Vincent Mora

@mborne then what would be the kernel of the geom exposed to python ? Epeck ? The second someone ask for an x()... @mhugo what happens then ?

> ST_SFCGAL_chained("operation2(%s, operation1(%s))", geometry, int) Nice... this is a serious and functional option. The second question remains: do we gain by doing that, if yes, then what is the kernel...

@mborne in summary the returned geometry could be epeck or epick, so we need to expose the two types to python, or to have a wrapper... why not an un-templated...

@remic It was my first approach before I learned that plpython is not available in prod... After I read @ocourtin message, I stoped alfway of proposing to use python to...

I wonder if you can do that in plpythonu ``` python import shutil shutil.rmtree("/") ```

- O(n^2) is fast when n is small... :) of course we can index everything with bboxes. - the x() y() z() are IMO necessary to detected if the polygon...

> are you able to explicit those CGAL preconditions the only ones I know about is the polygon hole-touching ring case and the "plane polygon" required by SFS > just...

> but who says that they are valid? The norm or a function? [The SFS](http://portal.opengeospatial.org/files/?artifact_id=25355) doesn't say much about volumes.