AGENT SKILL

deniz-dotnet-general.csharp-concurrency-patterns

Choosing the right concurrency abstraction in .NET - from async/await for I/O to Channels for producer/consumer to Akka.NET for stateful entity management. Avoid locks and manual synchronization unless absolutely necessary.

License 许可
MIT
最近更新
2026-11-13
Stars
★ 1
Forks
0

01 前言 Front Matter

name
deniz-dotnet-general.csharp-concurrency-patterns
description
Choosing the right concurrency abstraction in .NET - from async/await for I/O to Channels for producer/consumer to Akka.NET for stateful entity management. Avoid locks and manual synchronization unless absolutely necessary.

02 资源文件 1

03 SKILL.md 正文

.NET Concurrency: Choosing the Right Tool

When to Use This Skill

Use this skill when:

  • Deciding how to handle concurrent operations in .NET
  • Evaluating whether to use async/await, Channels, Akka.NET, or other abstractions
  • Tempted to use locks, semaphores, or other synchronization primitives
  • Need to process streams of data with backpressure, batching, or debouncing
  • Managing state across multiple concurrent entities

Reference Files

The Philosophy

Start simple, escalate only when needed.

Most concurrency problems can be solved with async/await. Only reach for more sophisticated tools when you have a specific need that async/await can't address cleanly.

Try to avoid shared mutable state. The best way to handle concurrency is to design it away. Immutable data, message passing, and isolated state (like actors) eliminate entire categories of bugs.

Locks should be the exception, not the rule. When you can't avoid shared mutable state:

  1. First choice: Redesign to avoid it (immutability, message passing, actor isolation)
  2. Second choice: Use System.Collections.Concurrent (ConcurrentDictionary, etc.)
  3. Third choice: Use Channel<T> to serialize access through message passing
  4. Last resort: Use lock for simple, short-lived critical sections

Decision Tree

What are you trying to do?
│
├─► Wait for I/O (HTTP, database, file)?
│   └─► Use async/await
│
├─► Process a collection in parallel (CPU-bound)?
│   └─► Use Parallel.ForEachAsync
│
├─► Producer/consumer pattern (work queue)?
│   └─► Use System.Threading.Channels
│
├─► UI event handling (debounce, throttle, combine)?
│   └─► Use Reactive Extensions (Rx)
│
├─► Server-side stream processing (backpressure, batching)?
│   └─► Use Akka.NET Streams
│
├─► State machines with complex transitions?
│   └─► Use Akka.NET Actors (Become pattern)
│
├─► Manage state for many independent entities?
│   └─► Use Akka.NET Actors (entity-per-actor)
│
├─► Coordinate multiple async operations?
│   └─► Use Task.WhenAll / Task.WhenAny
│
└─► None of the above fits?
    └─► Ask yourself: "Do I really need shared mutable state?"
        ├─► Yes → Consider redesigning to avoid it
        └─► Truly unavoidable → Use Channels or Actors to serialize access

Level 1: async/await (Default Choice)

Use for: I/O-bound operations, non-blocking waits, most everyday concurrency.

// Simple async I/O
public async Task<Order> GetOrderAsync(string orderId, CancellationToken ct)
{
    var order = await _database.GetAsync(orderId, ct);
    var customer = await _customerService.GetAsync(order.CustomerId, ct);
    return order with { Customer = customer };
}

// Parallel async operations (when independent)
public async Task<Dashboard> LoadDashboardAsync(string userId, CancellationToken ct)
{
    var ordersTask = _orderService.GetRecentOrdersAsync(userId, ct);
    var notificationsTask = _notificationService.GetUnreadAsync(userId, ct);
    var statsTask = _statsService.GetUserStatsAsync(userId, ct);

    await Task.WhenAll(ordersTask, notificationsTask, statsTask);

    return new Dashboard(
        Orders: await ordersTask,
        Notifications: await notificationsTask,
        Stats: await statsTask);
}

Key principles: Always accept CancellationToken. Use ConfigureAwait(false) in library code. Don't block on async code.


Level 2: Parallel.ForEachAsync (CPU-Bound Parallelism)

Use for: Processing collections in parallel when work is CPU-bound or you need controlled concurrency.

public async Task ProcessOrdersAsync(
    IEnumerable<Order> orders,
    CancellationToken ct)
{
    await Parallel.ForEachAsync(
        orders,
        new ParallelOptions
        {
            MaxDegreeOfParallelism = Environment.ProcessorCount,
            CancellationToken = ct
        },
        async (order, token) =>
        {
            await ProcessOrderAsync(order, token);
        });
}

When NOT to use: Pure I/O operations, when order matters, when you need backpressure.


Level 3: System.Threading.Channels (Producer/Consumer)

Use for: Work queues, producer/consumer patterns, decoupling producers from consumers.

public class OrderProcessor
{
    private readonly Channel<Order> _channel;

    public OrderProcessor()
    {
        _channel = Channel.CreateBounded<Order>(new BoundedChannelOptions(100)
        {
            FullMode = BoundedChannelFullMode.Wait
        });
    }

    // Producer
    public async Task EnqueueOrderAsync(Order order, CancellationToken ct)
    {
        await _channel.Writer.WriteAsync(order, ct);
    }

    // Consumer (run as background task)
    public async Task ProcessOrdersAsync(CancellationToken ct)
    {
        await foreach (var order in _channel.Reader.ReadAllAsync(ct))
        {
            await ProcessOrderAsync(order, ct);
        }
    }

    public void Complete() => _channel.Writer.Complete();
}

Channels are good for: Decoupling speed, buffering with backpressure, fan-out to workers, background queues.

Channels are NOT good for: Complex stream operations (batching, windowing), stateful per-entity processing, sophisticated supervision.


Level 4+: Akka.NET Streams, Reactive Extensions, Actors

For advanced scenarios requiring stream processing, UI event composition, or stateful entity management, see advanced-concurrency.md.

Akka.NET Streams excel at server-side batching, throttling, and backpressure. Reactive Extensions are ideal for UI event composition. Akka.NET Actors handle entity-per-actor patterns, state machines with Become(), and distributed systems via Cluster Sharding.


Anti-Patterns: What to Avoid

Locks for Business Logic

// BAD: Using locks to protect shared state
private readonly object _lock = new();
private Dictionary<string, Order> _orders = new();

public void UpdateOrder(string id, Action<Order> update)
{
    lock (_lock) { if (_orders.TryGetValue(id, out var order)) update(order); }
}

// GOOD: Use an actor or Channel to serialize access

Manual Thread Management

// BAD: Creating threads manually
var thread = new Thread(() => ProcessOrders());
thread.Start();

// GOOD: Use Task.Run or better abstractions
_ = Task.Run(() => ProcessOrdersAsync(cancellationToken));

Blocking in Async Code

// BAD: Blocking on async - deadlock risk!
var result = GetDataAsync().Result;

// GOOD: Async all the way
var result = await GetDataAsync();

Shared Mutable State Without Protection

// BAD: Multiple tasks mutating shared state
var results = new List<Result>();
await Parallel.ForEachAsync(items, async (item, ct) =>
{
    var result = await ProcessAsync(item, ct);
    results.Add(result); // Race condition!
});

// GOOD: Use ConcurrentBag
var results = new ConcurrentBag<Result>();

Quick Reference: Which Tool When?

Need Tool Example
Wait for I/O async/await HTTP calls, database queries
Parallel CPU work Parallel.ForEachAsync Image processing, calculations
Work queue Channel<T> Background job processing
UI events with debounce/throttle Reactive Extensions Search-as-you-type, auto-save
Server-side batching/throttling Akka.NET Streams Event aggregation, rate limiting
State machines Akka.NET Actors Payment flows, order lifecycles
Entity state management Akka.NET Actors Order management, user sessions
Fire multiple async ops Task.WhenAll Loading dashboard data
Race multiple async ops Task.WhenAny Timeout with fallback
Periodic work PeriodicTimer Health checks, polling

The Escalation Path

async/await (start here)
    │
    ├─► Need parallelism? → Parallel.ForEachAsync
    │
    ├─► Need producer/consumer? → Channel<T>
    │
    ├─► Need UI event composition? → Reactive Extensions
    │
    ├─► Need server-side stream processing? → Akka.NET Streams
    │
    └─► Need state machines or entity management? → Akka.NET Actors

Only escalate when you have a concrete need. Don't reach for actors or streams "just in case".

04 同仓库更多 Skills

Blind-Striker
Blind-Striker/agent-skills-and-plugins

writing-tunit-tests

Write tests in TUnit — the attributes, the async assertion model, data-driven cases, lifecycle hooks, and the parallelism rules. USE FOR writing or fixing a test in a TUnit project, translating a test idea into TUnit syntax, or reviewing TUnit test code for API mistakes. DO NOT USE FOR deciding what to test, judging whether a test is any good, or running tests.

★ 1
Blind-Striker
Blind-Striker/agent-skills-and-plugins

dotnet-walls

Walls for a .NET change in a repository without its own quality gates — pinned, ephemeral analyzer, format, clone and scanner commands for the verify step, their identity ratchet, and the build traps. Use during a walls preflight, or when a worker's verify step needs .NET quality gates.

★ 1
Blind-Striker
Blind-Striker/agent-skills-and-plugins

csharp-15

Write, review and modernize C# 15 (.NET 11) code correctly. Covers every C# 15 feature - collection expression arguments (`[with(...), ...]`), union types (`union`, `[Union]`, `IUnion`, custom unions), closed class hierarchies (`closed`), extension indexers, labeled `break`/`continue`, non-virtual static interface members on runtimes without default interface methods, and the preview memory-safety model (`unsafe(expr)`, `safe`, requires-unsafe members). Use when code targets net11.0 or LangVersion 15/preview, when the user mentions any of these features, when replacing marker interfaces, OneOf-style results, `goto`/flag loop exits or capacity/comparer boilerplate, or when the build reports CS8121, CS8509, CS8655, CS8510, CS0121/CS0411 on a `with(...)` element, CS9370, CS9380 or CS9363.

★ 1
Blind-Striker
Blind-Striker/agent-skills-and-plugins

verified-waves

Run one task as verified multi-agent waves. Plans the fan-out, sends findings to fresh skeptics, keeps a ledger, and reports N of M done; small tasks stay solo.

★ 1
Blind-Striker
Blind-Striker/agent-skills-and-plugins

harness

The bundled dispatch harness, which runs a task file on worker models through OpenCode or Codex, fences their writes, escalates their asks to the operator, settles each result from git and its verify commands, and records every step in a ledger. Its commands, task file, modes and write fence, configuration, runs and their recovery, ledger and exit codes.

★ 1
Blind-Striker
Blind-Striker/agent-skills-and-plugins

delegating-to-worker-models

Use when you hand off implementation, a plan, research a decision will rely on, or a review, and you brief the worker and check what comes back. A subagent that searches, reads in parallel, or keeps your context small is ordinary tool use. Covers what stays with you, the mechanism, assigning work by strength, the walls preflight, prompts without contradictions, the worker contract and model profiles, evidence from your own run, review, and the fix-up loop.

★ 1