[postgis-users] beginner questions re. SRID

[strk at refractions.net]

I exclude performance being hit by SRID in any way, except
by the fact that a geometry with SRID=-1 will be smaller
thus faster to load in memory (hardly noticeable).
--strk;

On Fri, Oct 21, 2005 at 04:34:19PM +0200, christomec wrote:
> I didn't bother setting my SRID when creating my tables. I thought I'd
> leave it for later, when I better understand this whole projection thing.
> 
> Do you mean to say that having all my SRID's as -1 influences my query
> performance.  I mainly make use of  the following type of queries.
> 
> PreparedStatement cadasteQuery = conn.prepareStatement(
>                     "SELECT ((('<path id=\"'::text ||
> cadastre.erf_number::text) || '\" d=\"'::text) ||
> assvg(intersection(cadastre.the_geom,
> expand(GeomFromText(?,-1), ?)), 0, 6)) ||
> 'z\"/>'::text AS path " +
>                     "FROM cadastre "+
>                     "WHERE the_geom && Expand(GeomFromText(?,-1),?) "+
>                     "AND Distance(GeomFromText(?,-1),the_geom) < ?;");
> 
> > Hi, Paul and Rick,
> >
> > Paul Ramsey wrote:
> >
> >>> Is there a more human-readable table that contains the SRIDs and
> >>> descriptions?
> >>
> >> More than spatial_ref_sys? Not really. The SRTEXT isn't completely
> >> *un*readable though.
> >
> > It may be that the EPSG database (where the spatial_ref_sys is derived
> > from) contains some more verbose descriptions, but I doubt that it will
> > be of much more use.
> >
> >> Yes! Confusing! But part of the standard.  The specification says  that
> >> all geometries have an SRID.  It also says that the  geometry_columns
> >> table contains a single SRID for each spatial column.
> >
> > Note, however, that in PostGIS you can avoid this constraint and create
> > geometry columns that contain mixed-srid data, or even geometries
> > without any specified SRID (represented as -1). PostGIS will work fine,
> > but you're not OpenGIS standard compliant any more, and index usage may
> > suffer.
> >
> >>> Can it properly compare an object with one SRID with an object with  a
> >>> different SRID?
> >>
> >> No, and most/many functions will error out if you try, because it is
> >> not really correct to compare geometries with different SRID (unless
> >> they happen to map to the same projection! ha!).
> >
> > One can use something like
> >
> >         intersection(geoma, transform(geomb, srid(geoma)))
> >
> > to compare geometries which (potentially) have different SRIDs.
> >
> > It works fine, but may be rather slow depending on your exact data and
> > query.
> >
> >
> > Markus
> > _______________________________________________
> > postgis-users mailing list
> > postgis-users at postgis.refractions.net
> > http://postgis.refractions.net/mailman/listinfo/postgis-users
> >
> > !DSPAM:4358f904214901269420936!
> >
> >
> >
> 
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users