BUG: Raise TypeError for scalar indexer in take

doc/source/whatsnew/v1.4.0.rst

@@ -230,6 +230,7 @@ Other enhancements
 - Implemented :meth:`IntervalArray.min`, :meth:`IntervalArray.max`, as a result of which ``min`` and ``max`` now work for :class:`IntervalIndex`, :class:`Series` and :class:`DataFrame` with ``IntervalDtype`` (:issue:`44746`)
 - :meth:`UInt64Index.map` now retains ``dtype`` where possible (:issue:`44609`)
 - :meth:`read_json` can now parse unsigned long long integers (:issue:`26068`)
+- :meth:`DataFrame.take` now raises a ``TypeError`` when passed a scalar for the indexer (:issue:`42875`)
 -
 
 

pandas/core/indexes/base.py

@@ -1066,6 +1066,8 @@ def take(
     ):
         if kwargs:
             nv.validate_take((), kwargs)
+        if is_scalar(indices):
+            raise TypeError("Expected indices to be array-like")
         indices = ensure_platform_int(indices)
         allow_fill = self._maybe_disallow_fill(allow_fill, fill_value, indices)
 

pandas/tests/indexes/test_indexing.py

@@ -70,6 +70,14 @@ def test_take(self, index):
             with pytest.raises(AttributeError, match=msg):
                 index.freq
 
+    def test_take_indexer_type(self):
+        # GH#42875
+        integer_index = Index([0, 1, 2, 3])
+        scalar_index = 1
+        msg = "Expected indices to be array-like"
+        with pytest.raises(TypeError, match=msg):
+            integer_index.take(scalar_index)
+
     def test_take_minus1_without_fill(self, index):
         # -1 does not get treated as NA unless allow_fill=True is passed
         if len(index) == 0: