Benchmarking some alternate implementations for checking if a BOM is present in a string and removing it.

benchmark.rb

require 'benchmark'

# you'll want this to be a largeish file that is representative of the kinds
# of CSS files you'll be using this on in the real world.  I used a compiled
# version of application.css from https://github.com/alphagov/government-frontend
css = File.binread "bom.css"

class BomBench
  def remove_bom_force_encoding_slice!(input)
    input.force_encoding('UTF-8').slice!(0)
    input
  end

  def remove_bom_force_encoding_slice(input)
    input.force_encoding('UTF-8').slice(1..-1)
  end

  def remove_bom_force_encoding_char_index(input)
    input.force_encoding('UTF-8')[1..-1]
  end

  def remove_bom_pack(input)
    input.bytes[3..-1].pack('c*').force_encoding('UTF-8')
  end

  def remove_bom_force_encoding_sub(input)
    input.force_encoding('UTF-8').sub(/^\xEF\xBB\xBF/, '')
  end

  def remove_bom_force_encoding_sub!(input)
    input.force_encoding('UTF-8').sub!(/^\xEF\xBB\xBF/, '')
  end

  def remove_bom_dup_force_encoding_slice!(input)
    out = input.dup.force_encoding('UTF-8')
    out.slice!(0)
    out
  end

  def remove_bom_dup_force_encoding_slice(input)
    out = input.dup.force_encoding('UTF-8')
    out.slice(1..-1)
    out
  end

  def remove_bom_dup_force_encoding_char_index(input)
    input.dup.force_encoding('UTF-8')[1..-1]
  end

  def remove_bom_dup_force_encoding_sub(input)
    input.dup.force_encoding('UTF-8').sub(/^\xEF\xBB\xBF/, '')
  end

  def remove_bom_dup_force_encoding_sub!(input)
    input.dup.force_encoding('UTF-8').sub!(/^\xEF\xBB\xBF/, '')
  end

  def bom_string_force_encoding_reset_full_regexp?(input)
    with_encoding_force_encoding_reset(input, 'UTF-8') { |utf_8_input| utf_8_input =~ /^\xEF\xBB\xBF/ }
  end

  def bom_string_dup_force_encoding_full_regexp?(input)
    with_encoding_dup_force_encoding(input, 'UTF-8') { |utf_8_input| utf_8_input =~ /^\xEF\xBB\xBF/ }
  end

  def bom_string_force_encoding_reset_slice_regexp?(input)
    with_encoding_force_encoding_reset(input, 'UTF-8') { |utf_8_input| utf_8_input[0..2] =~ /^\xEF\xBB\xBF/ }
  end

  def bom_string_dup_force_encoding_slice_regexp?(input)
    with_encoding_dup_force_encoding(input, 'UTF-8') { |utf_8_input| utf_8_input[0..2] =~ /^\xEF\xBB\xBF/ }
  end

  def bom_string_force_encoding_reset_slice_equality?(input)
    with_encoding_force_encoding_reset(input, 'UTF-8') { |utf_8_input| utf_8_input[0] == "\xEF\xBB\xBF" }
  end

  def bom_string_dup_force_encoding_slice_equality?(input)
    with_encoding_dup_force_encoding(input, 'UTF-8') { |utf_8_input| utf_8_input[0] == "\xEF\xBB\xBF" }
  end

  def bom_string_byte_check?(input)
    with_encoding_byte_check(input, 'UTF-8') { |x| x.bytes[0..2] == [0xEF, 0xBB, 0xBF] }
  end

  def with_encoding_force_encoding_reset(string, encoding)
    old_encoding = string.encoding
    begin
      return (yield string.force_encoding(encoding))
    ensure
      string.force_encoding(old_encoding)
    end
  end

  def with_encoding_dup_force_encoding(string, encoding)
    return (yield string.dup.force_encoding(encoding))
  end

  def with_encoding_byte_check(string, _encoding)
    yield string
  end
end

how_many = 10_000
compressor = BomBench.new

Benchmark.bmbm do |x|
  x.report("check: bytes") { how_many.times { compressor.bom_string_byte_check?(css.dup) } }
  x.report("check: dup + force_encoding + slice equality") { how_many.times { compressor.bom_string_dup_force_encoding_slice_equality?(css.dup) } }
  x.report("check: dup + force_encoding + slice regexp") { how_many.times { compressor.bom_string_dup_force_encoding_slice_regexp?(css.dup) } }
  x.report("check: dup + force_encoding + full regexp") { how_many.times { compressor.bom_string_dup_force_encoding_full_regexp?(css.dup) } }
  x.report("check: force_encoding + reset + slice equality") { how_many.times { compressor.bom_string_force_encoding_reset_slice_equality?(css.dup) } }
  x.report("check: force_encoding + reset + slice regexp") { how_many.times { compressor.bom_string_force_encoding_reset_slice_regexp?(css.dup) } }
  x.report("check: force_encoding + reset + full regexp") { how_many.times { compressor.bom_string_force_encoding_reset_full_regexp?(css.dup) } }
end

Benchmark.bmbm do |x|
  x.report("remove bom: pack") { how_many.times { compressor.remove_bom_pack(css.dup) } }

  x.report("remove bom: force encoding + slice!") { how_many.times { compressor.remove_bom_force_encoding_slice!(css.dup) } }
  x.report("remove bom: force encoding + slice") { how_many.times { compressor.remove_bom_force_encoding_slice(css.dup) } }
  x.report("remove bom: force encoding + char index") { how_many.times { compressor.remove_bom_force_encoding_char_index(css.dup) } }
  x.report("remove bom: force encoding sub!") { how_many.times { compressor.remove_bom_force_encoding_sub!(css.dup) } }
  x.report("remove bom: force encoding sub") { how_many.times { compressor.remove_bom_force_encoding_sub(css.dup) } }

  x.report("remove bom: dup + force encoding + slice!") { how_many.times { compressor.remove_bom_dup_force_encoding_slice!(css.dup) } }
  x.report("remove bom: dup + force encoding + slice") { how_many.times { compressor.remove_bom_dup_force_encoding_slice(css.dup) } }
  x.report("remove bom: dup + force encoding + char index") { how_many.times { compressor.remove_bom_dup_force_encoding_char_index(css.dup) } }
  x.report("remove bom: dup + force encoding sub!") { how_many.times { compressor.remove_bom_dup_force_encoding_sub!(css.dup) } }
  x.report("remove bom: dup + force encoding sub") { how_many.times { compressor.remove_bom_dup_force_encoding_sub(css.dup) } }
end