BUG: Raise ValueError for non numerical join columns in merge_asof

doc/source/whatsnew/v1.1.0.rst

@@ -936,6 +936,7 @@ Reshaping
 - Bug in :meth:`DataFrame.replace` casts columns to ``object`` dtype if items in ``to_replace`` not in values (:issue:`32988`)
 - Ensure only named functions can be used in :func:`eval()` (:issue:`32460`)
 - Fixed bug in :func:`melt` where melting MultiIndex columns with ``col_level`` > 0 would raise a ``KeyError`` on ``id_vars`` (:issue:`34129`)
+- :meth:`merge_asof` raises ``ValueError`` instead of cryptic ``TypeError`` in case of non numerical merge columns (:issue:`29130`)
 
 Sparse
 ^^^^^^

pandas/core/reshape/merge.py

@@ -1730,6 +1730,9 @@ def flip(xs) -> np.ndarray:
         )
         tolerance = self.tolerance
 
+        if is_object_dtype(right_values.dtype) or is_object_dtype(left_values.dtype):
+            raise TypeError("Both inputs have to be from numeric dtype")
+
         # we require sortedness and non-null values in the join keys
         if not Index(left_values).is_monotonic:
             side = "left"

pandas/tests/reshape/merge/test_merge_asof.py

@@ -1339,3 +1339,15 @@ def test_merge_index_column_tz(self):
             index=pd.Index([0, 1, 2, 3, 4]),
         )
         tm.assert_frame_equal(result, expected)
+
+
+@pytest.mark.parametrize(
+    "data",
+    [["2019-06-01 00:09:12", "2019-06-01 00:10:29"], [1.0, "2019-06-01 00:10:29"]],
+)
+def test_non_numerical_dtype(data):
+    # GH 29130
+    left = pd.DataFrame({"x": data})
+    right = pd.DataFrame({"x": data})
+    with pytest.raises(TypeError, match="Both inputs have to be from numeric dtype"):
+        pd.merge_asof(left, right, on="x")