C# Generics — Boxing-Induced OutOfMemory in Trading Engines

Generics are everywhere in C#. List<T>, Dictionary<TKey, TValue>, Task<T> — you use them daily without thinking about what's happening under the hood. That's fine until something breaks. A cast fails. Boxing spikes your GC. A constraint you didn't add forces a runtime workaround. That's when knowing how generics actually work saves you a late-night debug session.

Before C# 2.0, developers used ArrayList and cast everything to and from object. The compiler couldn't catch type mismatches — a bug that should fail at compile time exploded as InvalidCastException at runtime. Generics fix this by letting you parameterise a class with a type placeholder. The compiler fills it in when you use the code, giving you full type checking without any runtime casting overhead.

By the end of this article you'll understand why generics exist at a language-design level, write your own generic classes with constraints, combine generics with interfaces for real architectural patterns, and avoid the three mistakes that trip up even experienced developers. You'll also walk away with sharp answers to the interview questions that separate juniors from mid-level engineers.

What Generics Actually Prevent — and Why Trading Engines Care

Generics in C# let you define classes, methods, and interfaces with a placeholder type that is specified at compile time. The core mechanic: the compiler generates specialized code for each value type used, eliminating boxing and unboxing overhead. For reference types, it shares a single internal instantiation via runtime trickery, but for structs — like decimal, int, or custom value types — each gets its own native code path.

In practice, this means a List stores integers directly in a contiguous array, not as boxed objects on the heap. A List storing ints boxes each value: allocation, pointer dereference, GC pressure. The difference is measurable: a tight loop over 10 million ints in a generic list runs at native speed; the boxed version allocates 80 MB of heap objects and triggers multiple Gen-2 collections.

Use generics whenever you need type-safe collections, algorithms, or data structures that work with value types. In trading engines, where latency is measured in microseconds and memory churn kills determinism, generics are not optional — they are the difference between a predictable 10μs path and a GC pause that blows your P99.

The Problem Generics Solve — Why object-Based Code Is a Time Bomb

Before you can appreciate generics, you need to feel the pain they eliminate. The classic approach before C# 2.0 was to write everything against the object type — the root of all C# types. It seemed clever: one method handles everything. In practice, it was a maintenance nightmare.

Every value you pulled out had to be cast back to its real type. The compiler had no idea what was actually in your collection. You could put a string in a list of integers and the code would compile fine — it would just blow up at runtime when some unsuspecting method tried to call .ToString() on what it assumed was an int.

There's also a performance cost. Value types like int and double must be 'boxed' — wrapped in a heap-allocated object — to be stored as object, then 'unboxed' when retrieved. In tight loops processing thousands of items, this garbage pressure is measurable.

Generics eliminate both problems. You declare the type once at the use site, the compiler enforces it everywhere, and value types are stored directly without boxing. You get the flexibility of writing reusable code AND the safety of a strongly-typed language — not a trade-off between them.

using System;
using System.Collections;
using System.Collections.Generic;

class BeforeAndAfterGenerics
{
    static void Main()
    {
        // BEFORE GENERICS
        var legacyScores = new ArrayList();
        legacyScores.Add(95);          // int gets BOXED onto the heap
        legacyScores.Add(87);
        legacyScores.Add("oops");      // compiler is fine with this string!

        try
        {
            foreach (object item in legacyScores)
            {
                int score = (int)item;  // UNBOXING — risky cast every time
                Console.WriteLine($"Legacy score: {score}");
            }
        }
        catch (InvalidCastException ex)
        {
            Console.WriteLine($"Runtime crash: {ex.Message}");
        }

        // AFTER GENERICS
        var modernScores = new List<int>();
        modernScores.Add(95);          // stored directly, no boxing
        modernScores.Add(87);
        // modernScores.Add("oops");   // COMPILE ERROR: cannot convert string to int

        foreach (int score in modernScores)  // no cast needed
        {
            Console.WriteLine($"Modern score: {score}");
        }
    }
}

Writing Your Own Generic Class — Building a Type-Safe Result Wrapper

The best way to deeply understand generics is to build something you'd actually use in production. A Result wrapper is a perfect example — it represents either a successful value or an error, without throwing exceptions for expected failure cases. This pattern is common in functional-leaning C# codebases and in every API layer that needs to communicate failure without polluting control flow with exceptions.

The T in Result is a type parameter — a placeholder that the compiler replaces with a concrete type when you instantiate the class. You can name it anything, but T is the convention for a single generic type. TKey and TValue are conventional for two parameters, as you see in Dictionary.

Notice how the class is defined once, but can hold a string result, an int result, or a complex User object result. The internal logic — storing the value, checking success, returning errors — is written exactly once. That's the core promise of generics: write the shape of the behaviour, defer the type decision to the caller.

The private constructor pattern combined with static factory methods also means you can never accidentally create a Result in an invalid state — a bonus architectural win that generics enable cleanly.

using System;

public class Result<T>
{
    public T? Value { get; }
    public string? ErrorMessage { get; }
    public bool IsSuccess { get; }

    private Result(T? value, string? error, bool isSuccess)
    {
        Value = value;
        ErrorMessage = error;
        IsSuccess = isSuccess;
    }

    public static Result<T> Success(T value)
        => new Result<T>(value, null, true);

    public static Result<T> Failure(string errorMessage)
        => new Result<T>(default, errorMessage, false);

    public override string ToString()
        => IsSuccess ? $"Success: {Value}" : $"Failure: {ErrorMessage}";
}

public class UserService
{
    public Result<string> FindUsername(int userId)
    {
        if (userId == 42)
            return Result<string>.Success("ada.lovelace");
        return Result<string>.Failure($"No user found with ID {userId}");
    }
}

class Program
{
    static void Main()
    {
        var service = new UserService();
        Result<string> found = service.FindUsername(42);
        if (found.IsSuccess)
            Console.WriteLine($"Found user: {found.Value}");
        else
            Console.WriteLine($"Error: {found.ErrorMessage}");

        Result<string> notFound = service.FindUsername(99);
        Console.WriteLine(notFound);

        Result<int> calculationResult = Result<int>.Success(1337);
        Console.WriteLine($"Calculation gave us: {calculationResult.Value + 1}");
    }
}

Generic Constraints — Teaching the Compiler What T Can Do

Here's the most powerful — and most misunderstood — feature of C# generics: constraints. Without them, T is a complete mystery to the compiler. It could be anything, so you can only call the methods that every single type in C# shares: ToString(), GetHashCode(), and Equals(). That's a pretty short list.

Constraints let you tell the compiler 'T is guaranteed to be at least this kind of thing'. Once you add a constraint, the compiler unlocks every method and property defined by that constraint. You get IntelliSense, type checking, and zero casting.

The where keyword is how you add constraints. The most common ones are: where T : class (T must be a reference type), where T : struct (T must be a value type), where T : new() (T must have a parameterless constructor), and where T : ISomeInterface (T must implement that interface). You can combine multiple constraints on the same type parameter.

The interface constraint is the one you'll use most in real codebases. It's how you write algorithms that are generic over behaviour, not type. A sorting method that works on anything sortable, a repository that works on anything with an ID — these are built with interface constraints.

using System;
using System.Collections.Generic;

public interface IEntity
{
    int Id { get; }
    string Describe();
}

public class Product : IEntity
{
    public int Id { get; init; }
    public string Name { get; init; } = string.Empty;
    public decimal Price { get; init; }
    public string Describe() => $"Product #{Id}: {Name} at ${Price:F2}";
}

public class Employee : IEntity
{
    public int Id { get; init; }
    public string FullName { get; init; } = string.Empty;
    public string Department { get; init; } = string.Empty;
    public string Describe() => $"Employee #{Id}: {FullName} in {Department}";
}

public class InMemoryRepository<T> where T : IEntity
{
    private readonly Dictionary<int, T> _store = new();

    public void Save(T entity)
    {
        _store[entity.Id] = entity;
        Console.WriteLine($"Saved: {entity.Describe()}");
    }

    public T? FindById(int id)
    {
        _store.TryGetValue(id, out T? entity);
        return entity;
    }

    public void PrintAll()
    {
        foreach (var entry in _store.Values)
            Console.WriteLine($"  -> {entry.Describe()}");
    }
}

public static class EntityMapper
{
    public static List<TResult> MapDescriptions<TSource, TResult>(
        IEnumerable<TSource> entities,
        Func<TSource, TResult> mapFunc)
        where TSource : IEntity
        where TResult : new()
    {
        var results = new List<TResult>();
        foreach (var entity in entities)
            results.Add(mapFunc(entity));
        return results;
    }
}

class Program
{
    static void Main()
    {
        var productRepo = new InMemoryRepository<Product>();
        productRepo.Save(new Product { Id = 1, Name = "Mechanical Keyboard", Price = 149.99m });
        productRepo.Save(new Product { Id = 2, Name = "USB-C Hub", Price = 49.95m });

        Console.WriteLine("\nAll products:");
        productRepo.PrintAll();

        var employeeRepo = new InMemoryRepository<Employee>();
        employeeRepo.Save(new Employee { Id = 101, FullName = "Grace Hopper", Department = "Engineering" });
        employeeRepo.Save(new Employee { Id = 102, FullName = "Alan Turing", Department = "Research" });

        Console.WriteLine("\nAll employees:");
        employeeRepo.PrintAll();

        Product? keyboard = productRepo.FindById(1);
        Console.WriteLine($"\nFound: {keyboard?.Describe() ?? "not found"}");
    }
}

Generic Interfaces and Covariance — The Pattern Behind LINQ and IEnumerable

Once you're comfortable writing generic classes, the next level is understanding generic interfaces and variance — specifically covariance (out) and contravariance (in). These aren't academic features; they're why you can assign a List<string> to an IEnumerable<string> variable, and why LINQ works seamlessly across all collection types.

Covariance means a generic type with a more derived type argument can be treated as a generic type with a base type. So IEnumerable<string> can be assigned to IEnumerable<object> because string derives from object, and IEnumerable<T> is declared with out T — meaning T is only ever produced (returned), never consumed. The out keyword is what tells the compiler it's safe to widen the type.

Contravariance is the reverse — Action<object> can be assigned to Action<string> because Action<T> uses in T, meaning T is only consumed (taken as input). If you can handle any object, you can certainly handle a string.

In practice, you'll consume covariant and contravariant interfaces far more often than you'll write them. But knowing WHY IEnumerable<T> uses out T explains why so much LINQ code just works, and it's the kind of deep knowledge that separates engineers who use the framework from those who understand it.

using System;
using System.Collections.Generic;

public interface IProducer<out T>
{
    T Produce();
}

public interface IConsumer<in T>
{
    void Consume(T item);
}

public class Animal
{
    public string Name { get; init; } = string.Empty;
    public virtual string Sound() => "...";
}

public class Dog : Animal
{
    public override string Sound() => "Woof";
}

public class DogProducer : IProducer<Dog>
{
    public Dog Produce() => new Dog { Name = "Rex" };
}

public class AnimalLogger : IConsumer<Animal>
{
    public void Consume(Animal animal)
        => Console.WriteLine($"Logging animal: {animal.Name} says {animal.Sound()}");
}

class Program
{
    static void Main()
    {
        // COVARIANCE: IProducer<Dog> -> IProducer<Animal>
        IProducer<Dog> dogProducer = new DogProducer();
        IProducer<Animal> animalProducer = dogProducer;
        Animal producedAnimal = animalProducer.Produce();
        Console.WriteLine($"Produced: {producedAnimal.Name} says {producedAnimal.Sound()}");

        // CONTRAVARIANCE: IConsumer<Animal> -> IConsumer<Dog>
        IConsumer<Animal> animalConsumer = new AnimalLogger();
        IConsumer<Dog> dogConsumer = animalConsumer;
        dogConsumer.Consume(new Dog { Name = "Buddy" });

        // REAL-WORLD COVARIANCE: IEnumerable<T>
        List<string> dogNames = new() { "Max", "Bella", "Charlie" };
        IEnumerable<object> objectNames = dogNames;
        Console.WriteLine("\nDog names as objects:");
        foreach (object name in objectNames)
            Console.WriteLine($"  {name}");
    }
}

Generic Methods and Type Inference — When the Compiler Deducing T Works (and When It Doesn't)

Generic methods are distinct from generic classes. You can have a generic method inside a non-generic class, and the type parameter is inferred from the arguments you pass. This is incredibly convenient — you don't need to specify the type unless the compiler can't figure it out. For example, when you write var result = Helper.Swap(ref a, ref b); the compiler infers T from the type of a.

But type inference has limits. If the method's type parameter appears only in the return type, the compiler cannot infer it — you must specify it explicitly. This is common in factory patterns: T Create() where T : new() requires you to call Create().

Another common gotcha is overload resolution. If two overloads differ only by a generic type parameter, the compiler may pick the wrong one or fail with an ambiguity error. In that case, explicitly specifying the type argument resolves the ambiguity.

Understanding when inference works and when it doesn't separates developers who fight the compiler from those who let it work for them.

using System;
using System.Collections.Generic;

public class Helper
{
    public static void Swap<T>(ref T a, ref T b)
    {
        T temp = a;
        a = b;
        b = temp;
    }

    public static T Create<T>() where T : new()
    {
        return new T();
    }

    public static void Process<T>(T item)
    {
        Console.WriteLine($"Single item of type {typeof(T).Name}: {item}");
    }

    public static void Process<T>(IEnumerable<T> items)
    {
        Console.WriteLine($"Collection of {typeof(T).Name}:");
        foreach (var item in items)
            Console.WriteLine($"  - {item}");
    }
}

public class Example
{
    public static void Main()
    {
        int x = 1, y = 2;
        Helper.Swap(ref x, ref y);
        Console.WriteLine($"Swapped: x={x}, y={y}");

        string s1 = "hello", s2 = "world";
        Helper.Swap(ref s1, ref s2);
        Console.WriteLine($"Swapped: s1={s1}, s2={s2}");

        var list = Helper.Create<List<int>>();
        list.Add(42);
        Console.WriteLine($"Created list with: {list[0]}");

        List<int> numbers = new() { 10, 20, 30 };
        Helper.Process(numbers);

        int[] array = { 1, 2, 3 };
        Helper.Process(array);

        Helper.Process<int[]>(array);
    }
}

Real-World Generic Patterns — Repository, Specification and Type-Safe Builders

Now that you understand constraints and variance, let's look at three real-world patterns that use generics to solve production problems. These aren't academic — they're patterns you'll find in every mature C# codebase.

The Generic Repository pattern keeps data access code consistent across entity types. With a constraint like where T : IEntity, you get a single implementation that handles Product, Order, User — any type with an identity. The pattern reduces duplication, but it also introduces a decision: do you build one grand repository or compose small ones? Generics let you do either.

The Specification pattern pairs with generics to build composable, testable query logic. A Specification<T> is a predicate wrapped in a class. You combine specifications with &&, ||, and ! operators. Pass them to a generic repository method: repository.Find(spec). The T makes the specification reusable across entity types without casting.

Type-safe Builders use generic methods to enforce a construction sequence at compile time. Instead of a builder that throws InvalidOperationException when you call Build() too early, you make each step return a new builder type. The compiler prevents you from creating an invalid object in the first place — no runtime checks needed.

using System;
using System.Linq.Expressions;
using System.Collections.Generic;

// Specification pattern — encapsulate query logic
public class Specification<T>
{
    public Expression<Func<T, bool>> Criteria { get; }
    public Specification(Expression<Func<T, bool>> criteria)
        => Criteria = criteria;

    public static Specification<T> operator &(Specification<T> left, Specification<T> right)
        => new Specification<T>(CombineAnd(left.Criteria, right.Criteria));

    private static Expression<Func<T, bool>> CombineAnd(
        Expression<Func<T, bool>> left,
        Expression<Func<T, bool>> right)
    {
        var param = Expression.Parameter(typeof(T));
        var body = Expression.AndAlso(
            Expression.Invoke(left, param),
            Expression.Invoke(right, param));
        return Expression.Lambda<Func<T, bool>>(body, param);
    }
}

public interface IEntity
{
    int Id { get; }
}

public class Order : IEntity
{
    public int Id { get; init; }
    public decimal Total { get; init; }
}

// Generic repository with specification support
public class Repository<T> where T : IEntity
{
    private readonly List<T> _store = new();

    public void Add(T entity) => _store.Add(entity);

    public IEnumerable<T> Find(Specification<T> spec)
    {
        var compiled = spec.Criteria.Compile();
        return _store.FindAll(item => compiled(item));
    }

    public IEnumerable<T> GetAll() => _store.AsReadOnly();
}

// Type-safe builder: compile-time construction enforcement
public class EmailBuilder
{
    private string _to = string.Empty;
    private string _subject = string.Empty;

    public EmailBuilder WithTo(string to)
    {
        _to = to;
        return this;
    }

    public EmailBuilder WithSubject(string subject)
    {
        _subject = subject;
        return this;
    }

    public Email Build() => new Email(_to, _subject);
}

public record Email(string To, string Subject);

class Program
{
    static void Main()
    {
        var repo = new Repository<Order>();
        repo.Add(new Order { Id = 1, Total = 99.95m });
        repo.Add(new Order { Id = 2, Total = 199.99m });

        var largeOrders = new Specification<Order>(o => o.Total > 100);
        foreach (var order in repo.Find(largeOrders))
            Console.WriteLine($"Large order #{order.Id}: ${order.Total}");

        var email = new EmailBuilder()
            .WithTo("user@example.com")
            .WithSubject("Your order confirmation")
            .Build();

        Console.WriteLine($"Email ready: {email}");
    }
}

Generic Delegates — The Silent Assassin in Event-Driven Systems

Delegates without generics are null-checks waiting to happen. In trading engines, a price-feed handler that casts from object will eventually blow up when someone passes an OrderBook instead of a Quote. Generic delegates — Action<T>, Func<T, TResult>, Predicate<T> — force the contract at compile time. Your event publisher says 'I expect a Quote' and the compiler enforces it. No runtime casting, no reflection overhead. The pattern is brutal: define a delegate that matches your payload exactly, then wire it with generics. When the market data feed changes schema, the compiler breaks every mismatched subscriber before CI finishes. That's not paranoia; that's production discipline. The junior who wires a multicast delegate with object is the same junior who gets paged at 3 AM because a JSON parser returned a different type than expected.

// io.thecodeforge
public delegate void QuoteHandler<T>(T quote) where T : IQuote;

public sealed class MarketDataFeed
{
    private readonly List<QuoteHandler<IQuote>> _subscribers = new();

    public void Subscribe(QuoteHandler<IQuote> handler) => _subscribers.Add(handler);

    public void Publish(IQuote quote)
    {
        foreach (var handler in _subscribers)
            handler(quote); // Compile-time safe – no casting
    }
}

public interface IQuote { decimal Price { get; } string Symbol { get; } }

Generic Constraints — The Nine Words That Saved a Weekend

Constraints aren't compiler pedantry; they're contracts that prevent runtime explosions. Without constraints, your generic repository's GetById<T> will happily accept T = StringBuilder, then fail when you try to cast it to an entity. The 'where T : class' constraint says 'only reference types'. The 'where T : new()' constraint promises the default constructor exists. Stack them: 'where T : class, IEntity, new()'. Now your generic cache actually works because every T has an Id property and a parameterless constructor. The junior who says 'I'll add constraints later' is the same junior who ship a generic serializer that throws on structs. In .NET 8, constraints also enable static abstract interface methods — your generic math finally compiles. 'where T : INumber<T>' unlocks Add, Subtract, Multiply without boxing. That is the difference between latency-critical code and toy examples.

// io.thecodeforge
public interface IEntity { int Id { get; } }

public sealed class EntityCache<T> where T : class, IEntity, new()
{
    private readonly Dictionary<int, T> _store = new();

    public T GetOrCreate(int id)
    {
        if (_store.TryGetValue(id, out var entity))
            return entity;
        
        entity = new T { Id = id }; // new() constraint guarantees this works
        _store[id] = entity;
        return entity;
    }
}

// Usage – compiler rejects anything that doesn't satisfy all constraints
// EntityCache<string> cache; // ERROR: string doesn't implement IEntity
// EntityCache<SomeRecord> cache; // ERROR: record structs may not have parameterless ctor

Generic Math Interfaces INumber (C# 11+)

C# 11 introduced generic math interfaces, enabling you to write generic algorithms that work with numeric types. The INumber<T> interface (and related interfaces like IAdditionOperators<T, T, T>) allows you to constrain generic type parameters to numeric types, eliminating the need for boxing or runtime type checks. For example, you can create a generic Sum method that works with int, double, decimal, or any type implementing INumber<T>. This is particularly valuable in trading engines where high-performance numeric computations are common. Without these interfaces, you'd either need to overload methods for each numeric type or use object and boxing, which causes allocations and GC pressure. With INumber<T>, the compiler generates efficient, type-specific code without boxing. The interfaces are part of the System.Numerics namespace and require .NET 7 or later. They also support operators like +, -, *, /, and comparisons, making them ideal for generic math libraries. However, note that these interfaces are not available in older .NET versions, so you may need to target .NET 7+ or use polyfill packages. In trading systems, using INumber<T> can reduce latency by avoiding boxing and enabling JIT optimizations like inlining and vectorization.

using System.Numerics;

public static class MathHelper
{
    public static T Sum<T>(IEnumerable<T> values) where T : INumber<T>
    {
        T total = T.Zero;
        foreach (var v in values)
        {
            total += v;
        }
        return total;
    }
}

// Usage
var intSum = MathHelper.Sum(new[] { 1, 2, 3 });
var doubleSum = MathHelper.Sum(new[] { 1.5, 2.3, 3.7 });

Covariant Return Types on Overrides

C# 9.0 introduced covariant return types, allowing an overriding method to return a more derived type than the base method. This is particularly useful in generic hierarchies where you want to preserve type specificity without casting. For example, if you have a base class Repository<T> with a method GetById(int id) returning T, a derived CustomerRepository can override it to return a more specific type like Customer (which is T anyway), but more importantly, if you have a non-generic base with a virtual method returning object, you can override it to return a concrete type. This reduces the need for runtime type checks and casting, which can cause boxing if the return type is a value type. In trading engines, where value types like Price or Quantity are common, covariant return types help avoid boxing when overriding methods. For instance, a base interface IOrderProcessor might have a method Process() returning object, but a concrete implementation can override it to return OrderResult (a struct), eliminating boxing. This feature works with both classes and interfaces, but interfaces require explicit implementation. Note that covariant return types are not the same as generic covariance (which uses out); they are a simpler mechanism for method overrides. They improve type safety and performance by removing unnecessary casts and boxing.

public abstract class OrderBase
{
    public abstract object Process();
}

public class Order<T> : OrderBase
{
    public override T Process() // Covariant return type (C# 9.0+)
    {
        // Implementation
        return default(T);
    }
}

// Usage
Order<int> order = new Order<int>();
int result = order.Process(); // No boxing, no cast

Generic Attributes in C# 11

C# 11 introduced generic attributes, allowing you to define attributes that accept type parameters. Previously, attributes could only take Type as a parameter, requiring runtime reflection to get the type. With generic attributes, you can specify the type at compile time, enabling better type safety and eliminating boxing when storing type information. For example, you can create a [Validator] attribute that validates a property against a specific type. In trading engines, this is useful for metadata-driven systems like serialization, validation, or mapping. Without generic attributes, you'd need to pass a Type object, which often involves boxing if the type is a value type. With generic attributes, the type parameter is preserved without boxing. However, generic attributes have limitations: they cannot be used on attributes that are applied to generic type parameters themselves, and they must be used with concrete types (not open generics). Also, the attribute class must be unsealed to allow inheritance? Actually, generic attributes can be sealed or unsealed. They work with reflection, but you need to use GetCustomAttribute with the constructed generic type. For example, typeof(MyClass).GetCustomAttribute>(). This feature is part of .NET 7+ and C# 11. In trading systems, generic attributes can reduce runtime overhead by moving type checks to compile time, and they avoid boxing when storing value type information in attributes.

public class ValidatorAttribute<T> : Attribute
{
    public T Min { get; }
    public T Max { get; }

    public ValidatorAttribute(T min, T max)
    {
        Min = min;
        Max = max;
    }
}

[Validator<int>(Min = 0, Max = 100)]
public class Order
{
    public int Quantity { get; set; }
}

// Retrieval
var attr = typeof(Order).GetCustomAttribute<Validator<int>>();
Console.WriteLine($"Min: {attr.Min}, Max: {attr.Max}");