ci: division by zero test failing for pyspark #2753
Description
the test introduced in #2636 fails for pyspark, which isn't run by default
we should update the pyspark code to use try_divide
https://github.com/narwhals-dev/narwhals/actions/runs/15945539506/job/44979105403
_________________ test_floordiv_int_by_zero[pyspark-0-0-None] __________________
left = 0, right = 0, expected = None
constructor = <function pyspark_lazy_constructor.<locals>._constructor at 0x7fe8efc98180>
request = <FixtureRequest for <Function test_floordiv_int_by_zero[pyspark-0-0-None]>>
@pytest.mark.parametrize(
("left", "right", "expected"),
[(-2, 0, float("-inf")), (0, 0, None), (2, 0, float("inf"))],
)
@pytest.mark.skipif(PANDAS_VERSION < (2, 0), reason="converts floordiv by zero to 0")
def test_floordiv_int_by_zero(
left: int,
right: int,
expected: float | None,
constructor: Constructor,
request: pytest.FixtureRequest,
) -> None:
data: dict[str, list[int]] = {"a": [left]}
df = nw.from_native(constructor(data))
# pyarrow backend floordiv raises divide by zero error
# ibis backend floordiv cannot cast value to inf or -inf
if any(x in str(constructor) for x in ["ibis", "pyarrow"]):
request.applymarker(pytest.mark.xfail)
# duckdb backend floordiv return None
if "duckdb" in str(constructor):
floordiv_result = df.select(nw.col("a") // right)
assert_equal_data(floordiv_result, {"a": [None]})
# polars backend floordiv returns null
elif "polars" in str(constructor) and "lazy" not in str(constructor):
floordiv_result = df.select(nw.col("a") // right)
assert all(floordiv_result["a"].is_null())
# polars lazy floordiv cannot be sliced and returns None
elif all(x in str(constructor) for x in ["polars", "lazy"]):
floordiv_result = df.select(nw.col("a") // right)
assert_equal_data(floordiv_result, {"a": [None]})
# pandas[nullable] backend floordiv always returns 0
elif all(x in str(constructor) for x in ["pandas", "nullable"]):
floordiv_result = df.select(nw.col("a") // right)
assert_equal_data(floordiv_result, {"a": [0]})
else:
floordiv_result = df.select(nw.col("a") // right)
> assert_equal_data(floordiv_result, {"a": [expected]})
tests/expr_and_series/division_by_zero_test.py:115:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
tests/utils.py:92: in assert_equal_data
result = result.collect(**kwargs.get(result.implementation, {}))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
narwhals/dataframe.py:2336: in collect
self._compliant_frame.collect(backend=eager_backend, **kwargs), level="full"
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
narwhals/_spark_like/dataframe.py:234: in collect
self._collect_to_arrow(),
^^^^^^^^^^^^^^^^^^^^^^^^
narwhals/_spark_like/dataframe.py:196: in _collect_to_arrow
return self.native.toArrow()
^^^^^^^^^^^^^^^^^^^^^
/opt/hostedtoolcache/Python/3.11.13/x64/lib/python3.11/site-packages/pyspark/sql/classic/dataframe.py:1789: in toArrow
return PandasConversionMixin.toArrow(self)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
/opt/hostedtoolcache/Python/3.11.13/x64/lib/python3.11/site-packages/pyspark/sql/pandas/conversion.py:249: in toArrow
batches = self._collect_as_arrow(
/opt/hostedtoolcache/Python/3.11.13/x64/lib/python3.11/site-packages/pyspark/sql/pandas/conversion.py:315: in _collect_as_arrow
with unwrap_spark_exception():
/opt/hostedtoolcache/Python/3.11.13/x64/lib/python3.11/contextlib.py:158: in __exit__
self.gen.throw(typ, value, traceback)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
@contextmanager
def unwrap_spark_exception() -> Iterator[Any]:
from pyspark import SparkContext
from py4j.protocol import Py4JJavaError
from py4j.java_gateway import is_instance_of
assert SparkContext._gateway is not None
gw = SparkContext._gateway
try:
yield
except Py4JJavaError as e:
je: "Py4JJavaError" = e.java_exception
if je is not None and is_instance_of(gw, je, "org.apache.spark.SparkException"):
converted = convert_exception(je.getCause())
if not isinstance(converted, UnknownException):
> raise converted from None
E pyspark.errors.exceptions.captured.ArithmeticException: [DIVIDE_BY_ZERO] Division by zero. Use `try_divide` to tolerate divisor being 0 and return NULL instead. If necessary set "spark.sql.ansi.enabled" to "false" to bypass this error. SQLSTATE: 22012
E == DataFrame ==
E "__truediv__" was called from
E /home/runner/work/narwhals/narwhals/narwhals/_spark_like/expr.py:403