Hobo Hobo → Hobo 1.0 - Final

Comments and changes to this ticket

• 

  Tom Locke September 9th, 2008 @ 01:05 PM

  • State changed from “new” to “open”
• 

  Tom Locke September 9th, 2008 @ 01:06 PM

  See also #247

• 

  Tom Locke March 12th, 2009 @ 11:36 AM

  • Milestone changed from “Hobo 1.0 - Final” to “Beyond Hobo 1.0”

• You flagged this item as spam.
  

  Betelgeuse December 9th, 2009 @ 02:21 PM

  Please make this part of Hobo 1.0 - RC 1 because this makes using <count:> in index pages have horrible
  performance:

  >> o.childs.loaded?
  => false
  >> o.childs.try.to_int
  => nil
  >> o.childs.loaded?
  => true
  >> HoboSupport::VERSION
  => "0.9.102"
  

  Because count has the following code it means using it causes a full load:

  c = this.try.to_int || this.try.total_entries || (this.try.loaded? && this.try.length) || th    is.try.count || this.try.length
  

• You flagged this item as spam.
  

  Betelgeuse December 9th, 2009 @ 07:41 PM

  Actually it doesn't seem to be respond_to? but method_missing that's causing the target to load as the association is trying to relay calls to the target:
  

  /Library/Ruby/Gems/1.8/gems/activerecord-2.3.5/lib/active_record/associations/association_proxy.rb:212:in method_missing'
  /Library/Ruby/Gems/1.8/gems/activerecord-2.3.5/lib/active_record/associations/association_collection.rb:371:inmethod_missing_without_paginate'
  /Library/Ruby/Gems/1.8/gems/mislav-will_paginate-2.3.11/lib/will_paginate/finder.rb:170:in method_missing'
  /Library/Ruby/Gems/1.8/gems/hobo-0.9.0/taglibs/rapid_core.dryml:515:incount'
  

• 

  Tom Locke December 10th, 2009 @ 04:16 PM

  I agree this is quite important, especially index pages have the count tag on them by default. But perhaps we could fix this in <count> in time for Hobo 1.0, and come back to the tougher problem?

• You flagged this item as spam.
  

  Betelgeuse December 10th, 2009 @ 04:38 PM

  It's one possibility but just fixing count could leave other pitfalls like this around. Here's the quick fix I used for count in my application in case someone else needs it:

    has_many :childs, :dependent => :destroy do
      def to_int; count; end
  
      def is_a?(type); false; end
    end
  

  If you need test code, here's what I used

    it "should not load all childs when rendering index" do
      get :index
      response.should be_success
      list = assigns[:this]
      list.size.should > 0
      list.each do |o|
        o.childs.loaded?.should be_false
      end
    end
  

• You flagged this item as spam.
  

  Betelgeuse December 12th, 2009 @ 03:18 PM

  I opened #584 to make sure at least count gets fixed before 1.0.

• 

  Matt Jones December 13th, 2009 @ 12:23 AM

  Some further notes on this issue:

  • the direct loading from calling respond_to? is a side-effect of the implementation AssociationCollection inherits from AssociationProxy; it's relatively straightforward to fix this and avoid loading records. For now, I'll put it in the file that's already tweaking bits of AssociationCollection.

  • on scopes, however, the 'try' method is getting delegated to proxy_found (in named_scope.rb), so that can cause association loads as well. Note that only scope proxies do this; the plumbing is different in the AssociationCollection case, for no particularly apparent reason.

  Not sure what the best fix for the latter is, as it's caused by the name clash between HoboSupport's try and Rails 2.3's try. Maybe a rename is in order?

• 

  Matt Jones December 13th, 2009 @ 12:38 AM

  (from [a289c139779b76bb31a9ef8db94ee839270cb179]) avoid unwanted DB loads on respond_to? and try. [#261] http://github.com/tablatom/hobo/commit/a289c139779b76bb31a9ef8db94e...

• 

  Matt Jones December 13th, 2009 @ 12:41 AM

  @Betelgeuse: can you try out the above commit and see if it's resolved the issue? It worked in my limited test setup, but there may be more places that are causing this problem.

  To correct my previous comment @ 12:23AM, I dealt with the try issue the same way someone already had in methodcall.rb (for AssociationCollection) - adding the method back explicitly.

• You flagged this item as spam.
  

  Betelgeuse December 13th, 2009 @ 10:28 AM

  Your commit passes my unit test for try but #584 remains open as that's caused by hitting method_missing. As this ticked is originally about respond_to? then this one can be marked as resolved but #584 needs further work.

• 

  Bryan Larsen December 14th, 2009 @ 10:38 PM

  • Milestone changed from “Beyond Hobo 1.0” to “Hobo 1.0 - Final”

  just marking 1.0 so we don't lose track of it.

• 

  Tom Locke December 16th, 2009 @ 10:17 PM

  Bryan - are you disagreeing with Betelgeuse that this is resolved now?

• 

  Bryan Larsen December 16th, 2009 @ 10:32 PM

  Nope, I'm just saying I haven't looked closely enough to close it myself, and was hoping Matt would do so.

• 

  Matt Jones December 16th, 2009 @ 11:12 PM

  • Tag cleared.
  • State changed from “open” to “resolved”

  #584 captures the other half of this (with is_a?), so I'll close it.