Test some ways to unstructure xarray data

test/test_cattrs.py

@@ -0,0 +1,58 @@
+import numpy as np
+import xarray as xr
+from attrs import define, field
+from cattrs import Converter, unstructure
+from numpy.typing import NDArray
+
+
+@define
+class Foo:
+    x: NDArray[np.float32] = field()
+
+
+@define
+class Bar:
+    x: xr.DataArray = field()
+
+
+@define
+class Baz:
+    x: xr.DataTree = field()
+
+
+def test_unstructure_numpy_array():
+    np_arr = np.array([1.0, 2.0, 3.0])
+    f = Foo(x=np_arr)
+    f_dict = unstructure(f)
+
+    # We expect that the default unstructure functionality keeps the numpy array as is.
+    # This helps when finally converting the dictionary to MF6 input files.
+    assert np_arr is f_dict["x"]
+
+
+def test_unstructure_xarray():
+    x_arr = xr.DataArray([1, 2, 3])
+    f = Bar(x=x_arr)
+    f_dict = unstructure(f)
+
+    # We expect that the default unstructure functionality keeps the xarray as is.
+    # This helps when finally converting the dictionary to MF6 input files.
+    assert x_arr is f_dict["x"]
+
+
+def test_unstructure_xarray_tree_to_ascii():
+    x_arr = xr.DataArray([1, 2, 3])
+    x_set = xr.Dataset({"x": x_arr})
+    x_tree = xr.DataTree(x_set)
+    f = Baz(x=x_tree)
+
+    converter = Converter()
+    converter.register_unstructure_hook(
+        Baz, lambda b: " ".join(b.x["x"].data.astype(str))
+    )
+    f_dict = converter.unstructure(f)
+
+    # The data is formatted in ascii format.
+    # The whole string will be in memory when doing this,
+    # duplicating the data.
+    assert f_dict == "1 2 3"