In this episode of Hanami Mastery, I will deep dive into the possible implementation of the parts from a business layer in your project and show you why typed structs can be useful in your system. To do this, I'll use dry-struct as a struct initialization engine.
Unlike regular structs, these give you a static type checking for free and will raise an error each time you try to initialize a struct with invalid attributes violating the schema definition.
One may think: "Awesome, a nice, easy-to-use validation gem for my project", but in the next moment your eyes will see, that within the dry family only there are two other validation gems available already! Those are dry-schema and dry-validation, both great and together fulfill all needs of anyone who needs to validate anything.
However, more importantly, on the official website, you can find a section addressing directly the question of validating data by using dry-struct!
It basically says, that validating data with
dry-struct is totally possible, but you're asked not to do it as it's designed to work with already valid data....
You may now be wondering in which scenario you could need a type check if the data is supposed to be valid? I can feel your confusion, so let me show you a few examples of actual applications when
dry-struct is useful.
The first example I came up with is when you want to get benefits from the event sourcing in your applications. In such systems, one of the most important rules you would like to follow is that "what happened, happened".
It means, when you publish an event, you should not remove it from the log, or even update its data or structure. Most of the event stores don't even allow such a thing to happen by not implementing an interface for updating published events.
Therefore, it's extremely important to be sure, that every event in your system has a valid structure and all data they carry on are of a proper type.
This is one case of when
dry-struct can be useful. No matter if there can be a bug in the validation logic of the API interface, or at any point of the processing, by using
dry-struct it is not possible to even instantiate an event with inappropriate data.
Let me show it to you by an example.
class BaseEvent < Dry::Struct
First I'll add a pre-setup script, where I require
dry-struct, then create my local types module, in case I'd want to extend it later and create the base event, including my types module in it.
class MoneySent < BaseEvent
attribute :sender_id, Types::Strict::String.constrained(uuid_v4: true)
attribute :reciever_id, Types::Strict::String.constrained(uuid_v4: true)
attribute :currency, Types::String
attribute :amount, Types::Strict::Decimal
Then I'm going to define my business event. Let's say, I want to track an event in the system, where money is sent to the user. I can then add an attribute named:
sender_id, and set a rule that it needs to be a UUID type of value.
I'll repeat the same for
reciever_id. This way, if I have UUIDs in my project to define resources, there is no way to have an invalid format of the identifier when the event is published.
Finally, for the currency, I'll just set a type of string, and the amount will be set to a big decimal. This one is pretty important, as floats are not exact, so if we'd allow floats to be passed into our system, we could end up with inconsistencies in total calculations.
event = MoneySent.new(
Now to test it out let me add the irb and the secure random libraries and run the script.
This allows me to initialize correct events in the system, that can be published to or read from the event log, but never allow inconsistent data to pass through.
event = MoneySent.new(sender_id: nil)
# lib/dry/types/constrained.rb:42:in `call_unsafe':
# nil violates constraints (type?(String, nil) AND str?(nil) failed) (Dry::Types::ConstraintError)
I never expose the interface for publishing events directly. Publishing an event is a result of the request call handled by my action, so the parameters, which are my input, had already been validated, and the request had been authorized.
At the point where an event is initialized, there should not be any validation issue whatsoever, and if there is a problem, it certainly is unexpected, so the error risen is completely accurate behavior.
This way I can model my business logic without worrying about validation errors, and just test each part of the system in encapsulation, being sure that it's not possible to publish anything that violates the application state.
Even if you don't want to make use of events in your system, it is possible you will be interested in using the CQRS pattern in your applications, or even just extract your business domain layer aside from the framework part and communicate with it via service objects.
The rule is similar. A command, or service, directly affecting your business domain state, is something that can be called from multiple places.
- api endpoints
- rake tasks
- background jobs
- process managers
- in rails, callbacks.
All of those cases already have input data pre-validated, as I have shown in episode 7 so again, it's completely fine to raise an error in case of calling a business commands with an invalid set of attributes.
Below I'll add a simple command and name it:
SubscribeToHanamiMastery. Then inside I'll set the command schema.
As in the previous example, it'll inherit from
Dry::Struct and include my types module. Then I set up the subscriber ID, type of UUID, and the email.
class Schema < Dry::Struct
attribute :subscriber_id, Types::Strict::String.constrained(uuid_v4: true)
attribute :email, Types::String
schema = Schema.new(input)
# your fancy newsletter subscription call.
puts "You've successfully subscribed to HanamiMastery"
With this, I'm going to define the interface to call my command. The
call method accepts the input argument and initializes my command schema as the first step.
In case of invalid input, we raise the error, as the input should be validated before, in the controller action. If the error is not raised, then we are hundred percent sure that we work with valid data types, which minimizes the possibility to call subscription clients with incorrect data.
subscribe = SubscribeToHanamiMastery.new
subscribe.call(subscriber_id: SecureRandom.uuid, email: 'firstname.lastname@example.org')
# => You've successfully subscribed to HanamiMastery
subscribe.call(subscriber_id: 'invalid id', email: 'email@example.com')
# => Traceback (most recent call last):
# /dry-types-1.5.1/lib/dry/types/constrained.rb:42:in `call_unsafe': "invalid id" violates constraints (uuid_v4?("invalid id") failed) (Dry::Types::ConstraintError)
When I'll call this, the command will return a successful value in case of correct data but raise an error otherwise.
One more useful use case for using
dry-struct, in my opinion, is a value object implementation.
I can't count how many times I have seen undefined method errors on nil class...
Having the type check verification on arguments passed into the value object practically eliminates such errors. It raises an error during the value object initialization but with a way more detailed error message, which helps a lot in debugging and testing your application.
Here is an example of a simple Gender value object, where you can initialize the object using integer and in case of passing an invalid value, you get an error, similar to Integer('invalid')
class Gender < Dry::Struct::Value
attribute :value, Types::Integer.constrained(gteq: 0, lt: 3)
By setting the type I can eliminate the possibility to set gender value yet unsupported by the system while keeping this thing easily extendable in the future.
For now, let's just add to_i and to_s methods.
when 0 then "undefined"
when 1 then "male"
when 2 then "female"
You can see, that setting the type validation check on the value attribute simplifies a lot the whole class. I don't need to care about casting values, and I don't need to consider the case, where value is out of supported range.
gender = Gender.new(value: 1)
gender.value.frozen? # true
gender.to_s # male
gender = Gender.new(value: 4) # error
The value objects allow us to eliminate from our system the primitive obsession code smell, and keep our systems way more reliable.
This is just a basic example, but I believe you get the idea.
Typed structs and deeply frozen structs are a very useful tool and I certainly barely scratched the surface of possible applications, to make use of them. I didn't even mention the reciepies or nested structs features but I strongly encourage you to discover them on your own!
I would love to see in what scenarios do you apply them in your projects, so feel free to send me some code samples or article references in the comments of this episode, and if you do want to see more advanced content, join to Hanami Mastery premium, where I dig much deeper into these topics.
I'd like to especially thank my existing github sponsors and a new sponsor, Andrzej Krzywda for supporting this project and the whole Hanami Mastery initiative! I appreciate that as without your financial support this project could not exist.
Also thanks to Kelly Sikkema - for a great cover image!
Thanks to all of you for being a part of the great Ruby community and for all the positive reactions you give. You're awesome!
Feel free to checkout my other episodes!
Add your suggestion to our discussion panel!
I'll gladly cover them in the future episodes! Thank you!