On Wed, Oct 24, 2018 at 10:07 AM Konstantin Knizhnik <
k.knizh...@postgrespro.ru> wrote:
>
> Real number of distinct value for this dataset is about 10 millions. For
> some reasons, sampling using random blocks and Vitter algorithm produces
> worser results than just examining first 3 rows of
Konstantin Knizhnik writes:
> I will be pleased if somebody (first of all Robert) can comment me
> "strange" results of distinct values estimation.
Estimating the number of distinct values from a small sample is a hard
problem; every algorithm is going to blow it in some cases.
> In my case the
Hello hackers,
I will be pleased if somebody (first of all Robert) can comment me
"strange" results of distinct values estimation.
There is the following code in analyze.c:
/*--
* Estimate the number of distinct values using the estimator
* propos
