Why self.__dict__ can trigger getattribute and instance.attribute can't?

Question

The proof verify that self.__dict__ can trigger __getattribute__ in descriptor.

class Room:
    def __init__(self,name):
        self.name = name
    def __getattribute__(self,attr):
        print('in __getattribute__',attr)
        return  object.__getattribute__(self,attr)
    def __setattr__(self,attr,value):
        print('in __setattr__',attr)
        self.__dict__[attr] = value
        print('over')

Create an instance,i add some comments starting with # in the output info:

x = Room('r1')
# self.name = name in __init__ trigger __setattr__
in __setattr__ name  
# python interpreter read part of self.__dict__[attr] = value,extract self.__dict__ to trigger __getattribute__.
in __getattribute__ __dict__  
over

We saw that python interpreter can extract part of string in a whole line to trigger __getattribute__ method.

class Sample:
    def __init__(self, name):
        self.name = name
    def __get__(self,instance,owner):
        print('get called')
    def __set__(self, instance, value):
        print('set called')

class Person:
    name = Sample('name')

Make an instance:

p = Person()
p.name
get called

p.name can trigger __get__

p.name = 'tom'
set called

For the command p.name = 'tom',why python interpreter do not extract p.name invoke __get__ and then make p.name = 'tom' invoke __set__? Why the output is not as below?

p.name = 'tom'
get called   # p.name  trigger
set called   # p.name = 'tom' trigger

@jsbueno,The default __setattr__ will, ordinarily, create an entry in the instance __dict__, but that one will be refereed to in native code, using the internal slots, and not be redirected through __gettattribute__.It is not true.

import time
class Room:
    def __init__(self,name):
        self.name = name
    def __getattribute__(self,attr):
        print('in __getattribute__',attr)
        return  object.__getattribute__(self,attr)
    def __setattr__(self,attr,value):
        print('in __setattr__',attr)
        time.sleep(3)
        self.__setattr__(attr, value)
        print('over')

If we write self.__setattr__(attr, value) instead of super.__setattr__(attr, value),python interpreter extract self.__setattr__ and redirect into __gettattribute__,then parse super.__setattr__(attr, value) in the whole as __setattr__ ,jump into __setattr__ again,repeat without endness until you stop it by ctl+c.

x = Room('r1')
in __setattr__ name
in __getattribute__ __setattr__
in __setattr__ name
in __getattribute__ __setattr__
in __setattr__ name
in __getattribute__ __setattr__
in __setattr__ name
in __getattribute__ __setattr__
in __setattr__ name
in __getattribute__ __setattr__
in __setattr__ name

We draw some conclusion on data descriptor from above codes.
1.expressions without triggering __getattribute__.

super().__setattr__(attr, value)  
object.__setattr__(self,attr, value)  # i have checked

2.expressions with triggering __getattribute__.

self.__setattr__(attr, value)
self.__dict__[attr] = value

When self.__setattr__ trigger __getattribute__,__setattr__ was parsed as attr.
When self.__dict__ trigger __getattribute__,__dict__ was parsed as attr.
Puzzles remain here.
Puzzle1: With same structure:

super().__setattr__(attr, value)  
object.__setattr__(self,attr, value)
self.__setattr__(attr, value)

,python interpreter parse the first two expressions as setting value for attribute,never triger get nor set,Why python interpreter read part of code snippet self.__setattr__ in self.__setattr__(attr, value),to trigger get (__getattribute__) in a hurry?

Puzzle2:
self.__dict__[attr] = value is an assignment in the whole,python interpreter extract part code snippet self.__dict__ ,to trigger get,why don't python extract self.name in self.name = name in __init__ to trigger get (__getattribute__) ,with same action ?

Puzzle3:
Dig more deeper on property management,more strange action shocked me.As the above codes show that self.__setattr__(attr, value) trigger __getattribute__ in data descriptor,it trigger set instead of get in below code:

import time
class Room:
     def __init__(self,value):
         self.name = value  
     def get_name(self):
         print('in get')
         return self.__dict__['name']       
     def set_name(self, value):
         print('in set')
         time.sleep(3)
         self.__setattr__('name', value)
     name = property(get_name, set_name) 
r = Room('r1')
in set
in set
in set
in set

So i feel that some rules or logic still do not be summarized.

Answer 1

Assignemts will trigger __setattr__ for ordinary objects, and call the __set__ method in descriptors, as you verified.

The only reason your first example runs __getattribute__ to retrieve the instance __dict__ is because your custom implementation of __setattr__ does so.

The default __setattr__ will, ordinarily, create an entry in the instance __dict__, but that one will be refereed to in native code, using the internal slots, and not be redirected through __gettattribute__.

You can verify this easily by using the default __setattr__ implementation by callign super(), instead of creating a dict entry yourself:

class Room:
    def __init__(self,name):
        self.name = name
    def __getattribute__(self,key):
        print('in __getattribute__',key)
        return  object.__getattribute__(self,key)
    def __setattr__(self,attr,value):
        print('in __setattr__',attr)
        super().__setattr__(attr, value)
        print('over')

And in the interactive interpreter:

In [35]: r = Room("blah")
in __setattr__ name
over
In [36]: r.name
in __getattribute__ name
Out[36]: 'blah'

In [37]: r.__dict__
in __getattribute__ __dict__
Out[37]: {'name': 'blah'}

So, the behavior you observed in your __set__ in the descriptor is the ordinary behavior.