Re: Fix gin index cost estimation

Le vendredi 2 décembre 2022, 13:58:27 CET Ronan Dunklau a écrit :
> Le vendredi 2 décembre 2022, 12:33:33 CET Alexander Korotkov a écrit :
> > Hi, Ronan!
> > Thank you for your patch. Couple of quick questions.
> > 1) What magic number 50.0 stands for? I think we at least should make
> > it a macro.
>
> This is what is used in other tree-descending estimation functions, so I
> used that too. Maybe a DEFAULT_PAGE_CPU_COST macro would work for both ? If
> so I'll separate this into two patches, one introducing the macro for the
> other estimation functions, and this patch for gin.

The 0001 patch does this.

>
> > 2) "We only charge one data page for the startup cost" – should this
> > be dependent on number of search entries?

In fact there was another problem. The current code estimate two different
pathes for fetching data pages: in the case of a partial match, it takes into
account that all the data pages will have to be fetched. So this is is now
taken into account for the CPU cost as well.

For the regular search, we scale the number of data pages by the number of
search entries.

Best regards,

--
Ronan Dunklau