[gdal-dev] Re: OGR FileGDB driver: Failed at creating table ... (General function failure.)

[Hermann Peifer]

AFAICT, the issue is not so much that the FileGDB only supports certain 
projections, but that it only accepts a very specific WKT string for a 
given SRS. A single character of difference can cause the FileGDB API to 
not recognise the SRS, then fall back to a "General function failure".

GDAL/OGR has its own WKT definition of a given SRS, which is slightly 
different to ESRI's WKT of the same SRS. I noted that when creating a 
FileGDB, GDAL/OGR tries to ESRI'ify its WKT through the morphToESRI() 
function. For EPSG:3035, the result is obviously not "ESRI-friendly" 
enough. In fact, it is not what I first thought: an issue with the 
parameter name (Central_Meridian versus longitude_of_center) but simply 
the SRS's string identifier which causes the problem.

My quick-fix:
I edited ../local/share/gdal/pcs.csv and changed "ETRS89 / LAEA Europe" 
into "ETRS 1989 / LAEA", ..et voilà: the error is gone. "ETRS 1989 / 
LAEA" is ESRI'ified into "ETRS_1989_LAEA", which is exactly the expected 
string for this SRS.

I am not sure if it is worth filing a ticket for this specific case, as 
it is part of a wider ESRI-WKT-versus-GDAL/OGR-WKT dilemma, which in 
return is known already, see e.g. Uwe's post from some days ago.

Hermann


On 15/09/2011 13:28, Smith, Michael ERDC-CRREL-NH wrote:
> I've run into this. It seems to be that there are only certain projections
> that are supported in FileGDB. I tried to convert data in to the Spherical
> Mercator projection and none of the projections I used (either epsg, esri
> epsg, or esri wkt) would work. I believe that Even R. determined that it
> was a limitation of the FileGDB API currently.
>
> A ticket should be filed so that this can be documented.
>
> Mike
>