(test) ? a:b

[Alain Ketterlin]

Marko Rauhamaa <marko at pacujo.net> writes:

> "BartC" <bc at freeuk.com>:

>>> x = [f, g][cond]()

>> It will select f or g (which should refer to functions), and call one of
>> those depending on cond. That's not a problem.
>>
>> The problem is it will still evaluate both f and g,
>
> That's not really the problem. The problem is in readability.

Readability is in the eye of the beholder. The evaluation is the central
concern here.

> However, the "[f, g][cond]()" technique is how pure lambda calculus
> implements conditional branching so it is interesting in its own
> right.

Yes, but lambda calculus constructs have no meaning without an
evaluation strategy ("reduction" strategy in lambda calculs parlance).

> IOW, you can do "short-circuiting" in purely functional programming:
>
>     j = j + 1 if j < 10 else 3
>     <=>
>     j = (lambda: 3, lambda: j + 1)[j < 10]()

This is just a way to delay evaluation *of the potential results*, i.e.,
instill a bit of lazyness. One could object that, again, both
expressions in the tuple have been evaluated (building two lambdas),
only one of which is finally called. I guess that's what BartC meant.

-- Alain.