Build-in types initialization

[Marcin Kalicinski]

Hi,

void f()
{
int i1; // i1 now holds undefined value
int i2 = int(); // i2 now holds 0
}

Is the above true?

[Sharad Kala]

Re: Build-in types initialization


"Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
news:c1i7fp$dal $1@korweta.task .gda.pl...[color=blue]
> Hi,
>
> void f()
> {
> int i1; // i1 now holds undefined value
> int i2 = int(); // i2 now holds 0
> }
>
> Is the above true?[/color]

Yes

[Sharad Kala]

Re: Build-in types initialization


"Sharad Kala" <no.spam_sharad k_ind@yahoo.com > wrote in message
news:c1i8bg$1ii hqh$1@ID-221354.news.uni-berlin.de...[color=blue]
>
> "Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
> news:c1i7fp$dal $1@korweta.task .gda.pl...[color=green]
> > Hi,
> >
> > void f()
> > {
> > int i1; // i1 now holds undefined value
> > int i2 = int(); // i2 now holds 0
> > }
> >
> > Is the above true?[/color][/color]

For more on this read an ongoing thread on c.l.c++.moderat ed "int(), float(),
bool()".

[Ivan Vecerina]

Re: Build-in types initialization

"Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
news:c1i7fp$dal $1@korweta.task .gda.pl...
| void f()
| {
| int i1; // i1 now holds undefined value
| int i2 = int(); // i2 now holds 0
| }
|
| Is the above true?

Yes.
This is useful to know and use especially in generic code:
T val = T(); // ensures that built-ins also get initialized

Consider also default-parameter values:
vector(size_typ e n, const T& v = T(), const A& al = A());


Regards,
Ivan
--
http://ivan.vecerina.com/contact/?subject=NG_POST <- e-mail contact form

[Marcin Kalicinski]

Re: Build-in types initialization

Hi all,

Thanks for your answer, but if it is true, how do I define my custom class
so that it follows the same schema? It seems impossible, because there's no
way to distinguish between these 2 types of initialization when writing
class definition. Both are handled by default constructor.

To be specific, I'd like to define class Vector3, which is a 3 dimensional
vector. If I initialize coordinates to zeroes in default constructor, I get
behavior 'i2'. If I do not initialize, I get behavior 'i1'. But I'd like to
get both if them, so that Vector3 does not differ in this important point
from built-in types. Additionally, Vector3 is to be used in
performance-critical code, so I'd like to avoid unnecesary initialization if
possible - so 'i1' behavior is probably a must to have.

Best regards,
Marcin

U¿ytkownik "Sharad Kala" <no.spam_sharad k_ind@yahoo.com > napisa³ w
wiadomoœci news:c1i8bg$1ii hqh$1@ID-221354.news.uni-berlin.de...

"Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
news:c1i7fp$dal $1@korweta.task .gda.pl...[color=blue]
> Hi,
>
> void f()
> {
> int i1; // i1 now holds undefined value
> int i2 = int(); // i2 now holds 0
> }
>
> Is the above true?[/color]

Yes

[Sharad Kala]

Re: Build-in types initialization


"Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
news:c1ibcm$e3b $1@korweta.task .gda.pl...[color=blue]
> Hi all,
>
> Thanks for your answer, but if it is true, how do I define my custom class
> so that it follows the same schema? It seems impossible, because there's no
> way to distinguish between these 2 types of initialization when writing
> class definition. Both are handled by default constructor.
>
> To be specific, I'd like to define class Vector3, which is a 3 dimensional
> vector. If I initialize coordinates to zeroes in default constructor, I get
> behavior 'i2'. If I do not initialize, I get behavior 'i1'. But I'd like to
> get both if them, so that Vector3 does not differ in this important point
> from built-in types. Additionally, Vector3 is to be used in
> performance-critical code, so I'd like to avoid unnecesary initialization if
> possible - so 'i1' behavior is probably a must to have.[/color]

You got it wrong.
The default constructor for int, float, bool initializes them to 0 /false.
But that does not mean that ints, floats etc in UDTs will also be default
initialized in the default constructor.

In this code -

#include <iostream>
using namespace std;
struct A{
int i;
A(){}
A(int I):i(I){}
};

int main(){

A a1 = A(); // Default const
cout << a1.i; // Print garbage..don't expect 0
A a2(10);
cout << a2.i; // Print 10
}

[Andrey Tarasevich]

Re: Build-in types initialization

Sharad Kala wrote:[color=blue]
> ...
> The default constructor for int, float, bool initializes them to 0 /false.
> But that does not mean that ints, floats etc in UDTs will also be default
> initialized in the default constructor.
> ...[/color]

There are different ways to formalize this behavior of scalar (and POD)
types in C++, but the standard way is to assume that scalar types have
no constructors at all, which is what C++ standard says.

--
Best regards,
Andrey Tarasevich

[Marcin Kalicinski]

Re: Build-in types initialization

Hi,

A a1 = A(); // Default const
cout << a1.i; // Print garbage..don't expect 0

The above is true. But if we replace A with a built-in type, say int, its
value will be initialized by 1st line. I'm worried, because this makes
user-defined types inferior in some way to built-in types. At least it makes
them behave differently, as there's no way to distinguish between these 2
construction types (available with built-in types):

1. Construct and leave members uninitialized
2. Construct and initialize members to default values, specific for each
member

Or at least there's no elegant way (by elegant I mean the way built-in types
are initialized).

Marcin


U¿ytkownik "Sharad Kala" <no.spam_sharad k_ind@yahoo.com > napisa³ w
wiadomoœci news:c1id0c$1i5 nop$1@ID-221354.news.uni-berlin.de...

"Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
news:c1ibcm$e3b $1@korweta.task .gda.pl...[color=blue]
> Hi all,
>
> Thanks for your answer, but if it is true, how do I define my custom class
> so that it follows the same schema? It seems impossible, because there's[/color]
no[color=blue]
> way to distinguish between these 2 types of initialization when writing
> class definition. Both are handled by default constructor.
>
> To be specific, I'd like to define class Vector3, which is a 3 dimensional
> vector. If I initialize coordinates to zeroes in default constructor, I[/color]
get[color=blue]
> behavior 'i2'. If I do not initialize, I get behavior 'i1'. But I'd like[/color]
to[color=blue]
> get both if them, so that Vector3 does not differ in this important point
> from built-in types. Additionally, Vector3 is to be used in
> performance-critical code, so I'd like to avoid unnecesary initialization[/color]
if[color=blue]
> possible - so 'i1' behavior is probably a must to have.[/color]

You got it wrong.
The default constructor for int, float, bool initializes them to 0 /false.
But that does not mean that ints, floats etc in UDTs will also be default
initialized in the default constructor.

In this code -

#include <iostream>
using namespace std;
struct A{
int i;
A(){}
A(int I):i(I){}
};

int main(){

A a1 = A(); // Default const
cout << a1.i; // Print garbage..don't expect 0
A a2(10);
cout << a2.i; // Print 10
}

[Sharad Kala]

Re: Build-in types initialization


"Marcin Kalicinski" <kalita@poczta. onet.pl> wrote in message
news:c1kaqj$c01 $1@korweta.task .gda.pl...[color=blue]
> Hi,[/color]
Hello,
Please don't top-post.
[color=blue]
> A a1 = A(); // Default const
> cout << a1.i; // Print garbage..don't expect 0
>
> The above is true. But if we replace A with a built-in type, say int, its
> value will be initialized by 1st line. I'm worried, because this makes
> user-defined types inferior in some way to built-in types. At least it makes
> them behave differently, as there's no way to distinguish between these 2
> construction types (available with built-in types):[/color]

Different, but I don't think that it makes them inferior.
[color=blue]
> 1. Construct and leave members uninitialized[/color]

This is your particular need. People usually want to initialize the members of
their class.
Furthermore intialization lists are there so that this happens efficiently.
[color=blue]
> 2. Construct and initialize members to default values, specific for each
> member
> Or at least there's no elegant way (by elegant I mean the way built-in types
> are initialized).[/color]

I can think of this solution, probably have a bool value in your constructor
something like -

#include <iostream>
using namespace std;
struct A{
int i;
A(bool init=false):i(( init==true)?int ():i){}
};

int main(){
A a1 = A(); // Default const
cout << a1.i; // Print garbage
A a2(true); // Iniatialize with defaults
cout << a2.i; // Print 0
}