could see a potentially well received use of Topologic in BlenderBIM: automatically create IfcSpaces from IfcWall/IfcSlab/IfcRoofs surfaces that surround a 'room' empty.
The problem is the vast quality variation of walls, slabs and roofs in real world data, some kind of processing is needed first to reduce them to simple planes - then topologic can create spaces. If your IFC data already has planes, or something that can be easily converted to planes (i.e. slabs are extruded profiles and walls have axis and height), then you could 'grow' these and hopefully they would intersect enough to define spaces.
I agree with @Topologic and @brunopostle take on this issue. what should change is the process of modelling architecture. Instead of having to work with a complex model (with all complex elements) we should work with simple models instead that relate to the base spatial composition and complexify from there, parametrically with Topologic.
The issue is that in theory this is true, but in practice this isn't easy to achieve. I think topologic will have a place until a certain point and then it will be classical modeling. In between I think Homemaker can be very useful... There's a long way ahead in terms of exploring a practical workflow.
That is exactly how we see Topologic as well, but it is not that it gets unused in later stages, it gets reused (as homemaker does currently) until the design is decided on and the next phase begins. The conceptual model will always be there if needed to step back. But we should also embed the final conceptual model in the IFC model so we can extract it later if needed.
But that, at the moment, seems out of this world to me. At the final stage of my projects a lot of direct modelling and 2D detailing is being carried out. At that point any relationship with a "mass model" that is being used as basis for topologic or homemaker is very far from being useful. In fact, as long as any direct modeling is involved, how can Topologic or homemaker be used without destroying it?
EDIT: Maybe you're thinking Sverchok can bridge both direct and parametric modelling actions?
It is hard and currently no tool supports it, but one can imagine that the direct modelling is topologically linked to the underlying conceptual model so it moves and evolves with it based on shape grammar rules. Think of it as shaders or detailing in an animation applied to your model. They always update correctly when you update the low level objects.
This is going to be hard, but it needs to work in some way (it is similar to the freecad 'topological naming problem'):
1. Generate a low-level IFC model from high level schematic geometry
2. Edit the IFC to add detail to some area of the building
3. Move or change some other part of the building at the schematic level
4. Regenerate the IFC model to respect the high level changes without clobbering the edited detail
If that could happen that would be killer, but I guess that would require everything to be setup as in FreeCAD sketches. Each geometry manipulation should be framed within a rule the stems from the original schematic level geometry.
Like layers of constraints that could depend on LOD to generate complex geometry.
A long time ago I played a little with Grasshopper-Archicad connection and it seemed very promising. From what I understand it connects basic shapes coming from grasshopper and converts them to the Archicad tools (wall, slabs, etc.). Each Archicad tool has input restrictions, e.g. walls have to be line and slabs have to be horizontal planes or polygons. Once they are in Archicad, they can be treated as BIM objects, but you can keep the connection with the simple Grasshopper model. It would be amazing to have this kind of workflow with Topologic-Sverchok and BlenderBIM.