Say I have a struct that looks like this (a POD):
struct Foo
{
int i;
double d;
};
What are the differences between the following two lines:
Foo* f1 = new Foo;
Foo* f2 = new Foo();
-
Note the related question Why is it an error to use an empty set of brackets to call a constructor with no arguments?.Troubadour– Troubadour2010-06-01 16:51:49 +00:00Commented Jun 1, 2010 at 16:51
Add a comment
|
3 Answers
The first one leaves the values uninitialised; the second initialises them to zero. This is only the case for POD types, which have no constructors.
3 Comments
criddell
Thank you! Is that the only difference?
Mike Seymour
Yes, that's the only difference.
Suma
And if I am not mistaken, this was only added in some newer revisions of the standard (2003?)
I suppose nothing at all. Foo() is allowed, even if it makes no sense... I've tried to change struct into class and tried a diff on the generated exe, and they resulted to be the same, meaning that a class without method is like a struct from a practical and "effective" point of view.
But: if you use only one of the alternative, keeping struct or class no matter, it happens that new Foo and new Foo() gives executables which differ! (At least using g++) I.e.
struct Foo { int i; double d; }
int main() { Foo *f1 = new Foo; delete f1; }
is compiled into somehing different from
struct Foo { int i; double d; }
int main() { Foo *f1 = new Foo(); delete f1; }
and the same happens with class instead of struct. To know where the difference is we should look at the generated code... and to know if it is a g++ idiosincracy or not, I should try another compiler but I have only gcc and no time now to analyse the asm output of g++...
Anyway from a "functional" (practical) point of view, it is the same thing.
Add
At the end it is always better to know or do deeper investigation for some common human problems on Q/A sites... the only difference in the code generated by g++ in () and no () cases,
movl $0, (%eax)
fldz
fstpl 4(%eax)
which is a fragment that initializes to 0/0.0 the int and the double of the struct... so Seymour knows it better (but I could have discovered it without knowing if I had taken a look at the asm first!)
2 Comments
Salgar
Why does using brackets make no sense, it is invoking the default constructor. If anything, a lack of parens makes no sense.
ShinTakezou
what are you saying? it's a test, plausible for the question and the fact that foo() normally means a call to a function/method; a not so useful test, rather than a reading of specs, this would have been a possible critic.Your comment means nothing to me.
Per the link I posted.
In C++ the only difference between a class and a struct is that class-members are private by default, while struct-members default to public. So structures can have constructors, and the syntax is the same as for classes.
1 Comment
Troubadour
The point is the use or omission of the curved brackets when constructing new objects. The question applies to classes as well as structs.