Skip to content
πŸ›  A toolkit for building GitHub Actions in Node.js
TypeScript JavaScript Dockerfile
Branch: master
Clone or download

Latest commit

Latest commit 937540f Apr 5, 2020

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.github Update workflow Apr 6, 2020
bin Replace standard/tslint with eslint Feb 23, 2020
scripts Replace standard/tslint with eslint Feb 23, 2020
src Merge remote-tracking branch 'origin/master' into octokit-17 Apr 4, 2020
tests Fix prepublish warning Apr 4, 2020
.eslintignore Replace standard/tslint with eslint Feb 23, 2020
.eslintrc.js Fix some linting issues Feb 23, 2020
.gitignore To typescript we go Nov 4, 2018
LICENSE Init to winit Nov 4, 2018
README.md fix broken actions events link Mar 19, 2020
codecov.yml Init to winit Nov 4, 2018
package-lock.json 4.0.0 Apr 4, 2020
package.json 4.0.0 Apr 4, 2020
tsconfig.json Remove a comment Feb 2, 2019

README.md

GitHub Actions Toolkit

An opinionated toolkit for building GitHub Actions in Node.js
Usage β€’ API β€’ How to test your Action β€’ FAQ

GitHub Actions status Codecov

This toolkit is an opinionated alternative to (and wrapper around) the official toolkit. actions/toolkit makes many features optional in the interest of performance, so you may prefer to use it instead of this library.

Usage

Installation

$ npm install actions-toolkit
const { Toolkit } = require('actions-toolkit')
const tools = new Toolkit()

Bootstrap a new action

$ npx actions-toolkit my-cool-action

This will create a new folder my-cool-action with the following files:

β”œβ”€β”€ Dockerfile
β”œβ”€β”€ action.yml
β”œβ”€β”€ index.js
β”œβ”€β”€ index.test.js
└── package.json

API

Toolkit options

event (optional)

An optional list of events that this action works with. If omitted, the action will run for any event - if present, the action will exit with a failing status code for any event that is not allowed.

const tools = new Toolkit({
  event: ['issues', 'pull_requests']
})

You can also pass a single string:

const tools = new Toolkit({
  event: 'issues'
})

And/or strings that include an action (what actually happened to trigger this event) for even more specificity:

const tools = new Toolkit({
  event: ['issues.opened']
})

secrets (optional)

You can choose to pass a list of secrets that must be included in the workflow that runs your Action. This ensures that your Action has the secrets it needs to function correctly:

const tools = new Toolkit({
  secrets: ['SUPER_SECRET_KEY']
})

If any of the listed secrets are missing, the Action will fail and log a message.

token (optional)

You can pass a custom token used for authenticating with the GitHub API:

const tools = new Toolkit({
  token: '1234567890abcdefghi'
})

process.env.GITHUB_TOKEN will be used if no token was passed.

Toolkit.run

Run an asynchronous function that receives an instance of Toolkit as its argument. If the function throws an error (or returns a rejected promise), Toolkit.run will log the error and exit the action with a failure status code.

The toolkit instance can be configured by passing Toolkit options as the second argument to Toolkit.run.

Toolkit.run(async tools => {
  // Action code
}, { event: 'push' })

tools.github

Returns an Octokit SDK client authenticated for this repository. See https://octokit.github.io/rest.js for the API.

const newIssue = await tools.github.issues.create({
  ...tools.context.repo,
  title: 'New issue!',
  body: 'Hello Universe!'
})

You can also make GraphQL requests:

const result = await tools.github.graphql(query, variables)

See https://github.com/octokit/graphql.js for more details on how to leverage the GraphQL API.


tools.log

This library comes with a slightly-customized instance of Signale, a great logging utility. Check out their docs for the full list of methods. You can use those methods in your action:

tools.log('Welcome to this example!')
tools.log.info('Gonna try this...')
try {
  risky()
  tools.log.success('We did it!')
} catch (error) {
  tools.log.fatal(error)
}

In the GitHub Actions output, this is the result:

β„Ή  info      Welcome to this example!
β„Ή  info      Gonna try this...
βœ–  fatal     Error: Something bad happened!
    at Object.<anonymous> (/index.js:5:17)
    at Module._compile (internal/modules/cjs/loader.js:734:30)

tools.inputs

GitHub Actions workflows can define some "inputs" - options that can be passed to the action:

uses: JasonEtco/example-action@master
with:
  foo: bar

You can access those using tools.inputs:

console.log(tools.inputs.foo) // -> 'bar'

Note! This is not a plain object, it's an instance of Proxy, so be aware that there may be some differences.


tools.command(command, (args, match) => Promise)

Respond to a slash-command posted in a GitHub issue, comment, pull request, pull request review or commit comment. Arguments to the slash command are parsed by minimist. You can use a slash command in a larger comment, but the command must be at the start of the line:

Hey, let's deploy this!
/deploy --app example --container node:alpine
tools.command('deploy', async (args: ParsedArgs, match: RegExpExecArray) => {
  console.log(args)
  // -> { app: 'example', container: 'node:alpine' }
})

The handler will run multiple times for each match:

/deploy 1
/deploy 2
/deploy 3
let i = 0
await tools.command('deploy', () => { i++ })
console.log(i)
// -> 3

tools.getPackageJSON()

Get the package.json file in the project root and returns it as an object.

const pkg = tools.getPackageJSON()

tools.getFile(path, [encoding = 'utf8'])

Get the contents of a file in the repository.

const contents = tools.getFile('example.md')

tools.runInWorkspace(command, [args], [ExecaOptions])

Run a CLI command in the workspace. This uses execa under the hood so check there for the full options. For convenience, args can be a string or an array of strings.

const result = await tools.runInWorkspace('npm', ['audit'])

tools.token

The GitHub API token being used to authenticate requests.


tools.workspace

A path to a clone of the repository.


tools.exit

A collection of methods to end the action's process and tell GitHub what status to set (success, neutral or failure). Internally, these methods call process.exit with the appropriate exit code. You can pass an optional message to each one to be logged before exiting. This can be used like an early return:

if (someCheck) tools.exit.neutral('No _action_ necessary!')
if (anError) tools.exit.failure('We failed!')
tools.exit.success('We did it team!')

tools.store

Actions can pass information to each other by writing to a file that is shared across the workflow. tools.store is a modified instance of flat-cache:

Store a value:

tools.store.set('foo', true)

Then, in a later action (or even the same action):

const foo = tools.store.get('foo')
console.log(foo)
// -> true

Note: the file is only saved to disk when the process ends and your action completes. This is to prevent conflicts while writing to file. It will only write to a file if at least one key/value pair has been set. If you need to write to disk, you can do so with tools.store.save().


tools.context

tools.context.action

The name of the action

tools.context.actor

The actor that triggered the workflow (usually a user's login)

tools.context.event

The name of the event that triggered the workflow

tools.context.payload

A JSON object of the webhook payload object that triggered the workflow

tools.context.ref

The Git ref at which the action was triggered

tools.context.sha

The Git sha at which the action was triggered

tools.context.workflow

The name of the workflow that was triggered.

tools.context.issue

The owner, repo, and number params for making API requests against an issue or pull request.

tools.context.repo

The owner and repo params for making API requests against a repository. This uses the GITHUB_REPOSITORY environment variable under the hood.

How to test your GitHub Actions

Similar to building CLIs, GitHub Actions usually works by running a file with node <file>; this means that writing a complete test suite can be tricky. Here's a pattern for writing tests using actions-toolkit, by mocking Toolkit.run:

index.js
const { Toolkit } = require('actions-toolkit')
Toolkit.run(async tools => {
  tools.log.success('Yay!')
})
index.test.js
const { Toolkit } = require('actions-toolkit')
describe('tests', () => {
  let action

  beforeAll(() => {
    // Mock `Toolkit.run` to redefine `action` when its called
    Toolkit.run = fn => { action = fn }
    // Require the index.js file, after we've mocked `Toolkit.run`
    require('./index.js')
  })

  it('logs successfully', async () => {
    // Create a fake instance of `Toolkit`
    const fakeTools = new Toolkit()
    // Mock the logger, or whatever else you need
    fakeTools.log.success = jest.fn()
    await action(fakeTools)
    expect(fakeTools.log.success).toHaveBeenCalled()
  })
})

You can then mock things by tweaking environment variables and redefining tools.context.payload. You can check out this repo's tests as an example.

Motivation

actions-toolkit is a wrapper around some fantastic open source libraries, and provides some helper methods for dealing with the GitHub Actions runtime. Actions all run in Docker containers, so this library aims to help you focus on your code and not the runtime. You can learn more about building Actions in Node.js to get started!

After building a GitHub Action in Node.js, it was clear to me that I was writing code that other actions will want to use. Reading files from the repository, making requests to the GitHub API, or running arbitrary executables on the project, etc.

So, I thought it'd be useful to build those out into a library to help you build actions in Node.js πŸŽ‰

FAQ

Aren't these just wrappers around existing functions?

Yep! I just didn't want to rewrite them for my next Action, so here we are.

**What's the difference between this and actions/toolkit?

This library was the inspiration for the official toolkit. Nowadays, it's an opinionated alternative. My goal for the library is to make building simple actions easy, while the official toolkit needs to support more complicated use-cases (like performance and scaling concerns).

You can’t perform that action at this time.