Question

So i have made my own dict-based named-tuple class:

class t(dict):
   def __getattr__(self, v):
      try:
         return self[v]
      except KeyError:
         raise AttributeError("Key " + str(v) + " does not exist.")
   def __init__(self, *args, **kwargs):
      for source in args:
         for i, j in source.items():
            self[i] = j
      for i, j in kwargs.items():
         self[i] = j

>>> thing = t(cow=10, moo='moooooooo')
>>> thing.cow
10
>>> thing.moo
'moooooooo'

The point of using a dict is so i can use the **splat operator, as well as so the whole thing can be json’d as a dict. The double for loop in init is so I can immediately re-construct it from the dict after deserialization.

The only thing I’m missing is the multi-assignment thing you can do with tuples, like

>>> thing = (1, 3, ('blargh', 'mooo'))
>>> a, b, (c, d) = thing
>>> a
1
>>> b
3
>>> c
'blargh'
>>> d
'mooo'

Is there any way I can get this sort of behavior with my own custom datatype? Is there any function that I can override or some class I can inherit from to get this behaviour?

Answer 1

Yes. Implement __iter__().

class unpackable_dict(dict):
    def __iter__(self):
        return (self[key] for key in sorted(self.keys()))

d = unpackable_dict(a=1, b=2)
a, b = d

The reason you normally can’t unpack values from a dict like you can a tuple is that dicts don’t have a defined order. I’ve used a generator expression to get the items out in the sorted order of their keys. You might consider using an OrderedDict instead, though.