Some interesting benchmark results for various ways of doing typical Ruby things

benchmarks.rb

require 'benchmark'
REP = 1000000

CACHED_REGEX = /mat[ac]?h/
OBJECT_REGEX = Regexp.new('mat[ac]?h')

Benchmark.bm 20 do |x|

  x.report 'Regex interpolated string' do
    REP.times do |index|
      word = 't'
      /ma#{word}[ac]?h/.match "Some math string"
    end
  end

  x.report 'Regex interpolated string with o flag' do
    REP.times do |index|
      word = 't'
      /ma#{word}[ac]?h/o.match "Some math string"
    end
  end

  x.report 'Regex literal' do
    REP.times do |index|
      /mat[ac]?h/.match "Some math string"
    end
  end

  x.report 'Regex cached' do
    REP.times do |index|
      CACHED_REGEX.match "Some math string"
    end
  end

  x.report 'Regex instantiation' do
    REP.times do |index|
      Regexp.new('mat[ac]?h').match "Some math string"
    end
  end

  x.report 'Regex object' do
    REP.times do |index|
      OBJECT_REGEX.match "Some math string"
    end
  end

  x.report 'Regexp out an end =, short string' do
    REP.times do |index|
      str = 'somestring='
      str.gsub(/\=$/)
    end
  end

  x.report 'Chomp out an end =, short string' do
    REP.times do |index|
      str = 'somestring='
      str.chomp('=')
    end
  end

  x.report 'Index out an end =, short string' do
    REP.times do |index|
      str = 'somestring='
      str = str[0..-2] if str[-1]=='='
    end
  end

  x.report 'Regexp out an end =' do
    REP.times do |index|
      str = 'somelongstringsomelongstringsomelongstring='
      str.gsub(/\=$/)
    end
  end

  x.report 'Chomp out an end =' do
    REP.times do |index|
      str = 'somelongstringsomelongstringsomelongstring='
      str.chomp('=')
    end
  end

  x.report 'Index out an end =' do
    REP.times do |index|
      str = 'somelongstringsomelongstringsomelongstring='
      str = str[0..-2] if str[-1]=='='
    end
  end
end

#                           user     system      total        real
# Regex interpolated string  7.810000   0.010000   7.820000 (  7.806186)
# Regex interpolated string with o flag  1.550000   0.000000   1.550000 (  1.548230)
# Regex literal         1.360000   0.000000   1.360000 (  1.361472)
# Regex cached          1.350000   0.000000   1.350000 (  1.354619)
# Regex instantiation   6.320000   0.010000   6.330000 (  6.321500)
# Regex object          1.410000   0.000000   1.410000 (  1.417464)
# Regexp out an end =, short string  0.510000   0.000000   0.510000 (  0.511453)
# Chomp out an end =, short string  0.420000   0.000000   0.420000 (  0.418580)
# Index out an end =, short string  0.530000   0.000000   0.530000 (  0.533264)
# Regexp out an end =   0.470000   0.000000   0.470000 (  0.476832)
# Chomp out an end =    0.600000   0.000000   0.600000 (  0.595940)
# Index out an end =    0.690000   0.000000   0.690000 (  0.696599)

Things that are unexpectedly expensive:
/#{interpolation}/ and Regexp.new, more than /a literal/
String indexing via a Range being the slowest of all
Chomp not actually being much faster at all
Strings over 15 characters seem to result in an extra performance hit