Get rid of "userdata"

[mgeier]

Currently, all methods to set client callbacks (jack.Client.set_*_callback()) have an optional userdata argument. This was meant to mimic the "void* arg" argument in the JACK API, but this argument isn't actually used in the implementation. Instead, the userdata argument is implemented by using a Python closure (callback_wrapper()), which avoids converting the value to a void* and back.

I think the argument userdata will probably never be used, because users of the library can just create a closure themselves and use it as callback function.

The advantage of dumping userdata would be that the function signatures of the callback functions would become much easier to understand.
Also, users wouldn't be forced to add the userdata argument to their callback functions even if they know they will never use it, which IMHO is quite annoying.

Are there any disadvantages?

I think the userdata argument only adds confusion and nuisance and doesn't help anyone.

Is there any conceivable use case which cannot be implemented without the userdata argument?

[amstan]

This is a good idea.

Is there any conceivable use case which cannot be implemented without the userdata argument?

I think userdata only exists in jack because in C you don't really have lambdas and such.