Sanjibukai

Elixir with Ecto (and Phoenix without Ecto)?

Hello everybody,

Tl;Dr: How to decouple the “web interface” from the “business logic” when a database is involved.

I watched the following talk from 2016 given by Lance Halvorsen https://www.youtube.com/watch?v=lDKCSheBc-8 about “Phoenix is not your application”.

I don’t know if this is something rather accepted in Elixir/Phoenix communities or if it’s only the thought of the speaker.. But I couldn’t agree more with this statement.
Particularly coming from Rails where Rails is considered your application (to quote DHH).. Which I never really agree with that.

I even heard about not using at all Ecto within Phoenix apps (applying --no-ecto) and letting a regular Elixir app taking care of the business logic. An example I heard is from the course Elixir for Programmers by Dave Thomas. But it seems (I didn’t took the course yet) that in this case the business logic doesn’t use a database.
And I somehow understand how the Phoenix web app will communicate with the Elixir “domain” app using OTP.

But, I wonder if this approach is still applicable to application that have a business logic that actually uses a database.
And so, if Ecto is intended to be used solely with Elixir? I mean without Phoenix..
Or put it in another way I want to know if Phoenix is doing a lot under the hood in using Ecto.

But I also took a look at the following talk from 2017 given by Chris McCord https://www.youtube.com/watch?v=tMO28ar0lW8 where he explains how Phoenix from 1.3 is distinguishing the web part (the web interface) from the logic/domain part.

So I wanted to ask you, Elixir/Phoenix folks, what’s the best approach you’ve experienced?

I’m asking this question in the context of an application that could become relatively large and complex regarding the logic/domain..

I’ve already had a bad experience with Rails working with large applications (very probably also because of a bad design too) and I really love the approach of having applications completely decoupled regarding the web interface and the business logic.

I hope this makes sense and I’m sorry if it’s rather confusing and also for the bad English..

2 comments

/phoenix #ecto #design #architecture #business-logic

1 1794 2

2019-12-27 15:24:14 UTC

Most Liked

outlog

No, we just tend to mix them all up - call the DB directly in controllers/channels etc. - have business logic in controllers/channels etc. etc.

It makes good sense to keep them seperate - say in MyApp(ecto and business logic) and MyAppWeb(phoenix that calls MyApp for data and data processing)

this is a great talk on DDD - but also look into various materials on using “Contexts” search this forum etc.

ps: in a large app you might even see multiple Phoenix apps..

Post #1

ityonemo

I have deployed elixir with no Phoenix or ecto, elixir with ecto and plug and no Phoenix. I’m currently deploying elixir with a phoenix admin panel, and a json API in plug (no ecto)… It all depends on what you need.

Post #2