Re: cost_sort() improvements

On 06/29/2018 04:36 PM, Teodor Sigaev wrote:
>
>
> Peter Geoghegan wrote:
>> On Thu, Jun 28, 2018 at 9:47 AM, Teodor Sigaev <teodor(at)sigaev(dot)ru> wrote:
>>> Current estimation of sort cost has following issues:
>>>   - it doesn't differ one and many columns sort
>>>   - it doesn't pay attention to comparison function cost and column
>>> width
>>>   - it doesn't try to count number of calls of comparison function on
>>> per
>>> column
>>>     basis
>>
>> I've been suspicious of the arbitrary way in which I/O for
>> external sorts is costed by cost_sort() for a long time. I'm not
>> 100% sure about how we should think about this question, but I am
>> sure that it needs to be improved in *some* way.
>>
> Agree, but I think it should be separated patch to attack this
> issue. And I don't have an idea how to do it, at least right now.
> Nevertheless, I hope, issue of estimation of disk-based sort isn't a
> blocker of CPU cost estimation improvements.
>

Yes, I agree this should be addressed separately. Peter is definitely
right that should look at costing internal vs. external sorts (after
all, it's generally foolish to argue with *him* about sorting stuff).

But the costing changes discussed here are due to my nagging from the
GROUP BY patch (and also the "incremental sort" patch). The internal vs.
external costing seems like an independent issue.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services