Ah, the solution from the FAQ. Unfortunally, the FAQ is wrong.
Here's one array where the FAQ solution will give the wrong
answer:
@a = (undef, "");
Ooops. The FAQ assumes the values map to unique strings. But
that isn't true, the undefined value maps to an empty string.
And then I haven't shown examples with references, tied variables
or objects with overloaded stringification that all can cause
the FAQ solution to give the wrong results.
Abigail | [reply]
[d/l] |
The problem is how to define equity once you start slinging around tied or overloaded objects. By equal value, by equal reference, or by entirely different semantics? It's a question with very individual replies depending on the situation, and there can't be a universal solution. As for the undef, which is actually a common case, you are right, I had forgotten about that, even though I've fixed it before.
my @new = do{
my (%seen, $seen_undef);
grep defined ? !$seen{$_}++ : !$seen_undef++, @a, @b, @c, @d;
};
Makeshifts last the longest. | [reply]
[d/l] |
my @new = do{local %h; @h{@a,@b,@c,@d} = (@a,@b,@c,@d); values %h};
Adapted from BrowserUk's code
--
integral, resident of freenode's #perl
| [reply]
[d/l] |
That doesn't solve the problem at all. It fails for exactly
the same reason:
$ perl -le '@a = (undef, ""); @u = do {@h {@a} = @a; values %h}; print
+ scalar @u'
1
$
Abigail | [reply]
[d/l] |