[
https://issues.apache.org/jira/browse/PHOENIX-1439?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14208611#comment-14208611
]
Mujtaba Chohan commented on PHOENIX-1439:
-----------------------------------------
Tried this with Phoenix 4.2 branch. Generated data using bin/performance.py
script and ran the following query:
{code}
select count(*) from performance_5000000 where host like 'E%' and domain like
'S%';
{code}
Explain plan shows skip scan is used:
{code}
CLIENT 2-CHUNK PARALLEL 1-WAY SKIP SCAN ON 1 RANGE OVER PERFORMANCE_5000000 ['E
','S'] - ['F ','T']
SERVER FILTER BY FIRST KEY ONLY
SERVER AGGREGATE INTO SINGLE ROW
{code}
Also another query that runs part of perf test
{code}
select count(*) from IN_LIKE_TABLE where (K1 in ('A','E','F','G','H','I','Z')
or K1 like 'E%') and (K2 like '%xx1%' or K2 like '%xxxxxx2%')
{code}
Explain plan:
{code}
CLIENT 6-CHUNK PARALLEL 1-WAY SKIP SCAN ON 6 RANGES OVER IN_LIKE_TABLE [A] - [Z]
SERVER FILTER BY FIRST KEY ONLY AND (K2 LIKE %xx1% OR K2 LIKE %xxxxxx2%)
SERVER AGGREGATE INTO SINGLE ROW
{code}
[~jamestaylor]
> LIKE doesn't appear to use skip scanning
> ----------------------------------------
>
> Key: PHOENIX-1439
> URL: https://issues.apache.org/jira/browse/PHOENIX-1439
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Brian Johnson
>
> We had a query using several like statements of the form LIKE 'prefix|%' and
> LIKE 'prefix2|%' ... and the performance was on par with a full table scan.
> When we replaced it with BETWEEN 'prefix|' and 'prefix|FFFFFFF...' the query
> time was substantially faster. It seems that the LIKE query should work with
> skip scan, but I don't think it's being used.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
