Go SDK for building Fastly Compute applications with Go (1.21+) and TinyGo (0.28.1+).
The Fastly Developer Hub has a great Quick Start guide for Go.
Alternatively, you can take a look at the Go Starter Kit.
If you're using TinyGo, you'll also want to take a look at our TinyGo Recommended Packages section, as this can help with the sharp edges of the SDK, like JSON support.
Compute builds on top of WebAssembly and the WebAssembly System Interface.
TinyGo supports WASI as a target, and Go does as of its 1.21 release.
Each toolchain has its own tradeoffs. Generally speaking, TinyGo produces smaller compiled artifacts and takes less RAM at runtime. Build times are generally longer, sometimes considerably. TinyGo does not support all of the Go standard library, and in particular support for the reflect package is incomplete. This means that some third-party packages may not work with TinyGo.
Runtime performance is mixed, with TinyGo faster on some applications and Go faster on others. If you have a performance-critical application, we recommend benchmarking both toolchains to see which works best for you.
To switch between TinyGo and Go, set the build command in the [scripts] section of your fastly.toml as follows:
[scripts]
build = "tinygo build -target=wasi -o bin/main.wasm ."
or
[scripts]
build = "GOARCH=wasm GOOS=wasip1 go build -o bin/main.wasm ."
You might need to adjust the actual build command depending on your project.
If you're using Go, download the latest Go release. For TinyGo, follow the TinyGo Quick install guide.
Then, you can install compute-sdk-go in your project by running:
go get github.com/fastly/compute-sdk-go
Examples can be found in the examples directory.
The Fastly Developer Hub has a collection of common use cases in VCL ported to Go which also acts as a great set of introductory examples of using Go on Compute.
The API reference documentation can be found on pkg.go.dev/github.com/fastly/compute-sdk-go.
Tests that rely on a Compute runtime can utilize Viceroy, our local development tool.
Install Viceroy and ensure the viceroy command is available in your path.
Write your tests as ordinary Go tests. Viceroy provides the Compute APIs locally, although be aware that not all platform functionality is available. You can look at the integration_tests directory for examples.
The fastly-compute.json file provides a TinyGo target to run Viceroy. (In the future, we will include this in the Go starter kits.)
To run your tests:
tinygo test -target=targets/fastly-compute-wasip1.json ./...
You can try it out and make sure your local Viceroy environment is set up correctly by running the integration tests in this repository:
tinygo test -target=targets/fastly-compute-wasip1.json ./integration_tests/...
To run tests with Viceroy and Go
GOARCH=wasm GOOS=wasip1 go test -exec "viceroy run -C fastly.toml" -v ./...
You can try it out and make sure your local Viceroy environment is set up correctly by running the integration tests in this repository:
GOARCH=wasm GOOS=wasip1 go test -exec "viceroy run -C fastly.toml" -v ./integration_tests/...
Logging can be done using a Fastly Compute Log Endpoint (example), or by using normal stdout like:
fmt.Printf("request received: %s\n", r.URL.String())
TinyGo is still a new project, which has yet to get a version 1.0.0. Therefore, the project is incomplete, but in its current state can still handle a lot of tasks on Compute. However, some languages features of Go are still missing.
To help with your adoption of compute-sdk-go, here are some recommended packages to help with some of the current missing language features:
TinyGo's reflect support (which is needed by encoding/json among other things) is still new. While most use cases should work, for performance or other compatibility reasons you might need to consider a third-party JSON package if the standard library doesn't meet your needs.
The changelog can be found here.
If you find any security issues, see the Fastly Security Reporting Page or send an email to: [email protected]
Note that communications related to security issues in Fastly-maintained OSS as described here are distinct from Fastly security advisories.