How make my pattern observer with metaclass work?

Question

I want set a pattern observer on my class property

I try to use @classmethod but it doesn't have setter property.

class dataframe():
    df = None
    @classmethod
    @property
    def weather(cls):
        return cls.df
    @classmethod
    @weather.setter
    def weather(cls,value):
        cls.df= value
        print("the weath was change {}".format(cls.df))

<ipython-input-119-7e26ac08cb26> in dataframe()
      6         return cls.df
      7     @classmethod
----> 8     @weather.setter
      9     def weather(cls,value):
     10         cls.df= value

AttributeError: 'classmethod' object has no attribute 'setter'

Then I tried to adapt the solution I found there to my problem Using property() on classmethods

class dataframe_meta(type):
    def __init__(cls, *args, **kwargs):
        cls.df = None

    @property
    def change(cls):
        return cls.df

    @change.setter
    def change(cls, value):
        cls.df = value
        print("the weath was change {}".format(cls.df))

class dataframe(metaclass=dataframe_meta):
    pass

dataframe.df = 5 It doesn't return any error but the print from the function setter was not displayed.

How make it work properly?

This answer in the question you linked might be more helpful to you https://stackoverflow.com/a/39542816/548562 — Iain Shelvington, Oct 07 '19 at 00:44
You won't see the print statement because you're not calling the setter, you're modifying the member directly. `dataframe.change = 5` would display the print statement you're looking for. That being said, I'm not sure the general approach is the best but it's hard to say without knowing more. — jedwards, Oct 07 '19 at 00:44

score 1 · Answer 1 · answered Oct 07 '19 at 00:49

If you'd like to watch changes made to the df property, you should name the method df, and name the attribute that stores the actual value of df something else, such as _df:

class dataframe_meta(type):    
    def __init__(cls, *args, **kwargs):
        cls._df = None

    @property
    def df(cls):
        return cls._df

    @df.setter
    def df(cls,value):
        cls._df= value
        print("the weath was change {}".format(cls._df))

score 0 · Answer 2 · answered Oct 07 '19 at 15:08

The code you posted,


class dataframe_meta(type):
    def __init__(cls, *args, **kwargs):
        cls.df = None

    @property
    def change(cls):
        return cls.df

    @change.setter
    def change(cls, value):
        cls.df = value
        print("the weath was change {}".format(cls.df))

class dataframe(metaclass=dataframe_meta):
    pass

will work as is - however, you named the property change not df - if you do dataframe.change = 5 you will see the print.

The access to dataframe.df itself is unguarded, however.

If you want dataframe.df itself to trigger the getter/setter methods, you have to name the property itself df, and store the result in an attribute with another name. And yes, metaclasses are a straightforward way for "property" to work on class attributes.

How make my pattern observer with metaclass work?

2 Answers2