On Tuesday, 19 March 2013 at 11:46:14 UTC, monarch_dodra wrote:
On Tuesday, 19 March 2013 at 10:08:43 UTC, Jonathan M Davis
wrote:
On Tuesday, March 19, 2013 09:25:43 timotheecour wrote:
we need a std.algorithm.equalRecurse(T1,T2)(T1 a, T2 b) that
compares recursively a and b;
its behavior should be:
if opEqual is defined, call it
else, if its a range, call std.algorithm.equal (ie compare nb
elements, then each element for equality)
else, if it's a class/struct, make sure types are same and
call
it recursively on each field.
else if it's a numerical type, call "=="
else (is there an else?)
just as std.algorithm.equal, we should have
equalRecurse([1],[1.0]);
If you want recursive equal, then do equal!equal. Granted,
that's only one
level of recursion, but how many levels deep are you really
going to have your
ranges? And you have to get to == eventually anyway in order
to compare the
deepest elements. Going beyond a range of ranges is likely to
be quite rare,
and when it does happen, you can simply nest equal as many
times as you need.
- Jonathan M Davis
"equal!equal(RoR1, RoR2)"
That looks cute, but I think it says something about how
powerful and expressive D can be, while being compile-time
optimized. It's those little things that still amaze me about D.
and then:
template NumDimensions (T) {
static if(is(ElementType!T == void))
const NumDimensions = 0;
else
const NumDimensions = 1 + NumDimensions!(ElementType!T);
}
bool rec_equal(R0, R1)(R0 r0, R1 r1)
if(NumDimensions!R0 == NumDimensions!R1)
{
mixin("return " ~ replicate("equal!", NumDimensions!(R0)-1) ~
"equal(r0, r1);");
}
obviously it requires some more checks, but it works nicely
(except if you feed it two integer literals, in which case the
compiler throws an out of memory error!).