www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Conditional Templates

reply Derek Parnell <derek nomail.afraid.org> writes:
Help! The D documentation makes me seem such a fool. 

Here is what I want to do ...

I want to create a template for a single class member method that behaves
differently depending on the type of the one parameter supplied to the
method. In other words, if the method is called with a integer type of
parameter, D should instantiate one form of the member method, and if the
parameter is a floating point data type, D should instantiate another form
of the member method, and finally if the parameter is a specific class, D
should instantiate yet another form of the member method.

So in 'pseudo' code ...

class Bar
{
 template Foo(T)
 {
    static if (T is an integer type) // byte, ubyte, short, ushort, ...
    {
        void Foo(T x)
        {
             // do something with the integer 'x'
        }
    }

    static if (T is a floating point type) // float, double, real
    {
        void Foo(T x)
        {
             // do something with the floating point value.
        }
    }

    static if (T is the class 'Bar')
    {
        void Foo(Bar x)
        {
             // do something with this object.
        }
    }

    // All other forms are illegal.
 }
}

I've searched the docs and failed to solve how this can be done, if at all.
I'm sure there are some examples in the newsgroup postings but its like
looking for something that you don't know what it actually looks like.

I did note that the docs say you can't use templates to add non-static
class members, but I tried and it works fine; so I don't know if the docs
are wrong, I'm wrong or DMD is wrong.

-- 
Derek
(skype: derek.j.parnell)
Melbourne, Australia
"Down with mediocrity!"
29/01/2007 5:17:20 PM
Jan 28 2007
next sibling parent reply Max Samukha <samukha voliacable.com> writes:
On Mon, 29 Jan 2007 17:28:51 +1100, Derek Parnell
<derek nomail.afraid.org> wrote:

Help! The D documentation makes me seem such a fool. 

Here is what I want to do ...

I want to create a template for a single class member method that behaves
differently depending on the type of the one parameter supplied to the
method. In other words, if the method is called with a integer type of
parameter, D should instantiate one form of the member method, and if the
parameter is a floating point data type, D should instantiate another form
of the member method, and finally if the parameter is a specific class, D
should instantiate yet another form of the member method.

So in 'pseudo' code ...

class Bar
{
 template Foo(T)
 {
    static if (T is an integer type) // byte, ubyte, short, ushort, ...
    {
        void Foo(T x)
        {
             // do something with the integer 'x'
        }
    }

    static if (T is a floating point type) // float, double, real
    {
        void Foo(T x)
        {
             // do something with the floating point value.
        }
    }

    static if (T is the class 'Bar')
    {
        void Foo(Bar x)
        {
             // do something with this object.
        }
    }

    // All other forms are illegal.
 }
}

I've searched the docs and failed to solve how this can be done, if at all.
I'm sure there are some examples in the newsgroup postings but its like
looking for something that you don't know what it actually looks like.

I did note that the docs say you can't use templates to add non-static
class members, but I tried and it works fine; so I don't know if the docs
are wrong, I'm wrong or DMD is wrong.

With a mixin (if you want your method in more than one class): template TFoo(T) { static if (is(T == float) || is(T == real)) { void Foo(T x) { writefln("In float foo: ", x); } } else static if (is(T == int) || is(T == short)) { void Foo(T x) { writefln("In integer foo: ", x); } } else static assert("Bad type"); } class Bar(T) { mixin TFoo!(T); } void main() { auto bar = new Bar!(int); bar.Foo(1); auto bar2 = new Bar!(real); bar2.Foo(2.345364); } Or directly in the class body: class Bar(T) { static if (is(T == float) || is(T == real)) { void Foo(T x) { writefln("In float foo: ", x); } } else static if (is(T == int) || is(T == short)) { void Foo(T x) { writefln("In integer foo: ", x); } } else static assert("Bad type"); } void main() { auto bar = new Bar!(int); bar.Foo(1); auto bar2 = new Bar!(real); bar2.Foo(2.345364); } You can also turn 'is' conditions into templates (I'm not sure if they are in the std library already) template IsFloat(T) { const bool IsFloat = is(T == float) || is(T == double) || ...; } and use it like this: static if (IsFloat!(T)) ... You could also try to put 'static if' in the method's body, which is cleaner: ... void Foo(T x) { static if () { } else static if () ... }
Jan 29 2007
parent reply Thomas Kuehne <thomas-dloop kuehne.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Max Samukha schrieb am 2007-01-29:
 On Mon, 29 Jan 2007 17:28:51 +1100, Derek Parnell
<derek nomail.afraid.org> wrote:

Help! The D documentation makes me seem such a fool. 

Here is what I want to do ...

I want to create a template for a single class member method that behaves
differently depending on the type of the one parameter supplied to the
method. In other words, if the method is called with a integer type of
parameter, D should instantiate one form of the member method, and if the
parameter is a floating point data type, D should instantiate another form
of the member method, and finally if the parameter is a specific class, D
should instantiate yet another form of the member method.

So in 'pseudo' code ...

class Bar
{
 template Foo(T)
 {
    static if (T is an integer type) // byte, ubyte, short, ushort, ...
    {
        void Foo(T x)
        {
             // do something with the integer 'x'
        }
    }

    static if (T is a floating point type) // float, double, real
    {
        void Foo(T x)
        {
             // do something with the floating point value.
        }
    }

    static if (T is the class 'Bar')
    {
        void Foo(Bar x)
        {
             // do something with this object.
        }
    }

    // All other forms are illegal.
 }
}

I've searched the docs and failed to solve how this can be done, if at all.
I'm sure there are some examples in the newsgroup postings but its like
looking for something that you don't know what it actually looks like.

I did note that the docs say you can't use templates to add non-static
class members, but I tried and it works fine; so I don't know if the docs
are wrong, I'm wrong or DMD is wrong.

With a mixin (if you want your method in more than one class): template TFoo(T) { static if (is(T == float) || is(T == real)) { void Foo(T x) { writefln("In float foo: ", x); } } else static if (is(T == int) || is(T == short)) { void Foo(T x) { writefln("In integer foo: ", x); } } else static assert("Bad type"); }

I'd rather sugest static if(is(T : long) || is(T : ulong)) // ... else static if(is(T : real)) // .. else Thomas -----BEGIN PGP SIGNATURE----- iD8DBQFFvciULK5blCcjpWoRAn5nAKCpvyONxqqBpPdamnb/+NzTc+3aTQCePURd jKCGiYXxaD3LH7nbUEJ0ysw= =OnR+ -----END PGP SIGNATURE-----
Jan 29 2007
parent Max Samukha <samukha voliacable.com> writes:
On Mon, 29 Jan 2007 09:18:54 +0000 (UTC), Thomas Kuehne
<thomas-dloop kuehne.cn> wrote:

I'd rather sugest

	static if(is(T : long) || is(T : ulong))
	// ...
	else static if(is(T : real))
	// ..
	else 

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFFvciULK5blCcjpWoRAn5nAKCpvyONxqqBpPdamnb/+NzTc+3aTQCePURd
jKCGiYXxaD3LH7nbUEJ0ysw=
=OnR+
-----END PGP SIGNATURE-----

Yes, that's better. Thanks
Jan 29 2007
prev sibling next sibling parent "Christian Kamm" <kamm nospam.de> writes:
 I did note that the docs say you can't use templates to add non-static=

 class members, but I tried and it works fine; so I don't know if the d=

 are wrong, I'm wrong or DMD is wrong.

It seems as if you can add non-static template member functions but not = = variables: http://d.puremagic.com/issues/show_bug.cgi?id=3D566 Christian
Jan 29 2007
prev sibling next sibling parent Frits van Bommel <fvbommel REMwOVExCAPSs.nl> writes:
Derek Parnell wrote:
 Help! The D documentation makes me seem such a fool. 
 
 Here is what I want to do ...
 
 I want to create a template for a single class member method that behaves
 differently depending on the type of the one parameter supplied to the
 method. In other words, if the method is called with a integer type of
 parameter, D should instantiate one form of the member method, and if the
 parameter is a floating point data type, D should instantiate another form
 of the member method, and finally if the parameter is a specific class, D
 should instantiate yet another form of the member method.
 
 So in 'pseudo' code ...
 
 class Bar
 {
  template Foo(T)
  {
     static if (T is an integer type) // byte, ubyte, short, ushort, ...
     {
         void Foo(T x)
         {
              // do something with the integer 'x'
         }
     }
 
     static if (T is a floating point type) // float, double, real
     {
         void Foo(T x)
         {
              // do something with the floating point value.
         }
     }
 
     static if (T is the class 'Bar')
     {
         void Foo(Bar x)
         {
              // do something with this object.
         }
     }
 
     // All other forms are illegal.
  }
 }
 
 I've searched the docs and failed to solve how this can be done, if at all.
 I'm sure there are some examples in the newsgroup postings but its like
 looking for something that you don't know what it actually looks like.

Try one of these two methods[1] (Foo1 & Foo2): ----- import std.stdio; class Bar { private template Foo1Impl(T) { static if (is(T : int)) { // byte, ubyte, short, ushort, ... void Foo1Impl(T x) { // do something with the integer 'x' writefln("Foo1!(%s : int)(%x)", typeid(T), x); } } else static if (is(T : float)) { // float, double, real void Foo1Impl(T x) { // do something with the floating point value. writefln("Foo1!(%s : float)(%f)", typeid(T), x); } } else static if (is(T == Bar)) { void Foo1Impl(Bar x) { // do something with this object. writefln("Foo1!(%s == Bar)(%s)", typeid(T), x); } } else { // All other forms are illegal. static assert(0, "Illegal instantiation of Bar with " ~ T.mangleof); } } void Foo1(T)(T x) { // FooImpl isn't a function template so we need to explicitly // instantiate it Foo1Impl!(T)(x); } void Foo2(T)(T x) { static if (is(T : int)) { // byte, ubyte, short, ushort, ... // do something with the integer 'x' writefln("Foo2!(%s : int)(%x)", typeid(T), x); } else static if (is(T : float)) { // float, double, real // do something with the floating point value. writefln("Foo2!(%s : float)(%f)", typeid(T), x); } else static if (is(T == Bar)) { // do something with this object. writefln("Foo2!(%s == Bar)(%s)", typeid(T), x); } else { // All other forms are illegal. static assert(0, "Illegal instantiation of Bar with " ~ T.mangleof); } } char[] toString() { return "[Some Bar]"; }; } // Some code to test this out: template Tuple(Ts...) { alias Ts Tuple; } alias Tuple!(char, wchar, dchar, byte, ubyte, short, ushort, int, uint, long, ulong, float, double, real, Bar) Types; void main() { Bar b = new Bar; foreach (T; Types) { T t; static if (is(T == Bar)) { t = new T; } else { t = 1; } b.Foo1(t); b.Foo2(t); } } ----- Is-expressions are very nice for this sort of thing. The ones above match (respectively) anything that implicitly converts to 'int', anything that explicitly converts to 'float', and Bar (and only Bar), and then repeat the list for Foo2. To match any subclasses of Bar as well, change the '==' to ':' in that is-expression. Anything else runs into a static assert (though in the first case it's redundant; if you leave it out there won't be a match for Foo1Impl!(Whatever).Foo1Impl and it'll still be an error). These do allow char, wchar and dchar as integers though, not sure if that was your intent. If not, you'll need to explicitly filter them out. Note also that order is important: the types that convert to int _also_ convert to float, so the int case needs to be before the float case.
 I did note that the docs say you can't use templates to add non-static
 class members, but I tried and it works fine; so I don't know if the docs
 are wrong, I'm wrong or DMD is wrong.

All I know is it works. I'm pretty sure you can't override them though. [1] No pun intended :p.
Jan 29 2007
prev sibling parent reply "Jarrett Billingsley" <kb3ctd2 yahoo.com> writes:
"Derek Parnell" <derek nomail.afraid.org> wrote in message 
news:1p0qyzhxblvjj$.xbog5ahy31gi$.dlg 40tude.net...
 Help! The D documentation makes me seem such a fool.

 Here is what I want to do ...

 I want to create a template for a single class member method that behaves
 differently depending on the type of the one parameter supplied to the
 method. In other words, if the method is called with a integer type of
 parameter, D should instantiate one form of the member method, and if the
 parameter is a floating point data type, D should instantiate another form
 of the member method, and finally if the parameter is a specific class, D
 should instantiate yet another form of the member method.

I'm probably missing something important, but can't this be done with normal function overloading?
Jan 29 2007
parent reply Frits van Bommel <fvbommel REMwOVExCAPSs.nl> writes:
Jarrett Billingsley wrote:
 "Derek Parnell" <derek nomail.afraid.org> wrote in message 
 news:1p0qyzhxblvjj$.xbog5ahy31gi$.dlg 40tude.net...
 Help! The D documentation makes me seem such a fool.

 Here is what I want to do ...

 I want to create a template for a single class member method that behaves
 differently depending on the type of the one parameter supplied to the
 method. In other words, if the method is called with a integer type of
 parameter, D should instantiate one form of the member method, and if the
 parameter is a floating point data type, D should instantiate another form
 of the member method, and finally if the parameter is a specific class, D
 should instantiate yet another form of the member method.

I'm probably missing something important, but can't this be done with normal function overloading?

All integer types in D can implicitly convert to floating-point types. And he wants integer and float types to behave differently. In order to solve this use overloading you'd have to introduce a separate overload for every integer type to avoid ambiguous overloads. That's a lot of overloads :p... And it wouldn't even work for the 'cent' and 'ucent' types when they get added (they're reserved keywords for 128-bit integer types).
Jan 29 2007
parent reply "Jarrett Billingsley" <kb3ctd2 yahoo.com> writes:
"Frits van Bommel" <fvbommel REMwOVExCAPSs.nl> wrote in message 
news:epkvl5$2vc1$1 digitaldaemon.com...
 All integer types in D can implicitly convert to floating-point types. And 
 he wants integer and float types to behave differently. In order to solve 
 this use overloading you'd have to introduce a separate overload for every 
 integer type to avoid ambiguous overloads. That's a lot of overloads :p...

Ahh.
Jan 29 2007
parent reply Derek Parnell <derek nomail.afraid.org> writes:
On Mon, 29 Jan 2007 18:48:44 -0500, Jarrett Billingsley wrote:

 "Frits van Bommel" <fvbommel REMwOVExCAPSs.nl> wrote in message 
 news:epkvl5$2vc1$1 digitaldaemon.com...
 All integer types in D can implicitly convert to floating-point types. And 
 he wants integer and float types to behave differently. In order to solve 
 this use overloading you'd have to introduce a separate overload for every 
 integer type to avoid ambiguous overloads. That's a lot of overloads :p...

Ahh.

Exactly. I started out by using function overloading but all the cut&paste code got a bit obvious. Thanks for all who stopped by to help me. I now have the class working (sort of). ** Things I have Learned ** (a) Walter, true to his admission, is not a good documenter. The current D documentation provided by DigitalMars is a productivity sinkhole. Things are hard, or nigh impossible to find, and if you do find the thing you are looking for, the explanation is usually too brief, and may or may not be accurate. (I'll be soon adding to the Wiki-Doc-comments on a few items ). (b) Templated member functions do not seem to be able to be used as properties. In the example code given by Frits van Bommel, if you replace the execution code ... b.Foo1(t); b.Foo2(t); with b.Foo1 = t; b.Foo2 = t; you get errors along the lines of ... Error: (b).Foo2(T) has no value Error: (b).Foo2 is not an lvalue (c) Current DMD is unable to help templates distinguish (accurately) between character and integer types. I added some code to detect the 'char' family of types, but this just caused all char and int types to fail. void Foo2(T)(T x) { static if (is(T : char)) { // do something with the char 'x' static assert(0, "No Char allowed"); writefln("Foo2!(%s : char)(%x)", typeid(T), x); } else static if (is(T : int)) { // byte, ubyte, short, ushort, ... // do something with the integer 'x' writefln("Foo2!(%s : int)(%x)", typeid(T), x); } else static if (is(T : float)) { // float, double, real // do something with the floating point value. writefln("Foo2!(%s : float)(%f)", typeid(T), x); } else static if (is(T == Bar)) { // do something with this object. writefln("Foo2!(%s == Bar)(%s)", typeid(T), x); } else { // All other forms are illegal. static assert(0, "Illegal instantiation of Bar with " ~ T.mangleof); } } Then tested it using 'char' and it failed as expected. I then removed the 'char' types and it still failed ("No Char allowed"). I next removed all the int types and then it worked. So it seems that D regards characters and integers as interchangeable entities, which is just not correct. I know that characters are implemented in memory as integers, but they are conceptually different things, and just sometimes coders need to treat them as different things. As a double-blind test, I went back to Frits' original code and replaced static if (is(T : int)) { with static if (is(T : char)) { and the test ran without failing. This indicates that D can't see any difference between integer and character literals. -- Derek (skype: derek.j.parnell) Melbourne, Australia "Down with mediocrity!" 30/01/2007 10:53:56 AM
Jan 29 2007
parent "Jarrett Billingsley" <kb3ctd2 yahoo.com> writes:
"Derek Parnell" <derek nomail.afraid.org> wrote in message 
news:2bjnrvcmm8o9$.k39bxr2x2c52$.dlg 40tude.net...

 So it seems that D regards characters and
 integers as interchangeable entities, which is just not correct. I know
 that characters are implemented in memory as integers, but they are
 conceptually different things, and just sometimes coders need to treat 
 them
 as different things.

I hate this too. I've ended up writing all kinds of ugly templates to deal with this problem. I also really like this one: static assert(is(bool : dchar)); This passes. This caused a very difficult-to-find bug in MiniD, because never in a million years would I have expected bool to be implicitly convertible to dchar.
Jan 29 2007