[Gdal-dev] python gdal 1.3.2 and geos

[Didrik Pinte]

Le lundi 04 septembre 2006 à 10:44 -0400, Frank Warmerdam a écrit :
> Didrik Pinte wrote:
> > Le lundi 04 septembre 2006 à 12:25 +0200, Didrik Pinte a écrit :
> >> Any hint to solve the problem ?
> > 
> > Because of a problem between gdalnumeric and python-numpy, I have
> > uninstalled the following packages and kept only python-numeric :
> > 
> > apt-get remove python-numpy python-numpy-ext python-scipy fftw3
> 
> Didrik,
> 
> Well, I wouldn't have imagined this connection!

Nor do I ;-)

> > It solved the problem ! That's strange. When looking a the org.py code
> > of Disjoint, I'm wondering where I can find a link between the python
> > libs and the _gdal code :
> > ----------------------------------------------------------------------
> >    def Disjoint( self, other_geom ):
> >         return _gdal.OGR_G_Disjoint( self._o, other_geom._o )
> > ----------------------------------------------------------------------
> 
> This would appear to be from the 'old style' python bindings.
> 
> For these there is a gdal.i interface file which swig (1.1) turns into a
> gdal_wrap.c with C-to-python wrappers for all the published functions.
> This compiles to _gdalmodule.so on unix.  So the python wrapper layer in
> ogr.py imports _gdal and then calls the C functions in it.
> 
> Does that answer your question?

For sure. Thank you for the clear explanation. That's what I thought
some minutes after having sent the message. 

I've suggested that the debian package python-gdal conflict with
python-numpy in order to get a working system in any case (this will
appear in the debian bts today).

Mateusz, did you had time to work on the ng bindings ? Do you need
help ?

Best regards,

Didrik