On Sat, 22 Oct 2011 23:01:17 -0400, bearophile <bearophileh...@lycos.com>
wrote:
Thanks for all the answers.
Steven Schveighoffer:
a ~ b should technically be assignable to char[], since it's alread new
memory. We may yet get there with pure functions being able to implicit
cast to immutable.
Isn't that kind of the opposite?
No, if a ~ b would call a pure function that returns mutable when
possible, then you could use it for mutable, immutable, or const.
Is this already in Bugzilla?
Somewhat, although I like using the pure function avenue to ensure
immutable does not violate const than the crazy rules we came up with here:
http://d.puremagic.com/issues/show_bug.cgi?id=1654
Note this was in the infancy of my D knowledge ;)
Some versions I have written, there is no very good solution, it seems:
import std.algorithm, std.exception;
void main() {
string a = "red";
string b = "green";
int[string] aa;
// Works in D1.
// aa[(a ~ b).sort] = 1;
// Works in D2, but will stop working.
aa[(a.dup ~ b.dup).sort.idup] = 1;
// Works in D2, but will stop working.
char[] ab = (a.dup ~ b.dup).sort;
aa[assumeUnique(ab)] = 2;
// If keys don't actually need to be strings
// the code gets a bit simpler.
int[immutable(ubyte[])] aa2;
// not good
// aa2[assumeUnique(sort(cast(ubyte[])(a ~ b)))] = 3;
// not good?
// aa2[assumeUnique(sort(cast(ubyte[])(a ~ b)).release())] = 3;
// just 1 copy, two visible casts, 3 lines
auto ab4 = cast(ubyte[])(a ~ b);
ab4.sort();
aa2[cast(immutable(ubyte[]))ab4] = 3;
// a visible cast and an hidden cast
ubyte[] ab2 = cast(ubyte[])(a ~ b);
ab2.sort();
aa2[assumeUnique(ab2)] = 3;
// 2 lines, more complex looking
auto ab3 = sort(cast(ubyte[])(a ~ b)).release();
aa2[assumeUnique(ab3)] = 3;
}
I honestly am not sure your specific problem is solvable without a lot of
special cases. It may be you have to write a sub-function that hides the
mess.
-Steve
