tdd（时差）

Test-Driven Development

Philosophy

Core principle: Tests should verify behavior through public interfaces, not implementation details. Code can change entirely; tests shouldn't.

Good tests are integration-style: they exercise real code paths through public APIs. They describe *what* the system does, not *how* it does it. A good test reads like a specification — "user can checkout with valid cart" tells you exactly what capability exists. These tests survive refactors because they don't care about internal structure.

Bad tests are coupled to implementation. They mock internal collaborators, test private methods, or verify through external means (like querying a database directly instead of using the interface). The warning sign: your test breaks when you refactor, but behavior hasn't changed. If you rename an internal function and tests fail, those tests were testing implementation, not behavior.

See tests.md for examples and mocking.md for mocking guidelines.

Anti-Pattern: Horizontal Slices

DO NOT write all tests first, then all implementation. This is "horizontal slicing" — treating RED as "write all tests" and GREEN as "write all code."

This produces bad tests:

• Tests written in bulk test *imagined* behavior, not *actual* behavior
• You end up testing the *shape* of things (data structures, function signatures) rather than user-facing behavior
• Tests become insensitive to real changes — they pass when behavior breaks, fail when behavior is fine
• You outrun your headlights, committing to test structure before understanding the implementation

Correct approach: Vertical slices via tracer bullets. One test → one implementation → repeat. Each test responds to what you learned from the previous cycle. Because you just wrote the code, you know exactly what behavior matters and how to verify it.

WRONG (horizontal):
  RED:   test1, test2, test3, test4, test5
  GREEN: impl1, impl2, impl3, impl4, impl5

RIGHT (vertical):
  RED→GREEN: test1→impl1
  RED→GREEN: test2→impl2
  RED→GREEN: test3→impl3
  ...

Reference Files

Workflow

1. Explore the codebase

Use a subagent to understand the existing context before writing any tests. Look for:

• Which test framework is in use? (jest, pytest, vitest, xunit, go test, etc.)
• How are test files structured and named? (e.g. *.test.ts, tests/, *_test.go)
• Are there existing similar tests to use as prior art?
• What is the module under test and what are its current dependencies?

Matching the existing patterns matters — tests that look alien to the codebase don't get maintained.

2. Plan

Before writing any code:

• Confirm with the user what interface changes are needed
• Confirm with the user which behaviors to test (prioritize)
• Identify opportunities for deep modules (small interface, deep implementation)
• Design interfaces for testability
• List the behaviors to test as plain-English descriptions, not implementation steps
• Get user approval on the plan before writing a line

You can't test everything. Confirm with the user exactly which behaviors matter most. Focus testing effort on critical paths and complex logic, not every possible edge case.

Existing code with no tests? Write characterization tests first — tests that encode the *current* behavior without changing it. These lock in the baseline and give you a safety net before any refactoring begins. Only add new behavior after the baseline is secured.

Ask: *"What should the public interface look like? Which behaviors are most important to test?"*

3. Tracer Bullet

Write ONE test that confirms ONE thing about the system:

RED:   Write test for first behavior → test fails
GREEN: Write minimal code to pass → test passes

The tracer bullet does more than verify one behavior — it proves the infrastructure works: the test runner finds tests, assertions resolve correctly, and the module under test can be imported. If it fails for non-logic reasons (import errors, configuration issues, missing test setup), you have a wiring problem to fix before any other tests will help. Isolating this risk first saves significant debugging time.

4. Incremental Loop

For each remaining behavior:

RED:   Write next test → fails
GREEN: Minimal code to pass → passes

Rules — and why each one matters:

• One test at a time — forces full understanding of one behavior before moving on; writing multiple tests at once leads to over-specifying the interface before you understand what it needs to be
• Only enough code to pass the current test — extra code is speculation about future requirements; without test pressure it becomes unverified complexity that accumulates into design debt
• Don't anticipate future tests — the interface should emerge from actual needs, not imagined ones; speculative design almost always has to be undone later
• Keep tests focused on observable behavior — tests tied to internal structure become a refactoring tax: every internal change requires updating tests even when no behavior changed

5. Refactor

After all tests pass, look for refactor candidates:

• Extract duplication
• Deepen modules (move complexity behind simple interfaces)
• Apply SOLID principles where natural
• Consider what the new code reveals about adjacent existing code
• Run tests after each refactor step — each step should stay green

Never refactor while RED. Get to GREEN first, then clean up.

Checklist Per Cycle

[ ] Test describes behavior, not implementation
[ ] Test uses public interface only
[ ] Test would survive an internal refactor
[ ] Code is minimal for this test
[ ] No speculative features added