Liu Liu created FLINK-41253:
-------------------------------
Summary: Python scalar UDFs produce incorrectly ordered columns
when mixing sync and async calls
Key: FLINK-41253
URL: https://issues.apache.org/jira/browse/FLINK-41253
Project: Flink
Issue Type: Bug
Components: API / Python
Affects Versions: 2.3.0
Reporter: Liu Liu
Mixing synchronous and asynchronous Python scalar UDFs in a projection can
produce columns in the wrong order. The issue affects both streaming and batch
execution.
For the minimal example below:
{code:java}
from pyflink.table import DataTypes, EnvironmentSettings, TableEnvironment
from pyflink.table.udf import udf
@udf(result_type=DataTypes.BIGINT())
def sync_add(value):
return value + 1
@udf(result_type=DataTypes.BIGINT())
async def async_add(value):
return value + 2
table = env.from_elements(
[(1,)],
DataTypes.ROW([DataTypes.FIELD("a", DataTypes.BIGINT())]),
)
with table.select(sync_add(table.a), async_add(table.a)).execute().collect() as
rows:
print([tuple(row) for row in rows]){code}
* {{select(sync_add(a), async_add(a))}} should return {{{}(2, 3){}}}, but
returns {{{}(3, 2){}}}.
* {{select(async_add(a), sync_add(a))}} correctly returns {{{}(3, 2){}}}.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
