Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Sqoop >> mail # dev >> Review Request 12806: SQOOP-890: ClassWriter Generates setField Casts Based on the Output Type, not the Input Type


Copy link to this message
-
Re: Review Request 12806: SQOOP-890: ClassWriter Generates setField Casts Based on the Output Type, not the Input Type

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/12806/
-----------------------------------------------------------

(Updated July 24, 2013, 5:44 p.m.)
Review request for Sqoop.
Changes
-------

You're right, it's not essential due to map-columns, but it's really quite convenient when you don't know what types a JDBC driver will map some of the more odd SQL types to (e.g. reals, decimals & numerics - even HSQLDB 1.8 returns doubles for a float column). Sqoop should definitely support silent widening (e.g. storing an int in a SQL long-typed column) but I understand your point re: narrowing. Would a warning when converting be sufficient? Or would you prefer that narrowing types had to be given explicitly on the command line? I don't know any libraries of the top of my head that will narrow (e.g.) a float to a double if that narrowing doesn't result in data loss. I've updated the patch to add a widening and narrowing case, both of which fail in 1.4.3 and pass with the patch.

Thanks -
Repository: sqoop-trunk
Description
-------

Sqoop should be more flexible when converting between number types.
Diffs (updated)
-----

  src/java/org/apache/sqoop/orm/ClassWriter.java 1bd2a41
  src/test/com/cloudera/sqoop/TestAvroExport.java 7016826

Diff: https://reviews.apache.org/r/12806/diff/
Testing
-------

Unit test included.
Thanks,

Nick White