Gossamer

Gossamer watches you browse a website and record screenshots of your UI, then recreates your browsing session and passes or fails tests depending on whether the UI has changed. It's a way of automating in-browser visual regression testing, using Gossamer to automate Selenium WebDriver, expose test statuses of pass/fail/error, and provide visual diffs of failing tests. You needn't write Selenium tests, or make and keep in sync static pages for testing UI: this tool will test full webpages run on a development or testing webserver. Gossamer can be integrated into your continuous integration either via the command-line interface, or, if you're testing a Python application, via Python unittest integration.

Usage

Gossamer is a command-line application, called with gossamer. You create tests you wish to record in a text Gossamerfile. For each test, a WebDriver window is opened and you interact with the browser as a normal user, going back to the command line when you wish to take a screenshot and pressing enter. Your screenshots, and a JSON record of your test, is written to a data directory. Playback is done by reading this directory, and comparing against 'good' screenshots. Gossamer assumes that Selenium Server is already running.

To start, create a file Gossamerfile and specify a name and URL to visit for every test.

[example]
url=http://www.example.com

You can also add in additional settings:

[example]
url=http://www.example.com
desc=Example.com hasn't changed
screensize=800x1000
browser=chrome
expect_redirect=false

By default, Gossamer looks for a file called Gossamerfile in the current directory, and stores data in ./gossamer with one directory per test. Each test directory contains a record.json containing the data to reproduce the test, as well as good screenshots, and in a sub-directory last, the last test run's (possibly failing) screenshots.

You can run your tests with:

gossamer --file Gossamerfile --data <data_dir> --record --save-diff

If you wish to run only a subset of tests in that file, specify those tests' names as positional arguments.

When you browse, wait for requests to finish and rendering to be complete before moving on to another action. If you navigate to a new page, you will need to take a screenshot before new events are observed.

If your UI has changed and you wish to update the screenshots to match, then run with --rerecord: the test will be rerun automatically, and new PNGs will be saved. To playback the tests, simply call without an -r/-rr flag.

If you're running Python tests, you can integrate your Gossamer tests like so:

# myapp/test.py
from gossamer import run_gossamerfile
run_gossamerfile(locals(), <filename>, <data_dir>)

This mutates your module's locals to include a unittest.TestCase instance for every test in the given Gossamerfile(s). Your test runner will then detect and run them. You will, however, need to ensure that your Selenium server and test webserver are up when your tests are run.

Installation

Your testing machine will need Pillow's system-level dependencies for PNG support. Gossamer can then be installed from PyPi with pip install gossamerui.

On that machine or another accessible to it you will need Selenium Server installed and running. Note that Selenium Server comes with Firefox by default, needing an additional system package for Chrome, and for Internet Explorer an IE-specific standalone version of Selenium Server.

You'll also need your 'target' webserver running on any machine.

Authors

See the file AUTHORS. Based on Facebook's Huxley, and rewritten.

License

Apache 2.0

Known Issues

• Scrolling is unreliable.
• Opening a slow iframe will likely timeout on playback.
• Internet Explorer < 11 (which is all Selenium supports) shouldn't work at the moment because we use MutationObservers, but an older way of observing changes can be added for IE<11 (see js.pageChangingObserver).

Issues

Please create issues and pull requests at the GitHub repository.

Contributing

• Once you have the repository, setup using make develop.
• Please add tests and use the included .pylintrc; you can run make test and make lint.
• If any breaking changes are made to data structures, increment constant.DATA_VERSION and modify util.import_recorded_run to handle both new and old data.
• Feel free to contribute any functionality you want.