Chicago TDD Pattern Skill

Core Philosophy

Chicago TDD = State-based testing with real collaborators

Verify observable behavior changes:

• Return values
• State mutations
• Side effects
• Actual system effects

NOT internal implementation, method calls, or mocks (except London TDD in tests).

AAA Pattern (Arrange-Act-Assert)

#[test]
fn test_feature() {
    // ARRANGE: Set up real objects (no mocks)
    let manager = LockfileManager::new(temp_dir.path());

    // ACT: Call the public API being tested
    manager.upsert("pkg", "1.0.0", "sha256", "url").unwrap();

    // ASSERT: Verify observable state changed
    let entry = manager.get("pkg").unwrap().unwrap();
    assert_eq!(entry.version, "1.0.0");  // State changed ✓
}

Test Organization

Unit Tests (Fast, Focused)

Location:

src/module.rs

with

#[cfg(test)]

module

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_cache_insertion() {
        // Arrange
        let mut cache = Cache::new();

        // Act
        cache.insert("key", "value");

        // Assert
        assert_eq!(cache.get("key"), Some("value"));
    }
}

Focus: Single responsibility, fast execution, deterministic results

Integration Tests (End-to-End)

Location:

crates/*/tests/

directory

// tests/integration_test.rs
#[test]
fn test_full_pipeline() {
    let config = Config::load("test.toml").unwrap();
    let result = generate(&config).unwrap();

    assert!(result.contains("generated code"));
    assert!(Path::new("output.rs").exists());
}

Focus: Real filesystem, file I/O, full workflows

E2E Tests (System-Level)

Location:

crates/ggen-e2e/tests/

with testcontainers

#[tokio::test]
async fn test_with_docker() {
    let container = RustContainer::default().start().await.unwrap();
    let output = container.exec("cargo make test").await.unwrap();

    assert!(output.contains("test result: ok"));
}

Focus: Real containers, actual system behavior, byte-for-byte validation

Test Quality Metrics

Mutation Score (Target: > 90%)

Tests should catch code mutations (changes):

cargo make test-audit mutations

# Mutation: Change == to !=
let x = 5;
if x == 5 { }  →  if x != 5 { }

# Good test catches this:
assert_eq!(result, 5);  // Would fail if == changed to !=

Surviving mutations indicate weak tests - add more assertions.

Assertion Density (Target: > 1 per function)

Count assertions per function:

#[test]
fn test_parsing() {
    // Arrange
    let input = "key: value";

    // Act
    let result = parse(input).unwrap();

    // Assert: Multiple assertions per test ✓
    assert_eq!(result.key, "key");      // Assertion 1
    assert_eq!(result.value, "value");  // Assertion 2
    assert!(result.valid());             // Assertion 3
}

Low assertion density = incomplete testing.

False Positives (Target: Zero)

Tests passing when implementation broken:

// ❌ BAD: Test passes even if add() is broken
#[test]
fn test_add() {
    let result = add(2, 3);
    // Missing assertion!
}

// ✓ GOOD: Test actually verifies behavior
#[test]
fn test_add() {
    let result = add(2, 3);
    assert_eq!(result, 5);  // Catches bugs ✓
}

Exemptions: unwrap/expect in Tests

ALLOWED in test/benchmark code:

#[test]
fn test_something() {
    let obj = Object::new().unwrap();  // ✓ ALLOWED
    assert_eq!(obj.value(), 42);
}

#[cfg(test)]
mod tests {
    #[test]
    fn test_error() {
        let result = parse("invalid");
        assert!(result.is_err());  // ✓ ALLOWED
    }
}

#[bench]
fn bench_iteration(b: &mut Bencher) {
    let data = setup().unwrap();  // ✓ ALLOWED in benches
    b.iter(|| process(&data))
}

Rationale: Tests should fail fast on setup errors. Don't hide test issues.

Test Organization Best Practices

1. One Assertion Per Concept

// ✓ Good: Each assertion tests one thing
#[test]
fn test_user_creation() {
    let user = User::new("Alice", "alice@example.com").unwrap();
    assert_eq!(user.name, "Alice");         // Tests name
    assert_eq!(user.email, "alice@example.com");  // Tests email
    assert!(user.valid());                  // Tests validity
}

// ❌ Bad: Combined assertions hide failures
#[test]
fn test_user_creation() {
    let user = User::new("Alice", "alice@example.com").unwrap();
    assert!(user.name == "Alice" && user.email == "alice@example.com");
    // If either fails, both fail
}

2. Test One Error Path Per Test

// ✓ Good: Separate tests for each error
#[test]
fn test_parse_empty_input_error() {
    let result = parse("");
    assert!(result.is_err());
    assert_eq!(result.unwrap_err(), Error::EmptyInput);
}

#[test]
fn test_parse_invalid_syntax_error() {
    let result = parse("invalid {syntax");
    assert!(result.is_err());
    assert_eq!(result.unwrap_err(), Error::SyntaxError);
}

// ❌ Bad: Multiple error paths in one test
#[test]
fn test_parse_errors() {
    // This test does too much
}

3. Use Descriptive Test Names

// ✓ Good: Name tells you what is tested
#[test]
fn test_cache_returns_most_recent_value() { }

#[test]
fn test_cache_evicts_lru_on_full() { }

#[test]
fn test_cache_handles_concurrent_access() { }

// ❌ Bad: Unclear what's being tested
#[test]
fn test_cache_1() { }

#[test]
fn test_cache() { }

Benchmarking with Criterion

use criterion::{black_box, criterion_group, criterion_main, Criterion};

fn bench_generation(c: &mut Criterion) {
    c.bench_function("generate_small", |b| {
        b.iter(|| generate(black_box("small_input")))
    });

    c.bench_function("generate_large", |b| {
        b.iter(|| generate(black_box("large_input")))
    });
}

criterion_group!(benches, bench_generation);
criterion_main!(benches);

Run benchmarks:

cargo make bench              # Run all benchmarks
cargo make bench-compare      # Compare versions

Property-Based Testing (proptest)

use proptest::prelude::*;

proptest! {
    #[test]
    fn test_parse_roundtrip(s in r#"[a-z]+""#) {
        let parsed = parse(&s).unwrap();
        let serialized = serialize(&parsed);
        prop_assert_eq!(s, serialized);
    }
}

Advantage: Tests many inputs automatically, catches edge cases.

Test Frameworks Used

Framework	Purpose	Use When
chicago-tdd-tools 1.4.0	AAA pattern	Writing unit tests
proptest 1.8	Property-based	Testing invariants
criterion 0.7	Benchmarking	Performance tracking
testcontainers 0.25	E2E with docker	Integration testing
assert_cmd 2.0	CLI testing	Command validation
assert_fs 1.1	Filesystem testing	File operations
mockall 0.13	London TDD mocking	Test-only mocking
insta 1.43	Snapshot testing	Golden files

Running Tests

# Quick unit tests only
cargo make test-unit        # ~10s

# All tests
cargo make test             # ~30s (120s escalation)

# Specific test
cargo test test_name        # Single test

# With output
cargo test -- --nocapture  # See println! output

# Benchmarks
cargo make bench            # All benchmarks
cargo bench bench_name      # Specific benchmark

# Mutation testing
cargo make test-audit mutations

Success Criteria

✓ All tests pass ✓ Mutation score > 90% ✓ Assertion density > 1 per function ✓ No false positives ✓ Error paths tested ✓ Chicago TDD pattern used ✓ No flaky tests ✓ SLO timeouts met

Key Principles

1. Real Objects: Use actual implementations, not mocks
2. Observable State: Test what you can see, not internals
3. Comprehensive: Test happy path, errors, edge cases
4. Deterministic: Same inputs = same results, always
5. Fast: Unit tests should run instantly
6. Focused: Each test has one purpose
7. Mutation-Resistant: Tests catch code changes

Common Testing Mistakes

❌ Testing Internals

// WRONG: Testing internal detail
#[test]
fn test_cache_internal_structure() {
    let cache = Cache::new();
    assert_eq!(cache.data.capacity(), 100);  // Wrong!
}

// CORRECT: Test observable behavior
#[test]
fn test_cache_stores_and_retrieves() {
    let cache = Cache::new();
    cache.insert("key", "value");
    assert_eq!(cache.get("key"), Some("value"));
}

❌ Using Mocks

// WRONG: Mocks hide real behavior
#[test]
fn test_with_mock() {
    let mut mock = MockService::new();
    mock.expect_call().returning(|| 42);
    // Mock doesn't test real behavior!
}

// CORRECT: Use real objects
#[test]
fn test_with_real_service() {
    let service = RealService::new();
    let result = service.call();
    assert_eq!(result, 42);  // Tests real behavior ✓
}

❌ No Assertions

// WRONG: Test runs but doesn't verify
#[test]
fn test_something() {
    let result = do_something();  // Missing assertion!
}

// CORRECT: Assert expected behavior
#[test]
fn test_something() {
    let result = do_something();
    assert_eq!(result, expected);  // Verifies behavior ✓
}

See Also

• reference.md

  - Detailed testing patterns

• examples.md

  - Real-world test examples
• Chicago TDD documentation
• criterion.rs benchmarking guide