On Sun, Nov 3, 2013 at 3:01 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Michael Paquier <michael.paquier@gmail.com> writes:
>> On Sat, Nov 2, 2013 at 10:28 PM, Andrew Dunstan <andrew@dunslane.net> wrote:
>>> What exactly would be the point? Indexes are automatically maintained by
>>> postgres. Something that isn't doesn't seem to me to qualify for the
>>> description of "index".
>
>> Perhaps an index without data that could be used by the planner for
>> automatic query tuning to evaluate how a query could run if the index
>> exists? Like the concept of hypothetical indexes or something like the
>> possibility to do a CREATE/ALTER INDEX ... WITH [ NO ] DATA.
>
> But we already provide support for hypothetical indexes via planner
> plugins. Why would you need an actual empty index underlying that?
Using only the planner hook? Forgive my lack of knowledge of the
planner, but this does not seem really straight-forward :)
One of the only documentation I found about hypothetical indexes was
this tutorial of PGCon 2010:
http://www.inf.puc-rio.br/~postgresql/conteudo/projeto1/Tutorial9/TUTORIAL_9_0_1.pdf
But in this case the grammar of CREATE INDEX has been modified, so
core code was changed as well.
Of course if Naman has something else in mind...
--
Michael
