now we're going to store all texture paths in .ifc as posix to keep them working if .ifc moved from one platform to another
previously it was saving all paths as absolute
Errors:
Traceback (most recent call last):
File "\Blender\3.6\scripts\addons\blenderbim\bim\module\ifcgit\operator.py", line 20, in poll
if not os.path.isfile(path_ifc):
File "\Blender\3.6\python\lib\genericpath.py", line 30, in isfile
st = os.stat(path)
TypeError: stat: path should be string, bytes, os.PathLike or integer, not NoneType
Traceback (most recent call last):
File "\Blender\3.6\scripts\addons\blenderbim\bim\module\ifcgit\operator.py", line 48, in poll
if not os.path.isfile(path_ifc):
File "\Blender\3.6\python\lib\genericpath.py", line 30, in isfile
st = os.stat(path)
TypeError: stat: path should be string, bytes, os.PathLike or integer, not NoneType
Crash happened because we were addressing `.id()` of already deleted `cost_item`.
Surprisingly for me on windows addressing `cost_item` was causing crash too but `cost_item.id()` worked fine.
Now we do not rely on small offset to keep using inner curves to create linings, if some part of the lining is not present (which happens on mullions and transoms) we'll create U or L shape extrusion instead of rectangle with inner curve.
Long story short - this should keep ifc for windows valid.
From this - https://i.imgur.com/hQuCcR3.png
To this - https://i.imgur.com/qZ2Lyfo.png
It's a bit hacky since we still use really small offsets to keep "The outer curve shall enclose all inner curves." rule for IfcArbitraryProfileDefWithVoids valid.
While these actions are short in the backend--taking about 0.01 seconds to run--the Blender UI lags a lot leading to 2-3 second pauses because of the "refresh_brick_viewer" function being slow. This should be investigated.
To test, go into add-on preferences and enable "Should Stream". Then load an IFC-SPF in advanced mode, and disable "Load Geometry" before loading (the geometry iterator doesn't work with streaming).
- Turns out, you need the keyword "graph=" in graph_at() to actually select a graph of that name from the collection, otherwise it just returns the entire collection, so I changed that, which correctly isolates the project from the collection for serialization now.
- With this same change, I opted turn BrickStore.VersionedGraphCollection simply into BrickStore.graph and create a new BrickStore.get_project() to return the isolated graph. - This meant I should remove the reload_graph() function because I was actually just loading the entire collection into it still, and its functionality breaks when it isn't the entire collection
Andrej730
