Robert,

Maybe you can fix this along the way:

import std.variant;
struct B
{
   int p, q, r, s;
}
typedef B C;

void main()
{
   B b;
   C c;
   b = c;   // ok
   Variant v = c;
   assert(v.convertsTo!(B));   // no dice
}

Steve

typedef shouldn't exist anymore, use alias.

Steve Teale Wrote:

> Robert,
> 
> Maybe you can fix this along the way:
> 
> import std.variant;
> struct B
> {
>    int p, q, r, s;
> }
> typedef B C;
> 
> void main()
> {
>    B b;
>    C c;
>    b = c;   // ok
>    Variant v = c;
>    assert(v.convertsTo!(B));   // no dice
> }
> 
> Steve

On Mon, 31 Oct 2011 11:50:42 -0400, Jesse Phillips wrote:

> typedef shouldn't exist anymore, use alias.
>

Jesse,

Maybe you want to do something like:

struct X
{
  int a;
  double b;
}
typedef X Y;
alias X Z;

void discriminate(T)(T t)
{
   if (is(T == X))
      writeln("It's an X");
   else
      writeln("No it's not");
}

void main()
{
   Y y;
   Z z;
   discriminate(y);
   discriminate(z);
}

Steve

I literally meant it is not supposed to exist. There was a suggestion to even axe it from 2.056 I believe.

A library solution is intended to be created which will allow for the type to act as either a parent or child to the type.

http://stackoverflow.com/questions/3576580/does-d-have-newtype/3577246#3577246

So I understand its usage, but you shouldn't use it because it won't be there.

> I literally meant it is not supposed to exist. There was a suggestion to even axe it from 2.056 I believe.

At least they prepared to actually deprecate it.


> A library solution is intended to be created which will allow for the type to act as either a parent or child to the type.

http://d.puremagic.com/issues/show_bug.cgi?id=5467

On Mon, 31 Oct 2011 21:03:39 +0100, Trass3r wrote:

>> I literally meant it is not supposed to exist. There was a suggestion to even axe it from 2.056 I believe.
> 
> At least they prepared to actually deprecate it.
> 
> 
>> A library solution is intended to be created which will allow for the type to act as either a parent or child to the type.
> 
> http://d.puremagic.com/issues/show_bug.cgi?id=5467

OK, my use case is that in the database interface modules, as far as possible I would like to infer an SQL type from a D type. There are often SQL types for TIME, DATE, DATETIME, and TIMESTAMP, but the latter two will usually correspond to D structs with the same members.

I can define two identical structs with the same set of protocol parsing and packing methods, but with different names, but to me this seems clunky, error-prone, and not a very good advert for the language.

There's was also a temptation to create typedefs of char[] like tinyString, mediumString and longString. I have not gone there - just as well.

Forgetting typedef, which I should assume does not exist, what is the elegant approach to this?

Steve

On Mon, 31 Oct 2011 06:30:40 -0400, Steve Teale <steve.teale@britseyeview.com> wrote:
> Robert,
>
> Maybe you can fix this along the way:
>
> import std.variant;
> struct B
> {
>    int p, q, r, s;
> }
> typedef B C;
>
> void main()
> {
>    B b;
>    C c;
>    b = c;   // ok
>    Variant v = c;
>    assert(v.convertsTo!(B));   // no dice
> }
>
> Steve

As others have mentioned, typedef is being deprecated. That said, variant now supports duck-typing. So 'v.to!B()' works. However, 'convertsTo's mandate is to test for implicit conversion, which isn't covered as there's no way in __traits to recover a typedef's base type.