Google Ajax Crawler

Rack Middleware adhering to the Google Ajax Crawling Scheme, using a headless browser to render JS heavy pages and serve a dom snapshot of the rendered state to a requesting search engine.

Details of the scheme can be found at: https://developers.google.com/webmasters/ajax-crawling/docs/getting-started

Using

install

gem install google_ajax_crawler

In your config.ru

require 'google_ajax_crawler'

use GoogleAjaxCrawler::Crawler do |config|
  config.page_loaded_js = "MyApp.isPageLoaded()"
end

app = -> env { [200, {'Content-Type' => 'text/plain'}, "b" ] }
run app

rails usage

create in the initializer folder :

google_ajax_crawler_middleware.rb

with

if defined?(Rails.configuration) && Rails.configuration.respond_to?(:middleware)
  require 'google_ajax_crawler'
  Rails.configuration.middleware.insert_before ActionDispatch::Static, GoogleAjaxCrawler::Crawler do |config|
    config.page_loaded_test = -> driver { driver.page.evaluate_script('document.getElementById("loading") == null') }
  end
end

GoogleAjaxCrawler has to be inserted into the middleware stack before ActionDispatch::Static. It is to prevent serving static pages with angular controllers on them before Crawler has change to react and serve snapshot.

Important

Concurrent requests must be enabled to allow your site to snapshot itself. If concurrent requests are not allowed, the site will simple hang on a crawler request.

In config/application.rb :

config.threadsafe!

Examples

In the examples folder, each driver has a rackup file (at the moment only one driver, capybara-webkit, exists), which can be launched:

rackup examples/capybara_webkit.ru

Examples for how to use the crawler with Backbone.JS, Angular.JS and plain ol javascript are accesible via:

• http://localhost:9292/backbone
• http://localhost:9292/angular
• http://localhost:9292/

Curl, or open a browser to http://localhost:9292/[framework]#!test and view source.... This is how a search engine will see your page before snapshotting. NOTE: don't look at the markup through a web inspector as it will most likely display dom elements rendered on the fly by js.

Change the url to http://localhost:9292/[framework]?\_escaped_fragment_=test , and then again curl or view source to see how the DOM state has been captured

Configuration Options

Page Loaded Tests

As determining when a page has completed rendering can depend on a number of qualitative factors (i.e. all ajax requests have responses, certain content has been displayed, or even when there are no loaders / spinners visible on the page), you can specify one of two ways to tell the crawler that your page has finished loading / rendering and to return a snapshot of the rendered dom at that time.

page_loaded_js (client side test)

Tell the crawler the client side javascript function (returning true/false) you have created, that determines when your page has finished loading / rendering.

use GoogleAjaxCrawler::Crawler do |config|
  config.page_loaded_js = "MyApp.isPageLoaded()"
end

page_loaded_test (server side test)

A server side test determining when your page has finished loading / rendering. The configured crawler driver is passed to the lambda to allow querying of the current page's dom state from the server side.

use GoogleAjaxCrawler::Crawler do |config|
  config.page_loaded_test = -> driver { driver.page.has_css?('.loading') == false }
end

timeout

The max time (in seconds) the crawler should wait before returning a response. After the timeout has been reached, a snapshot of the DOM in its current state is returned. Defaults to 30 seconds.

driver

The configured google ajax crawler driver used to query the current page state. Defaults to capybara_webkit.

poll_interval

How often (in seconds) to test the page state with the configured page_loaded_test. Defaults to 0.5 seconds.

response_headers

What response headers shoudl be returned with the dom snapshot. Default headers specify the content-type text/html.

requested_route_key

The parameter name used by a search bot to idenitfy which client side route to snapshot. Defaults to escaped_fragment.

Identifing Search Engine Requests

Snapshot requests are passed an additional query string param (?search_engine=true), allowing you to optionally execute client side code. This is particularly handy should you have stats tracking code (i.e. Google Analytics), which you don't want executed / included when search engines are trawling your site.

License

All free - Use, modify, fork to your hearts content... See LICENSE.txt for further details.