[postgis-users] Detecting wrong geometries

[Nicolas Ribot]

On 12 August 2010 18:51, Andrea Peri <aperi2007 at gmail.com> wrote:
> Hi,
>
> I needed to test a shapefile to find eventually wrong geometries.
>
> My first idea was to use postgis to do a
> select *from table_from_shapefile where ST_IsValid(geom)=false;
>
> So I load the shapefile with shp2pgsql in postgis (1.5.1)
> and do this query.
> The results was 1 only geometry wrong.
>
> After some days I repeat the same test with Grass.
>
> Grass with my big surprise report me more geometry wrong.
> Infact in the shapefile was detected many geometries with wrong orientation.
>
> Now my question is why this was not detected from postgis ?
> I don't know why,
> but my idea is that perhaps shp2pgsql, when find a geometry wrong oriented
> correct it automatically .
>
> Is this correct ?
>
> Thx,
>

Hi,

Polygon orientation is not important with respect to OGC
specifications and does not define an invalid polygon.
That's why st_isvalid() returns true in Postgis.
You can, however, force polygon orientation using the st_forceRHR()
function (http://postgis.refractions.net/docs/ST_ForceRHR.html).

Nicolas

PS. the Jump software (http://www.vividsolutions.com/jump/) has a nice
validation tool in which you can enable this constraint when
validating a layer.
With the postgis plugin, you can perform this validation directly on a
postgis table.
It graphically spots invalid geometries and also generate a new layer,
with attributes telling the reason of invalidity.