Le 12/12/2021 à 19:32, Marco Neumann a écrit :
Just when I say this I get automated test failures for the British
National Grid EPSG:27700 transformations with 1.1 that run fine in 1.0
Are you aware of any deliberate changes to the grid databases here?
I'm not aware of significant changes for this code in the EPSG database.
On the implementation side, this is a "Transverse Mercator" projection,
whose formulas have been modified in SIS 1.1 for better stability. I
would not expect a difference so large however, unless the coordinates
are far from the CRS domain of validity. Can you confirm that the
geographic coordinates in the test are between 9°W and 2°E? (a few
degrees outside should not be a problem).
The new Transverse Mercator projection has been tested using Karney
(2009) Test data for the transverse Mercator projection
(https://zenodo.org/record/32470). Results compared to SIS 1.0 may
differ significantly for points that far outside the domain of validity
(e.g. at a distance of more than 40° of longitude), but should not
differ significantly for points inside the domain of validity.
If points are inside the domain of validity, can you execute a code like
below with SIS 1.0 and SIS 1.1 and check if the strings differ? Source
CRS and target CRS would be the CRS used by the
SRSInfo.buildDomainEnvelope(…) method implementation.
System.out.println(CRS.findOperation(sourceCRS, targetCRS, null));
Martin
