actionview/lib/action_view/renderer/streaming_template_renderer.rb

Summary

Maintainability
A
1 hr
Test Coverage
# frozen_string_literal: true

require "fiber"

module ActionView
  # == TODO
  #
  # * Support streaming from child templates, partials and so on.
  # * Rack::Cache needs to support streaming bodies
  class StreamingTemplateRenderer < TemplateRenderer # :nodoc:
    # A valid Rack::Body (i.e. it responds to each).
    # It is initialized with a block that, when called, starts
    # rendering the template.
    class Body # :nodoc:
      def initialize(&start)
        @start = start
      end

      def each(&block)
        begin
          @start.call(block)
        rescue Exception => exception
          log_error(exception)
          block.call ActionView::Base.streaming_completion_on_exception
        end
        self
      end

      private
        # This is the same logging logic as in ShowExceptions middleware.
        def log_error(exception)
          logger = ActionView::Base.logger
          return unless logger

          message = +"\n#{exception.class} (#{exception.message}):\n"
          message << exception.annotated_source_code.to_s if exception.respond_to?(:annotated_source_code)
          message << "  " << exception.backtrace.join("\n  ")
          logger.fatal("#{message}\n\n")
        end
    end

    # For streaming, instead of rendering a given a template, we return a Body
    # object that responds to each. This object is initialized with a block
    # that knows how to render the template.
    def render_template(view, template, layout_name = nil, locals = {}) # :nodoc:
      return [super.body] unless layout_name && template.supports_streaming?

      locals ||= {}
      layout   = find_layout(layout_name, locals.keys, [formats.first])

      Body.new do |buffer|
        delayed_render(buffer, template, layout, view, locals)
      end
    end

    private
      def delayed_render(buffer, template, layout, view, locals)
        # Wrap the given buffer in the StreamingBuffer and pass it to the
        # underlying template handler. Now, every time something is concatenated
        # to the buffer, it is not appended to an array, but streamed straight
        # to the client.
        output  = ActionView::StreamingBuffer.new(buffer)
        yielder = lambda { |*name| view._layout_for(*name) }

        ActiveSupport::Notifications.instrument(
          "render_template.action_view",
          identifier: template.identifier,
          layout: layout && layout.virtual_path,
          locals: locals
        ) do
          outer_config = I18n.config
          fiber = Fiber.new do
            I18n.config = outer_config
            if layout
              layout.render(view, locals, output, &yielder)
            else
              # If you don't have a layout, just render the thing
              # and concatenate the final result. This is the same
              # as a layout with just <%= yield %>
              output.safe_concat view._layout_for
            end
          end

          # Set the view flow to support streaming. It will be aware
          # when to stop rendering the layout because it needs to search
          # something in the template and vice-versa.
          view.view_flow = StreamingFlow.new(view, fiber)

          # Yo! Start the fiber!
          fiber.resume

          # If the fiber is still alive, it means we need something
          # from the template, so start rendering it. If not, it means
          # the layout exited without requiring anything from the template.
          if fiber.alive?
            content = template.render(view, locals, &yielder)

            # Once rendering the template is done, sets its content in the :layout key.
            view.view_flow.set(:layout, content)

            # In case the layout continues yielding, we need to resume
            # the fiber until all yields are handled.
            fiber.resume while fiber.alive?
          end
        end
      end
  end
end