C# and .NET

Purpose

Build .NET applications that are null-safe by contract, correctly asynchronous end to end, and free of the ORM traps that quietly turn one query into a thousand.

When to Use

• Writing or reviewing C# on .NET 8 or later.
• Enabling nullable reference types on an existing project.
• Designing minimal APIs or ASP.NET Core services.
• Diagnosing async deadlocks or thread-pool starvation.
• Fixing EF Core N+1 queries and tracking overhead.

Capabilities

• Nullable reference types, records, pattern matching, required members.
• Async correctness: ConfigureAwait, cancellation tokens, IAsyncEnumerable.
• Minimal APIs, endpoint filters, model validation, problem details.
• Dependency injection lifetimes and scope correctness.
• EF Core: projections, split queries, no-tracking reads, compiled queries.

Inputs

• Solution or project files and target framework.
• Whether the project is a library, web API, or worker service.
• Data-access layer and database engine.

Outputs

• Code compiling with <Nullable>enable</Nullable> and <TreatWarningsAsErrors>true</TreatWarningsAsErrors>.
• Async paths that flow a CancellationToken from request to database.
• EF Core queries with explicit projection and tracking behavior.

Workflow

1. Enable the gates — Nullable reference types and warnings-as-errors, project-wide.
2. Model — Records for immutable values; required members instead of constructor sprawl.
3. Thread cancellation — Every async method takes a CancellationToken and passes it down.
4. Shape the queries — Project to DTOs; never materialize entities you will not mutate.
5. Gate — Build with warnings as errors, run analyzers, run the test suite.

Best Practices

• Never call .Result or .Wait() on a Task. That is how deadlocks and starvation happen — async all the way down.
• Use ConfigureAwait(false) in library code; it is unnecessary in ASP.NET Core application code.
• Register DbContext as scoped. Injecting it into a singleton is a correctness bug, not a style issue.
• Read-only queries use AsNoTracking(). Tracking every row you only intend to display wastes memory and time.
• Return Results.Problem(...) / RFC 7807 payloads rather than bare status codes.
• Validate at the endpoint, not in the domain. The domain should be able to assume its inputs are valid.

Examples

Minimal API endpoint with cancellation and projection:

app.MapGet("/orders/{id:guid}", async (
    Guid id,
    AppDbContext db,
    CancellationToken ct) =>
{
    var order = await db.Orders
        .AsNoTracking()
        .Where(o => o.Id == id)
        .Select(o => new OrderSummary(
            o.Id,
            o.Customer.Name,
            o.Lines.Count,
            o.Lines.Sum(l => l.Price * l.Quantity)))
        .SingleOrDefaultAsync(ct);

    return order is null
        ? Results.Problem(statusCode: 404, title: "Order not found")
        : Results.Ok(order);
});

Notes

• IAsyncEnumerable<T> streams results without buffering the full set — use it for large exports.
• EF Core's split-query mode avoids the cartesian explosion of multiple Includes, at the cost of extra round trips. Measure both.
• Source generators (e.g. System.Text.Json) remove reflection at startup and matter a great deal for AOT and cold-start latency.