hadoop-mapreduce-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Maxim Ulanovskiy <mulanovs...@iponweb.net>
Subject Re: reuse of MultipleOutputFormat with new API
Date Fri, 23 Oct 2009 13:28:33 GMT
Thanks for answer Amareshwari,

For now I have my own ByDateTimeOutputFormat class which extends
MultipleOutputFormat. It generates custom target file name and also choose
RecordWriter based on some configuration params. I've seen new
MultipleOutputs but not quite understand how to achive the same
functionality with a new API...

-- 
Best Regards,
Max

2009/10/22 Amareshwari Sri Ramadasu <amarsri@yahoo-inc.com>

> With new api, functionality of MultipleOutputFormat and MultipleOutputs is
> combined into one as org.apache.hadoop.mapreduce.lib.output.MultipleOutputs
> in branch 0.21.
> You can use the same.
>
> Thanks
> Amareshwari
>
>
> Sean Owen wrote:
>
>> FWIW this same sort of thing is blocking Apache Mahout from progress
>> on implementations using Hadoop. I imagine the whole migration is far
>> more involved than it appears so it makes sense it is taking time. But
>> yeah making all the new APIs compatible with the new APIs would be a
>> great step forward for 0.20.x
>>
>> On Wed, Oct 21, 2009 at 1:00 PM, Maxim Ulanovskiy
>> <mulanovskiy@iponweb.net> wrote:
>>
>>
>>> Hello, fellows!
>>>
>>> Have a problem with migration to Hadoop 0.20: I need to register
>>> MyOutputFormatClass which extends MultipleOutputFormat in Job object. But
>>> since Job.setOutputFormatClass(Class<? extends OutputFormat> cls)
>>> requires
>>> new version of OutputFormat class and MultipleOutputFormat implements old
>>> version of OutputFormat I can't register it. So is there any replacement
>>> of
>>> MultipleOutputFormat in the new API or I should implement some kind of
>>> adapter?
>>>
>>>
>>> --
>>> Best Regards,
>>> Maxim Ulanovskiy
>>> java-developer
>>>
>>> IPonWeb
>>>
>>>
>>>
>>

Mime
View raw message