Reject functions without bytecode from _can_be_jitted in GroupBy Ap…

…ply (rapidsai#13429)

Closes rapidsai#13426

Authors:
  - https://github.com/brandon-b-miller

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)
  - Bradley Dice (https://github.com/bdice)

URL: rapidsai#13429

python/cudf/cudf/core/udf/groupby_utils.py

@@ -209,6 +209,10 @@ def _can_be_jitted(frame, func, args):
     by attempting to compile just the function to PTX using the
     target set of types
     """
+    if not hasattr(func, "__code__"):
+        # Numba requires bytecode to be present to proceed.
+        # See https://github.com/numba/numba/issues/4587
+        return False
     np_field_types = np.dtype(
         list(
             _supported_dtypes_from_frame(

python/cudf/cudf/tests/test_groupby.py

@@ -8,6 +8,7 @@
 import string
 import textwrap
 from decimal import Decimal
+from functools import partial
 
 import numpy as np
 import pandas as pd
@@ -583,6 +584,25 @@ def f(group):
     assert precompiled.currsize == 3
 
 
+def test_groupby_apply_no_bytecode_fallback():
+    # tests that a function which contains no bytecode
+    # attribute, but would still be executable using
+    # the iterative groupby apply approach, still works.
+
+    gdf = cudf.DataFrame({"a": [0, 1, 1], "b": [1, 2, 3]})
+    pdf = gdf.to_pandas()
+
+    def f(group):
+        return group.sum()
+
+    part = partial(f)
+
+    expect = pdf.groupby("a").apply(part)
+    got = gdf.groupby("a").apply(part, engine="auto")
+
+    assert_groupby_results_equal(expect, got)
+
+
 @pytest.mark.parametrize("func", [lambda group: group.x + group.y])
 def test_groupby_apply_return_col_from_df(func):
     # tests a UDF that consists of purely colwise