Stability in Unstability: Self-Healing Test Automation
summary
A prototype self-healing test automation system that repairs broken locators by combining image recognition with DOM-comparison algorithms (including graph edit distance and a modified LCS) and uses a synthetic DOM generator for testing. Evaluated on a 1.5k-test suite where ~40% failures were due to missing elements; the prototype can find moved elements and keep tests running but currently has high memory use, is slow on large DOMs, and can produce false positives, with planned performance improvements and open-sourcing.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 4:41Unstable locators problems
- 7:59Self-healing test idea
- 8:20Neural networks in testing
- 12:44Visual recognition for stability
- 16:20Dynamic DOM locators
- 19:12Existing tools overview
- 23:10New solution requirements
- 23:56Edit Distance algorithm
- 27:53Longest Common Subsequence
- 30:19Algorithm demo
- 34:22Approach pros and cons
- 35:27Development plans
- 36:51Quick implementation hack
- 44:24Q&A
related talks
55mStability in Unstability: Self‑Healing Test Automation
30m'Self-Healing Tests' - The holy grail of test automation?
32mSelf-healing system for UI tests using ML
25mSelf-Healing in Test Automation
44mSelf-healing Test Automation 2.0: The Future
43mSmart and Stable Automated Testing Using AI Based Dynamic Locators
31mSelf-Healing or Self-Killing?
15mLLMs and Test Automation: A New Era of Self Healing