Python is vs ==: identity vs equality

a == b asks whether two objects have the same value (it calls __eq__); a is b asks whether they are the same object in memory (id(a) == id(b)). Use == for comparing data, and is only for singletons: x is None, x is not None, and sentinel objects.

The difference in practice

cat > cmp.py <<'EOF'
a = [1, 2, 3]
b = [1, 2, 3]
c = a
print("a == b:", a == b)    # same value
print("a is b:", a is b)    # different objects
print("a is c:", a is c)    # c is another name for a
x = None
print("x is None:", x is None)
print(int("256") is int("256"), int("257") is int("257"))
EOF
python cmp.py

a == b: True
a is b: False
a is c: True
x is None: True
True False

The last line shows a CPython implementation detail: integers from -5 to 256 are cached, so two of them are the same object, while larger ones created at runtime usually are not. Code must never rely on that.

When == and is disagree

Because == runs code from the object's class, the two operators can give opposite answers even for None, and floating-point NaN is not equal to itself:

cat > eq.py <<'EOF'
class Always:
    def __eq__(self, other):
        return True

obj = Always()
print("obj == None:", obj == None)
print("obj is None:", obj is None)

nan = float("nan")
print("nan == nan:", nan == nan, " nan is nan:", nan is nan)
print("nan in [nan]:", nan in [nan])

built = "".join(["err", "codex"])
literal = "errcodex"
print("built == literal:", built == literal, " built is literal:", built is literal)
EOF
python eq.py

obj == None: True
obj is None: False
nan == nan: False  nan is nan: True
nan in [nan]: True
built == literal: True  built is literal: False

• obj == None is True because Always.__eq__ says so, which is exactly why PEP 8 asks for is None.
• nan == nan is False by the IEEE 754 rules, yet nan in [nan] is True: container membership tests (in, list.index, list.count) check identity first and only then call ==.
• A string built at runtime equals the literal but is a different object, so is fails where == works.

Criterion by criterion

Python warns about is with a literal

python -c 'x = 1000; print(x is 1000)'

<string>:1: SyntaxWarning: "is" with 'int' literal. Did you mean "=="?
True

Since Python 3.8, is against a literal emits a SyntaxWarning that suggests ==.

Which one to use

Use == (and !=) whenever you care about the data: numbers, strings, lists, dicts, dataclasses, anything a user typed or a file contained. Use is (and is not) only when you need to know that a name refers to one particular object: None, NotImplemented, Ellipsis, enum members, and private sentinel objects such as _missing = object() used as a default argument. A sentinel works with is precisely because nothing else can be identical to it, even an object that compares equal.

A missing is None check is the usual cause of AttributeError: 'NoneType' object has no attribute: test for None before using a result that may be absent.

FAQ