Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


What's This?

still_life is a testing framework enhancements for test-unit, minitest, RSpec, and Capybara that records all HTML response body texts that are rendered during E2E or unit test executions.

So What?

You can compare actually rendered HTML results before and after any app updates.

For What?

By comparing all these HTML files that are processed before and after any kind of code change, you can make sure that you did not (or you did) introduce any new user-facing incompatibilities. This may greatly help you for example, refactoring your app, replacing external libraries, or upgrading libraries. My personal use case that made me gemifying still_life was that I wanted to make sure that my own template engine renders the same HTML as the one that I was using.

But indeed, the real sweet spot of this tiny library is IMO "Rails upgrade". In fact, The first original version of this tool was implemented as an RSpec monkeypatch while we were upgrading a huge Rails application from Rails 2 to Rails 3.


Bundle still_life gem to your Rails app's :test environment.

gem 'still_life', group: :test


Run tests with an envvar STILL_LIFE. Then still_life creates some HTML files under tmp/html/#{ENV['STILL_LIFE']}/ directory. Each .html file is named from the location in your test code where the request was made.

For instance, if you run the tests against a simple scaffold app, the generated files will be like this:

% STILL_LIFE=rails52 rails test:system test
% tree tmp/html
└── rails52
    └── test
        ├── controllers
        │   ├── users_controller_test.rb-14.html
        │   ├── users_controller_test.rb-20.html
        │   ├── users_controller_test.rb-27.html
        │   ├── users_controller_test.rb-32.html
        │   ├── users_controller_test.rb-37.html
        │   ├── users_controller_test.rb-43.html
        │   └── users_controller_test.rb-9.html
        └── system
            ├── users_test.rb-14.html
            ├── users_test.rb-18.html
            ├── users_test.rb-21.html
            ├── users_test.rb-25.html
            ├── users_test.rb-26.html
            ├── users_test.rb-29.html
            ├── users_test.rb-32.html
            ├── users_test.rb-36.html
            ├── users_test.rb-37.html
            └── users_test.rb-9.html

4 directories, 17 files

And each file content is just an HTML.

% cat tmp/html/rails52/test/system/users_test.rb-18.html
<!DOCTYPE html><html xmlns=""><head>

    <link rel="stylesheet" media="all" href="/assets/application-35729bfbaf9967f119234595ed222f7ab14859f304ab0acc5451afb387f637fa.css" data-turbolinks-track="reload" />
    <script src="/assets/application-3c2e77f06bf9a01c87fc8ca44294f3d3879d89483d83b66a13a89fc07412dd59.js" data-turbolinks-track="reload"></script>

    <p id="notice">User was successfully created.</p>


<a href="/users/980190963/edit">Edit</a> |
<a href="/users">Back</a>



Usage Scenario

Consider you have a well-tested Rails 5.2 app, and you want to upgrade its Rails version to 6.0 without introducing any user-facing incompatibilities. Then the workflow will be as follows:

1. Draw a still_life with the 5.2 app

% STILL_LIFE=rails52 rails test:system test

2. Do the upgrade job

% bundle u
% rails app:update

and push some more commits...

3. Draw another still_life with the 6.0 app

% STILL_LIFE=rails60 rails test:system test

4. Compare the results, and make sure there's no unexpected diffs

% git diff --no-index --color-words tmp/html/rails52/ tmp/html/rails60/


git diff

As written in the above example, git diff --no-index --color-words should perfectly work for recursively comparing two output directories.

Random Values

If your response includes some kind of random values, the test results may change between each test runs. In such case, maybe you could specify a random seed, or mock the random source in your app.

TODOs / Known Issues

  • The Capybara monkeypatch sometimes fails to get the page.body due to Capybara timing problem
  • Support older versions of Rails, Capybara, and Ruby
  • Fix the CI with RSpec + headless Chrome


Pull requests are welcome on GitHub at


The original idea of this library was implemented as a 10 LOC anonymous module by @hotchpotch at Cookpad Inc. back in 2011 as written in this slide.


The gem is available as open source under the terms of the MIT License.