Yeah that's the eventually goal. Currently the user has to manually convert the TIMESTAMP to String and partition on String. Since we do NOT allow range partitioning, this should provide exactly the same semantics. The only thing is that user needs to convert TIMESTAMP to STRING (and back when needed) by himself for now.
Zheng On Thu, Feb 12, 2009 at 3:00 PM, Shyam Sarkar <shyam_sar...@yahoo.com>wrote: > So eventually UDFs need to be modified to operate on TIMESTAMP data type ! > > Can we partition a table based on TIMESTAMP ? We should because it is > really important for users in many cases. > > Shyam > > > > --- On Thu, 2/12/09, Zheng Shao <zsh...@gmail.com> wrote: > > > From: Zheng Shao <zsh...@gmail.com> > > Subject: Re: Server time zone ! > > To: hive-dev@hadoop.apache.org, shyam_sar...@yahoo.com > > Date: Thursday, February 12, 2009, 2:31 PM > > Not yet. I think JVM take the environment variable TZ for > > that purpose for > > all java-related functions. > > > > In build-common.xml you should find the following: > > <env key="TZ" > > value="US/Pacific"/> > > > > This makes sure all current UDFs related to time/date are > > using the > > US/Pacific time zone. (Note: all these UDFs are operating > > on String and > > Integer types right now) > > > > Zheng > > > > On Thu, Feb 12, 2009 at 2:21 PM, Shyam Sarkar > > <shyam_sar...@yahoo.com>wrote: > > > > > Is there any 'Server's time zone' > > implementation inside Hive? For proper > > > implementation of TIMESTAMP data type, this is > > necessay to translate from > > > stored string type. I am focusing on MySQL 6.0 (with > > limited properties) for > > > TIMESTAMP. > > > > > > http://dev.mysql.com/doc/refman/6.0/en/timestamp.html > > > > > > Thanks, > > > Shyam > > > > > > > > > > > > > > > > > > -- > > Yours, > > Zheng > > > > -- Yours, Zheng
