You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@phoenix.apache.org by "Istvan Toth (Jira)" <ji...@apache.org> on 2023/06/16 07:04:00 UTC
[jira] [Assigned] (PHOENIX-6980) Push TopN down into joins where applicable
[ https://issues.apache.org/jira/browse/PHOENIX-6980?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Istvan Toth reassigned PHOENIX-6980:
------------------------------------
Assignee: Istvan Toth
> Push TopN down into joins where applicable
> ------------------------------------------
>
> Key: PHOENIX-6980
> URL: https://issues.apache.org/jira/browse/PHOENIX-6980
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Istvan Toth
> Assignee: Istvan Toth
> Priority: Major
>
> This came up with an non-merging hinted uncovered index query, which may not be directly applicable in the latest versions, but the same problem is reproducible with a simple IN subquery, which uses the same plan:
> {noformat}
> CREATE TABLE A(A_ID INTEGER NOT NULL PRIMARY KEY, A_V1 INTEGER);
> CREATE TABLE B(B_ID INTEGER NOT NULL PRIMARY KEY, B_V1 INTEGER);
> explain select * from A where (A_ID, A_V1) in (select B_ID, B_V1 from b ) order by a_v1 limit 10;
> +-----------------------------------------------------------------+----------------+---------------+-------------+
> | PLAN | EST_BYTES_READ | EST_ROWS_READ | EST_INFO_TS |
> +-----------------------------------------------------------------+----------------+---------------+-------------+
> | CLIENT 1-CHUNK PARALLEL 1-WAY FULL SCAN OVER A | null | null | null |
> | SERVER TOP 10 ROWS SORTED BY [A.A_V1] | null | null | null |
> | CLIENT MERGE SORT | null | null | null |
> | CLIENT LIMIT 10 | null | null | null |
> | SKIP-SCAN-JOIN TABLE 0 | null | null | null |
> | CLIENT 1-CHUNK PARALLEL 1-WAY FULL SCAN OVER B | null | null | null |
> | SERVER AGGREGATE INTO DISTINCT ROWS BY [B_ID, B_V1] | null | null | null |
> | CLIENT MERGE SORT | null | null | null |
> | DYNAMIC SERVER FILTER BY A.A_ID IN ($48.$50) | null | null | null |
> +-----------------------------------------------------------------+----------------+---------------+-------------+
> {noformat}
> This is super inefficient, the topN part should be pushed down into the inner query.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)