Question

So I’m wondering if anyone can help me out with this issue I’m having.

Lets assume I have a dictionary:

d = {1: {2: 3}, 4: 5}

I want to create a dictionary of any contained dictionaries:

wanted_result = {2: 3}

what I am trying is this:

e = {inner_key: d[key][inner_key] for key in d.keys() for inner_key in d[key].keys() if isinstance(d[key], dict)}

However this is causing me to get an error saying that ints don’t have keys, which I know, but I thought my conditional would exclude say 4 from my example from being included in the comprehension.

Answer 1

My first idea was something like this:

d = {1: {2: 3}, 4: 5, 6: {7: 8}}
generator = (x.items() for x in d.itervalues() if type(x) == dict)
s = dict(sum( generator, list() )) 
# s = {2: 3, 7: 8}

But, in order to avoid letting sum() construct a large temporary list of all the items() you may use itertools.chain and iteritems() instead:

# chain "concatenates" iterables
from itertools import chain
d = {1: {2: 3}, 4: 5, 6: {7: 8}}
s = dict(chain( *(x.iteritems() for x in d.itervalues() if type(x) == dict) ))
# s = {2: 3, 7: 8}

You should be aware of edge cases. Especially if identical keys are in the part-dictionaries these will necessarily collapse somehow.

---

If you also want to support subclasses of dict to be collected from your d.values(), you may indeed want to use isinstance instead of type. See this question for details.