ObjectSpace searching

ruby-talk
1

I wrote some code that searches ObjectSpace prior to initializing an
object to see if one of that type and some attributes is already
present and uses that instead of initializing. (yes it's active record
and i am trying to avoid stale object errors). How horrible is this? I
am searching by specific class name so in practice it ends up looking
at at most a couple of instances but is there something inherently
wrong with doing this?

def find_in_space(search_id, search_version = nil)
  retval = nil
  ObjectSpace.each(Person){|person|
      retval = person if (p.id == search_id) &&
                               search_version &&
                               p.lock_version == search_version
  }
  return retval || Person.find(search_id)
end

Mark

2

For one, you could fail early:

s = 'foo'
obj = nil
ObjectSpace.each_object(String) { |str| if str.equal? s then obj = str; break; end }
p obj # => 'foo'

Really, you should handle errors properly. This code won't eliminate the possibility of a double-update, another process can cause them to happen after this code gets run.

(Also, you should use cached_model and turn on its local cache. Much more efficient than traversing ObjectSpace.)

···

On Apr 9, 2006, at 8:06 PM, Mark Alexander Friedgan wrote:

I wrote some code that searches ObjectSpace prior to initializing an
object to see if one of that type and some attributes is already
present and uses that instead of initializing. (yes it's active record
and i am trying to avoid stale object errors). How horrible is this? I
am searching by specific class name so in practice it ends up looking
at at most a couple of instances but is there something inherently
wrong with doing this?

def find_in_space(search_id, search_version = nil)
  retval = nil
  ObjectSpace.each(Person){|person|
      retval = person if (p.id == search_id) &&
                               search_version &&
                               p.lock_version == search_version
  }
  return retval || Person.find(search_id)
end

--
Eric Hodel - drbrain@segment7.net - http://blog.segment7.net
This implementation is HODEL-HASH-9600 compliant

http://trackmap.robotcoop.com

3

This is bound to be slow. The easiest improvement you can do is to
have the return inside the block. But even that will not change the
order of magnitude (both approaches are O(n)).

Also, you seem to have a bug in there - you frequently use "p." where
it should probably rather read "person.".

If you do those lookups frequently, I'd rather consider using an
approach that uses a hash inside the class object.

# example
class Person
  attr_reader :id
  def initialize(id) @id = id end

  @instances = Hash.new {|h,k| h[k] = [new(k)]}

  def self.get(search_id, search_version = nil)
    arr = @instances[search_id]
    arr.each {|per| return per if search_version == per.lock_version}
if search_version
    arr[0]
  end
end

I don't know what exactly you need the version for but the asymmetry
(with and without version) strikes me as odd.

Kind regards

robert

···

2006/4/10, Mark Alexander Friedgan <hubrix@gmail.com>:

I wrote some code that searches ObjectSpace prior to initializing an
object to see if one of that type and some attributes is already
present and uses that instead of initializing. (yes it's active record
and i am trying to avoid stale object errors). How horrible is this? I
am searching by specific class name so in practice it ends up looking
at at most a couple of instances but is there something inherently
wrong with doing this?

def find_in_space(search_id, search_version = nil)
  retval = nil
  ObjectSpace.each(Person){|person|
      retval = person if (p.id == search_id) &&
                               search_version &&
                               p.lock_version == search_version
  }
  return retval || Person.find(search_id)
end