[
https://issues.apache.org/jira/browse/SPARK-61403?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jungtaek Lim resolved SPARK-61403.
----------------------------------
Fix Version/s: 4.5.0
Resolution: Fixed
Issue resolved by pull request 60313
[https://github.com/apache/spark/pull/60313]
> Support external lookup joins in real-time mode
> -----------------------------------------------
>
> Key: SPARK-61403
> URL: https://issues.apache.org/jira/browse/SPARK-61403
> Project: Spark
> Issue Type: Task
> Components: SQL, Structured Streaming
> Affects Versions: 4.4.0
> Reporter: Jungtaek Lim
> Assignee: Jungtaek Lim
> Priority: Major
> Labels: pull-request-available
> Fix For: 4.5.0
>
>
> Part of SPARK-60031. Real-time mode rejects the external lookup join today,
> because `ExternalLookupJoinExec` is not in `RealTimeModeAllowlist`.
> In scope:
> - Allow `ExternalLookupJoinExec` in real-time mode. The operator is
> stateless, runs in its input's task and keeps its input's determinism level,
> so a stateful operator after it can read it through a pipelined shuffle.
> - Tests in real-time mode: inner, left outer, left semi and left anti joins;
> the cache within a batch and its reset at the next batch; a streaming
> aggregation after the join; a shuffle on the stream side before it; a stop
> during a lookup request.
> - Known limitations, pinned by a test and documented:
> - a stream row waits until its lookup request is full or the batch ends;
> - the output of a full request waits for the next stream row;
> - a cache hit waits the same way;
> - the cache starts empty at every batch.
> Out of scope:
> - A time-based flush of a lookup request that is not full.
> - Asynchronous lookup requests.
> - A cache that lives across batches.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]