CODEWITHSUNDEEP
python
Ian Fiddes
Ian Fiddes

Reputation: 3011

A namespace with additional significance parameters

I am trying to define a namespace-like object that can have significance parameters. This is useful so that I can implement a hashing function for the object that will only hash members tagged as significant. I would like the API to be as close to the namespace style as possible, but as far as I can think there is no easy way to do something like

n = MyNamespace()
n.a = (1, True)

And have 1 be stored in __dict__ while True would be stored in some other dictionary that maintains the knowledge that 1 is significant.

I can think of a way of doing it with a setter, like so, but that is a bit clunky and non-pythonic and requires a ugly paired tuple way of direct instantiation.

class SignificantNamespace(object):
    def __init__(self, *args):
        self.significant = {}
        for name, val, significance in args:
            setattr(self, name, val)
            self.significant[name] = significance

    def set(self, name, val, significant=True):
        setattr(self, name, val)
        self.significant[name] = significant

What do you guys think? Is there a better/cleaner way to do this?

Upvotes: 1

Views: 97

Answers (1)

Jack
Jack

Reputation: 21183

If I understand you correctly you could do something like this:

class MyNamespace(object):

    def __hash__(self):
        """Only hash dict key pairs that are significant"""
        d_keys = sorted(self.__dict__.keys())
        to_hash = []
        for k in d_keys:
            v = self.__dict__[k]
            if v[1]:
                to_hash.append((k, v[0]))
        return hash(frozenset(to_hash))

    def get_signifcant_dict(self):
      return {k: v[0] for k, v in self.__dict__.items() if v[1]}

Usage example:

>>> my = MyNamespace()
>>> my.a = (1, True)
>>> my.__hash__()
-2550060783245333914
>>> my.b = (2, False)
>>> my.__hash__()
-2550060783245333914 # hash didn't change since `b` is not signifcant
>>> my.get_signifcant_dict()
{'a': 1}

Upvotes: 1

Related Questions