Android Paging 3

Google has done a good job when introducing Paging 3 in Android 11 which has so many features that simplified complicated process creating RecyclerView with paging.

If you’re looking for Paging Implementation using Coroutines and Flow, you can visit their code labs here

This tutorial serve as alternative for other who doesn’t want to use Coroutines and prefer RxJava instead.

Paging 3 Architecture

Paging 3 was designed to follow Android Architecture Component using Repository, ViewModel and UI Layer.

Paging 3 Architecture

A brief explanation for each of the layer:


Use this if you only have 1 source data, e.g.: Network or Local Storage or File


Use this if you want to load data from network and save it to your local storage. RemoteMediator will take care of getting data for you. For example during loading data, it will check the local storage first, if no data found and next page is available, it will get data from network. RemoteMediator is using 1 single source of truth for data source, that is: your local storage.


Pager will turn PagingData from your repository into stream. In this case into Flowable or Observable.

You can also do data transformation in ViewModel. For example: you want to filter data based on specific condition or you want to add section separator.


PagingDataAdapter extend RecyclerView.Adapter and is specifically created to support PagingData on UI Layer. Simply call adapter.submitData(lifecycle, list) and it will handle item insertion, update and deletion for you.


In this tutorial, we’re going to connect to TMDB to get list of movies using Paging 3 approach. Here’s the final result of the app


At the time of writing, i am using the latest version of the following libraries:

Paging: 3.0.0-alpha3

Room: 2.3.0-alpha2

Retrofit: 2.7.2

Android Studio: 4.0.1

Let’s Start!

Open up your android studio, create empty Activity. Open up your app builds.gradle and paste this code

We’re going to start the journey from Repository up to UI Layer.


Before going directly to RxPagingSource and RxRemoteMediator, lets start by preparing our Network APImodel and Room Database.

Create retrofit, response, model and mapper

We are going to load list of favourite movie from TMDB using retrofit and map it to model using mapper

Now lets create MoviesResponseMovie object and Mapper code

Here we’re creating Movies Model which has 2 inner class: Movie and MovieRemoteKeys. We are going to keep page tracking in separate table.

You can also put page tracking in one single table, in this case Movie Table. By separating it into 2 tables we are following Separation of Concern Principle.

Next we will create Image Model class to get image based on its size.

and last, here’s the mapper.

How it works

All above steps should be pretty self explanatory since its pretty standard step to get data from API and map it into Object Model

Preparing Room Database

Lets start creating 2 Dao, one for Movies and the other one for MovieRemoteKeys

Note that we’re using PagingSource when selecting movies. Yes, Room also support Paging 3 out of the box. It will automatically select movies based on previous or next page provided without extra code!

You need to pay attention also on this code:

Make sure that your query returned the same order as your API response or you will get incorrect page result when using RxRemoteMediator

Since we’re using RxJava, make sure you don’t accidentally use suspend when creating DAO

Next, we’re going to create Room Converters because we have custom data type in our model: Image and Date

Now, lets create our Database Class

How its works

Again all above step are pretty straight. We’re creating DAO for both Moviesand MovieRemoteKeys, create TypeConverter to convert object into sqlitedata type, and last we create Database Class.


If you only want to load data from 1 data source, e.g.: Remote, Local Storage or File, you can simply use RxPagingSource to achieve that.

RxPagingSource need us to override loadSingle and return Single stream.

How it works

In this class we are calling service to get favourite movies which also return Single , then we run it on background thread by calling .subscribeOn(, and we map the response into Movies Model, last we map it again into LoadResult.Page

LoadResult have 2 type parameters, the first one is the page navigation key type, usually Int or String, and the second one are the data itself, that is: list of movies.

In LoadResult.Page we assign list of movies into first parameters, and prevKey & nextKey parameter to track page.

TMDB use querystring page=1, page=2, page=3 to navigate to next or previous page, so we can simply substract current page by 1 to go to previous page, and add current page by 1 to go to next page. If your API is using different type of navigation, you can always change it to prevKey = movies.prevKey and nextKey = movies.nextKey


You can use RxRemoteMediator as alternatives if you want to use multiple layers of data, such as: Network and Local Storage. RxRemoteMediator is using your local storage as 1 single source of truth to present data. Which means all data from network will be saved to your local storage first and then shown to user via that storage.

Same as RxPagingSourceRxRemoteMediator need us to override loadSingle and return Single stream. The different between RxPagingSource is that we need to keep track of the page and also insert the result from API into database. Let’s run through the code.

How it works

  • We convert loadType into Single by calling Single.just
  • We want all process to run in background thread by calling .subscribeOn(
  • LoadType have 3 enum values: REFRESHPREPEND and APPENDREFRESH will be called during initial load, or every-time the data need to be refreshed. PREPEND will be called if user scroll to near top and APPENDwill be called if user scroll to near bottom, depends on paging configuration. Here in this code we’re getting page variable based on which enum provided to us from Movie Table.
  • getRemoteKeyClosesToCurrentPosition will search for page closes to current scroll position, if return null means this is the initial page load
  • getRemoteKeyForFirstItem will try to get first remote key found in the first movie data. This method will be called during PREPEND event, means that we should provide previous key to load movie data before scroll to top ended
  • getRemoteKeyForLastItem will try to get last remote key found in the last movie data. This method will be called during APPEND event, means that we should provide next key to load movie data before scroll to bottom ended
  • If remoteKeys.prevKey or remoteKeys.nextKey is null we will assume that user already reached the end of the page and just return INVALID_PAGE
  • After getting the key, we call flatmap to get response from API. But first we validate the page whether it’s a valid page or not, if not just tell the Mediator that the list has ended
  • Call to API is the same as PagingSource. We call the service, map it to Movies Model, and then insert the result to database by calling insertToDb and tell Mediator if this is the end of a page or not by calling MediatorResult.Success
  • during insertion, we will check if loadType enum is REFRESH or not. If yes, we clear the both table, and then we continue to insert PagingDatainto MovieRemoteKeys Table and Movie into Movies Table
  • Any exception found during the process will call MediaResult.Error


After we’ve created the Single streams, its time to setup the PagingDataand turn it into Flowable. First lets create the interface to cater our repository implementation.

Next we created a repository for RxPagingSource

Here we are creating Pager class which has public API to turn PagingDatainto Flowable Streams. We also supply the configuration during the creation. Below are the list of configuration available:

  • pageSizeMandatory, if your API has query string to show how many data will be shown use this to pass it to API. In this tutorial TMDB doesn’t have the feature to load list of movies using custom pageSize, so we use their default value.
  • enablePlaceholders: Defines whether PagingData may display null placeholders if PagingSource provides them
  • maxSize: Default MAX_SIZE_UNBOUNDED. Defines the maximum number of items that may be loaded into PagingData before pages should be dropped. Value must be at least pageSize + (2 * prefetchDistance)
  • prefetchDistance: Prefetch distance which defines how far from the edge of loaded content an access must be to trigger further loading. Typically should be set several times the number of visible items onscreen
  • initialLoadSize: Default is pageSize * 3. Defines requested load size for initial load from PagingSource, typically larger than pageSize, so on first load data there’s a large enough range of content loaded to cover small scrolls.
  • jumpThreshold: Default is COUNT_UNDEFINED. Threshold for number of items scrolled outside the bounds of loaded items to trigger invalidate.

Next we create repository for RxRemoteMediator

The setup is almost the same with RxPagingSource’s Repository, the only differences is we need to add remoteMediator and provide our database as pagingSourceFactory

We’re done in repository layer, lets move to ViewModel


in ViewModel we can do data transformation if needed and cache the transforamtion so it won’t be re-run during configuration change.

In this tutorial we only want to show all movies with posters. Here’s how you can do it using map and filter.

Thats all in ViewModel part, now lets move to UI Layer



PagingDataAdapter are created to make implementation of Paging 3 very easy in this layer. Most of the time using PagingDataAdapter is enough, however if you want to do a lot of custom modification, you can always create your own Adapter by implementing AsyncPagingDataDiffer manually.

Create a class implementing PagingDataAdapter by providing COMPARATOR as constructor parameter. Adapter are using COMPARATOR to index list and item diffing. The rest of implementations are the same as standard Adapter.

Below are the layout and ViewHolder code

I’m using coil to display the image btw 🙂

How it works

In this UI Layer we created Adapter that extend PagingDataAdapter and use MovieGridViewHolder to display the movie poster, nothing fancy.

Now let’s move to the last part of tutorial.

Activity / Fragment

As you can see from the code above, the implementation is very clean, no over complicated logic to show next or previous page. All you have to do is subscribe to the ViewModel and call mAdapter.submitData by passing fragment’s lifecycle and the PagingData


Paging 3 is great, although its rewritten completely using Coroutines and Flow, we still can use it using RxJava.

The implementation is very clear using Android Architecture Component and create a standard for all developers who want to implement Paging in their RecyclerView.

You can check full source code here:

Look for Rx dir for RxJava and Flow for Coroutines