my @top26 = my($a, $b, $c, $d, $e, $f, $g, $h, $i, $j, $k, $l, $m, $n,
+ $o, $p, $q, $r, $s, $t, $u, $v, $w, $x, $y, $z) = sort @foo
[download]

my @top26 = top (26, @foo);
[download]

And you'd need to write tricky evals if you don't know which top N to take during compile time, but only at run time (for instance, because N is user input).

Having tracked down (part of) the p5p thread demerphq mentioned, the suggestion was that you would also be able to employ the short-cicuited sort by using a slice:

@most[ 0 .. 6 ] = sort @ary; # only sort 7 entries
[download]

I'm not familiar with how the RHS hints are derived, but that would seem to address both your concerns.

That said, I think topNbs() as posted by tall_man could be added to List::Util quite easily and would probably be easier to get accepted because of the lower risk.

Then again, it doesn't really use a List, so maybe it is time for an Array::Util package.

---

Examine what is said, not who speaks.

Silence betokens consent.

Love the truth but pardon error.

The code would need more work before it could go into List::Util, since it only works for integer values, positive only, and only for the N highest (not lowest, or any arbitrary sorting function). It's a good idea, though.

Since the basic algorithm needs only a size-N array, there ought to be some way to wrap this as a package that takes an arbitrary input set of unlimited size, not just an array ref as we have here.

my $acc = new NGreatest($n);
$acc->insert(\@arr);
while <> {
   $acc->insert(chomp $_);
}
my @top = $acc->result();
[download]

I don't suppose you recall how they were going to determine how many values to produce?

Is that information --ie. the number of values required on the right-hand side of a list assignment--generally available to XS code, or was the intention to make a special case for the construct?

IIRC the idea was to apply much the same type of logic as occurs with split. Ie the logic that implicitly sets the third argument of the split to be N+1 where N is the number of scalar slots on the LHS of the assignment. I hazzily recall discussion on whether using an explicit slice would also do the same. I think if you trawl the p5p archives for sort and optimize youll find it. I think the basic idea was that

my ($x,$y,$z)=sort @foo;
my @top=(sort @foo)[1..$n];
[download]

would be special cased somehow.

As for your question about XS, I really have no idea right now. Sorry.

---
demerphq

Note that your first construct works for split but your second would give no hints to split.

As to how this is done, I'd look at Want if I were curious.

- tye