When should I use list.count(0), and how do I to discount the "False" item?

Question

a.count(0) always returns 11, so what should I do to discount the False and return 10?

a = ["a",0,0,"b",None,"c","d",0,1,False,0,1,0,3,[],0,1,9,0,0,{},0,0,9]

Answer 1

Python 2.x interprets False as 0 and vice versa. AFAIK even None and "" can be considered False in conditions. Redefine count as follows:

sum(1 for item in a if item == 0 and type(item) == int)

or (Thanks to Kevin, and Bakuriu for their comments):

sum(1 for item in a if item == 0 and type(item) is type(0))

or as suggested by ozgur in comments (which is not recommended and is considered wrong, see this), simply:

sum(1 for item in a if item is 0)  

it may (“is” operator behaves unexpectedly with integers) work for small primary types, but if your list contains objects, please consider what is operator does:

From the documentation for the is operator:

The operators is and is not test for object identity: x is y is true if and only if x and y are the same object.

More information about is operator: Understanding Python's "is" operator

Answer 2

This can be done with sum and a generator expression that compares each element to False:

>>> a = ["a",0,0,"b",None,"c","d",0,1,False,0,1,0,3,[],0,1,9,0,0,{},0,0,9]
>>> sum((x == 0 and x is not False) for x in a)
10

This works because bool is a subclass of int - False equals 0 and True equals 1.

Answer 3

You need to filter out the Falses yourself.

>>> a = ["a",0,0,"b",None,"c","d",0,1,False,0,1,0,3,[],0,1,9,0,0,{},0,0,9]
>>> len([x for x in a if x == 0 and x is not False])
10

---

Old answer is CPython specific and it's better to use solutions that work on all Python implementations.

Since CPython keeps a pool of small integer objects, zero included, you can filter the list with the is operator.

This of course shouldn't be used for value comparisons but in this case it works since zeros are what we want to find.

>>> a = ["a",0,0,"b",None,"c","d",0,1,False,0,1,0,3,[],0,1,9,0,0,{},0,0,9]
>>> [x for x in a if x is 0]
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
>>> len(_)
10

Answer 4

How about this functional style solution:

print zip(a, map(type, a)).count((0, int))
>>> 10

Timing this against some of the other answers here, this also seems to be one of the quickest:

t0 = time.time()
for i in range(100000):
    zip(a, map(type, a)).count((0, int))
print time.time() - t0
>>> 0.275855064392

t0 = time.time()
for i in range(100000):
    sum(1 for item in a if item == 0 and type(item) == int)
print time.time() - t0
>>> 0.478030204773

t0 = time.time()
for i in range(100000):
    sum(1 for item in a if item == 0 and type(item) is type(0))
print time.time() - t0
>>> 0.52236700058

t0 = time.time()
for i in range(100000):
    sum((x==0 and x is not False) for x in a)
print time.time() - t0
>>> 0.450266122818

Answer 5

Solving this problem more generally, you could make your own subclass of list:

class key_equality_list(list):
    def __init__(self, key, items):
        super(key_equality_list, self).__init__(items)
        self._f = key

    def __contains__(self, x):
        return any(self._f(x) == self._f(item) for item in self)

    def count(self, x):
        return sum(self._f(x) == self._f(item) for item in self)

And then define a key function that checks types:

type_and_val = lambda x: (x, type(x))

So your usage becomes:

>>> a = ["a",0,0,"b",None,"c","d",0,1,False,0,1,0,3,[],0,1,9,0,0,{},0,0,9]
>>> a = key_equality_list(type_and_val, a)
>>> a.count(0)
10
>>> a.count(False)
1

Answer 6

Since count returns the number of items equal to its input, and since 0 is equal to False in Python, one way to approach this is to pass in something that is equal to 0 but is not equal to False.

Python itself provides no such object, but we can write one:

class Zero:
    def __eq__(self, other):
        return other == 0 and other is not False

print([1, 2, 0, False].count(Zero()))

Be aware that there are other things in Python that, like False, are equal to 0. I can think of 0.0, decimal.Decimal(0), and fractions.Fraction(0,1). Furthermore, anybody can subclass int to create other "fake 0 values", just like False. The code above counts all of these, treating False as the only special case, but you might want to do something different in your __eq__.

Of course "under the covers" this is very similar to just doing sum(other == 0 and other is not False for other in a). But considering how count() is defined, it should come as no surprise that a.count(x) is similar to sum(x == y for y in a).

You would only do this in the rare case where you have for some reason already decided to use count (perhaps because you're using some third-party code that's going to call it), and need to come up with a suitable object to pass it. Basically, it treats count as a function that calls an arbitrary function we write (requiring only that this function be the __eq__ function of the argument) and counts the number of true return values. This is an accurate description of what count does, but not really the point of count since sum is more explicit.