[postgis-devel] The index is not optimal. To optimize it, contact a developer, or try to use the column as the second one in the CREATE INDEX command.

[Sandro Santilli]

On Fri, Feb 04, 2022 at 01:52:30PM +0100, Sandro Santilli wrote:
> On Fri, Feb 04, 2022 at 01:07:42PM +0100, Sandro Santilli wrote:
> > On Fri, Feb 04, 2022 at 02:26:48PM +0300, Darafei "Komяpa" Praliaskouski wrote:
> > 
> > > What branch is it on?
> > 
> > Master branch as of commit 0ffe2870eade1a26a1573eb11e7540eacc207783.
> > 
> > > I touched picksplit on master but it passed the tests fine.
> > 
> > It's not an error, just a DEBUG with an HINT.
> > 
> > > This normally can happen if you're pushing dozens of same values into
> > > index. I'd also expect it on EMPTY's. May happen if you're having a lot of
> > > updates on one row without vacuum collecting stuff.
> > 
> > Topology building does have a lot of updates of the same rows over and
> > over again (and never expliclty runs vacuum). Maybe this would be good
> > to change.
> 
> Is there now a way to decide which picksplit method to use for a given
> index ? Because during execution of topology building function there's
> no way to VACUUM becuase changes are all still not committed, so the
> index will always (for such batch loads).

For the record: I've filed https://trac.osgeo.org/postgis/ticket/5083
to evaluate a change of tables layout to reduce this problem.

--strk;