How fast is `reverse!` technique.

benchmark.rb

# conversation: https://github.com/rack/rack/commit/734a00c5f4bb46e9a5e6e2677d89a2f285dcc185
#
# user     system      total        real
# reverse:   0.010000   0.010000   0.020000 (  0.024657)
# join:      0.050000   0.010000   0.060000 (  0.056571)
# regexp:    0.050000   0.000000   0.050000 (  0.048339)

require 'benchmark'

def join(session_data)
  *data, digest = session_data.split('--')
  session_data = data.join('--')
  [session_data, digest]
end

def reverse(session_data)
  digest, session_data = session_data.reverse.split('--', 2)
  digest.reverse! if digest
  session_data.reverse! if session_data
  [session_data, digest]
end

REGEXP = /\A(.*)--(.*)\z/m
def regexp(session_data)
  session_data, digest = REGEXP.match(session_data).captures
  [session_data, digest]
end

n = 1000
data = (1..100).to_a.map!{|i| i.to_s * 100}.join('--')

Benchmark.bm(8) do |x|
  x.report('reverse:') { n.times { reverse(data) } }
  x.report('join:') { n.times { join(data) } }
  x.report('regexp:') { n.times { regexp(data) } }
end