Hi Chuck I believe Praveenesh was adding his thought to the discussion on preprocessing the data using mapreduce itself. If you go with hadoop streaming you can use the python script in the mapper and that will do the preprocessing parallely on large volume data. Then this preprocessed data can be loaded into hive table.
Regards Bejoy KS Sent from handheld, please excuse typos. -----Original Message----- From: "Connell, Chuck" <chuck.conn...@nuance.com> Date: Sat, 8 Sep 2012 12:18:33 To: user@hive.apache.org<user@hive.apache.org> Reply-To: user@hive.apache.org Subject: RE: How to load csv data into HIVE I would like to hear more about this "hadoop streaming to Hive" idea. I have used streaming jobs as mappers, with a python script as map.py. Are you saying that such a streaming mapper can load its output into Hive? Can you send some example code? Hive wants to load "files" not individual lines/records. How would you do this? Thanks very much, Chuck ________________________________ From: praveenesh kumar [praveen...@gmail.com] Sent: Saturday, September 08, 2012 7:54 AM To: user@hive.apache.org Subject: Re: How to load csv data into HIVE You can use hadoop streaming that would be much faster... Just run your cleaning shell script logic in map phase and it will be done in just few minutes. That will keep the data in HDFS. Regards, Praveenesh On Fri, Sep 7, 2012 at 8:37 PM, Sandeep Reddy P <sandeepreddy.3...@gmail.com<mailto:sandeepreddy.3...@gmail.com>> wrote: Hi, Thank you all for your help. I'll try both ways and i'll get back to you. On Fri, Sep 7, 2012 at 11:02 AM, Mohammad Tariq <donta...@gmail.com<mailto:donta...@gmail.com>> wrote: I said this assuming that a Hadoop cluster is available since Sandeep is planning to use Hive. If that is the case then MapReduce would be faster for such large files. Regards, Mohammad Tariq On Fri, Sep 7, 2012 at 8:27 PM, Connell, Chuck <chuck.conn...@nuance.com<mailto:chuck.conn...@nuance.com>> wrote: I cannot promise which is faster. A lot depends on how clever your scripts are. From: Sandeep Reddy P [mailto:sandeepreddy.3...@gmail.com<mailto:sandeepreddy.3...@gmail.com>] Sent: Friday, September 07, 2012 10:42 AM To: user@hive.apache.org<mailto:user@hive.apache.org> Subject: Re: How to load csv data into HIVE Hi, I wrote a shell script to get csv data but when i run that script on a 12GB csv its taking more time. If i run a python script will that be faster? On Fri, Sep 7, 2012 at 10:39 AM, Connell, Chuck <chuck.conn...@nuance.com<mailto:chuck.conn...@nuance.com>> wrote: How about a Python script that changes it into plain tab-separated text? So it would look like this… 174969274<tab>14-mar-2006<tab>3522876<tab> <tab>14-mar-2006<tab>500000308<tab>65<tab>1<newline> etc… Tab-separated with newlines is easy to read and works perfectly on import. Chuck Connell Nuance R&D Data Team Burlington, MA 781-565-4611<tel:781-565-4611> From: Sandeep Reddy P [mailto:sandeepreddy.3...@gmail.com<mailto:sandeepreddy.3...@gmail.com>] Subject: How to load csv data into HIVE Hi, Here is the sample data "174969274","14-mar-2006"," 3522876","","14-mar-2006","500000308","65","1"| "174969275","19-jul-2006","3523154","","19-jul-2006","500000308","65","1"| "174969276","31-dec-2005","3530333","","31-dec-2005","500000308","65","1"| "174969277","14-apr-2005","3531470","","14-apr-2005","500000308","65","1"| How to load this kind of data into HIVE? I'm using shell script to get rid of double quotes and '|' but its taking very long time to work on each csv which are 12GB each. What is the best way to do this? -- Thanks, sandeep -- Thanks, sandeep
