[
https://issues.apache.org/jira/browse/PHOENIX-1436?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14214334#comment-14214334
]
rajeshbabu commented on PHOENIX-1436:
-------------------------------------
[~xufeng]
Sorry for late reply. I have checked it now. It's bug only. Currently the index
selection is not so intelligent like it should give first priority for index
having conditioned columns in the primary key. Currently we get the plans for
the indexes having all the columns to project and within that we use the plan
having less number of columns. In your case the both indexes have same number
of columns so selected INDEX_TEST_TABLE_INDEX_D.
Thanks.
> the choice of index by phoneix is not correct where do querying
> ---------------------------------------------------------------
>
> Key: PHOENIX-1436
> URL: https://issues.apache.org/jira/browse/PHOENIX-1436
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.1
> Environment: phoenix4.1
> HBase 0.98.6
> hadoop 1.2.1
> Reporter: xufeng
> Assignee: rajeshbabu
>
> 1.create INDEX_TEST_TABLE:
> create table index_test_table (a varchar not null,b varchar not null,c
> varchar not null,d varchar,e varchar, f varchar constraint pk primary
> key(a,b,c));
> 2.upsert some data into table.
> 3.create index on 'D' and 'F'
> create index INDEX_TEST_TABLE_INDEX_D on INDEX_TEST_TABLE(A,D)
> include(B,C,E,F);
> 5 rows affected (0.747 seconds)
> 0: jdbc:phoenix:10.20.14.35:2181> create index INDEX_TEST_TABLE_INDEX_F on
> INDEX_TEST_TABLE(A,F) include(B,C,D,E);
> 5 rows affected (0.748 seconds)
> 4.query data on 'D'
> explain select * from INDEX_TEST_TABLE where A in ('1','2','3','4','5') and D
> in ('1111','2222','3333');
> 5. it is ok:
> CLIENT PARALLEL 1-WAY SKIP SCAN ON 15 KEYS OVER INDEX_TEST_TABLE_INDEX_D
> ['1','1111'] - ['5','3333']
> 6. but query data on 'F'
> explain select * from INDEX_TEST_TABLE where A in ('1','2','3','4','5') and F
> in ('1111','2222','3333');
> 7. also use the index on 'D' and set filter on 'F'
> CLIENT PARALLEL 1-WAY SKIP SCAN ON 5 KEYS OVER INDEX_TEST_TABLE_INDEX_D ['1']
> - ['5'] |
> SERVER FILTER BY F IN ('1111','2222','3333') |
> is this a bug in 4.1 version??
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
