I'm writing the destruction for a templated class.  It has a member variable of type T (the template's type).  I want to somehow detect if this is a pointer type, and set it to null if it is.  Is there some way to detect if it's a point type, or will I have to make a specialized version of the class to handle it?

Owen

resistor@mac.com wrote:
>
> I'm writing the destruction for a templated class.  It has a member variable of
> type T (the template's type).  I want to somehow detect if this is a pointer
> type, and set it to null if it is.  Is there some way to detect if it's a point
> type, or will I have to make a specialized version of the class to handle it?

template IsPointer( T )
{
    static const bit IsPointer = false;
}

tempalte IsPointer( T : T* )
{
    static const bit IsPointer = true;
}

if( IsPointer!( T ) )
    t = NULL;

You stull might get a compiler error trying to set T to NULL depending on what T is.  You may want to do something like this instead:

template SetNULL( T )
{
    void SetNULL( T val ) {}
}

template SetNULL( T : T* )
{
    void SetNULL( T val )
    {
        val = NULL;
    }
}

SetNULL!(T)( t );

That seems rather...hackish.  I'm sure it would work, but it seems rather strange to have to do that.  Has nobody else ever needed to write a ~this() for a templated class?

Owen

In article <c1m53v$2gf5$1@digitaldaemon.com>, Sean Kelly says...
>
>resistor@mac.com wrote:
> >
>> I'm writing the destruction for a templated class.  It has a member variable of type T (the template's type).  I want to somehow detect if this is a pointer type, and set it to null if it is.  Is there some way to detect if it's a point type, or will I have to make a specialized version of the class to handle it?
>
>template IsPointer( T )
>{
>     static const bit IsPointer = false;
>}
>
>tempalte IsPointer( T : T* )
>{
>     static const bit IsPointer = true;
>}
>
>if( IsPointer!( T ) )
>     t = NULL;
>
>You stull might get a compiler error trying to set T to NULL depending on what T is.  You may want to do something like this instead:
>
>template SetNULL( T )
>{
>     void SetNULL( T val ) {}
>}
>
>template SetNULL( T : T* )
>{
>     void SetNULL( T val )
>     {
>         val = NULL;
>     }
>}
>
>SetNULL!(T)( t );
>

resistor@mac.com wrote:

> That seems rather...hackish.  I'm sure it would work, but it seems rather
> strange to have to do that.  Has nobody else ever needed to write a ~this() for
> a templated class?

If it's for a destructor, would foo=T.init work?
Sam

On Fri, 27 Feb 2004 04:04:03 +0000, resisto wrote:

> That seems rather...hackish.  I'm sure it would work, but it seems rather strange to have to do that.  Has nobody else ever needed to write a ~this() for a templated class?
> 
> Owen
> 
> In article <c1m53v$2gf5$1@digitaldaemon.com>, Sean Kelly says...
>>

I thought it to be rather novel and creative. It's an interesting example of the flexibility of templates.

Sorry if that sounded insulting.  I didn't mean to be.  I just meant that it seemed like something that shouldn't require such arcane antics to achieve.  I certainly am impressed with his solution.  It was more a comment on the need to do that at all than on his solution.

Sorry if I offended anyone.

Owen

In article <pan.2004.02.27.06.39.55.658007@telus.net>, John Reimer says...
>
>On Fri, 27 Feb 2004 04:04:03 +0000, resisto wrote:
>
>> That seems rather...hackish.  I'm sure it would work, but it seems rather strange to have to do that.  Has nobody else ever needed to write a ~this() for a templated class?
>> 
>> Owen
>> 
>> In article <c1m53v$2gf5$1@digitaldaemon.com>, Sean Kelly says...
>>>
>
>I thought it to be rather novel and creative. It's an interesting example of the flexibility of templates.

In article <c1njd1$1r5t$1@digitaldaemon.com>, resistor@mac.com says...
>
>Sorry if that sounded insulting.  I didn't mean to be.  I just meant that it seemed like something that shouldn't require such arcane antics to achieve.  I certainly am impressed with his solution.  It was more a comment on the need to do that at all than on his solution.
>
>Sorry if I offended anyone.
>
>Owen
>

No, no, no.  Don't worry about it. You were not insulting at all.  You are perfectly entitled to your opinion. For your purposes, the solution may have been more convoluted than desired.  I was just balancing out the perspective: if not a practical solution, it certainly was original :-).

Later,

John

resistor@mac.com wrote:

> That seems rather...hackish.  I'm sure it would work, but it seems rather
> strange to have to do that.  Has nobody else ever needed to write a ~this() for
> a templated class?

Not sure I understand.  If your pointer is a class member then it will disappear when the dtor exits and the gc will wipe out any orphaned data.

Sean

>> That seems rather...hackish.  I'm sure it would work, but it seems rather strange to have to do that.  Has nobody else ever needed to write a ~this() for a templated class?
>> 
>> Owen
>> 
>> In article <c1m53v$2gf5$1@digitaldaemon.com>, Sean Kelly says...
>>>
>
>I thought it to be rather novel and creative. It's an interesting example of the flexibility of templates.

This is very common technique in C++. Actualy this was the simplest kind of template-meta-programming.

On Mon, 01 Mar 2004 14:48:22 +0000, Matthias Becker wrote:

>>> That seems rather...hackish.  I'm sure it would work, but it seems rather strange to have to do that.  Has nobody else ever needed to write a ~this() for a templated class?
>>> 
>>> Owen
>>> 
>>> In article <c1m53v$2gf5$1@digitaldaemon.com>, Sean Kelly says...
>>>>
>>
>>I thought it to be rather novel and creative. It's an interesting example of the flexibility of templates.
> 
> This is very common technique in C++. Actualy this was the simplest kind of template-meta-programming.

Oh dear, I guess I better study up on my template-meta-programming then :-).