Python 3 - Unpacking Generalizations

Python 3.5 added more support for Unpacking Generalizations in PEP 448. According to the PEP, it added extended usages of the * iterable unpacking operator and ** dictionary unpacking operators to allow unpacking in more positions, an arbitrary number of times, and in additional circumstances. What this means is that we can now make calls to functions with an arbitrary number of unpackings. Let's take a look at a dict() example:

>>> my_dict = {'1':'one', '2':'two'}
>>> dict(**my_dict, w=6)
{'1': 'one', '2': 'two', 'w': 6}
>>> dict(**my_dict, w='three', **{'4':'four'})
{'1': 'one', '2': 'two', 'w': 'three', '4': 'four'}

Interestingly, if the keys are something other then strings, the unpacking doesn't work:

>>> my_dict = {1:'one', 2:'two'}
>>> dict(**my_dict)
Traceback (most recent call last):
  File "", line 1, in 
    dict(**my_dict)
TypeError: keyword arguments must be strings

Update: One of my readers was quick to point out that the reason this doesn't work is because I was trying to unpack into a function call (i.e. dict()). If I had done the unpacking using just dict syntax, the integer keys would have worked fine. Here's what I'm talking about:

>>> {**{1: 'one', 2:'two'}, 3:'three'}
{1: 'one', 2: 'two', 3: 'three'}

One other interesting wrinkle to dict unpacking is that later values will always override earlier ones. There's a good example in the PEP that demonstrates this:

>>> {**{'x': 2}, 'x': 1}
{'x': 1}

I thought that was pretty neat. You can do the same sort of thing with ChainMap from the collections module, but this is quite a bit simpler.

However this new unpacking also works for tuples and lists. Let's try combining some items of different types into one list:

>>> my_tuple = (11, 12, 45)
>>> my_list = ['something', 'or', 'other']
>>> my_range = range(5)
>>> combo = [*my_tuple, *my_list, *my_range]
>>> combo
[11, 12, 45, 'something', 'or', 'other', 0, 1, 2, 3, 4]

Before this unpacking change, you would have had do something like this:

>>> combo = list(my_tuple) + list(my_list) + list(my_range)
[11, 12, 45, 'something', 'or', 'other', 0, 1, 2, 3, 4]

I think the new syntax is actually quite handy for these kinds of circumstances. I've actually run into this a time or two in Python 2 where this new enhancement would have been quite useful.

Wrapping Up

There are lots of other examples in PEP 448 that are quite interesting to read about and try in Python's interpreter. I highly recommend checking it out and giving this feature a try. I am hoping to start using some of these features in my new code whenever we finally move to Python 3.