TopolearnGrouping records by a key is everywhere: orders by customer, files by extension, edges by node.
A defaultdict is a dict that calls a factory function to create a value for a missing key on first access:
from collections import defaultdict
by_ext = defaultdict(list)
by_ext[".py"].append("app.py") # list created automaticallyThe argument is the factory itself, not a value: pass list, set, int or a lambda, without calling it. Common choices:
defaultdict(list)groups items into lists.defaultdict(set)groups unique items.defaultdict(int)counts (thoughCounteris usually nicer).defaultdict(lambda: "n/a")for any other default.
Any access with d[key] inserts the default, even if you're only reading. That can silently grow the dict; use key in d or d.get(key) for lookups that shouldn't create entries. get never calls the factory.
Compared with dict.setdefault(key, []).append(x), defaultdict is clearer and doesn't build an unused empty list on every call.
A defaultdict prints as defaultdict(<class 'list'>, {...}). Convert with dict(d) before returning it from an API or serialising, so callers don't get the auto-insert behaviour unexpectedly.
Example
from collections import defaultdict
files = ["app.py", "README.md", "test_app.py", "setup.cfg", "notes.md"]
by_ext = defaultdict(list)
for name in files:
by_ext[name.rsplit(".", 1)[-1]].append(name)
print(dict(by_ext))
# {'py': ['app.py', 'test_app.py'], 'md': ['README.md', 'notes.md'], 'cfg': ['setup.cfg']}
graph = defaultdict(set)
for a, b in [("A", "B"), ("A", "C"), ("B", "C")]:
graph[a].add(b)
graph[b].add(a)
print(sorted(graph["C"])) # ['A', 'B']
print(graph.get("Z")) # None: get doesn't insert
print(graph["Z"], "Z" in graph) # set() True: [] does insertRemember
defaultdict(factory)creates missing values by calling the factory- Pass the factory uncalled:
list, notlist() d[key]inserts on read; useinorgetto check without inserting
Watch out for
- Writing
defaultdict([]); the argument must be a callable such aslist.