Hi,

I recently noticed that my compiler (VC++7.1) allows me to declare
variables inside the expression for an if-statement, like so:

if (int i = 0)
{
i = 1;
}

Now I was wondering if this is valid code (reference to the
appropriate section/paragraph in the Standard would be nice, as I do not
really know where to look for this kind of thing) and where else I am
able to declare variables (except the obvious "for (int i ...)" case).

For the record, I am asking this out of interest, not because I want
to start declaring my variables in all kinds of funny places :)

Thanks in advance!
--
jb

(reply address in rot13, unscramble first)

Jakob Bieling wrote:[color=blue]
> Hi,
>
> I recently noticed that my compiler (VC++7.1) allows me to declare
> variables inside the expression for an if-statement, like so:
>
> if (int i = 0)
> {
> i = 1;
> }[/color]

if ((int x = calcSomething()) == y)
{
doSomethingWith(x);
}

.... code not accessing x.

You could even do something like this:

int x = 5;

if ((int x = calcSomething()) == y)
{
doSomethingWithCalced(x);
}

assert(x == 5);

.... This code runs...;

Standard defines 'x' to have scope only within the following block of
code and it overrides the upper-scope definitions.
[color=blue]
>
> reference to the
> appropriate section/paragraph in the Standard would be nice, as I do not
> really know where to look for this kind of thing[/color]

3.3.2 paragraph 4 is one such reference but only explains scope.

Jakob Bieling wrote:[color=blue]
> Hi,
>
> I recently noticed that my compiler (VC++7.1) allows me to declare
> variables inside the expression for an if-statement, like so:
>
> if (int i = 0)
> {
> i = 1;
> }
>
> Now I was wondering if this is valid code (reference to the
> appropriate section/paragraph in the Standard would be nice, as I do not
> really know where to look for this kind of thing) and where else I am
> able to declare variables (except the obvious "for (int i ...)" case).
>
> For the record, I am asking this out of interest, not because I want
> to start declaring my variables in all kinds of funny places :)
>
> Thanks in advance!
> --
> jb
>
> (reply address in rot13, unscramble first)[/color]

That is certainly legal. I don't have the standard handy, but see
Stroustrup _TC++PL_ 3rd ed. on dynamic_cast. He does something like
this, which is where the construction is most useful:

if( Derived* pd = dynamic_cast<Derived*>( pBase ) )
{
// pd is valid; do something with it here
}
else
{
// pd is 0 since the dynamic_cast failed;
// do something else here, e.g.
pd = new Derived;
// ...
}
// pd is now out of scope

Cheers! --M

Jakob Bieling wrote:[color=blue]
> Hi,
>
> I recently noticed that my compiler (VC++7.1) allows me to declare
> variables inside the expression for an if-statement, like so:
>
> if (int i = 0)
> {
> i = 1;
> }[/color]

Keep in mind though that the above code does nothing worthwhile as the
value of "int i = 0" is the assigned value; in this case 0, which is
"false". So, i is created, initialized to 0, compared to 0, and then
goes away. The compiler might even toss all that code out.

Jakob Bieling wrote:[color=blue]
> Hi,
>
> I recently noticed that my compiler (VC++7.1) allows me to declare
> variables inside the expression for an if-statement, like so:
>
> if (int i = 0)
> {
> i = 1;
> }
>
> Now I was wondering if this is valid code (reference to the
> appropriate section/paragraph in the Standard would be nice, as I do not
> really know where to look for this kind of thing) and where else I am
> able to declare variables (except the obvious "for (int i ...)" case).[/color]

See 6.4/3.

"Jakob Bieling" <argfhesNGtzkQBGarg@rot13.com> schrieb im Newsbeitrag
news:dsg986$i06$01$1@news.t-online.com...[color=blue]
> Now I was wondering if this is valid code (reference to the appropriate
> section/paragraph in the Standard would be nice, as I do not really know
> where to look for this kind of thing) and where else I am able to declare
> variables (except the obvious "for (int i ...)" case).
>
> For the record, I am asking this out of interest, not because I want to
> start declaring my variables in all kinds of funny places :)[/color]

There is only one funny place to declare/define a variable -- at the top of
a function body. Variables should be defined as late as possible, and that
is the time of its first use. And of cause a variable should always be
initialized immediately.

That's one of the reason why in C++ you can define a variable almost
everywhere, where you can write a statement, and even at some other places
for statements. The only thing you must not do, you must never jump over a
definition. So the following code is not valid:

switch (something)
{
case 1:
int i = 1;
...
break;
case 2:
...
}

You cannot define a variable inside a switch, but if you write clean code
and such a variable is only used in one case, you can make it valid by using
a compound statement, which restricts the scope of such a variable:

switch (something)
{
case 1:
{
int i = 1;
...
}
break;
case 2:
...
}

Regards
Heinz

Heinz Ozwirk wrote:[color=blue]
> That's one of the reason why in C++ you can define a variable almost
> everywhere, where you can write a statement, and even at some other places
> for statements. The only thing you must not do, you must never jump over a
> definition. So the following code is not valid:
>
> switch (something)
> {
> case 1:
> int i = 1;
> ...
> break;
> case 2:
> ...
> }[/color]

That's interesting.
[color=blue]
> You cannot define a variable inside a switch, but if you write clean code
> and such a variable is only used in one case, you can make it valid by using
> a compound statement, which restricts the scope of such a variable:
>
> switch (something)
> {
> case 1:
> {
> int i = 1;
> ...
> }
> break;
> case 2:
> ...
> }[/color]

I have a habit of creating scope here anyway, in much the same way that
do it for the if construct.

Ben Pope
--
I'm not just a number. To many, I'm known as a string...