Rails bi-directional accepts_nested_attributes_for leads to unexpected validation result

[rsalehi2013]

Steps to reproduce

script: https://gist.github.com/rsalehi2013/7cf9c458af9923bcd3e2f3053cefcc55

Setup
Migration:

create_table :posts do |t|
  t.string :title
end

create_table :comments do |t|
  t.integer :post_id
  t.string :body
end

post.rb

class Post < ActiveRecord::Base
  has_many :comments, inverse_of: :post
  accepts_nested_attributes_for :comments

  validates :title, presence: true
end

comment.rb

class Comment < ActiveRecord::Base
  belongs_to :post, inverse_of: :comments
  accepts_nested_attributes_for :post
  
  validates :body, presence: true
end

Procedure
post = Post.new(title: 'foo', comments_attributes: [{body: 'bar'}, {body: nil}, {body: 'baz'}])
post.valid?
post.errors

Expected behavior

valid? -> false
errors -> @messages={:"comments.body"=>["can't be blank"]}, @details={:"comments.body"=>[{:error=>:blank}]}
user is prevented from persisting to the database

Actual behavior

valid? -> true
errors -> @messages={}, @details={}
user is able to persist the invalid records to the database

NOTE: if the invalid comments attribute is at the end of the comments_attributes list, or if there is only one item in the comments_attributes list, the post is correctly marked as invalid

System configuration

Rails version: Rails 5.2.4.5

Ruby version: ruby 2.5.8p224 (2020-03-31 revision 67882) [x86_64-linux]

[intrip]

What happens if you set validate: true on the belongs_to ? belongs_to :post, inverse_of: :comments, validate: true

Also would be great if you can put your steps into a runnable repro script :)

[rsalehi2013]

@intrip adding validate: true to the association definition on Comments appears to have no effect on the validity in this case
here's a reproduction script: https://gist.github.com/rsalehi2013/7cf9c458af9923bcd3e2f3053cefcc55

[intrip]

@rsalehi2013 thanks I'll take a look

[intrip]

@rsalehi2013 #41878 fixes this issue: please check.

Waiting for feedback from the core team now 😄

[rails-bot]

This issue has been automatically marked as stale because it has not been commented on for at least three months.
The resources of the Rails team are limited, and so we are asking for your help.
If you can still reproduce this error on the 6-1-stable branch or on main, please reply with all of the information you have about it in order to keep the issue open.
Thank you for all your contributions.

[intrip]

Still relevant, PR attached.

[Dwimcore]

Hello, any news on a fix ?

[intrip]

There is an attached PR waiting for review

[Dwimcore]

@intrip I've seen it, thank you. I was more asking if it would be merged sometime soon, my bad.
I have to thank you, we temporarily applied your solution as a monkey patch and it seems to work well :)

[donquxiote]

Hi @intrip, I had a quick question about your fix. As the code stands in the PR now, should the fix be backwards compatible with Rails 6.1? I'm trying to monkey patch it in an app that's on Rails 6.1, but I'm not seeing the change I'd expect.
I see you force pushed a few different commits but couldn't find a way to view the older commits. Was there significant differences in a fix for Rails 6.1 verses the current PR? (That is if the current PR is only for Rails 7.) While it's possible my issue is different or I monkey patched your solution improperly, I just want to check all options.

Nevermind, my issue was different and project specific!

[goulvench]

Hi, this issue looks a lot like this Stack Overflow bug report, where calling update and passing nested attributes returns true even though at least one nested record is invalid.

The workaround indicated in the accepted answer is easy and I confirm that it works: load the has_many before updating them. Here would be a way to apply this to the situation described:

# app/models/post.rb
class Post < ActiveRecord::Base
  has_many :comments, inverse_of: :post
  accepts_nested_attributes_for :comments

  validates :title, presence: true

  # WORKAROUND STARTS HERE
  def comments_attributes(params)
    # Preload comments to ensure validation state remains in memory
    comments.load
    super
  end
end