Write rule in Cancan or model validation? - ruby-on-rails

Update: After reading the answers, I think I should rephrase my question (as question 3)
From time to time I get confused as to where I should write a some conditional check: in Cancan ability or in ActiveRecord model validation?
As the first example: Say I have a folder model which can be nested. I want prevent deletion of a folder if it is the only child of the parent folder.
This should probably be model logic (as a before_destroy callback). However I would also want to hide the delete button(and block controller action), which seems like the realm of Cancan.
As the second example: I want to prevent deletion of a folder not owned by me.
This will need the use of current_user which is stored in the session. I have the impression that session related condition should not touch the model itself, so this is for Cancan. Is it correct?
Question 3:
If deleting a folder requires both:
current_user is owner check (written as Cancan ability)
folder is not the only child check (written in model as destroyable?())
Should the Cancan ability also call model.destroyable?(), or should I call model.destroyable?() separately (in view and in controller)?

IMO Cancan is about authorization: is the user allowed to delete a given resource based on who they are. Restricting resource deletion based on other criteria falls outside that purview.
This sounds like a combination of authorization and business logic. A view helper might check both if the user can? delete the resource, and that the resource is deletable?.

For the first example, I'd do both: put a before_destroy callback that guards against destroying the last record, and also show the button based on CanCan ability. Just hiding the button would not prevent a POST request being sent to your server and deleting the record.
For the second scenario, put the CanCan check on the controller action, so that the destroy action cannot be called without authorization. It does not belong in the model.

Related

How does cancancan set the model instance, and how can I inspect it?

I noticed users could access an action they shouldn't be able to access.
I debugged in the rails console with something like
user = User.first
physician = Physician.first
ability = Ability.new(user)
ability.can?(:send_message, physician)
# => false
The above says that user can't access the send_message action for that physician, which is the desired behaviour, yet I know they can in the app!
I think this narrows down the cause to a problem with cancancan loading the wrong model instance for some reason. And that's hinted to in the cancan docs too:
Note: this assumes that the model instance is being loaded properly.
But the problem is I'm not sure how to diagnose the problem from here, since the console says it should be working. I don't know how to view the model instance that cancancan has set, and I don't know what else to try.
Any ideas?
Update
I did manage to work around this by using authorize! :send_message, physician in the controller, but since I only stumbled upon this behaviour by chance, I think it's much more important to figure out why the wrong model instance was being loaded (especially so I can see if that was happening elsewhere too).
I figured out why it was probably happening (but I still don't know how to disagnose it)
I think this was happening because I had many custom actions, and some had #physician = Physician.find(current_user.physician.id) (i.e they're the current user), whereas others were more like #physician = Physician.find_by_id(physician_params[:id]). I'm not sure how cancan sets the instance model, but I do know it's not psychic, so it wouldn't know whether to set it to the current user's physician instance, or the physician instance for the physician id passed in.
What remains?
How does cancancan set the model instance for custom methods (I presume it tries something, and if that doesn't work, tries something else, etc etc)?
Small notes that help:
load_and_authorize_resource does attempt to load the model instance for non RESTful actions
Some useful info in the docs
This may have something to do with what I experienced:
When I returned slug it breaks this behaviour and I can edit all pokemons.
Leaving my notes here in case they are helpful to anyone else.
TL;DR, there are a lot of nuanced assumptions cancancan makes, which you won't know about from the outset. I discovered many of them by thoroughly reading the comments in the cancancan readme, code, and defining abilities docs
So here goes..
How cancancan works
if you call authorize! in the controller action itself, cancancan will look for an instance variable in each controller action.
if you instead simply add load_and_authorize_resource at the start of your controller, that will do two things:
Load an instance variable that cancancan thinks should be loaded, and
Checks for authorization on that model instance
Note that for custom actions, load_and_authorize_resource will still try to load a model instance, but how does it know what to load? It doesn't, it guesses, which, for me, I do not like, so be aware of that.
For me, I prefer to do the work of load_and_authorize_resource myself in two separate steps, so I know exactly what's going on.
Ensure #article is generated via a before action for each controller action (or #articles for index action)
Simply have a line at the top of the controller saying load_and_authorize_resource after the before action that sets the model instance
Note that the only difference is now the developer is responsible for loading the right model instance, and cancancan is not trying to guess it. I prefer this approach because it only takes one mistake to accidentally allow access where it shouldn't be granted.
Also remember that load_and_authorize_resource should always go after any before actions that set the model instance variable
Random notes that may also help
The name of the instance variable depends on the action. If we have an articles controller, then:
For the index action, authorize looks for #articles
For all other actions, authorize looks for #article
It then checks to see if the user is allowed access to that resource.
load_and_authorize_resource checks to see if the model instance exists, and if not, creates one. So if you have a before action that creates #article/#articles, then load_and_authorize_resource won't do it for you (i.e. it won't overwrite it), but if you didn't set one, cancan will try to set one. See here for more on that.
An ability rule will override a previous one. (see here for an example)
Just one last thing, never use current_user in ability.rb, it will error silently (!!), so be sure to use user instead :)
Here's what is happening: https://github.com/CanCanCommunity/cancancan/blob/585e5ea54c900c6afd536f143cde962ccdf68607/lib/cancan/controller_additions.rb#L342-L355
# Creates and returns the current user's ability and caches it. If you
# want to override how the Ability is defined then this is the place.
# Just define the method in the controller to change behavior.
#
# def current_ability
# # instead of Ability.new(current_user)
# #current_ability ||= UserAbility.new(current_account)
# end
#
# Notice it is important to cache the ability object so it is not
# recreated every time.
def current_ability
#current_ability ||= ::Ability.new(current_user)
end

Rails 5 access profile data anywhere in session without querying database each time

I've a user profile (with name, logo, about_me) which is created after user creation(using Devise). Profile table uses user_id as Primary key.
Now I want that whenever the user creates/updates a post, while filling in form some details are taken from profile, so profile data or #profile be available in post form as I cannot expose my model in form.
To set post.myname attribute in create and #update I'm doing this:
#myprofile = Profile.find_by_user_id(current_user)
write_attribute(:myname, #myprofile.name)
I read from various sources but what's the best solution of the 4 given and if anyone can back with easy code as I do not want to do something extensive? Thanks in advance.
1)Form Hidden fields - Like get the profile data as above in hash in #edit and then pass through form and access fields in #update but that way we will pass each field separately. Can one #myprofile be passed?
2)Session - I feel if profile data is stored in a session and someone updates profile then updated data won't be available in that session.So not sure if it is plausible.
3)Caching - easy way to do that?
4)polymorphic profile---tried it but I didnot get relevant example. I was stuck with what to put as profileable id and type and how to use them in the code.
If your Profile and User models have a one-to-one relationship with each other, the simplest solution is to remove the Profile model altogether and move its fields into the User model.
Devise already queries the database to obtain the current_user object. So, your example would like this:
write_attribute(:myname, current_user.name)
Which wouldn't hit the database (after Devise has retrieved the current_user object).
If you're forced to keep the Profile model, in looking at your four scenarios ...
You could use a session variable. Something like:
session[:profile_name] ||= #myprofile.name
This would go in a controller action.
The trick here is that you will want to redefine the each relevant session variable if the profile gets updated. And because you don't have access to the session in the model, you'd be best to perform that action in the controller. So, not pretty, but it could work.
You could also use low-level caching, and save the profile relationship on the user. In general, you could have a method like this in your user model:
def profile_cached
Rails.cache.fetch(['Profile', profile.id]) do
profile
end
end
Here, too, you will have to know when to expire the cache. The benefit of this approach is that you can put this code in the model, which means you can hook its expiration in a callback.
Read more about this in Caching with Rails.
I would avoid hidden fields and I'm not sure how a polymorphic relationship would solve you not hitting the database. So, #2 and #3 are options, but if you can combine the two models into one, that should simplify it.

Different update / edit methods available to different users

I have a model Post, which is submitted and graded by different Users. The submitter and grader are identified by submitter_id and grader_id in Post model. Note that an user is both a submitter himself and a grader to others.
I want to make sure that the submitter can only edit the content of the Post but not the grade. Likewise, the grader can only edit the grade but not the content.
Is multiple edit methods the way to go? How should I accomplish this otherwise?
You can have a role column in your users table, and the role can be either submitter or grader. Not sure what you are using for authentication, but in case you are using devise, you can access the currently logged in user with current_user helper (in case you are using something else, figure this part out, or add a new helper).
Now in your update method, you can do something like this:
# Controller
# scope post to current user, so that a user cannot edit someone else's post. A crude way to achieve this is post = Post.find(params[:id])
post = current_user.posts.find(params[:id])
post.content = params[:content] if post.submitter?(current_user.id)
post.grade = params[:grade] if post.grader?(current_user.id)
post.save!
# Model - Post.rb
def submitter?(user_id)
self.submitter_id == user_id
end
def grader?(user_id)
self.grader_id == user_id
end
The advantage of keeping those methods in the model is that in case you permission logic changes (who is submitter, or a grader), you need to change it at a single location. DRY.
You can modify the above approach to show error messages, and do other similar stuff. In case you are looking for more granular authorization control, you can look into cancan gem:
https://github.com/ryanb/cancan
Your post model should only be concerned with persisting data. Better to use plain old ruby objects to encapsulate the higher order behavior of grading and submitting. Consider using service objects or form objects.
Each service or form object can then include ActiveModel::Model(rails > v4) to get its own validations.
See more about service and form objects here: http://blog.codeclimate.com/blog/2012/10/17/7-ways-to-decompose-fat-activerecord-models/
If you only have one submit action and one grade action, its probably ok to keep in one controller. But if you start having multiple actions that are related to submitted, and multiple actions that are related to grading, this sounds like they would make great resources controllers on their own.

Setting Pundit role for user from Devise Registrations New View / Controller

I have both Pundit and Devise setup and working correctly in my rails app. However I am unsure about how to let the user decide their role when signing up.
At the moment:
I have a URL param which is passed to the Devise new view.
In the form_for I set a hidden field called role to the value of the param.
This works.
But I am concerned that a malicious user could change this param to say "Admin" and now they are an admin.
How should I handle this? I don't want to put a restriction in the model as that will cause issues when I want to create an admin. Should I override the devise registrations controller to put a check in there?
You don't need to override Devise's RegistrationsController for what you're trying to do.
If you want admins to be able to create users that have an arbitrary role set, you could simply use your own controller. Devise still makes it easy to create a user yourself, so you'll just have to make a controller handling this. Of course, don't forget to protect it using Pundit so only admins can use this functionality.
This approach still works if you use the Confirmable module. As no confirmation e-mail will be sent on user creation, though, you'll either have to call user.confirm! after saving the model to immediately unlock the account, or manually send the confirmation e-mail using user.send_confirmation_instructions.
Edit:
This Pundit policy may or may not work for what you're trying to do. You will have to override the create action of Devise's RegistrationsController here in order to use Pundit's authorize method. For dryness' sake, you should also move the roles list elsewhere, perhaps into the model.
class UserPolicy < Struct.new(:current_user, :target_user)
def create?
registration_roles.include?(target_user.role) if current_user.nil?
end
private
def registration_roles
%w(RED BLU Spectator)
end
end
After a fair amount of googling I have an answer. First stick some validation in your model for the roles Active Record Validations Guide: See 2.6 inclusion: validator option
After this your roles are validated to ensure they are correct, you could of course have a lookup table as well. Then you have two options:
Use a conditional before_save Callback for new records. Then check if the new record has the role your protecting and if so raise an error. To catch later (in an overridden devise controller (see second option).
Override the Devise registrations controller See this SO question. And put some checks in a completely overridden create action. Use the session to store the url param passed to the new action (also needs to be completely overridden). Then if the create action fails and redirects to new you still have access to the role in the session (as the param will be cleared from the URL unless you manipulate it).
So either way you need to override the registrations controller, its just a case of how much.
I suspect there is a way to do this with just Pundit. But I have yet to be able to get it to work.

Best practice for observing RESTful routing in Rails and coupling resource models and controllers?

Strict REST proponents might say that if you ever find yourself defining an action on a controller that isn't CRUD, then you should strongly consider creating a new resource and defining your action as a CRUD operation on the new resource.
An example might be operations that change state on a model - say Purchase. In this example, instead of defining a 'complete' action on PurchaseController you might create a CompletePurchasesController and use the create action to update the purchase's state to completed.
Assuming the above, you obviously don't persist PurchaseState directly to the database.
My question is when do you couple the Controllers to Models? When do you define a PurchaseState model (which isn't persisted) and when do you simply work directly with Purchase.
Is it a question of complexity and the number of loosely associated models you are interacting with in the Controller actions?
'Complete' is a state transition event on (an existing) purchase. I find it counter-intuitive to conceptualize this action as a create on a virtual resource rather than an update action on a controller coupled to the Purchase model, in fact PurchaseController itself.
I would define individual update-like actions for such state transitions. I think this way you can leverage rails structure in the most economical way, including model-initialisation, view dispatching, routing, access control. Let's assume you use inherited_resources, cancan by simply adding
# routes.rb
resource :purchase do
put :complete, :on => :member
end
# purchase_controller.rb
def complete
#purchase.complete!
end
# cancan ability (entry already there for basic crud)
can :manage, Purchase, :user_id => user.id
you are already done implementing the entire UI (view/model logic aside).
How utterly cool is that in rails.
If your typical usecase is that purchase is only updated by state transitions, especially all having the same access rights and redirect views, then I would even use the update action of PurchaseController with state_event attributes. See
Can somebody give an active record example for pluginaweek - statemachine?
Strict REST-ists, bite me! :)
For updating the purchase state, you probably only need a PurchasesController update action, which you would define in your routes file as a 'put' or 'patch' method.
If all that happens on update is changing the state field on your purchase object, you can probably just do that right in the update action.
If there's some business logic around some of the state transitions, but ultimately you're only changing that purchase object, you probably want to put that in your Purchase model.
If other tables are also updated, or you're also doing things like queuing up an email to your user congratulating them on their new purchase, I think that's when you might add a separate PurchaseComplete or PurchaseAbort models / service objects. These seem to come into play most naturally when the logic for the action is more complex, and/or you have changes to more than one model, or are doing something else.

Resources