Why __getattr__ triggers __setattr__ in the following python code?

class D:
    def __init__(self):
        self._attr = 1
        self._attr2 = 2

    def __getattr__(self, name):
        if name == 'data':
            return (self._attr, self._attr2)

    def __setattr__(self, name, value):
        print('__setattr__')
        if name == 'data':
            self._attr, self._attr2 = value
        else:            
            super().__setattr__(name, value)

d = D()
print(d.data)  
#OUTPUT:
#__setattr__
#__setattr__
#(1,2)

For some reason the line return (self._attr, self._attr2) behaves as an assignment to self._attr and self._attr2, but why?

I consulted Google and Gemini, the Mark Lutz book, but couldn't figure it out.

It's not coming from the print(d.data) line. Comment it out. It is from setting the attributes in __init__. Extend the print in __setattr__ to print(f'__setattr__({name=}, {value=})') to see more information:

class D:
    def __init__(self):
        self._attr = 1
        self._attr2 = 2

    def __getattr__(self, name):
        if name == 'data':
            return (self._attr, self._attr2)

    def __setattr__(self, name, value):
        print(f'__setattr__({name=}, {value=})')
        if name == 'data':
            self._attr, self._attr2 = value
        else:
            super().__setattr__(name, value)

d = D()
#print(d.data)

Output:

__setattr__(name='_attr', value=1)
__setattr__(name='_attr2', value=2)

Note you can also get the same behavior implementing data as a property and have more clear code:

class D:
    def __init__(self):
        self._attr = 1
        self._attr2 = 2

    @property
    def data(self):
        return (self._attr, self._attr2)

    @data.setter
    def data(self, value):
        self._attr, self._attr2 = value

d = D()
print(d.data)
d.data = 4,5
print(d.data)

Output:

(1, 2)
(4, 5)