Learning Python with Advent of Code Walkthroughs

Dazbo's Advent of Code solutions, written in Python

The Python Journey - Defaultdict

Useful Links



Page Contents


I’ll assume you already know what a Python dictionary is. If not, review this first.

If you try to reference a key that doesn’t exist in a Python dictionary, you’ll be met with a KeyError. E.g.

my_dict = {}
my_dict["some_key"] = 200



Traceback (most recent call last):
  File "c:\Users\djl\localdev\Python\Advent-of-Code\src\snippets\snippet.py", line 5, in <module>
KeyError: 'some_other_key'

The defaultdict is cool because it never raises a KeyError; instead, it assigns a default value to any key that is referenced, but which has not been explicitly initialised.

“Initialised to what?”, I hear you say!

Well, that depends on the default factory that we have assigned to our defaultdict when we create it.

Default Factory

We can assign many types of default factory to a defaultdict, such as:


It’s really useful to have a default factory of int, when we want to increment dictionary values, but we also want to initialise a key (e.g. to 0) if we reference the key for the first time.

For example:

from collections import defaultdict

d = defaultdict(int) 
L = [1, 2, 3, 4, 2, 4, 1, 2] 
# Iterate through the list for keeping the count 
for i in L: 
    # The default value is 0 
    # so there is no need to initialise the key first 
    d[i] += 1


defaultdict(<class 'int'>, {1: 2, 2: 3, 3: 1, 4: 2})

This code:

If we were not using a defaultdict, but rather a normal dict, then we would have to start each loop iteration by checking whether there is already a key for this value. And if there isn’t a key, we would explicitly add it. Then we can increment it.

This is very cumbersome. You can see why the defaultdict(int) is so awesome!


We can supply a default factory of set, when we want a new set to be created implicit for each key. That way, we can simply call add() to add items to the set with each iteration. We don’t need to check whether the set already exists, and we don’t need to explicitly create the empty sets.

A good example is when we build an adjacency dictionary for an undirected, unweighted graph.

from collections import defaultdict

edges = set() # store edge edge as a tuple of (node_a, node_b)

# Imagine we're only provided with the edges.
edges.add((1, 2))
edges.add((1, 5))
edges.add((2, 1))
edges.add((2, 3))
edges.add((3, 2))
edges.add((3, 4))
edges.add((4, 3))
edges.add((4, 5))
edges.add((4, 6))
edges.add((5, 1))
edges.add((5, 2))
edges.add((5, 4))
edges.add((6, 4))

# Now we build the adjacency map
node_map = defaultdict(set) # Use a set in order to filter out duplicate connections
for x, y in edges:
    node_map[x].add(y)  # a list of vertices that link to x
    node_map[y].add(x)  # a list of vertices that link to y

for vertex, connections in node_map.items():
    print(f"{vertex}: {connections}")


1: {2, 5}
2: {1, 3, 5}
3: {2, 4}
4: {3, 5, 6}
5: {1, 2, 4}
6: {4}


This is useful if we want to create a dictionary of dictionaries. Each outer dict value is itself a dict created implicitly and initially empty.

A handy use case is where we want to build an adjacency dictionary for nodes edges that are weighted. E.g.

from collections import defaultdict

edges = dict() # store edge edge as a tuple of (node_a, node_b)

# Imagine we're only provided with edges and their weights
edges[('a', 'b')] = 10   # a is connected to b, with weight of 10
edges[('a', 'c')] = 20   # a is connected to c, with weight of 20. And so on
edges[('a', 'd')] = 30
edges[('b', 'c')] = 40
edges[('b', 'a')] = 50
edges[('c', 'a')] = 60

# Now we build the adjacency map
node_map = defaultdict(dict) 
for (x, y), weight in edges.items():
    node_map[x][y] = weight  # a list of vertices that link to x

for k, v in node_map.items():
    vals = "; ".join("to " + k + " with weight " + str(v) for k, v in v.items()) 
    print(f"From node {k}: {vals}")

The output looks like this:

From node a: to b with weight 10; to c with weight 20; to d with weight 30
From node b: to c with weight 40; to a with weight 50
From node c: to a with weight 60

This is great!

But what if we had just used a regular dict. I.e. instead of this:

# Now we build the adjacency map
node_map = defaultdict(dict) 

We could have instead written this:

# Now we build the adjacency map
node_map = {} 

Well, if we do that, then the output of our program is this:

    node_map[x][y] = weight
KeyError: 'a'

Why? Because the first time we reference node_map[a] is when we’re trying to do this:

node_map['a']['b'] = 10

But this throws a KeyError, because we’ve never assigned a value to node_map['a']. Python does not know that we node_map['a'] needs to be initialised to an empty dictionary, so that we can add a K:V pair to it.

When we use a defaultdict(dict), this is all taken care of!

A Custom Factory

We can actually initialise our defaultdict to any type and any initial value we like. For example:

from collections import defaultdict

def def_value():
    return "Not defined"

dd = defaultdict(def_value)
dd["some_key"] = "Some value"

print(dd["some_other_key"]) # Not explicitly set; so default is used

The output:

Some value
Not defined