[
https://issues.apache.org/jira/browse/SQOOP-2612?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14996792#comment-14996792
]
Jarek Jarcec Cecho commented on SQOOP-2612:
-------------------------------------------
I see a value in having ability to map entire type to something else rather
then individual columns. In fact my original patch that added
{{\-\-map\-column\-java}} has that functionality back in SQOOP-342, but it
wasn't committed as the community felt that it won't be much helpful at the
time. I feel that this changed as I see many people struggling with the
mapping, so I'm +1 on adding that functionality.
It seems that the column mapping is primarily implemented in {{ClassWriter}}
class - any reason why the type based mapping is implemented elsewhere
[~proftodd]? I'm concerned that it will be confusing to maintain when very
similar functionality is twice on different places.
> Allow rules-based column mapping
> --------------------------------
>
> Key: SQOOP-2612
> URL: https://issues.apache.org/jira/browse/SQOOP-2612
> Project: Sqoop
> Issue Type: New Feature
> Components: codegen, connectors, test, tools
> Affects Versions: 1.4.6
> Reporter: John Todd
> Priority: Minor
> Attachments: SQOOP-2612.patch
>
>
> Allow users to map source columns to Java and Avro types based on rules
> rather than a list.
> About half the data types on my current project are NUMERIC, and many of
> those have a SCALE of 0. It would be helpful to be able to process those as
> Integers or Longs. The map-column-java feature is helpful, but it requires
> continuous surveillance of the source tables and maintenance of lists of
> columns and what type they should be mapped to.
> It would be helpful to be able to create a rules-based interpretation of the
> type inference provided by Sqoop.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
