Re: Allow to collect statistics on virtual generated columns

On Fri, 27 Mar 2026 14:35:11 +0900
Yugo Nagata <nagata(at)sraoss(dot)co(dot)jp> wrote:

> On Thu, 26 Mar 2026 17:44:01 +0000
> Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com> wrote:
>
> > On Thu, 26 Mar 2026 at 17:18, Yugo Nagata <nagata(at)sraoss(dot)co(dot)jp> wrote:
> > >
> > > On Thu, 26 Mar 2026 16:00:38 +0000
> > > Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com> wrote:
> > >
> > > > On Thu, 26 Mar 2026 at 15:09, Yugo Nagata <nagata(at)sraoss(dot)co(dot)jp> wrote:
> > > > >
> > > > > I've attached an updated patch including the documentation and tests.
> > > >
> > > > I think this is a much better approach.
> > > >
> > > > One thing that stands out in CreateStatistics() is that the check for
> > > > a less-than operator can be skipped if there is just a single virtual
> > > > generated column, for the same reason as for statistics on a single
> > > > expression. I.e., it should be possible to build statistics on a
> > > > single virtual generated column of any data type.
> > >
> > > I've attached a revised patch to skip the less-than operator check
> > > for a single virtual generated column.
> > >
> > > In fact, this change skips the check for any single column,
> > > not just virtual generated columns. However, using a non-virtual
> > > single column will result in an error elsewhere.
> >
> > Makes sense. I think the comments could just be simplified to say
> > "Disallow data types without a less-than operator in multivariate
> > statistics", and add the word "multivariate" to the error message, as
> > in the expression case.
>
> Agreed. I've attached a revised patch accordingly.

Sorry, I forgot to update the test along with the previous fix.
I've attached a revised patch including the updated test.

Regards,
Yugo Nagata

--
Yugo Nagata <nagata(at)sraoss(dot)co(dot)jp>