Writing tests with AI

Tests are how you stop fixing the same bug twice. They're also how you safely let AI refactor your code: if the tests still pass, the behavior is intact. You don't need to test everything — test the logic that would hurt if it broke.

What's worth testing in DayOne

• The String.firstLine helper — pure logic, easy to get subtly wrong.
• Decoding Insights from a sample JSON string — proves your struct matches the model's output.
• The 'Save is disabled when empty' rule — a behavior your users rely on.

import Testing
@testable import DayOne

struct DayOneTests {
    @Test func firstLineReturnsFirstLine() {
        #expect("Hello\nWorld".firstLine == "Hello")
        #expect("".firstLine == "Untitled")
    }

    @Test func decodesInsightsJSON() throws {
        let json = #"{"title":"A quiet win","mood":"calm","reflection":"What helped?"}"#
        let insights = try JSONDecoder().decode(Insights.self, from: Data(json.utf8))
        #expect(insights.mood == "calm")
    }
}

This uses Swift Testing (the modern #expect style). Each @Test function checks one claim. Run them with ⌘U — green means the claim holds. If you later change firstLine and it breaks, the test fails loudly instead of the bug reaching users.