IfcOpenShell invalid cast exception about xbimgeometry

[dontbmh]

S2.zip
I've attached my test ifc file here.

[dontbmh]

is it appropriate to replace IIfcGeometricRepresentationItem with IIfcRepresentationItem all the way down to the XbimGeometryCreator?

[CBenghi]

Hello there,
I've made a couple of changes to geometry to handle the error more gracefully and load most of the model. There are still problems with the file and I've added a test case to geometry to keep track of the problem.
I'll involve @SteveLockley here to make sure that we don't introduce regressions and come back to you.
Best,
Claudio

[CBenghi]

Hello,
I'm back with bad news I'm afraid.
The reason of the problem seems to sit with an invalid structure of the file.
I've extracted the minimal case to reproduce the error in this file:

https://github.com/xBimTeam/XbimGeometry/blob/22f8c436f927c38dc76dfb75ec1814c9f92b599e/Xbim.Geometry.Engine.Tests/SolidTestFiles/CatiaIfcOpenShellTest.ifc

But then it emerged that the problem is in the file.

IfcShapeRepresentations have specific clauses in the ifc schema that constrain its usage and the one in question (Entity #1494003) fails two requirements:

WR22 : No topological representation item shall be directly used for shape representations, with the exception of IfcVertexPoint, IfcEdgeCurve, IfcFaceSurface. And,
WR24 : Checks the proper use of Items according to the RepresentationType.

This is actually a problem that you should notify the good folks of Catia; They should export an IFCSHELLBASEDSURFACEMODEL between the IFCSHAPEREPRESENTATION and the IfcOpenShell.

The fix on the model is simple and it's as follows:
#1494003 needs to be changed and #1775737 to be added from scratch.

#1494003=IFCSHAPEREPRESENTATION(#14,'BODY','SurfaceModel',(#1775737));
#1775737=IFCSHELLBASEDSURFACEMODEL((#1503330));

you can download the fixed file from https://github.com/xBimTeam/XbimGeometry/blob/b97b017cfb3c8e32b0a9bd191b2a0763cd72b348/Xbim.Geometry.Engine.Tests/SolidTestFiles/CatiaIfcOpenShellTest.ifc

I hope this helps.
Claudio

[dontbmh]

Thanks for your detail reply.

According to buildingsmart's specs, I believe you are right.
Any topological representation directly used for shape representation seems violating the ifc definition.
I would make a temporarily "hack" on my fork in order to complete my export work and contact catia later.

Cheers.
Dirk