[
https://issues.apache.org/jira/browse/CRUNCH-501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14334557#comment-14334557
]
Gabriel Reid commented on CRUNCH-501:
-------------------------------------
+1 to the patch. It seems unfortunate that Aggregators.toCombineFn is public
(and so necessitates deprecation), but I guess there's no easy way around that.
About providing the PType for every DoFn, one problem I can see is that for
shared instances of DoFns (Identity.getInstance for example), there would be an
issue because multiple PTypes are used by the same instance. I'm not sure how
often that would come up, but at least in terms of IdentityFn I believe it
would break.
[~aj987] is the use case for having access to the PType only related to
detaching values, or are there other reasons to need access to a PType in a
DoFn?
> Object reuse issue in combineValues(Aggregator)
> -----------------------------------------------
>
> Key: CRUNCH-501
> URL: https://issues.apache.org/jira/browse/CRUNCH-501
> Project: Crunch
> Issue Type: Bug
> Reporter: Brandon Davis
> Attachments: CRUNCH-501.patch
>
>
> I'm trying to use combineValues on a PGroupedTable. I am using
> Aggregators.FIRST_N. If I have 20 keys in my PGroupedTable, then I only get
> 20 distinct values because the AggregatorCombineFn and FirstNAggregator don't
> detach the values from the incoming iterator.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
