What is the difference between __slots__ and a regular class's default __dict__-based attribute storage, and how does __slots__ improve memory efficiency?
slots stores attributes in fixed slots instead of a per-instance dict
By default, each instance of a class gets a dict, a hash table that maps attribute names to values. That is flexible but expensive: the dict itself has overhead, and every instance carries it even if the class has two attributes. Defining slots replaces the per-instance dict with a fixed set of C-level descriptors, one per slot, so an instance stores only the declared attributes in pre-allocated offsets. The savings are significant for classes created in large numbers: often 40 to 60 percent less memory per instance. The costs are real too: you cannot add arbitrary attributes, slots do not participate in dict-based tooling like vars(), and multiple inheritance with non-empty slots is restricted to one base with non-empty slots and the rest with empty slots. Choose slots for data-heavy value objects created in bulk, and avoid them when you need dynamic attributes or heavy introspection.
- 1
Slots eliminate the per-instance dict, saving memory and slightly speeding up attribute access.
- 2
Declare slots as a tuple of names to save a few bytes over a list.
- 3
Include weakref and dict explicitly in slots if you need weak references or dynamic attributes.
- 4
dataclasses support slots=True since 3.10, which is the cleanest way to get both.
- 5
Trade-off: slots remove flexibility. Dynamic attributes, monkey-patching, and some serialization libraries break.
- 6
Common mistake: defining slots in a subclass while the parent still has dict, which keeps the dict and negates most savings.
- 7
Common mistake: assuming slots makes instances immutable. It does not; assignment to declared slots still works.
- 8
Version note: dataclass slots=True arrived in 3.10. The basic slot semantics have been stable since Python 2.2.
Scenario Questions
0-2 years experience
- 1Why does a class with __slots__ use less memory per instance?
- 2Can you add a new attribute to an instance of a class with __slots__?
2-5 years experience
- 1You add __slots__ to a subclass but memory does not improve. Why?
- 2You need both slots and weak references on an instance. How do you declare it?
5-8 years experience
- 1You create millions of small records in a pipeline and memory is the bottleneck. How do you apply slots and measure the improvement?
- 2A serialization library breaks after you add slots. How do you reconcile serialization with slots?
8+ years experience
- 1Design a data model for high-volume ingestion that minimizes per-instance overhead while remaining introspectable and serializable.
- 2Compare __slots__, dataclasses with slots, and namedtuples on memory, attribute access speed, and compatibility with typing and pickling.
Follow-up Questions
- What happens if a parent class has __dict__ and a child defines __slots__?
- When would __slots__ make attribute access slower rather than faster?