Does Python evaluate if's conditions lazily?

Yes, Python evaluates boolean conditions lazily.

The docs say,

The expression x and y first evaluates x; if x is false, its value is returned; otherwise, y is evaluated and the resulting value is returned.

The expression x or y first evaluates x; if x is true, its value is returned; otherwise, y is evaluated and the resulting value is returned.

Yes, Python evaluates lazily, so foo2 will not be checked.

I use this all the time for grabbing items from dictionary-like objects if I don't know if the key exists:

if 'key' in mydict and mydict['key'] == 'heyyo!':
do_stuff()

See @unutbu's answer for a fuller explanation.

This isn't technically lazy evaluation, it's short-circuit boolean expressions.

Lazy evaluation has a somewhat different connotation. For example, true lazy evaluation would likely allow this

def foo(arg) :
print "Couldn't care less"


foo([][0])

But Python doesn't.

Python is also nice in that it "echos" it's boolean arguments. For example, an or condition returns either it's first "truthy" argument or the last argument (if all arguments are "falsey"). An and condition does the inverse.

So "echo argument" booleans means

2 and [] and 1

evaluates to [], and

[] or 1 or 2

evaluates to 1

It is really the or part that is short circuited:

>>> 1 or 1/0  #also 0 and 1/0
1
>>> 0 or 1/0  #also 1 and 1/0


Traceback (most recent call last):
File "<pyshell#1240>", line 1, in <module>
0 or 1/0
ZeroDivisionError: integer division or modulo by zero

Python's laziness can be proved by the following code:

def foo():
print('foo')
return False


def bar():
print('bar')
return False


foo() and bar()         #Only 'foo' is printed

On the other hand,

foo() or bar()

would cause both 'foo' and 'bar' to be printed.

and or is lazy

& | is not lazy

A short demo would be to compare the time difference between

all(xrange(1,1000000000))

and

any(xrange(1,1000000000))

The all() has to check every single value, whilst the any() can give up after the first True has been found. The xrange, being a generator, therefore also gives up generating things as soon as the evaluator is done. For this reason, the all will consume large amounts of RAM and take ages, whilst the any will use just a few bytes and return instantaneously.