[ 
https://issues.apache.org/jira/browse/DRILL-4862?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15437838#comment-15437838
 ] 

Zelaine Fong commented on DRILL-4862:
-------------------------------------

Info from [~chunhui.L]

The reason for this bug is that the original implementation copies the 
transferred results back into the source buffer and use part of source buffer 
as output buffer, so the same source buffer cannot be used a second time. 

> wrong results - use of convert_from(binary_string(key),'UTF8') in filter 
> results in wrong results
> -------------------------------------------------------------------------------------------------
>
>                 Key: DRILL-4862
>                 URL: https://issues.apache.org/jira/browse/DRILL-4862
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Execution - Flow
>    Affects Versions: 1.8.0
>         Environment: 4 node cluster CentOS
>            Reporter: Khurram Faraaz
>
> These results do not look right, i.e when the predicate has 
> convert_from(binary_string(key),'UTF8')
> Apache drill 1.8.0-SNAPSHOT git commit ID: 57dc9f43
> {noformat}
> [root@centos-0x drill4478]# cat f1.json
> {"key":"\\x30\\x31\\x32\\x33"}
> {"key":"\\x34\\x35\\x36\\x37"}
> {"key":"\\x38\\x39\\x30\\x31"}
> {"key":"\\x30\\x30\\x30\\x30"}
> {"key":"\\x31\\x31\\x31\\x31"}
> {"key":"\\x35\\x35\\x35\\x35"}
> {"key":"\\x38\\x38\\x38\\x38"}
> {"key":"\\x39\\x39\\x39\\x39"}
> {"key":"\\x41\\x42\\x43\\x44"}
> {"key":"\\x45\\x46\\x47\\x48"}
> {"key":"\\x49\\x41\\x44\\x46"}
> {"key":"\\x4a\\x4b\\x4c\\x4d"}
> {"key":"\\x57\\x58\\x59\\x5a"}
> {"key":"\\x4e\\x4f\\x50\\x51"}
> {"key":"\\x46\\x46\\x46\\x46"}
> {noformat}
> results without the predicate - these are correct results
> {noformat}
> 0: jdbc:drill:schema=dfs.tmp> select convert_from(binary_string(key),'UTF8') 
> from `f1.json`;
> +---------+
> | EXPR$0  |
> +---------+
> | 0123    |
> | 4567    |
> | 8901    |
> | 0000    |
> | 1111    |
> | 5555    |
> | 8888    |
> | 9999    |
> | ABCD    |
> | EFGH    |
> | IADF    |
> | JKLM    |
> | WXYZ    |
> | NOPQ    |
> | FFFF    |
> +---------+
> 15 rows selected (0.256 seconds)
> {noformat}
> results with a predicate - these results don't look correct
> {noformat}
> 0: jdbc:drill:schema=dfs.tmp> select convert_from(binary_string(key),'UTF8') 
> from `f1.json` where convert_from(binary_string(key),'UTF8') is not null;
> +----------+
> |  EXPR$0  |
> +----------+
> | 0123123  |
> | 4567567  |
> | 8901901  |
> | 0000000  |
> | 1111111  |
> | 5555555  |
> | 8888888  |
> | 9999999  |
> | ABCDBCD  |
> | EFGHFGH  |
> | IADFADF  |
> | JKLMKLM  |
> | WXYZXYZ  |
> | NOPQOPQ  |
> | FFFFFFF  |
> +----------+
> 15 rows selected (0.279 seconds)
> {noformat}



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

Reply via email to