On Sat, Sep 5, 2015 at 8:36 AM, Marc Mamin <m.ma...@intershop.de> wrote:

>
>    array_not_nulls(array_agg(case when type='A' then link end ))as ar_a,
>    array_not_nulls(array_agg(case when type='B' then link end ))as ar_b,
>    array_not_nulls(array_agg(case when type NOT IN ('A', 'B') then link
> end)) as ar_others
>
>
​Yeah.  I'm not sure why I didn't tend toward the array_agg approach since
I have made use of it previously.  It cleans up a bit but not as much as I
was hoping - but also not seeing.  The preponderance of case statements​

​ and array slicing is still present.  I image trying to write some kind of
"shift" function but that doesn't mesh well with set theory; it seems to be
a procedural concept which I suspect would be considerably worse performing.

Thanks for the suggestion and once I get back on that project I'm intended
to do a barebones performance comparison of the two which I'll follow up
with.

David J.
​

Reply via email to