Design Patterns (GoF)¶

★★★★★ Intermediate

Classic design patterns from the Gang of Four. A pattern describes a general concept for solving a common problem - use only when necessary, unnecessary patterns add complexity.

SOLID Principles¶

Principle	Rule	Violation Sign
Single Responsibility	One class = one reason to change	Class does logging AND business logic
Open/Closed	Open for extension, closed for modification	Modifying existing class for new behavior
Liskov Substitution	Subtypes substitutable for base types	Override breaks base class contract
Interface Segregation	Many specific interfaces > one general	Implementing unused methods
Dependency Inversion	Depend on abstractions, not implementations	Direct instantiation of dependencies

Meta-principles: favor composition over inheritance; program to interfaces, not implementations.

Creational Patterns¶

Pattern	Purpose	When to Use
Factory Method	Interface for creating objects, subclasses decide type	Don't know types in advance, need extension point
Abstract Factory	Create families of related objects	Multiple families of products
Builder	Construct complex objects step by step	Many optional parameters, different representations
Prototype	Create by cloning existing objects	Expensive construction, avoid subclassing
Singleton	Exactly one instance, global access	DB pool, config. Warning: violates SRP, hides dependencies

Structural Patterns¶

Pattern	Purpose	When to Use
Adapter	Convert incompatible interface	Legacy/third-party integration
Bridge	Separate abstraction from implementation	Multiple orthogonal dimensions, runtime switch
Composite	Tree structures, uniform treatment	Part-whole hierarchies (file system, UI, org chart)
Decorator	Attach behavior dynamically via wrapping	Add responsibilities without subclassing
Facade	Simplified interface to complex subsystem	Reduce coupling, convenient defaults
Flyweight	Share common state between many objects	Huge number of similar objects eating memory
Proxy	Surrogate controlling access	Lazy loading, access control, caching, logging

Behavioral Patterns¶

Pattern	Purpose	When to Use
Chain of Responsibility	Pass request along handler chain	Multiple handlers, varying order, runtime chain
Command	Encapsulate request as object	Undo, queueing, logging of operations
Iterator	Traverse collection without exposing internals	Complex structures (tree, graph) needing sequential access
Mediator	Centralize communication	Many interdependencies between components
Memento	Capture/restore state without violating encapsulation	Undo/redo, snapshots (editor history, game saves)
Observer	Subscription-based event notification	Changes in one object require updating others
State	Behavior changes based on internal state	Many conditionals based on state
Strategy	Interchangeable algorithms	Switch algorithms at runtime, many variants
Template Method	Algorithm skeleton, subclasses override steps	Shared structure, different specific steps
Visitor	New operations without modifying classes	Operations change frequently, structure is stable

Pattern Selection Guide¶

Creating objects:
  One type           -> Factory Method
  Family of types    -> Abstract Factory
  Complex build      -> Builder
  Clone existing     -> Prototype

Structuring:
  Incompatible APIs  -> Adapter
  Dynamic behavior   -> Decorator
  Tree structures    -> Composite
  Simplify complex   -> Facade

Communication:
  Event notification -> Observer
  Centralized comms  -> Mediator
  Sequential handling -> Chain of Responsibility
  Encapsulated ops   -> Command

Algorithm variation:
  Interchangeable    -> Strategy
  Skeleton + hooks   -> Template Method
  State-dependent    -> State

OOP Relationships¶

Relationship	Meaning	Strength
Dependency	Temporarily uses	Weakest
Association	Uses (has reference)	Weak
Aggregation	Has, can exist independently	Medium
Composition	Owns, lifecycle dependent	Strong
Inheritance	Is-a	Strongest
Implementation	Realizes interface	-

Gotchas¶

Singleton is the most misused pattern - often used where simple dependency injection suffices
Pattern overuse adds complexity. If code works cleanly without a pattern, don't force one
Decorator vs Inheritance - decorator is more flexible but creates many small objects. Inheritance is simpler when hierarchy is stable
Observer memory leaks - forgotten subscriptions keep objects alive. Always unsubscribe
Strategy vs State - Strategy client chooses algorithm; State transitions happen internally