What is the difference between a dictionary and a defaultdict?

Question

Accepted Answer

A dict is Python’s built-in hash map: key-value pairs where accessing a missing key raises KeyError . A collections.defaultdict is a subclass of dict that automatically creates a default value the first time a missing key is read. How defaultdict differs: you pass a zero-argument factory callable at construction. When you access a key that isn’t there, the factory is called, the result is stored under that key, and that value is returned — all in one step. from collections import defaultdict counts = defaultdict(int) # factory = int, produces 0 for word in "the quick brown fox".split(): counts[word] += 1 # no KeyError; first access creates 0 groups = defaultdict(list) # factory = list, produces [] for word in words: groups[len(word)].append(word) # auto-creates the list The equivalent with plain dict : doable but noisier — you use d.setdefault(key, factory()) or if k not in d: d[k] = [] . groups = {} for word in words: groups.setdefault(len(word), []).append(word) Common factories: int — counters (0 default). list — grouping values into buckets. set — collecting unique values per key. lambda: 0.0 or lambda: "N/A" — custom scalar defaults. A class — lazily instantiate rich objects. Behavior to know: Reading a missing key inserts and returns the factory result. This can surprise code that expects a “pure” read to leave the dict unchanged. in and .get() do not trigger the factory — use them if you want to probe without auto-insert. The factory is .default_factory ; setting it to None makes a defaultdict behave like a normal dict again. Pickling/unpickling and JSON dumps work fine; JSON serializes the current contents as a regular object. When to reach for each: Use dict for general key-value storage or when you explicitly want missing-key errors to surface bugs. Use defaultdict when you’re building per-key collections (counters, groupings, adjacency lists) and want to drop the if-exists boilerplate. For pure counting, collections.Counter is even more direct: Counter(words) . Gotcha to watch: because reading inserts, logging d["missing"] in an error path changes the dict . Use d.get("missing") for pure reads. Interview-ready summary: a plain dict raises on missing keys; a defaultdict(factory) auto-creates and stores a default value. Use defaultdict to simplify grouping, counting, and bucketing code — and use Counter when all you’re doing is counting. Show the before/after: verbose dict counting vs clean defaultdict counting. Know the three common factories: int (counting), list (grouping), set (unique grouping). Mention Counter as the purpose-built solution for counting.