Rekha commented on PIG-626:


This is nice.I need to trap the below data per script and save it into database 
as record of the run.


In pig 2.2 I see similar info in the console logs.Can the pulling be done from 

In anycase , I do not want stats run to be an offline process as seen in the 
Test files in the package.
Ideally would like to call/execute the stats module right from pig scripts 
itself as -
stats.record(); //first line
stats.flush('out<jobid>.txt');//last line

The push to db can be a cron process.

Any suggestions?

Thanks & Regards,

> Statistics (records read by each mapper and reducer)
> ----------------------------------------------------
>                 Key: PIG-626
>                 URL: https://issues.apache.org/jira/browse/PIG-626
>             Project: Pig
>          Issue Type: New Feature
>          Components: impl
>    Affects Versions: 0.2.0
>            Reporter: Shubham Chopra
>            Assignee: Shubham Chopra
>            Priority: Minor
>             Fix For: 0.3.0
>         Attachments: PIG-626.patch, pigStats.patch, pigStats.patch, 
> pigStats.patch, pigStats.patch, pigStats.patch, 
> TEST-org.apache.pig.test.TestBZip.txt
> This uses the counters framework that hadoop has. Initially, I am just 
> interested in finding out the number of records read by each mapper/reducer 
> particularly for the last job in any script. A sample code to access the 
> statistics for the last job:
> String reducePlan = 
> stats.getPigStats().get(stats.getLastJobID()).get("PIG_STATS_REDUCE_PLAN");
>         if(reducePlan == null) {
>             System.out.println("Records written : " + 
> stats.getPigStats().get(stats.getLastJobID()).get("PIG_STATS_MAP_OUTPUT_RECORDS"));
>         } else {
>             System.out.println("Records written : " + 
> stats.getPigStats().get(stats.getLastJobID()).get("PIG_STATS_REDUCE_OUTPUT_RECORDS"));
>         }
> The patch contains 7 test cases. These include tests PigStorage and 
> BinStorage along with one for multiple MR jobs case.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

Reply via email to