Initialization inside a class

[Roman]

Hi,

Does anyone know why in C++, it is ok to initialize a const static int
inside a class, but you can't initialize a float?

From what I understand, initialization of const static float can only
be done inside a constructor.

For example:

class A
{
public:

const static int b = 123; // OK
const static double c = 123.456; // ERROR

A();
~A();
};
Roman

[Alf P. Steinbach]

* Roman:

Does anyone know why in C++, it is ok to initialize a const static int
inside a class, but you can't initialize a float?

Nobody _knows_ why, but the decision was made to support integral
constants so that they could be used in compile time expressions,
without generalizing the feature.

Generalization could be difficult.

For now, you can do one of two things if you want that value specified
in a header file:

* Place the constant outside the class, with internal linkage.
* Use the template constant trick.

Here's the template constant trick (perhaps I should apply for a
patent?):

template< typename T >
struct StrawberryConst ants { static const double pi; };

template< typename T >
double StrawberryConst ants<T>::pi = 3.14;

struct Dummy {};

class Strawberry: public StrawberryConst ants<Dummy>
{
// Whatever.
};

One may of course argue that the template constant trick means there's
no _technical_ reason why the language can't support that directly.
After all, any C++ compiler is required to be able to deal with it.
Perhaps it's just that these features were introduced in the "wrong"
order, and nobody's yet found it important enough to champion a clean-up
operation vis-a-vis the standardization committee.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[red floyd]

Roman wrote:

Hi,

Does anyone know why in C++, it is ok to initialize a const static int
inside a class, but you can't initialize a float?
Because the Standard says so.

From what I understand, initialization of const static float can only
be done inside a constructor. Incorrect. It needs to be done in the declaration of the static const
member.
For example:

class A
{
public:

const static int b = 123; // OK const static double c;
A();
~A();
};

const double A::c = 123.456;

See, that's not inside a constructor.

[Jim Langston]

"Alf P. Steinbach" <al***@start.no > wrote in message
news:43******** ********@news.i ndividual.net.. .

* Roman:

Does anyone know why in C++, it is ok to initialize a const static int
inside a class, but you can't initialize a float?
Nobody _knows_ why, but the decision was made to support integral
constants so that they could be used in compile time expressions,
without generalizing the feature.

Generalization could be difficult.

For now, you can do one of two things if you want that value specified
in a header file:

* Place the constant outside the class, with internal linkage.
* Use the template constant trick.

Here's the template constant trick (perhaps I should apply for a
patent?):

template< typename T >
struct StrawberryConst ants { static const double pi; };

template< typename T >
double StrawberryConst ants<T>::pi = 3.14;

struct Dummy {};

I find little benifit of having to delcare Dummy{}; then to just have the
statement
StrawberryConst ants::pi = 3.14;

class Strawberry: public StrawberryConst ants<Dummy>
{
// Whatever.
};

One may of course argue that the template constant trick means there's
no _technical_ reason why the language can't support that directly.
After all, any C++ compiler is required to be able to deal with it.
Perhaps it's just that these features were introduced in the "wrong"
order, and nobody's yet found it important enough to champion a clean-up
operation vis-a-vis the standardization committee.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[Jim Langston]

"Jim Langston" <ta*******@rock etmail.com> wrote in message
news:k5******** *******@fe03.lg a...

"Alf P. Steinbach" <al***@start.no > wrote in message
news:43******** ********@news.i ndividual.net.. .

* Roman:

Does anyone know why in C++, it is ok to initialize a const static int
inside a class, but you can't initialize a float?
Nobody _knows_ why, but the decision was made to support integral
constants so that they could be used in compile time expressions,
without generalizing the feature.

Generalization could be difficult.

For now, you can do one of two things if you want that value specified
in a header file:

* Place the constant outside the class, with internal linkage.
* Use the template constant trick.

Here's the template constant trick (perhaps I should apply for a
patent?):

template< typename T >
struct StrawberryConst ants { static const double pi; };

template< typename T >
double StrawberryConst ants<T>::pi = 3.14;

struct Dummy {};

I find little benifit of having to delcare Dummy{}; then to just have the
statement
StrawberryConst ants::pi = 3.14;

Of course I meant
const double StrawberryConst ants::pi = 3.14;

class Strawberry: public StrawberryConst ants<Dummy>
{
// Whatever.
};

One may of course argue that the template constant trick means there's
no _technical_ reason why the language can't support that directly.
After all, any C++ compiler is required to be able to deal with it.
Perhaps it's just that these features were introduced in the "wrong"
order, and nobody's yet found it important enough to champion a clean-up
operation vis-a-vis the standardization committee.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[Alf P. Steinbach]

* Jim Langston:

I find little benifit of having to delcare Dummy{}; then to just have the
statement
StrawberryConst ants::pi = 3.14;

You don't have to.

Use "int" or whatever, although that will reduce readability by implying
the type matters (some libraries provide a ready-made Dummy type).

Btw. that should be "benefit" (your fault) and "const" (my fault).
Cheers,

- Alf

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[Axter]

Roman wrote:

Hi,

Does anyone know why in C++, it is ok to initialize a const static int
inside a class, but you can't initialize a float?

From what I understand, initialization of const static float can only
be done inside a constructor.

For example:

class A
{
public:

const static int b = 123; // OK
const static double c = 123.456; // ERROR

A();
~A();
};

A simple workaround solution is to use an inline static function with a
static local variable.

class A
{
public:
const static int b = 123; // OK
static inline double get_c(){
const static double c = 123.456;
return c;
}
A();
~A();
};