Bug 10046 – Wrong insertion of Tuple in associative array

import std.stdio: writeln; import std.typecons: Tuple; void main () { alias Foo = Tuple!string; int[Foo] aa; auto ka = Foo("foo".idup); aa[ka] = 1; auto kb = Foo("foo".idup); aa[kb] = 2; aa.writeln; } Output dmd 2.063beta: [Tuple!(string)("foo"):1, Tuple!(string)("foo"):2] Expected output: [Tuple!(string)("foo"):2]

Very annoying bug since it basically kills "multiindexing" using associative arrays. As far as I know it's simply because Tuple hasn't implemented toHash. It can be worked around by using an structure with a toHash function (WARNING: The documentation on associative arrays says it's called opHash. It's called toHash!). Here's my drop-in replacement for 2-tuples. Feel free to use it freely for whatever you want. It works in my (string, string) case but I haven't tested it for other types, but it should work. **** // Workaround for http://d.puremagic.com/issues/show_bug.cgi?id=10046 static struct MyTuple(A, B) { private A first; private B second; auto ref opIndex(size_t index) { switch (index) { case 0: return first; case 1: return second; default: assert(0); } } const hash_t toHash() { return typeid(A).getHash(&first) ^ typeid(B).getHash(&second); } const bool opEquals(ref const MyTuple!(A, B) s) { return first == s.first && second == s.second; } const int opCmp(ref const MyTuple!(A, B) s) { if (first != s.first) return (first < s.first) ? -1 : 1; if (second != s.second) return (second < s.second) ? -1 : 1; return 0; } }

To rephrase more drastically: std.typecons.Tuple is incompatible with associative arrays because it does not implement toHash and opCmp.

Can we get a pull for this?

*** This issue has been marked as a duplicate of issue 11037 ***