You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Nitin Pawar (JIRA)" <ji...@apache.org> on 2017/03/27 10:30:41 UTC

[jira] [Commented] (DRILL-5099) support OFFSET values greater than one, for LEAD & LAG window functions

    [ https://issues.apache.org/jira/browse/DRILL-5099?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15943019#comment-15943019 ] 

Nitin Pawar commented on DRILL-5099:
------------------------------------

[~khfaraaz] I am almost done with changes for this JIRA
I am currently stuck at the point when copyNext has argument from next partition. 

Can any dev from MapR spare 15 minutes to help me resolve the issue.

> support OFFSET values greater than one, for LEAD & LAG window functions
> -----------------------------------------------------------------------
>
>                 Key: DRILL-5099
>                 URL: https://issues.apache.org/jira/browse/DRILL-5099
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Execution - Flow
>    Affects Versions: 1.9.0
>            Reporter: Khurram Faraaz
>            Priority: Minor
>              Labels: window_function
>
> Provide support for OFFSET values greater than one, for LEAD & LAG window functions.
> Adding [~adeneche] comments from the dev list here
> {noformat}
> Things that need to be done to make Lag (or Lead) support offsets other
> than 1:
> - WindowFunction.Lead should extract the offset value from its FunctionCall
> argument, you can look at WindowFunctionNtile.numTilesFromExpression() for
> and example on how to do that.
> - make sure calls to copyNext() and copyPrev() in NoFrameSupportTemplate
> use the offset and not the hard coded value (you already figured that out)
> - finally make sure you update UnsupportedOperatorsVisitor to no longer
> throw an exception when we pass an offset value other than 1 to Lead or
> Lag. Just search for DRILL-3596 in that class and you will find the if
> block that need to be removed
> I think this should be enough to get it to work in the general case.
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)