Code Triage

What is Triage?

When patients come into the emergency room, they don't see a doctor immediately, they go to a triage nurse. The nurse knows enough about medical problems to properly assign that person to the doctor that can help them the quickest. Since the doctors are the most limited resource, triage nurses help to assign them as effectively as possible. Triage in open source means looking at open issues and adding useful information for maintainers. While you might not maintain a repository, you can help those who do by diagnosing issues, reviewing pull requests.

Why Triage?

Triage is an important part of open source. It can be difficult to keep up with bugs and assess the validity of contributions. Code introduced to fix one problem can easily generate more problems than it solves, so it's important for maintainers to look closely at bug reports and code contributions. Unfortunately as the size of a project grows, the demands placed on the maintainers grow. This means they are forced to choose between spending enormous amounts of time reviewing each GitHub issue, only skimming over issues, or worse, ignoring issues.

As a non-maintainer, you can help an open source project by triaging issues. When issues come in, they are assigned out to triage. If you get assigned an issue, you should look closely at it, and provide feedback to make a maintainer's life easier. If there is a bug reported, try to reproduce it and then give the results in the issue comments. If code is included in the issue, review the code, see if it makes sense. Code submitted should have a clear use case, be in the same style as the project, and not introduce failures into the test system. If the code is good, leave a comment explaining why you believe it is good. +1's are great, but leave no context and don't help maintainers much. If you don't like an issue you need to explain why as well. Either way leave a comment with the issue.

How Does it Work?

You sign up to follow a repository, once a day you'll be emailed with an open issue from that repository, and instructions on how to triage the issue in a helpful way. In the background we use Sidekiq to grab issues from GitHub's API, we then use another background task to assign users who subscribe to a repository one issue each day.

Run Code Triage

Dependencies

Make sure you have bundler, then install the dependencies:

$ gem install bundler
$ bundle install

Database

• Create a database (default is PostgreSQL)
• Copy database.yml and if you need, setup your own database credentials
• Run migrations

$ cp config/database.example.yml config/database.yml
$ bin/rake db:create
$ bin/rake db:migrate

Install Redis

Code Triage requires Redis for background processing.

Homebrew

If you're on OS X, Homebrew is the simplest way to install Redis:

$ brew install redis phantomjs memcached
$ redis-server

You now have Redis running on 6379.

Other

See the Download page on Redis.io for steps to install on other systems: http://redis.io/download

Environment

If you want your users to sign up with Github, create a GitHub Client Application. The urls you are asked to provide will be something like this:

• URL: http://localhost:3000
• Callback URL: http://localhost:3000/users/auth/github/callback

Then add the credentials to your .env file:

$ echo GITHUB_APP_ID=foo >> .env
$ echo GITHUB_APP_SECRET=bar >> .env
$ echo PORT=3000 >> .env

Running the app

Start your app using Foreman

$ foreman start
08:19:22 web.1    | started with pid 6347
08:19:22 worker.1 | started with pid 6348

Code Triage should now be running at http://localhost:3000

Tests

$ bin/rake db:create RAILS_ENV=test
$ bin/rake db:schema:load RAILS_ENV=test

You may need a github API token to run tests locally. You can get this by spinning your local server, clicking the "sign in" button and going through the OAuth flow.

Once you've done this spin down your server and run this:

$ bin/rails c
> `echo GITHUB_API_KEY=#{User.last.token} >> .env`

Make sure it shows up in your .env:

> puts File.read(".env")

Now you should be able to run tests

$ bin/rake test

Writing tests

If you need to mock out some github requests please use VCR. Put anything that may create an external request inside of a vcr block:

VCR.use_cassette('my_vcr_cassette_name_here') do
  # ... code goes here
end

Make sure to name your cassette something unique. The first time you run tests you'll need to set a record mode. This will make a real-life request to github using your GITHUB_API_KEY you specified in the .env and record the result. The next time you run your tests it should use your "cassette" instead of actually hitting github. All secrets including your GITHUB_API_KEY are filtered out, so you can safely commit the resultant. When running on travis the VCR cassettes are used to eliminate/minimize actual calls to Github.

The benefit of using VCR over stubbing/mocking methods is that we could swap out implementations if we wanted.

Make sure to remove any record modes from your VCR cassette before committing.

Modifying Tests

If you need to modify a test locally and the VCR cassette doesn't have the correct information, you can run with new episodes enabled.

Make sure to remove any record modes from your VCR cassette before committing.

Flow

• A user subscribes to a repo
• Consume API: Once a day, find all the repos that haven't been updated in 24 hours, produce issue subscription.
• Issue Assigning [repo]: Find all users subscribed to that repo that haven't been assigned an issue in 24 hours, pick a random issue that the user is not a part of and send them an email.

Contact

Richard Schneeman @Schneems

Licensed under MIT License. Copyright (c) 2012 Schneems. See LICENSE.txt for further details.