G'day Sarah I'm afraid I'm no expert on geography types, but yes, in some cases st_area can process invalid geometries (like you have described), but they are after all invalid geometries, so it returns 0 . For example, looking at your second example :
=# SELECT ST_AREA(the_geom), ST_ISVALID(the_geom), ST_ISVALIDREASON(the_geom), ST_SUMMARY(the_geom) FROM (SELECT ST_GEOMFROMTEXT('POLYGON((15 -5, 16 0, 15 0, 16 -2, 15 -5))') the_geom) as foo; NOTICE: Self-intersection at or near point 15.7143 -1.42857 st_area | st_isvalid | st_isvalidreason | st_summary ---------+------------+-------------------------------------------------------+------------------------- 0 | f | Self-intersection[15.7142857142857 -1.42857142857143] | + | | | Polygon[B] with 1 rings+ | | | ring 0 has 5 points + | | | (1 row) it is not because one side rules is the inverse of the other that the area is 0, it just returns 0 not an error. My guess is that the invalid geography is causing an uncaught error, but there is no st_isvalid(geography), hence using it on geometry type. This would be my approach to finding the invalid geometries. How you repair them...? I guess it would depend on how they were originally collected, how much each one cost to collect or what they meant. cheers Ben On 25/07/2011, at 5:27 AM, Sarah Berke wrote: > I've figured out a little more detail about my ST_Area problem, perhaps this > might help someone give me some advice for troubleshooting. To provide a > little more background, I've generated several thousand polygons using > ST_ConvexHull from collections of points (I've tried both > ST_ConvexHull(ST_Collect(points)) and ST_ConvexHull(ST_Union(points)) ) and > when I try to calcuate their areas I get this error: > > ERROR: ptarray_area_spheroid: cannot handle ptarray that crosses equator > CONTEXT: SQL function "st_area" statement 1 > > The problem seems to be that ST_Area cannot handle some polygons with > internal crosses (i.e. bowtie-shaped) that share a border with the equator, > even though it can handle them elsewhere. For example: > > SELECT ST_AREA(ST_GEOGFROMTEXT('POLYGON((15 -5, 15 5, 16 0, 16 -5, 15 -5))')) > ; > works just fine--this is just a simple polygon that crosses the equator. By > the same token, a simple polygon that borders the equator works: > > SELECT ST_AREA(ST_GEOGFROMTEXT('POLYGON((15 -5, 15 0, 16 0, 16 -5, 15 -5))')) > ; > > Switching the 2nd and 3rd coordinates, however, makes a bowtie-shaped polygon > which is valid, but it shares a border with the equator and ST_Area cannot > take it: > > SELECT ST_AREA(ST_GEOGFROMTEXT('POLYGON((15 -5, 16 0, 15 0, 16 -5, 15 -5))')) > ; > ERROR: ptarray_area_spheroid: cannot handle ptarray that crosses equator > CONTEXT: SQL function "st_area" statement 1 > The same kind of formation works fine, however, if it does not share a border > with the equator : > > SELECT ST_AREA(ST_GEOGFROMTEXT('POLYGON((15 -5, 16 1, 15 0, 16 -5, 15 -5))')) > ; > SELECT ST_AREA(ST_GEOGFROMTEXT('POLYGON((15 -4, 16 1, 15 1, 16 -4, 15 -4))')) > ; > > > Again, here are my system details: > PostgreSQL 8.4.8 on x86_64-pc-linux-gnu, compiled by GCC gcc-4.4.real > (Ubuntu 4.4.3-4ubuntu5) 4.4.3, 64-bit > POSTGIS="1.5.2" GEOS="3.3.0-CAPI-1.7.0" PROJ="Rel. 4.7.1, 23 September > 2009" LIBXML="2.7.6" USE_STATS > > I assume that one or more of the polygons I made with ConvexHull is a > bowtie-shape sharing a border with the equator, but why would ST_ConvexHull > generate a shape like that? How can I avoid this problem? Short of visually > examining the map of every one of these several thousand polygons, how do I > even figure out which one is the culprit? Is there a better way to go from > points or multipoints to polygons? > > Any insight would be really great! Thanks! > > _________________________________ > Sarah K Berke > Postdoctoral Scholar > Department of the Geophysical Sciences > University of Chicago > 5734 S. Ellis Ave > Chicago, IL 60637 > > > Date: Fri, 22 Jul 2011 14:39:52 -0500 > From: Sarah Berke <skbe...@uchicago.edu> > Subject: [postgis-users] ERROR: ptarray_area_spheroid: cannot handle > ptarraythat crosses equator > To: postgis-users@postgis.refractions.net > Message-ID: > <cajupzdeptykhb4log1szea0msprgikpfhm70fh_atbbuxso...@mail.gmail.com> > Content-Type: text/plain; charset="iso-8859-1" > > Hello, > > I found a short thread on this subject from February 2011--I am now having > the same problem that Marcello had, and I wonder if anyone has found a fix > or workaround for this problem? Using the same example code that Marcello > provided: > > SELECT 161109 polygon ,ST_AREA(ST_GEOGFROMTEXT('POLYGON((15.8327132977611 > 0.25,16 0.25,16 0,15.8566682147637 0,15.8566112518311 > 0.00356000009924173,15.8485898971558 0.124770000576973,15.8327132977611 > 0.25))')) gsareageo; > > I get > > ERROR: ptarray_area_spheroid: cannot handle ptarray that crosses equator > CONTEXT: SQL function "st_area" statement 1 > > I am using: > > PostgreSQL 8.4.8 on x86_64-pc-linux-gnu, compiled by GCC gcc-4.4.real > (Ubuntu 4.4.3-4ubuntu5) 4.4.3, 64-bit > POSTGIS="1.5.2" GEOS="3.3.0-CAPI-1.7.0" PROJ="Rel. 4.7.1, 23 September > 2009" LIBXML="2.7.6" USE_STATS > > >From the earlier thread, it sounds as if this version of postgis should run > the code just fine? I saw that Marcello was able to work around the problem > by changing 0 lats to being very near-0; but this of course affected the > calculated areas. Can anyone guess why this is a problem, or has anyone > figured out a better solution? > > Thanks very much! > _________________________________ > Sarah K Berke > Postdoctoral Scholar > Department of the Geophysical Sciences > University of Chicago > 5734 S. Ellis Ave > Chicago, IL 60637 > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: > <http://postgis.refractions.net/pipermail/postgis-users/attachments/20110722/ec2f1ed8/attachment.html> > > _______________________________________________ > postgis-users mailing list > postgis-users@postgis.refractions.net > http://postgis.refractions.net/mailman/listinfo/postgis-users
_______________________________________________ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users