Re: an algorithm with interruption in the middle
Seungbeom Kim wrote:
Andrei Alexandrescu (See Website For Email) wrote:
The scope rules remain the same for variables defined within
expressions: from the point of definition to the end of the current
scope. Expressions don't introduce scope, so the issues remain
orthogonal.
Oh, I see. But that's inconsistent with the declarations within
conditions, isn't it?
while (int x = pop()) { /* ... */ }
// The scope of x ends here
foo(int x = pop());
// The scope of x continues
I don't find it inconsistent more than the visuals that in one case a
name escapes a paren, and in the other a name does not. Other than that,
function call is an expression, compared to while, for, if etc., which
are (compound) statements, so there is not much consistency required in
the first place.
Variables defined in the control part of an if, while, switch, for, last
for the duration of these respective statements (as it happens now,
unfortunately with annoyingly crippled limitations - yes, you can define
a variable, it can only be compared against zero).
Hence my proposal. :)
(Yes, I know, I was not very serious, but who knows?)
Defining a variable on the right-hand side of a built-in operator|| or
operator&& is not allowed. Hardly a guess why :o).
How would that mix with overloaded operator&& and operator||?
Should declarations in the right-hand side operands be banned even if
both operands are guaranteed to be evaluated.. just for consistency?
Who cares? After we're done with this thread, we'll all forget all about
it as promptly as possible. :o)
And what about operator?:
cond ? (int x = f()) : (int y = g())
now, what's in scope and what's not?
The sound thing to do is to disallow variable definition in ?:, just as
with || and &&. I forgot about it.
Declarations and scopes are static concepts, so mixing them with runtime
behaviour seems too confusing.
I think this is an exaggeration. There is no mixing going on. We have
runtime stuff and we have lexical stuff. The two are governed by
different rules.
Besides, I use the feature in Perl all the time and there's not once I
got confused.
My proposal separates the declarations and the expressions used as
conditions, so it doesn't produce the subtle problems mentioned above.
(Yes, it may seem too radical... but what wasn't, in the early days?)
It's not that it's radical that I mind.
And I feel that most of cases where I needed declarations buried in
expressions have been in conditional expressions with if, while, etc.
Otherwise, you can just use a simple-declaration, without introducing
any change of the scope.
Can you give an example of the "many other instances"?
cin >> float f;
scanf("%f", &(float f));
Trim(cin >> string line);
factory.CreateInstance(&(Object * p = 0));
Object * p = new Object[size_t newLength = length * 2];
Andrei
--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]