Android Architecture: MV?

Written on April 1, 2015

How do the MV* ways of thinking relate to Android? Why is this now such a popular way of thinking about Android architecture?

The Problem

I’m a contractor and I’m often brought on to projects have long since passed their first commit. I feel like I have had a good insight how to the average dev approaches Android app architecture from this, and also from open-source projects, android blog posts, the Android developer site and more. One thing that strikes me about most of the code I see is that there is not much of a notion of a clean, overarching application architecture. By this I mean that people seem genuinely happy to shoehorn their code into Activitys & Fragments with the occasional class named something like MyCrazyController. Why is this an issue? The problem with this is that it results in

  • hard to test code
  • hard to read code
  • hard to refactor code
  • hard to reuse code
  • hard to handle-config changes code
  • hard to write asynchronous code
  • hard to not create loads of edge-case code (i.e. its buggy)
  • hard to hand-over code

Its an easy thing to fall into doing because

  • ~99% of all Android examples are written in this way
  • It sort of works most of the time
  • there are framework classes (i.e. Loaders) and librarys (i.e. RoboSpice) around now to work around some of the issues that manifest from using this stock approach. While I have used these approaches in the past I find they can make a code base more complicated than it needs to be.
  • People have got used to hearing that “Android is hard to test” or “just lock it to portrait” and “restart / reconnect to the web service call on rotation” and just think “well thats how it is”

I admit, I would also include my past-self into this group, which is why I am here, attempting to atone for my coding sins and mine and others unknowingly devout following of the book of anti-patterns.

How has this come to be?

Confused Droid

MV?, who’s in control?

I believe a big part of this is that not much thought has gone into the actual generic app architecture approach that devs could use. In my opinion there is a vast lack of some sort of (and I use this term extremely loosely) Controller for application components.

SNAP QUIZ: What does the term Controller mean to you?

There seem to be a few general camps of thought relating to MVC architecture and Android

  1. “What’s this MVC?”
  2. XML is the View, Activity / Fragment is the Controller, SQLite / in memory data is the Model
  3. Activity / Fragment is the View, some crazy class is the Controller with many mixed responsibilities, SQLite / in memory data is the Model
  4. I use some open source library and guidelines to structure my app
  5. I have my own solution

~98% of the apps I have seen fall into 1,2 & 3. Maybe my sample size is too small, or maybe people just don’t have time to think about this stuff, either way, there is a lot of room for improvement here.

A brief overview of MV* Architectures

My next post will introduce an approach that works well for me. The rest of this post will outline the MV~ family of terms. The goal of this is to show that they are ambigous and should serve only as a rough starting point in discussion, and are certainly not anything close to a design pattern.

MVC (Model-View-Controller)

Wiki link

First things first. The term MVC is in my opinion highly subjective and highly abstract. The pain point seems to be around the definition of Controller. Really its more of a concept that a design pattern. Put simply and by someone (Martin Fowler) who knows a bucketload more than myself

Take Model-View-Controller as an example. It’s often referred to as a pattern, but I don’t find it terribly useful to think of it as a pattern because it contains quite a few different ideas. Different people reading about MVC in different places take different ideas from it and describe these as ‘MVC’. If this doesn’t cause enough confusion you then get the effect of misunderstandings of MVC that develop through a system of Chinese whispers. (From GUI Architectures)

Reading up on the history of MVC, by the fact it was invented in the 70’s at the time when UI stuff was in its infancy its not surprising it has taken on many different faces over the years when encountering new tech stacks and frameworks.

MVC can be interpreted in two ways it seems

  1. (less-often) An over-arching term, which is a superset of all MV* architectures that separate some View from some Model and other middle-ish layer than may or may not communicated with the View & Model components directly.
  2. (more-often) A slightly more concrete notion of a pattern that relates the M V and C components in a triangular relationship which does not allow the user to interact with the View directly but only via a View and / or Controller. View displays the Model and Model is manipulated via the Controller.

Unfortunately in practice I feel (from code I have audited) is that some devs feel like naming a class Controller is enough to ensure ‘good’ architecture, even if that controller class breaks every rule in the book of OOP (ok, I’m thinking of one project only here, but this is also the only project I saw where the dev even attempted to pull the ‘controller code’ outside of the Activity) including leaking views via static references in the Controller code.

Anyhow, my advise would be not to actually think about MVC very much. IMHO its too abstract and therefore loses one of the major powers of design-patterns in the first place, which is to quickly convey a structural concept between developers. It is likely to cause more confusion than a more explicit term.

MVP (Model-View-Presenter)

Wiki link

MVP is again quite loose. Even from its inception people had different ideas of what responsibilities the Presenter component had.

One of the variations in thinking about MVP is the degree to which the presenter controls the widgets in the view. On one hand there is the case where all view logic is left in the view and the presenter doesn’t get involved in deciding how to render the model. This style is the one implied by Potel. The direction behind Bower and McGlashan was what I’m calling Supervising Controller, where the view handles a good deal of the view logic that can be described declaratively and the presenter then comes in to handle more complex cases.

You can also move all the way to having the presenter do all the manipulation of the widgets. This style, which I call Passive View isn’t part of the original descriptions of MVP but got developed as people explored testability issues. I’m going to talk about that style later, but that style is one of the flavors of MVP. (Also from GUI Architectures)

The thing that all MVP approaches seem to have in common is that the Presentor has some notion of what the the UI can do and the state of the UI elements. Depending on where you look this can be anything from the Presenter object having the same interface in code as the UI does to the user to a slight notion of what colors things should display etc.

Unlike the triangular relationship of components in the general MVC way of thinking MVP is linear, meaning the Presenter sits between the Model and the View.

For more info I would recommend you check out another Martin Fowler article Retirement note for Model View Presenter Pattern.

MVVM (Model-View-ViewModel)

Wiki link

This is a popular one in the Microsoft (& Xamarin) world. The thing that all MVVM architectures have in common is binding!

This can be Data-binding, which in a nutshell means the ViewModel can define the binding between Views and data propertie. If bound two ways then changes to the view update the model and changes to the model update the view, pretty much for free.

Or this can be Operation-binding in the form of Commands. See Commands in MVVM for more info.

There are some MVVM librarys for Android, but I have never played with them, or spoken to anyone who has. If you interested your can check out

EDIT (29/05/15): IO15 announced a new data-binding lib

MVA (Model-View-Adapter)

Wiki link

This one is interesting as its pretty simple. This is like a linear MVC (the same as MVP above), meaning that there is no communication between the View and the Model, the Adapter sits in-between them like a burger in a bun.

From my own reading this seems like it can easily be the same as an MVP pattern if the Presenter/Adapater isn’t storing much View state. As you can see, its all a bit of a gray area!



The above are the main ones you see around at present. I have come to the conclusion that whenever you hear about MV* to actually look at how the existing code or framework is actually doing the things above, to get a handle on whats really going on.

What Now?

I have outlined a few common architectural terms and is laying the way for me to talk about a pattern than I have found works for me that fixes the issues outlined at the beginning of this post.

Check out my next post Android Architecture: Introducing Dynamo

Down the rabbit hole (links)

While MV* approaches are often talked about, the following are also used on Android and the subject of the occasional discussion.

Appendix 1: iOS ViewControllers

I have extremely limited iOS exposure but from poking around it seems that the iOS version of MVC is closer to MVP as the Controller directly manipulates the Views and also contains ‘traditional’ controller logic (even though posts and docs say its MVA). Would be interested to hear an iOS devs take on this. Quite a lot of responsibilities which it seems earnt the title of Massive View Controller.