mirror of https://github.com/sinatra/sinatra synced 2023-03-27 23:18:01 -04:00

Closes #1786

2023-01-21 09:06:18 +09:00

70 KiB

Raw Blame History

Sinatra

Sinatra is a DSL for quickly creating web applications in Ruby with minimal effort:

# myapp.rb
require 'sinatra'

get '/' do
  'Hello world!'
end

Install the gem:

gem install sinatra
gem install puma # or any other server

And run with:

ruby myapp.rb

View at: http://localhost:4567

The code you changed will not take effect until you restart the server. Please restart the server every time you change or use a code reloader like rerun or rack-unreloader.

It is recommended to also run gem install puma, which Sinatra will pick up if available.

Sinatra

Routes

In Sinatra, a route is an HTTP method paired with a URL-matching pattern. Each route is associated with a block:

get '/' do
  .. show something ..
end

post '/' do
  .. create something ..
end

put '/' do
  .. replace something ..
end

patch '/' do
  .. modify something ..
end

delete '/' do
  .. annihilate something ..
end

options '/' do
  .. appease something ..
end

link '/' do
  .. affiliate something ..
end

unlink '/' do
  .. separate something ..
end

Routes are matched in the order they are defined. The first route that matches the request is invoked.

Routes with trailing slashes are different from the ones without:

get '/foo' do
  # Does not match "GET /foo/"
end

Route patterns may include named parameters, accessible via the params hash:

get '/hello/:name' do
  # matches "GET /hello/foo" and "GET /hello/bar"
  # params['name'] is 'foo' or 'bar'
  "Hello #{params['name']}!"
end

You can also access named parameters via block parameters:

get '/hello/:name' do |n|
  # matches "GET /hello/foo" and "GET /hello/bar"
  # params['name'] is 'foo' or 'bar'
  # n stores params['name']
  "Hello #{n}!"
end

Route patterns may also include splat (or wildcard) parameters, accessible via the params['splat'] array:

get '/say/*/to/*' do
  # matches /say/hello/to/world
  params['splat'] # => ["hello", "world"]
end

get '/download/*.*' do
  # matches /download/path/to/file.xml
  params['splat'] # => ["path/to/file", "xml"]
end

Or with block parameters:

get '/download/*.*' do |path, ext|
  [path, ext] # => ["path/to/file", "xml"]
end

Route matching with Regular Expressions:

get /\/hello\/([\w]+)/ do
  "Hello, #{params['captures'].first}!"
end

Or with a block parameter:

get %r{/hello/([\w]+)} do |c|
  # Matches "GET /meta/hello/world", "GET /hello/world/1234" etc.
  "Hello, #{c}!"
end

Route patterns may have optional parameters:

get '/posts/:format?' do
  # matches "GET /posts/" and any extension "GET /posts/json", "GET /posts/xml" etc
end

Routes may also utilize query parameters:

get '/posts' do
  # matches "GET /posts?title=foo&author=bar"
  title = params['title']
  author = params['author']
  # uses title and author variables; query is optional to the /posts route
end

By the way, unless you disable the path traversal attack protection (see below), the request path might be modified before matching against your routes.

You may customize the Mustermann options used for a given route by passing in a :mustermann_opts hash:

get '\A/posts\z', :mustermann_opts => { :type => :regexp, :check_anchors => false } do
  # matches /posts exactly, with explicit anchoring
  "If you match an anchored pattern clap your hands!"
end

It looks like a condition, but it isn't one! These options will be merged into the global :mustermann_opts hash described below.

Conditions

Routes may include a variety of matching conditions, such as the user agent:

get '/foo', :agent => /Songbird (\d\.\d)[\d\/]*?/ do
  "You're using Songbird version #{params['agent'][0]}"
end

get '/foo' do
  # Matches non-songbird browsers
end

Other available conditions are host_name and provides:

get '/', :host_name => /^admin\./ do
  "Admin Area, Access denied!"
end

get '/', :provides => 'html' do
  haml :index
end

get '/', :provides => ['rss', 'atom', 'xml'] do
  builder :feed
end

provides searches the request's Accept header.

You can easily define your own conditions:

set(:probability) { |value| condition { rand <= value } }

get '/win_a_car', :probability => 0.1 do
  "You won!"
end

get '/win_a_car' do
  "Sorry, you lost."
end

For a condition that takes multiple values use a splat:

set(:auth) do |*roles|   # <- notice the splat here
  condition do
    unless logged_in? && roles.any? {|role| current_user.in_role? role }
      redirect "/login/", 303
    end
  end
end

get "/my/account/", :auth => [:user, :admin] do
  "Your Account Details"
end

get "/only/admin/", :auth => :admin do
  "Only admins are allowed here!"
end

Return Values

The return value of a route block determines at least the response body passed on to the HTTP client or at least the next middleware in the Rack stack. Most commonly, this is a string, as in the above examples. But other values are also accepted.

You can return an object that would either be a valid Rack response, Rack body object or HTTP status code:

An Array with three elements: [status (Integer), headers (Hash), response body (responds to #each)]
An Array with two elements: [status (Integer), response body (responds to #each)]
An object that responds to #each and passes nothing but strings to the given block
A Integer representing the status code

That way we can, for instance, easily implement a streaming example:

class Stream
  def each
    100.times { |i| yield "#{i}\n" }
  end
end

get('/') { Stream.new }

You can also use the stream helper method (described below) to reduce boilerplate and embed the streaming logic in the route.

Custom Route Matchers

As shown above, Sinatra ships with built-in support for using String patterns and regular expressions as route matches. However, it does not stop there. You can easily define your own matchers:

class AllButPattern
  def initialize(except)
    @except = except
  end

  def to_pattern(options)
    return self
  end

  def params(route)
    return {} unless @except === route
  end
end

def all_but(pattern)
  AllButPattern.new(pattern)
end

get all_but("/index") do
  # ...
end

Note that the above example might be over-engineered, as it can also be expressed as:

get /.*/ do
  pass if request.path_info == "/index"
  # ...
end

Static Files

Static files are served from the ./public directory. You can specify a different location by setting the :public_folder option:

set :public_folder, __dir__ + '/static'

Note that the public directory name is not included in the URL. A file ./public/css/style.css is made available as http://example.com/css/style.css.

Use the :static_cache_control setting (see below) to add Cache-Control header info.

Views / Templates

Each template language is exposed via its own rendering method. These methods simply return a string:

get '/' do
  erb :index
end

This renders views/index.erb.

Instead of a template name, you can also just pass in the template content directly:

get '/' do
  code = "<%= Time.now %>"
  erb code
end

Templates take a second argument, the options hash:

get '/' do
  erb :index, :layout => :post
end

This will render views/index.erb embedded in the views/post.erb (default is views/layout.erb, if it exists).

Any options not understood by Sinatra will be passed on to the template engine:

get '/' do
  haml :index, :format => :html5
end

You can also set options per template language in general:

set :haml, :format => :html5

get '/' do
  haml :index
end

Options passed to the render method override options set via set.

Available Options:

locals: List of locals passed to the document. Handy with partials. Example: erb "<%= foo %>", :locals => {:foo => "bar"}
default_encoding: String encoding to use if uncertain. Defaults to settings.default_encoding.
views: Views folder to load templates from. Defaults to settings.views.
layout: Whether to use a layout (true or false). If it's a Symbol, specifies what template to use. Example: erb :index, :layout => !request.xhr?
content_type: Content-Type the template produces. Default depends on template language.
scope: Scope to render template under. Defaults to the application instance. If you change this, instance variables and helper methods will not be available.
layout_engine: Template engine to use for rendering the layout. Useful for languages that do not support layouts otherwise. Defaults to the engine used for the template. Example: set :rdoc, :layout_engine => :erb
layout_options: Special options only used for rendering the layout. Example: set :rdoc, :layout_options => { :views => 'views/layouts' }

Templates are assumed to be located directly under the ./views directory. To use a different views directory:

set :views, settings.root + '/templates'

One important thing to remember is that you always have to reference templates with symbols, even if they're in a subdirectory (in this case, use: :'subdir/template' or 'subdir/template'.to_sym). You must use a symbol because otherwise rendering methods will render any strings passed to them directly.

Literal Templates

get '/' do
  haml '%div.title Hello World'
end

Renders the template string. You can optionally specify :path and :line for a clearer backtrace if there is a filesystem path or line associated with that string:

get '/' do
  haml '%div.title Hello World', :path => 'examples/file.haml', :line => 3
end

Available Template Languages

Some languages have multiple implementations. To specify what implementation to use (and to be thread-safe), you should simply require it first:

require 'rdiscount'
get('/') { markdown :index }

Haml Templates

Dependency	haml
File Extension	`.haml`
Example	`haml :index, :format => :html5`

Erb Templates

Dependency	erubi or erb (included in Ruby)
File Extensions	`.erb`, `.rhtml` or `.erubi` (Erubi only)
Example	`erb :index`

Builder Templates

Dependency	builder
File Extension	`.builder`
Example	`builder { \|xml\| xml.em "hi" }`

It also takes a block for inline templates (see example).

Nokogiri Templates

Dependency	nokogiri
File Extension	`.nokogiri`
Example	`nokogiri { \|xml\| xml.em "hi" }`

It also takes a block for inline templates (see example).

Liquid Templates

Dependency	liquid
File Extension	`.liquid`
Example	`liquid :index, :locals => { :key => 'value' }`

Since you cannot call Ruby methods (except for yield) from a Liquid template, you almost always want to pass locals to it.

Markdown Templates

Dependency	Anyone of: RDiscount, RedCarpet, kramdown, commonmarker pandoc
File Extensions	`.markdown`, `.mkd` and `.md`
Example	`markdown :index, :layout_engine => :erb`

It is not possible to call methods from Markdown, nor to pass locals to it. You therefore will usually use it in combination with another rendering engine:

erb :overview, :locals => { :text => markdown(:introduction) }

Note that you may also call the markdown method from within other templates:

%h1 Hello From Haml!
%p= markdown(:greetings)

Since you cannot call Ruby from Markdown, you cannot use layouts written in Markdown. However, it is possible to use another rendering engine for the template than for the layout by passing the :layout_engine option.

RDoc Templates

Dependency	RDoc
File Extension	`.rdoc`
Example	`rdoc :README, :layout_engine => :erb`

It is not possible to call methods from RDoc, nor to pass locals to it. You therefore will usually use it in combination with another rendering engine:

erb :overview, :locals => { :text => rdoc(:introduction) }

Note that you may also call the rdoc method from within other templates:

%h1 Hello From Haml!
%p= rdoc(:greetings)

Since you cannot call Ruby from RDoc, you cannot use layouts written in RDoc. However, it is possible to use another rendering engine for the template than for the layout by passing the :layout_engine option.

AsciiDoc Templates

Dependency	Asciidoctor
File Extension	`.asciidoc`, `.adoc` and `.ad`
Example	`asciidoc :README, :layout_engine => :erb`

Since you cannot call Ruby methods directly from an AsciiDoc template, you almost always want to pass locals to it.

Markaby Templates

Dependency	Markaby
File Extension	`.mab`
Example	`markaby { h1 "Welcome!" }`

It also takes a block for inline templates (see example).

RABL Templates

Dependency	Rabl
File Extension	`.rabl`
Example	`rabl :index`

Slim Templates

Dependency	Slim Lang
File Extension	`.slim`
Example	`slim :index`

Yajl Templates

Dependency	yajl-ruby
File Extension	`.yajl`
Example	`yajl :index, :locals => { :key => 'qux' }, :callback => 'present', :variable => 'resource'`

The template source is evaluated as a Ruby string, and the resulting json variable is converted using #to_json:

json = { :foo => 'bar' }
json[:baz] = key

The :callback and :variable options can be used to decorate the rendered object:

var resource = {"foo":"bar","baz":"qux"};
present(resource);

Accessing Variables in Templates

Templates are evaluated within the same context as route handlers. Instance variables set in route handlers are directly accessible by templates:

get '/:id' do
  @foo = Foo.find(params['id'])
  haml '%h1= @foo.name'
end

Or, specify an explicit Hash of local variables:

get '/:id' do
  foo = Foo.find(params['id'])
  haml '%h1= bar.name', :locals => { :bar => foo }
end

This is typically used when rendering templates as partials from within other templates.

Templates with `yield` and nested layouts

A layout is usually just a template that calls yield. Such a template can be used either through the :template option as described above, or it can be rendered with a block as follows:

erb :post, :layout => false do
  erb :index
end

This code is mostly equivalent to erb :index, :layout => :post.

Passing blocks to rendering methods is most useful for creating nested layouts:

erb :main_layout, :layout => false do
  erb :admin_layout do
    erb :user
  end
end

This can also be done in fewer lines of code with:

erb :admin_layout, :layout => :main_layout do
  erb :user
end

Currently, the following rendering methods accept a block: erb, haml, liquid, slim . Also, the general render method accepts a block.

Inline Templates

Templates may be defined at the end of the source file:

require 'sinatra'

get '/' do
  haml :index
end

__END__

@@ layout
%html
  != yield

@@ index
%div.title Hello world.

NOTE: Inline templates defined in the source file that requires Sinatra are automatically loaded. Call enable :inline_templates explicitly if you have inline templates in other source files.

Named Templates

Templates may also be defined using the top-level template method:

template :layout do
  "%html\n  =yield\n"
end

template :index do
  '%div.title Hello World!'
end

get '/' do
  haml :index
end

If a template named "layout" exists, it will be used each time a template is rendered. You can individually disable layouts by passing :layout => false or disable them by default via set :haml, :layout => false:

get '/' do
  haml :index, :layout => !request.xhr?
end

Associating File Extensions

To associate a file extension with a template engine, use Tilt.register. For instance, if you like to use the file extension tt for Haml templates, you can do the following:

Tilt.register Tilt[:haml], :tt

Adding Your Own Template Engine

First, register your engine with Tilt, then create a rendering method:

Tilt.register MyAwesomeTemplateEngine, :myat

helpers do
  def myat(*args) render(:myat, *args) end
end

get '/' do
  myat :index
end

Renders ./views/index.myat. Learn more about Tilt.

Using Custom Logic for Template Lookup

To implement your own template lookup mechanism you can write your own #find_template method:

configure do
  set :views, [ './views/a', './views/b' ]
end

def find_template(views, name, engine, &block)
  Array(views).each do |v|
    super(v, name, engine, &block)
  end
end

Filters

Before filters are evaluated before each request within the same context as the routes will be and can modify the request and response. Instance variables set in filters are accessible by routes and templates:

before do
  @note = 'Hi!'
  request.path_info = '/foo/bar/baz'
end

get '/foo/*' do
  @note #=> 'Hi!'
  params['splat'] #=> 'bar/baz'
end

After filters are evaluated after each request within the same context as the routes will be and can also modify the request and response. Instance variables set in before filters and routes are accessible by after filters:

after do
  puts response.status
end

Note: Unless you use the body method rather than just returning a String from the routes, the body will not yet be available in the after filter, since it is generated later on.

Filters optionally take a pattern, causing them to be evaluated only if the request path matches that pattern:

before '/protected/*' do
  authenticate!
end

after '/create/:slug' do |slug|
  session[:last_slug] = slug
end

Like routes, filters also take conditions:

before :agent => /Songbird/ do
  # ...
end

after '/blog/*', :host_name => 'example.com' do
  # ...
end

Helpers

Use the top-level helpers method to define helper methods for use in route handlers and templates:

helpers do
  def bar(name)
    "#{name}bar"
  end
end

get '/:name' do
  bar(params['name'])
end

Alternatively, helper methods can be separately defined in a module:

module FooUtils
  def foo(name) "#{name}foo" end
end

module BarUtils
  def bar(name) "#{name}bar" end
end

helpers FooUtils, BarUtils

The effect is the same as including the modules in the application class.

Using Sessions

A session is used to keep state during requests. If activated, you have one session hash per user session:

enable :sessions

get '/' do
  "value = " << session[:value].inspect
end

get '/:value' do
  session['value'] = params['value']
end

Session Secret Security

To improve security, the session data in the cookie is signed with a session secret using HMAC-SHA1. This session secret should optimally be a cryptographically secure random value of an appropriate length which for HMAC-SHA1 is greater than or equal to 64 bytes (512 bits, 128 hex characters). You would be advised not to use a secret that is less than 32 bytes of randomness (256 bits, 64 hex characters). It is therefore very important that you don't just make the secret up, but instead use a secure random number generator to create it. Humans are extremely bad at generating random values.

By default, a 32 byte secure random session secret is generated for you by Sinatra, but it will change with every restart of your application. If you have multiple instances of your application, and you let Sinatra generate the key, each instance would then have a different session key which is probably not what you want.

For better security and usability it's recommended that you generate a secure random secret and store it in an environment variable on each host running your application so that all of your application instances will share the same secret. You should periodically rotate this session secret to a new value. Here are some examples of how you might create a 64-byte secret and set it:

Session Secret Generation

$ ruby -e "require 'securerandom'; puts SecureRandom.hex(64)"
99ae8af...snip...ec0f262ac

Session Secret Generation (Bonus Points)

Use the sysrandom gem to use the system RNG facilities to generate random values instead of userspace OpenSSL which MRI Ruby currently defaults to:

$ gem install sysrandom
Building native extensions.  This could take a while...
Successfully installed sysrandom-1.x
1 gem installed

$ ruby -e "require 'sysrandom/securerandom'; puts SecureRandom.hex(64)"
99ae8af...snip...ec0f262ac

Session Secret Environment Variable

Set a SESSION_SECRET environment variable for Sinatra to the value you generated. Make this value persistent across reboots of your host. Since the method for doing this will vary across systems this is for illustrative purposes only:

# echo "export SESSION_SECRET=99ae8af...snip...ec0f262ac" >> ~/.bashrc

Session Secret App Config

Set up your app config to fail-safe to a secure random secret if the SESSION_SECRET environment variable is not available.

For bonus points use the sysrandom gem here as well:

require 'securerandom'
# -or- require 'sysrandom/securerandom'
set :session_secret, ENV.fetch('SESSION_SECRET') { SecureRandom.hex(64) }

Session Config

If you want to configure it further, you may also store a hash with options in the sessions setting:

set :sessions, :domain => 'foo.com'

To share your session across other apps on subdomains of foo.com, prefix the domain with a . like this instead:

set :sessions, :domain => '.foo.com'

Choosing Your Own Session Middleware

Note that enable :sessions actually stores all data in a cookie. This might not always be what you want (storing lots of data will increase your traffic, for instance). You can use any Rack session middleware in order to do so, one of the following methods can be used:

enable :sessions
set :session_store, Rack::Session::Pool

Or to set up sessions with a hash of options:

set :sessions, :expire_after => 2592000
set :session_store, Rack::Session::Pool

Another option is to not call enable :sessions, but instead pull in your middleware of choice as you would any other middleware.

It is important to note that when using this method, session based protection will not be enabled by default.

The Rack middleware to do that will also need to be added:

use Rack::Session::Pool, :expire_after => 2592000
use Rack::Protection::RemoteToken
use Rack::Protection::SessionHijacking

See 'Configuring attack protection' for more information.

Halting

To immediately stop a request within a filter or route use:

halt

You can also specify the status when halting:

halt 410

Or the body:

halt 'this will be the body'

Or both:

halt 401, 'go away!'

With headers:

halt 402, {'Content-Type' => 'text/plain'}, 'revenge'

It is of course possible to combine a template with halt:

halt erb(:error)

Passing

A route can punt processing to the next matching route using pass:

get '/guess/:who' do
  pass unless params['who'] == 'Frank'
  'You got me!'
end

get '/guess/*' do
  'You missed!'
end

The route block is immediately exited and control continues with the next matching route. If no matching route is found, a 404 is returned.

Triggering Another Route

Sometimes pass is not what you want, instead, you would like to get the result of calling another route. Simply use call to achieve this:

get '/foo' do
  status, headers, body = call env.merge("PATH_INFO" => '/bar')
  [status, headers, body.map(&:upcase)]
end

get '/bar' do
  "bar"
end

Note that in the example above, you would ease testing and increase performance by simply moving "bar" into a helper used by both /foo and /bar.

If you want the request to be sent to the same application instance rather than a duplicate, use call! instead of call.

Check out the Rack specification if you want to learn more about call.

Setting Body, Status Code, and Headers

It is possible and recommended to set the status code and response body with the return value of the route block. However, in some scenarios, you might want to set the body at an arbitrary point in the execution flow. You can do so with the body helper method. If you do so, you can use that method from thereon to access the body:

get '/foo' do
  body "bar"
end

after do
  puts body
end

It is also possible to pass a block to body, which will be executed by the Rack handler (this can be used to implement streaming, see "Return Values").

Similar to the body, you can also set the status code and headers:

get '/foo' do
  status 418
  headers \
    "Allow"   => "BREW, POST, GET, PROPFIND, WHEN",
    "Refresh" => "Refresh: 20; https://ietf.org/rfc/rfc2324.txt"
  body "I'm a teapot!"
end

Like body, headers and status with no arguments can be used to access their current values.

Streaming Responses

Sometimes you want to start sending out data while still generating parts of the response body. In extreme examples, you want to keep sending data until the client closes the connection. You can use the stream helper to avoid creating your own wrapper:

get '/' do
  stream do |out|
    out << "It's gonna be legen -\n"
    sleep 0.5
    out << " (wait for it) \n"
    sleep 1
    out << "- dary!\n"
  end
end

This allows you to implement streaming APIs, Server Sent Events, and can be used as the basis for WebSockets. It can also be used to increase throughput if some but not all content depends on a slow resource.

Note that the streaming behavior, especially the number of concurrent requests, highly depends on the webserver used to serve the application. Some servers might not even support streaming at all. If the server does not support streaming, the body will be sent all at once after the block passed to stream finishes executing. Streaming does not work at all with Shotgun.

If the optional parameter is set to keep_open, it will not call close on the stream object, allowing you to close it at any later point in the execution flow. This only works on evented servers, like Rainbows. Other servers will still close the stream:

# config.ru
require 'sinatra/base'

class App < Sinatra::Base
  connections = []

  get '/subscribe', provides: 'text/event-stream'  do
    # register a client's interest in server events
    stream(:keep_open) do |out|
      connections << out
      # purge dead connections
      connections.reject!(&:closed?)
    end
  end

  post '/' do
    connections.each do |out|
      # notify client that a new message has arrived
      out << "data: #{params[:msg]}\n\n"

      # indicate client to connect again
      out.close
    end

    204 # response without entity body
  end
end

run App

# rainbows.conf
Rainbows! do
  use :EventMachine
end

Run:

rainbows -c rainbows.conf

It's also possible for the client to close the connection when trying to write to the socket. Because of this, it's recommended to check out.closed? before trying to write.

Logging

In the request scope, the logger helper exposes a Logger instance:

get '/' do
  logger.info "loading data"
  # ...
end

This logger will automatically take your Rack handler's logging settings into account. If logging is disabled, this method will return a dummy object, so you do not have to worry about it in your routes and filters.

Note that logging is only enabled for Sinatra::Application by default, so if you inherit from Sinatra::Base, you probably want to enable it yourself:

class MyApp < Sinatra::Base
  configure :production, :development do
    enable :logging
  end
end

To avoid any logging middleware to be set up, set the logging option to nil. However, keep in mind that logger will in that case return nil. A common use case is when you want to set your own logger. Sinatra will use whatever it will find in env['rack.logger'].

Mime Types

When using send_file or static files you may have mime types Sinatra doesn't understand. Use mime_type to register them by file extension:

configure do
  mime_type :foo, 'text/foo'
end

You can also use it with the content_type helper:

get '/' do
  content_type :foo
  "foo foo foo"
end

Generating URLs

For generating URLs you should use the url helper method, for instance, in Haml:

%a{:href => url('/foo')} foo

It takes reverse proxies and Rack routers into account - if present.

This method is also aliased to to (see below for an example).

Browser Redirect

You can trigger a browser redirect with the redirect helper method:

get '/foo' do
  redirect to('/bar')
end

Any additional parameters are handled like arguments passed to halt:

redirect to('/bar'), 303
redirect 'http://www.google.com/', 'wrong place, buddy'

You can also easily redirect back to the page the user came from with redirect back:

get '/foo' do
  "<a href='/bar'>do something</a>"
end

get '/bar' do
  do_something
  redirect back
end

To pass arguments with a redirect, either add them to the query:

redirect to('/bar?sum=42')

Or use a session:

enable :sessions

get '/foo' do
  session[:secret] = 'foo'
  redirect to('/bar')
end

get '/bar' do
  session[:secret]
end

Cache Control

Setting your headers correctly is the foundation for proper HTTP caching.

You can easily set the Cache-Control header like this:

get '/' do
  cache_control :public
  "cache it!"
end

Pro tip: Set up caching in a before filter:

before do
  cache_control :public, :must_revalidate, :max_age => 60
end

If you are using the expires helper to set the corresponding header, Cache-Control will be set automatically for you:

before do
  expires 500, :public, :must_revalidate
end

To properly use caches, you should consider using etag or last_modified. It is recommended to call those helpers before doing any heavy lifting, as they will immediately flush a response if the client already has the current version in its cache:

get "/article/:id" do
  @article = Article.find params['id']
  last_modified @article.updated_at
  etag @article.sha1
  erb :article
end

It is also possible to use a weak ETag:

etag @article.sha1, :weak

These helpers will not do any caching for you, but rather feed the necessary information to your cache. If you are looking for a quick reverse-proxy caching solution, try rack-cache:

require "rack/cache"
require "sinatra"

use Rack::Cache

get '/' do
  cache_control :public, :max_age => 36000
  sleep 5
  "hello"
end

Use the :static_cache_control setting (see below) to add Cache-Control header info to static files.

According to RFC 2616, your application should behave differently if the If-Match or If-None-Match header is set to *, depending on whether the resource requested is already in existence. Sinatra assumes resources for safe (like get) and idempotent (like put) requests are already in existence, whereas other resources (for instance post requests) are treated as new resources. You can change this behavior by passing in a :new_resource option:

get '/create' do
  etag '', :new_resource => true
  Article.create
  erb :new_article
end

If you still want to use a weak ETag, pass in a :kind option:

etag '', :new_resource => true, :kind => :weak

Sending Files

To return the contents of a file as the response, you can use the send_file helper method:

get '/' do
  send_file 'foo.png'
end

It also takes options:

send_file 'foo.png', :type => :jpg

The options are:

filename: File name to be used in the response, defaults to the real file name.
last_modified: Value for Last-Modified header, defaults to the file's mtime.
type: Value for Content-Type header, guessed from the file extension if missing.
disposition: Value for Content-Disposition header, possible values: nil (default), :attachment and :inline
length: Value for Content-Length header, defaults to file size.
status: Status code to be sent. Useful when sending a static file as an error page. If supported by the Rack handler, other means than streaming from the Ruby process will be used. If you use this helper method, Sinatra will automatically handle range requests.

Accessing the Request Object

The incoming request object can be accessed from request level (filter, routes, error handlers) through the request method:

# app running on http://example.com/example
get '/foo' do
  t = %w[text/css text/html application/javascript]
  request.accept              # ['text/html', '*/*']
  request.accept? 'text/xml'  # true
  request.preferred_type(t)   # 'text/html'
  request.body                # request body sent by the client (see below)
  request.scheme              # "http"
  request.script_name         # "/example"
  request.path_info           # "/foo"
  request.port                # 80
  request.request_method      # "GET"
  request.query_string        # ""
  request.content_length      # length of request.body
  request.media_type          # media type of request.body
  request.host                # "example.com"
  request.get?                # true (similar methods for other verbs)
  request.form_data?          # false
  request["some_param"]       # value of some_param parameter. [] is a shortcut to the params hash.
  request.referrer            # the referrer of the client or '/'
  request.user_agent          # user agent (used by :agent condition)
  request.cookies             # hash of browser cookies
  request.xhr?                # is this an ajax request?
  request.url                 # "http://example.com/example/foo"
  request.path                # "/example/foo"
  request.ip                  # client IP address
  request.secure?             # false (would be true over ssl)
  request.forwarded?          # true (if running behind a reverse proxy)
  request.env                 # raw env hash handed in by Rack
end

Some options, like script_name or path_info, can also be written:

before { request.path_info = "/" }

get "/" do
  "all requests end up here"
end

The request.body is an IO or StringIO object:

post "/api" do
  request.body.rewind  # in case someone already read it
  data = JSON.parse request.body.read
  "Hello #{data['name']}!"
end

Attachments

You can use the attachment helper to tell the browser the response should be stored on disk rather than displayed in the browser:

get '/' do
  attachment
  "store it!"
end

You can also pass it a file name:

get '/' do
  attachment "info.txt"
  "store it!"
end

Dealing with Date and Time

Sinatra offers a time_for helper method that generates a Time object from the given value. It is also able to convert DateTime, Date and similar classes:

get '/' do
  pass if Time.now > time_for('Dec 23, 2016')
  "still time"
end

This method is used internally by expires, last_modified and akin. You can therefore easily extend the behavior of those methods by overriding time_for in your application:

helpers do
  def time_for(value)
    case value
    when :yesterday then Time.now - 24*60*60
    when :tomorrow  then Time.now + 24*60*60
    else super
    end
  end
end

get '/' do
  last_modified :yesterday
  expires :tomorrow
  "hello"
end

Looking Up Template Files

The find_template helper is used to find template files for rendering:

find_template settings.views, 'foo', Tilt[:haml] do |file|
  puts "could be #{file}"
end

This is not really useful. But it is useful that you can actually override this method to hook in your own lookup mechanism. For instance, if you want to be able to use more than one view directory:

set :views, ['views', 'templates']

helpers do
  def find_template(views, name, engine, &block)
    Array(views).each { |v| super(v, name, engine, &block) }
  end
end

Another example would be using different directories for different engines:

set :views, :haml => 'templates', :default => 'views'

helpers do
  def find_template(views, name, engine, &block)
    _, folder = views.detect { |k,v| engine == Tilt[k] }
    folder ||= views[:default]
    super(folder, name, engine, &block)
  end
end

You can also easily wrap this up in an extension and share it with others!

Note that find_template does not check if the file really exists but rather calls the given block for all possible paths. This is not a performance issue, since render will use break as soon as a file is found. Also, template locations (and content) will be cached if you are not running in development mode. You should keep that in mind if you write a really crazy method.

Configuration

Run once, at startup, in any environment:

configure do
  # setting one option
  set :option, 'value'

  # setting multiple options
  set :a => 1, :b => 2

  # same as `set :option, true`
  enable :option

  # same as `set :option, false`
  disable :option

  # you can also have dynamic settings with blocks
  set(:css_dir) { File.join(views, 'css') }
end

Run only when the environment (APP_ENV environment variable) is set to :production:

configure :production do
  ...
end

Run when the environment is set to either :production or :test:

configure :production, :test do
  ...
end

You can access those options via settings:

configure do
  set :foo, 'bar'
end

get '/' do
  settings.foo? # => true
  settings.foo  # => 'bar'
  ...
end

Configuring attack protection

Sinatra is using Rack::Protection to defend your application against common, opportunistic attacks. You can easily disable this behavior (which will open up your application to tons of common vulnerabilities):

disable :protection

To skip a single defense layer, set protection to an options hash:

set :protection, :except => :path_traversal

You can also hand in an array in order to disable a list of protections:

set :protection, :except => [:path_traversal, :session_hijacking]

By default, Sinatra will only set up session based protection if :sessions have been enabled. See 'Using Sessions'. Sometimes you may want to set up sessions "outside" of the Sinatra app, such as in the config.ru or with a separate Rack::Builder instance. In that case, you can still set up session based protection by passing the :session option:

set :protection, :session => true

Available Settings

absolute_redirects: If disabled, Sinatra will allow relative redirects, however, Sinatra will no longer conform with RFC 2616 (HTTP 1.1), which only allows absolute redirects.; Enable if your app is running behind a reverse proxy that has not been set up properly. Note that the url helper will still produce absolute URLs, unless you pass in false as the second parameter.; Disabled by default.
add_charset: Mime types the content_type helper will automatically add the charset info to. You should add to it rather than overriding this option: settings.add_charset << "application/foobar"
app_file: Path to the main application file, used to detect project root, views and public folder and inline templates.
bind: IP address to bind to (default: 0.0.0.0 or localhost if your `environment` is set to development). Only used for built-in server.
default_content_type: Content-Type to assume if unknown (defaults to "text/html"). Set to nil to not set a default Content-Type on every response; when configured so, you must set the Content-Type manually when emitting content or the user-agent will have to sniff it (or, if nosniff is enabled in Rack::Protection::XSSHeader, assume application/octet-stream).
default_encoding: Encoding to assume if unknown (defaults to "utf-8").
dump_errors: Display errors in the log. Enabled by default unless environment is "test".
environment: Current environment. Defaults to ENV['APP_ENV'], or "development" if not available.
logging: Use the logger.
lock: Places a lock around every request, only running processing on request per Ruby process concurrently.; Enabled if your app is not thread-safe. Disabled by default.
method_override: Use _method magic to allow put/delete forms in browsers that don't support it.
mustermann_opts: A default hash of options to pass to Mustermann.new when compiling routing paths.
port: Port to listen on. Only used for built-in server.
prefixed_redirects: Whether or not to insert request.script_name into redirects if no absolute path is given. That way redirect '/foo' would behave like redirect to('/foo'). Disabled by default.
protection: Whether or not to enable web attack protections. See protection section above.
public_dir: Alias for public_folder. See below.
public_folder: Path to the folder public files are served from. Only used if static file serving is enabled (see static setting below). Inferred from app_file setting if not set.
quiet: Disables logs generated by Sinatra's start and stop commands. false by default.
reload_templates: Whether or not to reload templates between requests. Enabled in development mode.
root: Path to project root folder. Inferred from app_file setting if not set.
raise_errors: Raise exceptions (will stop application). Enabled by default when environment is set to "test", disabled otherwise.
run: If enabled, Sinatra will handle starting the web server. Do not enable if using rackup or other means.
running: Is the built-in server running now? Do not change this setting!
server: Server or list of servers to use for built-in server. Order indicates priority, default depends on Ruby implementation.
server_settings: If you are using a WEBrick web server, presumably for your development environment, you can pass a hash of options to server_settings, such as SSLEnable or SSLVerifyClient. However, web servers such as Puma do not support this, so you can set server_settings by defining it as a method when you call configure.
sessions: Enable cookie-based sessions support using Rack::Session::Cookie. See 'Using Sessions' section for more information.
session_store: The Rack session middleware used. Defaults to Rack::Session::Cookie. See 'Using Sessions' section for more information.
show_exceptions: Show a stack trace in the browser when an exception happens. Enabled by default when environment is set to "development", disabled otherwise.; Can also be set to :after_handler to trigger app-specified error handling before showing a stack trace in the browser.
static: Whether Sinatra should handle serving static files.; Disable when using a server able to do this on its own.; Disabling will boost performance.; Enabled by default in classic style, disabled for modular apps.
static_cache_control: When Sinatra is serving static files, set this to add Cache-Control headers to the responses. Uses the cache_control helper. Disabled by default.; Use an explicit array when setting multiple values: set :static_cache_control, [:public, :max_age => 300]
threaded: If set to true, will tell server to use EventMachine.defer for processing the request.
traps: Whether Sinatra should handle system signals.
views: Path to the views folder. Inferred from app_file setting if not set.
x_cascade: Whether or not to set the X-Cascade header if no route matches. Defaults to true.

Environments

There are three predefined environments: "development", "production" and "test". Environments can be set through the APP_ENV environment variable. The default value is "development". In the "development" environment all templates are reloaded between requests, and special not_found and error handlers display stack traces in your browser. In the "production" and "test" environments, templates are cached by default.

To run different environments, set the APP_ENV environment variable:

APP_ENV=production ruby my_app.rb

You can use predefined methods: development?, test? and production? to check the current environment setting:

get '/' do
  if settings.development?
    "development!"
  else
    "not development!"
  end
end

Error Handling

Error handlers run within the same context as routes and before filters, which means you get all the goodies it has to offer, like haml, erb, halt, etc.

Not Found

When a Sinatra::NotFound exception is raised, or the response's status code is 404, the not_found handler is invoked:

not_found do
  'This is nowhere to be found.'
end

Error

The error handler is invoked any time an exception is raised from a route block or a filter. But note in development it will only run if you set the show exceptions option to :after_handler:

set :show_exceptions, :after_handler

The exception object can be obtained from the sinatra.error Rack variable:

error do
  'Sorry there was a nasty error - ' + env['sinatra.error'].message
end

Custom errors:

error MyCustomError do
  'So what happened was...' + env['sinatra.error'].message
end

Then, if this happens:

get '/' do
  raise MyCustomError, 'something bad'
end

You get this:

So what happened was... something bad

Alternatively, you can install an error handler for a status code:

error 403 do
  'Access forbidden'
end

get '/secret' do
  403
end

Or a range:

error 400..510 do
  'Boom'
end

Sinatra installs special not_found and error handlers when running under the development environment to display nice stack traces and additional debugging information in your browser.

Rack Middleware

Sinatra rides on Rack, a minimal standard interface for Ruby web frameworks. One of Rack's most interesting capabilities for application developers is support for "middleware" -- components that sit between the server and your application monitoring and/or manipulating the HTTP request/response to provide various types of common functionality.

Sinatra makes building Rack middleware pipelines a cinch via a top-level use method:

require 'sinatra'
require 'my_custom_middleware'

use Rack::Lint
use MyCustomMiddleware

get '/hello' do
  'Hello World'
end

The semantics of use are identical to those defined for the Rack::Builder DSL (most frequently used from rackup files). For example, the use method accepts multiple/variable args as well as blocks:

use Rack::Auth::Basic do |username, password|
  username == 'admin' && password == 'secret'
end

Rack is distributed with a variety of standard middleware for logging, debugging, URL routing, authentication, and session handling. Sinatra uses many of these components automatically based on configuration so you typically don't have to use them explicitly.

You can find useful middleware in rack, rack-contrib, or in the Rack wiki.

Testing

Sinatra tests can be written using any Rack-based testing library or framework. Rack::Test is recommended:

require 'my_sinatra_app'
require 'minitest/autorun'
require 'rack/test'

class MyAppTest < Minitest::Test
  include Rack::Test::Methods

  def app
    Sinatra::Application
  end

  def test_my_default
    get '/'
    assert_equal 'Hello World!', last_response.body
  end

  def test_with_params
    get '/meet', :name => 'Frank'
    assert_equal 'Hello Frank!', last_response.body
  end

  def test_with_user_agent
    get '/', {}, 'HTTP_USER_AGENT' => 'Songbird'
    assert_equal "You're using Songbird!", last_response.body
  end
end

Note: If you are using Sinatra in the modular style, replace Sinatra::Application above with the class name of your app.

Sinatra::Base - Middleware, Libraries, and Modular Apps

Defining your app at the top-level works well for micro-apps but has considerable drawbacks when building reusable components such as Rack middleware, Rails metal, simple libraries with a server component, or even Sinatra extensions. The top-level assumes a micro-app style configuration (e.g., a single application file, ./public and ./views directories, logging, exception detail page, etc.). That's where Sinatra::Base comes into play:

require 'sinatra/base'

class MyApp < Sinatra::Base
  set :sessions, true
  set :foo, 'bar'

  get '/' do
    'Hello world!'
  end
end

The methods available to Sinatra::Base subclasses are exactly the same as those available via the top-level DSL. Most top-level apps can be converted to Sinatra::Base components with two modifications:

Your file should require sinatra/base instead of sinatra; otherwise, all of Sinatra's DSL methods are imported into the main namespace.
Put your app's routes, error handlers, filters, and options in a subclass of Sinatra::Base.

Sinatra::Base is a blank slate. Most options are disabled by default, including the built-in server. See Configuring Settings for details on available options and their behavior. If you want behavior more similar to when you define your app at the top level (also known as Classic style), you can subclass Sinatra::Application:

require 'sinatra/base'

class MyApp < Sinatra::Application
  get '/' do
    'Hello world!'
  end
end

Modular vs. Classic Style

Contrary to common belief, there is nothing wrong with the classic style. If it suits your application, you do not have to switch to a modular application.

The main disadvantage of using the classic style rather than the modular style is that you will only have one Sinatra application per Ruby process. If you plan to use more than one, switch to the modular style. There is no reason you cannot mix the modular and classic styles.

If switching from one style to the other, you should be aware of slightly different default settings:

Setting	Classic	Modular	Modular
app_file	file loading sinatra	file subclassing Sinatra::Base	file subclassing Sinatra::Application
run	$0 == app_file	false	false
logging	true	false	true
method_override	true	false	true
inline_templates	true	false	true
static	true	File.exist?(public_folder)	true

Serving a Modular Application

There are two common options for starting a modular app, actively starting with run!:

# my_app.rb
require 'sinatra/base'

class MyApp < Sinatra::Base
  # ... app code here ...

  # start the server if ruby file executed directly
  run! if app_file == $0
end

Start with:

ruby my_app.rb

Or with a config.ru file, which allows using any Rack handler:

# config.ru (run with rackup)
require './my_app'
run MyApp

Run:

rackup -p 4567

Using a Classic Style Application with a config.ru

Write your app file:

# app.rb
require 'sinatra'

get '/' do
  'Hello world!'
end

And a corresponding config.ru:

require './app'
run Sinatra::Application

When to use a config.ru?

A config.ru file is recommended if:

You want to deploy with a different Rack handler (Passenger, Unicorn, Heroku, ...).
You want to use more than one subclass of Sinatra::Base.
You want to use Sinatra only for middleware, and not as an endpoint.

There is no need to switch to a config.ru simply because you switched to the modular style, and you don't have to use the modular style for running with a config.ru.

Using Sinatra as Middleware

Not only is Sinatra able to use other Rack middleware, any Sinatra application can, in turn, be added in front of any Rack endpoint as middleware itself. This endpoint could be another Sinatra application, or any other Rack-based application (Rails/Hanami/Roda/...):

require 'sinatra/base'

class LoginScreen < Sinatra::Base
  enable :sessions

  get('/login') { haml :login }

  post('/login') do
    if params['name'] == 'admin' && params['password'] == 'admin'
      session['user_name'] = params['name']
    else
      redirect '/login'
    end
  end
end

class MyApp < Sinatra::Base
  # middleware will run before filters
  use LoginScreen

  before do
    unless session['user_name']
      halt "Access denied, please <a href='/login'>login</a>."
    end
  end

  get('/') { "Hello #{session['user_name']}." }
end

Dynamic Application Creation

Sometimes you want to create new applications at runtime without having to assign them to a constant. You can do this with Sinatra.new:

require 'sinatra/base'
my_app = Sinatra.new { get('/') { "hi" } }
my_app.run!

It takes the application to inherit from as an optional argument:

# config.ru (run with rackup)
require 'sinatra/base'

controller = Sinatra.new do
  enable :logging
  helpers MyHelpers
end

map('/a') do
  run Sinatra.new(controller) { get('/') { 'a' } }
end

map('/b') do
  run Sinatra.new(controller) { get('/') { 'b' } }
end

This is especially useful for testing Sinatra extensions or using Sinatra in your own library.

This also makes using Sinatra as middleware extremely easy:

require 'sinatra/base'

use Sinatra do
  get('/') { ... }
end

run RailsProject::Application

Scopes and Binding

The scope you are currently in determines what methods and variables are available.

Application/Class Scope

Every Sinatra application corresponds to a subclass of Sinatra::Base. If you are using the top-level DSL (require 'sinatra'), then this class is Sinatra::Application, otherwise it is the subclass you created explicitly. At the class level, you have methods like get or before, but you cannot access the request or session objects, as there is only a single application class for all requests.

Options created via set are methods at class level:

class MyApp < Sinatra::Base
  # Hey, I'm in the application scope!
  set :foo, 42
  foo # => 42

  get '/foo' do
    # Hey, I'm no longer in the application scope!
  end
end

You have the application scope binding inside:

Your application class body
Methods defined by extensions
The block passed to helpers
Procs/blocks used as a value for set
The block passed to Sinatra.new

You can reach the scope object (the class) like this:

Via the object passed to configure blocks (configure { |c| ... })
settings from within the request scope

Request/Instance Scope

For every incoming request, a new instance of your application class is created, and all handler blocks run in that scope. From within this scope you can access the request and session objects or call rendering methods like erb or haml. You can access the application scope from within the request scope via the settings helper:

class MyApp < Sinatra::Base
  # Hey, I'm in the application scope!
  get '/define_route/:name' do
    # Request scope for '/define_route/:name'
    @value = 42

    settings.get("/#{params['name']}") do
      # Request scope for "/#{params['name']}"
      @value # => nil (not the same request)
    end

    "Route defined!"
  end
end

You have the request scope binding inside:

get, head, post, put, delete, options, patch, link and unlink blocks
before and after filters
helper methods
templates/views

Delegation Scope

The delegation scope just forwards methods to the class scope. However, it does not behave exactly like the class scope, as you do not have the class binding. Only methods explicitly marked for delegation are available, and you do not share variables/state with the class scope (read: you have a different self). You can explicitly add method delegations by calling Sinatra::Delegator.delegate :method_name.

You have the delegate scope binding inside:

The top-level binding, if you did require "sinatra"
An object extended with the Sinatra::Delegator mixin

Have a look at the code for yourself: here's the Sinatra::Delegator mixin being extending the main object.

Command Line

Sinatra applications can be run directly:

ruby myapp.rb [-h] [-x] [-q] [-e ENVIRONMENT] [-p PORT] [-o HOST] [-s HANDLER]

Options are:

-h # help
-p # set the port (default is 4567)
-o # set the host (default is 0.0.0.0)
-e # set the environment (default is development)
-s # specify rack server/handler (default is puma)
-q # turn on quiet mode for server (default is off)
-x # turn on the mutex lock (default is off)

Multi-threading

Paraphrasing from this StackOverflow answer by Konstantin

Sinatra doesn't impose any concurrency model but leaves that to the underlying Rack handler (server) like Puma or WEBrick. Sinatra itself is thread-safe, so there won't be any problem if the Rack handler uses a threaded model of concurrency. This would mean that when starting the server, you'd have to specify the correct invocation method for the specific Rack handler. The following example is a demonstration of how to start a multi-threaded Rainbows server:

# config.ru

require 'sinatra/base'

class App < Sinatra::Base
  get '/' do
    "Hello, World"
  end
end

run App

# rainbows.conf

# Rainbows configurator is based on Unicorn.
Rainbows! do
  use :ThreadSpawn
end

To start the server, the command would be:

rainbows -c rainbows.conf

Requirement

The following Ruby versions are officially supported:

Ruby 2.6: 2.6 is fully supported and recommended. There are currently no plans to drop official support for it.
Rubinius: Rubinius is officially supported (Rubinius >= 2.x). It is recommended to gem install puma.
JRuby: The latest stable release of JRuby is officially supported. It is not recommended to use C extensions with JRuby. It is recommended to gem install trinidad.

Versions of Ruby before 2.6 are no longer supported as of Sinatra 3.0.0.

We also keep an eye on upcoming Ruby versions. Expect upcoming 3.x releases to be fully supported.

Sinatra should work on any operating system supported by the chosen Ruby implementation.

Running Sinatra on a not officially supported Ruby flavor means that if things only break there we assume it's not our issue but theirs.

The Bleeding Edge

If you would like to use Sinatra's latest bleeding-edge code, feel free to run your application against the master branch, it should be rather stable.

We also push out prerelease gems from time to time, so you can do a

gem install sinatra --pre

to get some of the latest features.

With Bundler

If you want to run your application with the latest Sinatra, using Bundler is the recommended way.

First, install bundler, if you haven't:

gem install bundler

Then, in your project directory, create a Gemfile:

source 'https://rubygems.org'
gem 'sinatra', :github => 'sinatra/sinatra'

# other dependencies
gem 'haml'                    # for instance, if you use haml

Note that you will have to list all your application's dependencies in the Gemfile. Sinatra's direct dependencies (Rack and Tilt) will, however, be automatically fetched and added by Bundler.

Now you can run your app like this:

bundle exec ruby myapp.rb

Versioning

Sinatra follows Semantic Versioning, both SemVer and SemVerTag.

70 KiB Raw Blame History

Sinatra

Table of Contents

Routes

Conditions

Return Values

Custom Route Matchers

Static Files

Views / Templates

Literal Templates

Available Template Languages

Haml Templates

Erb Templates

Builder Templates

Nokogiri Templates

Liquid Templates

Markdown Templates

RDoc Templates

AsciiDoc Templates

Markaby Templates

RABL Templates

Slim Templates

Yajl Templates

Accessing Variables in Templates

Templates with yield and nested layouts

Inline Templates

Named Templates

Associating File Extensions

Adding Your Own Template Engine

Using Custom Logic for Template Lookup

Filters

Helpers

Using Sessions

Session Secret Security

Session Config

Choosing Your Own Session Middleware

Halting

Passing

Triggering Another Route

Setting Body, Status Code, and Headers

Streaming Responses

Logging

Mime Types

Generating URLs

Browser Redirect

Cache Control

Sending Files

Accessing the Request Object

Attachments

Dealing with Date and Time

Looking Up Template Files

Configuration

Configuring attack protection

Available Settings

Environments

Error Handling

Not Found

Error

Rack Middleware

Testing

Sinatra::Base - Middleware, Libraries, and Modular Apps

Modular vs. Classic Style

Serving a Modular Application

Using a Classic Style Application with a config.ru

When to use a config.ru?

Using Sinatra as Middleware

Dynamic Application Creation

Scopes and Binding

Application/Class Scope

Request/Instance Scope

Delegation Scope

Command Line

Multi-threading

Requirement

The Bleeding Edge

With Bundler

Versioning

Further Reading

70 KiB

Raw Blame History

Templates with `yield` and nested layouts