Faster stringify_keys and symbolize_keys

hash_test.rb

require_relative "lib/active_support/core_ext/hash/keys"
require "benchmark"

[1, 10, 100, 1_000, 10_000, 100_000, 1_000_000].each do |count|
  puts
  puts '*' * 80
  puts "Hash with #{count} entries"
  puts

  strings = (1..count).map(&:to_s)
  symbols = strings.map(&:to_sym)

  hash1 = Hash[symbols.zip(strings)]
  hash2 = hash1.dup

  hash3 = Hash[strings.zip(strings)]
  hash4 = hash3.dup

  def hash2.transform_keys
    result = {}
    each_key do |key|
      result[yield(key)] = self[key]
    end
    result
  end

  def hash4.transform_keys
    result = {}
    each_key do |key|
      result[yield(key)] = self[key]
    end
    result
  end

  raise unless hash1.stringify_keys == hash2.stringify_keys
  raise unless hash3.symbolize_keys == hash4.symbolize_keys

  Benchmark.bm(17) do |x|
    x.report "original stringify" do
      hash1.stringify_keys
    end

    x.report "each_key stringify" do
      hash2.stringify_keys
    end

    x.report "original symbolize" do
      hash3.symbolize_keys
    end

    x.report "each_key symbolize" do
      hash4.symbolize_keys
    end
  end

  puts

  GC.disable

  count1 = ObjectSpace.each_object {}

  hash1.stringify_keys

  count2 = ObjectSpace.each_object {}

  hash2.stringify_keys

  count3 = ObjectSpace.each_object {}

  hash3.symbolize_keys

  count4 = ObjectSpace.each_object {}

  hash4.symbolize_keys

  count5 = ObjectSpace.each_object {}

  puts "original stringify  #{count2 - count1} objects created"
  puts "each_key stringify  #{count3 - count2} objects created"
  puts "original symbolize  #{count5 - count3} objects created"
  puts "each_key symbolize  #{count5 - count4} objects created"
end

output

grantmac:activesupport grant$ ruby --version
ruby 1.9.3p194 (2012-04-20 revision 35410) [x86_64-darwin11.3.0]
grantmac:activesupport grant$ ruby hash_test.rb 

********************************************************************************
Hash with 1 entries

                        user     system      total        real
original stringify  0.000000   0.000000   0.000000 (  0.000010)
each_key stringify  0.000000   0.000000   0.000000 (  0.000007)
original symbolize  0.000000   0.000000   0.000000 (  0.000007)
each_key symbolize  0.000000   0.000000   0.000000 (  0.000006)

original stringify  4 objects created
each_key stringify  3 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

********************************************************************************
Hash with 10 entries

                        user     system      total        real
original stringify  0.000000   0.000000   0.000000 (  0.000016)
each_key stringify  0.000000   0.000000   0.000000 (  0.000014)
original symbolize  0.000000   0.000000   0.000000 (  0.000020)
each_key symbolize  0.000000   0.000000   0.000000 (  0.000014)

original stringify  22 objects created
each_key stringify  21 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

********************************************************************************
Hash with 100 entries

                        user     system      total        real
original stringify  0.000000   0.000000   0.000000 (  0.000104)
each_key stringify  0.000000   0.000000   0.000000 (  0.000114)
original symbolize  0.000000   0.000000   0.000000 (  0.000104)
each_key symbolize  0.000000   0.000000   0.000000 (  0.000096)

original stringify  202 objects created
each_key stringify  201 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

********************************************************************************
Hash with 1000 entries

                        user     system      total        real
original stringify  0.000000   0.000000   0.000000 (  0.001030)
each_key stringify  0.000000   0.000000   0.000000 (  0.000955)
original symbolize  0.000000   0.000000   0.000000 (  0.000985)
each_key symbolize  0.000000   0.000000   0.000000 (  0.000925)

original stringify  2002 objects created
each_key stringify  2001 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

********************************************************************************
Hash with 10000 entries

                        user     system      total        real
original stringify  0.010000   0.000000   0.010000 (  0.011323)
each_key stringify  0.010000   0.000000   0.010000 (  0.011096)
original symbolize  0.010000   0.000000   0.010000 (  0.011501)
each_key symbolize  0.010000   0.000000   0.010000 (  0.011780)

original stringify  20002 objects created
each_key stringify  20001 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

********************************************************************************
Hash with 100000 entries

                        user     system      total        real
original stringify  0.150000   0.010000   0.160000 (  0.153056)
each_key stringify  0.130000   0.010000   0.140000 (  0.139755)
original symbolize  0.160000   0.000000   0.160000 (  0.161441)
each_key symbolize  0.140000   0.000000   0.140000 (  0.147776)

original stringify  200002 objects created
each_key stringify  200001 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

********************************************************************************
Hash with 1000000 entries

                        user     system      total        real
original stringify  2.460000   0.090000   2.550000 (  2.554547)
each_key stringify  2.190000   0.080000   2.270000 (  2.279136)
original symbolize  2.400000   0.040000   2.440000 (  2.429980)
each_key symbolize  2.080000   0.030000   2.110000 (  2.118653)

original stringify  2000002 objects created
each_key stringify  2000001 objects created
original symbolize  3 objects created
each_key symbolize  1 objects created

Motive?

To speed up Rails for everyone, everywhere.

rails/rails#6886

Interesting, I run this locally on my machine and consistently get results that show the original is faster.

How about object creation? I think that's more important in the long term than the speed. Less GC over time.