Automerge-kt

Automatically update and merge all github pull requests (PRs) in the specified repos with the Automerge label from oldest to newest.

Configuration

Set the GITHUB_USER_TOKEN environment variable to your-user-token

(Optional) Set the AUTOMERGE_LABEL environment variable with whatever label you're using to automerge. By default Automerge will be used.

(Optional) Set the PRIORITY_LABEL environment variable with whatever label you're using for tagging something as high priority. By default Priority Automerge will be used. The Priority Automerge label allows a PR to take precedence in the merge order.

(Optional) Set the MERGE_TYPE environment variable to your desired merge type (Valid options are "merge", "squash", and "rebase"). This will determine the type of merge that is performed when merging pull requests. By default, "squash" will be used

(Optional) Set the IGNORE_OPTIONAL_STATUSES environment variable to true if you want to have the bot merge your pull requests in the event of a failing optional status. If unspecified, the bot will remove the label and leave a comment.

Replace the repository url in the config.yml with the information relative to your project (i.e. a list of the urls to as many repos as you want to run it across in parallel).

For local running only: If you want to change the rate that it runs, update the INTERVAL constant in Main.kt. Note: This does not apply for running the job on AWS.

Running it

gradle run

Running it on AWS Lambda

Set your repositories in the config.yml file
Build a jar ./gradlew shadowJar
Upload the jar in the "Function Code" section on lambda, choose "Java 8" as the runtime, and set the handler to "MainKt::handleLambda"
Set your environment variables (Minimum GITHUB_USER_TOKEN)
Set your memory usage (I'm testing with 256 MB)

Recommended: Set the project to run on a periodic basis

Add a "CloudWatch Event trigger"
Set the rule type to "schedule expression" and set the expression (e.g. rate(5, minute) This runs the function once every five minutes)

How it works

It runs on an interval against the repositories specified in the config, and for each repository it runs the following asynchronously (so that it can run all of the specified repositories in parallel):

Finds the oldest PR with the Automerge label on it (Or if there is a PR with the Priority Automerge label, it will grab the oldest PR with that label instead)
If that PR has merge conflicts or a github status has failed (e.g. CI check failure or no approvals), the label on that PR is removed, and a comment is left on the PR with a guess about why the label was removed.
If that PR still has outstanding github statuses (that is, they are currently still running), nothing is done while it waits for the statuses to come back.
If that PR has no merge conflicts and all github statuses have passed, the PR is squash-merged and the branch is deleted.

Note: It only works on a single PR per repository at a time so that it doesn't blow up your CI pipeline with builds for every labeled pull request (Including this because someone thought that it was calling update branch on every single labeled PR at once rather than dealing with PRs one-by-one from oldest to newest).

Acknowledgements