将Python Fire与数据类相结合

Question

我有一个像下面这样的代码，里面有多个函数用于各种计算。

我使用 python fire 来传递参数，而不是定义 argparse 并从 cli 调用函数。每次添加新参数时，我都必须在 init 中为其添加

self

。我正在寻找更好的方法。

我发现 python dataclasses 可以解决这个问题。我研究了 python fire 命令分组和多个命令。

class MyClass:
    def __init__(
        self,
        input_path: str,
        output_path: str = '',
        same_size: bool = False,
        crop_size: int = 300,
        padding: int = 20,
        write_json: bool = False,
        write_image: bool = False,
        line_thickness: int = 2,
        side_color: Tuple = (255, 255, 0),
        top_color: Tuple = (255, 0, 0),
    ) -> None:
        super(MyClass, self).__init__()
        self.input_path = input_path
        self.output_path = output_path
        self.same_size = same_size
        self.crop_size = crop_size
        self.padding = padding
        self.write_json = write_json
        ...
        ...

    def something(
        self,
    ) -> None:
        ...
        ...

if __name__ == '__main__':
    fire.Fire(MyClass)

我的问题是如何使用 python fire 正确获取

MyDataClass

中的

MyClass

数据类值？

@dataclass
class MyDataClass:
    input_path: str
    output_path: str = ''
    same_size: bool = False
    crop_size: int = 300
    padding: int = 20
    write_json: bool = False
    write_image: bool = False
    line_thickness: int = 2
    side_color: Tuple = (255, 255, 0)
    top_color: Tuple = (255, 0, 0)

class MyClass:
    # init is not here anymore.

    def something(
        self,
    ) -> None:
        ...
        ...

Answer 1

如果开箱即用火支持就更好了。尽管如此，您可以在运行时定义一个方法来汇总所有数据类属性。该方法可以输入到

Fire

。

from dataclasses import dataclass, fields
from fire import Fire

@dataclass
class MyDataClass:
    input_path: str =''
    output_path: str = ''
    same_size: bool = False
    crop_size: int = 300
    padding: int = 20
    write_json: bool = False
    write_image: bool = False
    line_thickness: int = 2
    side_color: tuple = (255, 255, 0)
    top_color: tuple = (255, 0, 0)

    @classmethod
    def generate_init(cls):
        init_args = []
        for f in fields(cls):
            default = f.default
            if not default and issubclass(f.type, str):
                default = '""'
            init_args.append(f'{f.name}:{f.type.__name__}={default}')
        code = f'def fn(self, {", ".join(init_args)}):\n'
        code += '\n'.join(f'    self.{f.name} = {f.name}' for f in fields(cls))
        loc = {}
        exec(code, None, loc)
        fn = loc["fn"]
        cls.fire = fn
        return


MyDataClass.generate_init()
dc = MyDataClass()
Fire(dc.fire)  # use CLI
# Fire(dc.fire, "--crop_size=999")  # interactive example
print("Result:", dc)