Ramy,
glad to hear from you !
AFAIK, posgresql doesnt' supports several indices for the same type.
I think this is a problem of optimizer. Probably other hackers know
better. I forward your message to -hackers mailing list which is a
relevant place for GiST discussion.
regards, Oleg
On Tue, 9 Nov 2004, Ramy M. Hassan wrote:
> Dear Oleg and Teodor,
> Thanks for offering help.
> I have a design question for now.
> Currently in the postgresql GiST implementation, I noticed that the way to
> have a GiST based index is to define an operator class for a certain type
> using GiST index. There is no new index type defined from the point of view
> of postgresql ( nothing is added to pg_am ). This means that for a certain
> type there could only be one GiST based index. I mean that there is no way in
> the same server to use gist to implement an xtree index and a ytree for the
> same type even if they index different fields in different relations. is
> that correct ?
> What about doing it the other way ( I am talking about SP-GiST now ) , by
> providing the extension writer with an API to use it to instantiate a
> standalone SP-GiST based index ( for example trie index ) that has a record
> in the pg_am relation. In my point of view this would give more flexibility,
> and also would not require the extension writer to learn the postgresql API (
> maybe oneday SP-GiST will be ported to another database engine ) he will
> just need to learn the SP-GiST API which will propably be less amount of
> study (and this is what GiST and SP-GiST is all about if I correctly
> understand ).
> Please let me know your opinions regarding to this.
>
> Thanks
>
> Ramy
>
Regards, Oleg
_____________________________________________________________
Oleg Bartunov, sci.researcher, hostmaster of AstroNet,
Sternberg Astronomical Institute, Moscow University (Russia)
Internet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/
phone: +007(095)939-16-83, +007(095)939-23-83