Python API fast enough for custom modeling only modifiers (deformers)?

[Deepak Seeni]

Hey Guys,

I've been learning Maya's python API and was keen on experimenting with simple non-real time custom deformers for modeling purposes. For non-real time deformers for modeling purposes, similar to the modifier stack in Max or Blender (Things like the edge split modifier) would the Python API be fast enough?

I wanted to start with an edge split deformer.

I'm assuming yes, since it's not for real time display, like on a rig or something. I'd use these tools mainly in modeling, similar to how I'd use wrap/shrink wrap/lattice.

What do you guys think?

[Michael Boon]

It really depends on the complexity of your meshes and your scenes. 

Speaking very generally

- Until you profile your code and understand what is causing any slowness, Python vs C++ will probably make less difference than the way your code is written.

- Sensibly-written Python might be (really roughly) 10x slower than C++, and still 10x faster than MEL.

- Bear in mind that Python plugins can't be run in parallel, while well-written C++ plugins will benefit from Maya parallel mode.

- Also bear in mind that the Python API is not quite complete. There are some things you can only do in C++.

I'd go for it though. If you're not exactly sure how you're going to implement your design, and not already comfortable in C++, you might benefit a lot from the faster turnaround and easier debugging of Python, and you can always port your code to C++ once it works.

[Deepak Seeni]

Thanks Michael!
Yeah It seems sensible then as a new tool developer to then prototype and design with Python and then branch out into C++ once I identify bottlenecks.