Duplicate insert on autosave for specific HABTM-has_one combo

[nbekirov]

Steps to reproduce

class Organisation < ActiveRecord::Base
  has_one :user
end

class User < ActiveRecord::Base
  has_and_belongs_to_many :roles
  belongs_to :organisation
end

class Role < ActiveRecord::Base
  has_and_belongs_to_many :users
end

user = User.new
user.organisation = Organisation.new
user.roles << Role.new
user.save!

Here is my Active Record Executable Test Case

This issue is related to Rolify: Duplicate Roles #228

Some observations:

• If I change the order of has_and_belongs_to_many and belongs_to in User there is no issue
• If I add autosave: false to has_and_belongs_to_many in User there is no issue

Here is an issue I've found that seems related but I don't think is quite the same: has_and_belongs_to_many autosaves duplicate records

Expected behavior

One User with one Organisation and one Role saved

INSERT INTO "organisations" DEFAULT VALUES
INSERT INTO "users" ("organisation_id") VALUES (?)  [["organisation_id", 1]]
INSERT INTO "roles" DEFAULT VALUES
INSERT INTO "roles_users" ("user_id", "role_id") VALUES (?, ?)  [["user_id", 1], ["role_id", 1]]

Actual behavior

One User with one Organisation and two Role-s saved

INSERT INTO "organisations" DEFAULT VALUES
INSERT INTO "users" ("organisation_id") VALUES (?)  [["organisation_id", 1]]
INSERT INTO "roles" DEFAULT VALUES
INSERT INTO "roles_users" ("user_id", "role_id") VALUES (?, ?)  [["user_id", 1], ["role_id", 1]]
INSERT INTO "roles_users" ("user_id", "role_id") VALUES (?, ?)  [["user_id", 1], ["role_id", 1]]

System configuration

Rails version:
N/A

Ruby version:

$ ruby -v
ruby 2.3.0p0 (2015-12-25 revision 53290) [x86_64-linux-gnu]

[meinac]

Also If you change has_one :user to has_many :users or ?has_one :users? it works as expected, interesting 😕

[betesh]

Introduced in c7adc61

[meinac]

@rafaelfranca I see that this has been added into 5.1 milestone, do you want to go with the existing pr or do you think something different.

[rafaelfranca]

Oh, I thought it was caused by a commit only present in 5.1, but seems this bug is also present in 4.2. I'm removing the milestone and I'll leave the PR review with @sgrif

[betesh]

There's definitely a regression from 5.0.2 to 5.1.0.beta1. Is @nbekirov's test case is failing on 4.2? If so, I'll try to put together another test case that demonstrates the issue.

Basically, what I'm seeing is that when upgrading from 5.0.2 to 5.1.0.beta1, I get ActiveRecord::StatementInvalid from a has_and_belongs_to_many with a unique index on the join table. autosave: false fixes it, as does moving the has_and_belongs_to_many to after all the belongs_to's.

[rafaelfranca]

Yes, it is failing in 4.2 https://gist.github.com/nbekirov/01c34d2b99ed5bcf9154#file-rolif_test-rb-L11.

[betesh]

Here's the regression, and this is what I was referring to in my comment above about it being introduced in c7adc61:

begin
  require 'bundler/inline'
rescue LoadError => e
  $stderr.puts 'Bundler version 1.10 or later is required. Please update your Bundler'
  raise e
end

gemfile(true) do
  source 'https://rubygems.org'
  # Activate the gem you are reporting the issue against.
  # gem 'activerecord', '5.1.0.beta1' # FAILS with this uncommented
  gem 'activerecord', '5.0.2' # PASSES with this uncommented
  gem 'sqlite3'
end

require 'active_record'
require 'minitest/autorun'
require 'logger'

# This connection will do for database-independent bug reports.
ActiveRecord::Base.establish_connection(adapter: 'sqlite3', database: ':memory:')
ActiveRecord::Base.logger = Logger.new(STDOUT)

ActiveRecord::Schema.define do
  create_table :organisations, force: true do |t|
  end

  create_table :users, force: true do |t|
    t.references :organisation, index: true
  end

  create_table :roles, force: true do |t|
  end

  create_join_table :roles, :users, force: true do |t|
    t.index [:user_id, :role_id], unique: true
  end
end

class Organisation < ActiveRecord::Base
  has_one :user
end

class User < ActiveRecord::Base
  has_and_belongs_to_many :roles
  belongs_to :organisation
  accepts_nested_attributes_for :organisation
end

class Role < ActiveRecord::Base
  has_and_belongs_to_many :users
end

class BugTest < ActiveSupport::TestCase
  def test_association
    assert_nothing_raised do
      User.create! organisation: Organisation.new, roles: [Role.new]
    end
  end
end

I do remember reading a bug report about using accepts_nested_attributes for a has_one relationship, but I don't remember the details and could not find it when I looked around just now.