Rails - Aug 2006 - Stale object errors

I''m running into a couple of stale object problems with methods that  
at first glance appear fairly atomic - eg Message.update(params[:id],  
{:body => params[:value]})
Of course, Rails has to instantiate the object & validate it before  
it updates the database, so it''s not actually atomic, hence the  
occasional staleobject error.

So I''ve been wondering about creating a block method looking  
something like:

def update_volatile_object
   num_attempts = 5
   while num_attempts>0
     begin
       yield   # the block should do something along the lines of  
Message.update(blah)
       break	  # If we got here, we didn''t raise a stale object error
     rescue ActiveRecord::StaleObjectError => e
       num_attempts -= 1
     end
   end

   raise DatabaseSeemsTooDamnBusy if num_attempts==0
end


...but it feels like I''m missing something - does Rails supply any  
functionality to do this for me?  Any suggestions on how to improve  
the above?

On 8/16/06, Jonathan del Strother <maillist@steelskies.com>
wrote:
>
> I''m running into a couple of stale object problems with methods
that
> at first glance appear fairly atomic - eg Message.update(params[:id],
> {:body => params[:value]})
> Of course, Rails has to instantiate the object & validate it before
> it updates the database, so it''s not actually atomic, hence the
> occasional staleobject error.
>
> So I''ve been wondering about creating a block method looking
> something like:
>
> def update_volatile_object
>    num_attempts = 5
>    while num_attempts>0
>      begin
>        yield   # the block should do something along the lines of
> Message.update(blah)
>        break      # If we got here, we didn''t raise a stale object
error
>      rescue ActiveRecord::StaleObjectError => e
>        num_attempts -= 1
>      end
>    end
>
>    raise DatabaseSeemsTooDamnBusy if num_attempts==0
> end

You shouldn''t be locking the table at all if you find yourself bending
over
backward to circumvent the lock.

If you''re working with frequently stale records, consider using
pessimistic
locking instead (or rethinking the behavior that requires locking at all).

  def next!
    transaction do
      task = find(:first, :conditions => [''claimed=?'',
false], :lock =>
true)
      task.claimed = true
      task.save!
      task
    end
  end

jeremy
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://wrath.rubyonrails.org/pipermail/rails/attachments/20060816/ebd3ad5f/attachment.html