On Fri, Dec 2, 2011 at 2:01 PM, Pavel Stehule <pavel.stehule@gmail.com> wrote:
>>> so bytea_agg - one param aggregate has sense
>>>
>>> it's very easy to implement it
>>
>> yup:
>>
>> create aggregate bytea_agg (bytea)
>> (
>> sfunc=byteacat,
>> stype=bytea
>> );
>
> this is workaround :)
>
> without a memory preallocating it has same speed like cycle in plpgsql.
sure, but I prefer to code against the workaround because it's cleaner
and it makes things easier to port over when such a feature makes it
into core. also, one liner sql has better chance of inlining as a
general rule.
merlin