# DRbIdConv without \_id2ref

**URL:** <https://www.ruby-forum.com/t/drbidconv-without-id2ref/178141>\
**Category:** Ruby\
**Created:** [October 29, 2009, 6:08pm UTC](https://www.ruby-forum.com/t/drbidconv-without-id2ref/178141 "2009-10-29T18:08:13Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![headius](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/h/7c8e57/32.png) [@headius](https://www.ruby-forum.com/u/headius)\
**Post date:** [October 29, 2009, 6:08pm UTC](https://www.ruby-forum.com/t/drbidconv-without-id2ref/178141/1 "2009-10-29T18:08:13Z")

</div>

As mentioned in a previous thread on ruby-talk, we are looking to  
remove ObjectSpace\_id2ref from JRuby. Matz explained that it was added  
to support WeakRef, for which we have our own (Java-based)  
implementation. Supporting it on the JVM (or any VM where objects move  
in memory) is rather expensive.

The only place in stdlib that uses _id2ref_ is this class in drb/drb.rb:

class DRbIdConv

```
# Convert an object reference id to an object.
#
# This implementation looks up the reference id in the local object
# space and returns the object it refers to.
def to_obj(ref)
  ObjectSpace._id2ref(ref)
end

# Convert an object into a reference id.
#
# This implementation returns the object's __id__ in the local
# object space.
def to_id(obj)
  obj.nil? ? nil : obj. __id__
end

```

end

I have come up with a possible replacement for this that uses WeakRef.  
I’d like a few more eyes on it. It’s not concurrency-safe, which might  
be a good addition. The ‘clean’ method is also not very efficient, but  
Ruby’s WeakRef doesn’t have the notion of a reference queue you can  
poll cheaply. Adding finalizers would be a heavy-weight way to add  
support for a reference queue.

Here’s the code I’ve come up with (plus a “require ‘weakref’” near the  
top of drb/drb.rb):

class DRbIdConv  
def initialize  
@id2ref = {}  
end

```
# Convert an object reference id to an object.
#
# This implementation looks up the reference id in the local object
# space and returns the object it refers to.
def to_obj(ref)
  _get(ref)
end

# Convert an object into a reference id.
#
# This implementation returns the object's __id__ in the local
# object space.
def to_id(obj)
  obj.nil? ? nil : _put(obj)
end

def _clean
  dead = []
  @id2ref.each {|id,weakref| dead << id unless 

```

weakref.weakref\_alive?}  
dead.each {|id| @id2ref.delete(id)}  
end

```
def _put(obj)
  _clean
  @id2ref[obj. __id__] = WeakRef.new(obj)
  obj. __id__
end

def _get(id)
  weakref = @id2ref[id]
  if weakref
    result = weakref. __getobj__ rescue nil
    if result
      return result
    else
      @id2ref.delete id
    end
  end
  nil
end
private :_clean, :_put, :_get

```

end

And here is a test for it:

require ‘test/unit’  
require ‘drb’  
require ‘java’

class TestWeakDrbIdConv \< Test::Unit::TestCase  
def test\_weak\_drb\_id\_conv  
conv = DRb::DRbIdConv.new  
obj\_ary = []  
id\_ary = []

```
# populate
100.times do
  obj = Object.new
  obj_ary << obj
  id_ary << conv.to_id(obj)
end

# confirm they're there
id_ary.each do |id|
  assert conv.to_obj(id)
end

# dereference objects and force GC
obj_ary = nil
2.times {java.lang.System.gc}

# confirm they're gone
id_ary.each do |id|
  assert !conv.to_obj(id)
end

```

end  
end

- Charlie

---

<div class="post-metadata">

**Author:** ![headius](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/h/7c8e57/32.png) [@headius](https://www.ruby-forum.com/u/headius)\
**Post date:** [October 30, 2009, 1:50am UTC](https://www.ruby-forum.com/t/drbidconv-without-id2ref/178141/2 "2009-10-30T01:50:05Z")

</div>

Hi,

At Fri, 30 Oct 2009 02:07:04 +0900,  
Charles Oliver N. wrote in [ruby-core:26414]:

> As mentioned in a previous thread on ruby-talk, we are looking to  
> remove ObjectSpace\_id2ref from JRuby. Matz explained that it was added  
> to support WeakRef, for which we have our own (Java-based)  
> implementation. Supporting it on the JVM (or any VM where objects move  
> in memory) is rather expensive.

I agree. Could you show the link to that thread? I have  
stopped ruby-talk for a while.

---

<div class="post-metadata">

**Author:** ![headius](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/h/7c8e57/32.png) [@headius](https://www.ruby-forum.com/u/headius)\
**Post date:** [October 30, 2009, 6:18am UTC](https://www.ruby-forum.com/t/drbidconv-without-id2ref/178141/3 "2009-10-30T06:18:08Z")

</div>

Hello!

On Thu, Oct 29, 2009 at 7:48 PM, Nobuyoshi N. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid)  
wrote:

> I agree. Â Could you show the link to that thread? Â I have  
> stopped ruby-talk for a while.

[http://blade.nagaokaut.ac.jp/cgi-bin/vframe.rb/ruby/ruby-talk/348859?348709-349085](http://blade.nagaokaut.ac.jp/cgi-bin/vframe.rb/ruby/ruby-talk/348859?348709-349085)

You may also be interested in the JRuby commit  
7ed527822b6d2ad3e2b071376a5cd8062392ef08.

- Charlie