On Jan24, 2014, at 08:47 , Dean Rasheed <dean.a.rasheed@gmail.com> wrote:
> I think it should probably be broken up. It might be overly ambitious
> to try to get all of this committed during this commitfest, and in any
> case, I suspect that the committer would probably choose to commit it
> in stages. Perhaps something like:
>
> Patch 1
> - Basic support for inverse transition functions, CREATE AGGREGATE
> support and doc updates. This should include test cases to validate
> that the underlying executor changes are correct, by defining custom
> aggregates such as sum(int) and array_agg() using inverse transition
> functions.
>
> Patch 2
> - Add built-in inverse transition functions for count, sum(int), and friends.
>
> Patch 3, 4...
> - Other related groups of built-in aggregates. By this point, it
> should be a fairly mechanical process.
>
> Splitting it up this way now should help to focus on getting patch 1
> correct, without being distracted by all the other aggregates that may
> or may not usefully be made to have inverse transition functions. I
> think the value of the feature has been proved, and it is good to see
> that it can be applied to so many aggregates, but let's not try to do
> it all at once.
Working on that now, will post individual patches later today.
best regards,
Florian Pflug