generative modeling for Rhino
I was thinking about this today after see the tri-panel tutorial that "Grasshopper for Rhino" post in Google+ (I've created a discussion about this there).
It may be some kind of relation between number of nodes, time per global definition calculation, number of conections and geometry or entities generated that could help you to optimize the code. When you write C# or vbNET code, is complicate to calculate that index, but here connections, variables and entities are easily measurable.
I'm not talking about something messy or complicated only and index that relation some properties of the code and tell you: "Hey man!, you are putting lot of nodes to calculate only 3 simple spheres" or something like that.
This number could be better than the time that Gh shows you in the bottom bar or if you activate "profiler".
there's no way of knowing what data a definition actually creates. If there's a VB or C# or Python or 3rd party component involved all kinds of data may be generated and send off to the hard-drive or the web or God knows where.
Also, "3 simple spheres" implies that it's possible to determine what "simple" is. Perhaps you really do need 250 components and a runtime of 20 seconds in order to find that single point coordinate that is vital to your design.
An approach which might work but I'm not sure warrants the investment would be to define specific groups of components. Something like "If A is connected to B, then A and B are connected to C and finally C is hooked up to D and E, then you may be able to get the same result using only component X and Y". Not only is this approach free from unknowns it also tries to help the user out. I'd much rather be told "why don't you try using a Key/Value search here?" than "You're a f*cking idiot mate."