Is there any pythonic way to combine two dicts (making a list for common values)?

Question

For example, I have two dicts.

A = {'a':1, 'b':10, 'c':2}
B = {'b':3, 'c':4, 'd':10}

I want a result like this:

{'a':1, 'b': [10, 3], 'c':[2, 4], 'd':10}

If a key appears in both the dicts, I want to list of both the values.

Answer 1

I'd make all values lists:

{k: filter(None, [A.get(k), B.get(k)]) for k in A.viewkeys() | B}

using dictionary view objects .

Demo:

>>> A = {'a':1, 'b':10, 'c':2}
>>> B = {'b':3, 'c':4, 'd':10}
>>> {k: filter(None, [A.get(k), B.get(k)]) for k in A.viewkeys() | B}
{'a': [1], 'c': [2, 4], 'b': [10, 3], 'd': [10]}

This at least keeps your value types consistent.

To produce your output, you need to use the set intersection and symmetric differences between the two dictionaries:

dict({k: [A[k], B[k]] for k in A.viewkeys() & B},
     **{k: A.get(k, B.get(k)) for k in A.viewkeys() ^ B})

Demo:

>>> dict({k: [A[k], B[k]] for k in A.viewkeys() & B},
...      **{k: A.get(k, B.get(k)) for k in A.viewkeys() ^ B})
{'a': 1, 'c': [2, 4], 'b': [10, 3], 'd': 10}

In Python 3, dict.keys() is a dictionary view, so you can just replace all .viewkeys() calls with .keys() to get the same functionality there.

Answer 2

I would second the notion of Martijn Pieters that you problably want to have the same type for all the values in your result dict.

To give a second option:

you could also use the defaultdict to achieve your result quite intuitively.

a defaultdict is like a dict, but it has a default constructor that is called if the key doesn't exist yet.

so you would go:

from collections import defaultdict

A = {'a':1, 'b':10, 'c':2}
B = {'b':3, 'c':4, 'd':10}

result = defaultdict(list)

for d in [A, B]:
    for k, v in d.items():
        result[k].append(v)

then in a later stage you still easily add more values to your result.

you can also switch to

defaultdict(set)

if you don't want duplicate values