Python __index__ 和评估顺序

Question

I was expecting the two code below to return the same result because of https://docs.python.org/3.7/reference/expressions.html#evaluation-order . However it looks like __index__ from Counter class is called after __call__ in the second case. Is it because __index__ gets called only when reaching ] ? Or is there another explanation ?

class Counter:
    def __init__(self, start=0):
        self._i = start

    def __call__(self, step:int=None):
        if step is not None:
            self._i += step
        return self._i

    def __index__(self):
        return self._i

data = list(range(0, 10))

i = Counter(0)
data[i():i(3)]

returns [0, 1, 2]

i = Counter(0)
data[i:i(3)]

returns []

PS: The goal of the Counter class is to allow assignment like syntax in python < 3.8

i = 0
data[i: (i := i+3)]

Answer 1

data[i:i(3)] is equivalent to data.__getitem__(slice(i, i(3))) ; i.__index__ doesn't get called until the starting element of the slice is actually needed, inside the body of data.__getitem__ , but after i(3) has already incremented i._step to 3. So the order of evaluation is something like:

1. data[i:i(3)]
2. Evaluate i(3) to 3
3. Build the slice slice(i,3)
4. Inside data.__getitem__ , the integer value of i is needed, so i.__index__ is called, which returns 3.

You can see this by disassembling the indexing operation:

>>> import dis
>>> dis.dis('data[i:i(3)]')
  1           0 LOAD_NAME                0 (data)
              2 LOAD_NAME                1 (i)
              4 LOAD_NAME                1 (i)
              6 LOAD_CONST               0 (3)
              8 CALL_FUNCTION            1
             10 BUILD_SLICE              2
             12 BINARY_SUBSCR
             14 RETURN_VALUE

First i(3) is called (at offset 8), then the slice is built (at offset 10), and finally data.__getitem__ is called (at offset 12).