[postgis-devel] Re: [postgis-users] Re: Buggy transform() function in 1.0RC1 ?

[strk at refractions.net]

Nicolas, that was an excellent test. Thank you !
For all the "equivalents" starting at 27561, 27571, 27581, 27591
and taking 4 slots, I changed lon_0 to 0 in the proj4text.

Now, what about the srtext value ? Should it be changed as well ?

--strk;

On Wed, May 18, 2005 at 02:16:05PM +0200, Nicolas Ribot wrote:
> Hi Sandro and Paul,
> 
> Here is a SQL file whose purpose is to test some french projections.
> 
> It is pretty verbose, but it helped me organize my thoughts concerning
> french projection.
> 
> Basically, it contains:
> - A description of the official french planar projection: Lambert Zone
> (France is cut in 4 zones).
> - 8 reference points, 2 on each zone, with their official coordinates
> (taken from IGN database)
> - SQL code that creates the reference table with reference points,
> then reprojects these points in Lambert 93 and Lambert zone, computing
> the range between projected points and reference values.
> 
> Tests were performed on Windows XP, SP2. Postgresql 8.0.2, Postgis
> 1.0.0, Geos 2.1.1, Proj4 4.4.9.
> 
> One can see: Lambert 93 definition is correct.
> Lambert Zone definitions are NOT correct, with big errors on X
> coordinates, due to a bad prime meridian.
> 
> To obtain correct values, proj4 parameter +lon_0=2.33722917 was
> changed to: +lon_0=0:
> Prime meridian for these projections is Paris, as stated in the +pm
> parameter, so it wrong to provide a shift for the longitude of origin.
> 
> There are many equivalent Lambert zone projections defined in
> spatial_ref_sys.sql, with SRID 27561 to 27564, 27571, 27581, 27591.
> Only the name of the projection changes.
> All these 16 definitions are wrong and should be corrected for the
> +lon_0 parameter in order to provide correct results.
> 
> I'm currently setting up a test case for another French spatial
> reference system: geographic NTF with grade units.
> I will add it to the file I send to you.
> 
> I don't know if it is the kind of test case you expected. Let me know.
> 
> Regards
> 
> Nicolas Ribot