[postgis-users] Partitioning spatial table

[David Bitner]

The table that I have that the bitmap indices excite me about is a
table of flight track records where the main two search criteria are
time (start, end as attributes-- not necessarily a 4th d) and place. 
In addition I also need to do dozens of queries using various other
fields (what airport, arrival or departure, aircraft type, etc...). It
is the huge list of unpredictable queries that make me interested in
the bitmap index.

My assumption is that the bitmap will deal with both GIST and BTREE so
it won't matter to mix and match (unlike creating a multi-column).  Is
this the case?

On 2/21/06, Markus Schaber <schabi at logix-tt.com> wrote:
> Hi, David,
>
> David Bitner wrote:
> >>By "multi-index capabilities" I wonder if David is referring to
> >>8.1's bitmap scans, which allow the planner to make use of disparate
> >>indexes.
> > Yep, that's exactly what I was referring to -- sorry for the lack of detail.
>
> You would need both a GIST index and a btree index on the integer column.
>
> However, in this case I think having partial indices is the more
> efficient case, especially because there are only a few combinations of
> queries, so we can "statically optimize" for them.
>
> For such a bitmap scan, the query planner has to scan both indices to
> create the bitmap in memory, and then access the table. With partial
> indices, it only has to access a single index and no in-memory bitmap.
>
> (The difference in planning time should be neglible, and can be
> eliminated by using prepared statements.)
>
> HTH,
> Markus
> --
> Markus Schaber | Logical Tracking&Tracing International AG
> Dipl. Inf.     | Software Development GIS
>
> Fight against software patents in EU! www.ffii.org www.nosoftwarepatents.org
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>