Timestamp handling, specifically TimestampWritable, has several bugs. The latest versions of Hive (11.0) has some fixes needed to make it work correctly.
On Mon, May 20, 2013 at 2:18 AM, Zhenhua Chai <chaizhen...@gmail.com> wrote: > hello, > I tried to import data from SQL to Hive RCFile. I use RCFile.Writer to > generate rcfile and upload to HIVE table directory. the rcfiles has > different columns: > c0_1.rc has 2 columns > c1_1.rc has 3 columns > c2_1.rc has 4 columns > > the outputs was strange: > > # All rc files are loaded: > hive> select * from simple; > OK > 1 foo NULL null > 2 bar NULL null > 3 foobar NULL null > 3 haliluya NULL null > 4 holy shit NULL null > 4 NULL null > 7 NULL null > > # c1_1.rc and c2_1.rc are loaded. > hive> select * from simple; > OK > 3 haliluya 2013-05-17 14:19:02 null > 4 holy shit 2013-05-17 14:19:20 null > 4 2013-05-17 14:19:45 null > 7 2013-05-17 14:20:02 null > > # only c2_1 is loaded. > hive> select * from simple; > OK > 4 2013-05-17 14:19:45 4Vx� > 7 2013-05-17 14:20:02 �P"�:mo�d �)�2�� > > > i create the table use the command ``` create table foo (k int, v text, t > timestamp, v1 binary)stored as rcfile; ``` > i am using hive-0.8.1. and the appendix contains the rc files. > > can anyone help to solve this? did it a known bug? or should i use a later > version of hive? thanks. >
