Did you test this with the sub{}-example I provided in order to show context dependency?

Nope. I know it does exactly what you said it does. What it doesn't do is show what jdporter's solution has a problem. That would surely involve using jdporter's solution.

where would preservation of %_ (jdporter mentioned $_) usually be of some importance?

jdporter mentioned *_, not $_.

And localizing %_ would be "of some importance" when %_ is changed, like in the parent of the post to which jdporter replied. The same applies to all global variables.

ikegami:  I know it does exactly what you said it does

Then I'll be glad to give some enlightment to you:


sub uinq {
   @_ = do { @$_{@_}=(); keys %$_ };
   @_
} 

my @array = (1,1,3,3,2,3,5,7,5,2);

%_ = ( f=>'oo', b=>'ar' );
print "@{[ uinq @array ]}\n";
print map "\t$_ => $_{$_}\n", keys %_;
[download]


ikegami:  jdporter mentioned *_, not $_.

Better read back again then: http://perlmonks.org/?node_id=640777

Regards & thanks
mwa

Oh, I see, sorry.

Actually, both jdporter and I got it wrong.
You don't modify $_ like jdporter thought.
You don't modify %_ like I thought.

Wow, that piece of code is so incredibly bad! You modify some random variable based on the current contents of $_. Whether it'll work or not is random. Most of the time, it won't.

for (
   [qw(a a b c)],
   [qw(d e e f)],
) {
   print(uinq(@$_), "\n");    # Can't coerce array into hash
}
[download]

for ('a a b c', 'd e e f') {
   print(uinq(split), "\n");  # Can't use string ("a a b c") as a HASH
+ ref while "strict refs" in use
}
[download]

Adding local $_; causes the a specific variable to be used. That alieviates some of the problem, but that variable needs to be localized too. Why not just use %_?

As an expression:

do { local %_; @_{@_}=(); keys %_ }
[download]

As a sub:

sub uinq { local %_; @_{@_}=(); keys %_ }
[download]