Swift on the Server in 2023
Three years in the past I began to deal with Vapor, the preferred web-framework written in Swift, which served me very nicely through the years, however now it’s time to begin a brand new chapter in my life.
As I realized increasingly about how servers work I spotted that Vapor has it is personal execs and cons. The neighborhood modified so much in the course of the previous 3 years, some core members left and new folks began to keep up the framework. I additionally had some struggles with the default template engine (Leaf) and not too long ago I began to show away from the summary database layer (Fluent) too. One other ache level for me is the rising variety of dependencies, I barely use websockets & multipart-kit, however Vapor has these dependencies by default and you may’t eliminate them. 😢
Vapor has some very nice issues to supply, and for the general public it is nonetheless going to be a fantastic selection for constructing backends for frontends (BFFs). For me, Vapor reached its limits and I wished to make use of one thing that feels a bit lighter. Somethings that’s modular, one thing that may be simply prolonged and suits my actual wants with out extra (unused) bundle dependencies.
This shiny new factor is known as Hummingbird and it seems to be very promising. It was created by Adam Fowler who’s a member of the SSWG and in addition the primary creator of the Soto library (AWS Swift).
Hummingbird has a complete documentation obtainable on-line and a pleasant instance repository containing varied demo apps written utilizing the Hummingbird Swift server framework. I consider that the perfect a part of the the framework is modularity & extensibility. By the way in which, Hummingbird works with out Basis, nevertheless it has extensions for Basis objects, this can be a large plus for me, however perhaps that is simply my private choice these days. Hummingbird might be prolonged simply, you will discover some very helpful extensions beneath the Hummingbird venture web page, lengthy story quick it really works with Fluent and it is comparatively straightforward to get together with it when you have some Vapor data… 🤔
Getting began with Hummingbird
To begin with, there isn’t any toolbox or command line utility to assist the kickoff course of, however you’ll be able to all the time obtain the examples repository and use one of many initiatives as a place to begin. Alternatively you’ll be able to set every part up by hand, that is what we’ll do now. 🔨
As a way to construct a brand new utility utilizing the Hummingbird framework you must create a brand new listing and initialize a brand new Swift bundle utilizing the next instructions:
mkdir server && cd $_
swift bundle init --type executable
open Bundle.swift
This may create a brand new Swift bundle and open the Bundle.swift file in Xcode. You should use your individual editor in case you do not like Xcode, however both means you may have so as to add Hummingbird to your bundle manifest file as a dependency. We’ll setup an App goal for the applying itself, and a Server goal for the primary executable, which can use the applying and configure it as wanted.
import PackageDescription
let bundle = Bundle(
title: "server",
platforms: [
.macOS(.v10_15),
],
dependencies: [
.package(
url: "https://github.com/hummingbird-project/hummingbird",
from: "1.0.0"
),
.package(
url: "https://github.com/apple/swift-argument-parser",
from: "1.0.0"
),
],
targets: [
.executableTarget(
name: "Server",
dependencies: [
.product(
name: "ArgumentParser",
package: "swift-argument-parser"
),
.target(name: "App"),
]
),
.goal(
title: "App",
dependencies: [
.product(
name: "Hummingbird",
package: "hummingbird"
),
.product(
name: "HummingbirdFoundation",
package: "hummingbird"
),
],
swiftSettings: [
.unsafeFlags(
["-cross-module-optimization"],
.when(configuration: .launch)
),
]
),
.testTarget(
title: "AppTests",
dependencies: [
.product(
name: "HummingbirdXCT",
package: "hummingbird"
),
.target(name: "App"),
]
),
]
)
Please create the mandatory file and listing construction, as listed beneath, earlier than you proceed to the following steps. It is vitally vital to call issues as they seem, in any other case SPM will not work and the venture will not compile. Anyway, the venture construction is kind-of Vapor-like as you’ll be able to see. 💧
.
├── Bundle.resolved
├── Bundle.swift
├── README.md
├── Sources
│ ├── App
│ │ └── HBApplication+Configure.swift
│ └── Server
│ └── essential.swift
└── Assessments
└── AppTests
└── AppTests.swift
The following step is to create the primary entry level for the applying. For this objective Hummingbird makes use of the Swift Argument Parser library. Place the next contents into the primary.swift file:
import ArgumentParser
import Hummingbird
import App
struct HummingbirdCommand: ParsableCommand {
@Choice(title: .shortAndLong)
var hostname: String = "127.0.0.1"
@Choice(title: .shortAndLong)
var port: Int = 8080
func run() throws {
let app = HBApplication(
configuration: .init(
tackle: .hostname(hostname, port: port),
serverName: "Hummingbird"
)
)
attempt app.configure()
attempt app.begin()
app.wait()
}
}
HummingbirdCommand.essential()
The HummingbirdCommand has two choices, you’ll be able to setup a customized hostname and port by offering these values as command line choices (I will present it afterward), the applying itself will setup the tackle utilizing the enter after which it will begin listening on the required port.
The configure methodology comes from the App goal, that is the place you’ll be able to customise your server occasion, register route handlers and stuff like that, similar to you’ll do that in Vapor. The principle distinction is that Hummingbird makes use of the HB namespace, which is fairly helpful, and the configure methodology is written as an extension. Let’s write it and register a fundamental route handler. 🧩
import Hummingbird
import HummingbirdFoundation
public extension HBApplication {
func configure() throws {
router.get("/") { _ in
"Hi there, world!"
}
}
}
That is it. Now you must be capable to run your server, you’ll be able to press the Play button in Xcode that’ll begin your utility or enter one of many following instructions into the Terminal utility:
# simply run the server
swift run Server
# customized hostname and port
swift run Server --port 3000
swift run Server --hostname 0.0.0.0 --port 3000
# quick model
swift run Server -p 3000
swift run Server -h 0.0.0.0 -p 3000
# set the log stage (https://github.com/apple/swift-log#log-levels)
LOG_LEVEL=discover swift run Server -p 3000
# make launch construct
swift construct -c launch
# copy launch construct to the native folder
cp .construct/launch/Server ./Server
# run the executable
LOG_LEVEL=discover ./Server -p 3000
You may set these values in Xcode too, simply click on on the server scheme and choose the Edit Scheme… menu merchandise. Just be sure you’re on the Run goal, displaying the Arguments tag. Merely provde the Arguments Handed On Launch choices to set a customized hostname or port and you may set the log stage by including a brand new merchandise into the Atmosphere Variables part.
If you would like to unit check your utility, I’ve received a excellent news for you. Hummingbird additionally comes with a pleasant utility instrument referred to as HummingbirdXCT, which you’ll be able to simply setup & use if you would like to run some assessments in opposition to your API. In our venture, merely alter the AppTests.swift file.
import Hummingbird
import HummingbirdXCT
import XCTest
@testable import App
last class AppTests: XCTestCase {
func testHelloWorld() throws {
let app = HBApplication(testing: .dwell)
attempt app.configure()
attempt app.XCTStart()
defer { app.XCTStop() }
attempt app.XCTExecute(uri: "/", methodology: .GET) { response in
XCTAssertEqual(response.standing, .okay)
let expectation = "Hi there, world!"
let res = response.physique.map { String(buffer: $0) }
XCTAssertEqual(res, expectation)
}
}
}
As an alternative of making the applying from the primary entry level, we will arrange a brand new HBApplication occasion, import the App framework and name the configure methodology on it. the XCT framework comes with a customized XCTStart and XCTStop methodology, and you may execute HTTP requests utilizing the XCTExecute perform. The response is out there in a completion block and it is attainable to look at the standing code and extract the physique utilizing a handy String initializer.
As you’ll be able to see Hummingbird is kind of much like Vapor, nevertheless it’s light-weight and you may nonetheless add these further issues to your server when it’s wanted. Hummingbird seems like the following iteration of Vapor. I actually do not know if Vapor 5, goes to repair the problems I am presently having with the framework or not, however I do not actually care, as a result of that launch will not occur anytime quickly.