ruby-marc-serialization-benchmarks.rb

# Code to benchmark various serializations of MARC records using ruby-marc
# Not included is XML -- serialization using ruby-marc is ridiculously slow and the # filesizes are bigger than anything else. Even with the lib-xml reader, 
# deserialization is also relatively slow
#
# I didn't bother to benchmark json/pure in later runs because it's just so damn
# slow that it would never be a good choice. 
#
# My results can be found at http://robotlibrarian.billdueber.com/sizespeed-of-various-marc-serializations-using-ruby-marc/

require 'marc'
require 'rubygems'
require 'benchmark'
require 'yajl'
require 'msgpack'


jsonsize = 0.0
marcsize = 0.0
mpsize   = 0.0


# Use Benchmark.measure
# sjptime =  Benchmark::Tms.new(0,0,0,0,0, "JSON Pure")
sytime =  Benchmark::Tms.new(0,0,0,0,0, "YAJL")
smtime =  Benchmark::Tms.new(0,0,0,0,0, "MARC")
smptime =  Benchmark::Tms.new(0,0,0,0,0, "Msgpack")

dytime =  Benchmark::Tms.new(0,0,0,0,0, "YAJL")
# djptime =  Benchmark::Tms.new(0,0,0,0,0, "JSON/Pure")
dmtime =  Benchmark::Tms.new(0,0,0,0,0, "MARC")
dmptime =  Benchmark::Tms.new(0,0,0,0,0, "Msgpack")



i = 0
iterations = 20

iterations.times do
  reader = MARC::XMLReader.new('topics.xml', :parser=>'libxml')
  
  reader.each_with_index do |r, i|
   marc = nil
   json = nil
   mp   = nil
 

   smtime += Benchmark.measure {marc = r.to_marc}
   dmtime += Benchmark.measure {copy = MARC::Record.new_from_marc(marc)}

   smptime += Benchmark.measure{mp = MessagePack.pack(r.to_hash)}
   dmptime += Benchmark.measure {copy = MARC::Record.new_from_hash(MessagePack.unpack(mp))}

   sytime += Benchmark.measure {json = Yajl::Encoder.encode(r.to_hash)}
   dytime += Benchmark.measure {copy = MARC::Record.new_from_hash(Yajl::Parser.parse(json))}

   # sjptime += Benchmark.measure {json = JSON.generate(r.to_hash)}
   # djptime += Benchmark.measure {copy = MARC::Record.new_from_hash(JSON.parse(json))}
 
 
   # marcsize += marc.size
   # jsonsize += json.size
   # mpsize += mp.size
 
   # break if i > 1000

  end
end

puts "Total of #{i} records run #{iterations} times"

puts "\nSERIALIZING"

base = smtime.total
puts '  %-15s %8.2f s (%3.0f%%)' % ['MARC', smtime.total, smtime.total / base * 100]
puts '  %-15s %8.2f s (%3.0f%%)' % ['YAJL', sytime.total, sytime.total / base * 100]
puts '  %-15s %8.2f s (%3.0f%%)' % ['Msgpack', smptime.total, smptime.total / base * 100]
#puts '  %-15s %8.2f s (%3.0f%%)' % ['Json Pure', sjptime.total, sjptime.total / base * 100]

base = dmtime.total
puts "\nDESERIALIZING"
puts '  %-15s %8.2f s (%3.0f%%)' % ['MARC', dmtime.total, dmtime.total / base * 100]
puts '  %-15s %8.2f s (%3.0f%%)' % ['YAJL', dytime.total, dytime.total / base * 100]
puts '  %-15s %8.2f s (%3.0f%%)' % ['Msgpack', dmptime.total, dmptime.total / base * 100]
#puts '  %-15s %8.2f s (%3.0f%%)' % ['Json Pure', djptime.total, djptime.total / base * 100]

base = dmtime.total + smtime.total
puts "\nSERIALIZE + DESERIALIZE"
puts '  %-15s %8.2f s (%3.0f%%)' % ['MARC', dmtime.total + smtime.total, (dmtime.total + smtime.total) / base * 100]
puts '  %-15s %8.2f s (%3.0f%%)' % ['YAJL', dytime.total + sytime.total, (dytime.total + sytime.total) / base * 100]
puts '  %-15s %8.2f s (%3.0f%%)' % ['Msgpack', dmptime.total + smptime.total, (dmptime.total + smptime.total) / base * 100]
#puts '  %-15s %8.2f s (%3.0f%%)' % ['Json Pure', djptime.total + sjptime.total, (djptime.total + sjptime.total) / base * 100]

# puts "\nSIZE"
# puts '  %-15s %8.2f MBytes (%3.0f%%)' % ['MARC', marcsize /  (1024.0 * 1024), marcsize / (1.0 * marcsize) * 100]
# puts '  %-15s %8.2f MBytes (%3.0f%%)' % ['Msgpack', mpsize /  (1024.0 * 1024), mpsize / (1.0 * marcsize) * 100]
# puts '  %-15s %8.2f MBytes (%3.0f%%)' % ['JSON', jsonsize /  (1024.0 * 1024), jsonsize / (1.0 * marcsize) * 100]
# puts '  %-15s %8.2f MBytes (%3.0f%%)' % ['XML', 97958300 /  (1024.0 * 1024),  97958300.0 / marcsize * 100]
#