Hacker News new | ask | show | jobs
The Decree Design Pattern (calebhearth.com)
23 points by caleb_thompson 1230 days ago
7 comments

I've been using this pattern of naming and structuring service objects over the past ~year. It's initially been greeted with some uncertainty and skepticism when I introduce it to new developers, but it tends to grow on folks as they give it a shot and look at how it's already been used.

I wrote this up partially to codify some of how I've been explaining it to folks ad-hoc, but also to share with the broader community and get input on this way of extracting processes in Ruby/Rails projects.

So a Decree is… a function?
Everything is a function if you untangle enough of it
Everything is a series of statements.
Aha, but can you define series?!
This seems similar to an itch I've had for a while on many languages.

I have quite often wanted the ability to easily collect the arguments to a function call together in a structure with the function itself, and defer the execution of the method. I want a closure essentially but with the ability to reflect and inspect the arguments and function, potentially make changes, and execute it later.

Of course the only real way to do this in most languages is to make every single thing i want to do a single-function class with public "argument" members. But that messes with the structure of the code.

It would be really nice to have some syntactic sugar for extracting a class for function calls and their arguments.

The decree pattern seems similar the solution I've described above but avoided implementing because it sounds crazy.

I’ve used functools partials in Python before and it felt seamless to me. Is that what you’re talking about?

https://stackoverflow.com/questions/3188048/how-to-bind-argu...

We use a similar pattern and have a strict namespace.

Player would be a namespace. It would be a module with no or limited code. Within it would be Controller, Model or the agent noun modules that take call: Creator, Inporter, Mailer - the “er” words indicate there will be a single call function.

We broke from the standard Rails directory structure. All the code for Player would be in one directory - including views.

Sounds like you reinvented Django but in Ruby.

Feature-slicing (directory per feature) vs Component-slicing (models, views, controllers each have their own directory regardless of which feature they contribute to).

I thought this was a sarcastic description of an antipattern to start with. Whenever I see classes with verbs in their name I see it as a bad smell.

In the ProcessPayment example he says the pattern is good because it lets you split the construction of the payment from its execution.

Soooo why not have a Payment class with a process method? Am I missing something?

I've been asking myself *this* for the year-ish I've been working with Ruby on Rails. The idea of encapsulating any piece of logic into a class function in order to call a single method is very weird, given that you can do the very same thing with just a method on a model class. Payment.process seems superior in many areas (understanding, readability, explainability) compared to PaymentProcessor.new(payment: payment).process.

It dawn on me that the latter is precisely what people at Java do, and I've reached the conclusion that this way of writing software is enforced by Java developers who realized they needed to learn Ruby because their job prospects are better, and are simply retrofitting what they used to do in a new programming language.

Behaviour is what needs to despatched, and behaviour is what benefits from applying eg. strategy pattern, delegation, orchestration.

Extracting important business behaviour into objects allows applying OO (ie, powerful despatch rules) where it is valuable.

Subtyping your model to achieve many of these possible usecases would give poor design results.

For example, if payment processing were in Payment.process() the Payment entity would need to be subtyped or composed for any of 1) a different payment gateway, 2) a sales tax in a new jurisdiction, 3) a new confirmation, or 4 a new payment flow. Having all that in your model entity is probably wrong.

Simple behaviour is fine in the model, major business TN usually not.

>For example, if payment processing were in Payment.process() the Payment entity would need to be subtyped or composed for any of 1) a different payment gateway, 2) a sales tax in a new jurisdiction, 3) a new confirmation, or 4 a new payment flow. Having all that in your model entity is probably wrong.

There's nothing at all wrong with having, say, a PaymentGateway and SalesTax classes and composing them with Payment in whatever way makes sense given their interdependent relationships. Code that has real life analogs to objects tends to be much cleaner, in fact.

That’s a design pattern I used in the early ‘90s. I called it “False Objects” (I think that I even wrote up a pattern doc on it). I got the idea from QuickDraw GX (about the only good I got from that tech).

I used it to leverage OOP from non-OOP languages (like C).

An SDK that I wrote in 1994, using it, was still in use, when I left the company, in 2017.

The most confusing things about OOP is inheritance vs subtyping vs union.

You have a dequeue class, the stack and queue inherit dequeue. But dequeue is a subtype of stack and queue.

About union, could i say an abstract class is a union (implicitly) of all its subclasses ?

No, because then all subclass members would be implicitly available on the base abstract type, I believe. So "union" here doesn't make sense.