Fix

Author: dcherian

xarray/core/extension_array.py

@@ -144,4 +144,11 @@ def __array__(
             return np.asarray(self.array, dtype=dtype)
 
     def __getattr__(self, attr: str) -> Any:
-        return getattr(self.array, attr)
+        #  without __deepcopy__ or __copy__, the object is first constructed and then the sub-objects are attached:
+        # "A shallow copy constructs a new compound object and then (to the extent possible) inserts references into it to the objects found in the original" +
+        # "A deep copy constructs a new compound object and then, recursively, inserts copies into it of the objects found in the original."
+        # So without array, this method then calls getattr for "array" again while looking for __setstate__
+        # (which is apparently the first thing sought in copy.copy from the under-construction copied object)
+        # because of self.array. __getattribute__ bypasses the lookup mechanism of __getattr__.
+        # See stackoverflow.com/questions/40583131/python-deepcopy-with-custom-getattr-and-setattr as well.
+        return getattr(super().__getattribute__("array"), attr)

xarray/tests/test_duck_array_ops.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import datetime as dt
+import pickle
 import warnings
 
 import numpy as np
@@ -1098,4 +1099,8 @@ def test_extension_array_repr(int1):
 
 def test_extension_array_attr():
     array = pd.Categorical(["cat2", "cat1", "cat2", "cat3", "cat1"])
-    assert (array.categories == PandasExtensionArray(array).categories).all()
+    wrapped = PandasExtensionArray(array)
+    assert (array.categories == wrapped.categories).all()
+    assert array.nbytes == wrapped.nbytes
+
+    assert (pickle.loads(pickle.dumps(wrapped)) == wrapped).all()