the following is not compilable:
enum type{
a,
b,
}
template XX(uint a, uint c)
{
uint XX(){ return (a*256+c);}
}
void main()
{
int k=XX(cast(uint)type.a,cast(uint)type.b); // can't match any template declaration while it should
}
but a workaround as following is working fine :
enum type{
a,
b,
}
template XX(uint a, uint c)
{
uint XX(){ return (a*256+c);}
}
void main()
{
int k=XX!(cast(uint)type.a,cast(uint)type.b)();
}
Comment #1 by thomas-dloop — 2007-04-06T04:14:43Z
# template X(uint a, uint c){
# uint X(){ return 0; }
# }
#
# X(1u,2u);
According to http://www.digitalmars.com/d/template.html this doesn't the above
code doesn't seem to be legal. The following code however is legal and
compiles:
# template X(T1, T2){
# uint X(T1 a, T2 b){ return 0; }
# }
#
# X(1u,2u);
and
# template X(T){
# uint X(T a, T b){ return 0; }
# }
#
# X(1u,2u);
Comment #2 by davidl — 2007-04-09T04:10:47Z
can u show me exact sentence in the doc prevent the template i use?
The problem is dmd prevent IFTI from compiling while it allows the EXPLICITL template form.
the uint a, uint c here are just some compile time literal integer
Comment #3 by fvbommel — 2007-04-09T05:24:14Z
How about:
=====
Function templates can be explicitly instantiated [snip] or implicitly, where the TemplateArgumentList is deduced from the types of the function arguments
=====
Note: The *types* of the function arguments, not their values.
IFTI only works for type arguments.
This code is pretty close to yours though:
-----
enum type{
a,
b,
}
template XX(uint a, uint c)
{
uint XX(){ return (a*256+c);}
}
void main()
{
int k=XX!(type.a,type.b);
}
-----
All in all, a one-character edit is all it took.
(The casts are redundant so I removed them, and the trailing parentheses in your "workaround" are optional because of property syntax. Both should be fine if you left them in though)
If you still don't understand what's wrong with your code after reading this, I suggest you post in digitalmars.D.learn instead of reopening this issue.
Comment #4 by davidl — 2007-04-09T10:58:34Z
umm, i get it. It's as designed , making IFTI deduce TemplateArgumentList in the way which i thought it were would cause confliction between template(t:uint) and template(uint a)