__contains__ pollutes __dict__ with empty Namespace values

[christian-storm]

from pytool.lang import Namespace
f = Namespace({'a': {'a1': [1]}, 'b': {'b1': [2]}, 'c': {'d': [1, 2, 3]}})
print(bool('e' in f))  # False but creates f.e = Namespace()
print(bool('e' in f.__dict__))  # True

Doesn't it make sense that 'e' would only be added in assignment and not in attribute lookup/contains?

The reason I found this was because I added a method knitems

    for name, value in self.__dict__.items():
            yield name, value

to be able to iterate one level at a time. I've had to add

        if isinstance(value, Namespace) and bool(value):

in order to make it work.

[shakefu]

Good find. I'll see if I can get a patch fix published this week unless you want to PR earlier.

[christian-storm]

I'm a bit flat out right now. I may be able to get to it in a week or so.

[christian-storm]

This seems to have fixed things for me but wanted your opinion on whether this is the right fix or not.
Basically, if the contains is going to return False just delete the empty vestigial namespace created by getattr.

What do you think?

    def __contains__(self, name):
        names = name.split('.')

        obj = self
        for name in names:
            obj = getattr(obj, name)

        if isinstance(obj, Namespace):
            if obj.__dict__:
                return True
            else:
                del self.__dict__[name]  # <--- I added this line
                return False
        else:
            return True

[shakefu]

@christian-storm This is fixed in master. Apologies for the long delay, GitHub notifications suck.