[Python] Why does more_itertools.windowed(xs, 0) yield a range of 1 empty tuple?

In the more_itertools library, we have the function windowed, which has this behavior:

>>> from more_itertools import windowed
>>> xs = [1, 2, 3, 4]
>>> list(windowed(xs, 3))
[(1, 2, 3), (2, 3, 4)]
>>> list(windowed(xs, 2))
[(1, 2), (2, 3), (3, 4)]
>>> list(windowed(xs, 1))
[(1,), (2,), (3,), (4,)]
>>> list(windowed(xs, 0))
[()]


I can understand the argument for windowed(xs, 0) being an empty range (consistent with zip() being empty) and there is a separate argument for windowed(xs, 0) being an infinite range (which would be consistent with zip() being an infinite range, which would be the mathematically correct identity element — Julia does this, for instance).

But why does windowed(xs, 0) specifically yield a range consisting of one empty tuple? In the source code, this is handled as a special case, it's not like it falls out from the implementation. What is the motivation for specifically just yield() here:

def windowed(seq, n, fillvalue=None, step=1):
# ... snipping docstring ...

if n < 0:
raise ValueError('n must be >= 0')
if n == 0:
yield ()
return
if step < 1:
raise ValueError('step must be >= 1')

# ... rest of implementation ...

Continue reading...