dims_dict, _parse_grid_type
 
 
+def _open_dataset_with_fallback(filename_or_obj, chunks=None, **kwargs):
+    """Internal utility function to open datasets with fallback to netcdf4 engine.
+
+    Attempts to use Xarray's default read engine first, which may be "h5netcdf"
+    or "scipy" after v2025.09.0. If that fails (typically for h5-incompatible files),
+    falls back to using the "netcdf4" engine.
+
+    Parameters
+    ----------
+    filename_or_obj : str, Path, file-like or DataStore
+        Strings and Path objects are interpreted as a path to a netCDF file
+        or an OpenDAP URL and opened with python-netCDF4, unless the filename
+        ends with .gz, in which case the file is gunzipped and opened with
+        scipy.io.netcdf (only netCDF3 supported).
+    chunks : int, dict, 'auto' or None, optional
+        If chunks is provided, it is used to load the new dataset into dask
+        arrays.
+    **kwargs
+        Additional keyword arguments passed to xr.open_dataset
+
+    Returns
+    -------
+    xr.Dataset
+        The opened dataset
+    """
+    try:
+        # Try opening with xarray's default read engine
+        return xr.open_dataset(filename_or_obj, chunks=chunks, **kwargs)
+    except Exception:
+        # If it fails, use the "netcdf4" engine as backup
+        return xr.open_dataset(
+            filename_or_obj, engine="netcdf4", chunks=chunks, **kwargs
+        )
+
+
 def _map_dims_to_ugrid(
     ds,
     _source_dims_dict,
@@ -69,7 +104,7 @@ def match_chunks_to_ugrid(grid_filename_or_obj, chunks):
         # No need to rename
         return chunks
 
-    ds = xr.open_dataset(grid_filename_or_obj, chunks=chunks)
+    ds = _open_dataset_with_fallback(grid_filename_or_obj, chunks=chunks)
     grid_spec, _, _ = _parse_grid_type(ds)
 
     source_dims_dict = _get_source_dims_dict(ds, grid_spec)

uxarray/grid/grid.py

@@ -18,6 +18,9 @@
 
 from uxarray.constants import INT_FILL_VALUE
 from uxarray.conventions import ugrid
+
+# Import the utility function for opening datasets with fallback
+from uxarray.core.utils import _open_dataset_with_fallback
 from uxarray.cross_sections import GridCrossSectionAccessor
 from uxarray.formatting_html import grid_repr
 from uxarray.grid.area import get_all_face_area_from_coords
@@ -348,7 +351,7 @@ def from_file(
             grid_ds, source_dims_dict = _read_geodataframe(filename)
 
         elif backend == "xarray":
-            dataset = xr.open_dataset(filename, **kwargs)
+            dataset = _open_dataset_with_fallback(filename, **kwargs)
             return cls.from_dataset(dataset)
 
         else: