Write tests. Not too many. Mostly integration.
summary
Argument claims write tests not too many mostly integration and explains why integration tests hit the best tradeoff between cost, speed and confidence. Demonstrations mention ESLint and Flow typing to eliminate trivial tests, Istanbul coverage reports and the dangers of chasing 100% code coverage, and browser end-to-end checks with Cypress. Anecdotes include a popular tweet (≈180 retweets) and a testing “trophy”; conclusions show what to build, what tools to run, and why integration tests actually work.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:22Talk Overview
- 1:02Slides
- 1:23Tweet
- 1:51Enhanced workflows
- 2:19Confidence
- 2:43Wakeboarding
- 2:58Why we write tests
- 3:23Code coverage
- 3:43Code coverage example
- 4:06Was it depends
- 4:17Good code coverage
- 5:44Testing implementation details
- 7:13Integration
- 7:34Gleb Pyramid
- 7:49Testing Dorito
- 8:48ES lint
- 9:30Flow lint
- 10:14Unit tests
- 10:36Integration tests
- 11:19Endtoend tests
- 13:07Test speed
- 13:52Simple problems
- 14:26Mostly integration
- 15:21Unit vs integration
- 15:42Poke fewer holes
- 16:19Test higher up
- 17:03Thank you Guillermo
related talks
30mWait, before you write your next test...
24mDon't write a lot of tests!
37mUnderstanding the successes and the pain points of different testing strategies
39mTesting, Testing, 1, 2, NaN
29mTesting Is All About Principles
31mWriting (Really) Good Tests
35mTesting in Frontend World
55mUnit tests are useless. Change my mind