DagNode.addChild crashes in certain circumstances about cmdx HOT 6 CLOSED

Thanks for reporting this. I'm glad you consider the performance impact of your suggestions. :)

What I would suggest is that you profile the change, before and after, identify what the cost is and if it's "too much" then we could find a way to work with that; e.g. either make it optional, or add a stern warning in the documentation.

To profile things like this, have a look at plot.py for inspiration.

from cmdx.

looks like it's safe. i ran several loops of 20000 parent commands and the difference between with or without is notable but still very small (about 50ms vs 200ms for each packet)

from cmdx.

If you can fit your test in plot.py, ideally using a similar/same convention, then that would be swell. Then you could make a note of how long it takes along that line, for anyone looking at this in the future wondering how costly that is. You'll find similar notes across the code, in a μs unit, for microseconds.

from cmdx.

hmm ok i see! here are more serious results (: is that what you had in mind?

New()

root1 = cmdx.createNode("transform")
root2 = cmdx.createNode("transform")
child = cmdx.createNode("transform", parent=root2)

melparent = 'parent -r "transform3" "transform1"; parent -r "transform3" "transform2";'
Test("mel", "parent", lambda: mel.eval(melparent), number=1, repeat=1000)
Test("cmds", "parent", lambda: [cmds.parent("transform3", parent, r=1) for parent in ("transform1", "transform2")], number=1, repeat=1000)
Test("cmdx", "parent", lambda: [cmdx.parent(child, parent) for parent in (root1, root2)], number=1, repeat=1000)
Test("cmdx safe", "parent", lambda: [cmdx.parent(child, parent, safe=True) for parent in (root1, root2)], number=1, repeat=1000)
root1, root2, child = pm.ls(map(str, (root1, root2, child)))
Test("PyMEL", "parent", lambda: [pm.parent(child, parent, r=1) for parent in (root1, root2)], number=1, repeat=1000)

parent mel: 115.5 ms (109.30 µs/call)
parent cmds: 119.1 ms (114.50 µs/call)
parent cmdx: 11.1 ms (10.60 µs/call)
parent cmdx safe: 19.3 ms (18.30 µs/call)
parent PyMEL: 601.7 ms (583.70 µs/call)

it looks like adding this security costs about twice the original command. still way cheaper than mel/cmds though

from cmdx.

Oh you are a star. :D Thanks for putting a smile on my face.

Yes, that is exactly what I had in mind. That safe= option is great, we could leave that in, and default it to True. That way, anyone wanting to squeeze performance could toggle that flag, and they'd be good to go.

from cmdx.

Implemented in #24. Thanks @wougzy!

from cmdx.