Android MVP persistence

Should I keep my data in the Repository or should I keep it within the presenter and persist it there somehow?

You have an interesting question which once I had as well when I started working with MVP. Here is my 2 cents:

There has been continuing discussions on the two options that you mentioned. I personally prefer the repository pattern. It is much cleaner IMO and the dealing with cases like screen rotation when a request is already in flight etc becomes easier with it. Also every time the device rotates, you need not fetch from the db. If you think that's expensive, maybe you can create a level of cache before db. There is a very interesting and detailed article: Presenters are not for persisting by Mike Nakhimovich, who during his time as Android Developer at New York Times worked on a library called Store which helps achieve the same. From the article:

I'd propose using data stores which abstract away data loading. Rather than the presenter needing to cache the data, we can leverage the Repository pattern and create Data Providers/Stores that can get data from the network and cache it for when we need it. For one thing, this approach makes it easier to persist your data. Another advantage is that we have a single source of truth for our data. Finally, we can eliminate things like Interactors and just have your Presenter subscribe to your Stores whenever they need something for a view.

Hope this helps a bit.

The Repository should only act as your gate for accessing data from different data sources.

For example

If you can fetch user data both from an API and and from disk (cached data), you isolate the logic of the decision of choosing which data source to get from in the Repository class. The presenter should know nothing about this. The presenter only asks for the user data

That being said

The presenter needs to hold a reference to the data it's querying (i.e. UserModel), the repo only retrieves the data from whatever data source and that's it. Same applies for any data-related operations (save, delete, update, retrieve, etc...). So if you need to hold an application scope reference to your data object, create a separate class let's say named InMemoryStore and mark it as application scope and keep the references in it. Use the Repository pattern to query the data from it.

Also important to note

Each layer should hold a different form of the data. The model used in the view layer (i.e. recycler view) should only hold the data needed for the view, nothing more nothing less. Same goes for the presenter and the model layer (usually the model layer holds the most info about the model)

This means that you will need a mapper class for each model class to convert from one form to the other. You can refer to this example for better understanding

Lastly

Since i use RxJava i didn't feel the need to use LiveData so i'm not aware enough of this topic.

You can check this full example for how to apply MVP with RxJava and Dagger

I hope this was helpful and clear :)