On large projects predict dense mesh stage can take 20s+ and reusing attribute value can save up to half of this time.
Using indices also helps but it's not that significant and sometimes it's the same time as using attribute names.
removed part of test_append_two_type_products_sharing_the_same_material_indirectly_via_a_material_set for ifc2x3 compatibility and removed part is already tested in test_append_two_type_products_sharing_the_same_material_with_properties
After ab5ea4c85 it was always throwing wrong singature errors like below even if TypeError was caused by some internal issues inside API - it was adding couple extra steps to traceback making errors more noisy.
TypeError: Incorrect function arguments provided for library.edit_library
attribute 'VersionDate' for entity 'IFC2X3.IfcLibraryInformation' is expecting value of type 'ENTITY INSTANCE', got 'str'.. You specified args (<ifcopenshell.file.file object at 0x0000027EB8E6BCD0>,) and settings {'library': #1=IfcLibraryInformation('Name','Version',$,$,$), 'attributes': {'Name': 'Name', 'Version': 'Version', 'VersionDate': 'VersionDate', 'Location': 'Location', 'Description': 'Description'}}
E
Correct signature is (file: ifcopenshell.file.file, library: ifcopenshell.entity_instance.entity_instance, attributes: dict[str, typing.Any]) -> None
See help(ifcopenshell.api.library.edit_library) for documentation.
When some test was creating an element and then removing it, it would also remove user and application as they wasn't used anywhere else.
`ifcopenshell.util.element.remove_deep2(file, history)` we use in every api for element deletion can possibly remove user and application which can be unsafe if `get_user` is returning some specific entity that then will become invalid.
Also fixed tests breaking due ifcownerhistory and user/application appearing in ifc2x3.
it was creating new applications and users every time `ifcopenshell.api.owner.settings.get_user` or `ifcopenshell.api.owner.settings.get_application` was called
We were incorrectly multiplying a shape matrix in SI units with a vertex in project units. We also didn't do a final check whether or not that final resultant coordinate was far away or not (for example, origin and vertex can cancel each other out)