You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@calcite.apache.org by "Atri Sharma (JIRA)" <ji...@apache.org> on 2018/05/16 07:34:00 UTC

[jira] [Commented] (CALCITE-873) Prevent sort when ORDER BY not necessary due to equality constraints

    [ https://issues.apache.org/jira/browse/CALCITE-873?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16476971#comment-16476971 ] 

Atri Sharma commented on CALCITE-873:
-------------------------------------

[~julianhyde] I am reviving this PR with running the added tests and have also added the rule to CalcitePrepareImpl. The tests pass successfully. Please see and let me know. Thanks

> Prevent sort when ORDER BY not necessary due to equality constraints
> --------------------------------------------------------------------
>
>                 Key: CALCITE-873
>                 URL: https://issues.apache.org/jira/browse/CALCITE-873
>             Project: Calcite
>          Issue Type: Improvement
>            Reporter: James Taylor
>            Assignee: Atri Sharma
>            Priority: Major
>
> We're working on an optimization in Phoenix to optimize away an ORDER BY when it is known based on equality expressions in the WHERE clause that it is not necessary (PHOENIX-2194). It'd be great if Calcite could do that as well.
> Here's a example, given the following schema:
> {code}
> CREATE TABLE T (
>   K1 VARCHAR,
>   K2 VARCHAR,
>   K3 VARCHAR,
>   CONSTRAINT pk PRIMARY KEY (K1, K2, K3));
> {code}
> In the following queries, no sort is necessary:
> {code}
> SELECT * FROM T WHERE K1='A' ORDER BY K2,K3;
> SELECT * FROM T WHERE K2='B' ORDER BY K1,K3;
> SELECT * FROM T WHERE K1='A' AND K2='B' ORDER BY K3;
> {code}
> There are also some edge cases where a function may be known to select a *prefix* of the column value where it's still ok to not sort:
> {code}
> SELECT * FROM T WHERE K1='A' AND SUBSTR(K2,1,3)='ABC' ORDER BY K2;
> {code}
> But if another column is included in the ORDER BY after the prefixing, a sort would still be necessary:
> {code}
> SELECT * FROM T WHERE K1='A' AND SUBSTR(K2,1,3)='ABC' ORDER BY K2,K3;
> {code}



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