Write Azure Functions in Swift.
This framework supports the new Azure Functions Custom Handlers (starting from 0.6.0) in addition to the traditional custom worker.
Disclaimer: This is a community open source project, not an official Azure project
Deploy a sample project to Azure!
Classic worker sample:
Custom Handler sample:
A Timer Function (Custom Handler):
import Foundation
import AzureFunctions
import Vapor
class TimerFunction: Function {
required init() {
super.init()
self.name = "TimerFunction"
self.functionJsonBindings =
[
[
"type" : "timerTrigger",
"name" : "myTimer",
"direction" : "in",
"schedule" : "*/5 * * * * *"
]
]
//or
//self.trigger = TimerTrigger(name: "myTimer", schedule: "*/5 * * * * *")
app.post([PathComponent(stringLiteral: name)], use: run(req:))
}
func run(req: Request) -> InvocationResponse {
var res = InvocationResponse()
res.appendLog("Its is time!")
return res
}
}
An HTTP Function (Classic Worker):
import Foundation
import AzureFunctions
class HttpFunction: Function {
required init() {
super.init()
self.name = "HttpFunction"
self.trigger = HttpRequest(name: "req")
}
override func exec(request: HttpRequest, context: inout Context, callback: @escaping callback) throws {
let res = HttpResponse()
var name: String?
if let data = request.body, let bodyObj: [String: Any] = try? JSONSerialization.jsonObject(with: data, options: []) as? [String: Any] {
name = bodyObj["name"] as? String
} else {
name = request.query["name"]
}
res.body = "Hello \(name ?? "buddy")!".data(using: .utf8)
return callback(res);
}
}
Swift installation: https://swift.org/getting-started/#installing-swift
Install the latest Azure Functions Core Tools.
Just like Core Tools, Swift Functions Tools make Swift functions development easier and much more convenient.
On macOS, you can install it from Homebrew 🍺
brew install salehalbuga/formulae/swift-func
on Linux,
Clone the repo the tools repo
git clone https://github.com/SalehAlbuga/azure-functions-swift-tools
Install
make install
It installs a CLI tool called swiftfunc
that can be used to create projects, functions and run them locally.
Run the init command to create a new Azure Functions application:
swiftfunc init myApp [-hw]
It will create a new app in a new folder, and a folder named functions
inside the Sources target where Functions should be (/myApp/Sources/myApp/functions).
The project created is a Swift package project with the Azure Functions framework dependency.
Pass -hw
or --http-worker
option to create the project with the Custom Handler template.
Inside the new directory of your project, run the following to create a new HTTP Function named hello
:
swiftfunc new http -n hello [-hw]
The new function file will be created in the following path Sources/myApp/functions/hello.swift
.
Similar to the init
command, pass -hw
or --http-worker
option to create the new function with the Custom Handler template.
Run swiftfunc run
in the project directory to run your Swift Functions project locally. It will compile the code and start the host for you (as if you were running func host start
). The host output should show you the URL of hello
function created above. Click on it to run the function and see output!
There are 2 methods to deploy Swift Functions to Azure
To deploy the Function App in a Container, you can either use the Functions Core Tool func deploy
command, where it will build the image, push it to a registry and set it in the destination Function App or you can do that manually as shown below.
Build the image (Dockerfile is provided when the project is created)
docker build -t <imageTag> .
If you're using DockerHub then the tag would be username/imageName:version
.
If you're using ACR (Azure Container Registry) or any other private registry the tag would be registryURL/imageName:version
Then push it
docker push <imageTag>
In Azure portal, create a new Function App with Docker Container as the Publish option. Under Hosting options make sure Linux is selected as OS.
Once the app is created or in any existing Container Function App, under Platform Features, select Container settings and set the registry and select image you pushed.
You can use the buttons below to deploy prebuilt sample project to your Azure subscription
Custom Handler sample:
First, you need to set the following App Setting in the Function App on Azure.
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/site/wwwroot/workers/swift/lib/
Then depending if you're developing on a Linux machine or a Mac:
Login to your Azure account from Azure CLI
az login
When Azure CLI finishes loading your subscription(s) info, run:
swiftfunc publish myswiftfunctions
Swift Function Tools publish command is going to compile, export and publish your Swift Functions project.
Publishing to a Function App in a Linux Consumption Plan from macOS requires the app to be build in a Linux container first, to do that you can use VSCode Dev Containers.
The project needs to be created with the -dc
or --dev-container
option to have the Swift Function Dev Container added (or you can create a new one and copy the .devcontainer folder to your project).
swiftfunc init myFunctionApp -hw -dc
Reopen the folder in dev container (Command-Shift-P, search for and select Remote-Containers: Reopen in Container)
Once the dev container is ready, follow the same Linux steps above to publish the app!
Azure Functions offer a variety of Bindings and Triggers
The trigger, input bindings and output bindings of a Function are set in its initializer. Azure Functions in Swift must subclass the Function class from the framework.
When using the Custom Handler mode you can use all Azure Functions bindings and triggers by setting the functionJsonBindings
property to the JSON config of the bindings/triggers in Azure Functions docs. You can also use the framework supported Trigger/Binding types listed below.
Currently the following are supported by this mode. More bindings will be implemented and many improvements will be made in the future.
Swift Type | Azure Functions Binding | Direction |
---|---|---|
HttpRequest | HTTP Trigger | in |
HttpResponse | Output HTTP Response | out |
TimerTrigger | Timer Trigger | in |
Message datatype String (binding defined by Table in constructor) | Input and Ouput Table | in, out |
Message datatype String (binding defined by Queue in constructor) | Output Queue Message | out |
Message datatype String (binding defined by Queue in constructor) | Queue Trigger | in |
Blob (the blob data prob is either String or Data) | Input Blob | in |
String or Data | Output Blob | out |
Blob | Blob Trigger | in |
ServiceBusMessage | Service Bus Output Message | out |
ServiceBusMessage | Service Bus Trigger | in |
import AzureFunctions
import Vapor
class QueueFunction: Function {
required init() {
super.init()
self.name = "QueueFunction"
self.functionJsonBindings = [
[
"connection" : "AzureWebJobsStorage",
"type" : "queueTrigger",
"name" : "myQueueTrigger",
"queueName" : "myqueue",
"direction" : "in"
]
]
// or
//self.trigger = Queue(name: "myQueueTrigger", queueName: "myqueue", connection: "AzureWebJobsStorage")
app.post([PathComponent(stringLiteral: name)], use: run(req:))
}
func run(req: Request) -> InvocationResponse {
...
import AzureFunctions
class HttpFunction: Function {
required init() {
super.init()
self.name = "HttpFunction"
self.trigger = HttpRequest(name: "req")
self.inputBindings = [Blob(name: "fileInput", path: "container/myBlob.json", connection: "AzureWebJobsStorage")]
self.outputBindings = [Queue(name: "queueOutput", queueName: "myQueue", connection: "AzureWebJobsStorage")]
}
override func exec(request: HttpRequest, context: inout Context, callback: @escaping callback) throws {
...
Based on your Function's trigger type the worker will call the appropriate exec
overload. For instance, if the Function is timer-triggered, then the worker will call
exec(timer:context:callback:)
If it was an HTTP-triggered one:
exec(request:context:callback:)
You can see the list of available overloads in Xcode.
Input and Output bindings are available in the context as Dictionaries, where you can access/set the values using the binding names specified in the constructor. For example:
let tableVal = context.inputBindings["myTableInput"]
context.outputBindings["myQueueOutput"] = "new item!"
The framework uses Vapor 4.0 HTTP server. The Function
class has the app
property, thats the Vapor app instance you can use to register your functions's HTTP route.
class myFunction: Function {
required init() {
super.init()
self.name = "myFunction"
self.functionJsonBindings = [
[
"connection" : "AzureWebJobsStorage",
"type" : "queueTrigger",
"name" : "myQueueTrigger",
"queueName" : "myqueue",
"direction" : "in"
]
]
app.post([PathComponent(stringLiteral: name)], use: run(req:))
}
func run(req: Request) -> InvocationResponse {
var res = InvocationResponse()
if let payload = try? req.content.decode(InvocationRequest.self) {
res.appendLog("Got \\(payload.Data?["myQueueTrigger"] ?? "")")
}
return res
}
}
The framework also provides the function invocation Request and Response models needed for Azure Function host, which conform to Content protocol from Vapor, along with helper methods.
Invocation Request:
/// Trigger/Bindings data (values).
var data: [String:AnyCodable]?
/// Trigger/Bindings metadata.
var metadata: [String:AnyCodable]?
Invocation Request:
/// Output bindings values dictionary
var outputs: [String:AnyCodable]?
/// Functions logs array. These will be logged when the Function is executed
var logs: [String] = []
/// The $return binding value
var returnValue: AnyCodable?
As the framework is being actively updated, update the framework and the tools if you're having any issues or want to have the latest features and improvements.
To update the framework:
swift package update
To update the tools on macOS
brew upgrade salehalbuga/formulae/swift-func
on Linux
git clone https://github.com/SalehAlbuga/azure-functions-swift-tools
make install
In the generated main.swift
you can define your debug AzureWebJobsStorage
and optionally any other connections/environment vars.
Additionally, you can change the default Extension Bundle id and version.
//
// main.swift
//
//
// Auto Generated by SwiftFunctionsSDK
//
// Only set env vars or register/remove Functions. Do Not modify/add other code
//
import AzureFunctions
let registry = FunctionRegistry()
registry.AzureWebJobsStorage = "yourConnection" //Remove before deploying. Do not commit or push any Storage Account keys
registry.EnvironmentVariables = ["queueStorageConnection": "otherConnection"]
// Optionally you can change the default ExtensionBundleId and version
registry.ExtensionBundleId = "Microsoft.Azure.Functions.ExtensionBundle"
registry.ExtensionBundleVersion = "[1.*, 2.0.0)"
registry.register(hello.self)
...
Be sure not to commit any debugging Storage Account keys to a repo
Traditional Worker (Classic)
You can log using the log method in context
object
context.log(_)
Custom Handler (HTTP Worker)
Logs are returned in the InvocationResponse obj. You can append logs:
res.appendLog(_)
Traditional Worker (Classic)
When your Function is done executing the logic you should call the provided callback passing the $return
output binding value or with true
if none.
callback(res)