Fixes to dataset equivalence testing on xarray loads.

docs/changelog_fragments/195.dev.rst

@@ -0,0 +1 @@
+Fixed xarray load tests for new behaviour of xarray.Dataset.identical.

tests/integration/test_xarray_load_and_save_equivalence.py

@@ -6,6 +6,7 @@
 (2) check equivalence of files : xarray -> file VS xarray->ncdata->file
 """
 
+import numpy as np
 import pytest
 import xarray
 from ncdata.netcdf4 import from_nc4, to_nc4
@@ -37,6 +38,35 @@ def use_xarraylock():
         yield
 
 
+def equivalence_fix_datasets(
+    ds_from: xarray.Dataset, ds_to: xarray.Dataset
+) -> (xarray.Dataset, xarray.Dataset):
+    """
+    Modify datasets in legitimate ways to make "ds_from.identical(ds_to)".
+
+    The key differences are due to coordinates remaining lazy in loading via ncdata, but
+    have data fetched in the "normal" load.
+    The coordinates apparently remain 'identical', but it affects the dataset indexes.
+
+    Minimum found necessary : where in 'ds_from' we find a lazy coordinate, which is a
+    real one in 'ds_to', remove the associated index from 'ds_to'.
+    """
+    drop_indices = []
+    for varname, var in ds_from.variables.items():
+        if hasattr(var.data, "compute"):
+            var_other = ds_to.variables.get(varname, None)
+            if isinstance(var_other.data, np.ndarray):
+                # This is lazy, but the reference var is real :  replace with real data.
+                if varname in ds_to.indexes:
+                    drop_indices.append(varname)
+
+    # NB drop_indexes is *not* an inplace operation!
+    # So replace returned 'ds_to' with new dataset.
+    ds_to = ds_to.drop_indexes(drop_indices)
+    # NB: as it currently is, we do *not* ever have to modify/replace 'ds_from'.
+    return ds_from, ds_to
+
+
 def test_load_direct_vs_viancdata(standard_testcase, use_xarraylock, tmp_path):
     source_filepath = standard_testcase.filepath
     ncdata = from_nc4(source_filepath)
@@ -51,7 +81,15 @@ def test_load_direct_vs_viancdata(standard_testcase, use_xarraylock, tmp_path):
     # Load same, via ncdata
     xr_ncdata_ds = to_xarray(ncdata)
 
-    # Treat as OK if it passes xarray comparison
+    # Check that datasets are "equal" : but NB this only compares values
+    assert xr_ds.equals(xr_ncdata_ds)
+
+    # 'Fix' equivalence, by making lazy vars real + removing missing indices.
+    # These are the expected differences due to ncdata passing lazy arrays.
+    # This should then make "Dataset.identical" true.
+    xr_ncdata_ds, xr_ds = equivalence_fix_datasets(
+        ds_from=xr_ncdata_ds, ds_to=xr_ds
+    )
     assert xr_ds.identical(xr_ncdata_ds)