[Road to Python Intermediate] Define __getattr__ function in class

Link to summary

https://qiita.com/ganariya/items/fb3f38c2f4a35d1ee2e8

Introduction

In order to study Python, I copied a swarm intelligence library called acopy.

In acopy, many interesting Python grammars and idioms are used, and it is summarized that it is convenient among them.

This time, let's define a __getattr__ special method in the class.

Instance attributes

As a way to retrieve attributes used inside an instance in Python There is a way to refer to __dict__.

The name of the namespace used inside the instance is defined as a dictionary in __dict __.

For example, take the following example.

class A:

    def __init__(self, x, y):
        self.x = x
        self.y = y


# {'x': 10, 'y': 20}
a = A(10, 20)
print(a.__dict__)

Instance a has names $ x, y $ You can retrieve them as a dictionary.

__getattr__

You can define a special method called __getattr__ in a Python class.

In a Python instance name call If called with an undefined name, it will return ** AttributeError **.

But if __getattr__ is defined, Instead of returning an AttributeError "on a call with an undefined name", __getattr__ is executed.

Let's look at an example.

class A:

    def __init__(self, x, y):
        self.x = x
        self.y = y

    def __getattr__(self, name):
        print(name, 'getattr')
        if name == 'hello':
            return self.world
        raise AttributeError

    def world(self, x):
        return x


'''
{'x': 10, 'y': 20}
hello getattr
<bound method A.world of <__main__.A object at 0x10e602150>>
hello getattr
20
'''
a = A(10, 20)
print(a.__dict__)
print(a.hello)
print(a.hello(20))

The __getattr__ method is defined in the A class above.

When I try to call ʻa.hello, the attribute hello is not defined in a. So the getattr` method is called and returns the world method instead.

In ʻa.hello (20)`, the world method is returned first, and the world method is given a value of 20 and executed. Therefore, it is executed when the name hello is confirmed.

Therefore, if you use the above properties, you will get the following useless code.

class A:

    def __init__(self, x, y):
        self.x = x
        self.y = y

    def __getattr__(self, name):
        print(name, 'getattr')
        if name == 'hello':
            return self.world
        if name == 'world':
            return self.great
        if name == 'great':
            return self.bad
        if name == 'bad':
            return 0
        raise AttributeError


'''
{'x': 10, 'y': 20}
hello getattr
world getattr
great getattr
bad getattr
0
'''
a = A(10, 20)
print(a.__dict__)
print(a.hello)

Application example

I can't think of many application examples, If there is no attribute, I think there is something like letting you do alternative rule processing. Also, it seems that the branch etc. may be changed depending on the call name.

class A:

    def __getattr__(self, name):
        self.nothing(name)

    def exist(self):
        print('exist!')

    def nothing(self, name):
        print(f'nothing {name}')


'''
exist!
nothing unnti
'''
a = A()
a.exist()
a.unnti

In addition, as each content as a separate article at a later date It seems that it can also be used by throwing an exception code when there is no name in the lambda expression.

References

-Python attribute handling mechanism