Parallel assignments vs. Serial assigments

Hey all

I was always under the impression that parallel assignments in Ruby
were faster than assigning variables individually.

Recently I was curious to see how much faster it was and decided to
test it:

class One

  def initialize(first_name, last_name)
    @first_name = first_name
    @last_name = last_name
  end

end

class Two

  def initialize(first_name, last_name)
    @first_name, @last_name = first_name, last_name
  end

end
require "rubygems"
require "benchmark"

Benchmark.bmbm do |test|
  test.report("serial") do
    10000.times { |n| var = One.new("gavin#{n}", "morrice")}
  end
  test.report("parallel") do
    10000.times { |n| var = Two.new("gavin#{n}", "morrice")}
  end
end

The results I get show that it's slower (in both Ruby 1.8.7 and Ruby
1.9.1)

Can anyone elaborate?

Thanks

(My results show that parallel assignment is slower in this test)

···

On Feb 15, 4:23 pm, Gavin <thinkersplaygro...@googlemail.com> wrote:

Hey all

I was always under the impression that parallel assignments in Ruby
were faster than assigning variables individually.

Recently I was curious to see how much faster it was and decided to
test it:

class One

def initialize(first_name, last_name)
@first_name = first_name
@last_name = last_name
end

end

class Two

def initialize(first_name, last_name)
@first_name, @last_name = first_name, last_name
end

end
require "rubygems"
require "benchmark"

Benchmark.bmbm do |test|
test.report("serial") do
10000.times { |n| var = One.new("gavin#{n}", "morrice")}
end
test.report("parallel") do
10000.times { |n| var = Two.new("gavin#{n}", "morrice")}
end
end

The results I get show that it's slower (in both Ruby 1.8.7 and Ruby
1.9.1)

Can anyone elaborate?

Thanks

I think it might be because the parallel assigment creates an array
under the hood?

class One

def initialize(first_name, last_name)
   @first_name = first_name
   @last_name = last_name
end

end

class Two

def initialize(first_name, last_name)
   @first_name, @last_name = first_name, last_name
end
end

class Three

def initialize(*args)
   @first_name, @last_name = *args
end
end

require "rubygems"
require "benchmark"

GC.disable

puts "Arrays before serial: #{ObjectSpace.each_object(Array){}}"
10000.times { |n| var = One.new("gavin#{n}", "morrice")}
puts "Arrays after serial: #{ObjectSpace.each_object(Array){}}"
10000.times { |n| var = Two.new("gavin#{n}", "morrice")}
puts "Arrays after parallel: #{ObjectSpace.each_object(Array){}}"

$ ruby test_parallel_assignment.rb
Arrays before serial: 3589
Arrays after serial: 3589
Arrays after parallel: 13589

Jesus.

···

On Mon, Feb 15, 2010 at 5:30 PM, Gavin <thinkersplayground@googlemail.com> wrote:

(My results show that parallel assignment is slower in this test)

On Feb 15, 4:23 pm, Gavin <thinkersplaygro...@googlemail.com> wrote:

Hey all

I was always under the impression that parallel assignments in Ruby
were faster than assigning variables individually.

Recently I was curious to see how much faster it was and decided to
test it:

class One

def initialize(first_name, last_name)
@first_name = first_name
@last_name = last_name
end

end

class Two

def initialize(first_name, last_name)
@first_name, @last_name = first_name, last_name
end

end
require "rubygems"
require "benchmark"

Benchmark.bmbm do |test|
test.report("serial") do
10000.times { |n| var = One.new("gavin#{n}", "morrice")}
end
test.report("parallel") do
10000.times { |n| var = Two.new("gavin#{n}", "morrice")}
end
end

The results I get show that it's slower (in both Ruby 1.8.7 and Ruby
1.9.1)

Can anyone elaborate?

I have a class Three cause I was also testing this other form, which
also creates arrays:

GC.disable
puts "Arrays before serial: #{ObjectSpace.each_object(Array){}}"
10000.times { |n| var = One.new("gavin#{n}", "morrice")}
puts "Arrays after serial: #{ObjectSpace.each_object(Array){}}"
10000.times { |n| var = Two.new("gavin#{n}", "morrice")}
puts "Arrays after parallel: #{ObjectSpace.each_object(Array){}}"
10000.times { |n| var = Three.new("gavin#{n}", "morrice")}
puts "Arrays after parallel with array: #{ObjectSpace.each_object(Array){}}"

$ ruby test_parallel_assignment.rb
Arrays before serial: 3589
Arrays after serial: 3589
Arrays after parallel: 13589
Arrays after parallel with array: 23589

Jesus.

···

2010/2/15 Jesús Gabriel y Galán <jgabrielygalan@gmail.com>:

I think it might be because the parallel assigment creates an array
under the hood?

class One

def initialize(first_name, last_name)
@first_name = first_name
@last_name = last_name
end

end

class Two

def initialize(first_name, last_name)
@first_name, @last_name = first_name, last_name
end
end

class Three

def initialize(*args)
@first_name, @last_name = *args
end
end

However it is done technically, parallel assignment needs more space because it has to evaluate *all* right hand sides before doing any assignments. Otherwise swapping would not be possible

a, b = b, a

So, yes, it's likely an Array under the hood but even if not the parallel assignment of two variables needs to store two object references while sequential assignments of an arbitrary number of elements gets away with space for a single reference (if you need it at all).

Kind regards

  robert

···

On 02/15/2010 05:50 PM, Jesús Gabriel y Galán wrote:

2010/2/15 Jesús Gabriel y Galán <jgabrielygalan@gmail.com>:

I think it might be because the parallel assigment creates an array
under the hood?

--
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/

I think it might be because the parallel assigment creates an array
under the hood?

However it is done technically, parallel assignment needs more space because
it has to evaluate *all* right hand sides before doing any assignments.
Otherwise swapping would not be possible

a, b = b, a

Good point.

So, yes, it's likely an Array under the hood

Yep, that was my thought and that's why I checked with ObjectSpace,
that indeed shows Arrays being created.
It was easier for me than checking the implementation :-).

Jesus.

···

On Tue, Feb 16, 2010 at 7:01 PM, Robert Klemme <shortcutter@googlemail.com> wrote:

On 02/15/2010 05:50 PM, Jesús Gabriel y Galán wrote:

2010/2/15 Jesús Gabriel y Galán <jgabrielygalan@gmail.com>: