[gdal-dev] Coordinate transformation of geometries between CRSs with different axis ordering (C API)

[Florian Katerndahl]

Hi Even,

thanks for your reply.

I don't pass the coordinates directly via `OCTTransform` but use 
`OGR_GeomTransformer_Transform` to transform polygon geometries. Based 
on the fact than when fetching the points that make up the input 
polygon, the x and y coordinates match the axis of the input CRS, my 
assumption was that I don't need to change the ordering for the input 
CRS. Looking at the points in the output geometry, the coordinates are 
flipped when not setting OAMS_TRADITIONAL_GIS_ORDER. This was the reason 
why I assumed I need to set the axis ordering depending on the CRSs I 
translate between.

But if I understand your reply correctly, I should also not need to 
touch the axis ordering for the output CRS as GDAL takes care of mapping 
the data to CRS axis for the output geometry. I suppose my understanding 
is not sufficient enough (yet) to understand the underlying issue then 
and I will need to do some further research on my side.


Kind regards,

Florian


On 4/1/26 16:26, Even Rouault wrote:
>
> Le 01/04/2026 à 16:04, Florian Katerndahl via gdal-dev a écrit :
>> Dear all,
>>
>> I'm currently working on a small program which allows users to 
>> specify vector inputs in different coordinates systems and that 
>> potentially "normalizes" them to EPSG:4326 internally to process them 
>> further. Maybe I'm too inexperienced and there's an easy solution but 
>> I got hung up on the coordinate transformation when the input CRS has 
>> a different axis ordering compared to the output CRS using 
>> GDAL's/OGR's C API.
>>
>> For example, transforming between EPSG:5556 and EPSG:4326, GDAL 
>> outputs a warning (forwarded from proj?) that the output's latitude 
>> is invalid. Setting the axis ordering of the target CRS to 
>> `OAMS_TRADITIONAL_GIS_ORDER` fixes this and results in correctly 
>> located polygons. My question is now: Is it sufficient/correct to 
>> compare the axis ordering of the input and output coordinate systems 
>> and set above mentioned flag (or a custom mapping when neither 
>> "x"/"y" are the first two axis) in case they differ?
>
> The axis ordering of the input CRS vs the one of the output CRS is 
> complelely unrelated w.r.t your decision about setting or not 
> OAMS_TRADITIONAL_GIS_ORDER. Setting it or not for the input totally 
> depends on your decision how you want to specify the coordinates.
>
>