Prevent pydap (dap4) to change string arrays to unicode type (testing). Fixed upstream (#10482)

Co-authored-by: Kai Mühlbauer <kai.muehlbauer@uni-bonn.de>

Author: Mikejmnez

ci/requirements/min-all-deps.yml

@@ -38,7 +38,7 @@ dependencies:
   - pandas=2.2
   - pint=0.24
   - pip
-  - pydap=3.5
+  - pydap=3.5.0
   - pytest
   - pytest-cov
   - pytest-env

doc/whats-new.rst

@@ -117,6 +117,8 @@ Bug fixes
   are passed (:pull:`10440`). By `Mathias Hauser <https://github.com/mathause>`_.
 - Fix :py:func:`testing.assert_equal` with ``check_dim_order=False`` for :py:class:`DataTree` objects
   (:pull:`10442`). By `Mathias Hauser <https://github.com/mathause>`_.
+- Fix Pydap backend testing. Now test forces string arrays to dtype "S" (pydap converts them to unicode type by default). Removes conditional to numpy version. (:issue:`10261`, :pull:`10482`)
+  By `Miguel Jimenez-Urias <https://github.com/Mikejmnez>`_.
 
 
 Documentation

xarray/tests/test_backends.py

@@ -5437,7 +5437,9 @@ def convert_to_pydap_dataset(self, original):
 
         ds = DatasetType("bears", **original.attrs)
         for key, var in original.data_vars.items():
-            ds[key] = BaseType(key, var.values, dims=var.dims, **var.attrs)
+            ds[key] = BaseType(
+                key, var.values, dtype=var.values.dtype.kind, dims=var.dims, **var.attrs
+            )
         # check all dims are stored in ds
         for d in original.coords:
             ds[d] = BaseType(d, original[d].values, dims=(d,), **original[d].attrs)
@@ -5449,9 +5451,9 @@ def create_datasets(self, **kwargs):
             # print("QQ0:", expected["bears"].load())
             pydap_ds = self.convert_to_pydap_dataset(expected)
             actual = open_dataset(PydapDataStore(pydap_ds))
-            if Version(np.__version__) < Version("2.3.0"):
-                # netcdf converts string to byte not unicode
-                expected["bears"] = expected["bears"].astype(str)
+            # netcdf converts string to byte not unicode
+            # fixed in pydap 3.5.6. https://github.com/pydap/pydap/issues/510
+            actual["bears"].values = actual["bears"].values.astype("S")
             yield actual, expected
 
     def test_cmp_local_file(self) -> None:
@@ -5495,9 +5497,6 @@ def test_compatible_to_netcdf(self) -> None:
             with create_tmp_file() as tmp_file:
                 actual.to_netcdf(tmp_file)
                 with open_dataset(tmp_file) as actual2:
-                    if Version(np.__version__) < Version("2.3.0"):
-                        # netcdf converts string to byte not unicode
-                        actual2["bears"] = actual2["bears"].astype(str)
                     assert_equal(actual2, expected)
 
     @requires_dask
@@ -5515,9 +5514,11 @@ def create_dap2_datasets(self, **kwargs):
         # in pydap 3.5.0, urls defaults to dap2.
         url = "http://test.opendap.org/opendap/data/nc/bears.nc"
         actual = open_dataset(url, engine="pydap", **kwargs)
+        # pydap <3.5.6 converts to unicode dtype=|U. Not what
+        # xarray expects. Thus force to bytes dtype. pydap >=3.5.6
+        # does not convert to unicode. https://github.com/pydap/pydap/issues/510
+        actual["bears"].values = actual["bears"].values.astype("S")
         with open_example_dataset("bears.nc") as expected:
-            # workaround to restore string which is converted to byte
-            expected["bears"] = expected["bears"].astype(str)
             yield actual, expected
 
     def output_grid_deprecation_warning_dap2dataset(self):
@@ -5530,7 +5531,8 @@ def create_dap4_dataset(self, **kwargs):
         actual = open_dataset(url, engine="pydap", **kwargs)
         with open_example_dataset("bears.nc") as expected:
             # workaround to restore string which is converted to byte
-            expected["bears"] = expected["bears"].astype(str)
+            # only needed for pydap <3.5.6 https://github.com/pydap/pydap/issues/510
+            expected["bears"].values = expected["bears"].values.astype("S")
             yield actual, expected
 
     def test_session(self) -> None: