Fix bug in aiida.engine.utils.instantiate_process

[sphuber]

If the process argument receives an unsupported type that is not a class, the final check issubclass(process, Process) would raise:

TypeError: issubclass() arg 1 must be a class

which is not very useful to the user. This is fixed by first checking that the argument is a class to begin with. At least now when an invalid type is provided, the exception contains the invalid type:

invalid process <class \'bool\'>, needs to be Process or ProcessBuilder'

[unkcpz]

@sphuber Thanks! The change looks good.
But I am curious where you encounter this, I can not think of a case to pass the bool type as process to this function. The input argument type hint also does not reveal this as a case. It makes more sense if the type check is at the beginning of the function to make sure the process pass in conforms with the type hint.

[sphuber]

@sphuber Thanks! The change looks good. But I am curious where you encounter this, I can not think of a case to pass the bool type as process to this function. The input argument type hint also does not reveal this as a case. It makes more sense if the type check is at the beginning of the function to make sure the process pass in conforms with the type hint.

The boolean is just for the test case. I don't remember exactly, but I think in the case where this was encountered a class instance of something besides a Process or ProcessBuilder was passed, or even a method. For example, I think the following would have hit the exception:

builder = load_code('pw@localhost').get_builder
submit(builder)

Note how the () call at the end of get_builder was forgotten, and so now the bound method is passed and issubclass raises, since it is not a class.

It was something like that, but seemed like an honest mistake that could happen and the current resulting error message was not helpful whatsoever.

[unkcpz]

That's fair, thanks for the explanation!