Hello hello - quick testing question I have a set of transforms which I'm applying using

daft

. These are packaged to support

np.generic

types which I run using

pytest.parametrize()

. For the

ray.data

version of these transforms, I've written a simple

session

scope fixture which initializes a

ray

cluster with some CPU cores and uninitializes it at the end of the

pytest

run. This is used by a

Test

class, making it easy to trace in code. For the

daft

test, I have an analogous

Test

class. I'd like to re-use this class and simply re-decorate it with another

session

or

module

scope fixture which sets and unsets the

daft.context.runner

. How do? Has the context setting bug been resolved? Is there another way to handle this which you use for tests internally that you can point me to? I would of course like to extend this to the new python streaming execution API as well, hopefully with another

session

or

module

scope fixture. But the

ray

runner is the highest priority right now.

Is there another way to handle this which you use for tests internally that you can point me to?

One thing we've tried doing within tests is to set the runner to

None

, which might accomplish what you want to do for the time being https://github.com/Eventual-Inc/Daft/blob/274f300fb86413dfea11d38fba1fe93e7db2f02b/tests/actor_pool/test_actor_cuda_devices.py#L131 We'll probably try to have a better solution in the near future