[
https://issues.apache.org/jira/browse/FLINK-19065?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17188427#comment-17188427
]
Aljoscha Krettek commented on FLINK-19065:
------------------------------------------
Thanks! I found the culprit:
https://github.com/apache/flink/blob/11b2fcab5841e38d712e8e919c11e6ecd4f69a5d/flink-streaming-java/src/main/java/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java#L332.
The problem is that we're simulating a two-input operation by tagging elements
from the two sides and then treating them differently in the real (one-input)
operation. Even though I wrote this code (5 years ago) I don't like that we're
using user-facing methods internally like this.
A simple fix would be to set some UIDs on the operations that we create
internally but we have to make sure that they are actually unique.
> java.lang.IllegalStateException: Auto generated UIDs have been disabled on
> join
> -------------------------------------------------------------------------------
>
> Key: FLINK-19065
> URL: https://issues.apache.org/jira/browse/FLINK-19065
> Project: Flink
> Issue Type: Bug
> Components: API / DataStream
> Affects Versions: 1.11.0, 1.11.1
> Reporter: Maris
> Priority: Major
>
> Join operation with AutoGeneratedUID disabled leads to
> {code:java}
> java.lang.IllegalStateException: Auto generated UIDs have been disabled but
> no UID or hash has been assigned to operator Map
> {code}
> code to reproduce
> {code:java}
> class JoinSpec extends AnyFlatSpec with Matchers with Serializable {
> it should "be able to join streams" in {
> val env = StreamExecutionEnvironment.getExecutionEnvironment
> env.getConfig.disableAutoGeneratedUIDs()
> val a = env.fromCollection(List("1", "2",
> "3")).name("a").uid("source-uid")
> val b = env.fromCollection(List("1", "2",
> "3")).name("b").uid("source-uid2")
> val c = a
> .join(b)
> .where(identity)
> .equalTo(identity)
> .window(TumblingProcessingTimeWindows.of(Time.seconds(30)))((a, b) =>
> a+b)
> .uid("joined").name("joined")
> c.addSink(s => println(s))
> .name("ab")
> .uid("ab")
> println(env.getExecutionPlan)
> env.execute
> succeed
> }
> }
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
