[postgis-users] PostGIS Geography and Rhumb Lines

[Sandro Santilli]

I would just cast the result to a geography:

  ST_Segmentize(...)::geography

--strk;

On Wed, Nov 28, 2012 at 11:44:23AM -0800, Derek Meyer wrote:
> Outstanding, thanks guys!
> 
> So am I correct that with something like the below I'd have a
> resulting geometry of coords out of a 2D plain, correct?  Assuming I
> got an acceptable SRID for Mercator there (3857?).
> 
> ST_Segmentize(ST_GeomFromText('LINESTRING(-29 -27,-30 -29.7,-36
> -31,-45 -33)'), 3857), 5)
> 
> Then when worked into a query that bundles everything up via
> ST_GeographyFromText, that each tiny segment would be technically
> great-circle again, but *overall* a reasonable rhumb line
> representation cast by the previous segmentize?
> 
> Makes sense to me.  My only other question would be probably a newbie
> syntactical question regarding how to get the above segmentize result
> within the SQL query.  Would I be using ST_AsText to return it to WKT
> and concating it somehow within my ST_GeographyFromText line?  I'm
> sure the below would probably break but, to see what I'm getting at...
> 
> ST_GeographyFromText('GEOMETRYCOLLECTION(LINESTRING(-28 -25, -27.6
> -25.5), ' .ST_AsText(ST_Segmentize(ST_GeomFromText('LINESTRING(-29
> -27,-30 -29.7,-36 -31,-45 -33)'), 3857), 5)). ')');
> 
> Just struggling to find an example online.  But given that, I'm just
> about there.  Thanks again!
> 
> - Derek
> 
> On Wed, Nov 28, 2012 at 8:47 AM, Paul Ramsey <pramsey at opengeo.org> wrote:
> > ST_Segmentize in geography will just segmentize along the great
> > circle. You'll have to choose whether you want your model in geometry
> > or geography. If geometry, then you can use Mercator as your
> > projection (rhumb lines are then straight line) and use
> > ST_Segmentize(geography) to convert your great circles into
> > approximations in mercator.
> > Or, you can use geography, and leave your great circle lines alone,
> > but project your rhumb lines into geometry in mercator, st_segmentize
> > them *there*, then convert htem back to geography.
> > Clear as mud?
> > P.
> >
> > On Wed, Nov 28, 2012 at 4:33 AM, Sandro Santilli <strk at keybit.net> wrote:
> >> PostGIS-2.1 (in development) has an ST_Segmentize implementation for
> >> "geography" type, which should help with rhumb lines.
> >>
> >> You may or may not need to segmentize depending on the operation you
> >> want to do on those lines. Anyway it would be practical to use in your
> >> SQL queries if you don't want to store the segmentization persistently.
> >>
> >> --strk;
> >>
> >> On Wed, Nov 28, 2012 at 01:39:09AM -0800, Derek Meyer wrote:
> >>> My current project involves creating a spatial database of world-wide
> >>> airspace data, where each known flight information region is a polygon
> >>> containing boundaries that are a combination of great circle lines,
> >>> rhumb lines, and arcs.
> >>>
> >>> It seems straight forward enough that these polys could be easily
> >>> inserted into the database via ST_GeographyFromText fed with a WKT
> >>> GEOMETRYCOLLECTION of these compound boundary types, but I'm wondering
> >>> about how to deal with rhumb lines in such a case since the Geography
> >>> functions, if I understand correctly, base everything around a round
> >>> earth model and utilize great-circle by nature.
> >>>
> >>> Is there any way to explicitly declare a rhumb line in an insert query
> >>> such as this?  I've seen some examples elsewhere where a temporary
> >>> projection to 2D is utilized and then segmentized to
> >>> render/interpolate coordinates for a linestring that would more
> >>> accurately represent said rhumb line.  Is this still the accepted
> >>> approach?  Practical?  If so, how would such an approach be
> >>> implemented in context of the SQL query?
> >>>
> >>> Any info/suggestions welcome and appreciated.
> >>>
> >>> Thank you for your time,