Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
Hadoop, mail # user - is it possible to specify an empty key-value separator for TextOutputFormat?


+
Luca Pireddu 2012-02-09, 10:27
+
hadoop hive 2012-02-09, 10:53
Copy link to this message
-
Re: is it possible to specify an empty key-value separator for TextOutputFormat?
Luca Pireddu 2012-02-09, 10:58
Hi Vikas,

On 02/09/2012 11:53 AM, hadoop hive wrote:
> hey luca,
>
> you can use
>
> conf.set("*mapred.textoutputformat.separator*", " ");
>
> hope it works fine
>
> regards
> Vikas Srivastava
I'm using pipes for this task.  I can't specify the configuration
property via Java code.  In addition, I don't want it to insert a space
as a separator :-)

Thanks,

Luca
> On Thu, Feb 9, 2012 at 3:57 PM, Luca Pireddu<[EMAIL PROTECTED]>  wrote:
>
>> Hello list,
>>
>> I'm trying to specify from the command line an empty string as the
>> key-value separator for TextOutputFormat.
>>
>> I'm specifying a blank as the value for the mapred.textoutputformat.**separator
>> property:
>>
>>   -D mapred.textoutputformat.**separator=""
>>
>> And, when I look at the job's configuration file (job.xml) this setting
>> seems to be picked up correctly.  However, the output format continues
>> using a TAB to separate key and value when I specify an empty string.
>>
>> Any way to convince it not to insert a separator (use the empty string)?
>>
>>
>> ps:  I'm using pipes so I can't specify a null key or value.
>>
>>
>> --
>> Luca Pireddu
>> CRS4 - Distributed Computing Group
>> Loc. Pixina Manna Edificio 1
>> 09010 Pula (CA), Italy
>> Tel: +39 0709250452