Slicing elements from a Python list using Boolean indexing

Question

I recently came across this way of indexing a list in Python. I've never seen this one before, so I would like to understand this clearly.

I have a list ["Peter", "James", "Mark"] and if I index it using the boolean value False it returns Peter and if I index it using True it returns James, as given below

• ["Peter", "James", "Mark"][False] => Peter
• ["Peter", "James", "Mark"][True] => James

I would like to know what happens here and what is this method called as?

Answer 1

The datamodel hook here is the __index__ magic method:

>>> True.__index__()
1
>>> False.__index__()
0

The value returned by on obj's __index__ is used when accessing with subscription, allowing arbitrary objects to be used for indexing and slicing:

x[obj]

This is somewhat independent of the fact that bool is a subclass of int! You may achieve the same with any object.

>>> class A:
...     def __index__(self):
...         return 1
...     
>>> 'ab'[A()]
'b'

Whether __index__ is resolved for int subclasses depends on implementation detail.

CPython 3.7.1:

>>> class MyInt(int):
...     def __index__(self):
...         return 1
... 
>>> '01'[MyInt(0)]
'0'

PyPy 5.0.1:

>>>> class MyInt(int):
....     def __index__(self):
....         return 1
....         
>>>> '01'[MyInt(0)]
'1'

PyPy behaves correctly according to the Python datamodel. Looks like CPython is taking a shortcut / performance optimization.