It was producing a bug with text rotation in elevation view
Added fix to make matrix normalization work for the cases when matrix is rotated and I've just found that mathutils.Matrix.normalize(d) is very conveniently ignoring 4th column (translation) (same for mathutils.Vector)
Also simplified create_text_tag a bit - now it adds fill_bg by just by copying text tags and adding fill_bg filter to them instead of going through the entire logic again.
- Important for Ifc External References and other data.
- Set predicate_uri of predicates of the Bnode to None, which is used to remove the ability to delete the relationship in the UI.
- Also added try/excepts to difficult edge cases such as the object_uri not having a "#" character in it (such as "<http://qudt.org/vocab/unit/>") and the object_uri not being a brick class, which breaks "view_brick_item".
- Removing a Bnode type from the entity's relation removes all relations of the Bnode also
Change from never show the operator if the relation is "type" to not showing it only if the "type" is the current class being viewed.
This may come in handy if other list views other than hierarchy are implemented.
It also implies why we don't want to show "view brick item" here: because we already viewing it.
This means that if you created a curve body, you couldn't switch to it, or if you deleted everything, you ended up with an empty object. If you then tried to switch to that, it would "sync" the empty object and leave you with an empty cartesian point list. E.g:
1. Create cube and assign IFC element.
2. Remove all faces only (leaving edges and verts). Save changes.
3. Switch to representation (bug!)
4. Delete everything (no verts, edges, or faces). Attempt to save.
5. "Revert" switch will fail, leading to empty mesh.
6. Manually try to switch. Empty mesh will sync, creating an empty point list.
Turned out there are ODS files (and turned out we're producing them ourselves from ifccsv) where column styles are not defined and therefore no width/style assigned to each column which was resulting in errors on building schedule.
Now, instead of trying to be clever and regenerating filling locations and opening types, we simply copy the existing scenario. This also means it'll be more stable for weird fillings coming from proprietary apps.
Gorgious56