thanks. it seems that as long as I use sequencefile as the storage format, there will be \t before the first column. If this output is continously used by hive, it is fine. The problem is that I may use a self-define map-reduce job to read these files. Does that mean I have to take care of this \t by myself?
is there any option that I can disable this \t in hive? At 2013-01-09 22:38:11,"Dean Wampler" <dean.wamp...@thinkbiganalytics.com> wrote: To add to what Nitin said, there is no key output by Hive in front of the tab. On Wed, Jan 9, 2013 at 3:07 AM, Nitin Pawar <nitinpawar...@gmail.com> wrote: you may want to look at the sequencefile format http://my.safaribooksonline.com/book/databases/hadoop/9780596521974/file-based-data-structures/id3555432 that tab is to separate key from values in the record (I may be wrong but this is how I interpreted it) On Wed, Jan 9, 2013 at 12:49 AM, Richard <codemon...@163.com> wrote: more information: if I set the format as textfile, there is no tab space. if I set the format as sequencefile and view the content via hadoop fs -text, I saw a tab space in the head of each line. At 2013-01-09 15:44:00,Richard <codemon...@163.com> wrote: hi there I have a problem with creating a hive table. no matter what field delimiter I used, I always got a tab space in the head of each line (a line is a record). something like this: \t f1 \001 f2 \001 f3 ... where f1 , f2 , f3 denotes the field value and \001 is the field separator. here is the clause I used 35 create external table if not exists ${HIVETBL_my_table} 36 ( 37 nid string, 38 userid string, 39 spv bigint, 40 sipv bigint, 41 pay bigint, 42 spay bigint, 43 ipv bigint, 44 sellerid string, 45 cate string 46 ) 47 partitioned by(ds string) 48 row format delimited fields terminated by '\001' lines terminated by '\n' 49 stored as sequencefile 50 location '${HADOOP_PATH_4_MY_HIVE}/${HIVETBL_my_table}'; thanks for help. Richard -- Nitin Pawar -- Dean Wampler, Ph.D. thinkbiganalytics.com +1-312-339-1330
