Why is "if not someobj:" better than "if someobj == None:" in Python?

In the first test, Python try to convert the object to a bool value if it is not already one. Roughly, we are asking the object : are you meaningful or not ? This is done using the following algorithm :

  1. If the object has a __nonzero__ special method (as do numeric built-ins, int and float), it calls this method. It must either return a bool value which is then directly used, or an int value that is considered False if equal to zero.

  2. Otherwise, if the object has a __len__ special method (as do container built-ins, list, dict, set, tuple, ...), it calls this method, considering a container False if it is empty (length is zero).

  3. Otherwise, the object is considered True unless it is None in which case, it is considered False.

In the second test, the object is compared for equality to None. Here, we are asking the object, "Are you equal to this other value?" This is done using the following algorithm :

  1. If the object has a __eq__ method, it is called, and the return value is then converted to a boolvalue and used to determine the outcome of the if.

  2. Otherwise, if the object has a __cmp__ method, it is called. This function must return an int indicating the order of the two object (-1 if self < other, 0 if self == other, +1 if self > other).

  3. Otherwise, the object are compared for identity (ie. they are reference to the same object, as can be tested by the is operator).

There is another test possible using the is operator. We would be asking the object, "Are you this particular object?"

Generally, I would recommend to use the first test with non-numerical values, to use the test for equality when you want to compare objects of the same nature (two strings, two numbers, ...) and to check for identity only when using sentinel values (None meaning not initialized for a member field for exemple, or when using the getattr or the __getitem__ methods).

To summarize, we have :

>>> class A(object):
...    def __repr__(self):
...        return 'A()'
...    def __nonzero__(self):
...        return False

>>> class B(object):
...    def __repr__(self):
...        return 'B()'
...    def __len__(self):
...        return 0

>>> class C(object):
...    def __repr__(self):
...        return 'C()'
...    def __cmp__(self, other):
...        return 0

>>> class D(object):
...    def __repr__(self):
...        return 'D()'
...    def __eq__(self, other):
...        return True

>>> for obj in ['', (), [], {}, 0, 0., A(), B(), C(), D(), None]:
...     print '%4s: bool(obj) -> %5s, obj == None -> %5s, obj is None -> %5s' % \
...         (repr(obj), bool(obj), obj == None, obj is None)
  '': bool(obj) -> False, obj == None -> False, obj is None -> False
  (): bool(obj) -> False, obj == None -> False, obj is None -> False
  []: bool(obj) -> False, obj == None -> False, obj is None -> False
  {}: bool(obj) -> False, obj == None -> False, obj is None -> False
   0: bool(obj) -> False, obj == None -> False, obj is None -> False
 0.0: bool(obj) -> False, obj == None -> False, obj is None -> False
 A(): bool(obj) -> False, obj == None -> False, obj is None -> False
 B(): bool(obj) -> False, obj == None -> False, obj is None -> False
 C(): bool(obj) ->  True, obj == None ->  True, obj is None -> False
 D(): bool(obj) ->  True, obj == None ->  True, obj is None -> False
None: bool(obj) -> False, obj == None ->  True, obj is None ->  True

PEP 8 -- Style Guide for Python Code recommends to use is or is not if you are testing for None-ness

- Comparisons to singletons like None should always be done with
  'is' or 'is not', never the equality operators.

On the other hand if you are testing for more than None-ness, you should use the boolean operator.


Because None is not the only thing that is considered false.

if not False:
    print "False is false."
if not 0:
    print "0 is false."
if not []:
    print "An empty list is false."
if not ():
    print "An empty tuple is false."
if not {}:
    print "An empty dict is false."
if not "":
    print "An empty string is false."

False, 0, (), [], {} and "" are all different from None, so your two code snippets are not equivalent.

Moreover, consider the following:

>>> False == 0
True
>>> False == ()
False

if object: is not an equality check. 0, (), [], None, {}, etc. are all different from each other, but they all evaluate to False.

This is the "magic" behind short circuiting expressions like:

foo = bar and spam or eggs

which is shorthand for:

if bar:
    foo = spam
else:
    foo = eggs

although you really should write:

foo = spam if bar else egg

These are actually both poor practices. Once upon a time, it was considered OK to casually treat None and False as similar. However, since Python 2.2 this is not the best policy.

First, when you do an if x or if not x kind of test, Python has to implicitly convert x to boolean. The rules for the bool function describe a raft of things which are False; everything else is True. If the value of x wasn't properly boolean to begin with, this implicit conversion isn't really the clearest way to say things.

Before Python 2.2, there was no bool function, so it was even less clear.

Second, you shouldn't really test with == None. You should use is None and is not None.

See PEP 8, Style Guide for Python Code. 

- Comparisons to singletons like None should always be done with
  'is' or 'is not', never the equality operators.

  Also, beware of writing "if x" when you really mean "if x is not None"
  -- e.g. when testing whether a variable or argument that defaults to
  None was set to some other value.  The other value might have a type
  (such as a container) that could be false in a boolean context!

How many singletons are there? Five: None, True, False, NotImplemented and Ellipsis. Since you're really unlikely to use NotImplemented or Ellipsis, and you would never say if x is True (because simply if x is a lot clearer), you'll only ever test None.

Tags:

Python

Recent Posts