Flowtex.DataAccess 1.0.4

Flowtex.DataAccess

A modern, streamlined data access library for Entity Framework Core that provides a clean alternative to the traditional Repository and Unit of Work patterns. This library simplifies data operations while maintaining performance and testability.

What This Library Does

Flowtex.DataAccess provides a unified interface for data access operations through a set of core abstractions:

• IReadStore - Read-only operations with optimized no-tracking queries, including FindAsync<T> for identity-map-cached key lookups
• IDataStore - Full CRUD operations including reads, writes, and transactions
• ISaveHandle - Deferred save handle returned by single-mutation methods; calling SaveAsync() flushes all pending context changes, not only the staged operation — see ISaveHandle notes below
• IUpdateBuilder<T> - Fluent API for efficient bulk update operations

The library is built around Entity Framework Core but abstracts away the complexity while preserving all the power and performance benefits.

Why Use This Instead of Repository and Unit of Work Patterns?

Problems with Traditional Patterns

Repository Pattern Issues:

• Creates unnecessary abstraction layers over already-abstracted EF Core
• Often leads to "leaky abstractions" where EF-specific code bleeds through
• Requires extensive mocking for unit testing
• Can hide EF Core's powerful query capabilities behind generic interfaces
• Often results in N+1 query problems due to oversimplified interfaces

Unit of Work Pattern Issues:

• EF Core's DbContext already implements Unit of Work
• Creates redundant transaction management
• Can lead to confusing nested transaction scenarios
• Adds complexity without meaningful benefits

Flowtex.DataAccess Advantages

1. Direct EF Core Integration (no abstraction overhead)

• Leverages EF Core's native capabilities instead of hiding them
• No performance penalties from unnecessary abstractions
• Access to full LINQ query power through IQueryable<T>

2. Explicit Save Control (Deferred Execution pattern)

• ISaveHandle provides clear control over when data is persisted
• Enables batching multiple mutations into a single database round-trip
• Eliminates accidental implicit saves

3. Performance Optimized (Fluent Builder for bulk ops)

• No-tracking queries by default for reads (IReadStore)
• Tracked queries available when needed (QueryTracked<T>)
• Bulk operations via ExecuteUpdateAsync/ExecuteDeleteAsync push a single UPDATE/DELETE statement to the database — no entities loaded into memory

4. Clean Testing (Command-Query Separation)

• IReadStore and IDataStore are easily mockable
• Services declare only the interface they need — a read-only service takes IReadStore; a write service takes IDataStore
• No complex repository hierarchies to maintain

5. Minimal Boilerplate (Template Method pattern)

• DataStoreBase implements all logic; concrete stores override a single protected abstract DbContext Context { get; } property
• Built-in transaction support via InTransactionAsync uses EF Core's execution strategy for automatic retry and rollback
• Consistent API across all entity types

Installation

Two packages are published to match clean architecture layering:

# Application/domain project (IDataStore, IReadStore, ISaveHandle, IUpdateBuilder<T>)
dotnet add package Flowtex.DataAccess.Abstractions

# Infrastructure project (DataStoreBase + EF Core)
dotnet add package Flowtex.DataAccess

Examples

Basic Setup

// Implementation example (in your Infrastructure layer)
// Only Context needs to be overridden — DataStoreBase calls Context.Set<T>() internally.
public class AppDataStore : DataStoreBase
{
    private readonly AppDbContext _context;
    
    public AppDataStore(AppDbContext context)
    {
        _context = context;
    }
    
    protected override DbContext Context => _context;
}

// Dependency injection setup
services.AddScoped<IDataStore, AppDataStore>();
services.AddScoped<IReadStore>(provider => provider.GetRequiredService<IDataStore>());

Reading Data

public class ProductQueryService
{
    private readonly IReadStore _readStore;
    
    public ProductQueryService(IReadStore readStore)
    {
        _readStore = readStore;
    }
    
    // Simple query
    public async Task<List<Product>> GetActiveProductsAsync()
    {
        return await _readStore.ListAsync<Product>(q => 
            q.Where(p => p.IsActive)
             .OrderBy(p => p.Name));
    }
    
    // Complex query with projection
    public async Task<List<ProductSummary>> GetProductSummariesAsync()
    {
        return await _readStore.ListAsync<Product, ProductSummary>(q =>
            q.Where(p => p.IsActive)
             .Select(p => new ProductSummary 
             { 
                 Id = p.Id, 
                 Name = p.Name, 
                 Price = p.Price 
             }));
    }
    
    // Direct IQueryable access for complex scenarios
    public IQueryable<Product> GetProductsQuery()
    {
        return _readStore.Query<Product>();
    }
}

Writing Data

public class ProductCommandService
{
    private readonly IDataStore _dataStore;
    
    public ProductCommandService(IDataStore dataStore)
    {
        _dataStore = dataStore;
    }
    
    // Adding single entity
    public async Task<Product> CreateProductAsync(Product product)
    {
        var saveHandle = await _dataStore.AddAsync(product);
        await saveHandle.SaveAsync();
        return product;
    }
    
    // Adding multiple entities
    public async Task AddProductsAsync(IEnumerable<Product> products)
    {
        var saveHandle = await _dataStore.AddRangeAsync(products);
        await saveHandle.SaveAsync();
    }
    
    // Updating entity
    public async Task UpdateProductAsync(Product product)
    {
        var saveHandle = _dataStore.Update(product);
        await saveHandle.SaveAsync();
    }
    
    // Bulk update
    public async Task DiscontinueProductsByCategoryAsync(int categoryId)
    {
        await _dataStore.ExecuteUpdateAsync<Product>(
            p => p.CategoryId == categoryId,
            builder => builder.SetConst(p => p.IsDiscontinued, true));
    }
    
    // Bulk delete
    public async Task DeleteExpiredProductsAsync()
    {
        await _dataStore.ExecuteDeleteAsync<Product>(
            p => p.ExpirationDate < DateTime.UtcNow);
    }
}

Transaction Management

InTransactionAsync wraps work in a database transaction and commits on success or rolls back on exception. It does not call SaveChangesAsync implicitly — callers are responsible for flushing pending changes inside the delegate.

public async Task<OrderResult> ProcessOrderAsync(Order order, List<OrderItem> items)
{
    return await _dataStore.InTransactionAsync(async dataStore =>
    {
        // Save the order first to obtain its auto-generated Id.
        var orderHandle = await dataStore.AddAsync(order);
        await orderHandle.SaveAsync();
        
        // Set the FK, then stage items + any other changes and flush once.
        foreach (var item in items)
            item.OrderId = order.Id;

        await dataStore.AddRangeAsync(items); // staged, not yet saved
        await dataStore.SaveChangesAsync();   // single round-trip for all staged changes
        
        // ExecuteUpdateAsync runs immediately as its own statement inside the transaction.
        await dataStore.ExecuteUpdateAsync<Product>(
            p => items.Select(i => i.ProductId).Contains(p.Id),
            builder => builder.Set(p => p.Stock, 
                p => p.Stock - items.First(i => i.ProductId == p.Id).Quantity));
        
        return new OrderResult { OrderId = order.Id, Success = true };
    });
}

ISaveHandle Notes

Because EF Core uses a single shared change tracker per DbContext instance, calling SaveAsync() on any ISaveHandle saves all pending changes in the context — not only the operation that returned the handle. This is intentional: it preserves EF's unit-of-work semantics.

When you need to stage multiple mutations and persist them in one round-trip, call the mutation methods without awaiting their handles, then call SaveChangesAsync() directly:

// Stage both mutations
await dataStore.AddAsync(entity1);  // handle ignored
dataStore.Update(entity2);           // handle ignored

// Persist both in a single database round-trip
await dataStore.SaveChangesAsync();

Testing Examples

Separate read and write concerns into distinct service classes so each depends only on the interface it needs — this keeps mocking simple and makes the dependency contract explicit.

// Read-only service — depends only on IReadStore
public class ProductQueryService
{
    private readonly IReadStore _readStore;
    public ProductQueryService(IReadStore readStore) => _readStore = readStore;

    public Task<List<Product>> GetActiveProductsAsync() =>
        _readStore.ListAsync<Product>(q => q.Where(p => p.IsActive).OrderBy(p => p.Name));
}

// Write-side service — depends only on IDataStore
public class ProductCommandService
{
    private readonly IDataStore _dataStore;
    public ProductCommandService(IDataStore dataStore) => _dataStore = dataStore;

    public async Task CreateProductAsync(Product product)
    {
        var handle = await _dataStore.AddAsync(product);
        await handle.SaveAsync();
    }
}

// Tests
public class ProductQueryServiceTests
{
    [Test]
    public async Task GetActiveProducts_ReturnsOnlyActiveProducts()
    {
        // Arrange
        var mockReadStore = new Mock<IReadStore>();
        var products = new List<Product>
        {
            new() { Id = 1, Name = "Active Product", IsActive = true },
            new() { Id = 2, Name = "Inactive Product", IsActive = false }
        }.AsQueryable();
        
        mockReadStore.Setup(x => x.ListAsync<Product>(
                It.IsAny<Func<IQueryable<Product>, IQueryable<Product>>>(),
                It.IsAny<CancellationToken>()))
            .ReturnsAsync((Func<IQueryable<Product>, IQueryable<Product>> shape, CancellationToken _) =>
                shape(products).ToList());
        
        var service = new ProductQueryService(mockReadStore.Object);
        
        // Act
        var result = await service.GetActiveProductsAsync();
        
        // Assert
        Assert.That(result.Count, Is.EqualTo(1));
        Assert.That(result[0].Name, Is.EqualTo("Active Product"));
    }
}

Future Samples

This section will be expanded with additional examples as the library evolves:

Planned Examples

• Advanced Query Patterns

  • Complex joins and includes
  • Dynamic query building
  • Specification pattern integration

• Performance Optimization

  • Query performance analysis
  • Batch operation strategies
  • Memory-efficient data processing

• Integration Patterns

  • CQRS implementation
  • Event sourcing scenarios
  • Microservices data patterns

• Advanced Transaction Scenarios

  • Distributed transactions
  • Saga pattern implementation
  • Compensation-based transactions

• Testing Strategies

  • Integration testing with TestContainers
  • Performance testing approaches
  • Mocking complex scenarios

Summary

Flowtex.DataAccess is a pragmatic, modern alternative to the Repository + Unit of Work pattern stack. It solves the real pain points — testability, batching, bulk operations, transactions — without adding an abstraction layer that fights EF Core. Because IQueryable<T> is a first-class citizen, nothing is hidden from you: complex joins, projections, raw queries, and EF-specific features all work exactly as they do natively.

The split into Abstractions vs. Infrastructure packages makes it a natural fit for clean/onion architecture. Domain and application code reference only interfaces with no EF Core coupling; infrastructure wires it all up. The example suite — covering Web API usage, transaction scenarios, and unit testing strategies — demonstrates each pattern clearly and is worth reading alongside this documentation.

The one behaviour to internalise: ISaveHandle.SaveAsync() flushes all pending changes on the context, not just the one it was created from. This is a natural consequence of EF Core's Unit of Work and is by design — but understanding it upfront prevents surprises.

Contributing

We welcome contributions! Please see our contributing guidelines for details on how to submit pull requests, report issues, and suggest improvements.

License

This project is licensed under the MIT License - see the LICENSE file for details.