Python dictionary keys() return a dict_keys object, instead of list, causing awkward behavior

2025-03-02   python pythonquirk til

---

Example

d = {"a": 1, "b": 2}

keys_view = d.keys()

print(keys_view)  # dict_keys(['a', 'b'])

d["c"] = 3  # Modify the dictionary

print(keys_view)  # dict_keys(['a', 'b', 'c']) -> Updates automatically!

Key reasons for this design choice (per chatgpt):

1. Efficient Memory Usage: A view object does not create a separate list of keys. Instead, it dynamically reflects the dictionary's keys without extra memory overhead. This is important for large dictionaries, where copying all keys into a list would be expensive.
   

2. Dynamic Updates (Live View): A dictionary view object updates automatically when the dictionary changes. If keys are added or removed, the view reflects these changes without needing to regenerate the list.
   

3. Set-Like Behavior: dict_keys behaves like a set, meaning it supports fast membership testing (in operation). This allows efficient operations like set intersections with other collections.
   

Also see Python Design and History FAQ (archived)

Couldn't find anything in Neopythonic - Guido Van Rossum's blog (archived) about this

---

---

---