Apparent error in docstring for Operator.__init__

[bwohlberg]

There appears to be an error in the docstring for Operator.__init__ :

scico/scico/operator/_operator.py

Lines 88 to 90 in 88026d2

	eval_fn: Function used in evaluating this :class:`.Operator`.
	Defaults to ``None``. If ``None``, then `self.__call__`
	must be defined in any derived classes.

If I'm not mistaken, it should state "... then self._eval must be defined ...".

@Michael-T-McCann: do you agree?

[Michael-T-McCann]

@Michael-T-McCann: do you agree?

Agreed.

Even with that fixed, I find it mightily confusing. Perhaps "Required unless __init__ is being called from a derived class with an _eval method." Whether this is a good design is a separate question.

Another positive change we could make: H = Operator((10,)) currently throws the error AttributeError: 'Operator' object has no attribute '_eval'. Instead, __init__ could check if eval_fn is not passed and self._eval is not defined and throw an error.

[bwohlberg]

Even with that fixed, I find it mightily confusing. Perhaps "Required unless __init__ is being called from a derived class with an _eval method."

That's quite a bit better.

Whether this is a good design is a separate question.

Agreed, but I'm inclined to leave that issue alone for now.

Another positive change we could make: H = Operator((10,)) currently throws the error AttributeError: 'Operator' object has no attribute '_eval'. Instead, __init__ could check if eval_fn is not passed and self._eval is not defined and throw an error.

I was thinking of that too. Will do.