project-test-loop

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Run automated TDD cycle: test → fix → refactor.

Note: Configure project-specific test/build commands in

CLAUDE.md

.claude/rules/

for automatic detection.

Steps:

Detect test command (if not already configured):
- Check
```
package.json
```
  for
```
scripts.test
```
  (Node.js)
- Check for
```
pytest
```
  or
```
python -m unittest
```
  (Python)
- Check for
```
cargo test
```
  (Rust)
- Check for
```
go test ./...
```
  (Go)
- Check
```
Makefile
```
  for test target
- If not found, ask user how to run tests

Run test suite:

bash

# Run detected test command
[test_command]

Analyze results:

If tests FAIL:
- Parse failure output
- Identify failing tests:
  - Which tests failed?
  - What assertions failed?
  - What was expected vs actual?
- Identify root cause:
  - Bug in implementation?
  - Missing implementation?
  - Incorrect test?
- Make minimal fix:
  - Fix only what's needed to pass the failing test
  - Don't add extra functionality
  - Don't fix tests (fix code instead)
- Re-run tests to confirm fix
- Loop back to step 2
If tests PASS:
- Check for refactoring opportunities:
  - Code duplication?
  - Unclear naming?
  - Long functions?
  - Complex logic that can be simplified?
  - Magic numbers/strings to extract?
- If refactoring identified:
  - Refactor while keeping tests green
  - Re-run tests after each refactoring
  - Ensure tests still pass
- If no refactoring needed:
  - Report success
  - Stop loop
Repeat until:
- All tests pass AND
- No obvious refactoring opportunities
OR stop if:
- User intervention needed
- Blocked by external dependency
- Unclear how to fix failure

Report results:

🧪 Test Loop Results:

Cycles: [N] iterations

Fixes Applied:
- [Fix 1]: [Brief description]
- [Fix 2]: [Brief description]

Refactorings Performed:
- [Refactor 1]: [Brief description]
- [Refactor 2]: [Brief description]

Current Status:
✅ All tests pass
✅ Code refactored
📝 Ready for commit

OR

⚠️ Blocked: [Reason]
📝 Next steps: [Recommendation]

TDD Cycle Details:

运行自动化TDD循环：测试→修复→重构。

注意：请在

CLAUDE.md

或

.claude/rules/

中配置项目特定的测试/构建命令，以便自动检测。

步骤:

检测测试命令（如果尚未配置）:
- 检查
```
package.json
```
  中的
```
scripts.test
```
  （Node.js）
- 检查是否存在
```
pytest
```
  或
```
python -m unittest
```
  （Python）
- 检查是否存在
```
cargo test
```
  （Rust）
- 检查是否存在
```
go test ./...
```
  （Go）
- 检查
```
Makefile
```
  中的测试目标
- 如果未找到，询问用户如何运行测试

运行测试套件:

bash

# Run detected test command
[test_command]

分析结果:

如果测试失败:
- 解析失败输出
- 识别失败的测试:
  - 哪些测试失败了?
  - 哪些断言失败了?
  - 预期结果与实际结果分别是什么?
- 识别根本原因:
  - 实现中存在Bug?
  - 缺少实现?
  - 测试用例不正确?
- 进行最小化修复:
  - 仅修复使失败测试通过所需的内容
  - 不要添加额外功能
  - 不要修改测试（而是修复代码）
- 重新运行测试以确认修复有效
- 回到步骤2循环
如果测试通过:
- 寻找重构机会:
  - 存在代码重复?
  - 命名不清晰?
  - 函数过长?
  - 复杂逻辑可简化?
  - 存在可提取的魔术数字/字符串?
- 如果识别出重构点:
  - 在保持测试通过的前提下进行重构
  - 每次重构后重新运行测试
  - 确保测试仍然通过
- 如果无需重构:
  - 报告成功
  - 终止循环
重复直到:
- 所有测试通过 AND
- 无明显重构机会
或在以下情况时终止:
- 需要用户干预
- 被外部依赖阻塞
- 不清楚如何修复失败问题

报告结果:

🧪 测试循环结果:

循环次数: [N] 次迭代

已应用的修复:
- [修复1]: [简要描述]
- [修复2]: [简要描述]

已执行的重构:
- [重构1]: [简要描述]
- [重构2]: [简要描述]

当前状态:
✅ 所有测试通过
✅ 代码已重构
📝 可提交

或

⚠️ 已阻塞: [原因]
📝 下一步: [建议]

TDD循环详情:

RED Phase (If starting new feature)

RED阶段（如果开始新功能）

Write failing test describing desired behavior
Run tests → Should FAIL (expected)
This command picks up from here

编写描述期望行为的失败测试
运行测试 → 应失败（符合预期）
本命令从此处开始执行

GREEN Phase (This command handles)

GREEN阶段（本命令处理）

Run tests
If fail → Make minimal fix
Re-run tests → Should PASS
Loop until all pass

运行测试
如果失败 → 进行最小化修复
重新运行测试 → 应通过
循环直到所有测试通过

REFACTOR Phase (This command handles)

REFACTOR阶段（本命令处理）

Tests pass
Look for improvements
Refactor
Re-run tests → Should STILL PASS
Loop until no improvements

Common Failure Patterns:

Pattern: Missing Implementation

Symptom:
```
undefined is not a function
```
,
```
NameError
```
, etc.
Fix: Implement the missing function/class/method
Minimal: Just the signature, return dummy value

Pattern: Wrong Return Value

Symptom:
```
Expected X but got Y
```
Fix: Update implementation to return correct value
Minimal: Don't add extra logic, just fix the return

Pattern: Missing Edge Case

Symptom: Test fails for specific input
Fix: Handle the edge case
Minimal: Add condition for this case only

Pattern: Integration Issue

Symptom: Test fails when components interact
Fix: Fix the integration point
Minimal: Fix just the integration, not entire components

Refactoring Opportunities:

Look for:

Duplicated code → Extract to function
Magic numbers → Extract to constants
Long functions → Break into smaller functions
Complex conditionals → Extract to well-named functions
Unclear names → Rename to be descriptive
Comments explaining code → Refactor code to be self-explanatory

Don't:

Change behavior
Add new functionality
Skip test runs
Make tests pass by changing tests

Auto-Stop Conditions:

Stop and report if:

All tests pass + no refactoring needed (SUCCESS)
Same test fails 3 times in a row (STUCK)
Error in test command itself (TEST SETUP ISSUE)
External dependency unavailable (BLOCKED)
Unclear how to fix (NEEDS USER INPUT)

Integration with Blueprint Development:

This command applies project-specific skills:

Testing strategies: Knows how to structure tests
Implementation guides: Knows how to implement fixes
Quality standards: Knows what to refactor
Architecture patterns: Knows where code should go

测试通过
寻找改进点
进行重构
重新运行测试 → 应仍然通过
循环直到无改进空间

常见失败模式:

模式: 缺少实现

症状:
```
undefined is not a function
```
,
```
NameError
```
, etc.
修复: 实现缺失的函数/类/方法
最小化原则: 仅实现签名，返回占位值

模式: 返回值错误

症状:
```
Expected X but got Y
```
修复: 更新实现以返回正确值
最小化原则: 不要添加额外逻辑，仅修复返回值

模式: 缺少边界情况处理

症状: 特定输入导致测试失败
修复: 处理该边界情况
最小化原则: 仅添加针对该情况的条件

模式: 集成问题

症状: 组件交互时测试失败
修复: 修复集成点
最小化原则: 仅修复集成问题，而非整个组件

重构机会:

寻找以下情况:

重复代码 → 提取为函数
魔术数字 → 提取为常量
过长函数 → 拆分为更小的函数
复杂条件 → 提取为命名清晰的函数
不清晰的命名 → 重命名为描述性名称
用于解释代码的注释 → 重构代码使其自解释

请勿:

改变行为
添加新功能
跳过测试运行
通过修改测试使测试通过

自动终止条件:

在以下情况时终止并报告:

所有测试通过 + 无需重构（成功）
同一测试连续失败3次（卡住）
测试命令本身出错（测试设置问题）
外部依赖不可用（已阻塞）
不清楚如何修复（需要用户输入）

与蓝图开发的集成:

本命令应用项目特定技能:

测试策略: 了解如何构建测试结构
实现指南: 了解如何实施修复
质量标准: 了解需要重构的内容
架构模式: 了解代码应放置的位置