On Friday, 14 June 2013 at 10:07:13 UTC, Andrea Fontana wrote:
string[] test_filter(string[] words)
{
static blackList =
[
"d", "c", "e", "a", "è", "é", "e"
].sort();
return words.filter!((a) =>
!blackList.assumeSorted.contains(a)).array;
}
test_filter(["a", "b", "test", "hello"]).writeln;
This code crash! It's just a useless trimmed-down version of a
more complex code, just to show you the bug.
If you remove accented letters from "blacklist" array it works
fine. Why?
sort() requires UTF32 input (a dstring/dchar[]) or it will fail.
UTF8 (i.e. string type in D) is a variable-length encoding and if
such input is given to sort it would refuse to compile since it
cannot sort it in-place (a guarantee sort() makes).
I guess that in case UTF8 input with 1-byte only characters is
given to sort() it can detect that there are no variable length
characters and that is probably the reason your code works when
you remove non-ascii characters.
