Re: The fastes way to sum array of integers

2012/12/16 aasat <satriani(at)veranet(dot)pl>:
> Hi,
>
> I sum values of array integers by query
>
> select sum(t)
> from unnest(array[1,2,3,4,5,6,7,8,9,0,1,2,3,4,5,6,7,8,9,0]) as t
> cross join generate_series(1, 1000000);
>
> Total runtime: 10020.699 ms
>
> But simmilar method to count array's elements is dramatically faster
>
> select sum(array_length(array[1,2,3,4,5,6,7,8,9,0,1,2,3,4,5,6,7,8,9,0], 1))
> from generate_series(1, 1000000)
>
> Total runtime: 464.042 ms
>
>
> It's as fast as simple query!
>
> select sum(1) from generate_series(1, 1000000)
> "Total runtime: 462.687 ms"
>
>
> It is possible that function e.g. array_sum written in C is as fast as
> bult-in array_length?
>

yes - specialized function can be significantly faster - and
aggregation can be calculated with significantly less iterations.

Regards

Pavel

>
> regards
> Wojciech
>
>
>
>
>
> --
> View this message in context: http://postgresql.1045698.n5.nabble.com/The-fastes-way-to-sum-array-of-integers-tp5736771.html
> Sent from the PostgreSQL - general mailing list archive at Nabble.com.
>
>
> --
> Sent via pgsql-general mailing list (pgsql-general(at)postgresql(dot)org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-general