Test

Writing Tests

Tests verify behavior, not implementation. A test that breaks when you rename a variable is a bad test.

Step 1: Understand what to test

Before writing any test:

• Read the function or module you are testing in full
• Identify its contract: given input X, what output or side effect is promised?
• Find existing tests to understand the pattern in this codebase

Step 2: Identify the cases

For every function or behavior, enumerate:

Happy path The normal input that produces the expected output.

Boundary cases

• Empty input (empty string, empty array, zero, null)
• Single element vs. many elements
• Min and max values
• Exact boundary values (e.g. >= vs >)

Error cases

• Invalid input: what should happen?
• Missing required fields
• Network failure, file not found, permission denied

Edge cases specific to the logic Think about what could go wrong given how the code works.

Step 3: Write the tests

Follow the existing test framework and style in the codebase. Do not introduce a new testing library.

Structure each test as:

1. Arrange: set up inputs and any mocks or state
2. Act: call the function
3. Assert: verify the outcome

it("returns null when the token is expired", () => {
  // Arrange
  const token = createToken({ expiresAt: new Date("2000-01-01") });

  // Act
  const result = validateToken(token);

  // Assert
  expect(result).toBeNull();
});

Rules:

• One assertion per test (or logically grouped assertions about the same outcome)
• Test names describe the behavior, not the implementation: "returns null when token is expired" not "test validateToken case 3"
• Do not test private implementation details
• Mock only external dependencies (network, filesystem, time), not your own code

Step 4: Verify

• Run the tests and confirm they pass
• Temporarily break the implementation to confirm each test actually catches the failure
• Check that test names read clearly in the output

Report format

Tests written

• path/to/file.test.ts — what behaviors are now covered

Cases covered List the scenarios now tested.

Cases not covered Any gaps you identified but did not write tests for, and why (e.g. requires integration setup, out of scope).