[
https://issues.apache.org/jira/browse/PHOENIX-3666?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15864340#comment-15864340
]
James Taylor commented on PHOENIX-3666:
---------------------------------------
I wouldn't want to impact performance with an extra sort (as perf is one of the
main reasons we're doing this). I think using 2 bytes is reasonable as if you
need more than 65K columns the sparseness is going to end up being a problem.
The client-side should be able to use the number of encoded bytes of the
concrete PTable for the length of the column qualifier in any reserved column
qualifiers. If you're stuck on a subquery issue, I'd ping [~maryannxue] and if
you're stuck on a local index issue, I'd ping [~rajeshbabu]. I don't think we
should wait any longer to figure it out for 4.10. Not surfacing the setting of
the number of bytes is fine for now.
> Make use of EncodedColumnQualifierCellsList for all column name mapping
> schemes
> -------------------------------------------------------------------------------
>
> Key: PHOENIX-3666
> URL: https://issues.apache.org/jira/browse/PHOENIX-3666
> Project: Phoenix
> Issue Type: Sub-task
> Reporter: Samarth Jain
> Assignee: Samarth Jain
> Attachments: PHOENIX-3666_wip.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
