This is the Android SDK for Confidence, a feature flagging and Experimentation system developed by Spotify.
The SDK allows you to consume feature flags and track events from your application.
The latest release of the SDK is available on Maven central.
Add the following dependency to your gradle file to use it:
implementation("com.spotify.confidence:confidence-sdk-android:0.2.1")
Where 0.2.1
is the most recent version of this SDK.
Released versions can be found under "Releases" within this repository.
You can create your Confidence
instance using the ConfidenceFactory
class like this:
val confidence = ConfidenceFactory.create(
context = app.applicationContext,
clientSecret = "<MY_SECRET>",
region = ConfidenceRegion.EUROPE
)
Where MY_SECRET
is an API key that can be generated in the Confidence UI.
Make the initial fetching of flags using the activateAndFetch
method. This is a suspending function that will fetch the flags from the server and activate them.
It needs to be run in a coroutine scope.
viewModelScope.launch {
confidence.fetchAndActivate()
}
The context is a key-value map that will be used for sampling and for targeting input in assigning feature flag values by the Confidence backend. It is also a crucial way to create dimensions for metrics generated by event data.
The Confidence SDK supports multiple ways to set the Context. Some of them are mutating the current context of the Confidence instance, others are returning a new instance with the context changes applied.
confidence.putContext("key", ConfidenceValue.String("value")) // this will mutate the context of the current Confidence instance
val otherConfidenceInstance = confidence.withContext("key", ConfidenceValue.String("value")) // this will return a new Confidence instance with the context changes applied but the context of the original instance is kept intact
Once the flags are fetched and activated, you can access their value using the getValue
method or the getFlag
method.
Both methods uses generics to return a type defined by the default value type.
The method getFlag
returns an Evaluation
object that contains the value
of the flag, the reason
for the value returned, and the variant
selected.
In the case of an error, the default value will be returned and the Evaluation
contains information about the error.
The method getValue
will simply return the assigned value or the default.
val message: String = confidence.getValue("flag-name.message", "default message")
val messageFlag: Evaluation<String> = confidence.getFlag("flag-name.message", "default message")
val messageValue = messageFlag.value
// message and messageValue are the same
Events are defined by a name
and a message
where the message is a key-value map of type <String, ConfidenceValue>
. You can track an event using the track
method.
All context
data set on the Confidence
instance will be appended to the event and its message.
confidence.track("button-tapped", mapOf("button_id" to ConfidenceValue.String("purchase_button")))
The Confidence SDK has support for EventProducer
. This is a way to programmatically emit context changes and events into streams
which can be consumed by the SDK to automatically emit events or to automatically update context data.
The Confidence SDK comes with a pre-defined event producer to emit some application lifecycle events: AndroidLifecycleEventProducer
. To use it:
import com.spotify.confidence.AndroidLifecycleEventProducer
confidence.track(
AndroidLifecycleEventProducer(
application = getApplication(),
trackActivities = false // or true
)
)
This SDK automatically emits apply
events to the Confidence backend once a flag is accessed. This allows Confidence to track who was exposed to what variant and when.
Note: the apply
event is only generated for flags that are successfully evaluated (i.e. default values returned due to errors don't generate apply
events).
Note: the apply
event reports which flag and variant was read by the application, but not which property the application has read from such variant's value.
To avoid generating redundant data, as long as the flags' data returned from the backend for a user remains unchanged, only the first time a flag's property is read will generate an apply
event. This is true also across restarts of the application.
The SDK stores apply
events on disk until they are emitted correctly, thus ensuring the apply data reaches the backend even if generated when there is no network available (assuming the device will ever re-connect to the network before the application is deleted by the user).
A Confidence Provider for the OpenFeature SDK.
The latest release of the Provider is available on Maven central.
Add the following dependency to your gradle file:
implementation("com.spotify.confidence:openfeature-provider-android:0.2.1")
Where 0.2.1
is the most recent version of the Provider. Released versions can be found under "Releases" within this repository.
The Provider is created using a confidence instance. The Provider is then set in the OpenFeature API using the setProvider
or setProviderAndWait
methods.
Please refer to the OpenFeature Kotlin SDK documentation for more information on OpenFeature specific APIs mentioned throughout this README (e.g. setProvider
, setProviderAndWait
, OpenFeature Events).
Basic usage:
val confidence = ConfidenceFactory.create(
context = app.applicationContext,
clientSecret = "<MY_SECRET>"
)
coroutineScope.launch {
OpenFeatureAPI.setProviderAndWait(
ConfidenceFeatureProvider.create(
confidence = confidence
)
)
val result = client.getBooleanValue("flag.my-boolean", false)
}
The evaluation context can be changed during the app session using OpenFeatureApi.setEvaluationContext(...)
.
After calling this method the new context is set for the provider, the flags will be fetched again and the cache and storage will be updated accordingly.
The event ProviderReady
will be emitted once the new flags are ready to be consumed by the application (note that the selected initialization strategy property doesn't play a role in this case).
Notes:
- If a flag can't be resolved from cache, the provider does NOT automatically resort to calling remote: refreshing the cache from remote only happens when setting a new provider and/or evaluation context in the global OpenFeatureAPI
- It's advised not to access flags while
setProvider
andsetEvaluationContext
are running: flag access might return the default value with reasonSTALE
during such operations. The eventProviderReady
can be used to guarantee correctness.