Backport PR #34733 on branch 1.0.x (BUG: Fixed Series.replace for EA with casting)

doc/source/whatsnew/v1.0.5.rst

@@ -24,6 +24,8 @@ Note this disables the ability to read Parquet files from directories on S3
 again (:issue:`26388`, :issue:`34632`), which was added in the 1.0.4 release,
 but is now targeted for pandas 1.1.0.
 
+- Fixed regression in :meth:`~DataFrame.replace` raising an ``AssertionError`` when replacing values in an extension dtype with values of a different dtype (:issue:`34530`)
+
 .. _whatsnew_105.bug_fixes:
 
 Bug fixes

pandas/core/internals/blocks.py

@@ -779,7 +779,11 @@ def replace(
             if is_object_dtype(self):
                 raise
 
-            assert not self._can_hold_element(value), value
+            if not self.is_extension:
+                # TODO: https://github.com/pandas-dev/pandas/issues/32586
+                # Need an ExtensionArray._can_hold_element to indicate whether
+                # a scalar value can be placed in the array.
+                assert not self._can_hold_element(value), value
 
             # try again with a compatible block
             block = self.astype(object)

pandas/tests/series/methods/test_replace.py

@@ -362,3 +362,8 @@ def test_replace_no_cast(self, ser, exp):
         expected = pd.Series(exp)
 
         tm.assert_series_equal(result, expected)
+
+    def test_replace_extension_other(self):
+        # https://github.com/pandas-dev/pandas/issues/34530
+        ser = pd.Series(pd.array([1, 2, 3], dtype="Int64"))
+        ser.replace("", "")  # no exception