hadoop-mapreduce-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shahab Yunus <shahab.yu...@gmail.com>
Subject Re: Number of records in an HDFS file
Date Mon, 13 May 2013 19:02:09 GMT
The count file will be a very small file, right? Once it is generated on
HDFS, you can automate its downloading or movement anywhere you want. This
should not take much time.

Regards,
Shahab


On Mon, May 13, 2013 at 2:58 PM, Mix Nin <pig.mixed@gmail.com> wrote:

> Hi,
>
> The final count file should reside in local directory, but not in HDFS
> directory. The above scripts will store text file in HDFS directory.
> The count file would need to be sent to other team who do not work on HDFS.
>
> Thanks
>
>
>
> On Mon, May 13, 2013 at 11:36 AM, Mohammad Tariq <dontariq@gmail.com>wrote:
>
>> If it is just counting the no. of records in a file then how about having
>> a short 3 liner :
>> LOGS= LOAD 'log';
>> LOGS_GROUP= GROUP LOGS ALL;
>> LOG_COUNT = FOREACH LOGS_GROUP GENERATE COUNT(LOGS);
>>
>> It did the trick for me.
>>
>> Warm Regards,
>> Tariq
>> cloudfront.blogspot.com
>>
>>
>> On Mon, May 13, 2013 at 11:57 PM, Shahab Yunus <shahab.yunus@gmail.com>wrote:
>>
>>> Not terribly efficient but at the top of my head: GROUP ALL and then do
>>> a COUNT (or COUNT (*). You can implement a follow-up script or add this in
>>> the existing script once the file has been generated.
>>>
>>> Regards,
>>> Shahab
>>>
>>>
>>> On Mon, May 13, 2013 at 2:16 PM, Mix Nin <pig.mixed@gmail.com> wrote:
>>>
>>>> Ok, let re modify my requirement. I should have specified in the
>>>> beginning itself.
>>>>
>>>> I need to get count of records in an HDFS file created by a PIG script
>>>> and the store the count in a text file. This should be done automatically
>>>> on a daily basis without manual intervention
>>>>
>>>>
>>>> On Mon, May 13, 2013 at 11:13 AM, Rahul Bhattacharjee <
>>>> rahul.rec.dgp@gmail.com> wrote:
>>>>
>>>>> How about the second approach , get the application/job id which the
>>>>> pig creates and submits to cluster and then find the job output counter
for
>>>>> that job from the JT.
>>>>>
>>>>> Thanks,
>>>>> Rahul
>>>>>
>>>>>
>>>>> On Mon, May 13, 2013 at 11:37 PM, Mix Nin <pig.mixed@gmail.com>
wrote:
>>>>>
>>>>>> It is a text file.
>>>>>>
>>>>>> If we want to use wc, we need to copy file from HDFS and then use
wc,
>>>>>> and this may take time. Is there a way without copying file from
HDFS to
>>>>>> local directory?
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>>
>>>>>> On Mon, May 13, 2013 at 11:04 AM, Rahul Bhattacharjee <
>>>>>> rahul.rec.dgp@gmail.com> wrote:
>>>>>>
>>>>>>> few pointers.
>>>>>>>
>>>>>>> what kind of files are we talking about. for text you can use
wc ,
>>>>>>> for avro data files you can use avro-tools.
>>>>>>>
>>>>>>> or get the job that pig is generating , get the counters for
that
>>>>>>> job from the jt of your hadoop cluster.
>>>>>>>
>>>>>>> Thanks,
>>>>>>>  Rahul
>>>>>>>
>>>>>>>
>>>>>>> On Mon, May 13, 2013 at 11:21 PM, Mix Nin <pig.mixed@gmail.com>wrote:
>>>>>>>
>>>>>>>> Hello,
>>>>>>>>
>>>>>>>> What is the bets way to get the count of records in an HDFS
file
>>>>>>>> generated by a PIG script.
>>>>>>>>
>>>>>>>> Thanks
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>

Mime
View raw message