Re: Reconcile stats in find_tabstat_entry() and get rid of PgStat_BackendFunctionEntry

Hi,

On 2/13/23 8:40 AM, Kyotaro Horiguchi wrote:
> At Mon, 13 Feb 2023 08:09:50 +0100, "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> wrote in
>>> I think this is useful beyond being able to generate those functions
>>> with
>>> macros. The fact that we had to deal with transactional code in
>>> pgstatfuncs.c
>>> meant that a lot of the relevant itnernals had to be exposed "outside"
>>> pgstat,
>>> which seems architecturally not great.
>>>
>> Right, good point.
> 
> Agreed.
> 
>> Removing the pfrees in V2 attached.
> 
> Ah, that sound good.
> 
>       if (!entry_ref)
> +    {
>           entry_ref = pgstat_fetch_pending_entry(PGSTAT_KIND_RELATION, InvalidOid, rel_id);
> +        return tablestatus;
> +    }
> 
> We should return something if the call returns a non-null value?

What we do is: if entry_ref is NULL then we return NULL (so that the caller returns 0).

If entry_ref is not NULL then we return a copy of entry_ref->pending (with or without subtrans).

> 
> So, since we want to hide the internal from pgstatfuncs, the
> additional flag should be gone. 

I think there is pros and cons for both but I don't have a strong opinion about that.

So also proposing V3 attached without the flag in case this is the preferred approach.

Regards,

-- 
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com