Scheduling tweets with GitHub Actions and Swift

RevenueCat logo
Relax, you can roll back your mobile release

No one is immune from shipping critical bugs to production, but Runway helps you limit the amount of havoc that can cause.

I have recently started a new tradition on Twitter where I post the week’s top three most-read articles from my blog every Friday:

I have been doing this manually by generating aggregate page view reports from Fathom, the tool I use to collect analytics from my site. As Fathom has an API that allows you to generate such reports, I saw an opportunity to automate this workflow and save myself some time.

I set myself a challenge for this week to write a cron job which runs every friday at 12:00 pm, finds the top 3 most-read articles form that week and then uses Twitter’s API to post the results in a suitable format. If you have read any articles in my blog, you will have guessed by now that I will be making all the tooling required for this cron job using Swift 🚀.

Creating a command-line tool

The first step to writing this cron job is to create a command-line tool which queries Fathom’s API to find the most-read articles and then uses Twitter’s API to share this report with my followers.

To write this in Swift, I first created an executable Swift package with no dependencies:

mkdir weekly-tweet-cron-job && cd weekly-tweet-cron-job
swift package init --type executable --name WeeklyTweetCronJob

Getting the most-read articles

I then used async/await and URLSession to get a report of the unique visitors for the article paths in my blog between two given dates using Fathom’s API.

First, I retrieved all the necessary environment variables needed to authenticate with the API and created a request with both all necessary query parameters to filter the right data using the URLComponents API and an Authorization header with a bearer token I retrieved from Fathom’s console.

public struct WeeklyTweetCronJob {
    public static func main() async {
        // 1
        guard let fathomEntity = ProcessInfo.processInfo.environment["FATHOM_ENTITY_ID"],
              let fathomToken = ProcessInfo.processInfo.environment["FATHOM_TOKEN"] else {
            print("Missing environment variables...")

        // 2
        let currentDate = Date()
        let aWeekAgoDate = Calendar(identifier: .iso8601).date(byAdding: .day, value: -7, to: currentDate)!

        // 3
        var components = URLComponents()
        components.scheme = "https" = ""
        components.path = "/v1/aggregations"
        components.queryItems = [
            URLQueryItem(name: "entity", value: "pageview"),
            URLQueryItem(name: "entity_id", value: fathomEntity),
            URLQueryItem(name: "aggregates", value: "uniques"),
            URLQueryItem(name: "field_grouping", value: "pathname"),
            URLQueryItem(name: "sort_by", value: "uniques:desc"),
            URLQueryItem(name: "timezone", value: "Europe/London"),
            URLQueryItem(name: "date_from", value: aWeekAgoDate.ISO8601Format()),
            URLQueryItem(name: "date_to", value: currentDate.ISO8601Format()),
            URLQueryItem(name: "filters", value: "[{\"property\":\"pathname\", \"operator\":\"is like\", \"value\":\"/*-*\"}]")

        // 4
        var request = URLRequest(url: components.url!)
        request.setValue("Bearer \(fathomToken)", forHTTPHeaderField: "Authorization")

If you’re a Fathom user and want to learn more about what each of the query parameters do, you can find them all documented in their site.

I then used URLSession’s async/await APIs to execute the request above and decode the response’s data into a Decodable model containing just the information I needed:

struct PageView: Decodable {
    var uniques: Int
    let pathname: String

    enum CodingKeys: CodingKey {
        case uniques
        case pathname

    init(from decoder: Decoder) throws {
        let container = try decoder.container(keyedBy: CodingKeys.self)
        let uniquesString = try container.decode(String.self, forKey: .uniques)
        self.uniques = Int(uniquesString)!
        self.pathname = try container.decode(String.self, forKey: .pathname)
            .replacingOccurrences(of: "/", with: "")

public static func main() async {
    // ...
    do {
        let (data, _) = try await request)
        let pageViews = try! JSONDecoder().decode([PageView].self, from: data)
    } catch {
        print("Something went wrong making the request: \(error.localizedDescription)")

Note that the Decodable model above implements custom decoding to convert the uniques proterty type from String to Int.

Now that I had a number of aggregated page view events over the course of a week, I had to make sure that all variations of a single path were being accounted for. For example the pathname for a specific page view could either be /my-blog-post or /my-blog-post/ and would therefore be counted as two separate articles. I made use of the reduce method in Swift to merge all variations of the same path into a single page view event with all the aggregated unique visitors:

public static func main() async {
    // ...
    do {
        let (data, _) = try await request)
        let pageViews = try! JSONDecoder().decode([PageView].self, from: data)
        let uniquedPageViews = pageViews.reduce(into: [PageView]()) { partialResult, pageView in
            if let index = partialResult.firstIndex(where: { $0.pathname == pageView.pathname }) {
                partialResult[index].uniques += pageView.uniques
            } else {
    } catch {
        print("Something went wrong making the request: \(error.localizedDescription)")

Writing the tweet

Once I had the data for how many unique visitors each of the articles in my blog had, I had to sort them in descending order and get the top 3:

public static func main() async {
    // ...
    let topArticles = uniquedPageViews
        .sorted(by: { lhs, rhs in lhs.uniques > rhs.uniques })

With the top articles available, I then had to write the tweet’s content, which turned into an interesting challenge.

I first had to generate a list of the top articles with their URLs prepended by their position in the rank in emoji form (1️⃣, 2️⃣ or 3️⃣). I could have done this in a number of easier ways but, since I wanted to learn more about unicode, I decided to make an emoji out of unicode characters based on the iteration’s index.

Finally, I used the join method with a new line as the separator to merge all article strings into a single one, which I then appended to the rest of the tweet’s static content:

public static func main() async {
    // ...
    let topArticlesList = uniquedPageViews.enumerated().map { index, pageView in
        let emoji = [
            UnicodeScalar(0x0031 + index),
        .compactMap { $0 }
        .map { String($0) }

        return "\(emoji)\(pageView.pathname)"
    .joined(separator: "\n")

    let tweet = """
    Happy Friday everyone! 👋

    Hope you've all had a great week. Here's a look back at the week's most read articles in my blog:


    #iosdev #swiftlang

Posting the tweet

To post the tweet using Twitter’s API, I had to generate and adapt a URLRequest with OAuth1 headers following Twitter’s API documentation and send a POST request to the endpoint with the tweet’s text as its body:

public static func main() async {
    // ...
    // Create a URL request
    let url = URL(string: "")!
    var request = URLRequest(url: url)
    request.httpMethod = "POST"
    request.httpBody = try JSONEncoder().encode(Tweet(text: tweet))

    // Add oAuth1 key-value pairs to `URLRequest` headers
    let oAuth1 = OAuth1(key: twitterAPIKey, secret: twitterAPISecret, token: twitterAPIToken, tokenSecret: twitterAPITokenSecret)
    let adaptedRequest = try oAuth1.adaptRequest(request)

    // Make the request
    _ = try await adaptedRequest)

And that was it! I did a couple of test runs of the command-line tool and verified that it posted the right content to Twitter.

Making a cron job

At this stage I had the command-line tool which generates a report and posts a tweet and all I was missing was a way to schedule it to run every Friday.

One of my favourite GitHub features is that you get unlimited GitHub Actions minutes for public repositories. This, in combination with its ability to define cron jobs to schedule workflows, made it a perfect candidate for the task I had in hand.

Creating a scheduled workflow is as simple as adding the schedule trigger to the .yml file with the desired frequency in the form of a cron expression:

name: Tweet most read posts

    - cron: 0 12 * * 5
    runs-on: macos-12

      - uses: actions/checkout@v3
      - run: swift run -c release WeeklyTweetCronJob
          FATHOM_TOKEN: ${{Secrets.FATHOM_TOKEN}}
          TWITTER_API_KEY: ${{Secrets.TWITTER_API_KEY}}

Note that I also added the ability of running the workflow manually using the workflow-dispatch trigger so that I could test it without having to wait for Friday at 12 pm.

If you want to dig into the code and find out more about the implementation details, I have made the repository public on GitHub.