Python无单一父类场景的类型提示：Sklearn缩放器标注咨询

1. 恰当的类型提示方式 #

有两种更优雅的方案，比直接写Union或硬编码具体类更合适：

• 基于公共父类组合：既然所有目标缩放器都继承自OneToOneFeatureMixin、TransformerMixin、BaseEstimator，可以直接用这三个类的组合作为类型提示（Python 3.10+支持|操作符表示类型并集）：

  from sklearn.base import OneToOneFeatureMixin, TransformerMixin, BaseEstimator
  from abc import ABC
  import pandas as pd
  
  class FeatureEngineering(ABC):
      def __init__(self, dataframe: pd.DataFrame, scaler: OneToOneFeatureMixin | TransformerMixin | BaseEstimator | None = None):
          self.dataframe = dataframe
          self.scaler = scaler
  

  这种方式直接利用现有继承关系，静态检查工具能准确识别合规的缩放器类型。

• 基于协议（Protocol）：定义一个仅关注核心行为的协议，只要类实现了缩放器必备的方法，就会被视为符合该类型，这种方式更灵活，不局限于sklearn的继承体系：

  from typing import Protocol
  import pandas as pd
  from abc import ABC
  
  class ScalerProtocol(Protocol):
      def fit(self, X: pd.DataFrame, y=None) -> "ScalerProtocol":
          ...
      def transform(self, X: pd.DataFrame) -> pd.DataFrame:
          ...
      def fit_transform(self, X: pd.DataFrame, y=None) -> pd.DataFrame:
          ...
  
  class FeatureEngineering(ABC):
      def __init__(self, dataframe: pd.DataFrame, scaler: ScalerProtocol | None = None):
          self.dataframe = dataframe
          self.scaler = scaler
  

  自定义缩放器只要实现上述方法，就能被正确识别为合规类型。

2. Python允许创建特定类型，实现方式如下 #

可以通过三种方式为该场景创建专属类型：

• 使用TypeAlias定义类型别名：轻量化地给现有类型组合起一个语义清晰的别名，静态检查时等价于原类型：

  from typing import TypeAlias
  from sklearn.base import OneToOneFeatureMixin, TransformerMixin, BaseEstimator
  
  GeneralScaler: TypeAlias = OneToOneFeatureMixin | TransformerMixin | BaseEstimator
  
  # 用法示例
  class FeatureEngineering(ABC):
      def __init__(self, dataframe: pd.DataFrame, scaler: GeneralScaler | None = None):
          self.dataframe = dataframe
          self.scaler = scaler
  

• 使用NewType创建语义类型：给类型赋予专属名称，仅用于静态类型检查，运行时不会做额外限制：

  from typing import NewType
  from sklearn.base import OneToOneFeatureMixin, TransformerMixin, BaseEstimator
  
  GeneralScaler = NewType("GeneralScaler", OneToOneFeatureMixin | TransformerMixin | BaseEstimator)
  
  # 用法示例
  class FeatureEngineering(ABC):
      def __init__(self, dataframe: pd.DataFrame, scaler: GeneralScaler | None = None):
          self.dataframe = dataframe
          self.scaler = scaler
  

• 使用TypeVar绑定约束类型：创建一个受约束的类型变量，限定只能是同时继承指定父类的类型，检查更严谨：

  from typing import TypeVar
  from sklearn.base import OneToOneFeatureMixin, TransformerMixin, BaseEstimator
  
  ScalerType = TypeVar("ScalerType", bound=OneToOneFeatureMixin & TransformerMixin & BaseEstimator)
  
  # 用法示例
  class FeatureEngineering(ABC):
      def __init__(self, dataframe: pd.DataFrame, scaler: ScalerType | None = None):
          self.dataframe = dataframe
          self.scaler = scaler
  

内容的提问来源于stack exchange，提问作者An old man in the sea.