Pattern matching is a major feature in software development. While pattern matching applies in several locations, its current usage is limited to switch case blocks. I want to compare the power of pattern matching across a couple of programming languages I'm familiar with in this post. switch case I assume that every reader is familiar with the switch case syntax inherited from C. In short: switch case The switch clause references a value-returning statement.
Each case clause sets another statement; if the value matches the statement, it executes the related block.
case clauses are evaluated in order. The first clause that matches gets its block executed.
.In C, case clauses are fall-through; you need to explicitly break to escape the switch, otherwise, the next case is evaluated The switch clause references a value-returning statement. switch Each case clause sets another statement; if the value matches the statement, it executes the related block. case case clauses are evaluated in order. The first clause that matches gets its block executed. case .In C, case clauses are fall-through; you need to explicitly break to escape the switch, otherwise, the next case is evaluated case break switch case Java's Pattern Matching I'll start with Java, as it was the first programming language I used in a professional context. Java's pattern matching has evolved a lot across its versions. Here's the "official" sample for version 23, slightly amended. It defines several Shape implementations and a static method to evaluate its perimeter. It's definitely not a good example of Object-Oriented Programming. Shape static not interface Shape { }
record Rectangle(double length, double width) implements Shape { }
record Circle(double radius) implements Shape { }

public class Main {
    static double getPerimeter(Shape s) throws IllegalArgumentException {
        return switch (s) {                                                 //1
            case Rectangle r when r.length() == r.width() -> {              //2
                System.out.println("Square detected");
                    yield 4 * r.length();
            }
            case Rectangle r ->                                             //3
                2 * r.length() + 2 * r.width();
            case Circle c ->                                                //4
                2 * c.radius() * Math.PI;
            default ->                                                      //5
                throw new IllegalArgumentException("Unrecognized shape");
        };
    }
} interface Shape { }
record Rectangle(double length, double width) implements Shape { }
record Circle(double radius) implements Shape { }

public class Main {
    static double getPerimeter(Shape s) throws IllegalArgumentException {
        return switch (s) {                                                 //1
            case Rectangle r when r.length() == r.width() -> {              //2
                System.out.println("Square detected");
                    yield 4 * r.length();
            }
            case Rectangle r ->                                             //3
                2 * r.length() + 2 * r.width();
            case Circle c ->                                                //4
                2 * c.radius() * Math.PI;
            default ->                                                      //5
                throw new IllegalArgumentException("Unrecognized shape");
        };
    }
} Reference the Shape parameter as s
Evaluate whether s is a Rectangle and whether the Rectangle is a square
Evaluate whether s is a Rectangle
Evaluate whether s is a Circle
If none of the previous clauses match, default to throwing an exception Reference the Shape parameter as s Shape s Evaluate whether s is a Rectangle and whether the Rectangle is a square s Rectangle and Rectangle Evaluate whether s is a Rectangle s Rectangle Evaluate whether s is a Circle s Circle If none of the previous clauses match, default to throwing an exception This Java version will be our baseline. Characteristics of the New switch Syntax switch The new syntax has some advantages over the legacy non-arrow switch inherited from C. switch C-style case clauses are fall-through. Once the runtime has executed a case block, it runs the next case block. To avoid it, you need to set a break explicitly. In the not-so-good old days, some developers leveraged this feature to avoid code duplication. case case case break However, experience has shown that it was too much of a cause of potential bugs: modern coding practices tend to avoid it. Yet, it's too easy to forget a break. With the new arrow syntax, the runtime only evaluates the corresponding case block. break case C-style case clauses only evaluate simple values. The community celebrated Java 7 as a huge boon when it allowed String values. Java 21 went even further: coupled with the arrow syntax, it allowed for switching on types. In the above example, we check the exact Shape type. case Shape Additionally, if switching on classes and if the class hierarchy is sealed, the compiler can automatically detect missing cases. Finally, version 24 added the additional optional when filter. and sealed when On the flip side, in the old C syntax, the runtime jumps directly to the correct case clause. The new arrow syntax evaluates them sequentially, exactly like if with if else statements. case if else In the next sections, we will port the code to other languages. Scala's Pattern Matching Scala's pattern matching has been second to none since its inception. Kotlin drew a lot of inspiration from it. trait Shape
case class Rectangle(length: Double, width: Double) extends Shape
case class Circle(radius: Double) extends Shape

def getPerimeter(s: Shape) = {
  s match {                                                                 //1
    case Rectangle(length, width) if length == width =>                     //2-3
      println("Square detected")
      4 * length
    case Rectangle(length, width) => 2 * length + 2 * width                 //3-4
    case Circle(radius) => 2 * radius * Math.PI                             //3-5
    case _ => throw new IllegalArgumentException("Unrecognized shape")      //6
  }
} trait Shape
case class Rectangle(length: Double, width: Double) extends Shape
case class Circle(radius: Double) extends Shape

def getPerimeter(s: Shape) = {
  s match {                                                                 //1
    case Rectangle(length, width) if length == width =>                     //2-3
      println("Square detected")
      4 * length
    case Rectangle(length, width) => 2 * length + 2 * width                 //3-4
    case Circle(radius) => 2 * radius * Math.PI                             //3-5
    case _ => throw new IllegalArgumentException("Unrecognized shape")      //6
  }
} Use the s reference directly
Evaluate whether s is a Rectangle and whether the Rectangle is a square
Scala additionally matches the class attributes
Evaluate whether s is a Rectangle
Evaluate whether s is a Circle
If none of the previous clauses match, default to throwing an exception Use the s reference directly s Evaluate whether s is a Rectangle and whether the Rectangle is a square s Rectangle Rectangle Scala additionally matches the class attributes Evaluate whether s is a Rectangle s Rectangle Evaluate whether s is a Circle s Circle If none of the previous clauses match, default to throwing an exception Kotlin's Pattern Matching Let's translate Java's code to Kotlin. For that, we must activate the experimental Xwhen-guards compilation feature described in KEEP-371. must experimental Xwhen-guards KEEP-371 interface Shape
data class Rectangle(val length: Double, val width: Double): Shape
data class Circle(val radius: Double): Shape

fun getPerimeter(s: Shape) = when (s) {                                     //1
        is Rectangle if s.length == s.width -> {                            //2
            println("Square detected")
            4 * s.length
        }
        is Rectangle -> 2 * s.length + 2 * s.width                          //3
        is Circle -> 2 * s.radius * Math.PI                                 //4
        else -> throw IllegalArgumentException("Unknown shape")             //5
    }
} interface Shape
data class Rectangle(val length: Double, val width: Double): Shape
data class Circle(val radius: Double): Shape

fun getPerimeter(s: Shape) = when (s) {                                     //1
        is Rectangle if s.length == s.width -> {                            //2
            println("Square detected")
            4 * s.length
        }
        is Rectangle -> 2 * s.length + 2 * s.width                          //3
        is Circle -> 2 * s.radius * Math.PI                                 //4
        else -> throw IllegalArgumentException("Unknown shape")             //5
    }
} Reference the Shape parameter as s
Evaluate whether s is a Rectangle and whether the Rectangle's width is equal to its height
Evaluate whether s is a Rectangle
Evaluate whether s is a Circle
If none of the previous clauses match, default to throwing an exception Reference the Shape parameter as s Shape s Evaluate whether s is a Rectangle and whether the Rectangle's width is equal to its height s Rectangle and Rectangle width height Evaluate whether s is a Rectangle s Rectangle Evaluate whether s is a Circle s Circle If none of the previous clauses match, default to throwing an exception Kotlin's pattern matching greatly resembles Java's, with slight syntax changes, e.g., if replaces when. e.g. if when The comparative evolution of Kotlin vs. Java in pattern matching is quite enlightening. Java lagged behind the legacy C syntax, while Kotlin could already match on concrete types from the beginning. In 2020, Java 14 reduced the gap with the arrow operator; one year later, Java 16 closed the gap completely with the ability to match on concrete types. Finally, Java 23 added the when clause. when The experimental nature of Kotlin's if means Java has overtaken Java–at least in this area! It's rare enough to be noted. if Python's Pattern Matching Before, Python didn't offer anything similar to the switch statement of the above JVM languages. From version 3.10, it does offer the same capability in an elegant way: switch class Shape:
    pass


class Rectangle(Shape):
    def __init__(self, length: float, width: float):
        self.length = length
        self.width = width


class Circle(Shape):
    def __init__(self, radius: float):
        self.radius = radius


def get_perimeter(s: Shape) -> float:
    match s:
        case Rectangle(length=l, width=w) if l == w:
            print("Square detected")
            return 4 * l
        case Rectangle(length=l, width=w):
            return 2 * l + 2 * w
        case Circle(radius=r):
            return 2 * pi * r
        case _:
            raise ValueError("Unknown shape") class Shape:
    pass


class Rectangle(Shape):
    def __init__(self, length: float, width: float):
        self.length = length
        self.width = width


class Circle(Shape):
    def __init__(self, radius: float):
        self.radius = radius


def get_perimeter(s: Shape) -> float:
    match s:
        case Rectangle(length=l, width=w) if l == w:
            print("Square detected")
            return 4 * l
        case Rectangle(length=l, width=w):
            return 2 * l + 2 * w
        case Circle(radius=r):
            return 2 * pi * r
        case _:
            raise ValueError("Unknown shape") The runtime evaluates the case clauses sequentially, as in the JVM languages. case Rust's Pattern Matching Rust's approach to memory management doesn't play well with checking types. In short, Rust offers two base concepts: Structures are structured data placeholders; their memory size is known at compile time
Traits are contracts, akin to interfaces
You can provide an implementation of a trait for a structure
Referencing a variable by its trait has consequences. Since the reference can point to structures of different sizes, the compiler can't know its size at compile-time and must place the variable on the heap instead of the stack. Structures are structured data placeholders; their memory size is known at compile time Traits are contracts, akin to interfaces You can provide an implementation of a trait for a structure Referencing a variable by its trait has consequences. Since the reference can point to structures of different sizes, the compiler can't know its size at compile-time and must place the variable on the heap instead of the stack. I tried to port the original code to Rust one-to-one for educational purposes. The original Java doesn't take advantage of polymorphism; it is not great. In Rust, it's even uglier. It's interesting to realize that while Rust isn't an Object-Oriented programming language, it leads you to use polymorphism. Warning: I don't recommend the following code and will deny any association with it! trait Shape: Any {
    fn as_any(&self) -> &dyn Any;                                     //1
}

struct Circle {
    radius: f64,
}

struct Rectangle {
    width: f64,
    height: f64,
}

impl Shape for Circle {
    fn as_any(&self) -> &dyn Any {                                    //2
        self
    }
}
impl Shape for Rectangle {
    fn as_any(&self) -> &dyn Any {                                    //2
        self
    }
}

fn get_perimeter(s: Box<dyn Shape>) -> f64 {
    match s.as_any() {                                                //3
        any if any.is::<Rectangle>() => {                             //4
            let rectangle = any.downcast_ref::<Rectangle>().unwrap(); //5
            if rectangle.width == rectangle.height {
                println!("Square matched");
                4.0 * rectangle.width
            } else {
                2.0 * (rectangle.width + rectangle.height)
            }
        }
        any if any.is::<Circle>() => {                                //4
            let circle = any.downcast_ref::<Circle>().unwrap();       //5
            2.0 * std::f64::consts::PI * circle.radius
        }
        _ => panic!()
    }
} trait Shape: Any {
    fn as_any(&self) -> &dyn Any;                                     //1
}

struct Circle {
    radius: f64,
}

struct Rectangle {
    width: f64,
    height: f64,
}

impl Shape for Circle {
    fn as_any(&self) -> &dyn Any {                                    //2
        self
    }
}
impl Shape for Rectangle {
    fn as_any(&self) -> &dyn Any {                                    //2
        self
    }
}

fn get_perimeter(s: Box<dyn Shape>) -> f64 {
    match s.as_any() {                                                //3
        any if any.is::<Rectangle>() => {                             //4
            let rectangle = any.downcast_ref::<Rectangle>().unwrap(); //5
            if rectangle.width == rectangle.height {
                println!("Square matched");
                4.0 * rectangle.width
            } else {
                2.0 * (rectangle.width + rectangle.height)
            }
        }
        any if any.is::<Circle>() => {                                //4
            let circle = any.downcast_ref::<Circle>().unwrap();       //5
            2.0 * std::f64::consts::PI * circle.radius
        }
        _ => panic!()
    }
} Rust's type system doesn't offer a way to get the type of a variable. We must create a dedicated function for that.
Implement the function for structures
Match on the underlying structure type
Check the type
Downcast to the underlying structure to use its fields Rust's type system doesn't offer a way to get the type of a variable. We must create a dedicated function for that. Implement the function for structures Match on the underlying structure type Check the type Downcast to the underlying structure to use its fields This artificial code port above misrepresents Rust's pattern-matching abilities. It applies in many places. many places Conclusion Among all languages described in the post, Scala was the first to provide pattern-matching in switch clauses. For many years, it was the Grail that others tried to catch up with; Kotlin and Java have finally reached this stage. switch Outside of the JVM, Python and Rust feature powerful pattern-matching capabilities. With destructuring, pattern matching is a huge help to developers who want to write readable and maintainable code. The complete source code for this post can be found on GitHub^. GitHub^ To go further: To go further: Pattern matching in Java 23
Pattern matching in Kotlin
Pattern matching in Scala
Match statement in Python
Pattern matching in Rust Pattern matching in Java 23 Pattern matching in Java 23 Pattern matching in Kotlin Pattern matching in Kotlin Pattern matching in Scala Pattern matching in Scala Match statement in Python Match statement in Python Pattern matching in Rust Pattern matching in Rust Originally published at A Java Geek on July 20th, 2025 Originally published at A Java Geek on July 20th, 2025 A Java Geek

GitHub

Understanding Git Push and Rebase Defaults: A Hands-On Guide

Home Assistant: How to Replace the Explicit Grouping of Devices by Areas

Read all my posts!

Comparing Pattern-Matching Across Different Languages: Java, Scala, and More

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

16 Best Practices For Securing Your APIs with Apache APISIX - Part 1

Why Being Likable At Work Matters

Windows Sticky Keys Exploit: The War Veteran That Never Dies

Is ‘bias for action’ making product managers lazier?

Zero People Charged With Online Pirating, Swedish Prosecutor's Office Reports

16 Best Practices For Securing Your APIs with Apache APISIX - Part 1

Why Being Likable At Work Matters

Windows Sticky Keys Exploit: The War Veteran That Never Dies

Is ‘bias for action’ making product managers lazier?

Zero People Charged With Online Pirating, Swedish Prosecutor's Office Reports

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps