[teiid-issues] [JBoss JIRA] (TEIID-3119) Improve full sort performance for dup removal and grouping

[Steven Hawkins (JIRA)]

    [ https://issues.jboss.org/browse/TEIID-3119?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13000505#comment-13000505 ] 

Steven Hawkins commented on TEIID-3119:
---------------------------------------

The grouping change is quite a bit more involved and will require a change the to user defined aggregate interface to work with custom functions.  The basic idea is that we'll maintain a tree keyed by the ordering and expose the accumulator state to be stored along with the key.  Then we'll walk the tree at the need to pull the final aggregate values.  Initially we'll probably only do this in non-distinct and non-rollup scenarios - although each can be done in this manner as well with some additional effort.

> Improve full sort performance for dup removal and grouping
> ----------------------------------------------------------
>
>                 Key: TEIID-3119
>                 URL: https://issues.jboss.org/browse/TEIID-3119
>             Project: Teiid
>          Issue Type: Enhancement
>          Components: Query Engine
>            Reporter: Steven Hawkins
>            Assignee: Steven Hawkins
>             Fix For: 8.9
>
>
> We perform a full sort for grouping/distinct reusing the existing sorting logic.  This is not as efficient as it could for larger data sets with smaller numbers of groups.  A hash implementation or combining the grouping with the sorting would perform better.



--
This message was sent by Atlassian JIRA
(v6.3.1#6329)