decorator - How can getattr() respect python class properties? - Stack Overflow

admin管理员组

文章数量:1122832

I have a class that performs some useful job in my project

class Test:
    def __init__(self):
        self.__value = None

    def set_value(self, value):
        print(f"set_value(): value={value}")
        self.__value = value

    def get_value(self):
        print(f"get_value()")
        return self.__value

    value = property(get_value, set_value)

Among regular functions it has some getters/setters that is handy to group into properties for simpler client code. E.g.

t.value = 123

Now, in some parts of the application I need to do some extra actions (a print() call in my case) on every access to the Test object, on each method call. So I did this:

class Wrapper:
    def __init__(self, wrapped):
        self.wrapped = wrapped

    def __getattr__(self, attr):
        print(f"Wrapper::__getattr__(): {attr}")
        return getattr(self.wrapped, attr)

This is very similar to what decorators do, but this approach has 2 benefits:

• Original class is untouched, and I can use it separately from the Wrapper
• I can add as many new methods to the Test class, without having to add new decorators in Wrapper

Now the usage code:

t = Test()
w = Wrapper(t)

w.set_value(123)
print(f"Value: {w.get_value()}")

w.value = 234
print(f"Value: {w.value}")

Direct set/get calls work as expected. Unfortunately w.value syntax no longer work, and silently assign a value of 234 to the value name, instead of calling getters and setters.

How this issue can be fixed, so that Wrapper respects Test class properties?

I have a class that performs some useful job in my project

class Test:
    def __init__(self):
        self.__value = None

    def set_value(self, value):
        print(f"set_value(): value={value}")
        self.__value = value

    def get_value(self):
        print(f"get_value()")
        return self.__value

    value = property(get_value, set_value)

Among regular functions it has some getters/setters that is handy to group into properties for simpler client code. E.g.

t.value = 123

Now, in some parts of the application I need to do some extra actions (a print() call in my case) on every access to the Test object, on each method call. So I did this:

class Wrapper:
    def __init__(self, wrapped):
        self.wrapped = wrapped

    def __getattr__(self, attr):
        print(f"Wrapper::__getattr__(): {attr}")
        return getattr(self.wrapped, attr)

This is very similar to what decorators do, but this approach has 2 benefits:

• Original class is untouched, and I can use it separately from the Wrapper
• I can add as many new methods to the Test class, without having to add new decorators in Wrapper

Now the usage code:

t = Test()
w = Wrapper(t)

w.set_value(123)
print(f"Value: {w.get_value()}")

w.value = 234
print(f"Value: {w.value}")

Direct set/get calls work as expected. Unfortunately w.value syntax no longer work, and silently assign a value of 234 to the value name, instead of calling getters and setters.

How this issue can be fixed, so that Wrapper respects Test class properties?

• python
• decorator
• python-decorators

Share Improve this question Follow asked Nov 22, 2024 at 19:29 Oleksandr MasliuchenkoOleksandr Masliuchenko 84522 gold badges99 silver badges1919 bronze badges 2

• Does it work if you use @property and @value.setter decorators? – Barmar Commented Nov 22, 2024 at 19:39
• 2 You need to override the __setattr__ magic function. Otherwise, when you set value on wrapper it just uses normal semantics to set the name on your wrapper and not your wrapped/test object. There will be some edge cases where the behaviour is not exactly reproduced. eg. Test.value = ... vs Wrapper.value = ... (ie class attributes). – Dunes Commented Nov 22, 2024 at 19:51

Add a comment  | 

2 Answers 2

Sorted by: Reset to default 3

You can define a __setattr__ method to manage assignments. Note that it has to treat the wrapped attribute separately to avoid infinite recursion in __init__.

class Test:
    def __init__(self):
        self.__value = None

    def set_value(self, value):
        print(f"set_value(): value={value}")
        self.__value = value

    def get_value(self):
        print(f"get_value()")
        return self.__value

value = property(get_value, set_value)


class Wrapper:
    def __init__(self, wrapped):
        self.wrapped = wrapped

    def __getattr__(self, attr):
        print(f"Wrapper::__getattr__(): {attr}")
        return getattr(self.wrapped, attr)
    
    def __setattr__(self, attr, value):
        if attr == 'wrapped':
            super().__setattr__('wrapped', value)
        else:
            print(f"Wrapper::__setattr__(): {attr}")
            setattr(self.wrapped, attr, value)
    

    
    
    t = Test()
    w = Wrapper(t)
    
    w.set_value(123)
    print(f"Value: {w.get_value()}")
    print()
    w.value = 234
    print(f"Value: {w.value}")

Output:

Wrapper::__getattr__(): set_value
set_value(): value=123
Wrapper::__getattr__(): get_value
get_value()
Value: 123

Wrapper::__setattr__(): value
set_value(): value=234
Wrapper::__getattr__(): value
get_value()
Value: 234

Properties are descriptors and descriptors are looked up in the class, not in an instance --> data model / descriptors.

Instance Binding

If binding to an object instance, a.x is transformed into the call: type(a).__dict__['x'].__get__(a, type(a)).

The easiest way to have Test descriptors (properties) in the Wrapper is to inherit them.

Further, there is a better place to monitor almost all access than __getattr__. It's the __getattribute__--> data model / attribute access

object.__getattribute__(self, name)

Called unconditionally to implement attribute accesses for instances of the class. If the class also defines __getattr__(), the latter will not be called unless __getattribute__() either calls it explicitly or raises an AttributeError.

---

Result:

...

class Wrapper(Test):
    def __getattribute__(self, attr):
        print(f"Wrapper::__getattribute__(): {attr}")
        return super().__getattribute__(attr)

w = Wrapper()

...

本文标签： decoratorHow can getattr() respect python class propertiesStack Overflow