[ 
https://issues.apache.org/jira/browse/KYLIN-3924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16852618#comment-16852618
 ] 

wangxiaojing commented on KYLIN-3924:
-------------------------------------

Hello ShaoFeng,

      I am sorry , I responded to you so late.
      The values of some fields in my lookup table are really large. 
For example, the memo field is a text type, but this may often be encountered 
in tables. 
Can we just check and build the dimension columns and the necessary metric 
column, and the others can be set as a null value?
    This way, on the one hand, it can avoid situations where tables due to 
large content fields (non-dimension fields) can not be used as lookup tables,on 
the other hand,it can reduce the time and size of the build dictionary.

> Build cube on step 'build Dimension Dictionary ', the contents of 
> non-dimensional columns are also checked and built
> --------------------------------------------------------------------------------------------------------------------
>
>                 Key: KYLIN-3924
>                 URL: https://issues.apache.org/jira/browse/KYLIN-3924
>             Project: Kylin
>          Issue Type: Bug
>          Components: Job Engine, Tools, Build and Test
>    Affects Versions: v2.6.1
>            Reporter: wangxiaojing
>            Priority: Major
>              Labels: build, patch
>         Attachments: builderr.log
>
>
> Hi all,
>  I use version 2.6.1 , when build cube on step 'build Dimension Dictionary' 
> for lookup tables ,it throws some exceptions,it seems checks the content size 
> of each field in the table while building Dimension Dictionary, and build all 
> fields in the table with a dictionary.
> Suggestion:
>  1. Large content fields may not be dimension columns, in fact, it is not 
> necessary to check the value size of non-dimension columns,only check 
> dimension columns.
> 2. If the value of some fields in the dimension table is too long, the 
> exception is not clear, can not be located the problem column quickly.
> 3. There is no need to build dictionaries for non-dimensional columns



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to