Scope difference between MyClass(expr); and MyClass myObject(expr);

Question

I have a statistics manager that examines my application performance by measuring the time it takes for a method to execute. It's usage is like this:

myStatManager.StartStat("Rendering");
Render();
myStatManager.StopStat("Rendering");

The output to screen will tell me how long that method took.

For convenience, I have written a dummy object, that calls these two methods as the object is created and destroyed. This allows me to use C++ scope rules to my advantage and only type the stat tracking lines once instead of twice as above.

class ScopedStat
{
    string label;
    ScopedStat(string inLabel): label(inLabel) { myStatManager.StartStat(label); }
    ~ScopedStat() { myStatManager.StopStat(label); }
}

The expected usage is as follows:

{
    ScopedStat("Rendering");
    Render();
}

However, this does not work, as perhaps the compiler or something has optimized the ScopedStat object out. The reported time is a fraction of a millisecond and nowhere near the time it should take to render. My question is, why does this way not work? Does this object not get destroyed at the end of scope?

Edit: I have found a workaround:

{
    ScopedStat ss("Rendering");
    Render();
}

This works as intended--the object gets destroyed only at the end of the curly bracket. Though, I'd still like to know why.

Note: Using Microsoft Visual Studio 2008 C++;

Edit2: Ah, i understand now that unless i bind my object to a variable it is destroyed after the expression is evaluated. Thanks for all your help.

Does anyone know why C++ is written this way? What use is a temporary variable if it gets destroyed immediately?

I think they're related but not duplicates. This question's about the difference between `MyClass(expr);` and `MyClass myObject(expr);` and the answer is about temporary variables for the first case. Whereas the potential duplicate is asking specifically about lifetime of temporaries. — TooTone, Apr 11 '14 at 21:40

score 4 · Answer 1 · answered Jul 17 '11 at 21:42

{
    ScopedStat("Rendering");
    Render();
}
However, this does not work, as perhaps the compiler or something has optimized the ScopedStat object out. The reported time is a fraction of a millisecond and nowhere near the time it should take to render. My question is, why does this way not work? Does this object not get destroyed at the end of scope?

You're not creating a normal variable that just happens to have no name. You're creating a temporary, and it ceases to exist at the end of the full-expression in which it was declared. (That is, before Render() is invoked, the ScopedStat has come and gone already.)

You'll have to give it a name.

To answer your final question, this makes perfect sense when you do something like:

doFoo(ScopedStat("Rendering"));

The temporary lives as long as does the call to doFoo which is just right.

score 3 · Answer 2 · answered Jul 17 '11 at 21:34

If you don't use an identifier for an object, it becomes a temporary object and gets destroyed after the expression it belongs to ends. In other words, it has the scope of that single expression.

You have to name your objects for them to live until the end of the bracket scope.

ScopedStat stat("Rendering");

score 1 · Answer 3 · answered Jul 17 '11 at 21:34

1

You have created a temp object which is out of scope immediately, that is when the next line starts to execute. so that start and stop are done before the render.

answered Jul 17 '11 at 21:34

roni bar yanai

1,492
10
12

Scope difference between MyClass(expr); and MyClass myObject(expr);

3 Answers3