Complete guide to implementing data persistence as an adapter in hexagonal architecture with entity mappings, repositories, Unit of Work, and migration workflows. Demonstrates one ORM approach; patterns apply to any data access technology.

version
1.0.0
GitHub stars
55

Install

/plugin install data-dotnet@dotnet-claude-code-skills

Run it inside Claude Code after adding the marketplace with /plugin marketplace add nesbo/dotnet-claude-code-skills.

Contents

Discovered from source

Skills 1

EF Core Data Persistence Layer - Hexagonal Architecture Adapter

Overview

This skill provides a comprehensive guide to implementing the data persistence layer using Entity Framework Core as an adapter in a hexagonal architecture (ports and adapters pattern). The data layer is a library that encapsulates its implementation internally and exposes DI registration publicly for wiring into applications.

This skill complements the ddd-dotnet-basics.md skill and assumes familiarity with Domain layer patterns.

Key Architecture Principles

┌─────────────────────────────────────────────┐
│        Domain (Ports)                       │
│  - IRepository<TAggregate, TKey>            │
│  - IQueryRepository<TAggregate>             │
│  - IUnitOfWork                              │
│  - Aggregates & Entities                    │
└─────────────────────────────────────────────┘
                     ▲
                     │ (implements)
                     │
┌─────────────────────────────────────────────┐
│        Data (Adapter - Infrastructure)      │
│  - DbContext                                │
│  - Entity Configurations                    │
│  - Repository Implementations               │
│  - UnitOfWork Implementation                │
│  - Service Registrations (DI)               │
└─────────────────────────────────────────────┘

Key Benefits:

  • Testability: In-memory database for unit/integration tests
  • Separation of Concerns: Domain knows nothing about EF Core
  • CQRS Support: Separate read/write repository base classes
  • Schema Organization: Multiple database schemas for bounded contexts
  • Convention-based Registration: Automatic repository discovery and registration

Project Structure

Pikot.LMP.Adapters.Data/
├── DataContext.cs                    # Main DbContext
├── UnitOfWork.cs                     # Transaction coordinator
├── ServiceRegistrations.cs           # DI setup (public API)
├── EntityConfigurations/             # Fluent configurations
│   ├── ConfigHelper.cs               # Reusable config utilities
│   ├── ShipperOrderConfiguration.cs
│   ├── PackageConfiguration.cs
│   └── ...
├── Repositories/                     # Repository implementations
│   ├── RepositoryBase.cs             # Write repository base
│   ├── QueryRepositoryBase.cs        # Read repository base
│   ├── ShipperOrderRepository.cs
│   ├── ShipperOrderQueryRepository.cs
│   └── ...
└── Migrations/                       # EF Core migrations
    └── ...

DbContext Configuration

Main DbContext

The DataContext is the central EF Core context with specific patterns for production and testing.

using Microsoft.EntityFrameworkCore;
using Microsoft.Extensions.DependencyInjection;
using Microsoft.Extensions.Logging;

namespace Pikot.LMP.Adapters.Data;

public class DataContext : DbContext
{
    public DataContext(DbContextOptions options) : base(options) { }

    public DataContext() { }  // Parameterless for design-time tools

    protected override void OnConfiguring(DbContextOptionsBuilder optionsBuilder)
    {
        // Fallback configuration for design-time tools (migrations)
        if (!optionsBuilder.IsConfigured)
        {
            optionsBuilder.UseNpgsql("Server=127.0.0.1;Database=lmp-nia;User Id=admin;Password=Password1!");
        }

        base.OnConfiguring(optionsBuilder);
    }

    protected override void OnModelCreating(ModelBuilder modelBuilder)
    {
        // Auto-discover and apply all IEntityTypeConfiguration<T> from assembly
        modelBuilder.ApplyConfigurationsFromAssembly(typeof(DataContext).Assembly);

        base.OnModelCreating(modelBuilder);
    }

    // Migration runner with retry logic (for application startup)
    public static async Task RunDbMigrationsAsync(IServiceProvider serviceProvider)
    {
        using var scope = serviceProvider.CreateScope();
        var logger = scope.ServiceProvider.GetRequiredService<ILogger<DataContext>>();

        const int maxAttempts = 5;
        var attempt = 1;
        var success = false;
        Exception? startupException = null;

        while (attempt < maxAttempts && !success)
        {
            try
            {
                logger.LogInformation("Attempting db migration run on startup. Attempt [{StartupMigrationAttempt}]",
                    attempt);
                using var dbScope = scope.ServiceProvider.CreateScope();
                var db = dbScope.ServiceProvider.GetRequiredService<DataContext>();
                await db.Database.MigrateAsync();
                logger.LogInformation("Db migration run successful");
                success = true;
                startupException = null;
            }
            catch (Exception ex)
            {
                startupException = ex;
                logger.LogError(ex, "Error accessing db for startup migration run on attempt [{StartupMigrationAttempt}]",
                    attempt);
                attempt++;
                await ExponentialBackoffAsync(attempt);
            }
        }

        if (attempt < maxAttempts || startupException == null)
        {
            return;
        }

        logger.LogCritical("Db migration run failed. Service run FAILED");
        throw new Exception("Critical stop", startupException);
    }

    private static Task ExponentialBackoffAsync(int attempt)
    {
        var waitInterval = TimeSpan.FromSeconds(Fibonacci(attempt));
        return Task.Delay(waitInterval);
    }

    private static int Fibonacci(int n)
    {
        if (n is 0 or 1)
        {
            return n;
        }
        return Fibonacci(n - 1) + Fibonacci(n - 2);
    }
}

Key Patterns:

  • Two constructors: Parameterized for runtime, parameterless for design-time
  • Conditional configuration: Fallback for migration tools
  • Assembly scanning: Auto-discovers entity configurations
  • Migration runner: Resilient startup with exponential backoff
  • Logging: Track migration attempts and failures

Entity Configurations

Entity configurations use the Fluent API to map domain objects to database tables. Each entity gets its own configuration class.

Configuration Helper

Reusable utilities for common configuration patterns:

using System.Linq.Expressions;
using Microsoft.EntityFrameworkCore.Metadata.Builders;

namespace Pikot.LMP.Adapters.Data.EntityConfigurations;

internal static class ConfigHelper
{
    // Bulk set multiple properties as required
    public static void SetRequired<TEntityType>(this EntityTypeBuilder<TEntityType> builder,
        params Expression<Func<TEntityType, object>>[] setters) where TEntityType : class
    {
        foreach (var setter in setters)
        {
            var member = setter.Body;
            if (member == null)
            {
                throw new ArgumentNullException(nameof(setter.Body));
            }

            var propInfo = member is UnaryExpression expression
                ? expression.Operand as MemberExpression
                : member as MemberExpression;

            builder.Property(propInfo.Member.Name).IsRequired();
        }
    }

    public static void SetMaxLength<TEntityType, TProperty>(this EntityTypeBuilder<TEntityType> builder,
        Expression<Func<TEntityType, TProperty>> property, int maxLength)
        where TEntityType : class
    {
        builder.Property(property).HasMaxLength(maxLength);
    }

    // Database schemas for bounded contexts
    public const string DispatcherScheme  = "Dispatcher";
}

Aggregate Configuration Example

Configuration for an aggregate root with child entities:

using Microsoft.EntityFrameworkCore;
using Microsoft.EntityFrameworkCore.Metadata.Builders;
using Pikot.LMP.Domain.Shipment;
using Pikot.LMP.Domain.Shipment.ShipperOrders;

namespace Pikot.LMP.Adapters.Data.EntityConfigurations;

internal class ShipperOrderConfiguration : IEntityTypeConfiguration<ShipperOrder>
{
    public void Configure(EntityTypeBuilder<ShipperOrder> builder)
    {
        // Table mapping with schema
        builder.ToTable("ShipperOrders", ConfigHelper.ShipmentScheme)
            .HasKey(p => p.Id);

        // Primary key with custom column name
        builder.Property(p => p.Id)
            .HasColumnName("ShipperOrderId");

        // Bulk required properties
        builder.SetRequired(
            d => d.ShipperId,
            d => d.ShipperCompanyName);

        // String length constraints
        builder.Property(d => d.ShipperCompanyName).HasMaxLength(128);
        builder.Property(d => d.OrderNumber).HasMaxLength(64);

        // Nullable properties
        builder.Property(d => d.DispatcherId);
        builder.Property(d => d.DispatcherCompanyName).HasMaxLength(128);

        // One-to-many relationships (owned collections)
        builder.HasMany(d => d.Changes)
            .WithOne()
            .HasForeignKey(c => c.ShipperOrderId);

        builder.HasMany(d => d.Versions)
            .WithOne()
            .HasForeignKey(c => c.ShipperOrderId);

        builder.HasMany(d => d.Packages)
            .WithOne()
            .HasForeignKey(p => p.ShipperOrderId);

        // Indexes
        builder.HasIndex(d => new { d.ShipperId, d.OrderNumber }).IsUnique();
        builder.HasIndex(d => d.Guid);
    }
}

Entity Configuration Example

Configuration for child entities:

using Microsoft.EntityFrameworkCore;
using Microsoft.EntityFrameworkCore.Metadata.Builders;
using Pikot.LMP.Domain.Shipment;
using Pikot.LMP.Domain.Shipment.ShipperOrders;

namespace Pikot.LMP.Adapters.Data.EntityConfigurations;

internal class PackageConfiguration : IEntityTypeConfiguration<Package>
{
    public void Configure(EntityTypeBuilder<Package> builder)
    {
        builder.ToTable("Packages", ConfigHelper.ShipmentScheme)
            .HasKey(p => p.Id);

        builder.Property(p => p.Id)
            .HasColumnName("PackageId");

        // Database-generated GUID
        builder.Property(p => p.Guid)
            .HasDefaultValueSql("gen_random_uuid()");

        // Custom column types
        builder.Property(p => p.Numeration)
            .HasColumnType("smallint");

        // Enum to byte conversion
        builder.Property(p => p.Status)
            .HasColumnType("smallint")
            .HasConversion<byte>();

        builder.SetRequired(
            d => d.Kg,
            d => d.ShipperOrderId,
            d => d.Guid,
            d => d.Numeration,
            d => d.Status);

        // Decimal precision
        builder.Property(d => d.Kg)
            .HasColumnType("decimal(10,2)");
    }
}

Simple Aggregate Configuration

Configuration for aggregates without child entities:

using Microsoft.EntityFrameworkCore;
using Microsoft.EntityFrameworkCore.Metadata.Builders;
using Pikot.LMP.Domain.Dispatcher;
using Pikot.LMP.Domain.Dispatcher.Dispatchers;

namespace Pikot.LMP.Adapters.Data.EntityConfigurations;

internal class DispatcherConfiguration : IEntityTypeConfiguration<Dispatcher>
{
    public void Configure(EntityTypeBuilder<Dispatcher> builder)
    {
        builder.ToTable("Dispatchers", ConfigHelper.DispatcherScheme)
            .HasKey(p => p.Id);

        builder.Property(p => p.Id)
            .HasColumnName("DispatcherId");

        builder.SetRequired(
            d => d.Name,
            d => d.Itin,
            d => d.CreatedAt,
            d => d.InsertedAt,
            d => d.IsActive);

        builder.Property(d => d.Name).HasMaxLength(128);
        builder.Property(d => d.Itin).HasMaxLength(15);
        builder.Property(d => d.ContactEmail).HasMaxLength(128);
        builder.Property(d => d.ContactName).HasMaxLength(64);
        builder.Property(d => d.ContactPhone).HasMaxLength(16);

        builder.HasIndex(d => d.Itin).IsUnique();
    }
}

One-to-One Relationship Configuration

using Microsoft.EntityFrameworkCore;
using Microsoft.EntityFrameworkCore.Metadata.Builders;
using Pikot.LMP.Domain.Shipper;
using Pikot.LMP.Domain.Shipper.Shippers;

namespace Pikot.LMP.Adapters.Data.EntityConfigurations;

internal class ShipperConfiguration : IEntityTypeConfiguration<Shipper>
{
    public void Configure(EntityTypeBuilder<Shipper> builder)
    {
        builder.ToTable("Shippers", ConfigHelper.ShipperScheme)
            .HasKey(p => p.Id);

        builder.Property(p => p.Id)
            .HasColumnName("ShipperId");

        builder.SetRequired(
            d => d.Name,
            d => d.Itin,
            d => d.CreatedAt,
            d => d.InsertedAt);

        builder.Property(d => d.Name).HasMaxLength(128);
        builder.Property(d => d.Itin).HasMaxLength(15);

        builder.HasIndex(d => d.Itin).IsUnique();

        // One-to-one relationship
        builder.HasOne(d => d.Settings)
            .WithOne()
            .HasForeignKey<ShipperSettings>(s => s.ShipperId);
    }
}

Configuration Best Practices:

  1. One class per entity - Separate configuration files
  2. Internal visibility - Configurations are implementation details
  3. Schema organization - Group by bounded context
  4. Custom column names for IDs - EntityNameId convention
  5. Bulk required properties - Use SetRequired helper
  6. Explicit string lengths - Prevent unlimited nvarchar(max)
  7. Type conversions - Enums, decimals, custom types
  8. Indexes - Unique constraints and performance indexes
  9. Database defaults - Use HasDefaultValueSql for GUIDs, timestamps

Repository Implementations

Write Repository Base Class

Base class for command-side (write) repositories:

using Microsoft.EntityFrameworkCore;
using Pikot.LMP.Domain;
using Pikot.LMP.Domain.Common;

namespace Pikot.LMP.Adapters.Data.Repositories;

internal abstract class RepositoryBase<TAggregate, TExistsKey>(DataContext context)
    : IRepository<TAggregate, TExistsKey>
    where TAggregate : class, IAggregate
{
    protected DataContext Context => context;

    protected virtual IQueryable<TAggregate> Query => Context.Set<TAggregate>();

    public abstract Task<bool> ExistsAsync(TExistsKey key, CancellationToken cancellationToken);

    public virtual Task SaveChangesAsync(CancellationToken cancellationToken)
    {
        return Context.SaveChangesAsync(cancellationToken);
    }

    public virtual Task<TAggregate?> GetAsync(int id, CancellationToken cancellationToken)
    {
        return Query.SingleOrDefaultAsync(x => x.Id == id, cancellationToken);
    }

    public virtual void Add(TAggregate entity)
    {
        Context.Add(entity);
    }
}

Key Features:

  • Internal visibility - Implementation detail
  • Primary constructor - Injects DataContext
  • Protected Query - Subclasses can customize
  • Virtual methods - Override for eager loading
  • Abstract ExistsAsync - Each repository defines its own unique key check

Query Repository Base Class

Base class for query-side (read) repositories:

using Microsoft.EntityFrameworkCore;
using Pikot.LMP.Domain;
using Pikot.LMP.Domain.Common;

namespace Pikot.LMP.Adapters.Data.Repositories;

internal abstract class QueryRepositoryBase<TAggregate>(DataContext dataContext) : IQueryRepository<TAggregate>
    where TAggregate : class, IAggregate
{
    protected DataContext DataContext => dataContext;

    protected IQueryable<TAggregate> Query => DataContext.Set<TAggregate>()
        .AsNoTracking();  // CRITICAL: No change tracking for reads

    public virtual Task<TAggregate?> GetAsync(int id, CancellationToken cancellationToken)
    {
        return Query.SingleOrDefaultAsync(x => x.Id == id, cancellationToken);
    }
}

Key Features:

  • AsNoTracking() - Critical for read performance
  • Simpler interface - No writes or existence checks
  • Virtual Get - Override for eager loading and includes

Write Repository Implementation

Example repository for write operations:

using Microsoft.EntityFrameworkCore;
using Pikot.LMP.Domain.Shipment;
using Pikot.LMP.Domain.Shipment.Repositories;
using Pikot.LMP.Domain.Shipment.ShipperOrders;

namespace Pikot.LMP.Adapters.Data.Repositories;

internal class ShipperOrderRepository(DataContext context)
    : RepositoryBase<ShipperOrder, int>(context), IShipperOrderRepository
{
    public override Task<bool> ExistsAsync(int key, CancellationToken cancellationToken)
    {
        return Query.AnyAsync(s => s.Id == key, cancellationToken);
    }

    public void Delete(ShipperOrder order)
    {
        Context.Set<ShipperOrder>().Remove(order);
    }

    // Override to include child collections
    public override Task<ShipperOrder?> GetAsync(int id, CancellationToken cancellationToken)
    {
        return Query
            .Include(s => s.Changes)
            .Include(s => s.Versions)
            .Include(s => s.Packages)
            .SingleOrDefaultAsync(s => s.Id == id, cancellationToken);
    }

    public Task<ShipperOrder[]> ListAsync(int[] ids, CancellationToken cancellationToken)
    {
        return Query
            .Where(s => ids.Contains(s.Id))
            .Include(s => s.Changes)
            .Include(s => s.Versions)
            .Include(s => s.Packages)
            .ToArrayAsync(cancellationToken);
    }
}

Query Repository Implementation

Example repository for read operations with complex queries:

using Microsoft.EntityFrameworkCore;
using Pikot.LMP.Domain.Common;
using Pikot.LMP.Domain.Shipment;
using Pikot.LMP.Domain.Shipment.Repositories;
using Pikot.LMP.Domain.Shipment.ShipperOrders;
using Pikot.LMP.Domain.Shipment.ShipperOrders.ListShipperOrders;

namespace Pikot.LMP.Adapters.Data.Repositories;

internal class ShipperOrderQueryRepository(DataContext dataContext)
    : QueryRepositoryBase<ShipperOrder>(dataContext), IShipperOrderQueryRepository
{
    public async Task<ShipperOrderPagedResult> ListByStatusAndShipperAsync(
        ShipperOrderStatusType status,
        ,int shipperId,
        QueryPaging paging,
        CancellationToken token)
    {
        var query = Query
            .Where(s => s.ShipperId == shipperId)
            .Where(s => s.Status == status);

        var total = await query.CountAsync(token);

        var result = await query
            .Include(c => c.Changes)
            .Include(c => c.Versions)
            .OrderBy(s => s.CreatedAt)
            .Skip(paging.PageSize * (paging.PageNumber - 1))
            .Take(paging.PageSize)
            .ToArrayAsync(token);

        return new ShipperOrderPagedResult(paging.PageNumber,
            paging.PageSize, total, result);
    }

    public override Task<ShipperOrder?> GetAsync(int id, CancellationToken cancellationToken)
    {
        return Query
            .Include(s => s.Versions)
            .Include(s => s.Changes)
            .Include(s => s.Packages)
            .SingleOrDefaultAsync(s => s.Id == id, cancellationToken);
    }

    public Task<ShipperOrder?> GetByGuidAsync(Guid guid, CancellationToken cancellationToken)
    {
        return Query
            .Include(s => s.Versions)
            .Include(s => s.Changes)
            .Include(s => s.Packages)
            .SingleOrDefaultAsync(s => s.Guid == guid, cancellationToken);
    }
}

Repository Best Practices:

  1. Override GetAsync - Always include necessary child collections
  2. Use Include() - Eager load related entities to avoid N+1 queries
  3. AsNoTracking - Query repositories should never track changes
  4. Specific methods - Add domain-specific query methods
  5. Pagination - Use Skip/Take for large result sets
  6. Filter early - Apply Where clauses before Include
  7. Count separately - Get total before pagination for page info

Unit of Work Implementation

Handles transactions across multiple repositories:

using Microsoft.EntityFrameworkCore.Storage;
using Pikot.LMP.Domain.Common;

namespace Pikot.LMP.Adapters.Data;

internal class UnitOfWork(DataContext context) : IUnitOfWork
{
    private IDbContextTransaction? _transaction;

    public async Task BeginTransactionAsync(CancellationToken cancellationToken)
    {
        _transaction = await context.Database
            .BeginTransactionAsync(cancellationToken);
    }

    public Task CommitTransactionAsync(CancellationToken cancellationToken)
    {
        if (_transaction == null)
        {
            throw new InvalidOperationException("Transaction has not been started");
        }

        return _transaction.CommitAsync(cancellationToken);
    }

    public Task RollbackTransactionAsync(CancellationToken cancellationToken)
    {
        if (_transaction == null)
        {
            throw new InvalidOperationException("Transaction has not been started");
        }

        return _transaction.RollbackAsync(cancellationToken);
    }

    public void Dispose()
    {
        _transaction?.Dispose();
        _transaction = null;
    }
}

Key Patterns:

  • Internal visibility - Implementation detail
  • Null checks - Validate transaction state
  • IDisposable - Proper transaction cleanup

Dependency Injection Registration

The public API of the data layer - the only public class:

using Microsoft.EntityFrameworkCore;
using Microsoft.Extensions.Configuration;
using Microsoft.Extensions.DependencyInjection;
using Pikot.LMP.Adapters.Data.Repositories;
using Pikot.LMP.Domain.Common;

namespace Pikot.LMP.Adapters.Data;

public static class ServiceRegistrations
{
    private static IServiceCollection RegisterRepositories(this IServiceCollection services)
    {
        var missingRegistrations = new List<string>();

        // Find all repository interfaces in Domain assembly
        var domainRepositoryInterfaces = typeof(IRepository<,>).Assembly.GetTypes()
            .Where(t => (typeof(IRepository<,>).IsAssignableFrom(t) || typeof(IQueryRepository<>).IsAssignableFrom(t))
                        && t is { IsInterface: true, IsGenericType: false })
            .ToArray();

        // Find all repository implementations in Data assembly
        var repositoryTypes = typeof(ServiceRegistrations).Assembly.GetTypes()
            .Where(t => t is { IsAbstract: false, IsInterface: false, BaseType.IsGenericType: true, IsNotPublic: true }
                        && (t.BaseType.GetGenericTypeDefinition() == typeof(RepositoryBase<,>)
                            || t.BaseType.GetGenericTypeDefinition() == typeof(QueryRepositoryBase<>)))
            .ToArray();

        // Verify all domain interfaces have implementations
        foreach (var repositoryInterface in domainRepositoryInterfaces)
        {
            if (!repositoryTypes.Any(r => r.IsAssignableFrom(repositoryInterface)))
            {
                missingRegistrations.Add($"{repositoryInterface.Name} has no registered implementations.");
            }
        }

        if (missingRegistrations.Count != 0)
        {
            throw new Exception($"Missing repository implementations: {string.Join(", ", missingRegistrations)}");
        }

        // Auto-register all implementations
        foreach (var repositoryType in repositoryTypes)
        {
            var domainType = repositoryType
                .GetInterfaces()
                .First(i => !i.IsGenericType);
            services.AddScoped(domainType, repositoryType);
        }

        return services;
    }

    public static IServiceCollection RegisterDataServices(
        this IServiceCollection services,
        IConfiguration configuration)
    {
        var connectionString = configuration.GetConnectionString("Postgres");

        return services
            .AddDbContext<DataContext>(options => options.UseNpgsql(connectionString))
            .RegisterRepositories()
            .AddScoped<IUnitOfWork, UnitOfWork>();
    }
}

DI Registration Features:

  1. Convention-based - Auto-discovers repositories by reflection
  2. Validation - Ensures all domain interfaces have implementations
  3. Fail-fast - Throws on startup if implementations are missing
  4. Public extension method - Only public API surface
  5. Configuration-driven - Connection string from appsettings
  6. Scoped lifetime - Proper DbContext scope management

Usage in Application Startup:

// In Program.cs or Startup.cs
builder.Services.RegisterDataServices(builder.Configuration);

Migration Workflow

Creating Migrations

# From the data project directory
dotnet ef migrations add MigrationName

# From solution root
dotnet ef migrations add MigrationName --project Pikot.LMP.Adapters.Data

Applying Migrations

Option 1: Automatic on Startup (Recommended for Development)

// In Program.cs
await DataContext.RunDbMigrationsAsync(app.Services);

Option 2: Manual via CLI

dotnet ef database update

Removing Last Migration

dotnet ef migrations remove

Database Schema Organization

Use schemas to organize tables by bounded context:

// In ConfigHelper.cs
public const string DispatcherScheme = "Dispatcher";

// In entity configuration
builder.ToTable("Orders", ConfigHelper.ShipmentScheme);
// Results in: Shipment.Orders table

Benefits:

  • Logical grouping - Related tables together
  • Clear boundaries - Visualizes bounded contexts
  • Permission management - Schema-level database permissions
  • Naming conflicts - Same table name in different schemas

Common Patterns & Best Practices

1. Eager Loading Child Collections

Always override GetAsync in repositories that need child collections:

public override Task<ShipperOrder?> GetAsync(int id, CancellationToken cancellationToken)
{
    return Query
        .Include(s => s.Changes)      // Load child collection
        .Include(s => s.Versions)     // Load child collection
        .Include(s => s.Packages)     // Load child collection
        .SingleOrDefaultAsync(s => s.Id == id, cancellationToken);
}

2. Handling Enums

// Store as byte/smallint
builder.Property(p => p.Status)
    .HasColumnType("smallint")
    .HasConversion<byte>();

3. Decimal Precision

builder.Property(d => d.Kg)
    .HasColumnType("decimal(10,2)");  // 10 digits, 2 decimal places

4. Database-Generated Values

// PostgreSQL UUID generation
builder.Property(p => p.Guid)
    .HasDefaultValueSql("gen_random_uuid()");

// SQL Server
builder.Property(p => p.Guid)
    .HasDefaultValueSql("NEWID()");

5. Unique Indexes

// Single column
builder.HasIndex(d => d.Itin).IsUnique();

// Composite
builder.HasIndex(d => new { d.ShipperId, d.OrderNumber }).IsUnique();

6. Query Performance Optimization

// BAD: N+1 queries
var orders = await context.ShipperOrders.ToListAsync();
foreach (var order in orders)
{
    var packages = order.Packages;  // Lazy load - separate query per order!
}

// GOOD: Single query with eager loading
var orders = await context.ShipperOrders
    .Include(o => o.Packages)
    .ToListAsync();

7. Filtering with Navigation Properties

// Access nested properties in queries
query = ordering.OrderBy switch
{
    OrderByType.PickupCity => query.OrderBy(
        s => s.Versions.OrderByDescending(v => v.Version).First().ShipperCity),
    _ => query.OrderBy(s => s.CreatedAt)
};

Quick Reference Templates

New Entity Configuration

using Microsoft.EntityFrameworkCore;
using Microsoft.EntityFrameworkCore.Metadata.Builders;
using Pikot.LMP.Domain.YourContext;

namespace Pikot.LMP.Adapters.Data.EntityConfigurations;

internal class YourEntityConfiguration : IEntityTypeConfiguration<YourEntity>
{
    public void Configure(EntityTypeBuilder<YourEntity> builder)
    {
        builder.ToTable("YourEntities", ConfigHelper.YourScheme)
            .HasKey(p => p.Id);

        builder.Property(p => p.Id)
            .HasColumnName("YourEntityId");

        builder.SetRequired(
            d => d.CreatedAt,
            d => d.InsertedAt);

        // Add specific configurations
    }
}

New Write Repository

using Microsoft.EntityFrameworkCore;
using Pikot.LMP.Domain.YourContext;
using Pikot.LMP.Domain.YourContext.Repositories;

namespace Pikot.LMP.Adapters.Data.Repositories;

internal class YourEntityRepository(DataContext context)
    : RepositoryBase<YourEntity, YourUniqueKey>(context), IYourEntityRepository
{
    public override Task<bool> ExistsAsync(YourUniqueKey key, CancellationToken cancellationToken)
    {
        return Query.AnyAsync(e => e.SomeProperty == key.Value, cancellationToken);
    }

    public override Task<YourEntity?> GetAsync(int id, CancellationToken cancellationToken)
    {
        return Query
            .Include(e => e.ChildCollection)
            .SingleOrDefaultAsync(e => e.Id == id, cancellationToken);
    }
}

New Query Repository

using Microsoft.EntityFrameworkCore;
using Pikot.LMP.Domain.YourContext;
using Pikot.LMP.Domain.YourContext.Repositories;

namespace Pikot.LMP.Adapters.Data.Repositories;

internal class YourEntityQueryRepository(DataContext dataContext)
    : QueryRepositoryBase<YourEntity>(dataContext), IYourEntityQueryRepository
{
    public override Task<YourEntity?> GetAsync(int id, CancellationToken cancellationToken)
    {
        return Query
            .Include(e => e.ChildCollection)
            .SingleOrDefaultAsync(e => e.Id == id, cancellationToken);
    }

    public async Task<YourEntityPagedResult> ListAsync(
        QueryPaging paging,
        CancellationToken cancellationToken)
    {
        var query = Query
            .Where(e => e.IsActive);

        var total = await query.CountAsync(cancellationToken);

        var result = await query
            .OrderBy(e => e.CreatedAt)
            .Skip(paging.PageSize * (paging.PageNumber - 1))
            .Take(paging.PageSize)
            .ToArrayAsync(cancellationToken);

        return new YourEntityPagedResult(paging.PageNumber, paging.PageSize, total, result);
    }
}

Checklist for New Aggregate

When adding a new aggregate to the data layer:

  • Create YourEntityConfiguration : IEntityTypeConfiguration<YourEntity>
  • Configure table name, schema, and primary key
  • Configure all properties (required, max length, types)
  • Configure relationships (HasMany, HasOne)
  • Configure indexes (unique constraints, performance)
  • Create YourEntityRepository : RepositoryBase<YourEntity, TKey>
  • Implement ExistsAsync with unique key logic
  • Override GetAsync with .Include() for child collections
  • Create YourEntityQueryRepository : QueryRepositoryBase<YourEntity>
  • Override GetAsync with .Include() for child collections
  • Add domain-specific query methods
  • Run dotnet ef migrations add YourEntityAdded
  • Review generated migration SQL
  • Test with in-memory database
  • Verify auto-registration in ServiceRegistrations

Common Pitfalls to Avoid

  1. Don't forget .Include() - Always eager load child collections
  2. Don't track queries - Use AsNoTracking() for query repositories
  3. Don't use lazy loading - Leads to N+1 queries
  4. Don't make repositories public - Keep them internal
  5. Don't skip unique indexes - Enforce business uniqueness at DB level
  6. Don't use unlimited strings - Always set HasMaxLength()
  7. Don't forget enum conversions - Specify column type explicitly
  8. Don't ignore decimal precision - Use HasColumnType("decimal(x,y)")
  9. Don't skip null checks in UnitOfWork - Validate transaction state
  10. Don't forget parameterless constructor - Needed for EF Core design-time tools

Integration with Domain Layer

The data layer implements ports defined in the domain:

Domain (Ports)                     Data (Adapters)
──────────────                     ───────────────
IRepository<T, TKey>      ────►    RepositoryBase<T, TKey>
IQueryRepository<T>       ────►    QueryRepositoryBase<T>
IUnitOfWork               ────►    UnitOfWork
IShipperOrderRepository   ────►    ShipperOrderRepository

Dependency flow:

  • Domain defines interfaces (ports)
  • Data implements interfaces (adapters)
  • Application/API depends on both, wires via DI

This skill document provides production-ready patterns for implementing EF Core data persistence in a hexagonal architecture with full CQRS support, testability, and clean separation of concerns.

Sponsors

Details

Marketplace
nesbo/dotnet-claude-code-skills
Source
data-dotnet
Category
Dotnet

More from nesbo/dotnet-claude-code-skills

All 4