[
https://issues.apache.org/jira/browse/PHOENIX-1837?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14490529#comment-14490529
]
James Taylor commented on PHOENIX-1837:
---------------------------------------
FWIW, once PHOENIX-1826 is implemented, the
TrackOrderPreservingExpressionCompiler could be used on the ORDER BY
expression, but it's probably better to be able to determine this early as
you're suggesting. Note also, somewhat related, but we can remove the need to
sort GROUP BY rows at all which would help performance quite a bit:
PHOENIX-1006.
> Detect ordered/unordered group-by
> ---------------------------------
>
> Key: PHOENIX-1837
> URL: https://issues.apache.org/jira/browse/PHOENIX-1837
> Project: Phoenix
> Issue Type: Sub-task
> Reporter: Maryann Xue
> Assignee: Maryann Xue
> Original Estimate: 120h
> Remaining Estimate: 120h
>
> With ordered group-by, we would do much less work. The existing Phoenix code
> depends on TrackOrderPreservingExpressionCompiler to calculate the flag. But
> given that we would no longer use any of the Phoenix parse tree nodes and the
> TrackOrderPreservingExpressionCompiler is a parse tree node visitor, we
> should now implement the same logic with Calcite RexNode.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
