Avery Qi created SPARK-59681:
--------------------------------
Summary: Guaranteed shuffle reuse for materialized/not inlined CTE
Key: SPARK-59681
URL: https://issues.apache.org/jira/browse/SPARK-59681
Project: Spark
Issue Type: Task
Components: Optimizer
Affects Versions: 4.4.0
Reporter: Avery Qi
A CTE referenced multiple times can be either inlined (recomputed per
reference) or materialized once. For a *non-deterministic* body (e.g.
{{{}rand(){}}}, {{{}monotonically_increasing_id(){}}}, non-deterministic UDFs),
recomputation is not just wasteful but *semantically wrong* — each reference
would observe different rows.
Currently spark already has CTE which is marked as `materialized` or
`forceSkipInline` but the CTE is being changed into Repartition over subplans.
Repartition leads to a normal shuffle which can be changed or removed in the
physical planning stages.
This task provide a guaranteed shuffle reuse for CTE which is marked as
`forceSkipInline`.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
