Skip to content


Subversion checkout URL

You can clone with
Download ZIP


Regression in 3.2.5: collection.clear does not call after_remove callback #6609

jplang opened this Issue · 9 comments

4 participants


In previous Rails versions (in Rails 2.3.x and up to Rails 3.2.3 at least), calling collection.clear for a HABTM association used to trigger after_remove callbacks. It's no longer the case with Rails 3.2.5.

class Developer < ActiveRecord::Base
  has_and_belongs_to_many :projects

class Project < ActiveRecord::Base
  has_and_belongs_to_many :developers, :after_add => :developer_added, :after_remove => :developer_removed

  def developer_added(developer)
    puts "#{} added"

  def developer_removed(developer)
    puts "#{} removed"

p = Project.create!(:name => 'Test')
d = Developer.create!(:name => 'John')
p.developers << d

With Rails 3.2.3, will output:
John added
John removed

With Rails 3.2.5, will output:
John added

In both cases, the association is cleared.


Also applies to :has_many association as well.


Callbacks are called as expected when using p.developers = [] instead of p.developers.clear.


Verified, git bisect reveals b98d1e2 causes the change in behavior.

It is possible the previous behavior should have never been expected, and that the model was being loaded when it shouldn't have been on a #clear. Perhaps @jonleighton would know better.


From the documentation: There is also a clear method which is the same as delete_all, so it would appear this behavior was wrong in 3.2.3, and is now correct, since delete_all should not load the records.


I don't think that 3.2.5 implements the expected behaviour. A test actually exists for that: test_has_and_belongs_to_many_remove_callback_on_clear.


@jplang Okay, it looks like delete may be intended to work a bit differently for HABTM. Let's see what Jon Leighton has to say regarding it.

@jonleighton jonleighton was assigned

Damn. Thanks for the report. I will look into it.


Isn't this issue bad enough to trigger an early 3.2.6 release?

@jonleighton jonleighton closed this issue from a commit
@jonleighton jonleighton Revert "Perf: Don't load the association for #delete_all."
This reverts commit b98d1e2.

Closes #6609



I have reverted the commit which introduced this regression.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.