tl;dr How can I get a single Sinatra app to start up very differently on different servers via customizations to config.ru?
Background
I have a single web application written using Sinatra that's run on different servers. Currently the codebase for these servers is forked because there are some non-trivial differences in the way (discrete) parts of them work. For example:
one server authenticates users via an intranet LDAP server, while another server uses a simpler local database table lookup.
one server uses an external cron job to periodically update some statistics, while another (Windows-based) server uses an internal sleepy Thread.
one server stores certain metadata in a local table, while another server pulls the metadata from an external Wiki via screen scraping (!).
…and so on.
I'd like to get these code bases completely shared (single Git repo). I envision that each server would have one slightly-differing configuration file that causes the app to be started up differently.
Abandoned Solutions
I could change the behavior of the app based on environment variables. As there are a not-tiny number of variations in behavior, I'd rather not hide the settings in environment variables.
I could create my own "server-settings.rb" file that is unique to each machine, require it in my app.rb, and then change the configuration there. However, this seems to possibly be re-inventing the wheel. I already have a file named config.ru for each server. Shouldn't I be using this?
The Current Code
My config.ru for the app currently is simply:
require ::File.join( ::File.dirname(__FILE__), 'app' )
run MyApp.new
And the app.rb that it requires is, in essence:
require 'sinatra'
require_relative 'helpers/login' # customized for LDAP lookup on this server
class MyApp < Sinatra::Application
use Rack::Session::Cookie, key:'foo.bar', path:'/', secret:'ohnoes'
set :protection, except: [:path_traversal, :session_hijacking]
configure :production do
# run various code that depends on server settings, e.g.
Snapshotter.start # there is no cron on this machine, so we do it ourselves
end
configure :development do
# run various code that depends on server settings
end
end
The Question
I'd like to make config.ru live up to its name, and have it look something like this:
require ::File.join( ::File.dirname(__FILE__), 'app' )
run MyApp.new( auth: :ldap, snapshot:false, metadata: :remote_wiki, … )
How can I modify my application to change its configuration behavior based on settings supplied via config.ru? Or is this an abuse of config.ru, trying to use it for totally the wrong thing?
As soon as I started reading the question the first answer to pop into my head was "environment variable" but you scotched that straight away :)
I'll go with a mixture of one of your coulds and the desired outcome code, as it's how I structure things…
Because I want to be able to test my applications more easily, I take most of the Ruby out of the config.ru and into a separate config.rb file and leave config.ru to be a bootstrap file. So my standard skel is:
config.ru
# encoding: UTF-8
require 'rubygems'
require 'bundler'
Bundler.setup
root = File.expand_path File.dirname(__FILE__)
require File.join( root , "./app/config.rb" )
# everything was moved into a separate module/file to make it easier to set up tests
map "/" do
run APP_NAME.app
end
app/config.rb
# encoding: utf-8
require_relative File.expand_path(File.join File.dirname(__FILE__), "../lib/ext/warn.rb")
require_relative "./init.rb" # config
require_relative "./main.rb" # routes and helpers
require 'encrypted_cookie'
# standard cookie settings
COOKIE_SETTINGS = {
:key => 'usr',
:path => "/",
:expire_after => 86400, # In seconds, 1 day
:secret => ENV["LLAVE"],
:httponly => true
}
module APP_NAME # overall name of the app
require 'rack/ssl' # force SSL
require 'rack/csrf'
if ENV["RACK_ENV"] == "development"
require 'pry'
require 'pry-nav'
end
# from http://devcenter.heroku.com/articles/ruby#logging
$stdout.sync = true
ONE_MONTH = 60 * 60 * 24 * 30
def self.app
Rack::Builder.app do
cookie_settings = COOKIE_SETTINGS
# more security if in production
cookie_settings.merge!( :secure => true ) if ENV["RACK_ENV"] == "production"
# AES encryption of cookies
use Rack::Session::EncryptedCookie, cookie_settings
if ENV["RACK_ENV"] == "production"
use Rack::SSL, :hsts => {:expires => ONE_MONTH}
end
# to stop XSS
use Rack::Csrf, :raise => true unless ENV["RACK_ENV"] == "test"
run App # the main Sinatra app
end
end # self.app
end # APP_NAME
The initial reason I did this was making it easy to run the app in specs:
shared_context "All routes" do
include Rack::Test::Methods
let(:app){ APP_NAME.app }
end
but it makes sense to me to keep this code with the rest of the application code, so to speak, as I can bundle things together, run other apps etc. I've used this to conditionally load different examples into the specs in a few projects (it helps cut down on duplicated effort and check the examples really work), so I don't see why you couldn't use it to conditionally load configurations.
This way you get to choose to use a conditional in the config.ru as to which config.rb file you would use, or use an env var in the config.rb as to which definiton of self.app to use , or pass in an options hash to self.app…
With your set up I'd rename the APP_NAME module to MyApp, and the Sinatra class to App (because quite often I'll have an website that runs a front end and an API, so the Sinatra classes get named by their function (App, API etc) and wrapped in a module named after the site) and end up with:
config.ru
map "/" do
run MyApp.app( auth: :ldap, snapshot:false, metadata: :remote_wiki )
end
config.rb
def self.app( opts={} )
opts = DEFAULT_OPTIONS.merge opts
# …
run App
end
It'll be interesting to see how other people tackle this.
Related
I use free heroku instance to run my Dashing project. In result, it looses the value passed previously, when my instance sleeps. I was recommended to use Redis to keep history. I tryed to follow the instruction given here. In result I got the following config.ru (as part of my dashing project):
require 'dashing'
require 'redis-objects'
require 'yaml'
configure do
set :auth_token, 'my-token'
set :default_dashboard, 'def' # https://github.com/Shopify/dashing/wiki/How-To:-Change-the-default-dashboard
helpers do
def protected!
# Put any authentication code you want in here.
# This method is run before accessing any resource.
end
end
end
def redis?
ENV.has_key? 'REDISTOGO_URL'
end
if redis?
redis_uri = URI.parse(ENV['REDISTOGO_URL'])
Redis.current = Redis.new(:host => redis_uri.host,
:port => redis_uri.port,
:password => redis_uri.password)
set :history, Redis::HashKey.new('dashing-history')
elsif File.exists?(settings.history_file)
set history: YAML.load_file(settings.history_file)
else
set history: {}
end
map Sinatra::Application.assets_prefix do
run Sinatra::Application.sprockets
end
run Sinatra::Application
and the following Gemfile:
source 'https://rubygems.org'
gem 'dashing'
gem 'redis-objects'
## Remove this if you don't need a twitter widget.
gem 'twitter', '>= 5.9.0'
But it didn't help. What I did incorrectly?
I also tried to use this tutorial. But it was giving me an error at line redis_uri = URI.parse(ENV["REDISTOGO_URL"]) (something like wrong url is given).
The problem was that the app requires the add-on Redis To Go
If Redis To Go is configured, REDISTOGO_URL is added to environment variables, it will work
For more information on how to setup Redis To Go, read the heroku article
Adding Redis to an application provides benefits, you may be using RedisToGo to power simple Resque or Sidekiq jobs, or using the raw power of Redis 2.6 Lua Scripting to do some crazy fast operations. Redis can be used a database, but it’s often used as a complementary datastore. With over 140 commands, the possibilities are endless.
This is a non-rails app, just a simple ruby script that uses rake etc. to automate some things.
My folder layout is this:
/scripts/Rakefile
/scripts/config/config.yml
/scripts/tasks/*.rake (various rake files with namespaces to organize them)
/scripts/lib/settings.rb
Now I want to create a Settings class that will load the config yaml file, and then expose properties/methods for the contents of the yaml file.
The yaml file has separate sections for development and production.
development:
scripts_path: '/dev/mygit/app1/scripts/'
production:
scripts_path: '/var/lib/app1/scripts/'
My rakefile so far looks like:
$LOAD_PATH.unshift File.expand_path('..', __FILE__)
#imports
require 'fileutils'
require 'rubygems'
require 'active_record'
require 'yaml'
require 'logger'
require 'ar/models'
require 'lib/app1'
env = ENV['ENV'] || 'development'
config = YAML::load(File.open('config/config.yml'))[env]
Dir.glob('tasks/*.rake').each { |r| import r }
I need help with the Settings.rb file, is this right?
module App1
class Settings
def initialize(config_path, env)
config = YAML.load(File.open(config_path))
end
def scripts_path
end
end
end
How can I pass in the env, and then read the correct value from the config for each method like scripts_path etc?
Now suppose each *.rake file needs to reference my Settings.rb file somehow (to get the config related information). How should I do this? Since my settings needs the path of the config.yml file, do I have to do this in each rake file?
Update
Sorry, this isn't a Rails app, just some ruby scripts.
I would do it quite simple.
You don't need a complex solution.
require 'ostruct'
require 'yaml'
MY_ENV = ENV['ENV'] || 'development'
CONFIG = OpenStruct.new(YAML.load_file("config/config.yml")[MY_ENV])
Stick this at the top of your rakefile
and CONFIG will be available in all rake tasks.
Just call CONFIG.scripts_path
Inside my applications I do something of this sort.
# config/application.yml
development:
some_variable: a string
production:
some_variable: a different string
Then in application.rb I load it up.
# config/application.rb
module MyApp
def self.config
#config ||= OpenStruct.new(YAML.load_file("config/application.yml")[Rails.env.to_s])
end
class Application < Rails::Application
...
In this case, anywhere the environment is loaded I can say
MyApp.config.some_variable
To get access to this inside a rake task, I just need to include environment
task :something => :environment do
MyApp.config.some_variable
# do something with it
end
I am trying to setup rack-offline in Sinatra, but I am having trouble setting it up. In rails it is prettty easy, but have no found any examples in Sinatra...
Basically, in your config.ru, map /application.manifest to Rack::Offline. (If you're not familiar with using config.ru with your Sinatra application, check out this part of Sinatra docs.) Here's an example, which caches all the files under directory public:
require 'your-app'
require 'rack/offline'
map "/application.manifest" do
offline = Rack::Offline.new :cache => true, :root => "public" do
# Cache all files under the directory public
Dir[File.join(settings.public, "**/*")].each do |file|
cache file.sub(File.join(settings.public, ""), "")
end
# All other files should be downloaded
network '/'
end
run offline
end
map "/" do
run Sinatra::Application
end
Remember to set manifest="/application.manifest" in your html tag and you should be good to go. You should take a look at rack-offline's README for more documentation and explanation of how it works.
I'm building out a medium-sized application using Sinatra and all was well when I had a single app.rb file and I followed Aslak's guidance up on Github:
https://github.com/cucumber/cucumber/wiki/Sinatra
As the app grew a bit larger and the app.rb file started to bulge, I refactored out a lot of of the bits into "middleware" style modules using Sinatra::Base, mapping things using a rack-up file (config.ru) etc.
The app works nicely - but my specs blew up as there was no more app.rb file for webrat to run against (as defined in the link above).
I've tried to find examples on how to work this - and I think I'm just not used to the internal guts of Cuke just yet as I can't find a single way to have it cover all the apps. I tried just pointing to "config.ru" instead of app.rb - but that doesn't work.
What I ended up doing - which is completely hackish - is to have a separate app.rb file in my support directory, which has all the requires stuff so I can at least test the model stuff. I can also specify routes in there - but that's not at all what I want to do.
So - the question is: how can I get Cucumber to properly work with the modular app approach?
Update to include dealing with multiple Sinatra apps
Require the file where your app comes together and change
def app
Sinatra::Application
end
to
def app
Rack::Builder.new do
map '/a' { run MyAppA }
map '/b' { run MyAppB }
end
end
and just test the app proper.
eg, if you define middleware in your config.ru that you want to test, maybe move loading those into your app's definition.
Thanks to Mr. BaroqueBobcat - the answer now, of course, seems so damn obvious :). Here's the env.rb (/features/support/env.rb):
require 'sinatra'
require 'test/unit'
require 'spec/expectations'
require 'rack/test'
require 'webrat'
require 'app1'
require 'app2'
require 'app3'
Webrat.configure do |config|
config.mode = :rack
end
class MyWorld
require 'test/unit'
set :environment, :test
include Rack::Test::Methods
include Webrat::Methods
include Webrat::Matchers
Webrat::Methods.delegate_to_session :response_code, :response_body, :response
def app
Rack::Builder.new do
map '/' do
run App1 #important - this is the class name
end
map '/app1' do
run App2
end
map '/app2' do
run App3
end
end
end
end
World do
MyWorld.new
end
https://gist.github.com/28d510d9fc25710192bc
def app
eval "Rack::Builder.new {( " + File.read(File.dirname(__FILE__) + '/../config.ru') + "\n )}"
end
I've set up Rack::Reload according to this thread
# config.ru
require 'rubygems'
require 'sinatra'
set :environment, :development
require 'app'
run Sinatra::Application
# app.rb
class Sinatra::Reloader < Rack::Reloader
def safe_load(file, mtime, stderr = $stderr)
if file == Sinatra::Application.app_file
::Sinatra::Application.reset!
stderr.puts "#{self.class}: reseting routes"
end
super
end
end
configure(:development) { use Sinatra::Reloader }
get '/' do
'foo'
end
Running with thin via thin start -R config.ru, but it only reloads newly added routes. When I change already existing route, it still runs the old code.
When I add new route, it correctly reloads it, so it is accessible, but it doesn't reload anything else.
For example, if I changed routes to
get '/' do
'bar'
end
get '/foo' do
'baz'
end
Than / would still serve foo, even though it has changed, but /foo would correctly reload and serve baz.
Is this normal behavior, or am I missing something? I'd expect whole source file to be reloaded. The only way around I can think of right now is restarting whole webserver when filesystem changes.
I'm running on Windows Vista x64, so I can't use shotgun because of fork().
You could try sinatra-reloader, which is known to work well on Windows (also, it's faster than shotgun).
This works:
# config.ru
require 'rubygems'
require 'app'
set :environment, :development
run Sinatra::Application
# app.rb
require 'sinatra'
class Sinatra::Reloader < Rack::Reloader
def safe_load(file, mtime, stderr = $stderr)
if file == File.expand_path(Sinatra::Application.app_file)
::Sinatra::Application.reset!
stderr.puts "#{self.class}: reseting routes"
end
super
end
end
configure(:development) { use Sinatra::Reloader }
get '/' do
'foo'
end
It matters from where you have the require statement. But I find the following solution more elegant and robust:
# config.ru
require 'rubygems'
require 'sinatra'
require 'rack/reloader'
require 'app'
set :environment, :development
use Rack::Reloader, 0 if development?
run Sinatra::Application
# app.rb
Sinatra::Application.reset!
get '/' do
'foo'
end
Does Shotgun not work on Windows?
From the README:
Shotgun
This is an automatic reloading version of the rackup command that's shipped with
Rack. It can be used as an alternative to the complex reloading logic provided
by web frameworks or in environments that don't support application reloading.
The shotgun command starts one of Rack's supported servers (e.g., mongrel, thin,
webrick) and listens for requests but does not load any part of the actual
application. Each time a request is received, it forks, loads the application in
the child process, processes the request, and exits the child process. The
result is clean, application-wide reloading of all source files and templates on
each request.
You can also try using Trinidad a JRuby Rack container based on Tomcat. In my experience it does change reloading by default without having to modify your source files. Bloody fast too. Obviously no good if you are using native libraries, but if you are deploying on Windows you are probably used to adopting a pure-ruby approach.
Its syntax is just as simple as the thin approach:
jruby -S trinidad -r config.ru
There is no Java specific yak shaving (i.e. creating web.xml or WARing up your Ruby app) and the gem is simple to install.