Poor man's visual regression testing

Let's assume you've made some big changes to some global css styles in your web application.
You're unsure whether these changes broke any of your meticulously crafted pages.

You could set up automatic visual regression testing on your CI pipeline with tools from the "Awesome Visual Regression Testing" list.
That would probably be the safest option. But it requires some setup and time to get used to the tools.

You don't have time for that right now? Then let me introduce you to...

Manual Regression Testing

The idea is: Take a screenshot from a known working state of your webapp. For example your production environment. Then take this screenshot and overlay it on top of your potentially broken state of your webapp.
In this example I'm using the chrome plugins

• screenshot-capture - to get a screenshot into my clipboard
• PerfectPixel - to overlay the screenshot on top of the development page

First I configured screenshot-capture to copy the screenshot into my clipboard and to use a different shortcut as the default one (OPTION + S) was already occupied by PerfectPixel.

screenshot-capture plugin configuration to copy into clipboard

Prepare your browser

Then I prepare my browser to have both (local and production environment) open in equal sized browser windows.
It's important those pages have the same viewport size as comparing differently sized images will just make the process more complicated.
This also means the devtools need to be closed during the screenshot and comparison.
If the screenshot doesn't have the same size as the viewport of your browser window, you would need to manually position the screenshot later.

Screenshot and compare

Next I take a screenshot of my production webapp running on https://idkshite.com (with OPTION + SHIFT + S).
This screenshot is now in my clipboard. I can switch to http://localhost:3000/ and open PerfectPixel to click on the paste from clipboard link.
Now the screenshot of the reference page (production environment) should sit on top of the viewport of your localhost browser window.
Feel free to play around with the opacity slider to get an impression of how those changes differ.
Or turn your screenshot into a negative with OPTION + i to get a stronger contrast.

🎥 Original post contains a Video available on idkshite.com

In the long run I would probably invest in some automated regression testing, but as a provisional arrangement it's way more reliable than just glancing over visual changes. 👨‍🎨

Comments

[Michael Mangialardi]

I like to open two instances of an application in separate tabs and then click back and forth between each tab. Remarkably, your eyes will easily catch any jumps even if its a few pixels.

[Rob OLeary]

Nice domain name. 😉

Would it not be easier to use an end-to-end testing library that does visual diffing e.g. Playwright?

[Lucca Sanwald]

Thanks 🙏
I agree it's easier in the long term to set up an end-to-end testing library and reap the benefits on every single commit (ergo test run). But short term, manual testing is arguably easier to setup and gives you results much faster. But it all depends on the use case I guess 😄