Workshop: Dependency Inversion

Introduction

DIP workshop — concrete PostgresUserRepository inside UserService — introduce port, fake, wire DI.

Business problem

Business pressure: Google's core platform must ship weekly while keeping incident rates below SLO. Without Workshop: Dependency Inversion, teams accumulate coupling — every feature touches the same modules, regressions correlate, and senior hiring loops fail because candidates cannot articulate refactor trade-offs.

• Velocity: Median PR size grows; cross-team coordination dominates delivery time.
• Quality: Production incidents cluster around modules with multiple reasons to change.
• Talent: Staff promotion cases require evidence of principled architecture decisions — not buzzwords.

Architecture motivation

Why architects care: Workshop: Dependency Inversion addresses maintainability, change isolation, and testability. The naive alternative (God classes, concrete dependencies, duplicated business rules) works until team size and traffic make every change expensive and risky.

• Force: Requirements change along different axes — tax law, UI, persistence, notifications.
• Constraint: Cannot pause feature development for a multi-year rewrite.
• Outcome: Clear boundaries, faster tests, principled code reviews, and ADRs that survive turnover.

Real production story

During a production incident at Google, a checkout regression traced to a God module that violated Workshop: Dependency Inversion: one tax-rule change broke unrelated notification code. The post-incident refactor did not rewrite the system — it applied Workshop: Dependency Inversion incrementally with characterization tests. PR cycle time on that domain dropped 40% within two quarters, and code review debates shifted from style to named principles.

Enterprise case study

Google — Workshop: Dependency Inversion in production: Platform team inherited an 80k-line module where every feature violated basic separation. Incremental extractions behind interfaces, package acyclicity in CI, and principle-based review rubrics reduced change-failure rate 45% without a big-bang rewrite.

• Before: Correlated regressions; reviews debated formatting.
• Decision: Apply smallest refactor that improves testability and names the principle.
• After: Faster unit tests, smaller PRs, staff-ready interview narratives.

Refactoring walkthrough

Incremental refactor sequence for Workshop: Dependency Inversion — Martin Fowler's safe steps, not a rewrite:

• 1. Characterize: Add tests capturing current behavior before moving code.
• 2. Extract: Move the smallest cohesive unit; rename to reveal intent.
• 3. Invert: Introduce interface at the boundary already mocked in tests.
• 4. Wire: DI container or factory registers implementations.
• 5. Document: ADR notes rejected alternatives and YAGNI deferrals.

// Workshop: Dependency Inversion — illustrative refactor sketch
// BEFORE: OrderFacade validates, prices, persists, emails (4 reasons to change)
// AFTER:
//   OrderValidator      — validation only (SRP)
//   PricingStrategy     — discount rules (OCP)
//   OrderRepository     — persistence port (DIP)
//   NotificationPort    — email/SMS (ISP)
// Tests: same public API; smaller units; fakes at boundaries

Failure scenario

What breaks when Workshop: Dependency Inversion is ignored or misapplied:

• Big-bang rewrite trap: Team freezes features for 9 months; business ships competitor features first.
• Golden hammer: Strategy pattern for every if/else — YAGNI and readability suffer.
• False DRY: Forced abstraction across unrelated domains — wrong coupling worse than duplication.

Scalability analysis

Scale dimensions: Workshop: Dependency Inversion decisions compound across team count, codebase size, and deployment frequency — not just lines of code.

• Team scale: Conway's Law — module boundaries should align with ownership.
• Code scale: Package cycles block parallel development; enforce acyclic graphs in CI.
• Change scale: Feature flags + incremental extraction beat monolithic refactors.

Security considerations

Security is structural: Workshop: Dependency Inversion affects blast radius, secret handling, and auditability.

• Boundary leaks: God classes mix auth logic with export code — privilege escalation paths hide in tangled methods.
• Test doubles: DIP enables security unit tests without hitting real payment APIs in CI.
• Supply chain: Stable abstractions reduce direct imports of vendor SDKs across the codebase.

Architecture review questions

• What is the single reason to change for the module under review (Workshop: Dependency Inversion)?
• Which principle is violated — and what is the smallest safe refactor?
• Are tests sufficient to characterize behavior before extraction?
• What trade-off are we accepting (indirection vs testability)?
• Does this change align with package dependency rules and team ownership?
• Is there an ADR if this is an architectural boundary shift?

Staff engineer notes

• Principle-based reviews convert subjective taste into teachable decisions — your job is to make the trade-off legible.
• At Google scale, Workshop: Dependency Inversion fails at boundaries: interfaces, test fakes, and package imports — not diagram aesthetics.
• Good architecture is boring on the happy path and explicit on the failure path — if runbooks are fiction, the design is not production-ready.

Interview questions

Hands-on refactoring exercise

Refactor UserService to depend on UserRepository interface. Ship InMemoryUserRepository for tests.

// BEFORE
class UserService {
  private repo = new PostgresUserRepository();
}
// AFTER
class UserService {
  constructor(private repo: UserRepository) {}
}

In-depth explanation

Core idea: Workshop: Dependency Inversion guides how you structure code so that maintainability, change isolation, and testability. It is a force balancer — not a rule to maximize at all costs.

• Definition: Workshop: Dependency Inversion — durable guideline for structuring maintainable software.
• Violation signals: God classes, shotgun surgery, feature envy, divergent change.
• Refactor moves: Extract class, introduce interface, move method, replace conditional with polymorphism.
• When to relax: Early prototype, throwaway spike, or truly trivial script — optimize for learning speed.

Visual explanation

Three diagrams: principle architecture, refactor workflow, and common violation patterns:

Structural view

Workshop: Dependency Inversion — structural view:

Violation detected (Workshop: Dependency Inversion)
   ↓
Name principle + axis of change
   ↓
Incremental refactor (Fowler catalog)
   ↓
Tests + simpler dependency graph

Execution workflow

Informative example

Example — Workshop: Dependency Inversion:

// Review prompt for Workshop: Dependency Inversion
// 1. What reasons to change does this module have?
// 2. Which dependencies point the wrong direction?
// 3. Smallest extraction that improves testability?
// 4. What behavior must tests lock before refactor?

Real-world use

Uncle Bob's SOLID and Clean Architecture popularized dependency direction for enterprise systems. The Pragmatic Programmer's DRY, orthogonality, and tracer bullets shape daily craft. Martin Fowler's refactoring catalog turns principles into safe, incremental steps. Package design rules from Robert C. Martin and Rebecca Wirfs-Brock keep monorepos navigable. These ideas appear in Google's code review culture, Google readability standards, and Stripe's API design norms.

Trade-offs

• Pro: clearer change isolation, faster principled reviews, and interview-ready narratives.
• Con: more types and indirection — hurts readability if over-applied.
• Con: premature abstraction violates YAGNI — balance with KISS.

Best practices

• Map each review comment to a principle, not personal taste.
• Characterization tests before extracting legacy God classes.
• Enforce acyclic package dependencies in CI where possible.

Anti-patterns

• SOLID buzzword review — no concrete smell or proposed fix.
• Golden hammer — Strategy pattern for every if/else.
• Big-bang rewrite instead of incremental principle-driven extractions.

Common mistakes

• SRP taken as "one public method" — misses axis of change.
• DRY across unrelated domains — wrong abstraction worse than duplication.

Summary

Workshop: Dependency Inversion at staff level means naming smells, proposing minimal tested refactors, and documenting trade-offs — the skill Google senior engineers demonstrate in every architecture review.

Key takeaways

• Workshop: Dependency Inversion is a force balancer — apply with evidence, not dogma.
• Production success depends on incremental refactors, tests, and explicit trade-offs.
• Use this vocabulary in reviews and staff interviews with concrete smells and fixes.