Use each_with_object instead of Hash[map]

[krzysiek1507]

No description provided.

[kpheasey]

@krzysiek1507 Is this a performance optimization?

[kpheasey]

@krzysiek1507 Is this a performance optimization?

[krzysiek1507]

Yes.

# frozen_string_literal: true

require 'benchmark/ips'
require 'benchmark/memory'

class DeepSymbolize
  class << self
    def deep_symbolize(collection)
      if collection.is_a? Hash
        Hash[collection.map do |k, v|
          [k.to_sym, deep_symbolize(v)]
        end]
      elsif collection.is_a? Array
        collection.map { |i| deep_symbolize(i) }
      else
        collection.to_sym
      end
    end
  end
end

class DeepSymbolizeOptimized
  class << self
    def deep_symbolize(collection)
      if collection.is_a? Hash
        collection.each_with_object({}) do |(k, v), hsh|
          hsh[k.to_sym] = deep_symbolize(v)
        end
      elsif collection.is_a? Array
        collection.map { |i| deep_symbolize(i) }
      else
        collection.to_sym
      end
    end
  end
end

hsh = { 'a' => 'v', 'b' => { 'b2' => {} } }
hsh = { :a => 'v', :b => { :b2 => {} }, c: [{ c1: '' }] }

[:ips, :memory].each do |type|
  puts "#{type} benchmark:"
  puts

  Benchmark.public_send(type) do |x|
    x.report('original') { DeepSymbolize.deep_symbolize hsh }
    x.report('patch') { DeepSymbolizeOptimized.deep_symbolize hsh }

    x.compare!
  end
end

ruby 2.6.3p62 (2019-04-16 revision 67580) [x86_64-darwin18]
ips benchmark:

Warming up --------------------------------------
            original    21.332k i/100ms
               patch    21.833k i/100ms
Calculating -------------------------------------
            original    217.069k (± 7.3%) i/s -      1.088M in   5.040936s
               patch    264.518k (± 9.4%) i/s -      1.332M in   5.085691s

Comparison:
               patch:   264518.5 i/s
            original:   217068.7 i/s - 1.22x  slower

memory benchmark:

Calculating -------------------------------------
            original     1.328k memsize (     0.000  retained)
                        14.000  objects (     0.000  retained)
                         0.000  strings (     0.000  retained)
               patch     1.168k memsize (     0.000  retained)
                        10.000  objects (     0.000  retained)
                         0.000  strings (     0.000  retained)

Comparison:
               patch:       1168 allocated
            original:       1328 allocated - 1.14x more