Module: Roda::RodaPlugins::Chunked

Defined in:
lib/roda/plugins/chunked.rb

Overview

The chunked plugin allows you to stream rendered views to clients. This can significantly improve performance of page rendering on the client, as it flushes the headers and top part of the layout template (generally containing references to the stylesheet and javascript assets) before rendering the content template.

This allows the client to fetch the assets while the template is still being rendered. Additionally, this plugin makes it easy to defer executing code required to render the content template until after the top part of the layout has been flushed, so the client can fetch the assets while the application is still doing the necessary processing in order to render the content template, such as retrieving values from a database.

There are a couple disadvantages of streaming. First is that the layout must be rendered before the content, so any state changes made in your content template will not affect the layout template. Second, error handling is reduced, since if an error occurs while rendering a template, a successful response code has already been sent.

To use chunked encoding for a response, just call the chunked method instead of view:

r.root do
  chunked(:index)
end

If you want to execute code after flushing the top part of the layout template, but before rendering the content template, pass a block to chunked:

r.root do
  chunked(:index) do
    # expensive calculation here
  end
end

You can also call delay manually with a block, and the execution of the block will be delayed until rendering the content template. This is useful if you want to delay execution for all routes under a branch:

r.on 'albums', Integer do |album_id|
  delay do
    @album = Album[album_id]
  end
  r.get 'info' do
    chunked(:info)
  end
  r.get 'tracks' do
    chunked(:tracks)
  end
end

If you want to chunk all responses, pass the :chunk_by_default option when loading the plugin:

plugin :chunked, chunk_by_default: true

then you can just use the normal view method:

r.root do
  view(:index)
end

and it will chunk the response. Note that you still need to call chunked if you want to pass a block of code to be executed after flushing the layout and before rendering the content template. Also, before you enable chunking by default, you need to make sure that none of your content templates make state changes that affect the layout template. Additionally, make sure nowhere in your app are you doing any processing after the call to view.

If you use :chunk_by_default, but want to turn off chunking for a view, call no_chunk!:

r.root do
  no_chunk!
  view(:index)
end

Inside your layout or content templates, you can call the flush method to flush the current result of the template to the user, useful for streaming large datasets.

<% (1..100).each do |i| %>
  <%= i %>
  <% sleep 0.1 %>
  <% flush %>
<% end %>

Note that you should not call flush from inside subtemplates of the content or layout templates, unless you are also calling flush directly before rendering the subtemplate, and also directly injecting the subtemplate into the current template without modification. So if you are using the above template code in a subtemplate, in your content template you should do:

<% flush %><%= render(:subtemplate) %>

If you want to use chunked encoding when rendering a template, but don’t want to use a layout, pass the layout: false option to chunked.

r.root do
  chunked(:index, layout: false)
end

In order to handle errors in chunked responses, you can override the handle_chunk_error method:

def handle_chunk_error(e)
  env['rack.logger'].error(e)
end

It is possible to set @_out_buf to an error notification and call flush to output the message to the client inside handle_chunk_error.

In order for chunking to work, you must make sure that no proxies between the application and the client buffer responses.

If you are using nginx and have it set to buffer proxy responses by default, you can turn this off on a per response basis using the X-Accel-Buffering header. To set this header or similar headers for all chunked responses, pass a :headers option when loading the plugin:

plugin :chunked, headers: {'X-Accel-Buffering'=>'no'}

By default, this plugin does not use Transfer-Encoding: chunked, it only returns a body that will stream the response in chunks. If you would like to force the use of Transfer-Encoding: chunked, you can use the :force_chunked_encoding plugin option. If using the :force_chunked_encoding plugin option, chunking will only be used for HTTP/1.1 requests since Transfer-Encoding: chunked is only supported in HTTP/1.1 (non-HTTP/1.1 requests will have behavior similar to calling no_chunk!).

The chunked plugin requires the render plugin, and only works for template engines that store their template output variable in injected into the layout template without modification.

If using the chunked plugin with the flash plugin, make sure you call the flash method early in your route block. If the flash method is not called until template rendering, the flash may not be rotated.

Defined Under Namespace

Modules: InstanceMethods Classes: Body, StreamBody

Class Method Summary collapse

Class Method Details

.configure(app, opts = OPTS) ⇒ Object

Set plugin specific options. Options:

:chunk_by_default

chunk all calls to view by default

:headers

Set default additional headers to use when calling view



159
160
161
162
163
164
165
 
# File 'lib/roda/plugins/chunked.rb', line 159

def self.configure(app, opts=OPTS)
  app.opts[:chunk_by_default] = opts[:chunk_by_default]
  app.opts[:force_chunked_encoding] = opts[:force_chunked_encoding]
  if opts[:headers]
    app.opts[:chunk_headers] = (app.opts[:chunk_headers] || {}).merge(opts[:headers]).freeze
  end
end

.load_dependencies(app, opts = OPTS) ⇒ Object

Depend on the render plugin



152
153
154
 
# File 'lib/roda/plugins/chunked.rb', line 152

def self.load_dependencies(app, opts=OPTS)
  app.plugin :render
end