I'm having trouble with creating a template in D that works:
pure T BSpline(int k:1, T)(in T x, in T[] t)
{
if (t[0] <= x && x < t[k])
return 1;
else
return 0;
}
pure T BSpline(int k, T)(in T x, in T[] t)
{
if (t[0] <= x && x < t[k])
{
T a = (x - t[0]) / (t[k-1] - t[0]);
T b = (t[k] - x) / (t[k] - t[1]);
return a * BSpline!(k-1,T)(x, t[0..k-1]) + b * BSpline!(k-1,T)(x, t[1..k]);
}
else
return 0;
}
And then my unit-tests:
unittest {
real a = .5;
real[] b = [0,0,1,2,3];
assert(BSpline!(1,real)(a,b[0..2]) == 0);
assert(BSpline!(2,real)(a,b[0..3]) == .5);
assert(BSpline!(3,real)(a,b[0..4]) == .625);
assert(BSpline!(4,real)(a,b[0..5]) == 0.260417);
}
which are failing with the following error:
bspline.d(18): Error: template BSplineBasis.BSpline(int k : 1,T) does not match any function template declaration
bspline.d(18): Error: template BSplineBasis.BSpline(int k : 1,T) cannot deduce template function from argument types !(1,real)(const(real),const(real[]))
bspline.d(18): Error: template instance errors instantiating template
I wouldn't be asking, however, I don't understand why D is having trouble deducing the template function from the rather explicit argument types...
What am I doing wrong.
If this should be in code review stack exchange rather than here, let me know, but I expect this to be a misunderstanding on my part about how templates work, rather than a bug.
I can't test it right now, but I think it's because in T[]
is scope const T[]
which is scope const(T[])
, which is a lot more annoying for everyone to deal with than scope const(T)[]
, and at the same time, has pretty much no benefits.
Try changing
in T x, in T[] t
to
scope T x, scope const(T)[] t
in the parameter list, to see if that solves the problem.