[ 
https://issues.apache.org/jira/browse/PIG-5166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15984513#comment-15984513
 ] 

Adam Szita commented on PIG-5166:
---------------------------------

This is a float precision problem and although {{floatpostprocess}} is turned 
on for this test case it doesn't help too much because it's only considering 3 
decimals...
Here are the "different" values produced by MR and Spark without float post 
processing, basically they are all the same, but the e2e test framework fails 
to compare them properly:
{code:title=MR full prec}
bob davidson    2.3175
bob ichabod     1.8125 
david xylophone 1.7865000000000002
rachel ellison  2.5294999999999996
zach ellison    2.4774999999999996
zach steinbeck  1.9785000000000004
zach white      1.6575
{code}
{code:title=SPRAK full prec}
bob davidson    2.3175000000000003
bob ichabod     1.8125000000000002
david xylophone 1.7864999999999998
rachel ellison  2.5295
zach ellison    2.4775 
zach steinbeck  1.9785
zach white      1.6575000000000002
{code}

I propose an *extension of floatpostprocessor.pl* so that a test case may 
define how many decimals it wants to use before comparing results. The default 
can stay 3 as it is right now, and this can be overridden by supplying for 
example {{'decimals' => 6}} as seen in [^PIG-5166.0.patch]

[~kellyzly] [~nkollar] let me know what you think

> GroupAggFunc_9 is failing with spark exec type
> ----------------------------------------------
>
>                 Key: PIG-5166
>                 URL: https://issues.apache.org/jira/browse/PIG-5166
>             Project: Pig
>          Issue Type: Sub-task
>          Components: spark
>            Reporter: Nandor Kollar
>            Assignee: Adam Szita
>             Fix For: spark-branch
>
>         Attachments: PIG-5166.0.patch
>
>
> results are different:
> {code}
> diff GroupAggFunc_9.out/out_sorted GroupAggFunc_9_benchmark.out/out_sorted
> 30c30
> < bob davidson        2.318
> ---
> > bob davidson        2.317
> 35c35
> < bob ichabod 1.813
> ---
> > bob ichabod 1.812
> 102c102
> < david xylophone     1.786
> ---
> > david xylophone     1.787
> 447c447
> < rachel ellison      2.530
> ---
> > rachel ellison      2.529
> 655c655
> < zach ellison        2.478
> ---
> > zach ellison        2.477
> 669c669
> < zach steinbeck      1.978
> ---
> > zach steinbeck      1.979
> 673c673
> < zach white  1.658
> ---
> > zach white  1.657
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Reply via email to