[postgis-devel] Call for 1.4.2 and 1.5.1 (Handling of Invalid Geometries)

[strk]

All in all I'm really not convinced the ST_MakeValid will
return a valid representation of what the input data intended
to express. I don't think any such process will ever exist.
As Paul recently said (was it him?) it's hard when the computers
start doing "what I mean" rather than "what I say".

And what you say to shp2pgsql is "import this data", not "clean
it up in an arbitrary way before importing it".

As a note, my client explicitly mentioned they would NOT want
the data to be automatically cleaned but would rather have
some triggers added to a table to *catch* the invalid geometries
to back them up in an specific table to send the original producer
with requests-to-cleanup. The actual ST_GeometryClean is just a
temporary solution they'd use until they have something "certified"
to be valid....

--strk;

On Wed, Feb 24, 2010 at 09:55:14AM -0800, Kevin Neufeld wrote:
> What if the shp loader were modified so that it didn't use 
> addgeometrycolumn but rather apply the constraints *after* the table was 
> created and loaded.
> 
> The loader could keep track of the geometry types it inserts and add an 
> appropriate common denominator geometrytype constraint.
> 
> (I was going to suggest the loader could use populate_geometry_columns, but 
> I just discovered a bug in the script when trying to apply constraints to 
> such a table.)
> 
> -- Kevin
> 
> On 2/24/2010 9:18 AM, strk wrote:
> >Mmm, I'm not convinced about this.
> >The ST_MakeValid function easily returns GEOMETRYCOLLECTION in its attempt
> >to keep all points. This mean the geometries will not be accepted in a 
> >table
> >crated with ST_AddGeometryColumn because that call will enforce a 
> >constraint
> >on geometry type.
> >
> 
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-devel

-- 

  ()   Free GIS & Flash consultant/developer
  /\   http://foo.keybit.net/~strk/services.html