Upon writing a piece of code for Maya, it might be interesting to know how it performs under different conditions, such as within scenes that are large or small, that define a deep DAG hiearchy or a flat one, that use many node types or only a few, and so on.

Following sets of user-provided commands, Revl can pseudo-randomly generate Maya scenes with different properties against which the behaviour of a piece of code can be observed.

The random nature of the process can also help revealing potential bugs by exposing edge cases that were not thought of, thus making it also a good tool for unit testing. See Wikipedia’s Fuzzing page.

Note that Revl does not provide any sort of profiling tool to measure performances. The built-in timeit module as well as other open-source packages can be used for this purpose.


  • generates scenes by running commands a given total number of times.
  • fine control over the probability distribution for each command.
  • scene generations are reproducible using a fixed seed.
  • extensible with custom commands.
  • allows for fuzz testing.
  • fast (using Maya’s API, not the command layer).


>>> import revl
>>> commands = [
...     (4.0, revl.createTransform,),
...     (1.0, revl.createPrimitive, (), {'parent': True}),
... ]
>>> count = 100
>>> revl.run(commands, count, seed=1.23)

In this example, Revl invokes a total of 100 evaluations inequally shared between the two distinct commands provided, leading to create approximatively 80% of transforms, and 20% of primitives (plus their associated transforms). Also, the primitive type is picked randomly, and each primitive’s transform is randomly parented under another transform from the scene, possibly creating a scene with a deep DAG hierarchy.

See also

The Tutorial section for more detailed examples and explanations on how to use Revl.