Prasad, I've confirmed that I don't see the issue with Hive 0.3 (and Hadoop 0.19.1), but I can still reproduce it with trunk. I can take a look at finding the exact revision that caused the problem later this week...
Neil On Sun, May 3, 2009 at 4:06 PM, Prasad Chakka <pcha...@facebook.com> wrote: > I am not able to reproduce the error. I just tried it on our cluster > (hadoop-0.17.2.1 & hive-0.3). > > > ________________________________ > From: Neil Conway <neil.con...@gmail.com> > Reply-To: <hive-dev@hadoop.apache.org> > Date: Sun, 3 May 2009 15:38:48 -0700 > To: <hive-dev@hadoop.apache.org> > Subject: Re: more ClassCastExceptions with joins, group by and non-strings > > Prasad, > > Thanks for the response. However, as far as I can tell, in the example > query Peter gave we are already casting the non-string columns to > string, but we're still seeing a ClassCastException. Any other > suggestions? > > Thanks, > > Neil > > On Sun, May 3, 2009 at 7:29 AM, Prasad Chakka <pcha...@facebook.com> wrote: >> I think this is a known bug. casting the non-string column should work. >> >> See the below email thread (and HIVE-405) >> http://mail-archives.apache.org/mod_mbox/hadoop-hive-user/200904.mbox/browser >> >> >> ________________________________ >> From: Neil Conway <neil.con...@gmail.com> >> Reply-To: <hive-dev@hadoop.apache.org> >> Date: Sat, 2 May 2009 23:15:06 -0700 >> To: <hive-dev@hadoop.apache.org> >> Subject: Re: more ClassCastExceptions with joins, group by and non-strings >> >> On Sat, Apr 25, 2009 at 4:47 PM, Peter Alvaro <palv...@gmail.com> wrote: >>> create table bar ( >>> bas string--, >>> --bat double >>> ) >>> ROW FORMAT DELIMITED FIELDS TERMINATED BY '\174'; >> >> "bat" shouldn't be commented-out, if you want to use the provided test data. >> >>> It appears that queries with all three of (join, group by, non-string >>> datatype) cause a crash in the serde code run at the reducer: >> >> I can reproduce this issue with the trunk Hive code (+ Hadoop 0.19.1). >> Has anyone from the Hive team had a chance to investigate this? >> >> Thanks, >> >> Neil >> >> > >
