It would be nice to have a Python to Node pipeline.
I've come across ideas that would be great under geometry nodes, or other functions that use nodes, but due to missing math functionality or some other missing features, make implementing in nodes impossible. In one recent instance that comes to mind is the lack of a SVG mesh import node. It doesn't exist and to make it exist currently would require a recompilation of Blender's code for myself and a long, drawn out process to get it implemented into a future release. It would be trivial to write in Python. Another one was a video I watched that used binary bit sum parity to generate a Koch fractal. Doing so in geometry nodes would require the ability to split a number into it's bits and then operate on those bits, as well as storing variables. Easy to do through Python.
This would open up Nodes across Blender to ideas that are currently not even on the developer's radar.
5 Comments
I am upvoting this, though I would probably suggest a language other than python. Python is not a very high performance language, especially with large data sets. And Geometry Nodes run over very large data sets (think points, vertices, faces, simulations, etc.)
I love Python as a language personally, but it is a general purpose language. But I agree that some sort of a text-based programming language is needed (similar to VEX in Houdini). VEX is pretty limited as a general purpose language, but it is well suited to the kinds of math you re proposing, and it is optimized to be very multi-threaded. I think it actually spawns one thread per point or primitive as I recall (I am basing this on stuff I have read in the past... I am no VEX expert).
So I am voting this up so that we can get some sort of an advanced expression language built into nodes.
Python is already deeply integrated into Blender, so it seems a natural choice. I'm not stuck on Python, but I think the idea of writing an entirely new language for Blender will be a hard no. Since Python is already a part of Blender, I doubt the developers will have any interest in something else. I could be wrong though. We'll see.
The Blender dev team is pretty amazing, so maybe they could make it work. Who knows. But the underlying need for some sort of an expression/programming need is real so I hope they see your suggestion.
I agree with that. We can already script with Python in Blender but geometry nodes requires a different language for simpler operations that essentially would allow users to type in custom "scripts" for their input properties instead of plugging a lot of math and read nodes together. I honestly really like building math operations with nodes since it's very intuitive to me, but having worked in Houdini with VEX, the comparison is huge. It takes a few seconds to type in a custom operation, but it takes a good minute in Blender to build the same thing with nodes.
Something comparable to VEX in geometry nodes would be game changer
This idea has potential to put Blender closer to Houdini, with much lower development cost in comparison to what actually needs to be done to achieve a fraction of Houdini's flexibility.