0
votes

I am running a SQL with multiple projections each are time consuming, for example:

SELECT UDF1(id), UDF2(id) 
FROM prod

The UDF1 and UDF2 may be time consuming function, but it looks like Flink SQL runs UDF1 and UDF2 sequentially, my question is whether UDF1 and UDF2 can be run in parallel to decrease latency?

1

1 Answers

0
votes

Evaluating scalar UDFs in parallel in different threads is not supported.

However, it might be an interesting feature. Please create a Jira issue or start a discussion on Flink's dev mailing list.