I have an embarrassing programming secret. Please don’t judge me harshly.
I don’t use automated tests.
I’m not a complete moron. Whatever I’m currently working on is thoroughly tested manually before being committed in git. And I commit pretty frequently with descriptive commit messages. I also fork and merge religiously.
I haven’t really felt the need to set up automated tests. Not til yesterday.
THAT ONE TIME I WRECKED IT
I’m currently building an API server for a project. An offline-capable smart spreadsheet browser app.
Made some updates to the client and wired it up to the API endpoint to test the feature and the server came crashing down. My changes were to the client, which was in a separate repo, so I didn’t change anything on the server. Restarted the API server and tested it again. Crash and burn.
What went wrong?
A quick investigation pointed to when I renamed a few variables to make things more readable. I didn’t change every instance of a particular variable. Two slipped by me. I didn’t even realize there were two instances til I fixed one, pushed the change up and the server crashed again.
It’s supremely easy to make mistakes or not check as thoroughly as you think, which introduces bugs into you code.
Manual testing is okay, but you can’t count on it to catch every single problem. And every single problem matters in production. It’s the difference between a well-running app and steaming pile.
No more relying on myself to make sure everything is working properly. I’m writing a complete test suite for the server before I build any further.
From now on, I’ll test it so I don’t wreck it.