What are dataclasses, and how do they reduce boilerplate compared to manually writing __init__, __eq__, and __repr__?
@dataclass generates init, repr, and eq from class annotations
A dataclass is a normal class that the @dataclass decorator augments by reading annotations and generating boilerplate methods. By default it generates init, repr, and eq. With frozen=True it also generates hash and makes instances immutable. With order=True it generates comparison methods. Fields can be customized with field(), which supports default_factory, repr=False, and comparison flags. The result is concise, readable value objects with correct semantics. The trade-offs are that generated init can be awkward with complex validation, and that dataclasses are mutable by default, so hashability requires an explicit choice.
- 1
@dataclass(eq=True, frozen=True) is the standard pattern for immutable value objects that need to be hashable.
- 2
@dataclass(order=True) generates ordering based on field order; be explicit about the field order.
- 3
field(default_factory=list) is required for mutable defaults. Using a list directly raises ValueError in 3.11+.
- 4
Trade-off: dataclasses generate methods you cannot easily customize per field. For complex validation or serialization, use Pydantic or attrs.
- 5
Common mistake: assuming dataclasses are immutable. They are mutable unless frozen=True.
- 6
Common mistake: defining a field with a mutable default and sharing it across instances.
- 7
Version note: dataclasses landed in 3.7. slots=True arrived in 3.10, kw_only in 3.10, and 3.11 added stricter mutable-default checks and improved error messages.
Scenario Questions
0-2 years experience
- 1How do you create a simple value object without writing __init__ and __repr__?
- 2What is the default value of eq and frozen in @dataclass?
2-5 years experience
- 1You need a dataclass field that is a list per instance. What do you use instead of a mutable default?
- 2You need ordering by a specific field order. How do you enable and control it?
5-8 years experience
- 1You need a dataclass with validation and a canonical serialization format. How do you extend dataclasses without losing ergonomics?
- 2You need a dataclass with slots to reduce memory. What changes in behavior and what breaks with inheritance?
8+ years experience
- 1Design a domain model library using dataclasses that supports immutability, validation, serialization, and versioning without duplicating field metadata.
- 2Compare dataclasses, attrs, and Pydantic on performance, validation, typing, and migration cost, and recommend a strategy for a large codebase.
Follow-up Questions
- How do you make a dataclass immutable and hashable?
- When would Pydantic or attrs be a better choice than a dataclass?