test-design-techniques

仓库: proffesor-for-testing/agentic-qe
安装量: 58
排名: #12906

安装

npx skills add https://github.com/proffesor-for-testing/agentic-qe --skill test-design-techniques

When designing test cases systematically:

  • APPLY Boundary Value Analysis (test at min, max, edges)

  • USE Equivalence Partitioning (one test per partition)

  • CREATE Decision Tables (for complex business rules)

  • MODEL State Transitions (for stateful behavior)

  • REDUCE with Pairwise Testing (for combinations)

Quick Design Selection:

  • Numeric ranges → BVA + EP

  • Multiple conditions → Decision Tables

  • Workflows → State Transition

  • Many parameters → Pairwise Testing

Critical Success Factors:

  • Systematic design finds more bugs with fewer tests

  • Random testing is inefficient

  • 40+ years of research backs these techniques

Quick Reference Card

When to Use

  • Designing new test suites

  • Optimizing existing tests

  • Complex business rules

  • Reducing test redundancy

Technique Selection Guide

| Numeric input ranges | BVA + EP

| Multiple conditions | Decision Tables

| Stateful workflows | State Transition

| Many parameter combinations | Pairwise

| All combinations critical | Full Factorial

Boundary Value Analysis (BVA)

Principle: Bugs cluster at boundaries.

Test at boundaries:

  • Minimum valid value

  • Just below minimum (invalid)

  • Just above minimum (valid)

  • Maximum valid value

  • Just above maximum (invalid)

// Age field: 18-120 valid
const boundaryTests = [
  { input: 17, expected: 'invalid' },  // Below min
  { input: 18, expected: 'valid' },    // Min boundary
  { input: 19, expected: 'valid' },    // Above min
  { input: 119, expected: 'valid' },   // Below max
  { input: 120, expected: 'valid' },   // Max boundary
  { input: 121, expected: 'invalid' }  // Above max
];

Equivalence Partitioning (EP)

Principle: One test per equivalent class.

// Discount rules:
// 1-10:  No discount
// 11-100: 10% discount
// 101+:   20% discount

const partitionTests = [
  { quantity: -1, expected: 'invalid' },  // Invalid partition
  { quantity: 5, expected: 0 },           // Partition 1: 1-10
  { quantity: 50, expected: 0.10 },       // Partition 2: 11-100
  { quantity: 200, expected: 0.20 }       // Partition 3: 101+
];

// 4 tests cover all behavior (vs 200+ if testing every value)

Decision Tables

Use for: Complex business rules with multiple conditions.

Loan Approval Rules:
┌──────────────┬───────┬───────┬───────┬───────┬───────┐
│ Conditions   │ R1    │ R2    │ R3    │ R4    │ R5    │
├──────────────┼───────┼───────┼───────┼───────┼───────┤
│ Age ≥ 18     │ Yes   │ Yes   │ Yes   │ No    │ Yes   │
│ Credit ≥ 700 │ Yes   │ Yes   │ No    │ Yes   │ No    │
│ Income ≥ 50k │ Yes   │ No    │ Yes   │ Yes   │ Yes   │
├──────────────┼───────┼───────┼───────┼───────┼───────┤
│ Result       │Approve│Approve│Reject │Reject │Reject │
└──────────────┴───────┴───────┴───────┴───────┴───────┘

// 5 tests cover all decision combinations

State Transition Testing

Model state changes:

States: Logged Out → Logged In → Premium → Suspended

Valid Transitions:
- Login: Logged Out → Logged In
- Upgrade: Logged In → Premium
- Payment Fail: Premium → Suspended
- Logout: Any → Logged Out

Invalid Transitions to Test:
- Logged Out → Premium (should reject)
- Suspended → Premium (should reject)

test('cannot upgrade without login', async () => {
  const result = await user.upgrade(); // While logged out
  expect(result.error).toBe('Login required');
});

Pairwise (Combinatorial) Testing

Problem: All combinations explode exponentially.

// Parameters:
// Browser: Chrome, Firefox, Safari (3)
// OS: Windows, Mac, Linux (3)
// Screen: Desktop, Tablet, Mobile (3)

// All combinations: 3 × 3 × 3 = 27 tests
// Pairwise: 9 tests cover all pairs

const pairwiseTests = [
  { browser: 'Chrome', os: 'Windows', screen: 'Desktop' },
  { browser: 'Chrome', os: 'Mac', screen: 'Tablet' },
  { browser: 'Chrome', os: 'Linux', screen: 'Mobile' },
  { browser: 'Firefox', os: 'Windows', screen: 'Tablet' },
  { browser: 'Firefox', os: 'Mac', screen: 'Mobile' },
  { browser: 'Firefox', os: 'Linux', screen: 'Desktop' },
  { browser: 'Safari', os: 'Windows', screen: 'Mobile' },
  { browser: 'Safari', os: 'Mac', screen: 'Desktop' },
  { browser: 'Safari', os: 'Linux', screen: 'Tablet' }
];
// Each pair appears at least once

Agent-Driven Test Design

// Auto-generate BVA tests
await Task("Generate BVA Tests", {
  field: 'age',
  dataType: 'integer',
  constraints: { min: 18, max: 120 }
}, "qe-test-generator");
// Returns: 6 boundary test cases

// Auto-generate pairwise tests
await Task("Generate Pairwise Tests", {
  parameters: {
    browser: ['Chrome', 'Firefox', 'Safari'],
    os: ['Windows', 'Mac', 'Linux'],
    screen: ['Desktop', 'Tablet', 'Mobile']
  }
}, "qe-test-generator");
// Returns: 9-12 tests (vs 27 full combination)

Agent Coordination Hints

Memory Namespace

aqe/test-design/
├── bva-analysis/*       - Boundary value tests
├── partitions/*         - Equivalence partitions
├── decision-tables/*    - Decision table tests
└── pairwise/*           - Combinatorial reduction

Fleet Coordination

const designFleet = await FleetManager.coordinate({
  strategy: 'systematic-test-design',
  agents: [
    'qe-test-generator',    // Apply design techniques
    'qe-coverage-analyzer', // Analyze coverage
    'qe-quality-analyzer'   // Assess test quality
  ],
  topology: 'sequential'
});

Remember

Systematic design > Random testing. 40+ years of research shows these techniques find more bugs with fewer tests than ad-hoc approaches.

Combine techniques for comprehensive coverage. BVA for boundaries, EP for partitions, decision tables for rules, pairwise for combinations.

With Agents: qe-test-generator applies these techniques automatically, generating optimal test suites with maximum coverage and minimum redundancy. Agents identify boundaries, partitions, and combinations from code analysis.

← 返回排行榜