# or using a slice (might be more efficient, but doesn't read as nicel
+y)
sub hash_slice {
    my ($hr, @k) = @_;
    my %ret;
    @ret{@k} = @$hr{@k};
    return %ret;
}
[download]

Personally, I think this one reads very nicely, and this is most likely the approach I would go with. Nice side-effect: in case I ever need to step through with "perl -d" (e.g. to check for mangled data), this versions provides a handy place for a breakpoint where I can dump %ret before it gets returned.