Re: Compatible defaults for LEAD/LAG

On 5/31/20 9:53 PM, Tom Lane wrote:
> Vik Fearing <vik@postgresfriends.org> writes:
>>   postgres=# SELECT LAG(n, 1, -99) OVER (ORDER BY n)
>>   postgres-# FROM (VALUES (1.1), (2.2), (3.3)) AS v (n)
>>   postgres-# ORDER BY n;
>>   ERROR:  function lag(numeric, integer, integer) does not exist
>>   LINE 1: SELECT LAG(n, 1, -99) OVER (ORDER BY n)
>>                ^
>
> Yeah, we have similar issues elsewhere.
>
>> Attached is a patch that fixes this issue using the new anycompatible
>> pseudotype.  I am hoping this can be slipped into 13 even though it
>> requires a catversion bump after BETA1.
>
> When the anycompatible patch went in, I thought for a little bit about
> trying to use it with existing built-in functions, but didn't have the
> time to investigate the issue in detail.  I'm not in favor of hacking
> things one-function-at-a-time here; we should look through the whole
> library and see what we've got.
>
> The main thing that makes this perhaps-not-trivial is that an
> anycompatible-ified function will match more cases than it would have
> before, possibly causing conflicts if the function or operator name
> is overloaded.  We'd have to look at such cases and decide what we
> want to do --- one answer would be to drop some of the alternatives
> and rely on the parser to add casts, but that might slow things down.
>
> Anyway, I agree that this is a good direction to pursue, but not in
> a last-minute-hack-for-v13 way.

Fair enough.  I put it in the commitfest app for version 14.
https://commitfest.postgresql.org/28/2574/
--
Vik Fearing