Model callbacks not working with self referential association - ruby-on-rails

I am having a model Evaluation that has many sub evaluations (self refential)
class Evaluation < ApplicationRecord
has_many :sub_evaluations, class_name: "Evaluation", foreign_key: "parent_id", dependent: :destroy
before_save :calculate_score
def calculate_score
# do something
end
end
I am creating and updating evaluation with sub evaluations as nested attributes.
calculate_score method is triggered on sub evaluation creation but not while updating. I have tried before_update and after_validation. But nothing seems to be working.
Evaluation form
= form_for #evaluation do |f|
...
= f.fields_for :sub_evaluations do |sub_evaluation|
...
What seems to be the issue?

This article helped me to fix the issue.
Child callback isn't triggered because the parent isn't "dirty".
The solution in the article is to "force" it to be dirty by calling attr_name_will_change! on a parent attribute that, in fact, does not change.
Here is the updated model code:
class Evaluation < ApplicationRecord
has_many :sub_evaluations, class_name: "Evaluation", foreign_key: "parent_id", dependent: :destroy
before_save :calculate_score
def calculate_score
# do something
end
def exam_id= val
exam_id_will_change!
#exam_id = val
end
end
See Active Model Dirty in the Rails API

Related

Perform callback after destroying child records only

class MyTask < ApplicationRecord
has_many :jobs, as: :ownerable, dependent: :destroy
accepts_nested_attributes_for :jobs, allow_destroy: true
before_save :set_some_data
end
class Job < ApplicationRecord
belongs_to :ownerable, polymorphic: true, optional: true
end
In the :set_some_data method, we actually take the values from all the jobs belonging to the MyTask object and perform some calculations and save the result in a column (actually, just a self.column_name = calculated_value, not actually calling save).
The problem is that the UPDATE on the column happens before any jobs marked for destruction ie with "_destroy" => 1 in the params. And so obviously, it includes data from the deleted jobs, which is incorrect.
I am currently doing the following - change callback to:
after_save :set_some_data
def set_some_data
#Do stuff
# WARNING: Don't use any method that will trigger an after_save callback. Infinite loop otherwise.
self.update_columns(column_name: calculated_value)
end
It does what I want. But is this a good solution? Can you suggest some better alternatives?
you can do this with with after_destroy and put the method in job.rb this will make sure when child deleted(job) it will call the parent to update the value
class Job < ApplicationRecord
belongs_to :ownerable, polymorphic: true, optional: true
after_destroy :update_parent
def update_parent
# check your parent model
self.ownerable.update_columns(column_name: calculated_value)
end
end
for more detail callback you can check this and

after_create doesn't have access to associated records created during before_created callback?

I am running into a weird issue, and reading the callbacks RoR guide didn't provide me an answer.
class User < ActiveRecord::Base
...
has_many :company_users, dependent: :destroy
has_many :companies, through: :company_users
has_many :user_teams, dependent: :destroy
has_many :teams, through: :user_teams
before_create :check_company!
after_create :check_team
def check_company!
return if self.companies.present?
domain = self.email_domain
company = Company.find_using_domain(domain)
if company.present?
assign_company(company)
else
create_and_assign_company(domain)
end
end
def check_team
self.companies.each do |company|
#do stuff
end
end
...
end
The after_create :check_team callback is facing issues because the line
self.companies.each do |company|
Here, self.companies is returning an empty array [] even though the Company and User were created and the User was associated with it. I know I can solve it by making it a before_create callback instead. But I am puzzled!
Why does the after_create callback not have access to self's associations after the commit?
Solution: Please read my comments in the accepted answer to see the cause of the problem and the solution.
inside before_create callbacks, the id of the record is not yet available, because it is before... create... So it is not yet persisting in the database to have an id. This means that the associated company_user record doesn't have a user_id value yet, precisely because the user.id is still nil at that point. However, Rails makes this easy for you to not worry about this "chicken-and-egg" problem, provided that you do it correctly:
I recreated your setup (Company, User, and CompanyUser models), and the following is what should work on your case (tested working):
class User < ApplicationRecord
has_many :company_users, dependent: :destroy
has_many :companies, through: :company_users
before_create :check_company!
after_create :check_team
def check_company!
# use `exists?` instead of `present?` because `exists?` is a lot faster and efficient as it generates with a `LIMIT 1` SQL.
return if companies.exists?
## when assigning an already persisted Company record:
example_company = Company.first
# 1) WORKS
companies << example_company
# 2) WORKS
company_users.build(company: example_company)
## when assigning and creating a new Company record:
# 1) WORKS (this company record will be automatically saved/created after this user record is saved in the DB)
companies.build(name: 'ahaasdfwer') # or... self.companies.new(name: 'ahaasdfwer')
# 2) DOES NOT WORK, because you'll receive an error `ActiveRecord::RecordNotSaved: You cannot call create unless the parent is saved`
companies.create(name: 'ahaasdfwer')
end
def check_team
puts companies.count
# => 1 if "worked"
puts companies.first.persisted?
# => true if "worked"
end
end

Prevent from raising ActiveRecord::RecordInvalid or adding twice on has_many association

I want to change has_many association behaviour
considering this basic data model
class Skill < ActiveRecord::Base
has_many :users, through: :skills_users
has_many :skills_users
end
class User < ActiveRecord::Base
has_many :skills, through: :skills_users, validate: true
has_many :skills_users
end
class SkillsUser < ActiveRecord::Base
belongs_to :user
belongs_to :skill
validates :user, :skill, presence: true
end
For adding a new skill we can easily do that :
john = User.create(name: 'John Doe')
tidy = Skill.create(name: 'Tidy')
john.skills << tidy
but if you do this twice we obtain a duplicate skill for this user
An possibility to prevent that is to check before adding
john.skills << tidy unless john.skills.include?(tidy)
But this is quite mean...
We can as well change ActiveRecord::Associations::CollectionProxy#<< behaviour like
module InvalidModelIgnoredSilently
def <<(*records)
super(records.to_a.keep_if { |r| !!include?(r) })
end
end
ActiveRecord::Associations::CollectionProxy.send :prepend, InvalidModelIgnoredSilently
to force CollectionProxy to ignore transparently adding duplicate records.
But I'm not happy with that.
We can add a validation on extra validation on SkillsUser
class SkillsUser < ActiveRecord::Base
belongs_to :user
belongs_to :skill
validates :user, :skill, presence: true
validates :user, uniqueness: { scope: :skill }
end
but in this case adding twice will raise up ActiveRecord::RecordInvalid and again we have to check before adding
or make a uglier hack on CollectionProxy
module InvalidModelIgnoredSilently
def <<(*records)
super(valid_records(records))
end
private
def valid_records(records)
records.with_object([]).each do |record, _valid_records|
begin
proxy_association.dup.concat(record)
_valid_records << record
rescue ActiveRecord::RecordInvalid
end
end
end
end
ActiveRecord::Associations::CollectionProxy.send :prepend, InvalidModelIgnoredSilently
But I'm still not happy with that.
To me the ideal and maybe missing methods on CollectionProxy are :
john.skills.push(tidy)
=> false
and
john.skills.push!(tidy)
=> ActiveRecord::RecordInvalid
Any idea how I can do that nicely?
-- EDIT --
A way I found to avoid throwing Exception is throwing an Exception!
class User < ActiveRecord::Base
has_many :skills, through: :skills_users, before_add: :check_presence
has_many :skills_users
private
def check_presence(skill)
raise ActiveRecord::Rollback if skills.include?(skill)
end
end
Isn't based on validations, neither a generic solution, but can help...
Perhaps i'm not understanding the problem but here is what I'd do:
Add a constraint on the DB level to make sure the data is clean, no matter how things are implemented
Make sure that skill is not added multiple times (on the client)
Can you show me the migration that created your SkillsUser table.
the better if you show me the indexes of SkillsUser table that you have.
i usually use has_and_belongs_to_many instead of has_many - through.
try to add this migration
$ rails g migration add_id_to_skills_users id:primary_key
# change the has_many - through TO has_and_belongs_to_many
no need for validations if you have double index "skills_users".
hope it helps you.

How to validate presence of associated records in case of has_many, through: association?

There are models with has has_many through association:
class Event < ActiveRecord::Base
has_many :event_categories
has_many :categories, through: :event_categories
validates :categories, presence: true
end
class EventCategory < ActiveRecord::Base
belongs_to :event
belongs_to :category
validates_presence_of :event, :category
end
class Category < ActiveRecord::Base
has_many :event_categories
has_many :events, through: :event_categories
end
The issue is with assigning event.categories = [] - it immediately deletes rows from event_categories. Thus, previous associations are irreversibly destroyed and an event becomes invalid.
How to validate a presence of records in case of has_many, through:?
UPD: please carefully read sentence marked in bold before answering.
Rails 4.2.1
You have to create a custom validation, like so:
validate :has_categories
def has_categories
unless categories.size > 0
errors.add(:base, "There are no categories")
end
end
This shows you the general idea, you can adapt this to your needs.
UPDATE
This post has come up once more, and I found a way to fill in the blanks.
The validations can remain as above. All I have to add to that, is the case of direct assignment of an empty set of categories. So, how do I do that?
The idea is simple: override the setter method to not accept the empty array:
def categories=(value)
if value.empty?
puts "Categories cannot be blank"
else
super(value)
end
end
This will work for every assignment, except when assigning an empty set. Then, simply nothing will happen. No error will be recorded and no action will be performed.
If you want to also add an error message, you will have to improvise. Add an attribute to the class which will be populated when the bell rings.
So, to cut a long story short, this model worked for me:
class Event < ActiveRecord::Base
has_many :event_categories
has_many :categories, through: :event_categories
attr_accessor :categories_validator # The bell
validates :categories, presence: true
validate :check_for_categories_validator # Has the bell rung?
def categories=(value)
if value.empty?
self.categories_validator = true # Ring that bell!!!
else
super(value) # No bell, just do what you have to do
end
end
private
def check_for_categories_validator
self.errors.add(:categories, "can't be blank") if self.categories_validator == true
end
end
Having added this last validation, the instance will be invalid if you do:
event.categories = []
Although, no action will have been fulfilled (the update is skipped).
use validates_associated, official documentaion is Here
If you are using RSpec as your testing framework, take a look at Shoulda Matcher. Here is an example:
describe Event do
it { should have_many(:categories).through(:event_categories) }
end

Rails association with multiple foreign keys

I want to be able to use two columns on one table to define a relationship. So using a task app as an example.
Attempt 1:
class User < ActiveRecord::Base
has_many :tasks
end
class Task < ActiveRecord::Base
belongs_to :owner, class_name: "User", foreign_key: "owner_id"
belongs_to :assignee, class_name: "User", foreign_key: "assignee_id"
end
So then Task.create(owner_id:1, assignee_id: 2)
This allows me to perform Task.first.owner which returns user one and Task.first.assignee which returns user two but User.first.task returns nothing. Which is because task doesn't belong to a user, they belong to owner and assignee. So,
Attempt 2:
class User < ActiveRecord::Base
has_many :tasks, foreign_key: [:owner_id, :assignee_id]
end
class Task < ActiveRecord::Base
belongs_to :user
end
That just fails altogether as two foreign keys don't seem to be supported.
So what I want is to be able to say User.tasks and get both the users owned and assigned tasks.
Basically somehow build a relationship that would equal a query of Task.where(owner_id || assignee_id == 1)
Is that possible?
Update
I'm not looking to use finder_sql, but this issue's unaccepted answer looks to be close to what I want: Rails - Multiple Index Key Association
So this method would look like this,
Attempt 3:
class Task < ActiveRecord::Base
def self.by_person(person)
where("assignee_id => :person_id OR owner_id => :person_id", :person_id => person.id
end
end
class Person < ActiveRecord::Base
def tasks
Task.by_person(self)
end
end
Though I can get it to work in Rails 4, I keep getting the following error:
ActiveRecord::PreparedStatementInvalid: missing value for :owner_id in :donor_id => :person_id OR assignee_id => :person_id
TL;DR
class User < ActiveRecord::Base
def tasks
Task.where("owner_id = ? OR assigneed_id = ?", self.id, self.id)
end
end
Remove has_many :tasks in User class.
Using has_many :tasks doesn't make sense at all as we do not have any column named user_id in table tasks.
What I did to solve the issue in my case is:
class User < ActiveRecord::Base
has_many :owned_tasks, class_name: "Task", foreign_key: "owner_id"
has_many :assigned_tasks, class_name: "Task", foreign_key: "assignee_id"
end
class Task < ActiveRecord::Base
belongs_to :owner, class_name: "User"
belongs_to :assignee, class_name: "User"
# Mentioning `foreign_keys` is not necessary in this class, since
# we've already mentioned `belongs_to :owner`, and Rails will anticipate
# foreign_keys automatically. Thanks to #jeffdill2 for mentioning this thing
# in the comment.
end
This way, you can call User.first.assigned_tasks as well as User.first.owned_tasks.
Now, you can define a method called tasks that returns the combination of assigned_tasks and owned_tasks.
That could be a good solution as far the readability goes, but from performance point of view, it wouldn't be that much good as now, in order to get the tasks, two queries will be issued instead of once, and then, the result of those two queries need to be joined as well.
So in order to get the tasks that belong to a user, we would define a custom tasks method in User class in the following way:
def tasks
Task.where("owner_id = ? OR assigneed_id = ?", self.id, self.id)
end
This way, it will fetch all the results in one single query, and we wouldn't have to merge or combine any results.
Extending upon #dre-hh's answer above, which I found no longer works as expected in Rails 5. It appears Rails 5 now includes a default where clause to the effect of WHERE tasks.user_id = ?, which fails as there is no user_id column in this scenario.
I've found it is still possible to get it working with a has_many association, you just need to unscope this additional where clause added by Rails.
class User < ApplicationRecord
has_many :tasks, ->(user) {
unscope(:where).where(owner: user).or(where(assignee: user)
}
end
Rails 5:
you need to unscope the default where clause
see #Dwight answer if you still want a has_many associaiton.
Though User.joins(:tasks) gives me
ArgumentError: The association scope 'tasks' is instance dependent (the scope block takes an argument). Preloading instance dependent scopes is not supported.
As it is no longer possible you can use #Arslan Ali solution as well.
Rails 4:
class User < ActiveRecord::Base
has_many :tasks, ->(user){ where("tasks.owner_id = :user_id OR tasks.assignee_id = :user_id", user_id: user.id) }
end
Update1:
Regarding #JonathanSimmons comment
Having to pass the user object into the scope on the User model seems like a backwards approach
You don't have to pass the user model to this scope.
The current user instance is passed automatically to this lambda.
Call it like this:
user = User.find(9001)
user.tasks
Update2:
if possible could you expand this answer to explain what's happening? I'd like to understand it better so I can implement something similar. thanks
Calling has_many :tasks on ActiveRecord class will store a lambda function in some class variable and is just a fancy way to generate a tasks method on its object, which will call this lambda. The generated method would look similar to following pseudocode:
class User
def tasks
#define join query
query = self.class.joins('tasks ON ...')
#execute tasks_lambda on the query instance and pass self to the lambda
query.instance_exec(self, self.class.tasks_lambda)
end
end
I worked out a solution for this. I'm open to any pointers on how I can make this better.
class User < ActiveRecord::Base
def tasks
Task.by_person(self.id)
end
end
class Task < ActiveRecord::Base
scope :completed, -> { where(completed: true) }
belongs_to :owner, class_name: "User", foreign_key: "owner_id"
belongs_to :assignee, class_name: "User", foreign_key: "assignee_id"
def self.by_person(user_id)
where("owner_id = :person_id OR assignee_id = :person_id", person_id: user_id)
end
end
This basically overrides the has_many association but still returns the ActiveRecord::Relation object I was looking for.
So now I can do something like this:
User.first.tasks.completed and the result is all completed task owned or assigned to the first user.
Since Rails 5 you can also do that which is the ActiveRecord safer way:
def tasks
Task.where(owner: self).or(Task.where(assignee: self))
end
My answer to Associations and (multiple) foreign keys in rails (3.2) : how to describe them in the model, and write up migrations is just for you!
As for your code,here are my modifications
class User < ActiveRecord::Base
has_many :tasks, ->(user) { unscope(where: :user_id).where("owner_id = ? OR assignee_id = ?", user.id, user.id) }, class_name: 'Task'
end
class Task < ActiveRecord::Base
belongs_to :owner, class_name: "User", foreign_key: "owner_id"
belongs_to :assignee, class_name: "User", foreign_key: "assignee_id"
end
Warning:
If you are using RailsAdmin and need to create new record or edit existing record,please don't do what I've suggested.Because this hack will cause problem when you do something like this:
current_user.tasks.build(params)
The reason is that rails will try to use current_user.id to fill task.user_id,only to find that there is nothing like user_id.
So,consider my hack method as an way outside the box,but don't do that.
Better way is using polymorphic association:
task.rb
class Task < ActiveRecord::Base
belongs_to :taskable, polymorphic: true
end
assigned_task.rb
class AssignedTask < Task
end
owned_task.rb
class OwnedTask < Task
end
user.rb
class User < ActiveRecord::Base
has_many :assigned_tasks, as: :taskable, dependent: :destroy
has_many :owned_tasks, as: :taskable, dependent: :destroy
end
In result, we can use it so:
new_user = User.create(...)
AssignedTask.create(taskable: new_user, ...)
OwnedTask.create(taskable: new_user, ...)
pp user.assigned_tasks
pp user.owned_tasks

Resources