Focussed on eliminating compiler hangs, gpf's and internal errors.

http://www.digitalmars.com/d/changelog.html

Why don't scalar constructors take parametars?
int* ip = new int;
*ip == 0;

int* ip = new int(5);
*ip still == 0;

int* ip = new int(5,5,"HA");
*ip still == 0;

?

"Walter" <newshound@digitalmars.com> wrote in message news:cesu05$1bmc$1@digitaldaemon.com...
> Focussed on eliminating compiler hangs, gpf's and internal errors.
>
> http://www.digitalmars.com/d/changelog.html
>
>
>

>Why don't scalar constructors take parametars?
>int* ip = new int;
>*ip == 0;
>
>int* ip = new int(5);
>*ip still == 0;
>
>int* ip = new int(5,5,"HA");
>*ip still == 0;

Well, structs don't have constructors. Scalar types seem not to have them, too.

"Ivan Senji" <ivan.senji@public.srce.hr> wrote in message news:cet2hu$1fbm$1@digitaldaemon.com...
> Why don't scalar constructors take parametars?
> int* ip = new int;
> *ip == 0;
>
> int* ip = new int(5);
> *ip still == 0;
>
> int* ip = new int(5,5,"HA");
> *ip still == 0;
>
> ?

I'm not sure what the case is for it. The reason for adding the current method is for Matthew's DTL.

Walter wrote:
> Focussed on eliminating compiler hangs, gpf's and internal errors.
> 
> http://www.digitalmars.com/d/changelog.html

The changelog doesn't include 0.98.  Otherwise, thanks much for the release!

Russ Lewis wrote:

> Walter wrote:
> 
>> Focussed on eliminating compiler hangs, gpf's and internal errors.
>>
>> http://www.digitalmars.com/d/changelog.html
> 
> 
> The changelog doesn't include 0.98.  Otherwise, thanks much for the release!
> 

Yes, it do (and it did). Maybe your browser had a problem that made it use the cache?

Lars Ivar Igesund

but you can overload OpCall for a struct and for a class to do the same thing
you cannot, however, overload OpCall for int unless I'm mistaken

so while you could make a common way to create either structs or classes , you have to specialize for all the basic types--and woe to you if you should encounter typedefs.

Matthias Becker wrote:
>>Why don't scalar constructors take parametars?
>>int* ip = new int;
>>*ip == 0;
>>
>>int* ip = new int(5);
>>*ip still == 0;
>>
>>int* ip = new int(5,5,"HA");
>>*ip still == 0;
> 
> 
> Well, structs don't have constructors. Scalar types seem not to have them, too.
> 
> 

In article <ceu40a$2355$1@digitaldaemon.com>, Daniel Horn says...
>
>but you can overload OpCall for a struct and for a class to do the same
>thing
>you cannot, however, overload OpCall for int unless I'm mistaken
>
>so while you could make a common way to create either structs or classes , you have to specialize for all the basic types--and woe to you if you should encounter typedefs.

I, too was wondering why you can't simply go "new int(42)", when its pretty much the most obvious constructor a scalar type could have (IMO).  Granted, it's pretty useless outside of templates, but it *looks* correct, right? Frustrating.

Now if global operators were overload-able, I suppose this would be the equivalent of simulating a constructor on a built-in scalar type:

# static int opCall(int value){
#     return(value);
# }

Which doesn't get us very close:

# int foo = int(42); // closer, but not close enough.

I wouldn't even begin to know how to roll a static 'new' operator for type int. Every time I try, the syntax folds up into something ambiguous. :(

Now, not to beat a dead horse, but is this something that a "typedef block" (or "extended typedef" i guess) could help solve?

# typedef int Integer{
#     this(int value){
#         this = value; //'this' is an 'int' within this particular typedef
#     }
#     this(int value,int value2,char[] msg){
#         this = value + value2;
#         writefln(msg,"[",this,"]");
#     }
# }

At least then you have something that will /cast/ to an int with no trouble at all:

# int* foo = new Integer(42);
# int* foo = new Integer(5,5,"HA");

Now what would really do the job is if we were allowed something like this in addition to the aforementioned sytnax:

# typedef .int int{ //extension of global type 'int'
#     this(.int value){
#         this = value;
#     }
# }

Any thoughts from the crowd?

- Pragma

"Walter" <newshound@digitalmars.com> wrote in message news:cettf5$1u0m$1@digitaldaemon.com...
>
> "Ivan Senji" <ivan.senji@public.srce.hr> wrote in message news:cet2hu$1fbm$1@digitaldaemon.com...
> > Why don't scalar constructors take parametars?
> > int* ip = new int;
> > *ip == 0;
> >
> > int* ip = new int(5);
> > *ip still == 0;
> >
> > int* ip = new int(5,5,"HA");
> > *ip still == 0;
> >
> > ?
>
> I'm not sure what the case is for it. The reason for adding the current method is for Matthew's DTL.

That's great, but I agree with the others that one should be able to parameterise the construction. :)

I also think structs should have ctors, but that's another issue. (Since we have faux-ctors by opCall, why not acecpt reality and give proper ctors?)

Walter wrote:

> "Ivan Senji" <ivan.senji@public.srce.hr> wrote in message
> news:cet2hu$1fbm$1@digitaldaemon.com...
> 
>> Why don't scalar constructors take parametars?
>> int* ip = new int;
>> *ip == 0;

Would I be right to figure that this is just syntactic sugar for

	int* ip = new int[1];

albeit with some generic programming advantage I'm not sure I see?

>> int* ip = new int(5);
>> *ip still == 0;
>>
>> int* ip = new int(5,5,"HA");
>> *ip still == 0;
>>
>> ?
> 
> I'm not sure what the case is for it. The reason for adding the current
> method is for Matthew's DTL.

Then what _does_ new int(5) mean at the moment?  Or is it just a bug that the compiler accepts it?

Stewart.

-- 
My e-mail is valid but not my primary mailbox.  Please keep replies on the 'group where everyone may benefit.