January 01, 2013
Greetings

I observed that struct type function call arguments do not implicitly
convert even when I define a valid constructor for type conversion. So DMD
complains for line 15 in the code listed below saying:
test.d(15): Error: function test.fiz (Bar _bar) is not callable using
argument types (Foo)
test.d(15): Error: cannot implicitly convert expression (foo) of type Foo
to Bar

Is this a bug, or is it intended behavior? If it is a bug, can somebody please point me to the issue number filed on Bugzilla so that I can track it. I made some effort to search it myself but could find it on Bugzilla.

And if it is intended behavior, I guess it is because of some issues with template function matching. But I find explicit casting while calling functions a bit of a stretch.

Regards
- Puneet

struct Foo { }                 // 1
                               // 2
struct Bar {                   // 3
  this (Foo _foo) { }          // 4
  void opAssign(Foo _foo) { }  // 5
}                              // 6
                               // 7
void fiz(Bar _bar) { }         // 8
                               // 9
void main() {                  // 10
  Foo foo;                     // 11
  // implicit conversion works // 12
  Bar bar = foo;               // 13
  // But not here              // 14
  fiz(foo);                    // 15
}                              // 16


January 01, 2013
On Tuesday, 1 January 2013 at 17:27:10 UTC, d coder wrote:
> I observed that struct type function call arguments do not implicitly
> convert even when I define a valid constructor for type conversion.

That's not how you define implicit conversions in D. It's a weird hack from C++.

In D, you use alias this for implicit type conversions.
http://dlang.org/class.html#AliasThis

If you change Foo to this:

struct Foo
{
  Bar asBar() { return Bar(this); }
  alias asBar this;
}

Your code will work.

Note however that alias this has a few bugs.
http://d.puremagic.com/issues/buglist.cgi?query_format=advanced&short_desc=alias%20this&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED&short_desc_type=allwordssubstr
January 01, 2013
> That's not how you define implicit conversions in D. It's a weird hack from C++.
>
>
Thanks for clarification. I knew about alias this, but I thought it was in addition to constructor path.

Regards
- Puneet


January 02, 2013
On Wednesday, January 02, 2013 00:37:32 d coder wrote:
> > That's not how you define implicit conversions in D. It's a weird hack from C++.
> 
> Thanks for clarification. I knew about alias this, but I thought it was in addition to constructor path.

Nope. It's the _only_ way to define implicit conversions in D. In general, D requires explicit conversions in order to avoid all of the problems that implicit conversions cause in C++ (e.g. C++ will do up to 3 implicit conversions when passing an argument to a function; that gets particularly nasty when combined with function overloading; you can quickly have no idea what's actually being called without stepping through the code).

- Jonathan M Davis
Top | Discussion index | About this forum | D home