CoreDesign.Data 1.1.0

CoreDesign.Data

A generic, reusable Entity Framework Core data access layer providing base entity infrastructure, repository abstractions, automatic audit field management, and a migration worker base class for .NET projects.

Requirements

• .NET 10.0
• Microsoft.EntityFrameworkCore 10.x
• Microsoft.EntityFrameworkCore.SqlServer 10.x
• Microsoft.Extensions.Hosting.Abstractions 10.x
• Ulid 1.4.x

What Is Included

Infrastructure

• BaseEntity - Base class all entities must inherit from. Provides Id (Ulid, auto-generated), CreatedAt, UpdatedAt, CreatedBy, UpdatedBy, and IsDeleted audit fields.
• BaseEntityConfiguration<T> - EF Core IEntityTypeConfiguration<T> base that wires up primary key, index, soft-delete query filter, and required audit field constraints.
• CoreDesignDbContext - Abstract DbContext base class that wires the AuditInterceptor automatically. Your DbContext must inherit from this instead of DbContext directly.
• AuditInterceptor - EF Core SaveChangesInterceptor that sets audit fields on every save. On insert it sets CreatedAt, CreatedBy, UpdatedAt, and UpdatedBy (skipped if CreatedAt is already set, allowing seeded data to preserve explicit values). On update it sets UpdatedAt and UpdatedBy. CreatedBy is never modified on update.
• ICurrentUserAccessor - Interface the consuming application implements to supply the current user's ID to the interceptor. Backed by IHttpContextAccessor in web apps.
• SystemUserAccessor - Built-in ICurrentUserAccessor implementation that returns Guid.Empty. Use this in migration services and test projects where there is no authenticated user.
• CoreDesignDataExtensions - Provides AddCoreDesignData<TCurrentUserAccessor>() to register ICurrentUserAccessor and AuditInterceptor in one call.
• ValueConverters - Provides GetUlidConverter() (Ulid to string) and GetEnumConverter<TEnum>() (enum to string) for use in entity configurations.
• MigrationWorker<TContext> - Concrete BackgroundService that ensures the database exists, applies pending EF Core migrations, seeds from JSON files in the configured seed directory (default: SeedData), then stops the host. No subclassing is required. Override the virtual SeedAsync method only when custom seed logic is needed.

Interfaces

• IReadRepository<TContext, T> - Read-only repository interface with GetAllAsync, GetAllAttachedAsync, GetAsync, and GetAttachedAsync.
• ICudRepository<TContext, T> - Create/Update/Delete repository interface. Methods: InsertAsync, InsertRangeAsync, UpdateAsync, UpdateRangeAsync, SoftDeleteAsync, SoftDeleteRangeAsync, SoftDeleteCascadeAsync, HardDeleteAsync, HardDeleteCascadeAsync.

Repositories

• ReadRepository<TContext, T> - Concrete read repository. All queries use AsNoTracking() by default. Supports optional where expressions, orderBy, and strongly typed includes.
• CudRepository<TContext, T> - Concrete CUD repository. Soft-delete methods set IsDeleted = true rather than removing rows. Hard-delete methods physically remove rows.

Setup

1. Install the NuGet package

Using the .NET CLI:

dotnet add package CoreDesign.Data

Or add directly to your .csproj:

<ItemGroup>
    <PackageReference Include="CoreDesign.Data" Version="*" />
</ItemGroup>

2. Define your entities

All entities must inherit from BaseEntity. The Id property is automatically initialized to a new Ulid at construction time:

public class Widget : BaseEntity
{
    public string Name { get; set; } = string.Empty;
}

3. Configure your entities

Inherit from BaseEntityConfiguration<T> and call base.Configure(builder) to apply the standard audit field and query filter setup:

public class WidgetConfiguration : BaseEntityConfiguration<Widget>
{
    public override void Configure(EntityTypeBuilder<Widget> builder)
    {
        base.Configure(builder);
        builder.Property(e => e.Name).IsRequired().HasMaxLength(100);
    }
}

4. Implement ICurrentUserAccessor

Create one implementation per application. For a web API, read the oid claim from the current HTTP context:

public class AppCurrentUserAccessor(IHttpContextAccessor httpContextAccessor) : ICurrentUserAccessor
{
    public Guid UserId
    {
        get
        {
            var oid = httpContextAccessor.HttpContext?.User.FindFirst("oid")?.Value;
            return Guid.TryParse(oid, out var id) ? id : Guid.Empty;
        }
    }
}

For migration services and background workers where there is no HTTP context, use the built-in SystemUserAccessor (returns Guid.Empty) or provide your own implementation.

5. Register services

Call AddCoreDesignData<T> once in Program.cs, before the DbContext is registered. Also register IHttpContextAccessor if your accessor depends on it. Then wire the AuditInterceptor when registering the DbContext:

builder.Services.AddHttpContextAccessor();
builder.Services.AddCoreDesignData<AppCurrentUserAccessor>();

builder.Services.AddDbContextPool<AppDbContext>((sp, options) =>
    options.UseSqlServer(connectionString)
        .AddInterceptors(sp.GetRequiredService<AuditInterceptor>()));

6. Inherit from CoreDesignDbContext

Your DbContext must inherit from CoreDesignDbContext instead of DbContext directly:

public class AppDbContext(DbContextOptions<AppDbContext> options)
    : CoreDesignDbContext(options)
{
    public DbSet<Widget> Widgets { get; set; }

    protected override void OnModelCreating(ModelBuilder modelBuilder)
    {
        modelBuilder.HasDefaultSchema("app");
        modelBuilder.ApplyConfigurationsFromAssembly(typeof(AppDbContext).Assembly);
    }
}

ApplyConfigurationsFromAssembly scans the assembly for every class that implements IEntityTypeConfiguration<T> and calls Configure on each one automatically. No manual registration is needed when new entity configuration classes are added.

7. Register repositories

Register the repositories in your DI container, once per entity type:

services.AddTransient<IReadRepository<AppDbContext, Widget>, ReadRepository<AppDbContext, Widget>>();
services.AddTransient<ICudRepository<AppDbContext, Widget>, CudRepository<AppDbContext, Widget>>();

Usage

Querying

// All widgets, no tracking
var all = await readRepository.GetAllAsync();

// Filtered and ordered, with related entities included
var active = await readRepository.GetAllAsync(
    whereExpression: w => w.Name.StartsWith("A"),
    orderBy: q => q.OrderBy(w => w.Name),
    includes: q => q.Include(w => w.Parts).ThenInclude(p => p.Supplier));

// Single entity
var widget = await readRepository.GetAsync(w => w.Id == id);

Writing

Audit fields (Id, CreatedAt, CreatedBy, UpdatedAt, UpdatedBy) are set automatically by the interceptor. No userId parameter is needed on any repository method:

// Insert
var widget = new Widget { Name = "Sprocket" };
await cudRepository.InsertAsync(widget, cancellationToken);

// Update
widget.Name = "Updated Sprocket";
await cudRepository.UpdateAsync(widget, cancellationToken);

// Soft delete (sets IsDeleted = true, row is not removed)
await cudRepository.SoftDeleteAsync(id, cancellationToken);

// Soft delete a parent and all child entities in the EF navigation graph
await cudRepository.SoftDeleteCascadeAsync(id, cancellationToken);

// Hard delete (physically removes the row; finds the entity even if soft-deleted)
await cudRepository.HardDeleteAsync(id, cancellationToken);

// Hard delete a parent and all child entities in the EF navigation graph
await cudRepository.HardDeleteCascadeAsync(id, cancellationToken);

Notes

• BaseEntityConfiguration applies a global query filter (e => !e.IsDeleted) to every entity. Soft-deleted rows are automatically excluded from all queries.
• HardDeleteAsync uses IgnoreQueryFilters() internally so it can remove rows that were previously soft-deleted.
• Cascade methods walk EF Core's navigation graph in memory. They load child collections before marking or removing them, so large graphs increase memory usage proportionally.
• GetAllAttachedAsync and GetAttachedAsync return tracked entities for use when you need EF Core to detect changes without an explicit Attach call.
• Seeded entities whose JSON includes explicit audit values (non-default CreatedAt) are preserved as-is. The interceptor only fills in audit fields when CreatedAt is at its default value.

Migration Worker

MigrationWorker<TContext> is a concrete BackgroundService that works in any .NET host application, including .NET Aspire migration services. It runs three steps in order when the host starts:

1. Ensure database — creates the database if it does not exist.
2. Migrate — applies all pending EF Core migrations via MigrateAsync.
3. Seed — scans the seed directory for *.json files and inserts any records that do not yet exist in the database.

When all steps complete, it calls IHostApplicationLifetime.StopApplication() and the process exits with code 0. If any step throws, the exception propagates and the process exits with a non-zero code, which blocks deployment pipelines from proceeding.

Both the ensure and migrate steps wrap their database calls in CreateExecutionStrategy() so transient SQL Server errors are retried automatically.

Registration

Register AddCoreDesignData before AddMigrationWorker. Migration services have no authenticated user, so use SystemUserAccessor:

builder.Services.AddCoreDesignData<SystemUserAccessor>();
builder.AddMigrationWorker<AppDbContext>();

builder.Services.AddOpenTelemetry()
    .WithTracing(t => t.AddSource(MigrationWorker<AppDbContext>.ActivitySourceName));

Seed file naming convention

By default, MigrationWorker<TContext> scans a directory named SeedData (relative to the working directory) for *.json files. Each file must be named after the fully qualified type name of the entity it seeds. The worker resolves the filename (without the .json extension) to a type in typeof(TContext).Assembly and deserializes the JSON as a List<T>. Files whose name cannot be resolved to a BaseEntity subclass are skipped with a warning.

For an entity MyApp.Orders.Models.Order the seed file must be named:

MyApp.Orders.Models.Order.json

Seed files may include all BaseEntity audit fields (Id, CreatedAt, CreatedBy, UpdatedAt, UpdatedBy, IsDeleted). When explicit values are present the interceptor leaves them unchanged. When they are absent the interceptor fills them in with Guid.Empty (the SystemUserAccessor value) and the current timestamp.

Overriding the seed directory

Pass a custom directory path as the second argument to AddMigrationWorker:

builder.AddMigrationWorker<AppDbContext>("ReferenceData");

The path is relative to the working directory. Absolute paths are also accepted. If the directory does not exist at runtime the worker logs a warning and skips seeding without throwing.

SeedEntitiesAsync

SeedEntitiesAsync<T> is a protected helper. It accepts a deserialized sequence of entities and inserts any that do not yet exist in the database, identified by BaseEntity.Id. Existence is checked with IgnoreQueryFilters() so soft-deleted rows count as existing and no duplicate-key errors occur on re-runs:

protected async Task SeedEntitiesAsync<T>(
    TContext dbContext,
    IEnumerable<T> entities,
    CancellationToken cancellationToken)
    where T : BaseEntity

The writes are wrapped in CreateExecutionStrategy() for the same transient-error retry behaviour as the migration steps.

Custom seed logic

Override SeedAsync to replace or extend the default directory-scanning behavior. Call SeedFromDirectoryAsync or SeedEntitiesAsync as needed:

public class AppMigrationWorker(
    IServiceProvider serviceProvider,
    IHostApplicationLifetime lifetime,
    ILogger<AppMigrationWorker> logger)
    : MigrationWorker<AppDbContext>(serviceProvider, lifetime, logger)
{
    protected override async Task SeedAsync(AppDbContext dbContext, CancellationToken ct)
    {
        // Seed from the default directory first, then apply supplemental data.
        await SeedFromDirectoryAsync(dbContext, "SeedData", typeof(AppDbContext).Assembly, ct);
        await SeedEntitiesAsync(dbContext, GetAdminUsers(), ct);
    }
}

Register the subclass directly as a hosted service:

builder.Services.AddHostedService(sp =>
    new AppMigrationWorker(
        sp,
        sp.GetRequiredService<IHostApplicationLifetime>(),
        sp.GetRequiredService<ILogger<AppMigrationWorker>>()));

PurgeBeforeSeed

By default the worker only inserts rows that do not yet exist (identified by BaseEntity.Id). This is safe to re-run as many times as needed but it means updated seed content is never pushed to a database that already has rows for that entity.

Pass entity type names to purgeBeforeSeed to clear specific tables before their seed file is applied. This lets you replace the content of a table on every migration run without having to manually delete rows first.

Short class names and fully qualified type names are both accepted:

builder.AddMigrationWorker<AppDbContext>(purgeBeforeSeed: ["SiteContent", "PolicyText"]);

The list is additive: tables not in the list use the default insert-only behavior. Purge uses ExecuteDeleteAsync on relational providers (one round-trip, no change-tracker overhead) and falls back to a load-then-remove approach on non-relational providers (in-memory databases used in tests).

Configuring PurgeBeforeSeed at runtime

The purge list can also be driven entirely from configuration so you can change it between runs without redeploying. Add a MigrationWorker:PurgeBeforeSeed array to appsettings.json:

{
  "MigrationWorker": {
    "PurgeBeforeSeed": [ "SiteContent" ]
  }
}

Or via environment variable (useful in Docker and Aspire):

MigrationWorker__PurgeBeforeSeed__0=SiteContent

Config-driven names and names passed directly to AddMigrationWorker are merged into a single set, so you can combine both. Remove an entry from config and the table reverts to insert-only behavior on the next run.

Running outside Aspire

MigrationWorker<TContext> has no dependency on the Aspire AppHost. It works in any hosted environment as long as a TContext is registered in the DI container. In a GitHub Actions pipeline, supply the connection string as an environment variable and run the migration project with dotnet run:

- name: Run migrations
  env:
    ConnectionStrings__my-db: ${{ secrets.AZURE_SQL_CONNECTION_STRING }}
  run: dotnet run --project src/MyApp.MigrationService --configuration Release --no-build

The process exits 0 on success and non-zero on failure, making it safe to use as a deployment gate step.

Feedback

Feedback on this package is welcome. If you run into a missing feature, an unexpected behavior, or something that required more effort than it should have, open an issue at github.com/codyskidmore/CoreDesign/issues or tag @codyskidmore. Suggestions about missing features and priority input are especially appreciated.