Amit V Shah wrote:
> Hi all,
>
> I have like a repository table with is very very huge with atleast a few
> hundreds of millions, may be over that. The information is stored in form of
> rows in these tables. I need to make that information wide based on some
> grouping and display them as columns on the screen.
>
> I am thinking of having a solution where I create views for each screen,
> which are just read only.
>
> However, I donot know if the query that creates the view is executed
> everytime I select something from the view. Because if that is the case,
> then I think my queries will again be slow. But if that is the way views
> work, then what would be the point in creating them ..
That's exactly how they work. You'd still want them because they let you
simplify access control (user A can only see some rows, user B can see
all rows) or just make your queries simpler.
Sounds like you want what is known as a "materialised view" which is
basically a summary table that is kept up to date by triggers. You query
the table instead of actually recalculating every time. Perhaps google
for "postgresql materialized view" (you might want a "z" or "s" in
materialised).
--
Richard Huxton
Archonet Ltd
