Bump the minimum version of h5netcdf to 1.4.0 in accordance with nep29

doc/whats-new.rst

@@ -23,6 +23,10 @@ New Features
   :py:class:`~xarray.indexes.PandasIndex` to perform the selection
   (:issue:`9703`, :pull:`11029`).
   By `Ian Hunt-Isaak <https://github.com/ianhi>`_.
+- The minimum supported version of ``h5netcdf`` is now 1.4. Version 1.4.0
+  brings improved alignment between h5netcdf and libnetcdf4 in the storage of
+  complex numbers (:pull:`11068`). By `Mark Harfouche
+  <https://github.com/hmaarrfk>`_.
 
 
 Breaking Changes

pixi.toml

@@ -117,7 +117,7 @@ cftime = "1.6.*"
 dask-core = "2024.6.*"
 distributed = "2024.6.*"
 flox = "0.9.*"
-h5netcdf = "1.3.*"
+h5netcdf = "1.4.*"
 # h5py and hdf5 tend to cause conflicts
 # for e.g. hdf5 1.12 conflicts with h5py=3.1
 # prioritize bumping other packages instead

pyproject.toml

@@ -37,7 +37,7 @@ accel = [
 complete = ["xarray[accel,etc,io,parallel,viz]"]
 io = [
   "netCDF4>=1.6.0",
-  "h5netcdf",
+  "h5netcdf>=1.4.0",
   "pydap",
   "scipy>=1.13",
   "zarr>=2.18",

xarray/tests/__init__.py

@@ -230,10 +230,6 @@ def _importorskip_h5netcdf_ros3(has_h5netcdf: bool):
     "netCDF4", "1.6.2"
 )
 
-has_h5netcdf_1_4_0_or_above, requires_h5netcdf_1_4_0_or_above = _importorskip(
-    "h5netcdf", "1.4.0.dev"
-)
-
 has_h5netcdf_1_7_0_or_above, requires_h5netcdf_1_7_0_or_above = _importorskip(
     "h5netcdf", "1.7.0.dev"
 )

xarray/tests/test_backends.py

@@ -74,7 +74,6 @@
     assert_identical,
     assert_no_warnings,
     has_dask,
-    has_h5netcdf_1_4_0_or_above,
     has_netCDF4,
     has_numpy_2,
     has_scipy,
@@ -89,7 +88,6 @@
     requires_dask,
     requires_fsspec,
     requires_h5netcdf,
-    requires_h5netcdf_1_4_0_or_above,
     requires_h5netcdf_1_7_0_or_above,
     requires_h5netcdf_or_netCDF4,
     requires_h5netcdf_ros3,
@@ -2124,20 +2122,14 @@ def test_encoding_enum__no_fill_value(self, recwarn):
                 )
                 v[:] = 1
             with open_dataset(tmp_file, engine="netcdf4") as original:
-                save_kwargs = {}
                 # We don't expect any errors.
                 # This is effectively a void context manager
                 expected_warnings = 0
                 if self.engine == "h5netcdf":
-                    if not has_h5netcdf_1_4_0_or_above:
-                        save_kwargs["invalid_netcdf"] = True
-                        expected_warnings = 1
-                        expected_msg = "You are writing invalid netcdf features to file"
-                    else:
-                        expected_warnings = 1
-                        expected_msg = "Creating variable with default fill_value 0 which IS defined in enum type"
-
-                with self.roundtrip(original, save_kwargs=save_kwargs) as actual:
+                    expected_warnings = 1
+                    expected_msg = "Creating variable with default fill_value 0 which IS defined in enum type"
+
+                with self.roundtrip(original) as actual:
                     assert len(recwarn) == expected_warnings
                     if expected_warnings:
                         assert issubclass(recwarn[0].category, UserWarning)
@@ -2147,14 +2139,6 @@ def test_encoding_enum__no_fill_value(self, recwarn):
                         actual.clouds.encoding["dtype"].metadata["enum"]
                         == cloud_type_dict
                     )
-                    if not (
-                        self.engine == "h5netcdf" and not has_h5netcdf_1_4_0_or_above
-                    ):
-                        # not implemented in h5netcdf yet
-                        assert (
-                            actual.clouds.encoding["dtype"].metadata["enum_name"]
-                            == "cloud_type"
-                        )
 
     @requires_netCDF4
     def test_encoding_enum__multiple_variable_with_enum(self):
@@ -2176,10 +2160,7 @@ def test_encoding_enum__multiple_variable_with_enum(self):
                     fill_value=255,
                 )
             with open_dataset(tmp_file, engine="netcdf4") as original:
-                save_kwargs = {}
-                if self.engine == "h5netcdf" and not has_h5netcdf_1_4_0_or_above:
-                    save_kwargs["invalid_netcdf"] = True
-                with self.roundtrip(original, save_kwargs=save_kwargs) as actual:
+                with self.roundtrip(original) as actual:
                     assert_equal(original, actual)
                     assert (
                         actual.clouds.encoding["dtype"] == actual.tifa.encoding["dtype"]
@@ -2192,14 +2173,6 @@ def test_encoding_enum__multiple_variable_with_enum(self):
                         actual.clouds.encoding["dtype"].metadata["enum"]
                         == cloud_type_dict
                     )
-                    if not (
-                        self.engine == "h5netcdf" and not has_h5netcdf_1_4_0_or_above
-                    ):
-                        # not implemented in h5netcdf yet
-                        assert (
-                            actual.clouds.encoding["dtype"].metadata["enum_name"]
-                            == "cloud_type"
-                        )
 
     @requires_netCDF4
     def test_encoding_enum__error_multiple_variable_with_changing_enum(self):
@@ -2235,17 +2208,6 @@ def test_encoding_enum__error_multiple_variable_with_changing_enum(self):
                     "u1",
                     metadata={"enum": modified_enum, "enum_name": "cloud_type"},
                 )
-                if not (self.engine == "h5netcdf" and not has_h5netcdf_1_4_0_or_above):
-                    # not implemented yet in h5netcdf
-                    with pytest.raises(
-                        ValueError,
-                        match=(
-                            r"Cannot save variable .*"
-                            r" because an enum `cloud_type` already exists in the Dataset .*"
-                        ),
-                    ):
-                        with self.roundtrip(original):
-                            pass
 
     @pytest.mark.parametrize("create_default_indexes", [True, False])
     def test_create_default_indexes(self, tmp_path, create_default_indexes) -> None:
@@ -4927,31 +4889,6 @@ def create_store(self):
         with create_tmp_file() as tmp_file:
             yield backends.H5NetCDFStore.open(tmp_file, "w")
 
-    @pytest.mark.skipif(
-        has_h5netcdf_1_4_0_or_above, reason="only valid for h5netcdf < 1.4.0"
-    )
-    def test_complex(self) -> None:
-        expected = Dataset({"x": ("y", np.ones(5) + 1j * np.ones(5))})
-        save_kwargs = {"invalid_netcdf": True}
-        with pytest.warns(UserWarning, match="You are writing invalid netcdf features"):
-            with self.roundtrip(expected, save_kwargs=save_kwargs) as actual:
-                assert_equal(expected, actual)
-
-    @pytest.mark.skipif(
-        has_h5netcdf_1_4_0_or_above, reason="only valid for h5netcdf < 1.4.0"
-    )
-    @pytest.mark.parametrize("invalid_netcdf", [None, False])
-    def test_complex_error(self, invalid_netcdf) -> None:
-        import h5netcdf
-
-        expected = Dataset({"x": ("y", np.ones(5) + 1j * np.ones(5))})
-        save_kwargs = {"invalid_netcdf": invalid_netcdf}
-        with pytest.raises(
-            h5netcdf.CompatibilityError, match="are not a supported NetCDF feature"
-        ):
-            with self.roundtrip(expected, save_kwargs=save_kwargs) as actual:
-                assert_equal(expected, actual)
-
     def test_numpy_bool_(self) -> None:
         # h5netcdf loads booleans as numpy.bool_, this type needs to be supported
         # when writing invalid_netcdf datasets in order to support a roundtrip
@@ -5105,7 +5042,6 @@ def test_byte_attrs(self, byte_attrs_dataset: dict[str, Any]) -> None:
         with pytest.raises(ValueError, match=byte_attrs_dataset["h5netcdf_error"]):
             super().test_byte_attrs(byte_attrs_dataset)
 
-    @requires_h5netcdf_1_4_0_or_above
     def test_roundtrip_complex(self):
         expected = Dataset({"x": ("y", np.ones(5) + 1j * np.ones(5))})
         with self.roundtrip(expected) as actual: