An attribute is a named value or behavior attached to an object, class, or module and accessed with dot syntax. For example, a Scoop instance can expose flavor, a class can expose max_scoops, and the os module exposes sep. Attributes let code keep data and operations close to the objects they describe. section 8.0 section 9.2
Use attributes in several related ways:
- Create an instance attribute in __init__, commonly by assigning self.flavor = flavor. Each instance then carries its own value. section 9.1 section 9.2
- Define a class attribute when one value belongs to the class as a whole. Bowl.max_scoops communicates a shared limit and avoids repeating a hardcoded value. section 9.11 section 9.12
- Access module attributes through a module name, such as os.sep, or import selected attributes into the current namespace. section 8.0
- Store other objects as attributes to express composition: a Bowl keeps Scoop instances in self.scoops rather than storing only their flavor strings. section 9.7 section 9.10
Attribute lookup follows the ICPO rule: Python checks the instance, then its class, then parent classes. Thus self.max_scoops lets one method work for Bowl and BigBowl while resolving 3 or 5 from the appropriate class. An instance may also shadow a class attribute by defining an attribute with the same name. section 9.12 section 9.17
Instance and class attributes
class Bowl: max_scoops = 3 def __init__(self): self.scoops = [] bowl = Bowl() print(bowl.max_scoops) # found on Bowl print(bowl.scoops) # found on bowl class BigBowl(Bowl): max_scoops = 5 big = BigBowl() print(big.max_scoops) # 5 A method that tests len(self.scoops) against self.max_scoops automatically uses the subclass limit without duplicating the method. section 9.12 section 9.17
Attributes can also represent behavior. Methods are attributes defined on classes, while special attributes such as __init__ and __repr__ use double underscores. Calling an instance method supplies the instance through the conventional first parameter, self. section 9.2 section 9.7
For data-focused classes, @dataclass reduces attribute-assignment boilerplate. Declare flavor: str at class level and the decorator supplies an initializer; annotations remain optional generally but are required for dataclass fields. External tools such as Mypy can check the declared types. section 9.10
@dataclass class Scoop(): flavor : str
Inspect the compact dataclass definition to see how a declared field becomes an instance attribute without writing __init__. section 9.10
