We believe that visual regression testing is only useful if it is done on every build, for every change, and fast enough to be integrated in your existing development workflow and CI services. We coined the phrase "continuous visual integration" to describe this more expansive view of visual testing and how Percy fits in the world of continuous integration.
First, Percy renders a screenshot of the page and then visually compares it, pixel by pixel, to a previous screenshot. The output of this comparison creates a third image---this is the "visual diff" or "perceptual diff" that shows exactly what pixels changed between the images:
Nope! You might be surprised to learn that under the hood, Percy is not designed to accept screenshots, but instead captures DOM snapshots and page assets (CSS, images, etc.). This is very intentional---we want to keep your environment fast and handle all the complexities of deterministic rendering and visual diffs in our infrastructure. We also want to support visual testing for any web content, regardless of the language or framework it was created in.
Nope! Percy is designed to keep all the computationally expensive screenshot rendering and visual diffing completely out of your systems. By accepting DOM snapshots and assets alone, we can highly optimize Percy to have lightweight clients in numerous languages and environments with little performance impact on your side.
The first build may be slower than usual while assets are first uploaded, but then assets are never uploaded unless they change.
Short answer: yes! We have a set of officially supported client libraries and are adding more. Our asset-centric design lets us support any web content no matter what language or framework it was created in. We want to encourage a developer ecosystem of tools that can easily integrate with Percy.
Yes! By default, both Firefox and Chrome are enabled for each new project, and browsers can be enabled or disabled at any time. Read more about cross-browser testing in the docs.
Short answer: we don't. If you have page content that changes (auto-generated test data, timestamps, dynamic test users, etc.) you may see visual diffs in every build. You can simply ignore them if you know that they are correct, or take steps to reduce the dynamic data present in your feature tests. For example, you could use static fixture data or inject static content just for the relevant tests.
We are exploring features to help with this problem. Please email us if you're interested.
No, unless your app is built using HTML and CSS.
Although this is not a feature in Percy itself, you can get this behavior by creating webpages that contain only the portion you want to snapshot. In many frameworks, you can do this programmatically within your integration tests before you take the snapshot.
// Execute within the browser context of the page to snapshot. let scopedContainer = document.querySelector(selector); document.querySelector('body').innerHTML = scopedContainer.outerHTML
You can see a working example of this approach in our sample
percy-puppeteer app: https://github.com/percy/example-percy-puppeteer/blob/master/tests/todomvc_spec.js#L66-L84
This is one of the hard parts of perceptual diffs---the diff itself is pixel-by-pixel, so the diff may end up looking like noise if the content shifts by any amount. You still get the benefits of knowing when and where your app has changed, but it may require more manual analysis. In Percy, you can simply click the diff to hide it and view the underlying screenshot to compare visually. Also, see Animations for info on how Percy freezes animations.
I'm glad you asked! Check out our RailsConf 2016 talk:
Updated about a year ago