From: Robert Dober <robert.dober@...>
Date: 2007-10-25T06:03:08+09:00
Subject: Re: pop/push, shift/unshift

On 10/24/07, Bob Hutchison <hutch@recursive.ca> wrote:
> Hi,
>
> On 24-Oct-07, at 3:09 PM, Robert Dober wrote:
>
> > Sorry for replying to my own post, can somebody confirm please that
> > the following code proves that Array#shift does not leak?
> > Thx in advance
> > def count
> >   i = 0
> >   ObjectSpace.each_object{ i += 1 }
> >   i
> > end
> >
> > a = []
> > 10.times do
> >   10_000.times do
> >     a << "a"
> >     a.shift
> >   end
> >   ObjectSpace.garbage_collect
> >   p count
> > end
> > 381
> > 382
> > 382
> > 382
> > 382
> > 382
> > 382
> > 382
> > 382
> > 382
> >
> > Robert
> >
>
>
> Try it this way:
>
> class Thing
> end
>
> def count
>    i = 0
>    ObjectSpace.each_object(Thing){ i += 1 }
>    i
> end
>
> a = []
> 10.times do
>    10_000.times do
>      a << Thing.new
>    end
>    ObjectSpace.garbage_collect
>    10_000.times do
>      a.shift
>    end
>    ObjectSpace.garbage_collect
>    puts "count(1): #{count} size: #{a.size}"
>    a.push(nil).pop
>    ObjectSpace.garbage_collect
>    puts "count(2): #{count} size: #{a.size}"
>
>    10_000.times do
>      a << Thing.new
>    end
>    ObjectSpace.garbage_collect
>    10_000.times do
>      a[0] = nil # <<<<<< !!!!
>      a.shift
>    end
>    ObjectSpace.garbage_collect
>    puts "count(3): #{count} size: #{a.size}"
> end
>
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
> count(1): 10000 size: 0
> count(2): 0 size: 0
> count(3): 0 size: 0
>
> ----
> Bob Hutchison                  -- tumblelog at http://
> www.recursive.ca/so/
> Recursive Design Inc.          -- weblog at http://www.recursive.ca/
> hutch
> http://www.recursive.ca/       -- works on http://www.raconteur.info/
> cms-for-static-content/home/
>
>
Amazing Bob,
it has nothing to do with nil though, but with the assignment to a[0]
change a[0] = nil to a[0]=42, you get the same results.
You can even do the following

def count
  i = 0
  ObjectSpace.each_object{ i += 1 }
  i
end

a = []
  10_000.times do
    a << Object.new
  end
  ObjectSpace.garbage_collect
  10_000.times do
    a.shift
  end
  ObjectSpace.garbage_collect
  puts "count(1): #{count} size: #{a.size}"
  a.push(nil).pop
  ObjectSpace.garbage_collect
  puts "count(2): #{count} size: #{a.size}"

  10_000.times do
    a << Object.new
  end
  ObjectSpace.garbage_collect
  10_000.times do
  a[0] = Object.new
    a.shift
  end
  ObjectSpace.garbage_collect
  puts "count(3): #{count} size: #{a.size}"

count(1): 10386 size: 0
count(2): 389 size: 0
count(3): 392 size: 0

There remains one thing to be clarified, is this really a bug?
I do not think so, I just ran this until I got 200 points on my terminal :-0

a = []
loop do
 a << Array.new( 1_000_000){ Object.new }
 a.shift
 $stdout.print "."
 $stdout.flush
end

on my Zenwalk box memory is allocated and released regualrely, I doubt
if ObjectSpace is the right tool to use.


Anyway forgive me to be blunt but everybody should just use Array#shift.

Cheers
Robert