www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Template overloading

reply Bill Baxter <dnewsgroup billbaxter.com> writes:
Does anyone have a really good understanding of the dummy=void hack for 
making two templates with the same name but different arguments?  Every 
time I try to use that I end up I just end up randomly adding in 
dummy=voids here and there till it works.  Is there some rationale for 
where it should go (end of parameter list? beginning?)?  And what about 
when you have 3 or more templates you need to differentiate?

Also, am I right in thinking that if you pass any of the "real" template 
arguments explicitly then suddenly the dummy args must be passed too?

Any advice on best practices in this area?

My most recent run-in with this was trying to write a matrix multiply 
template that could take arguments of either Mat!(T),Mat!(T), or 
Mat!(T),Vec!(T).

--bb
Nov 25 2007
parent reply Bruno Medeiros <brunodomedeiros+spam com.gmail> writes:
Bill Baxter wrote:

 Does anyone have a really good understanding of the dummy=void hack for 
 making two templates with the same name but different arguments?  Every 
 time I try to use that I end up I just end up randomly adding in 
 dummy=voids here and there till it works.  Is there some rationale for 
 where it should go (end of parameter list? beginning?)?  And what about 
 when you have 3 or more templates you need to differentiate?
 


What do you mean "3 or more templates you need to differentiate"? Do you 
remember this: 
http://d.puremagic.com/issues/show_bug.cgi?format=multiple&id=337 ?


 Also, am I right in thinking that if you pass any of the "real" template 
 arguments explicitly then suddenly the dummy args must be passed too?
 


Yes, the dummy hack only works implicitly for IFTI.


 Any advice on best practices in this area?
 
 My most recent run-in with this was trying to write a matrix multiply 
 template that could take arguments of either Mat!(T),Mat!(T), or 
 Mat!(T),Vec!(T).
 
 --bb


What exactly are you trying to accomplish? This works:

void mult2(T)(Mat!(T) a, Mat!(T) b) {
	pragma(msg, "Mat, Mat");
}

void mult2(T)(Mat!(T) a, Vec!(T) b) {
	pragma(msg, "Mat, Vec");
}


void main() {
     mult2(new Mat!(int), new Mat!(int));
     mult2(new Mat!(int), new Vec!(int));
}

But that seems too simple a solution, was that you were trying to do?

-- 
Bruno Medeiros - MSc in CS/E student
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D
Dec 05 2007
parent Bill Baxter <dnewsgroup billbaxter.com> writes:
Bruno Medeiros wrote:

 Bill Baxter wrote:

 Does anyone have a really good understanding of the dummy=void hack 
 for making two templates with the same name but different arguments?  
 Every time I try to use that I end up I just end up randomly adding in 
 dummy=voids here and there till it works.  Is there some rationale for 
 where it should go (end of parameter list? beginning?)?  And what 
 about when you have 3 or more templates you need to differentiate?


 
 What do you mean "3 or more templates you need to differentiate"? Do you 
 remember this: 
 http://d.puremagic.com/issues/show_bug.cgi?format=multiple&id=337 ?
 

 Also, am I right in thinking that if you pass any of the "real" 
 template arguments explicitly then suddenly the dummy args must be 
 passed too?


 
 Yes, the dummy hack only works implicitly for IFTI.
 

 Any advice on best practices in this area?

 My most recent run-in with this was trying to write a matrix multiply 
 template that could take arguments of either Mat!(T),Mat!(T), or 
 Mat!(T),Vec!(T).

 --bb


 
 What exactly are you trying to accomplish? This works:
 
 void mult2(T)(Mat!(T) a, Mat!(T) b) {
     pragma(msg, "Mat, Mat");
 }
 
 void mult2(T)(Mat!(T) a, Vec!(T) b) {
     pragma(msg, "Mat, Vec");
 }
 
 
 void main() {
     mult2(new Mat!(int), new Mat!(int));
     mult2(new Mat!(int), new Vec!(int));
 }
 
 But that seems too simple a solution, was that you were trying to do?


Ah, ok.  Actually in the real code this was taken from there are integer 
parameters that give the dimensions of the Mat and Vec.  But DMD seems 
to have some issues with inference on integer parameters wrapped in a 
type, so I'm guessing that was probably why the straightforward solution 
didn't work.

Actually the two should be:

void mult2(T, int M, int N, int P)(Mat!(T,M,N) a, Mat!(T,N,P) b) {
     pragma(msg, "Mat, Mat");
}

void mult2(T,int M, int N)(Mat!(T,M,N) a, Vec!(T,N) b) {
     pragma(msg, "Mat, Vec");
}

--bb
Dec 05 2007