[jira] [Commented] (PHOENIX-2961) select on a composite primary key fails when DESC is set in the table

["James Taylor (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/PHOENIX-2961?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15310905#comment-15310905 ] 

James Taylor commented on PHOENIX-2961:
---------------------------------------

My first guess is that the full row key(s) we produce for our skip scan when we're doing a point lookup don't have the trailing 0xFF byte as required for the trailing DESC PK column, [~samarthjain].

> select on a composite primary key fails when DESC is set in the table
> ---------------------------------------------------------------------
>
>                 Key: PHOENIX-2961
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-2961
>             Project: Phoenix
>          Issue Type: Bug
>    Affects Versions: 4.7.0
>         Environment: linux ubuntu 14.0.4.4   
> phoenix-4.7.0-HBase-0.98-client
> HBase Version	0.98.19
>            Reporter: Dhiraj Hegde
>            Assignee: Samarth Jain
>              Labels: SFDC
>             Fix For: 4.8.0
>
>         Attachments: test1.sql
>
>
> I created a table with a primary key composed of 4 fields. The table specifies DESC for one of those fields (please see the test script provided). When the table is queried using any combination of 3 of the primary key fields, it selects rows correctly. However, when all 4 fields are specified, no rows are selected.  Here is the sql script that reproduces this issue:
> ----------------------------------------------------------------
> CREATE TABLE IF NOT EXISTS  tests1 (
>    ACCOUNT_ID  VARCHAR NOT NULL,
>   BUCKET_ID     VARCHAR NOT NULL,
>   OBJECT_ID      VARCHAR NOT NULL,
>   OBJECT_VERSION   VARCHAR NOT NULL,
>       LOC           VARCHAR,
> CONSTRAINT PK PRIMARY KEY (ACCOUNT_ID, BUCKET_ID, OBJECT_ID, OBJECT_VERSION DESC));
> UPSERT INTO tests1  (ACCOUNT_ID, BUCKET_ID, OBJECT_ID, OBJECT_VERSION, LOC) VALUES ('acct1', 'bucket1', 'obj1', '1111', 'loc1');
> select ACCOUNT_ID, BUCKET_ID, OBJECT_VERSION  from tests1  WHERE ACCOUNT_ID = 'acct1' and BUCKET_ID = 'bucket1' and OBJECT_VERSION = '1111';
> select ACCOUNT_ID, BUCKET_ID, OBJECT_VERSION  from tests1  WHERE ACCOUNT_ID = 'acct1' and BUCKET_ID = 'bucket1' and OBJECT_ID = 'obj1';
> select ACCOUNT_ID, BUCKET_ID, OBJECT_VERSION  from tests1  WHERE ACCOUNT_ID = 'acct1' and BUCKET_ID = 'bucket1' and OBJECT_VERSION = '1111'  and OBJECT_ID = 'obj1';
> ----------------------------------------------
> The first two select statements selects the row that was inserted, whereas the third one fails to select any rows.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)