Parallelize two or more async functions

Fork is a Swift library that allows for parallelizing multiple async functions. It provides a Fork struct that takes a single input and splits it into two separate async functions that return different outputs. The two functions can then be merged into one which returns a single output.

Asynchronous programming in Swift can be made easier with the async-await syntax, but it can still be challenging to parallelize multiple functions. Fork simplifies this by allowing developers to create parallel tasks with ease.

The Swift Book has the following example for downloading multiple images.

let firstPhoto = await downloadPhoto(named: photoNames[0])
let secondPhoto = await downloadPhoto(named: photoNames[1])
let thirdPhoto = await downloadPhoto(named: photoNames[2])

let photos = [firstPhoto, secondPhoto, thirdPhoto]
show(photos)

Now the code above is still asynchronous, but will only run one function at a time. In the example above, firstPhoto will be set first, then secondPhoto, and finally thirdPhoto.

To run these three async functions in parallel we need to change the code to this following example.

async let firstPhoto = downloadPhoto(named: photoNames[0])
async let secondPhoto = downloadPhoto(named: photoNames[1])
async let thirdPhoto = downloadPhoto(named: photoNames[2])

let photos = await [firstPhoto, secondPhoto, thirdPhoto]
show(photos)

The above code will now download all three photos at the same time. When all the photos have been downloaded it will show the photos.

Now, using Fork we could simiplfy this to just a couple of lines!

let photos = try await photoNames.asyncMap(downloadPhoto(named:))
show(photos)

When using Fork, functions will be ran in parallel and higher order forks will also be ran in parallel.

• Fork: Using a single input create two separate async functions that return LeftOutput and RightOutput.
• ForkedArray: Using a single array and a single async function, parallelize the work for each value of the array.
• BatchedForkedArray: Using a single array and a single async function, batch the parallelized work for each value of the array
• ForkedActor: Using a single actor create two separate async functions that are passed the actor.
  • KeyPathActor: A generic Actor that uses KeyPaths to update and set values.

import Fork

let fork = Fork(
    value: 10,
    leftOutput: { $0.isMultiple(of: 2) },
    rightOutput: { "\($0)" }
)
        
let leftOutput = try await fork.left()
let rightOutput = try await fork.right()

XCTAssertEqual(leftOutput, true)
XCTAssertEqual(rightOutput, "10")
        
let output: String = try await fork.merged { bool, string in
    if bool {
        return string + string
    }
        
    return string
}
        
let output = await mergedFork()

XCTAssertEqual(output, "1010")

A ForkedArray makes it easy to perform an asynchronous function on all of the elements in an Array. ForkedArray helps with the example above.

let forkedArray = ForkedArray(photoNames, map: downloadPhoto(named:))
let photos = try await forkedArray.output()

The BatchedForkedArray allows you to efficiently parallelize and batch process an array of values using an async function. It provides methods for both resolving the parallelized array in a single output as well as streaming the batches of the resolved array.

let batchedForkedArray = BatchedForkedArray(photoNames, batch: 3, map: downloadPhoto(named:))
let photos = try await forkedArray.output()

In the above example, we create an instance of BatchedForkedArray with a batch size of 3 and the downloadPhoto function as the map closure.

To resolve the batched array, we use the output() method, which executes the downloadPhoto function on each batch of photo names in parallel. After the resolution is complete, the photos array will contain the downloaded photos in the order they were processed.

let photoNames = [Int](0 ..< 100)

let batchedForkedArray = BatchedForkedArray(
    photoNames,
    batch: 5,
    map: downloadPhoto(named:)
)

for try await batch in batchedForkedArray.stream() {
    for photo in batch {
        // Perform operations on each photo in the batch
        print(photo)
    }
}

In this example, we create an instance of BatchedForkedArray with a batch size of 5 and the downloadPhoto(named:) function as the map closure. By using the stream() method, we can iterate over batches of photo names asynchronously.

Within the for-await loop, each batch of photo names is processed asynchronously. We then iterate over each photo in the batch and perform operations accordingly. This allows for efficient processing of large datasets in batches while controlling the number of parallel processes running at once.

actor TestActor {
    var value: Int = 0
    
    func increment() {
        value += 1
    }
}

let forkedActor = ForkedActor(
    actor: TestActor(),
    leftOutput: { actor in
        await actor.increment()
    },
    rightOutput: { actor in
        try await actor.fork(
            leftOutput: { await $0.increment() },
            rightOutput: { await $0.increment() }
        )
        .act()
    }
)

let actorValue = await forkedActor.act().value

XCTAssertEqual(actorValue, 3)

let forkedActor = ForkedActor(
    value: 0,
    leftOutput: { actor in
        await actor.update(to: { $0 + 1 })
    },
    rightOutput: { actor in
        try await actor.fork(
            leftOutput: { actor in
                await actor.update(to: { $0 + 1 })
            },
            rightOutput: { actor in
                await actor.update(\.self, to: { $0 + 1 })
            }
        )
        .act()
    }
)

let actorValue = try await forkedActor.act().value

XCTAssertEqual(actorValue, 3)

• Vapor ForkedActor Example
• ForkedArray Pictures Example

Concurrently generates an App Icon for your iOS app from an SVG, in the accompanying XCAssets file structure.