Re: [RFC] speed up count(*)

On 10/21/21 4:29 PM, Joe Conway wrote:
> On 10/21/21 4:23 PM, Robert Haas wrote:
>> On Thu, Oct 21, 2021 at 4:19 PM Joe Conway <mail(at)joeconway(dot)com> wrote:
>>> That is a grossly overstated position. When I have looked, it is often
>>> not that terribly far off. And for many use cases that I have heard of
>>> at least, quite adequate.
>>
>> I don't think it's grossly overstated. If you need an approximation it
>> may be good enough, but I don't think it will often be exactly correct
>> - probably only if the table is small and rarely modified.
>
> meh -- the people who expect this to be impossibly fast don't
> typically need or expect it to be exactly correct, and there is no way
> to make it "exactly correct" in someone's snapshot without doing all
> the work.
>
> That is why I didn't suggest making it the default. If you flip the
> switch, you would get a very fast approximation. If you care about
> accuracy, you accept it has to be slow.
>

I don't think we really want a switch for "inaccurate results
acceptable", and I doubt the standard would accept an approximation for
count(*).

But something else that gave a fast approximate answer
("count_estimate(*)"?) would be useful to many. Not portable but still
useful, if someone could come up with a reasonable implementation.

cheers

andrew

--
Andrew Dunstan
EDB: https://www.enterprisedb.com