flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-4937) Add incremental group window aggregation for streaming Table API
Date Sun, 13 Nov 2016 09:14:59 GMT

    [ https://issues.apache.org/jira/browse/FLINK-4937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15661167#comment-15661167
] 

ASF GitHub Bot commented on FLINK-4937:
---------------------------------------

Github user wuchong commented on a diff in the pull request:

    https://github.com/apache/flink/pull/2792#discussion_r87708469
  
    --- Diff: flink-libraries/flink-table/src/main/scala/org/apache/flink/api/table/runtime/aggregate/AggregateUtil.scala
---
    @@ -61,25 +61,108 @@ object AggregateUtil {
        * }}}
        *
        */
    -  def createOperatorFunctionsForAggregates(
    +    def createOperatorFunctionsForAggregates(
    --- End diff --
    
    It is a little confused that what's the difference between these two `createOperatorFunctionsForAggregates`
.  
    
    Could we remove the first one, and call `transformToAggregateFunctions` explicitly before
`createOperatorFunctionsForAggregates` in DataSetAggregate ? 


> Add incremental group window aggregation for streaming Table API
> ----------------------------------------------------------------
>
>                 Key: FLINK-4937
>                 URL: https://issues.apache.org/jira/browse/FLINK-4937
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Table API & SQL
>    Affects Versions: 1.2.0
>            Reporter: Fabian Hueske
>            Assignee: sunjincheng
>
> Group-window aggregates for streaming tables are currently not done in an incremental
fashion. This means that the window collects all records and performs the aggregation when
the window is closed instead of eagerly updating a partial aggregate for every added record.
Since records are buffered, non-incremental aggregation requires more storage space than incremental
aggregation.
> The DataStream API which is used under the hood of the streaming Table API features [incremental
aggregation|https://ci.apache.org/projects/flink/flink-docs-release-1.2/dev/windows.html#windowfunction-with-incremental-aggregation]
using a {{ReduceFunction}}.
> We should add support for incremental aggregation in group-windows.
> This is a follow-up task of FLINK-4691.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message