Nested Defaultdict (tree) In Python

14 July 2014

Warning If KeyError is useful for your debugging, defaultdict may not be something you want to use.

The collections.defaultdict is very useful for well tested code as it cleans up a lot of try-except blocks when dealing with dictionaries. Singly nested dictionary can be done by collections.defaultdict(collections.defaultdict). But what about multi-layer nested dictionary?

Essentially I want to do this:

data_dict = nested_defaultdict()
data_dict['A']['B']['C']['D'] = 1

without having to initialise data_dict['A'], data_dict['A']['B'] etc.

A recursive function would do. And this is pickle-able too.

import collections


def nested_defaultdict():
    return collections.defaultdict(nested_defaultdict)

Update (Dec 8, 2014): I just learned that this is called autovivification. Very interesting. The Python example there is basically the same as the one presented here.