functional-csharp

name	functional-csharp
description	Use when applying functional patterns in C# — Result types, railway-oriented programming, or immutability.
when_to_use	Use when applying functional patterns in C# — Result types, railway-oriented programming, or immutability.
metadata	{"category":"core","agent":"dotnet-architect"}

Functional Programming in C#

C# is a multi-paradigm language. Use FP where it shines, OOP where it fits. This skill covers practical functional patterns that integrate naturally with idiomatic .NET code -- not Haskell or F# transplants.

Core Principles

• Immutability -- prefer data that cannot change after creation (records, init, readonly)
• Pure functions -- given the same inputs, always produce the same output with no side effects
• Composition over inheritance -- build complex behavior by chaining small functions, not deep class hierarchies
• Expressions over statements -- prefer switch expressions, ternary operators, and LINQ over imperative if/else blocks
• Explicit error handling -- use Result<T> to make failure a first-class return value, not an exception

When to Use

• Data transformation pipelines (mapping, filtering, projecting)
• Error handling that must compose across multiple steps (validation, orchestration)
• Validation chains where each rule is independent and combinable
• LINQ query composition (Select, Where, Aggregate)
• Pure business logic that is easy to unit test without mocking
• Configuration builders and fluent APIs

When NOT to Use

• Stateful UI component logic (Blazor components, WinForms) -- OOP fits better
• Complex mutation workflows where entities change through multiple lifecycle stages
• Teams unfamiliar with FP concepts -- the learning curve hurts more than the pattern helps
• Simple CRUD operations -- a service class with straightforward methods is clearer
• Performance-critical hot paths where allocations from closures and delegates matter
• EF Core tracking scenarios that rely on mutable entity state

Patterns

Result Pattern

A lightweight discriminated union that makes success and failure explicit in the type system. No exceptions for expected failures.

public sealed record Result<T>
{
    private readonly T? _value;
    private readonly string? _error;

    private Result(T value) { _value = value; IsSuccess = true; }
    private Result(string error) { _error = error; IsSuccess = false; }

    public bool IsSuccess { get; }
    public bool IsFailure => !IsSuccess;
    public T Value => IsSuccess ? _value! : throw new InvalidOperationException(_error);
    public string Error => IsFailure ? _error! : throw new InvalidOperationException("No error");

    public static Result<T> Success(T value) => new(value);
    public static Result<T> Failure(string error) => new(error);

    public TOut Match<TOut>(Func<T, TOut> onSuccess, Func<string, TOut> onFailure) =>
        IsSuccess ? onSuccess(_value!) : onFailure(_error!);
}

Usage:

public Result<Order> CreateOrder(CreateOrderRequest request)
{
    if (string.IsNullOrWhiteSpace(request.CustomerName))
        return Result<Order>.Failure("Customer name is required");

    if (request.Items.Count == 0)
        return Result<Order>.Failure("Order must have at least one item");

    var order = new Order(Guid.NewGuid(), request.CustomerName, request.Items);
    return Result<Order>.Success(order);
}

// Caller uses Match to handle both paths
var response = CreateOrder(request).Match(
    onSuccess: order => Results.Created($"/orders/{order.Id}", order),
    onFailure: error => Results.BadRequest(error));

Railway-Oriented Programming

Chain operations that each return Result<T>. If any step fails, the pipeline short-circuits and carries the error forward -- like a train switching to the failure track.

public static class ResultExtensions
{
    public static Result<TOut> Map<TIn, TOut>(
        this Result<TIn> result, Func<TIn, TOut> map) =>
        result.IsSuccess ? Result<TOut>.Success(map(result.Value)) : Result<TOut>.Failure(result.Error);

    public static Result<TOut> Bind<TIn, TOut>(
        this Result<TIn> result, Func<TIn, Result<TOut>> bind) =>
        result.IsSuccess ? bind(result.Value) : Result<TOut>.Failure(result.Error);

    public static async Task<Result<TOut>> BindAsync<TIn, TOut>(
        this Result<TIn> result, Func<TIn, Task<Result<TOut>>> bind) =>
        result.IsSuccess ? await bind(result.Value) : Result<TOut>.Failure(result.Error);
}

Chaining a multi-step workflow:

public async Task<Result<OrderConfirmation>> ProcessOrderAsync(CreateOrderRequest request)
{
    return ValidateRequest(request)
        .Bind(CalculatePricing)
        .Bind(CheckInventory)
        .Bind(ApplyDiscounts)
        .Map(order => new OrderConfirmation(order.Id, order.Total));
}
// If ValidateRequest fails, the rest never execute -- error propagates automatically.

Pure Functions

A pure function has no side effects and always returns the same result for the same input. Pure functions are trivially testable and safe to parallelize.

// PURE -- no side effects, deterministic, testable
public static decimal CalculateDiscount(decimal subtotal, CustomerTier tier) =>
    tier switch
    {
        CustomerTier.Gold when subtotal > 500 => subtotal * 0.15m,
        CustomerTier.Gold => subtotal * 0.10m,
        CustomerTier.Silver => subtotal * 0.05m,
        _ => 0m
    };

// IMPURE -- reads DateTime.Now, writes to database, throws exceptions
public decimal CalculateDiscount(Order order)
{
    var tier = _dbContext.Customers.Find(order.CustomerId)!.Tier;
    if (DateTime.Now.DayOfWeek == DayOfWeek.Friday)
        tier = CustomerTier.Gold; // surprise promotion
    _dbContext.SaveChanges();
    return order.Total * GetRate(tier);
}

Guidelines for pure functions:

• Accept all data as parameters -- do not read from fields, databases, or environment
• Return the result -- do not mutate input objects or external state
• Use static methods when possible to signal purity
• Push I/O to the edges: impure shell calls pure core

Immutability

Immutable data eliminates entire categories of bugs: race conditions, unintended aliasing, and stale state.

// Records are immutable by default
public sealed record OrderLine(string ProductName, int Quantity, decimal UnitPrice)
{
    public decimal Total => Quantity * UnitPrice;
}

// Use 'with' expressions to create modified copies
var updated = originalLine with { Quantity = 5 };

// Init-only properties on classes
public sealed class AppSettings
{
    public required string ConnectionString { get; init; }
    public required int MaxRetries { get; init; }
    public TimeSpan Timeout { get; init; } = TimeSpan.FromSeconds(30);
}

// Immutable collections from System.Collections.Immutable
using System.Collections.Immutable;

ImmutableList<OrderLine> lines = [..initialLines];
ImmutableList<OrderLine> withNew = lines.Add(new OrderLine("Widget", 2, 9.99m));
// 'lines' is unchanged -- 'withNew' is a new list

// ReadOnlyCollection for API boundaries
public IReadOnlyList<OrderLine> Lines => _lines.AsReadOnly();

When to use which immutable type:

Type	Use When
record	DTOs, value objects, command/query messages
init properties	Configuration objects, builder results
ImmutableList<T>	Collections that grow/shrink in pure pipelines
IReadOnlyList<T>	Public API surface to prevent external mutation
ReadOnlySpan<T>	High-performance slicing without allocation
FrozenSet<T> / FrozenDictionary<K,V>	Lookup collections built once, read many (.NET 8+)

Pattern Matching

Pattern matching replaces verbose type checks and conditional chains with concise, exhaustive expressions.

// Switch expression with multiple pattern types
public static string Classify(object value) => value switch
{
    null => "nothing",
    int n when n < 0 => "negative integer",
    int n => $"integer: {n}",
    string { Length: 0 } => "empty string",
    string s => $"string: {s}",
    IEnumerable<int> nums => $"collection of {nums.Count()} ints",
    _ => $"unknown: {value.GetType().Name}"
};

// Property patterns for domain logic
public static decimal CalculateShipping(Order order) => order switch
{
    { Total: >= 100, ShippingAddress.Country: "US" } => 0m,      // free US shipping over $100
    { Total: >= 100 } => 9.99m,                                   // discounted international
    { ShippingAddress.Country: "US" } => 5.99m,                   // standard US
    _ => 19.99m                                                    // standard international
};

// Relational and logical patterns
public static string RateTemperature(double celsius) => celsius switch
{
    < 0 => "freezing",
    >= 0 and < 15 => "cold",
    >= 15 and < 25 => "comfortable",
    >= 25 and < 35 => "warm",
    >= 35 => "hot"
};

// List patterns (C# 11+)
public static string DescribeRoute(string[] segments) => segments switch
{
    ["api", var version, .. var rest] => $"API {version}: /{string.Join('/', rest)}",
    ["admin", ..] => "Admin area",
    [var single] => $"Root: {single}",
    [] => "Home"
};

LINQ as Functional Composition

LINQ is the most natural FP surface in C#. Chain small, focused transformations instead of writing imperative loops.

// Functional pipeline: filter, transform, aggregate
public static OrderSummary Summarize(IEnumerable<Order> orders)
{
    var completedOrders = orders
        .Where(o => o.Status == OrderStatus.Delivered)
        .Where(o => o.CompletedAt >= DateOnly.FromDateTime(DateTime.UtcNow).AddMonths(-3))
        .Select(o => new
        {
            o.Id,
            o.CustomerName,
            LineCount = o.Lines.Count,
            Total = o.Lines.Sum(l => l.Quantity * l.UnitPrice)
        })
        .OrderByDescending(o => o.Total)
        .ToList();

    return new OrderSummary(
        OrderCount: completedOrders.Count,
        TotalRevenue: completedOrders.Sum(o => o.Total),
        AverageOrderValue: completedOrders.Count > 0
            ? completedOrders.Average(o => o.Total)
            : 0m,
        TopCustomer: completedOrders.FirstOrDefault()?.CustomerName ?? "N/A");
}

// Aggregate for custom reductions
var csv = names.Aggregate(
    seed: new StringBuilder(),
    func: (sb, name) => sb.Length == 0 ? sb.Append(name) : sb.Append(',').Append(name),
    resultSelector: sb => sb.ToString());

OOP vs FP Decision Matrix

Scenario	Prefer OOP	Prefer FP	Why
Error handling across steps		X	Result<T> with Bind chains composes cleanly without try/catch nesting
Entity with rich behavior	X		Domain entities encapsulate state + invariants in methods
Data transformation pipeline		X	LINQ/Select chains express intent declaratively
State machine with transitions	X		Class with methods and guards models lifecycle clearly
Validation pipeline		X	Chain independent validators, collect all errors
UI component with lifecycle	X		Stateful components need mutable properties and event handlers
Business rule calculation		X	Pure functions are trivially testable and composable
Repository / data access	X		Interfaces + DI fit the OOP service pattern naturally
Configuration building		X	Immutable records with with expressions, init properties
Event sourcing projections		X	Fold events into state with pure aggregate functions

Anti-Patterns

Problem	Why It Hurts	Correct Approach
Wrapping every method in Result<T>	Noise -- not every operation can fail meaningfully	Use Result<T> for operations with expected domain failures; use exceptions for truly exceptional cases
Deeply nested Match calls	Unreadable pyramid of doom	Use Bind/Map railway chaining instead
Impure functions disguised as pure	Hidden database calls or DateTime.Now break testability	Extract I/O to parameters; inject time as TimeProvider
Mutable fields inside records	Records look immutable but mutate through reference fields	Use ImmutableList<T> or IReadOnlyList<T> for collection properties
Forcing FP on the entire codebase	Team confusion, inconsistent style	Apply FP to data pipelines and business logic; keep OOP for services and infrastructure
Using exceptions for control flow	Expensive, invisible flow, no composition	Return Result<T> from validation and business operations
Ignoring Result.Error	Silent failures that propagate bad state	Always Match or Bind -- never access .Value without checking
Reimplementing a full monad library	Over-engineering for C# -- not Haskell	Keep Result<T> simple; add Map/Bind/BindAsync and stop

Detect Existing Patterns

1. Search for Result< or Result<T> types in the codebase
2. Look for OneOf, ErrorOr, or FluentResults NuGet packages in .csproj files
3. Check for record usage in domain or DTO layers
4. Scan for LINQ chains in service/handler classes
5. Look for ImmutableList, ImmutableDictionary usage
6. Check for existing railway-style extension methods (Bind, Map, Then)
7. If a Result library exists, use it -- do not introduce a competing type

Adding to Existing Project

1. Start with Result -- introduce for new validation and business operations
2. Convert DTOs to records -- immutable by default, with expressions for copies
3. Extract pure functions -- pull calculation logic out of services into static methods
4. Chain with LINQ -- replace imperative loops with Select/Where/Aggregate pipelines
5. Add Map/Bind extensions -- enable railway-oriented composition on Result
6. Use ImmutableList where collections must not be modified after creation
7. Keep services as classes -- DI, logging, and I/O stay in OOP-style service classes
8. Do not mix Result libraries -- pick one (ErrorOr, FluentResults, or custom) and standardize

References

• https://learn.microsoft.com/en-us/dotnet/csharp/fundamentals/functional/pattern-matching
• https://learn.microsoft.com/en-us/dotnet/architecture/microservices/microservice-ddd-cqrs-patterns/
• https://learn.microsoft.com/en-us/dotnet/api/system.collections.immutable