[geotk] Is WKT always sufficient to define a CRS?

[Martin Desruisseaux]

Hello Jon

Le 02/08/11 18:38, Jon Blower a écrit :
> * If I'm using the embedded EPSG database within Geotoolkit, can I get hold of the Proj4 string in my application?
We don't have yet a mechanism for formatting Proj4 string from a CRS. However 
this has been a request for a long time, something we will have to do soon or 
later. In the main time, I think that something along that line would work for 
ProjectedCRS:

private final Properties OGCtoProj4 = ...; // TODO: This is the tedious part, see below.

private String getProj4Name(IdentifiedObject object) {
     return OGCtoProj4.get(IdentifiedObjects.getIdentifiers(object, Citations.OGC).getCode());
}

public String getProj4String(ProjectedCRS crs) {
     StringBuilder proj4String = new StringBuilder();
     proj4String.append("proj=").append(getProj4Name(crs));
     for (GeneralParameterValue param : crs.getConversionFromBase().getParametersValues().values()) {
         double value = ((ParameterValue) param).doubleValue();
         proj4String.append(" +").append(getProj4Name(param)).append('=').append(value);
     }
     return proj4String.toString();
}


The part which would be a bit tedious would be to fill a map of "OGC to Proj4" 
name, something like that:

Albers_Conic_Equal_Area = eae
standard_parallel_1 = lat_1
standard_parallel_2 = lat_2
latitude_of_center = lat_0

etc...

There is a high probability that this mapping already exists somewhere. With 
such mapping, I think it would be trivial to add Proj4 formatting support 
straight into Geotk.

> * Do Proj4 string contain all the info from the EPSG database, or do they fall short in the ways you describe below?
As far as I know, they have the same limitation than WKT: no information about 
area of validity, scope, accuracy, aliases and remarks.

There is also one more thing which is unique to the EPSG database: for some 
pairs of CRS, the database specifies explicitly which transformation path should 
be used. Many transformation paths exist, for example a datum shift could be 
performed by "Geocentric translation", "Molodensky" or "Abridged Molodensky" 
methods among others. Without EPSG database, the method choosen is 
implementation-dependent. With an EPSG database, it is formaly defined. Simple 
syntax like WKT don't do that (well, a database of transformation paths with WKT 
would be possible, but I'm not aware of anyone doing that). Proj4 don't do 
neither, or at least the PostGIS "spatial_ref_sys" table doesn't contain any 
information about transformation paths.

     Regards,

         Martin