Singleton Design Pattern in Java: Part 2

Recap from Part 1

In Part 1, we covered:

• The concept of the Singleton design pattern
• Eager initialization approaches
• Basic lazy loading (not thread-safe)
• Synchronized method (thread-safe but slow)

Now let’s explore more efficient thread-safe implementations!

Double-Checked Locking Singleton

This pattern is explained in the Wikipedia article on Double-checked locking.

public class DoubleCheckedLockingSingleton {

    private static volatile DoubleCheckedLockingSingleton instance;

    private DoubleCheckedLockingSingleton() {}

    public static DoubleCheckedLockingSingleton getInstance() {
        if (instance == null) {                              // First check (no locking)
            synchronized (DoubleCheckedLockingSingleton.class) {
                if (instance == null) {                      // Second check (with locking)
                    instance = new DoubleCheckedLockingSingleton();
                }
            }
        }
        return instance;
    }
}

How It Works

Instead of synchronizing the entire method, we only synchronize the instantiation block:

1. First check - Quick check without synchronization (fast path for already-initialized case)
2. Synchronized block - Only entered if instance might be null
3. Second check - Prevents multiple instantiations when two threads pass the first check simultaneously

Why Two Null Checks?

Consider this scenario:

Thread T1: Passes first null check
Thread T2: Passes first null check
Thread T1: Enters synchronized block, creates instance
Thread T2: Enters synchronized block...
         → Without second check, T2 would create another instance!

The Critical Role of volatile

Without volatile, this pattern is broken!

Here’s why:

1. The compiler/JIT can reorder instructions
2. Object construction involves multiple steps:
   • Allocate memory
   • Initialize fields
   • Assign reference to variable

Without volatile, the reference might be assigned before initialization completes:

Thread T1: Allocates memory, assigns to instance (partially constructed!)
Thread T2: Sees instance != null, returns partially constructed object
         → CRASH or undefined behavior!

The volatile keyword prevents this reordering, ensuring the object is fully constructed before the reference is visible to other threads.

For more details, see The “Double-Checked Locking is Broken” Declaration.

Bill Pugh Singleton (Initialization-on-demand Holder)

Bill Pugh’s solution is the most elegant and efficient approach. It uses a static inner class to achieve lazy initialization.

public class BillPughSingleton {

    private BillPughSingleton() {}

    private static class SingletonHolder {
        private static final BillPughSingleton INSTANCE = new BillPughSingleton();
    }

    public static BillPughSingleton getInstance() {
        return SingletonHolder.INSTANCE;
    }
}

Why This Works

This pattern leverages the JVM’s class loading mechanism:

1. The SingletonHolder class is not loaded until getInstance() is called - Classes in Java are loaded on first use
2. Class loading is thread-safe - The JVM guarantees that class initialization is atomic
3. No synchronization overhead - Subsequent calls don’t require any locking

Benefits

• Safe: Thread-safe without explicit synchronization
• Efficient: No synchronization overhead on reads
• Lazy: Instance created only when first accessed
• Simple: No volatile, no synchronized blocks

This is considered the best approach for implementing singletons in Java.

Performance Comparison

Approach	First Access	Subsequent Access	Complexity
Synchronized Method	Slow	Slow (always locks)	Simple
Double-Checked Locking	Medium	Fast	Medium
Bill Pugh	Fast	Fast	Simple

Enum Singleton (Bonus)

Joshua Bloch recommends using an enum for singletons:

public enum EnumSingleton {
    INSTANCE;

    public void doSomething() {
        System.out.println("Singleton operation");
    }
}

Advantages:

• Thread-safe by default
• Serialization handled automatically
• Reflection-proof
• Simplest implementation

Disadvantage:

• Cannot extend other classes (enums implicitly extend Enum)

Summary

Implementation	Thread-Safe	Lazy	Performance	Recommended
Eager	Yes	No	Good	Simple cases
Synchronized	Yes	Yes	Poor	Avoid
Double-Checked	Yes	Yes	Good	Legacy code
Bill Pugh	Yes	Yes	Best	General use
Enum	Yes	No	Good	When possible

Conclusion

This two-part series covered the Singleton design pattern comprehensively:

• What is the singleton pattern
• Where it’s used (Spring, Runtime, etc.)
• When to use it
• How to implement it correctly

Best Practices:

• Use the Bill Pugh pattern for most cases
• Consider enums when you don’t need lazy initialization
• Avoid the synchronized method approach
• Always make the constructor private
• Be careful with serialization (can break singleton)

Thanks for reading! See you in the next post.