Go
This Backend Go SDK is designed to help you integrate with Unleash and evaluate feature flags inside your application.
You can use this client with Unleash Enterprise or Unleash Open Source.
Go Version
The client is currently tested against Go 1.21.x, 1.22.x, 1.23.x and 1.24.x. These versions will be updated as new versions of Go are released.
The client may work on older versions of Go as well, but is not actively tested.
Getting Started
1. Install unleash-go-sdk
To install the latest version of the client use:
2. Initialize unleash
The easiest way to get started with Unleash is to initialize it early in your application code:
Asynchronous initialization example:
Synchronous initialization example:
Preloading Feature Toggles
If you’d like to prebake your application with feature toggles (maybe you’re working without persistent storage, so Unleash’s backup isn’t available), you can replace the defaultStorage implementation with a BootstrapStorage. This allows you to pass in a reader to where data in the format of /api/client/features can be found.
Bootstrapping From File
Bootstrapping from file on disk is then done using something similar to:
Bootstrapping From S3
Bootstrapping from S3 is then done by downloading the file using the AWS library and then passing in a Reader to the just downloaded file:
Bootstrapping From Google
Since the Google Cloud Storage API returns a Reader, implementing a Bootstrap from GCS is done using something similar to
3. Use unleash
After you have initialized the unleash-client you can easily check if a feature toggle is enabled or not.
4. Stop unleash
To shut down the client (turn off the polling) you can simply call the destroy-method. This is typically not required.
unleash.Close()
Built in Activation Strategies
The Go client comes with implementations for the built-in activation strategies provided by unleash.
- DefaultStrategy
- UserIdStrategy
- FlexibleRolloutStrategy
- GradualRolloutUserIdStrategy
- GradualRolloutSessionIdStrategy
- GradualRolloutRandomStrategy
- RemoteAddressStrategy
- ApplicationHostnameStrategy
Read more about activation strategies in the docs.
Unleash Context
In order to use some of the common activation strategies you must provide an
unleash-context.
This client SDK allows you to send in the unleash context as part of the isEnabled call:
Caveat
This client uses go routines to report several events and doesn’t drain the channel by default. So you need to either register a listener using WithListener or drain the channel “manually” (demonstrated in this example).
Feature Resolver
FeatureResolver is a FeatureOption used in IsEnabled via the WithResolver.
The FeatureResolver can be used to provide a feature instance in a different way than the client would normally retrieve it. This alternative resolver can be useful if you already have the feature instance and don’t want to incur the cost to retrieve it from the repository.
An example of its usage is below:
Impression Data
When impression data is enabled on a flag, the SDK will emit impression events during evaluation. You can hook into these with a custom listener to collect insights or integrate with analytics. You can read more about impression data in Unleash’s documentation.
Development
To override dependency on unleash-go-sdk github repository to a local development folder (for instance when building a local test-app for the SDK),
you can add the following to your apps go.mod:
Steps to Release
- Update the clientVersion in
client.go - Tag the repository with the new tag
- Create the release manually in Github
Adding Client Specifications
In order to make sure the unleash clients uphold their contract, we have defined a set of client specifications that define this contract. These are used to make sure that each unleash client at any time adhere to the contract, and define a set of functionality that is core to unleash. You can view the client specifications here.
In order to make the tests run please do the following steps.
Requirements:
- make
- golint (go get -u golang.org/x/lint/golint)
Run tests:
make
Run lint check:
make lint
Run code-style checks:(currently failing)
make strict-check
Run race-tests:
make test-race
Benchmarking
You can benchmark feature toggle evaluation by running:
Here’s an example of how the output could look like:
In this example the benchmark was run on a MacBook Pro (M1 Pro, 2021) with 16GB RAM.
We can see a result of 854.3 ns/op, which means around 101.131 billion feature toggle evaluations per day.
Note: The benchmark is run with a single CPU core, no parallelism.
Design Philosophy
This feature flag SDK is designed according to our design philosophy. You can read more about that here.