Open GuyOmer opened 2 months ago
hey @GuyOmer , thanks for reaching out.
this is because greatest_ignore_nulls
is not implemented in the snowpark python library, for now the patch only works for functions available in snowflake.snowpark.functions
.
one workaround is to manually add a fake greatest_ignore_nulls
function into the functions module as below:
import snowflake.snowpark.functions
from snowflake.snowpark import Session, Column
from snowflake.snowpark._internal.type_utils import ColumnOrName
from snowflake.snowpark.functions import call_function
from snowflake.snowpark.mock import patch, ColumnEmulator, ColumnType
from snowflake.snowpark.types import IntegerType
session = Session.builder.configs({"local_testing": True}).create()
# this is just a placeholder, implementation does not matter
def fake_greatest_ignore_nulls(*columns: ColumnOrName) -> Column:
return None
@patch("greatest_ignore_nulls")
def mock_GREATEST_IGNORE_NULLS(*columns) -> ColumnEmulator:
return ColumnEmulator([1], sf_type=ColumnType(IntegerType(), nullable=False))
# dynamically add function to module so that local test can find it
snowflake.snowpark.functions.greatest_ignore_nulls = fake_greatest_ignore_nulls
df = session.create_dataframe(
[1, 2, 3, 4],
schema=["a"]
)
df.select(call_function("greatest_ignore_nulls", df["a"])).show()
I agree we should provide a better way to patch functions not defined in the functions.py module
What is the current behavior?
Trying to mock a non snowpark function (like
greatest_ignore_nulls
) results in an exception.What is the desired behavior?
Same mocking API.
If this is not an existing feature in
snowflake-snowpark-python
. How would this impact/improve non-local testing mode?This will be a complementary feature to using
call_function
.References, Other Background
Additionally, it is not possible to mock
call_function
.