error-handling-patterns（错误处理模式）

Error Handling Patterns

Build resilient applications with robust error handling strategies.

Error Handling Philosophies

.NET Exception Patterns

Custom Exception Hierarchy

public class ApplicationException : Exception
{
    public string Code { get; }
    public ApplicationException(string message, string code) : base(message)
    {
        Code = code;
    }
}

public class ValidationException : ApplicationException
{
    public ValidationException(string message)
        : base(message, "VALIDATION_ERROR") { }
}

public class NotFoundException : ApplicationException
{
    public NotFoundException(string resource, Guid id)
        : base($"{resource} not found: {id}", "NOT_FOUND") { }
}

ABP BusinessException

// Use ABP's BusinessException for domain errors
throw new BusinessException(
    code: ClinicManagementSystemDomainErrorCodes.PatientNotFound,
    message: "Patient not found")
    .WithData("PatientId", patientId);

Result Type Pattern

type Result<T, E = Error> =
    | { ok: true; value: T }
    | { ok: false; error: E };

function Ok<T>(value: T): Result<T, never> {
    return { ok: true, value };
}

function Err<E>(error: E): Result<never, E> {
    return { ok: false, error };
}

// Usage
function parseJSON<T>(json: string): Result<T, SyntaxError> {
    try {
        return Ok(JSON.parse(json) as T);
    } catch (error) {
        return Err(error as SyntaxError);
    }
}

.NET Resilience with Polly

HTTP Retry with Exponential Backoff

public IAsyncPolicy<HttpResponseMessage> BuildHttpRetryPolicy(int retryCount = 3)
{
    return HttpPolicyExtensions
        .HandleTransientHttpError()
        .Or<TimeoutException>()
        .WaitAndRetryAsync(
            retryCount: retryCount,
            sleepDurationProvider: retryAttempt =>
            {
                var exponentialDelay = TimeSpan.FromSeconds(Math.Pow(2, retryAttempt));
                var jitter = TimeSpan.FromMilliseconds(Random.Shared.Next(0, 1000));
                return exponentialDelay + jitter;
            },
            onRetryAsync: async (outcome, timespan, retryAttempt, context) =>
            {
                _logger.LogWarning(
                    "[Retry {Attempt}/{Total}] Waiting: {Delay:F2}s",
                    retryAttempt, retryCount, timespan.TotalSeconds);
            });
}

Database Retry for Transient Errors

public IAsyncPolicy BuildDatabaseRetryPolicy(int retryCount = 3)
{
    return Policy
        .Handle<DbUpdateConcurrencyException>()
        .Or<DbUpdateException>(ex =>
            ex.InnerException is NpgsqlException npgsqlEx &&
            IsTransientPostgresException(npgsqlEx))
        .WaitAndRetryAsync(
            retryCount: retryCount,
            sleepDurationProvider: retryAttempt =>
                TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)));
}

private static bool IsTransientPostgresException(NpgsqlException ex)
{
    var transientCodes = new[] { "40001", "40P01", "55P03", "57014", "53300", "08000" };
    return transientCodes.Contains(ex.SqlState);
}

Combined Policy (Retry + Circuit Breaker + Timeout)

public IAsyncPolicy<HttpResponseMessage> BuildResilientPolicy()
{
    var timeout = Policy.TimeoutAsync<HttpResponseMessage>(TimeSpan.FromSeconds(30));

    var retry = HttpPolicyExtensions
        .HandleTransientHttpError()
        .WaitAndRetryAsync(3, attempt => TimeSpan.FromSeconds(Math.Pow(2, attempt)));

    var circuitBreaker = HttpPolicyExtensions
        .HandleTransientHttpError()
        .CircuitBreakerAsync(
            handledEventsAllowedBeforeBreaking: 5,
            durationOfBreak: TimeSpan.FromSeconds(30));

    return Policy.WrapAsync(timeout, retry, circuitBreaker);
}

DI Registration

services.AddHttpClient<IMyApiClient, MyApiClient>()
    .AddPolicyHandler((provider, _) =>
    {
        var retryService = provider.GetRequiredService<IRetryPolicyService>();
        return retryService.BuildHttpRetryPolicy();
    });

Circuit Breaker Pattern

States:
  CLOSED  → Normal operation, tracking failures
  OPEN    → Failing, reject all requests
  HALF_OPEN → Testing recovery with limited requests

Flow:
  CLOSED --[failure threshold]--> OPEN
  OPEN --[timeout]--> HALF_OPEN
  HALF_OPEN --[success]--> CLOSED
  HALF_OPEN --[failure]--> OPEN

Graceful Degradation

// Polly Fallback Policy
public IAsyncPolicy<T> BuildFallbackPolicy<T>(Func<Task<T>> fallbackAction)
{
    return Policy<T>
        .Handle<Exception>()
        .FallbackAsync(
            fallbackAction: async (context, cancellationToken) =>
            {
                _logger.LogWarning("Primary operation failed, using fallback");
                return await fallbackAction();
            },
            onFallbackAsync: async (exception, context) =>
            {
                _logger.LogError(exception.Exception, "Fallback triggered");
            });
}

// Usage
var fallbackPolicy = BuildFallbackPolicy(() => FetchFromDatabaseAsync(userId));
var profile = await fallbackPolicy.ExecuteAsync(() => FetchFromCacheAsync(userId));

When to Retry

✅ Retry for:

• HTTP API calls (transient network errors)
• Database operations (deadlocks, connection timeouts)
• External service integrations
• File I/O operations

❌ Don't retry:

• Authentication failures
• Validation errors
• Business logic errors
• Non-idempotent operations without safeguards

Best Practices

1. Fail Fast - Validate input early
2. Preserve Context - Include stack traces, metadata
3. Meaningful Messages - Explain what and how to fix
4. Log Appropriately - Error = log, expected = don't spam
5. Handle at Right Level - Catch where you can meaningfully handle
6. Clean Up Resources - Use try-finally, using statements
7. Don't Swallow Errors - Log or re-throw, don't ignore
8. Type-Safe Errors - Use typed errors when possible

Detailed References

For comprehensive patterns, see:

• references/exception-hierarchy-design.md
• references/polly-advanced-patterns.md
• references/async-error-handling.md