phoenix-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Lars Hofhansl (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PHOENIX-5123) Avoid using MappedByteBuffers for server side GROUP BY
Date Tue, 05 Feb 2019 21:25:00 GMT

    [ https://issues.apache.org/jira/browse/PHOENIX-5123?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16761236#comment-16761236
] 

Lars Hofhansl commented on PHOENIX-5123:
----------------------------------------

A "hack" in that I didn't try it, yet - other than running SpillableGroupByIT.
One part I don't understand in the original code: How can this ever store any key larger than
4k?!
It gets "pages" represented by MappedByteBuffers that have an exact size of 4k, so you can't
write past 4k, or read past 4k - would throw a BufferOverflowException.

I think in my new code I should add an explicit safety check to not read from or write past
4k.
Perhaps then it's ok... 

> Avoid using MappedByteBuffers for server side GROUP BY
> ------------------------------------------------------
>
>                 Key: PHOENIX-5123
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-5123
>             Project: Phoenix
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Priority: Major
>         Attachments: 5123-4.x-W.I.P.txt
>
>
> Like PHOENIX-5120 but for GROUP BY.
> Solution is a bit more tricky, since outline for sorting the access here is truly random.
> [~apurtell] suggests to perhaps just use a RandomAccessFile for this.
> (I'm not sure that uses under the hood, though)



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message