"Sergey E. Koposov" <math@sai.msu.ru> writes:
> Some small clean-up of the patch...
> + implementing the Tom's idea of minimizing the copying of the data inside
> advance_transition_function by using the temporary FunctionCallInfoData
> (now the computed arguments of the aggregates are putted directly into
> proper fcinfo.args fields, ready for the transition function call).
I've committed the core parts of this, but not yet the newly added
standard aggregates (it seemed like a good idea to break the thing
into smaller chunks for review). Will work on those next.
regards, tom lane