[EMAIL PROTECTED] writes:
> Tom, I posted a message about a week ago (I forget the name) about a
> persistent reference index, sort of like CTID, but basically a table
> lookup. The idea is to simulate a structure that ISAM sort of techniques
> can work in PostgreSQL.
> Eliminating the bitmap index issue for a moment, how hard would it be to
> create a reference table like index?
I didn't see the point. You cannot actually use CTID that way (at least
not without fundamentally redesigning our MVCC machinery) and anything
else you might come up with is effectively just a random unique ID that
has to be mapped through an index to find the row. I don't see the
advantage compared to any ordinary application-defined primary key.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 8: explain analyze is your friend
