Re: JSON constructors and window functions

Hi,

On 4/4/22 2:19 PM, Andrew Dunstan wrote:
> 
> On 4/4/22 12:33, Andres Freund wrote:

> It can, as Jaime's original post showed.
> 
> But on further consideration I'm thinking this area needs some rework.
> ISTM that it might be a whole lot simpler and comprehensible to generate
> the json first without bothering about null values or duplicate keys and
> then in the finalizer check for null values to be skipped and duplicate
> keys. That way we would need to keep far less state for the aggregation
> functions, although it might be marginally less efficient. Thoughts?

This is still on the open items list[1]. Given this is a 
user-triggerable crash and we are approaching PG15 Beta 1, I wanted to 
check in and see if there was any additional work required to eliminate 
the crash, or if the work at this point is just optimization.

If the latter, I'd suggest we open up a new open item for it.

Thanks,

Jonathan

[1] https://wiki.postgresql.org/wiki/PostgreSQL_15_Open_Items