(This was originally posted as a comment on Reddit)
This blog post has been translated into Japanese, thanks to @hachi8833!
I’ve been toying around with Hanami for a few weeks now and imo it feels better than Rails. If you want to see some code rather than text, here’s my little example app I’ve been building with Hanami.
Here are the top three things that I like about Hanami, in long form:
Repository pattern instead of Active Record
Documentation link: [Hanami | Guides - Models Overview](http://hanamirb.org/guides/1.1/models/overview/) |
The Active Record pattern encourages you to throw everything into the model. Think: callbacks, validations, persistence, database logic (queries) and business logic.
In contrast, the repository pattern (through ROM) leads to a cleaner separation between your database and your application. I especially like how validations are handled by dry-validation
, a completely separate library to ROM. This enforces a separation between your models, persistence and validation layers, allowing you to pick and choose which ones you wish to use.
My favourite part about the repository pattern is that it’s incredibly difficult to make database queries from a view. This helps prevent things like N+1 queries.
Also: ROM has no such thing as callbacks, and that’s a huge plus in my eyes too. Callbacks are way too magical and – using DHH’s words – too sharp a knife for developers to have. Explicitness in code leads to a much better understanding.
Action classes
Documentation link: [Hanami | Guides - Actions Overview](http://hanamirb.org/guides/1.1/actions/overview/) |
Rather than throwing all your actions into the same controller, they’re separated out into their own classes. This helps keep each action isolated from one another, leading to overall a cleaner application architecture.
If you want to share things across actions, it’s very easy to create a module for that functionality and to include it into those action classes. For instance, if you wanted a find_post
method for show
, edit
, update
and destroy
actions, you can define this in a module and only include it in the actions you want.
A Rails controller is typically made messier by the addition of “helper” methods like this. It’s not uncommon to have Rails controllers of multiple hundreds of lines, due to the complexity of the actions and these helper methods. Separating out each action into its own class indeed makes them easier to work with.
You’re also able to test the action in complete isolation from the routing layer, as it is simply a class. I think this is what was attempted with Rails’ controller specs… but those never really felt “right” to me and I typically go to Request Specs these days instead. Hanami brings me back to testing the actions easily. I like that.
Params validation for actions
Documentation link: [Hanami | Guides - Action Parameters](http://hanamirb.org/guides/1.1/actions/parameters/) |
Somewhat related to the previous point:
Actions in Hanami also use parameter validation (with the help of dry-validation
, I think). This means that each action can uniquely validate the parameters. Rather than having create_project_params
and update_project_params
defined within a controller – as you might do in Rails – you can define what constitutes valid parameters right there in the action class.
In my Hanami experiments, I’ve not found a use for this yet, but I can think of several places where I’ve needed this sort of thing in a Rails app.
View classes
Documentation link: Hanami | Guides - Views Overview
Similarly to Action classes, Hanami also has the concept of a view class. I like these because they separate the logic of view “helper” methods and the templates very cleanly.
In a Rails application, you define a helper in a module named after the controller. I never liked this approach for two main reasons: 1) normally I would only ever use the helper in one particular view and 2) the helper is made globally available across all views in the application, which means I can’t have a similarly named method in another helper module. A quite horrible design decision.
In Hanami, the view classes are a great place to put these little helper methods and because they’re just small, isolated classes, you can test these without involving the router, controller or model.
Conclusion
As I said at the beginning: Hanami feels right. It’s “familiar” enough – still somewhat following the Model-View-Controller architecture we know-and-love Rails – that it feels comfortable to use. There’s a few places where your mind has to change into the “Hanami Way”, but ultimately once your mind does that you’ll realise that the Hanami way is better.
Hanami has demonstrably learned the very hard lessons of big Rails application development and presents a robust, well-developed alternative web framework for Ruby.
In fact, it is so robust and well-developed that if I was to develop a Ruby web application from scratch for production use, I would use Hanami over Rails – despite my 10 years Rails experience.