Difference in decimal values for variance in Presto.

[Jatin Demla]

Hi community,

I was working on presto bug 995 (
https://issues.apache.org/jira/browse/CARBONDATA-995) in which variance of
decimal values is calculated. When there is lesser number of decimal values
to calculate variance in carbondata and presto then there being no
difference in result. But when the dataset become large the result value
varies to minor difference (in my case i am calculating it with 2000
bigdecimal values the result difference between presto and carbondata is
0.1938... ). So, should i continue looking into this issue or this much
difference is fine.


--
Thanks & Regards
Jatin

[Erlu Chen]

Hi

So what is the behave of same operation in hive ?


Regards.
Chenerlu.

[Jatin Demla]

Hi

I was getting the same result in hive as in carbondata. Also, I was getting
the same difference in values when I query presto using hive connector.
On Thu, Jun 29, 2017 at 3:03 PM, Erlu Chen <[hidden email]> wrote:

> Hi
>
> So what is the behave of same operation in hive ?
>
>
> Regards.
> Chenerlu.
>
>
>
> --
> View this message in context: http://apache-carbondata-dev-
> mailing-list-archive.1130556.n5.nabble.com/Difference-in-
> decimal-values-for-variance-in-Presto-tp16496p16711.html
> Sent from the Apache CarbonData Dev Mailing List archive mailing list
> archive at Nabble.com.
>

--
Thanks & Regards
Jatin

[Erlu Chen]

Hi

I think you can continue look into this issue and confirm whether it is carbondata issue or presto issue.

Thanks.


Regards.
Chenerlu.

[geetikagupta]

Hi,

I was debugging into the code and found out that for variance calculation
presto casts the decimal values in double due to which we are getting some
difference in decimal values. Most probably this is an issue with Presto.

--
Regards,
Geetika Gupta

On Fri, Jun 30, 2017 at 7:02 AM, Erlu Chen <[hidden email]> wrote:

> Hi
>
> I think you continue looking into this issue and confirm whether it is
> carbondata issue or presto issue.
>
> Thanks.
>
>
> Regards.
> Chenerlu.
>
>
>
> --
> View this message in context: http://apache-carbondata-dev-
> mailing-list-archive.1130556.n5.nabble.com/Difference-in-
> decimal-values-for-variance-in-Presto-tp16496p16839.html
> Sent from the Apache CarbonData Dev Mailing List archive mailing list
> archive at Nabble.com.
>

--
Regards,
Geetika Gupta

[Erlu Chen]

Hi

OK, thanks very much.

If you find something wrong in carbondata, we can discuss here.

Regards.
Chenerlu.