Creational Design Patterns

Section 1 — Notebook Summary · hand-copy this

Section 2 — Detailed Study Reference

§0What a design pattern actually is

A design pattern is a tried-and-true, reusable solution to a problem that keeps showing up in software design.

Think of it as a cookbook for developers. You don't copy-paste a pattern the way you copy a library; you follow its recipe and adapt it to your kitchen. It's a template — a shared vocabulary — not a chunk of code. When one engineer says "just use a Builder here," a whole design is communicated in three words.

Patterns sort into three families based on what they organize:

Everything in this lesson lives in the Creational box. And every one of them is really an application of a single principle, so let's meet that first.

The principle beneath everything: Encapsulate What Varies

Find the part of your code that changes, and separate it from the part that stays the same.

Analogy — the restaurant kitchen. The dining room never changes: tables, waiters, the act of ordering. What varies is the menu. A well-run restaurant keeps the menu on a separate card, not carved into the tables. Swap the card, keep the room. In code, you push the volatile part (which compiler? which fields? how many instances?) behind a stable interface so the rest of your program never has to be edited when it changes.

§1Factory Method

Create an object without forcing the caller to name the exact class being created.

Analogy — ordering coffee. You walk up to the counter and say "a latte, please." You don't reach behind the bar, grab the specific machine, and operate it yourself. The barista (the factory) hears your request and produces the right drink. You care about the result — a latte — not about which machine made it. Factory Method puts a barista between your code and the messy business of construction.

The scenario that motivates it

Design a compiler for LeetCode that handles Java, C++, and Python. The public API is fixed:

class LeetcodeCompiler {
    public void compile(String code, String language) { }
}

The naïve instinct is to stuff a giant if/else or switch right inside compile(), mixing "figure out which compiler" with "actually run the compiler." Every new language then means editing this one method — which does two unrelated jobs at once. The language choice is what varies. Encapsulate it.

The solution: pull creation into a Factory

// Product interface — the stable contract
public interface Compilation {
    void compile(String code);
}

// The factory — the ONLY place that knows concrete classes
class CompilationFactory {
    public static Compilation getInstance(String language) {
        Compilation op = null;
        switch (language) {
            case "JAVA":   op = new JavaCompiler();   break;
            case "PYTHON": op = new PythonCompiler(); break;
            case "CPP":    op = new CPPCompiler();    break;
        }
        return op;
    }
}

// Concrete products
class JavaCompiler implements Compilation {
    public void compile(String code) { /* ... */ }
}
// CppCompiler, PythonCompiler ... same shape

// The client — blissfully ignorant of concrete classes
class LeetCodeCompiler {
    public void compile(String code, String language) {
        Compilation instance = CompilationFactory.getInstance(language);
        instance.compile(code);
    }
}

What this buys you

• Effortless creation of the right compiler object from a simple string.
• Total abstraction — the client never imports or names JavaCompiler.
• Loose coupling through the interface: swap implementations freely.
• Scalable — adding a language touches the factory, not the client.

§2Builder

Construct a complex object step by step, keeping it readable and immutable, instead of drowning in constructor arguments.

Analogy — the sandwich shop. At Subway you don't shout your entire order in one breath. You walk the line: pick the bread, add cheese, add veggies, skip the onions, choose a sauce. Each choice is a small step, in any order, and only the ones you want. At the end, it's wrapped — sealed, done, unchangeable. That's a Builder: a construction line for objects.

The scenario: a Student Enrollment Form

The form has two mandatory fields (first name, last name) and many optional ones (age, gender, isGraduate, hasExperience, city, state, isEarning…). Once filled, details cannot be changed — so every field is private final (that's what makes the object immutable). Now: how do we let callers set only the fields they care about, while keeping immutability? We try three approaches; watch two of them fail.

Failed attempt #1 — Telescoping constructors

Write one constructor per combination, each calling the next:

public EnrollmentForm(String firstName, String lastName) {
    this(firstName, lastName, 0);
}
public EnrollmentForm(String firstName, String lastName, int age) {
    this(firstName, lastName, age, null);
}
// ... and on, and on ...

This works for a handful of fields, but it can't express "first name, last name, and gender" without yet another constructor — and with 10 optional fields you'd need an unmanageable number of overloads. Not scalable, and unreadable at the call site.

Failed attempt #2 — JavaBean (setters)

Use a small constructor for mandatory fields, then setters for the rest:

public void setAge(int age) { this.age = age; }
public void setGender(String gender) { this.gender = gender; }

Flexible, yes — but a setter, by definition, lets you change a field after creation. That destroys immutability: someone can call form.setGender(...) tomorrow. Requirement violated.

The solution — a static inner Builder class

public class EnrollmentForm {
    private final String firstName;
    private final String lastName;
    private final int age;
    private final String gender;

    // private: only the Builder may construct us
    private EnrollmentForm(Builder builder) {
        this.firstName = builder.firstName;
        this.lastName  = builder.lastName;
        this.age       = builder.age;
        this.gender    = builder.gender;
    }

    public static class Builder {
        private final String firstName;  // mandatory
        private final String lastName;
        private int age;                  // optional
        private String gender;

        public Builder(String firstName, String lastName) {
            this.firstName = firstName;
            this.lastName = lastName;
        }
        public Builder age(int age)     { this.age = age;    return this; }
        public Builder gender(String g)  { this.gender = g;  return this; }

        public EnrollmentForm build() {
            return new EnrollmentForm(this);
        }
    }
}

And the call site reads like a sentence:

EnrollmentForm form = new EnrollmentForm.Builder("John", "Doe").age(25).gender("Male").build();
// EnrollmentForm{firstName='John', lastName='Doe', age=25, gender='Male'}

Run through the requirements and it passes all three: create with just the two names? Yes. Create with names + gender only? Yes. Change gender afterward? No — there's no setter on the finished object, only on the builder. That last "no" is the win.

When to reach for Builder

Rule of thumb: reach for Builder when an object needs more than 4–5 fields to construct. Its three advantages are maintainability (readable at the call site), robustness (the client only ever receives a fully-constructed object), and immutability.

@Getter @Builder
public class EnrollmentForm {
    private final String firstName;
    private final String lastName;
    private final int age;
    // ... rest of the fields, no boilerplate
}

§3Singleton

Guarantee a class has exactly one instance, and give the whole program a single shared access point to it.

Analogy — the country's president. No matter who asks, at any given moment there is exactly one sitting president, and everyone refers to the same one. You can't spin up a second president for convenience. A Singleton is that: one instance, globally reachable, no duplicates allowed.

The scenario: a Logger

A logger records messages to a log file. The catch: the whole application must write to a single log file. If two logger objects each open the file, you get access conflicts and inconsistent output. So we need to forbid creating more than one Logger. How?

Building it up, one obstacle at a time

We derive the pattern by hitting walls:

• Problem: as long as the class has a public constructor, anyone can write new Logger() as many times as they like. Fix: make the constructor private.
• New problem: now nobody can create even one instance. Fix: a private constructor is still callable from inside the class — so add a getInstance() method that lives inside the class and does the creating.
• New problem: we can't call getInstance() without an object, which we can't make. Fix: make getInstance() static so it belongs to the class, not an instance.
• New problem: callers can still invoke getInstance() repeatedly and get multiple instances. Fix: keep a static reference; only create if it's null, otherwise return the one that exists.

public class Logger {
    private static Logger instance;    // the one and only

    private Logger() { }                 // 1. no public construction

    public static Logger getInstance() { // 2. static access point
        if (instance == null) {         // 3. create once, reuse forever
            instance = new Logger();
        }
        return instance;
    }
}

The recipe in three steps

1. Make the constructor private.
2. Create a static method to hand out the instance.
3. Inside it, check if the instance already exists — return it if so, otherwise create a new one.

Why Singleton is worth it (for a Logger)

• Centralized control over a shared resource or configuration.
• Resource management — one file handle / connection, created and cleaned up once, no conflicts.
• Consistency — every part of the app logs with the same config and log level, so output is uniform.
• Lower memory — one object with one set of handlers, shared everywhere, instead of many.

How the three fit together

All three are creational, all three are "Encapsulate What Varies" wearing different clothes — but they answer different questions:

Interview & Practice Q&A

Tap a question to reveal a 2–4 sentence answer you can read aloud.

Key Takeaways