I am very interested in this patch because it allows to use partial indexes to ... speed up inserts.
I have implemented "ALTER INDEX ... WHERE ..." construction which allows to change predicate of partial index without
necessityto fully rebuild it.
So it is not necessary to insert new records in index immediately (if new records do not match partial index
conditions).
It can be done later in background (or at night). My experiments show that it allows to increase insert speed five
times(for either partial indexes).
At the same time we do not loose RDBMS requirement that result of query should not depend on presence of indexes. And
itis applicable to all indexes: B-Tree, GIN, GIST,...
But such optimization makes sense only of partial indexes can be used without extra overhead, first of all for
index-onlyscans.
And it is impossible without this patch.
On 01/31/2016 03:34 PM, Alvaro Herrera wrote:
> Tomas Vondra wrote:
>
>> On 12/24/2015 04:05 AM, Michael Paquier wrote:
>>> Tomas, are you still working on that? This thread is stalling for 3 weeks.
>> I haven't discovered anything interesting during the testing, so I guess the
>> "needs review" state is appropriate. Let's move the patch to the next
>> commitfest.
> Not sure what to do here, since this patch got no feedback at all in
> this CF. The right thing to do, ISTM, is to just move it again to the
> next CF. But it'd be really useful if someone can have it a look and
> verify at least whether it doesn't need a rebase (requiring a further
> submission) so that other people can play with it. Of course, if
> Horiguchi-san or anyone has more review comments, that would be even
> better.
>
> Tomas said he'd do more testing, but we never got a report on whether
> anything turned up.
>
> (At this point I'm not sure if either Kyotaro or Tomas should be
> considered the patch author ... maybe both?)
>
--
Konstantin Knizhnik
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company
