Optimized Array#skip

[asterite]

Enumerable#skip doesn't know the size of the resulting array, plus it's much efficient to do a mem-copy here.

A benchmark I did:

require "benchmark"

a = Array.new(10_000) { |i| i }

Benchmark.ips do |x|
  x.report("old skip") do
    a.skip(1)
  end
  x.report("new skip 2") do
    a.skip2(1)
  end
  x.report("new skip 3") do
    a.skip3(1)
  end
end

where skip is the current one, skip2 just called [count..-1] and skip3 is the version in this PR.

Results:

  old skip  15.75k ( 63.48µs) (± 3.89%)  99072 B/op   2.62× slower
  new skip 2 41.28k ( 24.23µs) (± 6.37%)  40032 B/op   1.00× slower
new skip 3  41.31k ( 24.21µs) (± 6.05%)  40032 B/op        fastest

[asterite]

BTW, maybe we should rename skip to drop, it reads better. Also for Iterator... I'm thinking that if we want a bang version for this, skip! reads badly, even skip reads badly already.

[yxhuvud]

"BTW, maybe we should rename skip to drop, it reads better."

Also, skip is not an antonym to the opposite, ie take. drop works better from that view too.

[straight-shoota]

Shouldn't this better repeat the description of Enumerable#skip?

[asterite]

Yes. Actually, we should have some thingy to inherit doc from the parent. Let's work on that next (there are already a few methods in Array that say they are an optimized version of something else)

[bew]

For now I think it's better indeed, but it would be nice to have a :previous_doc: and/or :super_doc: to automatically copy the doc from previous def or super def or that same method

[asterite]

I just copied the doc.

[asterite]

@yxhuvud Yeah, take is something else we need. Right now we have first. The problem is that if we have take and first they will be aliases... maybe it's not a big deal. Maybe having aliases is not that bad after all.

[yxhuvud]

Ah right. We have first instead of take, but still have take_while. That is a bit confusing.