• Stars
    star
    128
  • Rank 281,044 (Top 6 %)
  • Language
    Ruby
  • License
    MIT License
  • Created almost 10 years ago
  • Updated almost 2 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Brave New World for Rails with more cohesion, less coupling and greater development speed for all

Aldous

The basic idea behind this gem is that Rails is missing a few key things that make our lives especially painful when our applications start getting bigger and more complex. With Aldous we attempt to address some of these shortcomings, namely:

  1. Bloated models that trample all over the single responsibility principle. God models that tend to appear despite our best intentions. Not to mention the complexity of testing objects that are so big and unwieldy.

Aldous attempts to address this, by introducing a light convention for creating service objects that conform to the same interface. See below, for the features that Aldous service objects have and how to use them in your app.

  1. Big and bloated controllers (despite our best intentions) that contain business logic and are hard to test. Not to mention the logic that is spread among the various before_actions requiring mental compilation to understand a controller action fully. Lastly the instance variables spread all over the controller that get inherited by the view templates as global variables.

Aldous addresses this by introducing the concept of a controller action object. The Rails controllers still exist, but they contain no logic in them. All the logic moves to the action classes with one action per class. Instance variables from these don't automatically get inherited by the view templates and the job of before_actions is taken over by precondition objects. See below for a more detailed overview of Aldous controller actions.

  1. Views that are not really views, but are instead view templates that inherit instance variables from controllers as globals and leave no good place for view-specific logic making us resort to hacky solutions like view helpers.

Aldous addresses this by introducing a concept of view objects which are actual ruby objects, these live alongside the standard Rails views (which we try to call templates from now on). These view objects provide a good place for view specific logic, which lets us remove much of the logic that Rails templates tend to accumulate. And being Ruby objects this logic can be tested like it should. See below for more details about Aldous view objects.

The key concepts that motivate Aldous are:

  • a greater number of light and small objects which are easy to understand and test
  • looser coupling between all parts of a larger application

No matter how big our applications get we want to maintain the rapid speed of feature development we usually only have at the start of a project. The two ideas above greatly facilitate this. As a side-effect we also achieve looser coupling with the web framework and more clearly defined business logic.

Installation

Add this line to your application's Gemfile:

gem 'aldous'

And then execute:

$ bundle

Or install it yourself as:

$ gem install aldous

Blog Posts About Aldous

It is difficult to explain all the motivations and thoughts and philosophy that have gone into this gem without bloating the Readme tremendously. So this Readme will remain focused on usage. Any extra stuff, such as how it all hangs together and why things are done a certain way will be delegated to blog posts which will be linked from here.

Usage

All the concepts in Aldous are designed to work well when used together, but we also don't want to be too prescriptive, so you can easily use just some of the concepts without using others. You can take advantage of the service objects without worrying about controller action or view objects or if you prefer to just use the controller action you can do that too. Here is an in-depth explanation for how to hook everything up one concept at a time (with recommendations for how to achieve the best results).

Service Objects

The first thing to do is to configure a folder for the services to live in, app/services is a good candidate. So go into your app/config/application.rb and add the following:

config.autoload_paths += %W(
  #{config.root}/app/services
)

config.eager_load_paths += %W(
  #{config.root}/app/services
)

The next thing to think about is how you will name your service objects. I prefer a *_service convention (create_user_service.rb), that way you can always tell you're looking at a service object regardless of where it is being used. Also remember it's a service object, so name it like you would a method instead of like you would a class (e.g. it's not UserCreatorService, it's CreateUserService).

A typical Aldous service object might look something like this:

class CreateUserService < Aldous::Service
  attr_reader :user_data_hash

  def initialize(user_data_hash)
    @user_data_hash = user_data_hash
  end

  def raisable_error
    MyApplication::Errors::UserError
  end

  def default_result_data
    {user: nil}
  end

  def perform
    user = User.new(user_data_hash)
    user.roles << Role.where(name: "account_holder").first

    if user.save
      Result::Success.new(user: user)
    else
      Result::Failure.new
    end
  end
end

A couple of things of note:

  • it inherits from Aldous::Service, this is important as it gives it some nice behaviour.
  • the raisable_error, default_result_data and perform methods are part of the public interface that is inherited from Aldous::Service

The only method you strictly have to override is perform, but I recommend you do all 3 for extra goodness that is explained below.

Here is how you would call it:

hash = {}
result = CreateUserService.perform(hash)
if result.success?
  # do success stuff
else #result.failure?
  # do failure stuff
end

or

hash = {}
begin
  result = CreateUserService.perform!(hash)
  if result.success?
    # do success stuff
  else #result.failure?
    # do failure stuff
  end
rescue MyApplication::Errors::UserError => e
end

or

hash = {}
service = CreateUserService.build(hash)
result = service.perform
if result.success?
  # do success stuff
else #result.failure?
  # do failure stuff
end

As you can see, you don't use the constructor to build these objects, use build or call perform directly (otherwise you'll be missing all the nice features).

An Aldous service object can be either error free or error raising, here is what that means. If you construct the service object correctly via the factory method and/or call perform. The method call is guaranteed to never raise an error. When implementing the perform method you have to remember to return Result::Success for any success cases and return Result::Failure for any failure cases. If the code you have written raises an error, it will be automatically caught and a Result::Failure will be returned. The idea is that a service either succeeds or fails and if an error occurs that's just a kind of failure.

However there are situations where you want the service to raise an error, for example if you're using it in a background job and an error means that the job will be retried (e.g. like sidekiq would do). For this purpose Aldous service objects automatically gain a perform! method. This will use the funtionality of your perform method, but if an error occurs it will be caught and then re-raised as the error type you specify in the raisable_error method. This way you always know that your services can return Result::Success, Result::Failure or raise the error type you specify depending on how you call them.

Service objects are immutable once they are constructed their state doens't change, but we may still want to return data to the user of the service. This is why Result::Success and Result::Failure are not just are not just semantic types they are also DTO (data transfer) objects. When you construct a Result::Success for example, it will automatically have reader methods defined on it with the names of the keys in default_result_data hash. An example to demonstrate:

class CreateUserService < Aldous::Service
  def default_result_data
    {user: nil}
  end

  def perform
    Result::Success.new
  end
end

result = CreateUserService.perform
result.user  # will be equal to nil

or

class CreateUserService < Aldous::Service
  def default_result_data
    {user: nil}
  end

  def perform
    Result::Success.new(user: user)
  end

  private

  def user
    User.new
  end
end

result = CreateUserService.perform
result.user  # will be a User instance

Hopefully you can see how this hangs together, the default_result_data method defines a hash of data with default values. This data will magically end up on the Result::Success and Result::Failure objects you construct in your perform method. If you pass a hash of data when you construct those object, that data will override the default values defined in the default_result_data method.

Controller Actions

The first thing to do is to configure a folder for the controller actions to live in, app/controller_actions is a good candidate. So go into your app/config/application.rb and add the following:

config.autoload_paths += %W(
  #{config.root}/app/controller_actions
)

config.eager_load_paths += %W(
  #{config.root}/app/controller_actions
)

Let's say we're creating a controller called TodosController, here is what it would look like now:

class TodosController < ApplicationController
  include Aldous::Controller

  controller_actions :index, :new, :create, :edit, :update, :destroy
end

After you've done this, the code for the index action should live in, app/controller_actions/todos_controller/index.rb - pretty intuitive.

A controller action might look like this:

class TodosController::Index < BaseAction
  def default_view_data
    super.merge({todos: todos})
  end

  def perform
    return view_builder.build(Home::ShowRedirect) unless current_user

    view_builder.build(Todos::IndexView)
  end

  private

  def todos
    Todo.where(user_id: current_user.id)
  end
end

The action above is also using Aldous view objects.

As you can see, the controller action lives under the namespace of the controller. I also recommend creating a BaseAction and inheriting from that for most of your actions. The BaseAction might look like this:

class BaseAction < ::Aldous::ControllerAction
  def default_view_data
    {
      current_user: current_user,
      current_ability: current_ability,
    }
  end

  def preconditions
    [Shared::EnsureUserNotDisabledPrecondition]
  end

  def default_error_handler(error)
    Defaults::ServerErrorView
  end

  def current_user
    @current_user ||= FindCurrentUserService.perform(session).user
  end

  def current_ability
    @current_ability ||= Ability.new(current_user)
  end
end

As you can see if inherits from Aldous::ControllerAction. The methods you should override are:

  • default_view_data - this hash of data will be available to all the aldous view objects
  • preconditions - this is used as a replacement for before_actions
  • default_error_handler - this view will be rendered if an unhandled error gets raised in the action code

Similar to services, for actions you implement the perform method (might as well keep things consistent). All action classes have the same constructor signature, they take a controller object and you have access to this controller object in your action classes. Also a few methods from the controller get exposed directly to your action (params, session, cookies, request, response) for convenience. You can expose others via configuration in an initializer, or you can grab them from the controller instance you have access to.

Controller action are automatically error free, so any error that gets raised in the perform method, automatically get caught and a view you specify in default_error_handler will get rendered to handle that error.

Lets look at a slightly bigger controller action:

class TodosController::Update < BaseAction
  def default_view_data
    super.merge({todo: todo})
  end

  def preconditions
    super.reject{|klass| klass == Shared::EnsureUserNotDisabledPrecondition}
  end

  def perform
    return view_builder.build(Home::ShowRedirect) unless current_user
    return view_builder.build(Defaults::BadRequestView, errors: [todo_params.error_message]) unless todo_params.fetch
    return view_builder.build(Todos::NotFoundView, todo_id: params[:id]) unless todo
    return view_builder.build(Defaults::ForbiddenView) unless current_ability.can?(:update, todo)

    if todo.update_attributes(todo_params.fetch)
      view_builder.build(Todos::IndexRedirect)
    else
      view_builder.build(Todos::EditView)
    end
  end

  private

  def todo
    @todo ||= Todo.where(id: params[:id]).first
  end

  def todo_params
    TodosController::TodoParams.build(params)
  end
end

Let's start with the perform method. As you can see it's fattish, but no fatter than regular controller methods if you take into account all the before_actions. The idea is to handle the special cases first and then perform the logic which will either succeed or fail. In our case we want to redirect to home if there is no current_user. The actual current_user method, comes from our BaseAction. If we can't fetch strong params we render a bad request view. If we couldn't find a todo we render a not found view. And if the current user isn't allowed to update the todo, we render a forbidden view. All of those are things we should think about and handle before we try to perform the logic of our update action. And all those things should be close to the logic of the action, so we can grok the full scope of the action easily without mental compilation. We then perform the logic and render or redirect based on the success or failure of that logic. As you might imagine, if the logic was more complex you would push it into a service object e.g.:

class SignUpsController::Create < BaseAction
  def perform
    return view_builder.build(Todos::IndexRedirect) if current_user
    return view_builder.build(Defaults::BadRequestView, errors: [user_params.error_message]) unless user_params.fetch

    if create_user_result.success?
      SignInService.perform!(session, create_user_result.user)
      view_builder.build(Todos::IndexRedirect)
    else
      view_builder.build(SignUps::NewView)
    end
  end

  private

  def create_user_result
    @create_user_result ||= CreateUserService.perform(user_params.fetch)
  end

  def user_params
    @user_params ||= ::SignUpsController::UserParams.build(params)
  end
end

Let's talk about default_view_data. We first define this method in our BaseAction we them augment it in our actual actions. Eventually that whole hash of data will get bundles up into a DTO object and be available in whichever view object we end up rendering. You always know which data your view object will have access to and you directly control it with pure ruby and very limited magic.

Let's talk about the preconditions method. We usually want to define it in our BaseAction and only override it in our actual actions if we want to get rid of a particular precondition, which should happen infrequently. This method defines an array of classes, this classes will get instantiated and executed in the order they are defined before the actual action is executed. In most respects a precondition class looks and behaves exactly like an action class. The precondition also has access to the action instance:

class Shared::EnsureUserNotDisabledPrecondition < BasePrecondition
  delegate :current_user, :current_ability, to: :action

  def perform
    if current_user && current_user.disabled && !current_ability.can?(:manage, :all)
      return view_builder.build(Defaults::ForbiddenView, errors: ['Your account has been disabled'])
    end
  end
end

I recommend having a base class for preconditions:

class BasePrecondition < ::Aldous::Controller::Action::Precondition
end

Preconditions should be applicable to all or most actions, and if you need to switch a precondition off for a particular action, you can do so using ruby collection methods (e.g. reject).

Aldous provides a small helper for bundling strong params logic into an object:

class SignUpsController::UserParams < Aldous::Params
  def permitted_params
    params.require(:user).permit(:email, :password)
  end

  def error_message
    'Missing param :user'
  end
end

::SignUpsController::UserParams.build(params)

When you have an instance of an Aldous::Params object, you just call fetch on it and it will either return the params hash you were after, or nil if something went wrong. You can put params object along side your actions, or if they are applicable across controllers, put it into app/controller_actions/shared just like you would with precondition objects.

View Objects

The first thing to do is to configure a folder for the view objects to live in, we already have app/views for our view templates, so lets just put our view objects right there next to the templates. So go into your app/config/application.rb and add the following:

config.autoload_paths += %W(
  #{config.root}/app/views
)

config.eager_load_paths += %W(
  #{config.root}/app/views
)

We have to understand that when we're talking about view objects, we're not necessarily talking about views as Rails would understand them. We're talking about any kind of output you would normally produce from a Rails controller (e.g. render view, redirect to location, send data, head etc.). In Aldous this concept is called - respondables and we have a different type of object to represent each one of these. There are Renderable object for standard view, Redirectable objects for redirects etc. The key thing is that a controller action always produces a respondable object and it's the responsibility of that object to do the right thing to produce the output that we expect. Example time, here is what a redirect object might look like:

class Home::ShowRedirect < Aldous::Respondable::Redirectable
  def location
    view_context.root_path
  end
end

As you can see it inherits from Aldous::Respondable::Redirectable and all you have to do is to provide the location method. Here is what a renderable object (for views) might look like:

class Home::ShowView < BaseView
  def template_data
    {
      template: 'home/show',
      locals: {}
    }
  end
end

with BaseView being:

class BaseView < ::Aldous::Respondable::Renderable
  def default_template_locals
    {
      current_user: current_user,
      header_view: header_view,
    }
  end

  def current_user
    view_data.current_user
  end

  private

  def header_view
    view_builder.build(Modules::HeaderView)
  end
end

As you can see a view object ultimately inherits from Aldous::Respondable::Renderable and the key method to override is template_data. This method needs to return a hash with template or partial that we want to render as well as the locals that we want to supply. If we have a BaseView which I recommend, then we can also override the default_template_locals method. When you do this, some Aldous magic will happen and all the things in that hash will be available in all the templates as locals which is very handy for data that's common across all or most templates. The key things here is that you control it.

You construct view objects either in your controller actions or in other view objects, to do this you use the view_builder object build method. This object is always available and you don't need to worry about providing it, but if you ever want to override it (e.g. in tests), you can inject it also. The only place where you may have to construct view objects directly is in tests. The constructor signature of a view object is:

def initialize(status, view_data, view_context, view_builder)
  @status = status
  @view_data = view_data
  @view_context = view_context
  @view_builder = view_builder
 end

The first variable is self explanatory, use it if you want to override the status for a particular view object. The second parameter is a DTO which contains all the data that comes from the default_view_data in your controller action. The last parameter is the view_context that you normally get from the controller. This is the context in which the view templates get executed in vanilla Rails.

You should only override the default_template_locals in a BaseView object. In the actual view object just provide the locals directly in the template_data method. Everything will get merged correctly by Aldous. You can also provide a default status for any view by overriding the default_status method.

Lets look at a more complex view object and the correspoding templates:

class Todos::IndexView < BaseView
  def template_data
    {
      template: 'todos/index',
      locals: {
        todo_views: todo_views,
      }
    }
  end

  private

  def todos
    view_data.todos
  end

  def todo_views
    todos.map do |todo|
      todo_view(todo)
    end
  end

  def todo_view(todo)
    view_builder.build(Todos::IndexView::TodoView, todo: todo)
  end
end

our BaseView is:

class BaseView < ::Aldous::Respondable::Renderable
  def default_template_locals
    {
      current_user: current_user,
      header_view: header_view,
    }
  end

  def current_user
    view_data.current_user
  end

  private

  def header_view
    view_builder.build(Modules::HeaderView)
  end
end

so our template index.html.slim has access to current_user, header_view, and todo_views as locals. Let's have a look at the template:

- provide :title, "View Todos"

= render header_view.template

h1 Your Todos

= link_to 'Create New Todo', new_todo_path
= button_to 'Delete Completed Todos', all_completed_todos_path, method: :delete

- todo_views.each do |todo_view|
  = render todo_view.template

You can see how we render partials inside a template:

= render header_view.template

If we needed to pass in extra locals for the partial (e.g. a form object), we could do:

= render header_view.template(f: f)

In our case header_view will be available to all templates, which includes the layout, so we can pull it out of this template and add it to the layout instead.

The idea here is for templates to have very little logic besides boilerplate stuff like the - todo_views.each above. There should also be very few (or none at all) demeter violations in the template, as this stuff can be pushed up into the view objects now and handled more robustly, tested etc.

Configuring Aldous

You can create an initializer for Aldous it would look like this:

Aldous.configuration do |aldous|
  aldous.logger = Rails.logger
  aldous.error_reporter = MyApplication::ErrorReporter
  aldous.controller_methods_exposed_to_action += [:current_user]
end

The most interesting config option initially is the ability to supply the logger so that you can see relevant things in your log files and also pick up on any issues with Aldous itself. The error_reporter is an object that responds to report:

module Aldous
  class DummyErrorReporter
    class << self
      def report(e, data = {})
        nil
      end
    end
  end
end

This can be used for things like sending errors to an error collection service or whatever else you want to do when exceptions occur.

You can also expose more methods from the Rails controller to the controller action by using controller_methods_exposed_to_action.

Why is it called Aldous?

When we initially started using some of the things that would later become Aldous, we called it Brave New World (BNW). So when a lot of that functionality migrated into a gem we called it Aldous cause Brave New World.

Examples

There is a basic example Rails app that lives in this repo. Run it up, and have a look at the code. Also one of the goals of Aldous was to make its own source code easy to understand read and modify for other people, to facilitate understading and cooperation. So, feel free to read the code itself to figure out how things hang together, start with the Aldous objects that you would inherit from e.g. Aldous::ControllerAction, Aldous::Service.

Contributing

  1. Fork it
  2. Create your feature branch (git checkout -b my-new-feature)
  3. Commit your changes (git commit -am 'Add some feature')
  4. Push to the branch (git push origin my-new-feature)
  5. Create new Pull Request

More Repositories

1

double_entry

A double-entry accounting system for Ruby applications.
Ruby
417
star
2

envato-wordpress-toolkit

Wordpress toolkit for authors of items available from the Envato Market sites.
PHP
373
star
3

wp-envato-market

WordPress Theme & Plugin management for the Envato Market.
PHP
340
star
4

zxcvbn-ruby

Ruby port of Dropbox's zxcvbn javascript lib
Ruby
314
star
5

stack_master

The missing CloudFormation tool
Ruby
290
star
6

envato-theme-check

The WordPress Theme Check plugin for Envato
PHP
205
star
7

pagerduty

📟 A Ruby gem for talking to the Pagerduty Events API
Ruby
99
star
8

event_sourcery

A library for building event sourced applications in Ruby
Ruby
84
star
9

safe_shell

Safely execute shell commands and get their output.
Ruby
74
star
10

react-breakpoints

Respond to changes in a DOM element's size. With React Breakpoints, element queries are no longer "web design's unicorn" 🦄
TypeScript
72
star
11

envato-wordpress-toolkit-library

WordPress Toolkit Library for Envato Marketplace hosted items
PHP
71
star
12

lumberg

Ruby library for the WHM & cPanel API; It's not a half day or anything like that.
Ruby
56
star
13

rails_4_session_flash_backport

Rails 4 Session Flash Backport
Ruby
52
star
14

ami-spec

Acceptance testing your AMIs
Ruby
50
star
15

heroku-deploy

Complete zero downtime deploys for Rails applications on Heroku
Ruby
41
star
16

rack-ecg

Health check page as Rack middleware
Ruby
39
star
17

splinter

Chop your specs in half with Splinter, a collection of helpers to submit forms with RSpec and Capybara.
Ruby
30
star
18

extensions-sketch-plugin

Envato Elements - Sketch Plugin (Beta)
JavaScript
26
star
19

unwrappr

🍫 bundle update PRs: Automated. Annotated. Announced daily.
Ruby
26
star
20

exiv2

A simple wrapper around exiv2
Ruby
25
star
21

jwt_signed_request

Request signing and verification made easy
Ruby
24
star
22

iodized

Iodine rich feature toggling. Lack of iodine can lead to development delays, and stunted growth.
Elixir
21
star
23

react-ab-experiment

A/B Experiment React Component
JavaScript
18
star
24

event_sourcery_todo_app

Example event_sourcery app
Ruby
15
star
25

awsraw

Minimal AWS client
Ruby
14
star
26

moo_moo

Implements OpenSRS XML API
Ruby
14
star
27

create-react-icon

Convert one or multiple svg files into React component
JavaScript
13
star
28

aspect_ratio

Image aspect ratio calculation utility
Ruby
13
star
29

event_sourcery-postgres

Postgres event store implementation for EventSourcery
Ruby
13
star
30

github-pull-request-buildkite-plugin

Open Github pull requests via your Buildkite builds
Shell
13
star
31

cookie-consent

Some helper functions to deal with cookie-consent
TypeScript
11
star
32

knuckle_cluster

Gem to interrogate and connect to ECS clusters, Spot fleets and Autoscaling groups
Ruby
10
star
33

cloudformation_rspec

Test your CloudFormation templates
Ruby
10
star
34

aws-account-concierge

Opensource AWS account management tool
Ruby
9
star
35

flv

FLV Parser
Ruby
9
star
36

ejsonkms

Integrates EJSON with AWS KMS
Go
9
star
37

subvalid

Subjective validation for Plain Old Ruby Objects
Ruby
8
star
38

forked

Forked manages long running worker processes
Ruby
8
star
39

react-resize-observer-hook

Observe multiple DOM elements with a single ResizeObserver.
TypeScript
8
star
40

heroku-buildpack-libsodium

libsodium/rbnacl buildpack for Heroku
Shell
8
star
41

foundation-design-system-tokens

Design Tokens for the Foundation Design System
JavaScript
8
star
42

guide

Document your application with a living component library and styleguide
Ruby
8
star
43

studio-client-uploader

JavaScript
8
star
44

outatime

Choose versioned S3 files from a point in time.
Ruby
7
star
45

wp-image-size-limit

Wordpress plugin for limiting the image file size
PHP
7
star
46

aws-s3-sync-buildkite-plugin

Shell
7
star
47

rails_session_key_rotator

Graceful secret key rotation for the signed cookie store in Rails.
Ruby
6
star
48

babushka-deps

Ruby
6
star
49

lambda-deploy-buildkite-plugin

A Buildkite plugin to deploy AWS Lambda function code
Shell
6
star
50

double_entry-reporting

Reporting on Double Entry accounts and transfers
Ruby
6
star
51

medusa

The most beautiful parallel build system, evar.
JavaScript
5
star
52

hamburglar

Hamburglar helps you prevent fraudulent orders
Ruby
5
star
53

fake_aws

Ruby
5
star
54

market-api-gateway-js

JavaScript SDK for marketplace api-gateway
JavaScript
5
star
55

ejson_wrapper

Combines EJSON with AWS KMS
Ruby
5
star
56

bundle-update-buildkite-plugin

Update Ruby gem dependencies in a Buildkite build!
Shell
5
star
57

stop-the-line-buildkite-plugin

Stop Buildkite pipelines based on build meta-data values.
Shell
5
star
58

encapsulate_as_money

Surprise me ;)
Ruby
5
star
59

event_sourcery_generators

An opinionated CLI tool for building event-sourced Ruby services with EventSourcery
Ruby
4
star
60

imageresize

Image resizing for Envato Sites author tools
Go
4
star
61

react-ab-experiment-example

React AB Experiment example app
JavaScript
4
star
62

packer-ami-copy-buildkite-plugin

A buildkite plugin which simplifies the process of building and copying an AMI with packer
Shell
4
star
63

create-datadog-event-buildkite-plugin

An experimental Buildkite plugin which creates Datadog events.
Shell
4
star
64

heroku-container-deploy-buildkite-plugin

Deploy pre-built docker images to Heroku using Heroku Container Registry
Shell
4
star
65

barricade

better locking
Ruby
4
star
66

google-analytics-for-wordpress

WordPress.org Plugin Mirror
PHP
4
star
67

build-failed-notify-slack-buildkite-plugin

@'s the creator of the failed build via a mapping file
Shell
4
star
68

browser-privacy-checker

Handy React 🎣 module to check whether can use user cookie and similar technologies from browser
TypeScript
4
star
69

unwrappr-demo

Demonstrating the power of Unwrappr
Ruby
3
star
70

marketplace-stats-as-a-service

A little micro service to cache and expose Envato Marketplace stats(total items and total users)
JavaScript
3
star
71

docker-cache-buildkite-plugin

Build a multi-stage compatible cache image and store it in S3
Shell
3
star
72

puppet-hubot

Puppet
3
star
73

sprockets-resilience

Stop Rails from killing your site when assets aren't precompiled
Ruby
3
star
74

rack_fake_s3

Ruby
3
star
75

mandrill-merge

We all hate spam, but sometimes you need to get an important message out to a large group of people. Mandrill is great at allowing you to define templates, sending emails out, and keeping track of sends and responses. But how do you get the data from your database over to Mandrill quickly and easily? That's where MandrillMerge comes in...
Ruby
3
star
76

chisel-scripts

Chisel scripts created to debug any issues
Lua
2
star
77

siren-call

A gem for interacting with Siren style REST APIs
Ruby
2
star
78

backbone.proxy-view

A proxy view is an instance of Backbone.View that doesn't render any of its own HTML
CoffeeScript
2
star
79

iodized_ruby_client

Ruby
2
star
80

eventory

A new take on event sourcing in ruby
Ruby
2
star
81

docker-size-annotation-buildkite-plugin

Annotates the build with a docker image size
Shell
2
star
82

thrift-erlang

copy + paste of Apache thrift erlang lib. Mirrored here so it can be used for rebar/mix etc
Erlang
2
star
83

fancy-confluence

CSS
2
star
84

spot_build

Agents are transient, Jobs are forever
Ruby
2
star
85

logstash-filter-dynamo_enrich

Enrich Logstash events with data from DynamoDB
Ruby
2
star
86

batch_retry

Batch send data with retries and exponential backoff
Python
2
star
87

rack_request_ip_strategies

Replacement for Rack::Request#ip
Ruby
2
star
88

heartbeat

Heartbeat plugin that exposes an API to monitor internal services
Ruby
2
star
89

wordpress-https

A fork of an older version of wordpress-https plugin that disables some super-slow functionality.
PHP
2
star
90

ejson2env-buildkite-plugin

A Buildkite plugin for exporting environment variables stored in ejson files using ejson2env
Shell
2
star
91

backbone.autocomplete-view

A template driven autocomplete view for Backbone
CoffeeScript
2
star
92

swarm

Ruby
2
star
93

cloudformation-output-buildkite-plugin

Collects cloudformation output from an AWS stack and add them as environment variables
Shell
2
star
94

studio-asset-service-uploader

An uploader for the studio asset service
JavaScript
1
star
95

puppet_module_provider

Ruby
1
star
96

babel-plugin-postcss-cssmodules-transform

Babel Plugin Transform PostCSS and extract static CSS
JavaScript
1
star
97

studio-asset-service-client-js

A javascript client for the studio asset service
JavaScript
1
star
98

reproduce-fluentd-cascading-buffer

Reproduce cascading failure when Fluentd's internal memory buffer becomes full
Shell
1
star
99

virtellus

A smaller, faster, less powerful Virtus
Ruby
1
star
100

auth0-logs-to-s3

JavaScript
1
star