Very dangerous bug that could create a situation when .blend will go out of sync with IFC leading to unpredictable issues.
Example issue:
1. Create a drawing and activate it.
2. Create a grid.
3. Activate the drawing again.
4. Undoing just drawing activation here would work fine - annotations would be still floating in IFC without objects, but that would be fixed on the next drawing activation
To make it more dangerous undo grid creation. Then you'll get the traceback similar to the one below. The most dangerous part of it, that user won't see an error, it will just pop up in the system console silently.
```
Traceback (most recent call last):
File "\bonsai\bim\handler.py", line 244, in undo_post
IfcStore.undo(until_key=props.last_transaction)
File "\bonsai\bim\ifc.py", line 533, in undo
transaction["rollback"](transaction["data"])
File "\bonsai\bim\ifc.py", line 449, in <lambda>
operator, rollback=lambda d: tool.Ifc.get().undo(), commit=lambda d: tool.Ifc.get().redo()
^^^^^^^^^^^^^^^^^^^^^
File "\ifcopenshell\file.py", line 345, in undo
transaction.rollback()
File "\ifcopenshell\file.py", line 161, in rollback
element = self.file.by_id(operation["id"])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "\ifcopenshell\file.py", line 498, in by_id
return self[id]
~~~~^^^^
File "\ifcopenshell\file.py", line 485, in __getitem__
return entity_instance(self.wrapped_data.by_id(key), self)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "\ifcopenshell\ifcopenshell_wrapper.py", line 8884, in by_id
return _ifcopenshell_wrapper.file_by_id(self, id)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: Instance 1565 not found
```
Noticed fixing #6670 that though example project in this issue had camera named "PERSPECTIVE" but it was saved to IFC as orthogonal and there is no straightforward way to change it, besides changing Blender camera props and updating representation.
Example - https://imgur.com/a/GQ9A3EK
It's kind of exposed to UI now, but there's still bunch of issues with it to address.
Similar issue to 82f25f5
Basically matrix is a little bit different every time you activate drawing even if user made no changes to camera position. My guess is basically all float values stored by Blender somewhere deep as float32 but when you access matrix world, Blender converts them to Python floats (which are float64) and some garbage values introduced along the way creating this noise.
To make it possible to run the tests with PYTEST_DISABLE_PLUGIN_AUTOLOAD set, which is very useful if you don't want to start Blender anytime you run any tests on your system.
Traceback errors give users an impression that some broke and maybe project now is in some invalid state and they should ctrl-z immediately, error reports are much more friendly.
Before - https://i.imgur.com/N4fUU1M.png
After - https://i.imgur.com/6On1Be0.png
Example error traceback:
```
Traceback (most recent call last):
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\bim\module\geometry\operator.py", line 1044, in execute
return OverrideDuplicateMove.execute_duplicate_operator(self, context, linked=False)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\bim\module\geometry\operator.py", line 1055, in execute_duplicate_operator
IfcStore.execute_ifc_operator(operator, context)
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\bim\ifc.py", line 458, in execute_ifc_operator
result = getattr(operator, "_execute")(context)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\bim\module\geometry\operator.py", line 1047, in _execute
return OverrideDuplicateMove.execute_ifc_duplicate_operator(self, context)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\bim\module\geometry\operator.py", line 1085, in execute_ifc_duplicate_operator
old_to_new, new_active_obj = tool.Geometry.duplicate_ifc_objects(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\tool\geometry.py", line 2131, in duplicate_ifc_objects
tool.Root.recreate_aggregate(old_to_new)
File "C:\Users\Andrej\AppData\Roaming\Blender Foundation\Blender\4.4\extensions\.local\lib\python3.11\site-packages\bonsai\tool\root.py", line 397, in recreate_aggregate
tool.Blender.select_and_activate_single_object(bpy.context, tool.Ifc.get_object(new_aggregate[0]))
~~~~~~~~~~~~~^^^
TypeError: 'NoneType' object is not subscriptable
```
So non MESH or CURVE objects are not listed in the Colour by property.
Now all unrelated empty objects are excluded from the coloring section as they were just taking space, see - https://imgur.com/sPA8I5V
Essentialy outliner.delete should do the same as object.delete, but in our case the main implementation was in object.delete while outliner.delete was missing very important parts - e.g. handling arrays.
So if user could have remove part of an array without errors, leaving pset data invalid. See #6590, #6547.
Found this by trying to debug issues above.
Now outliner.delete will call object.delete explicitly, so the two stay connected.
I also had an idea that may be it somehow can produce #6632/#6619, but couldn't find a way to reproduce it. So it's still uncertain.
Andrej730
