Re: [HACKERS] Missing array support

Tom Lane wrote:
> If I were you I'd file this on the to-fix-later list and concentrate
> on polymorphic aggregates during the next couple days. If that's not
> done by Tuesday it will be a tough sell to put in during beta.

Attached is a patch that implements polymorphic aggregates.

Included in the patch, I changed SQL language functions so that they
could be declared with and use polymorphic types. This was necessary to
facilitate my testing, and I had wanted to implement that all along
anyway (in fact I'd still like to allow PL/pgSQL to use polymorphic
types, but I'll try to do that separately).

The attached compiles cleanly and passes all regression tests. I've also
attached a test script and its output that I used to verify appropriate
behavior of CREATE AGGREGATE. The script attempts to cover all possible
combinations of inputs and outputs (wrt polymorphic vs non-polymorphic).
As far as I can see, the behaviors look correct.

Not sure if it makes sense to do so, but I could either add the test
script to an existing regression test, or create a new one with it if
desired.

If there are no objections, please apply.

Thanks,

Joe