Android Components > Service > Fretboard
An Android framework for segmenting users in order to run A/B tests and rollout features gradually.
Use Gradle to download the library from maven.mozilla.org (Setup repository):
implementation "org.mozilla.components:service-fretboard:{latest-version}"
In order to use the library, first you have to create a new Fretboard
instance. You do this once per app launch
(typically in your Application
class onCreate
method). You simply have to instantiate the Fretboard
class and
provide the ExperimentStorage
and ExperimentSource
implementations, like this:
class SampleApp : Application() {
lateinit var fretboard: Fretboard
override fun onCreate() {
fretboard = Fretboard(
experimentSource,
experimentStorage
)
}
}
Fretboard includes a default source implementation for a Kinto backend, which you can use like this:
// Specify which HTTP (Fetch) client to use
val httpClient = GeckoViewFetchClient(context)
val fretboard = Fretboard(
KintoExperimentSource(baseUrl, bucketName, collectionName, httpClient),
experimentStorage
)
Fretboard includes support for flat JSON files as storage mechanism out of the box:
val fretboard = Fretboard(
experimentSource,
FlatFileExperimentStorage(File(context.filesDir, "experiments.json"))
)
After instantiating Fretboard
, in order to load the list of already downloaded
experiments from disk, you need to call the loadExperiments
method (don't call it
on the UI thread, this example uses a coroutine):
launch(CommonPool) {
fretboard.loadExperiments()
}
Fretboard provides two ways of updating the downloaded experiment list from the server: the first one is to directly
call updateExperiments
on a Fretboard
instance, which forces experiments to be updated immediately and synchronously
(do not call this on the main thread), like this:
fretboard.updateExperiments()
The second one is to use the provided JobScheduler
-based scheduler, like this:
val scheduler = JobSchedulerSyncScheduler(context)
scheduler.schedule(EXPERIMENTS_JOB_ID, ComponentName(this, ExperimentsSyncService::class.java))
Where ExperimentsSyncService
is a subclass of SyncJob
you create like this, providing the Fretboard
instance via the
getFretboard
method:
class ExperimentsSyncService : SyncJob() {
override fun getFretboard(): Fretboard {
return fretboard
}
}
And then you have to register it on the manifest, just like any other JobService
:
<service android:name=".ExperimentsSyncService"
android:exported="false"
android:permission="android.permission.BIND_JOB_SERVICE">
In order to check if a user is part of a specific experiment, Fretboard provides two APIs: a Kotlin-friendly
withExperiment
API and a more Java-like isInExperiment
. In both cases you pass an instance of ExperimentDescriptor
with the name of the experiment you want to check:
val descriptor = ExperimentDescriptor("first-experiment-name")
fretboard.withExperiment(descriptor) {
someButton.setBackgroundColor(Color.RED)
}
otherButton.isEnabled = fretboard.isInExperiment(descriptor)
Fretboard allows experiments to carry associated metadata, which can be retrieved using the Kotlin-friendly
withExperiment
API or the more Java-like getExperiment
API, like this:
val descriptor = ExperimentDescriptor("first-experiment-name")
fretboard.withExperiment(descriptor) {
toolbar.setColor(Color.parseColor(it.payload?.get("color") as String))
}
textView.setText(fretboard.getExperiment(descriptor)?.payload?.get("message"))
Fretboard allows you to force activate / deactivate a specific experiment via setOverride
, you
simply have to pass true to activate it, false to deactivate:
val descriptor = ExperimentDescriptor("first-experiment-name")
fretboard.setOverride(context, descriptor, true)
You can also clear an override for an experiment or all overrides:
val descriptor = ExperimentDescriptor("first-experiment-name")
fretboard.clearOverride(context, descriptor)
fretboard.clearAllOverrides(context)
Fretboard allows you to specify the following filters:
- Buckets: Every user is in one of 100 buckets (0-99). For every experiment you can set up a min and max value (0 <= min <= max <= 100). The bounds are [min, max).
- Both max and min are optional. For example, specifying only min = 0 or only max = 100 includes all users
- 0-100 includes all users (as opposed to 0-99)
- 0-0 includes no users (as opposed to just bucket 0)
- 0-1 includes just bucket 0
- Users will always stay in the same bucket. An experiment targeting 0-25 will always target the same 25% of users
- appId (regex): The app ID (package name)
- version (regex): The app version
- country (regex): country, pulled from the default locale
- lang (regex): language, pulled from the default locale
- device (regex): Android device name
- manufacturer (regex): Android device manufacturer
- region: custom region, different from the one from the default locale (like a GeoIP, or something similar).
- release channel: release channel of the app (alpha, beta, etc)
For region and release channel to work you must provide a ValuesProvider
implementation when creating the Fretboard
instance, as detailed below
Additionally, Fretboard allows you to specify a custom ValuesProvider
object in order to return a custom region,
different from the one of the current locale (perhaps doing a GeoIP or something like that), or the app
relase channel (alpha, beta, etc). It also allows you to override the values for other experiment properties
(such as the appId, country, etc):
val fretboard = Fretboard(
experimentSource,
experimentStorage,
object : ValuesProvider {
override fun getRegion() {
return custom_region
}
override fun getReleaseChannel() {
return app_channel
}
}
)
You can create a custom experiment source simply by implementing the ExperimentSource
interface:
class MyExperimentSource : ExperimentSource {
override fun getExperiments(snapshot: ExperimentsSnapshot): ExperimentsSnapshot {
// ...
return updatedSnapshot
}
}
The getExperiments
method takes an ExperimentsSnapshot
object, which contains the list of already downloaded experiments and
a last_modified date, and returns another ExperimentsSnapshot
object with the updated list of experiments.
As the getExperiments
receives the list of experiments from storage and a last_modified date, it allows you
to do diff requests, if your storage mechanism supports it (like Kinto does).
You can create a custom experiment storage simply by implementing the ExperimentStorage
interface, overriding
the save and retrieve methods, which use ExperimentsSnapshot
objects with the list of experiments and a last_modified date:
class MyExperimentStorage : ExperimentStorage {
override fun save(snapshot: ExperimentsSnapshot) {
// save snapshot to disk
}
override fun retrieve(): ExperimentsSnapshot {
// load snapshot from disk
return snapshot
}
}
The provided implementation for Kinto expects the experiments in the following JSON format:
{
"data":[
{
"name": "",
"match":{
"lang":"",
"appId":"",
"regions":[],
"country":"",
"version":"",
"device":"",
"manufacturer":"",
"region":"",
"release_channel":""
},
"buckets": {
"min": "0",
"max": "100"
},
"description":"",
"id":"",
"last_modified":1523549895713
}
]
}
This Source Code Form is subject to the terms of the Mozilla Public
License, v. 2.0. If a copy of the MPL was not distributed with this
file, You can obtain one at http://mozilla.org/MPL/2.0/