Keep your Selenium browser open while you use Capybara to talk to it. Save seconds and sanity.
Go to file
2015-01-09 14:55:26 -05:00
bin some cleanup, docs too 2013-01-25 10:21:17 -05:00
lib Split up Cucumber and RSpec support 2015-01-09 14:55:26 -05:00
skel/features/support add the ability to install a cucumber helper 2013-01-25 10:56:56 -05:00
.gitignore initial commit 2012-11-26 11:36:43 -05:00
Gemfile initial commit 2012-11-26 11:36:43 -05:00
LICENSE initial commit 2012-11-26 11:36:43 -05:00
persistent_selenium.gemspec change for capybara 2.1 2013-04-09 11:37:14 -04:00
Rakefile initial commit 2012-11-26 11:36:43 -05:00
README.md update readme 2013-04-09 11:29:44 -04:00

Now you can keep that precious browser window open when doing continuous integration testing. Save seconds, and sanity, with every test re-run!

Also, the browser stays open at its last state so you can inspect it and more easily fix your tests and/or code.

Start an instance:

persistent_selenium [ --port 9854 ] [ --browser firefox ]

Tell Capybara to use it:

# features/support/env.rb

require 'persistent_selenium/driver'
Capybara.default_driver = :persistent_selenium

If you're using Cucumber, you can also install that hook:

persistent_selenium install

Should work just the same as if you used the standard Capybara Selenium driver, except for these two differences:

  • The browser starts up first thing and sticks around, so you don't pay the startup/shutdown penalty with each test run.
  • The last page you were on before your tests passed/failed stays there, so you can inspect it and adjust your tests.

The browser's cache is disabled, and cookies are reset before the next test runs, so you still get the state cleared out before your next set of tests.

Under the hood

It's DRb, which mostly Just Works (tm), and has a little reshuffling of the default Capybara Selenium driver's code.

When DRb doesn't Just Work (tm)

You're most likely using all and invoking an action on one of the nodes within, I'd wager. If you need to find a node to perform an action on, it's best to stick with find, since it's less likely that node will go out of ObjectSpace that quickly. If you need to examine the document for particular properties, and all seems like the best way to do it, instead try parsing the document body with Nokogiri and using its finders. That way, all your node searching will be done on the client end.