[postgis-devel] PostGIS 1.1.7

Markus Schaber schabi at logix-tt.com
Thu Dec 14 02:31:00 PST 2006

Hi, Mark,

Mark Leslie <mark at refractions.net> wrote:

> The conversion of curve to linestring will be done simply by
> approximating the curve by a given number of segments per quarter
> circle.  This is the same approach as is used by the buffer command,
> only likely with a higher default number of segments.
> The conversion back will be done by examining sets of points and looking
> for sequences of segments of equal length with the same angle from the
> previous segment.  This way, if the original curve is only partially
> included in the final geometry, that bit would be recognized and a new
> curve constructed.

Ok. You should add some rounding-error tolerance to this recogniton
algorithm, and see that the first and last segment could be shorter
when the "cutting point" is between two of your vertices. But I see
that it can work.

> >> Eventually GEOS will be upgraded to support them properly.
> > So GEOS will finally fork from JTS?
> I don't expect so.  This step is still a ways away, but it's expected
> that the upgrade will first go into JTS, then be ported back into GEOS.

Ah, that's fine, then I can hook my PLJava mapping (which is only
implemented for JTS currently) directly into this extensions.

> >> As for the WKT/WKB, full parsing support isn't there yet.
> > 
> > What's missing there?
> The EWKT parser will accept the new types (eg. CIRCULARSTRING(0 0,0 1,1
> 1)) but the higher dimensional definitions aren't up to the SQL-MM spec
> yet.  The parser will accept CIRCULARSTRING(0 0 0, 0 1 3, 1 1 6) for
> example, while the spec demands CIRCULARSTRING Z (0 0 0,0 1 3,1 1 6).

Ah, I see.

Is the SQL-MM spec "downwards compatible" to OGC WKT / WKB and without
conflicts with our EWKB / EWKT?


Markus Schaber | Logical Tracking&Tracing International AG
Dipl. Inf.     | Software Development GIS

Fight against software patents in Europe! www.ffii.org

More information about the postgis-devel mailing list