Develop Reference

How does Python turn a function into a method?

I know that functions are just descriptors, like this:
def func(self):
print(self.name)
class C:
def __init__(self, name):
self.name = name
C.func = func
c = C("foo")
c.func()
I thought at first that c.func equals C.func.__get__(c)，yes，C.func.__get__(c) return a bound method. But when I set the __get__ of func to None, c.func still returns a bound method.
def func(self):
print(self.name)
class C:
def __init__(self, name):
self.name = name
func.__get__ = None
C.func = func
c = C("foo")
c.func
output:
<bound method func of <__main__.C object at 0x0000027EB23BF088>>
So I'm confused. Moreover, I found that when calling a function from an instance, Python actually calls the class's ___getAttribute__ method, which returns a bound method.
def func(self):
print(self.name)
func.__get__ = None
class C:
def __getattribute__(self, name):
r = super().__getattribute__(name)
print(r) # r is a bound method already
return r
def __init__(self, name):
self.name = name
C.func = func
c = C("foo")
c.func
output:
<bound method func of <__main__.C object at 0x0000027EB243D1C8>>
func.__get__ doesn't seem to have any effect. So, What happended in __getattribute__? How does Python turn a function into a method? I've Googled and done some research, but I still can't find the answer.
Maybe I'm making things complicated, In my understanding, function is itself a descriptor, but just like the code below, I set the func to None, it works normally:
class C:
def func(self):
print('hello world')
func.__get__ = None
c = C()
c.func()
but if it's a descriptor, it will raise TypeError:
class C:
class D:
def __get__(self, inst, cls):
if inst is None:
return self
return 'hello world'
D.__get__ = None
func = D()
c = C()
c.func

Well, if I understand correctly from what I found. (Since I didn't know the descriptors, that's exactly why I like to help, still learning)
First, let's look at __getattr__ and __getattribute__.
Let's have an empty class A
class A:
pass
If I initialize an object and try to call a property, because there is none at the moment, we get AttributeError.
a = A()
a.some_property
The following occurs:
Simple check of flow:
class FlowDemo:
def __init__(self):
self.inited_property = True
def __getattribute__(self, item):
if item in ('__class__', '__len__') : # For less spam of getting this attribute, if you want, you can remove condition.
print('Get Attribute', item)
# Call default behavior
return super().__getattribute__(item)
def __getattr__(self, item):
print('Get Attr', item)
if item == 'some_magic_name':
return "It's magic!"
raise AttributeError
fd = FlowDemo()
fd.inited_property
# Get Attribute inited_property
# True
fd.some_magic_property
# Get Attribute some_magic_name
# Get Attr some_magic_name
# "It's magic!"
fd.some_property
# Get Attribute some_property
# Get Attr some_property
# Traceback (most recent call last):
# File "<input>", line 1, in <module>
# File "stack-class-property-and-descriptors.py", line 67, in # __getattr__
# raise AttributeError
# AttributeError
This is probably understandable, including the use. But to be sure, I'll give an example. This logic is used as a dynamic representation of the result from the databases (mapping of attributes to ordinary dict, list, etc.).
But it can also be just logic for accessing an attribute (property), such as an access counter or validation (but this applies to __setattr__ and __setattribute__)
And what about descriptors?
First let's look at data-descriptors, they are easier for me to understand.
This is a class or decoder that has __get__ and one or both of __set__ and __delete__.
Once this is defined, python, when used in the property definition with it and then does not return a class but the value it obtains through __get__, does not overwrite an already declared class when declaring a value, but uses its __set__.
Example:
class WeekDayDescriptor:
def __init__(self):
self.__week_day = 0
def __get__(self, instance, owner=None):
return self.__week_day
def __set__(self, instance, value):
if not isinstance(value, int):
raise TypeError('Value must be int')
if not (0 <= value < 6):
raise ValueError('Value must be in range 0 - 6')
self.__week_day = value
class Calendar:
week_day = WeekDayDescriptor()
def __init__(self, week_day):
self.week_day = week_day
Demo:
c = Calendar(9)
# ValueError: Value must be in range 0-6
c = Calendar('6')
# TypeError: Value must be int
c = Calendar(3)
c.week_day = 6
c.week_day = 10
# ValueError: Value must be in range 0-6
c.week_day = 'monday'
# TypeError: Value must be int
Decorator descriptor:
class Calendar:
#property
def week_day(self):
return self.__week_day
#week_day.setter
def week_day(self, week_day):
if not isinstance(week_day, int):
raise TypeError('Value must be int')
if not (0 <= week_day < 6):
raise ValueError('Value must be in range 0 - 6')
self.__week_day = week_day
def __init__(self, week_day):
self.week_day = week_day
pass
And now for non-data descriptors...
A non-data descriptor is one that has only __get__.
As I understand it, each method automatically has its own descriptor, thanks to which the functions get references to the object - self.
We can write our own descriptor for a function / method, but it's not that straightforward, we have to help ourselves and get around it a bit.
def function_as_method(self, value):
print(self, value)
class HelperDescriptor:
def __get__(self, instance, owner):
def wrapper(*args, **kwargs):
return function_as_method(instance, *args, **kwargs)
return wrapper
class Foo:
baz = HelperDescriptor()
>>> bar = Foo()
>>> bar.baz(1)
<__main__.Foo object at 0x7f64f7768b70> 1
Source of last code block, but in czech lang.
And finally, your mentioned problem, when we set __get__ to None and you still get a reference to the function.
It's simple, python doesn't directly distinguish between primitive data types and functions, it's all a variable (or attribute / property) that has a value. Whether it's just value or it's callable is a different matter.
def f(): return True
print(type(f), f())
# <class 'function'> True
f = 123
print(type(f), f)
# <class 'int'> 123
Therefore, when we ask for the obj.func method or call it obj.func() directly, the first two changed magic is called first - __getattribute__ and __getattr__.
And if we call a method, it is called only after we get a reference to a function in memory.
Again a simple example:
def func(self, value):
print('Printing:', value)
class PrintDescriptor:
def __init__(self, name):
self.name = name
def __get__(self, instance, owner):
def wrapper(*args, **kwargs):
print(f"Calling '{self.name}' method")
return func(instance, *args, **kwargs)
return wrapper
class B:
foo = PrintDescriptor('foo')
bar = PrintDescriptor('bar')
def __getattribute__(self, item):
if item not in ('__len__', '__class__', '__dict__'):
print('Get Attribute', item)
return super().__getattribute__(item)
Demo:
b = B()
b.foo
# Get Attribute foo
# <function PrintDescriptor.__get__.<locals>.wrapper at 0x7f774a782ee0>
b.foo(2)
# Get Attribute foo
# Calling 'foo' method
# Printing: 2
b.bar(4)
# Get Attribute bar
# Calling 'bar' method
# Printing: 4
Sources:
https://www.datacamp.com/community/tutorials/python-descriptors#above1
https://blog.milde.cz/post/319-pokrocile-techniky-v-pythonu-deskriptory/
Python Doc, __get__
Python Docs, __getattribute__
Python Docs, __getattr__

Add an automated function call to each method

Is it possible to create a "constructor".. or rather "Initializer" to each function, instead of having to manually write it at the top of each function in class?
So, each time a function in a class is called, the other assigned function (unknown to caller) is always called first (called pre_check in below example).
An example using super(), but I then have to manually copy it inside each function.
class Helper():
def pre_check(self):
print("Helper fcn")
class Parent(Helper):
def __init__(self):
print("Initializer")
def foo(self):
super().pre_check() # <---- new code
# ... existing code here ...
def bar(self):
super().pre_check() # <---- new code
# ... existing code here ...
def many_more_functions(self):
super().pre_check() # <---- new code
# ... existing code here ...
m = Parent()
m.foo()
m.bar()
Note how __init__ in Parent is not supposed to run pre_check.

You can use a decorator for the class that will in turn decorate all public methods defined in the class:
def addhelper(helpmethod):
def deco(cls):
def decomethod(method):
def inner(self, *args, **kwargs):
helpmethod(self)
return method(self, *args, **kwargs)
# copy signature, doc and names from the original method
inner.__signature__ = inspect.signature(method)
inner.__doc__ = method.__doc__
inner.__name__ = method.__name__
inner.__qualname__ = method.__qualname__
return inner
# search all methods declared in cls with a name not starting with _
for name, meth in inspect.getmembers(
cls,lambda x: inspect.isfunction(x)
and not x.__name__.startswith('_')
and x.__qualname__.startswith(cls.__name__)):
# replace each method with its decoration
setattr(cls, name, decomethod(meth))
return cls
return deco
class Helper():
def pre_check(self):
print("Helper fcn")
#addhelper(Helper.pre_check)
class Parent(Helper):
def __init__(self):
print("Initializer")
def foo(self):
# super().pre_check() # <----
print('in foo')
def bar(self):
# super().pre_check() # <----
print('in bar')
def many_more_functions(self):
# super().pre_check() # <----
print('in many_more_functions')
We can now use it:
>>> p = Parent()
Initializer
>>> p.foo()
Helper fcn
in foo
>>> p.bar()
Helper fcn
in bar
>>> p.many_more_functions()
Helper fcn
in many_more_functions

Use __init_subclass__ to change subclasses as they are created. You can wrap the methods of subclasses:
class Helper():
def __init_subclass__(cls):
for field, value in cls.__dict__.items():
# add additional checks as desired, e.g. exclude __special_methods__
if inspect.isfunction(value) and not getattr(value, 'checked', False):
setattr(cls, field, cls._check(value)) # wrap method
#classmethod
def _check(cls, fcn):
"""Create a wrapper to inspect the arguments passed to methods"""
#functools.wraps(fcn)
def checked_fcn(*args, **kwargs):
print(fcn, "got", args, kwargs)
return fcn(*args, **kwargs)
return checked_fcn
class Parent(Helper):
def __init__(self):
print("Initializer")
def foo(self):
print("Foo")
Note that this will wrap all methods, including special methods such as __init__:
>>> Parent().foo()
<function Parent.__init__ at 0x1029b2378> got (<__main__.Parent object at 0x102c09080>,) {}
Initializer
<function Parent.foo at 0x1029b2158> got (<__main__.Parent object at 0x102c09080>,) {}
Foo
You can extend the check in __init_subclass__ with arbitrary rules to filter out functions. For example, field[:2] == field[-2:] == "__" excludes special methods.

You can use metaclass and define a decorator for each method in the instance of that metaclass
Code :
def decorate(f):
def do_something(self, a):
if (f(self, a) > 18) :
return ("Eligible to vote")
else :
return ("Not eligible to vote")
return do_something
class Meta(type):
def __new__(cls, name, bases, namespace, **kwds):
namespace = {k: v if k.startswith('__') else decorate(v) for k, v in namespace.items()}
return type.__new__(cls, name, bases, namespace)
class MetaInstance(metaclass=Meta):
def foo1(self, val):
return val + 15
def foo2(self, val):
return val + 9
obj1 = MetaInstance()
print(obj1.foo1(5))
print(obj1.foo2(2))

Using #classmethod with #property [duplicate]

This question already has answers here:
Using property() on classmethods
(19 answers)
Closed 3 years ago.
In python I can add a method to a class with the #classmethod decorator. Is there a similar decorator to add a property to a class? I can better show what I'm talking about.
class Example(object):
the_I = 10
def __init__( self ):
self.an_i = 20
#property
def i( self ):
return self.an_i
def inc_i( self ):
self.an_i += 1
# is this even possible?
#classproperty
def I( cls ):
return cls.the_I
#classmethod
def inc_I( cls ):
cls.the_I += 1
e = Example()
assert e.i == 20
e.inc_i()
assert e.i == 21
assert Example.I == 10
Example.inc_I()
assert Example.I == 11
Is the syntax I've used above possible or would it require something more?
The reason I want class properties is so I can lazy load class attributes, which seems reasonable enough.

Here's how I would do this:
class ClassPropertyDescriptor(object):
def __init__(self, fget, fset=None):
self.fget = fget
self.fset = fset
def __get__(self, obj, klass=None):
if klass is None:
klass = type(obj)
return self.fget.__get__(obj, klass)()
def __set__(self, obj, value):
if not self.fset:
raise AttributeError("can't set attribute")
type_ = type(obj)
return self.fset.__get__(obj, type_)(value)
def setter(self, func):
if not isinstance(func, (classmethod, staticmethod)):
func = classmethod(func)
self.fset = func
return self
def classproperty(func):
if not isinstance(func, (classmethod, staticmethod)):
func = classmethod(func)
return ClassPropertyDescriptor(func)
class Bar(object):
_bar = 1
#classproperty
def bar(cls):
return cls._bar
#bar.setter
def bar(cls, value):
cls._bar = value
# test instance instantiation
foo = Bar()
assert foo.bar == 1
baz = Bar()
assert baz.bar == 1
# test static variable
baz.bar = 5
assert foo.bar == 5
# test setting variable on the class
Bar.bar = 50
assert baz.bar == 50
assert foo.bar == 50
The setter didn't work at the time we call Bar.bar, because we are calling
TypeOfBar.bar.__set__, which is not Bar.bar.__set__.
Adding a metaclass definition solves this:
class ClassPropertyMetaClass(type):
def __setattr__(self, key, value):
if key in self.__dict__:
obj = self.__dict__.get(key)
if obj and type(obj) is ClassPropertyDescriptor:
return obj.__set__(self, value)
return super(ClassPropertyMetaClass, self).__setattr__(key, value)
# and update class define:
# class Bar(object):
# __metaclass__ = ClassPropertyMetaClass
# _bar = 1
# and update ClassPropertyDescriptor.__set__
# def __set__(self, obj, value):
# if not self.fset:
# raise AttributeError("can't set attribute")
# if inspect.isclass(obj):
# type_ = obj
# obj = None
# else:
# type_ = type(obj)
# return self.fset.__get__(obj, type_)(value)
Now all will be fine.

If you define classproperty as follows, then your example works exactly as you requested.
class classproperty(object):
def __init__(self, f):
self.f = f
def __get__(self, obj, owner):
return self.f(owner)
The caveat is that you can't use this for writable properties. While e.I = 20 will raise an AttributeError, Example.I = 20 will overwrite the property object itself.

[answer written based on python 3.4; the metaclass syntax differs in 2 but I think the technique will still work]
You can do this with a metaclass...mostly. Dappawit's almost works, but I think it has a flaw:
class MetaFoo(type):
#property
def thingy(cls):
return cls._thingy
class Foo(object, metaclass=MetaFoo):
_thingy = 23
This gets you a classproperty on Foo, but there's a problem...
print("Foo.thingy is {}".format(Foo.thingy))
# Foo.thingy is 23
# Yay, the classmethod-property is working as intended!
foo = Foo()
if hasattr(foo, "thingy"):
print("Foo().thingy is {}".format(foo.thingy))
else:
print("Foo instance has no attribute 'thingy'")
# Foo instance has no attribute 'thingy'
# Wha....?
What the hell is going on here? Why can't I reach the class property from an instance?
I was beating my head on this for quite a while before finding what I believe is the answer. Python #properties are a subset of descriptors, and, from the descriptor documentation (emphasis mine):
The default behavior for attribute access is to get, set, or delete the
attribute from an object’s dictionary. For instance, a.x has a lookup chain
starting with a.__dict__['x'], then type(a).__dict__['x'], and continuing
through the base classes of type(a) excluding metaclasses.
So the method resolution order doesn't include our class properties (or anything else defined in the metaclass). It is possible to make a subclass of the built-in property decorator that behaves differently, but (citation needed) I've gotten the impression googling that the developers had a good reason (which I do not understand) for doing it that way.
That doesn't mean we're out of luck; we can access the properties on the class itself just fine...and we can get the class from type(self) within the instance, which we can use to make #property dispatchers:
class Foo(object, metaclass=MetaFoo):
_thingy = 23
#property
def thingy(self):
return type(self).thingy
Now Foo().thingy works as intended for both the class and the instances! It will also continue to do the right thing if a derived class replaces its underlying _thingy (which is the use case that got me on this hunt originally).
This isn't 100% satisfying to me -- having to do setup in both the metaclass and object class feels like it violates the DRY principle. But the latter is just a one-line dispatcher; I'm mostly okay with it existing, and you could probably compact it down to a lambda or something if you really wanted.

If you use Django, it has a built in #classproperty decorator.
from django.utils.decorators import classproperty
For Django 4, use:
from django.utils.functional import classproperty

I think you may be able to do this with the metaclass. Since the metaclass can be like a class for the class (if that makes sense). I know you can assign a __call__() method to the metaclass to override calling the class, MyClass(). I wonder if using the property decorator on the metaclass operates similarly.
Wow, it works:
class MetaClass(type):
def getfoo(self):
return self._foo
foo = property(getfoo)
#property
def bar(self):
return self._bar
class MyClass(object):
__metaclass__ = MetaClass
_foo = 'abc'
_bar = 'def'
print MyClass.foo
print MyClass.bar
Note: This is in Python 2.7. Python 3+ uses a different technique to declare a metaclass. Use: class MyClass(metaclass=MetaClass):, remove __metaclass__, and the rest is the same.

As far as I can tell, there is no way to write a setter for a class property without creating a new metaclass.
I have found that the following method works. Define a metaclass with all of the class properties and setters you want. IE, I wanted a class with a title property with a setter. Here's what I wrote:
class TitleMeta(type):
#property
def title(self):
return getattr(self, '_title', 'Default Title')
#title.setter
def title(self, title):
self._title = title
# Do whatever else you want when the title is set...
Now make the actual class you want as normal, except have it use the metaclass you created above.
# Python 2 style:
class ClassWithTitle(object):
__metaclass__ = TitleMeta
# The rest of your class definition...
# Python 3 style:
class ClassWithTitle(object, metaclass = TitleMeta):
# Your class definition...
It's a bit weird to define this metaclass as we did above if we'll only ever use it on the single class. In that case, if you're using the Python 2 style, you can actually define the metaclass inside the class body. That way it's not defined in the module scope.

def _create_type(meta, name, attrs):
type_name = f'{name}Type'
type_attrs = {}
for k, v in attrs.items():
if type(v) is _ClassPropertyDescriptor:
type_attrs[k] = v
return type(type_name, (meta,), type_attrs)
class ClassPropertyType(type):
def __new__(meta, name, bases, attrs):
Type = _create_type(meta, name, attrs)
cls = super().__new__(meta, name, bases, attrs)
cls.__class__ = Type
return cls
class _ClassPropertyDescriptor(object):
def __init__(self, fget, fset=None):
self.fget = fget
self.fset = fset
def __get__(self, obj, owner):
if self in obj.__dict__.values():
return self.fget(obj)
return self.fget(owner)
def __set__(self, obj, value):
if not self.fset:
raise AttributeError("can't set attribute")
return self.fset(obj, value)
def setter(self, func):
self.fset = func
return self
def classproperty(func):
return _ClassPropertyDescriptor(func)
class Bar(metaclass=ClassPropertyType):
__bar = 1
#classproperty
def bar(cls):
return cls.__bar
#bar.setter
def bar(cls, value):
cls.__bar = value
bar = Bar()
assert Bar.bar==1
Bar.bar=2
assert bar.bar==2
nbar = Bar()
assert nbar.bar==2

I happened to come up with a solution very similar to #Andrew, only DRY
class MetaFoo(type):
def __new__(mc1, name, bases, nmspc):
nmspc.update({'thingy': MetaFoo.thingy})
return super(MetaFoo, mc1).__new__(mc1, name, bases, nmspc)
#property
def thingy(cls):
if not inspect.isclass(cls):
cls = type(cls)
return cls._thingy
#thingy.setter
def thingy(cls, value):
if not inspect.isclass(cls):
cls = type(cls)
cls._thingy = value
class Foo(metaclass=MetaFoo):
_thingy = 23
class Bar(Foo)
_thingy = 12
This has the best of all answers:
The "metaproperty" is added to the class, so that it will still be a property of the instance
Don't need to redefine thingy in any of the classes
The property works as a "class property" in for both instance and class
You have the flexibility to customize how _thingy is inherited
In my case, I actually customized _thingy to be different for every child, without defining it in each class (and without a default value) by:
def __new__(mc1, name, bases, nmspc):
nmspc.update({'thingy': MetaFoo.services, '_thingy': None})
return super(MetaFoo, mc1).__new__(mc1, name, bases, nmspc)

If you only need lazy loading, then you could just have a class initialisation method.
EXAMPLE_SET = False
class Example(object):
#classmethod
def initclass(cls):
global EXAMPLE_SET
if EXAMPLE_SET: return
cls.the_I = 'ok'
EXAMPLE_SET = True
def __init__( self ):
Example.initclass()
self.an_i = 20
try:
print Example.the_I
except AttributeError:
print 'ok class not "loaded"'
foo = Example()
print foo.the_I
print Example.the_I
But the metaclass approach seems cleaner, and with more predictable behavior.
Perhaps what you're looking for is the Singleton design pattern. There's a nice SO QA about implementing shared state in Python.

Overwrite several methods of a class with a loop

I would like to know if there is an easy way to do some identical edits on several methods of a class. An example :
class Dog():
def __init__(self):
self.name = 'abc'
self.age = 1
def setName(self, newValue):
self.name = newValue
def setAge(self, newValue):
self.age = newValue
class TalkingDog(Dog):
def __init__(self):
super().__init__()
# The end is in pseudo code :
for method in TalkingDog.allMethods :
method = method + 'print('I have been edited !')'
I know that I can also overwrite each method but in a situation with tens of methods, that will be a little boring...
So I tried this :
class TalkingDog(Dog):
def __init__(self):
super().__init__()
for method in self.__dir__():
if method.startswith('set'):
oldMethod = getattr(self, method)
def _newMethod(newValue):
oldMethod(newValue)
print('I have been edited !')
setattr(self, method, _newMethod)
a = TalkingDog()
print(a.setName) >>> <function TalkingDog.__init__.<locals>._newMethod at 0x0000000002C350D0>
That almost works but setName is not anymore a method. It's an attribute which contains a function. I completely understand why but I'm trying to get a cleaner result. With that result, I risk of having problems later. For example I can't use the library pickle with that object (got the error _pickle.PicklingError: Can't pickle <function TalkingDog.__init__.<locals>._newMethod at 0x00000000003DCBF8>: attribute lookup _newMethod on __main__ failed).

The Pythonic way to do this is probably to use the descriptor protocol, which is also what properties use:
class VocalAttribute:
def __init__(self, name, feedback):
"""Called when you first create the descriptor."""
self.name = name # the name of the attribute 'behind' the property
self.feedback = feedback # the feedback to show when the value changes
def __get__(self, obj):
"""Called when you get the descriptor value."""
return getattr(obj, self.name)
def __set__(self, obj, value):
"""Called when you set the descriptor value."""
prev = getattr(obj, self.name, None)
if value != prev:
setattr(obj, self.name, value)
print(self.feedback)
def __delete__(self, obj):
"""Called when you delete the descriptor value."""
delattr(obj, self.name)
class Foo:
bar = VocalAttribute('_bar', 'I have been edited!')
foo = Foo()
print('1.')
foo.bar = 'hello'
print('2.')
foo.bar = 'hello'
print('3.')
foo.bar = 'world'
Output:
1.
I have been edited!
2.
3.
I have been edited!
Note that this only gives feedback when the new value is different to the old one - you can tweak the behaviour as needed in __set__. It also means you can directly read from and assign to foo.bar, rather than needing to call getters and setters (what is this, Java?)

since decorator could explicit called here a way to use it:
def updater(obj, call_back, call_back_args=(), call_back_kw=None, replace=False):
# ability to be called on the fly with different args and kw for the callback
# now it returns the updated obj (instance or class)
# but could a be factory returning a new obj in this case make a copy of obj, update this coy and return it
def update_function(fn, *args, **kw):
def wrapper(*args, **kw):
if replace:
# call only the callback
res = call_back(*call_back_args, **call_back_kw)
else:
res = fn(*args, **kw)
call_back(*call_back_args, **call_back_kw)
return res
return wrapper
# get all methods of the obj
# and apply update_function (a decorator) to all methods
for name, m in inspect.getmembers(
obj, predicate=lambda x: inspect.isfunction(x) or inspect.ismethod(x)):
# make the selection here
# could be made on the name for instance
if not name.startswith('_'):
new_m = update_function(m)
setattr(obj, name, new_m)
return obj
# declare a callback
def call_back(*args, **kw):
# simple callback
print("I have been edited and called with %r args and %r kw " % (args, kw))
a = Dog()
# could be called on instance or class
# apply the callback on all "public" methods
updater(
a,
call_back,
call_back_args=(2, 3, 4),
call_back_kw={"kw1": "v_1"}
)

Finding a function's parameters in Python

I want to be able to ask a class's __init__ method what it's parameters are. The straightforward approach is the following:
cls.__init__.__func__.__code__.co_varnames[:code.co_argcount]
However, that won't work if the class has any decorators. It will give the parameter list for the function returned by the decorator. I want to get down to the original __init__ method and get those original parameters. In the case of a decorator, the decorator function is going to be found in the closure of the function returned by the decorator:
cls.__init__.__func__.__closure__[0]
However, it is more complicated if there are other things in the closure, which decorators may do from time to time:
def Something(test):
def decorator(func):
def newfunc(self):
stuff = test
return func(self)
return newfunc
return decorator
def test():
class Test(object):
#Something(4)
def something(self):
print Test
return Test
test().something.__func__.__closure__
(<cell at 0xb7ce7584: int object at 0x81b208c>, <cell at 0xb7ce7614: function object at 0xb7ce6994>)
And then I have to decide if I want to the parameters from decorator or the parameters from the original function. The function returned by the decorator could have *args and **kwargs for its parameters. What if there are multiple decorators and I have to decide which is the one I care about?
So what is the best way to find a function's parameters even when the function may be decorated? Also, what is the best way to go down a chain of decorators back to the decorated function?
Update:
Here is effectively how I am doing this right now (names have been changed to protect the identity of the accused):
import abc
import collections
IGNORED_PARAMS = ("self",)
DEFAULT_PARAM_MAPPING = {}
DEFAULT_DEFAULT_PARAMS = {}
class DICT_MAPPING_Placeholder(object):
def __get__(self, obj, type):
DICT_MAPPING = {}
for key in type.PARAMS:
DICT_MAPPING[key] = None
for cls in type.mro():
if "__init__" in cls.__dict__:
cls.DICT_MAPPING = DICT_MAPPING
break
return DICT_MAPPING
class PARAM_MAPPING_Placeholder(object):
def __get__(self, obj, type):
for cls in type.mro():
if "__init__" in cls.__dict__:
cls.PARAM_MAPPING = DEFAULT_PARAM_MAPPING
break
return DEFAULT_PARAM_MAPPING
class DEFAULT_PARAMS_Placeholder(object):
def __get__(self, obj, type):
for cls in type.mro():
if "__init__" in cls.__dict__:
cls.DEFAULT_PARAMS = DEFAULT_DEFAULT_PARAMS
break
return DEFAULT_DEFAULT_PARAMS
class PARAMS_Placeholder(object):
def __get__(self, obj, type):
func = type.__init__.__func__
# unwrap decorators here
code = func.__code__
keys = list(code.co_varnames[:code.co_argcount])
for name in IGNORED_PARAMS:
try: keys.remove(name)
except ValueError: pass
for cls in type.mro():
if "__init__" in cls.__dict__:
cls.PARAMS = tuple(keys)
break
return tuple(keys)
class BaseMeta(abc.ABCMeta):
def __init__(self, name, bases, dict):
super(BaseMeta, self).__init__(name, bases, dict)
if "__init__" not in dict:
return
if "PARAMS" not in dict:
self.PARAMS = PARAMS_Placeholder()
if "DEFAULT_PARAMS" not in dict:
self.DEFAULT_PARAMS = DEFAULT_PARAMS_Placeholder()
if "PARAM_MAPPING" not in dict:
self.PARAM_MAPPING = PARAM_MAPPING_Placeholder()
if "DICT_MAPPING" not in dict:
self.DICT_MAPPING = DICT_MAPPING_Placeholder()
class Base(collections.Mapping):
__metaclass__ = BaseMeta
"""
Dict-like class that uses its __init__ params for default keys.
Override PARAMS, DEFAULT_PARAMS, PARAM_MAPPING, and DICT_MAPPING
in the subclass definition to give non-default behavior.
"""
def __init__(self):
pass
def __nonzero__(self):
"""Handle bool casting instead of __len__."""
return True
def __getitem__(self, key):
action = self.DICT_MAPPING[key]
if action is None:
return getattr(self, key)
try:
return action(self)
except AttributeError:
return getattr(self, action)
def __iter__(self):
return iter(self.DICT_MAPPING)
def __len__(self):
return len(self.DICT_MAPPING)
print Base.PARAMS
# ()
print dict(Base())
# {}
At this point Base reports uninteresting values for the four contants and the dict version of instances is empty. However, if you subclass you can override any of the four, or you can include other parameters to the __init__:
class Sub1(Base):
def __init__(self, one, two):
super(Sub1, self).__init__()
self.one = one
self.two = two
Sub1.PARAMS
# ("one", "two")
dict(Sub1(1,2))
# {"one": 1, "two": 2}
class Sub2(Base):
PARAMS = ("first", "second")
def __init__(self, one, two):
super(Sub2, self).__init__()
self.first = one
self.second = two
Sub2.PARAMS
# ("first", "second")
dict(Sub2(1,2))
# {"first": 1, "second": 2}

Consider this decorator:
def rickroll(old_function):
return lambda junk, junk1, junk2: "Never Going To Give You Up"
class Foo(object):
#rickroll
def bar(self, p1, p2):
return p1 * p2
print Foo().bar(1, 2)
In it, the rickroll decorator takes the bar method, discards it, replaces it with a new function that ignores its differently-named (and possibly numbered!) parameters and instead returns a line from a classic song.
There are no further references to the original function, and the garbage collector can come and remove it any time it likes.
In such a case, I cannot see how you could find the parameter names p1 and p2. In my understanding, even the Python interpreter itself has no idea what they used to be called.