I agree that we need to improve UI tests. Happy to make that a priority during 3.1.x and 3.2 development.
Personally, I will start with expanding unit tests first. But that doesn't mean others can't get started on E2E tests. I lean towards using typescript for playwright. With E2E we are testing the UI more than we are testing the API so I think the code should sit closer to the UI code. On Tue, Sep 23, 2025 at 9:38 AM Vincent Beck <[email protected]> wrote: > Not answering any of these questions (sorry) but strong +1 on adding > end-to-end (E2E) tests. This will significantly make our UI more robust but > also will speed up some of our work. An example is front-end dependency > upgrades by dependabot. Today we have no way of knowing whether these PRs > break the UI, in order to merge these PRs we need to compile the front-end > assets and test manually whether the UI work. Having end-to-end tests on > the front-end and making them run in the CI would speed up this very > example (on top of increasing code robustness, increase our confidence etc). > > On 2025/09/23 11:03:40 Rahul Vats wrote: > > Hi all, > > > > With Airflow 3.0, the UI has undergone a complete revamp, and new changes > > are naturally more prone to regressions. We've already seen more UI > > regressions that are hard to catch with our current limited test > coverage. > > Manual testing is also challenging given the wide variety of workflows > and > > screens in the UI. > > > > Currently, the Airflow UI has a few unit tests using Vitest and React > > Testing Library, but coverage is quite limited — only about 4 test files > > covering major pages like DagsList and TaskInstance. Many critical > > pages (Connections, > > Variables, Pools, Dashboard, etc.) have no test coverage at all. > > > > I’m exploring a two-pronged approach to improve UI testing: > > > > 1. > > > > Expand unit test coverage for individual components and pages > > > > 2. > > > > Add end-to-end (E2E) tests using Playwright for complete user > workflows > > > > > > This would give us both component-level safety and full user-journey > > testing across browsers, helping catch regressions early. > > > > A few questions: > > > > - > > > > Should we prioritize expanding unit test coverage first, or work on > unit > > and E2E tests in parallel? > > > > - > > > > For E2E testing, is Playwright a good choice, or are there other > > preferred tools we should explore? > > > > - > > > > If Playwright, should we use TypeScript or Python as language for > tests? > > > > - > > > > Any other suggestions to strengthen our UI testing strategy? > > > > > > I’d love to hear your thoughts on whether this would be valuable for the > > project before diving into implementation details. > > > > Regards, > > Rahul Vats > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
