Use pattern matching to build your class behavior for better code

[アーティクル]
12/04/2024

The pattern matching features in C# provide syntax to express your algorithms. You can use these techniques to implement the behavior in your classes. You can combine object-oriented class design with a data-oriented implementation to provide concise code while modeling real-world objects.

In this tutorial, you learn how to:

Express your object oriented classes using data patterns.
Implement those patterns using C#'s pattern matching features.
Leverage compiler diagnostics to validate your implementation.

Prerequisites

You need to set up your machine to run .NET. Download Visual Studio 2022 or the .NET SDK.

Build a simulation of a canal lock

In this tutorial, you build a C# class that simulates a canal lock. Briefly, a canal lock is a device that raises and lowers boats as they travel between two stretches of water at different levels. A lock has two gates and some mechanism to change the water level.

In its normal operation, a boat enters one of the gates while the water level in the lock matches the water level on the side the boat enters. Once in the lock, the water level is changed to match the water level where the boat leaves the lock. Once the water level matches that side, the gate on the exit side opens. Safety measures make sure an operator can't create a dangerous situation in the canal. The water level can be changed only when both gates are closed. At most one gate can be open. To open a gate, the water level in the lock must match the water level outside the gate being opened.

You can build a C# class to model this behavior. A CanalLock class would support commands to open or close either gate. It would have other commands to raise or lower the water. The class should also support properties to read the current state of both gates and the water level. Your methods implement the safety measures.

Define a class

You build a console application to test your CanalLock class. Create a new console project for .NET 5 using either Visual Studio or the .NET CLI. Then, add a new class and name it CanalLock. Next, design your public API, but leave the methods not implemented:

public enum WaterLevel
{
    Low,
    High
}
public class CanalLock
{
    // Query canal lock state:
    public WaterLevel CanalLockWaterLevel { get; private set; } = WaterLevel.Low;
    public bool HighWaterGateOpen { get; private set; } = false;
    public bool LowWaterGateOpen { get; private set; } = false;

    // Change the upper gate.
    public void SetHighGate(bool open)
    {
        throw new NotImplementedException();
    }

    // Change the lower gate.
    public void SetLowGate(bool open)
    {
        throw new NotImplementedException();
    }

    // Change water level.
    public void SetWaterLevel(WaterLevel newLevel)
    {
        throw new NotImplementedException();
    }

    public override string ToString() =>
        $"The lower gate is {(LowWaterGateOpen ? "Open" : "Closed")}. " +
        $"The upper gate is {(HighWaterGateOpen ? "Open" : "Closed")}. " +
        $"The water level is {CanalLockWaterLevel}.";
}

The preceding code initializes the object so both gates are closed, and the water level is low. Next, write the following test code in your Main method to guide you as you create a first implementation of the class:

// Create a new canal lock:
var canalGate = new CanalLock();

// State should be doors closed, water level low:
Console.WriteLine(canalGate);

canalGate.SetLowGate(open: true);
Console.WriteLine($"Open the lower gate:  {canalGate}");

Console.WriteLine("Boat enters lock from lower gate");

canalGate.SetLowGate(open: false);
Console.WriteLine($"Close the lower gate:  {canalGate}");

canalGate.SetWaterLevel(WaterLevel.High);
Console.WriteLine($"Raise the water level: {canalGate}");

canalGate.SetHighGate(open: true);
Console.WriteLine($"Open the higher gate:  {canalGate}");

Console.WriteLine("Boat exits lock at upper gate");
Console.WriteLine("Boat enters lock from upper gate");

canalGate.SetHighGate(open: false);
Console.WriteLine($"Close the higher gate: {canalGate}");

canalGate.SetWaterLevel(WaterLevel.Low);
Console.WriteLine($"Lower the water level: {canalGate}");

canalGate.SetLowGate(open: true);
Console.WriteLine($"Open the lower gate:  {canalGate}");

Console.WriteLine("Boat exits lock at upper gate");

canalGate.SetLowGate(open: false);
Console.WriteLine($"Close the lower gate:  {canalGate}");

Next, add a first implementation of each method in the CanalLock class. The following code implements the methods of the class without concern to the safety rules. You add safety tests later:

// Change the upper gate.
public void SetHighGate(bool open)
{
    HighWaterGateOpen = open;
}

// Change the lower gate.
public void SetLowGate(bool open)
{
    LowWaterGateOpen = open;
}

// Change water level.
public void SetWaterLevel(WaterLevel newLevel)
{
    CanalLockWaterLevel = newLevel;
}

The tests you wrote so far pass. You implemented the basics. Now, write a test for the first failure condition. At the end of the previous tests, both gates are closed, and the water level is set to low. Add a test to try opening the upper gate:

Console.WriteLine("=============================================");
Console.WriteLine("     Test invalid commands");
// Open "wrong" gate (2 tests)
try
{
    canalGate = new CanalLock();
    canalGate.SetHighGate(open: true);
}
catch (InvalidOperationException)
{
    Console.WriteLine("Invalid operation: Can't open the high gate. Water is low.");
}
Console.WriteLine($"Try to open upper gate: {canalGate}");

This test fails because the gate opens. As a first implementation, you could fix it with the following code:

// Change the upper gate.
public void SetHighGate(bool open)
{
    if (open && (CanalLockWaterLevel == WaterLevel.High))
        HighWaterGateOpen = true;
    else if (open && (CanalLockWaterLevel == WaterLevel.Low))
        throw new InvalidOperationException("Cannot open high gate when the water is low");
}

Your tests pass. But, as you add more tests, you add more if clauses and test different properties. Soon, these methods get too complicated as you add more conditionals.

Implement the commands with patterns

A better way is to use patterns to determine if the object is in a valid state to execute a command. You can express if a command is allowed as a function of three variables: the state of the gate, the level of the water, and the new setting:

New setting	Gate state	Water Level	Result
Closed	Closed	High	Closed
Closed	Closed	Low	Closed
Closed	Open	High	Closed
~~Closed~~	~~Open~~	~~Low~~	~~Closed~~
Open	Closed	High	Open
Open	Closed	Low	Closed (Error)
Open	Open	High	Open
~~Open~~	~~Open~~	~~Low~~	~~Closed (Error)~~

The fourth and last rows in the table have strike through text because they're invalid. The code you're adding now should make sure the high water gate is never opened when the water is low. Those states can be coded as a single switch expression (remember that false indicates "Closed"):

HighWaterGateOpen = (open, HighWaterGateOpen, CanalLockWaterLevel) switch
{
    (false, false, WaterLevel.High) => false,
    (false, false, WaterLevel.Low) => false,
    (false, true, WaterLevel.High) => false,
    (false, true, WaterLevel.Low) => false, // should never happen
    (true, false, WaterLevel.High) => true,
    (true, false, WaterLevel.Low) => throw new InvalidOperationException("Cannot open high gate when the water is low"),
    (true, true, WaterLevel.High) => true,
    (true, true, WaterLevel.Low) => false, // should never happen
};

Try this version. Your tests pass, validating the code. The full table shows the possible combinations of inputs and results. That means you and other developers can quickly look at the table and see that you covered all the possible inputs. Even easier, the compiler can help as well. After you add the previous code, you can see that the compiler generates a warning: CS8524 indicates the switch expression doesn't cover all possible inputs. The reason for that warning is that one of the inputs is an enum type. The compiler interprets "all possible inputs" as all inputs from the underlying type, typically an int. This switch expression only checks the values declared in the enum. To remove the warning, you can add a catch-all discard pattern for the last arm of the expression. This condition throws an exception, because it indicates invalid input:

_  => throw new InvalidOperationException("Invalid internal state"),

The preceding switch arm must be last in your switch expression because it matches all inputs. Experiment by moving it earlier in the order. That causes a compiler error CS8510 for unreachable code in a pattern. The natural structure of switch expressions enables the compiler to generate errors and warnings for possible mistakes. The compiler "safety net" makes it easier for you to create correct code in fewer iterations, and the freedom to combine switch arms with wildcards. The compiler issues errors if your combination results in unreachable arms you didn't expect, and warnings if you remove a needed arm.

The first change is to combine all the arms where the command is to close the gate; that's always allowed. Add the following code as the first arm in your switch expression:

(false, _, _) => false,

After you add the previous switch arm, you'll get four compiler errors, one on each of the arms where the command is false. Those arms are already covered by the newly added arm. You can safely remove those four lines. You intended this new switch arm to replace those conditions.

Next, you can simplify the four arms where the command is to open the gate. In both cases where the water level is high, the gate can be opened. (In one, it's already open.) One case where the water level is low throws an exception, and the other shouldn't happen. It should be safe to throw the same exception if the water lock is already in an invalid state. You can make the following simplifications for those arms:

(true, _, WaterLevel.High) => true,
(true, false, WaterLevel.Low) => throw new InvalidOperationException("Cannot open high gate when the water is low"),
_ => throw new InvalidOperationException("Invalid internal state"),

Run your tests again, and they pass. Here's the final version of the SetHighGate method:

// Change the upper gate.
public void SetHighGate(bool open)
{
    HighWaterGateOpen = (open, HighWaterGateOpen, CanalLockWaterLevel) switch
    {
        (false, _,    _)               => false,
        (true, _,     WaterLevel.High) => true,
        (true, false, WaterLevel.Low)  => throw new InvalidOperationException("Cannot open high gate when the water is low"),
        _                              => throw new InvalidOperationException("Invalid internal state"),
    };
}

Implement patterns yourself

Now that you've seen the technique, fill in the SetLowGate and SetWaterLevel methods yourself. Start by adding the following code to test invalid operations on those methods:

Console.WriteLine();
Console.WriteLine();
try
{
    canalGate = new CanalLock();
    canalGate.SetWaterLevel(WaterLevel.High);
    canalGate.SetLowGate(open: true);
}
catch (InvalidOperationException)
{
    Console.WriteLine("invalid operation: Can't open the lower gate. Water is high.");
}
Console.WriteLine($"Try to open lower gate: {canalGate}");
// change water level with gate open (2 tests)
Console.WriteLine();
Console.WriteLine();
try
{
    canalGate = new CanalLock();
    canalGate.SetLowGate(open: true);
    canalGate.SetWaterLevel(WaterLevel.High);
}
catch (InvalidOperationException)
{
    Console.WriteLine("invalid operation: Can't raise water when the lower gate is open.");
}
Console.WriteLine($"Try to raise water with lower gate open: {canalGate}");
Console.WriteLine();
Console.WriteLine();
try
{
    canalGate = new CanalLock();
    canalGate.SetWaterLevel(WaterLevel.High);
    canalGate.SetHighGate(open: true);
    canalGate.SetWaterLevel(WaterLevel.Low);
}
catch (InvalidOperationException)
{
    Console.WriteLine("invalid operation: Can't lower water when the high gate is open.");
}
Console.WriteLine($"Try to lower water with high gate open: {canalGate}");

Run your application again. You can see the new tests fail, and the canal lock gets into an invalid state. Try to implement the remaining methods yourself. The method to set the lower gate should be similar to the method to set the upper gate. The method that changes the water level has different checks, but should follow a similar structure. You might find it helpful to use the same process for the method that sets the water level. Start with all four inputs: The state of both gates, the current state of the water level, and the requested new water level. The switch expression should start with:

CanalLockWaterLevel = (newLevel, CanalLockWaterLevel, LowWaterGateOpen, HighWaterGateOpen) switch
{
    // elided
};

You have 16 total switch arms to fill in. Then, test and simplify.

Did you make methods something like this?

// Change the lower gate.
public void SetLowGate(bool open)
{
    LowWaterGateOpen = (open, LowWaterGateOpen, CanalLockWaterLevel) switch
    {
        (false, _, _) => false,
        (true, _, WaterLevel.Low) => true,
        (true, false, WaterLevel.High) => throw new InvalidOperationException("Cannot open low gate when the water is high"),
        _ => throw new InvalidOperationException("Invalid internal state"),
    };
}

// Change water level.
public void SetWaterLevel(WaterLevel newLevel)
{
    CanalLockWaterLevel = (newLevel, CanalLockWaterLevel, LowWaterGateOpen, HighWaterGateOpen) switch
    {
        (WaterLevel.Low, WaterLevel.Low, true, false) => WaterLevel.Low,
        (WaterLevel.High, WaterLevel.High, false, true) => WaterLevel.High,
        (WaterLevel.Low, _, false, false) => WaterLevel.Low,
        (WaterLevel.High, _, false, false) => WaterLevel.High,
        (WaterLevel.Low, WaterLevel.High, false, true) => throw new InvalidOperationException("Cannot lower water when the high gate is open"),
        (WaterLevel.High, WaterLevel.Low, true, false) => throw new InvalidOperationException("Cannot raise water when the low gate is open"),
        _ => throw new InvalidOperationException("Invalid internal state"),
    };
}

Your tests should pass, and the canal lock should operate safely.

Summary

In this tutorial, you learned to use pattern matching to check the internal state of an object before applying any changes to that state. You can check combinations of properties. Once you built tables for any of those transitions, you test your code, then simplify for readability and maintainability. These initial refactorings might suggest further refactorings that validate internal state or manage other API changes. This tutorial combined classes and objects with a more data-oriented, pattern-based approach to implement those classes.

次の方法で共有