[ 
https://issues.apache.org/jira/browse/TEZ-4498?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

László Bodor updated TEZ-4498:
------------------------------
    Description: 
This problem is present for a long time ago, we maintain 2 completely different 
groups of classes for shuffling:
UnorderedKVInput -> *ShuffleManager -> Fetcher*
OrderedGroupedKVInput -> *Shuffle -> ShuffleScheduler -> FetcherOrderedGrouped*

I don't think any of us would start to merge them in a single patch (and if so, 
we can wait for a complete review and fight with conflicts all the time), 
instead, we can make smaller refactors that make sense on their own.

(I'm 100% this is not the first ticket addressing this but I cannot find the 
earlier ones at the moment)

  was:
This problem is present for a long time ago, we maintain 2 completely different 
groups of classes for shuffling:
UnorderedKVInput -> *ShuffleManager -> Fetcher*
OrderedGroupedKVInput -> *Shuffle -> ShuffleScheduler -> FetcherOrderedGrouped*

I don't think any of us would start to merge them in a single patch, instead, 
we can make smaller refactors that makes sense on their own.

(I'm 100% this is not the first ticket addressing this but I cannot find the 
earlier ones)


> Refactor/merge ordered and unordered shuffle
> --------------------------------------------
>
>                 Key: TEZ-4498
>                 URL: https://issues.apache.org/jira/browse/TEZ-4498
>             Project: Apache Tez
>          Issue Type: Improvement
>            Reporter: László Bodor
>            Priority: Major
>
> This problem is present for a long time ago, we maintain 2 completely 
> different groups of classes for shuffling:
> UnorderedKVInput -> *ShuffleManager -> Fetcher*
> OrderedGroupedKVInput -> *Shuffle -> ShuffleScheduler -> 
> FetcherOrderedGrouped*
> I don't think any of us would start to merge them in a single patch (and if 
> so, we can wait for a complete review and fight with conflicts all the time), 
> instead, we can make smaller refactors that make sense on their own.
> (I'm 100% this is not the first ticket addressing this but I cannot find the 
> earlier ones at the moment)



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to