Permalink
Browse files

Merge pull request #9252 from senny/8423_hmt_preloading_bug

don't cache invalid subsets when preloading hmt associations
  • Loading branch information...
2 parents a254e4b + 4a4ff50 commit cdbd64d1a3afd5a08c393403da61d8c7c1ae6c72 @rafaelfranca rafaelfranca committed Feb 14, 2013
@@ -1,5 +1,27 @@
## Rails 4.0.0 (unreleased) ##
+* Preloading `has_many :through` associations with conditions won't
+ cache the `:through` association. This will prevent invalid
+ subsets to be cached.
+ Fixes #8423.
+
+ Example:
+
+ class User
+ has_many :posts
+ has_many :recent_comments, -> { where('created_at > ?', 1.week.ago) }, :through => :posts
+ end
+
+ a_user = User.includes(:recent_comments).first
+
+ # this is preloaded
+ a_user.recent_comments
+
+ # fetching the recent_comments through the posts association won't preload it.
+ a_user.posts
+
+ *Yves Senn*
+
* Don't run after_commit callback when creating through an association
if saving the record fails.
@@ -31,7 +31,8 @@ def through_records_by_owner
through_records = Array.wrap(owner.send(through_reflection.name))
# Dont cache the association - we would only be caching a subset
- if reflection.options[:source_type] && through_reflection.collection?
+ if (through_scope != through_reflection.klass.unscoped) ||
+ (reflection.options[:source_type] && through_reflection.collection?)
owner.association(through_reflection.name).reset
end
@@ -1162,4 +1162,10 @@ def test_deep_including_through_habtm
Post.where('1 = 0').scoping { Comment.preload(:post).find(1).post }
)
end
+
+ test "preloading does not cache has many association subset when preloaded with a through association" do
+ author = Author.includes(:comments_with_order_and_conditions, :posts).first
+ assert_no_queries { assert_equal 2, author.comments_with_order_and_conditions.size }
+ assert_no_queries { assert_equal 5, author.posts.size, "should not cache a subset of the association" }
+ end
end

0 comments on commit cdbd64d

Please sign in to comment.