[ https://issues.apache.org/jira/browse/FLINK-25683?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17479152#comment-17479152 ]
Dawid Wysakowicz commented on FLINK-25683: ------------------------------------------ Sorry, I have not spotted that before. Actually, I am not sure anymore, the described situation is the root problem. In pure {{DataStream}} program, the trailing keys are supposed to be flushed by the {{MAX_WATERMARK}} emitted from sources. However, as I looked into the code in the {{InputConversionOperator}} we swallow all watermarks for the given program. See {{InputConversionOperator#processWatermark}}. [~twalthr] As I am not 100% familiar with that logic. Would it be possible to forward at least the {{MAX_WATERMARK}} there? > wrong result if table transfrom to DataStream then window process in batch > mode > ------------------------------------------------------------------------------- > > Key: FLINK-25683 > URL: https://issues.apache.org/jira/browse/FLINK-25683 > Project: Flink > Issue Type: Bug > Components: Table SQL / API, Table SQL / Runtime > Affects Versions: 1.14.2 > Environment: mac book pro m1 > jdk 8 > scala 2.11 > flink 1.14.2 > idea 2020 > Reporter: zhangzh > Assignee: Yao Zhang > Priority: Major > Labels: pull-request-available > Attachments: TableToDataStreamBatchWindowTest.scala, pom.xml > > > I have 5 line datas, > i first need to transform current data with SQL > then mix current data and historical data which is batch get from hbase > for some special reason the program must run in batch mode > i think the correct result should be like this: > (BOB,1) > (EMA,1) > (DOUG,1) > (ALICE,1) > (CENDI,1) > but the result is : > (EMA,1) > > if i set different parallelism ,the result is different. -- This message was sent by Atlassian Jira (v8.20.1#820001)