conditional expressions

[Alex Martelli]

Pedro RODRIGUEZ wrote:
        ...
>>> But why use grotesque hacks in the first place?  "a ? b : c" is a
        ...
> Ada has two short-circuits operators : "and then" and "or else" that
> do the trick :
>     if A and then B then
>         ...
>     end if;
> 
> Looks cleaner to me.

Ada's "and then" is like Python's "and"; Ada;s "or else" is like
Python's "or".  Neither is like C's "a?b:c" (ternary operator),
though with some tricks boolean shortcircuiting operators can be
used to simulate ternary operators (a bit cumbersome in the
general case).

The reason
        zz = a and b or c
is not the same thing as
        zz = a ? b : c
is that, when a is true and b is false, the former returns c,
while the latter returns b.

The tricks deal with ways to work around this problem, by
ensuring that and's RHO can never be false, e.g.
        zz = (a and [b] or [c])[0]


The other traditional emulation of ternary operator is by
indexing, e.g.:
        zz = (b, c)[not a]
and here the defect is that this isn't short-circuiting --
both the b and c sub-expressions are evaluated anyway, no
matter what's the value of a (this is a problem when,
e.g., b is actually a function call such as "fb()").

Here, the workarounds can use and/or if it's OK to
evaluate _a_ multiple times:
        zz = (a and b, not a and c)[not a]
or lambdas:
        zz = (lambda:b, lambda:c)[not a]()


The deeper problem with all of these "solutions" is that
they're unholy kludges, of course.


Alex