BUG: Take method of NumpyExtensionArray now returns another extension array with the correct dtype.

doc/source/whatsnew/v3.0.0.rst

@@ -1171,6 +1171,7 @@ ExtensionArray
 - Bug in :class:`Categorical` when constructing with an :class:`Index` with :class:`ArrowDtype` (:issue:`60563`)
 - Bug in :meth:`.arrays.ArrowExtensionArray.__setitem__` which caused wrong behavior when using an integer array with repeated values as a key (:issue:`58530`)
 - Bug in :meth:`ArrowExtensionArray.factorize` where NA values were dropped when input was dictionary-encoded even when dropna was set to False(:issue:`60567`)
+- Bug in :meth:`NDArrayBackedExtensionArray.take` which produced arrays whose dtypes didn't match their underlying data, when called with integer arrays (:issue:`62448`)
 - Bug in :meth:`api.types.is_datetime64_any_dtype` where a custom :class:`ExtensionDtype` would return ``False`` for array-likes (:issue:`57055`)
 - Bug in comparison between object with :class:`ArrowDtype` and incompatible-dtyped (e.g. string vs bool) incorrectly raising instead of returning all-``False`` (for ``==``) or all-``True`` (for ``!=``) (:issue:`59505`)
 - Bug in constructing pandas data structures when passing into ``dtype`` a string of the type followed by ``[pyarrow]`` while PyArrow is not installed would raise ``NameError`` rather than ``ImportError`` (:issue:`57928`)

pandas/core/arrays/_mixins.py

@@ -40,6 +40,7 @@
 from pandas.core.dtypes.dtypes import (
     DatetimeTZDtype,
     ExtensionDtype,
+    NumpyEADtype,
     PeriodDtype,
 )
 from pandas.core.dtypes.missing import array_equivalent
@@ -173,6 +174,17 @@ def take(
             fill_value=fill_value,
             axis=axis,
         )
+        if self.dtype in [
+            NumpyEADtype(np.uint8),
+            NumpyEADtype(np.uint16),
+            NumpyEADtype(np.uint32),
+            NumpyEADtype(np.uint64),
+            NumpyEADtype(np.int8),
+            NumpyEADtype(np.int16),
+            NumpyEADtype(np.int32),
+            NumpyEADtype(np.int64),
+        ]:
+            return type(self)(new_data)
         return self._from_backing_data(new_data)
 
     # ------------------------------------------------------------------------

pandas/tests/arrays/numpy_/test_numpy.py

@@ -325,6 +325,15 @@ def test_factorize_unsigned():
     tm.assert_extension_array_equal(res_unique, NumpyExtensionArray(exp_unique))
 
 
+@pytest.mark.parametrize("dtype", [np.uint32, np.uint64, np.int32, np.int64])
+def test_take_assigns_correct_dtype(dtype):
+    array = NumpyExtensionArray(np.array([1, 2, 3], dtype=dtype))
+
+    result = array.take([-1], allow_fill=True)
+
+    assert result.dtype == NumpyEADtype(np.float64)
+
+
 # ----------------------------------------------------------------------------
 # Output formatting