48
votes

My repositories deal with and provide persistence for a rich domain model. I do not want to expose the anemic, Entity Framework data entity to my business layers, so I need some way of mapping between them.

In most cases, constructing a domain model instance from a data entity requires the use of parameterised constructors and methods (since it is rich). It is not as simple as a property/field match. AutoMapper could be used for the opposite situation (mapping to data entities) but not when creating domain models.

Below is the core of my repository pattern.

The EntityFrameworkRepository class works with two generic types:

  • TDomainModel: The rich domain model
  • TEntityModel: The Entity Framework data entity

Two abstract methods are defined:

  • ToDataEntity(TDomainModel): To convert to data entities (for Add() and Update() methods)
  • ToDomainModel(TEntityModel): To construct domain models (for the Find() method).

Concrete implementations of these methods would define the mapping required for the repository in question.

public interface IRepository<T> where T : DomainModel
{
    T Find(int id);
    void Add(T item);
    void Update(T item);
}

public abstract class EntityFrameworkRepository<TDomainModel, TEntityModel> : IRepository<TDomainModel>
    where TDomainModel : DomainModel
    where TEntityModel : EntityModel
{
    public EntityFrameworkRepository(IUnitOfWork unitOfWork)
    {
        // ...
    }

    public virtual TDomainModel Find(int id)
    {
        var entity = context.Set<TEntityModel>().Find(id);

        return ToDomainModel(entity);
    }

    public virtual void Add(TDomainModel item)
    {
        context.Set<TEntityModel>().Add(ToDataEntity(item));
    }

    public virtual void Update(TDomainModel item)
    {
        var entity = ToDataEntity(item);

        DbEntityEntry dbEntityEntry = context.Entry<TEntityModel>(entity);

        if (dbEntityEntry.State == EntityState.Detached)
        {
            context.Set<TEntityModel>().Attach(entity);

            dbEntityEntry.State = EntityState.Modified;
        }
    }

    protected abstract TEntityModel ToDataEntity(TDomainModel domainModel);
    protected abstract TDomainModel ToDomainModel(TEntityModel dataEntity);
}

Here is a basic example of a repository implementation:

public interface ICompanyRepository : IRepository<Company>
{
    // Any specific methods could be included here
}

public class CompanyRepository : EntityFrameworkRepository<Company, CompanyTableEntity>, ICompanyRepository
{
    protected CompanyTableEntity ToDataEntity(Company domainModel)
    {
        return new CompanyTable()
        {
            Name = domainModel.Name,
            City = domainModel.City
            IsActive = domainModel.IsActive
        };
    }

    protected Company ToDomainModel(CompanyTableEntity dataEntity) 
    {
        return new Company(dataEntity.Name, dataEntity.IsActive)
        {
            City = dataEntity.City
        }
    }
}

Problem:

A Company might be composed of many Departments. If I want to eagerly load these from the CompanyRepository when fetching a Company then where would I define the mapping between a Department and a DepartmentDataEntity?

I could provide more mapping methods in the CompanyRepository, but this will soon get messy. There would soon be duplicated mapping methods across the system.

What is a better approach to the above problem?

6
1. define the Company stricly according to the business concept. 2. Model the entity use cases 3. Note that if done properly you won't be having the problem you have now. And ignore anything related to persistenceMikeSW
AutoMapper could be used for the opposite situation [...] but not when creating domain models. Why is it not possible both ways?Thomas Weller
@ThomasWeller, there is an answer by creator of AutoMapper, Jimmmy Bogard - The case for two-way mapping in AutoMapper.Ilya Palkin
@davenewza, did you decide to use the approach described in your question or did you map directly to the domain model?w0051977

6 Answers

34
votes

My repositories deal with and provide persistence for a rich domain model. I do not want to expose the anemic, Entity Framework data entity to my business layers, so I need some way of mapping between them.

If you you use Entity Framework, it can map Rich Domain Model itself.

I've answered the similar question "Advice on mapping of entities to domain objects" recently.

I've been using NHibernate and know that in Entity Framework you can also specify mapping rules from DB tables to your POCO objects. It is an extra effort to develop another abstraction layer over Entity Framework entities. Let the ORM be responsible for all of the mappings, state tracking, unit of work and identity map implementation, etc. Modern ORMs know how to handle all these issues.

AutoMapper could be used for the opposite situation (mapping to data entities) but not when creating domain models.

You are completely right.

Automapper is useful when one entity can be mapped into another without additional dependencies (e.g. Repositories, Services, ...).

... where would I define the mapping between a Department and a DepartmentDataEntity?

I would put it into DepartmentRepository and add method IList<Department> FindByCompany(int companyId) in order to retreive company's departments.

I could provide more mapping methods in the CompanyRepository, but this will soon get messy. There would soon be duplicated mapping methods across the system.

What is a better approach to the above problem?

If it is needed to get list of Departments for another entity, a new method should be added to DepartmentRepository and simply used where it is needed.

5
votes

Let's say you have the following data access object...

public class AssetDA
{        
    public HistoryLogEntry GetHistoryRecord(int id)
    {
        HistoryLogEntry record = new HistoryLogEntry();

        using (IUnitOfWork uow = new NHUnitOfWork())
        {
            IReadOnlyRepository<HistoryLogEntry> repository = new NHRepository<HistoryLogEntry>(uow);
            record = repository.Get(id);
        }

        return record;
    }
}

which returns a history log entry data entity. This data entity is defined as follows...

public class HistoryLogEntry : IEntity
{
    public virtual int Id
    { get; set; }

    public virtual int AssetID 
    { get; set; }

    public virtual DateTime Date
    { get; set; }

    public virtual string Text
    { get; set; }

    public virtual Guid UserID
    { get; set; }

    public virtual IList<AssetHistoryDetail> Details { get; set; }
}

You can see that the property Details references another data entity AssetHistoryDetail. Now, in my project I need to map these data entities to Domain model objects which are used in my business logic. To do the mapping I have defined extension methods...I know it's an anti-pattern since it is language specific, but the good thing is that it isolate and breaks dependencies between each other...yeah, that's the beauty of it. So, the mapper is defined as follows...

internal static class AssetPOMapper
{
    internal static HistoryEntryPO FromDataObject(this HistoryLogEntry t)
    {
        return t == null ? null :
            new HistoryEntryPO()
            {
                Id = t.Id,
                AssetID = t.AssetID,
                Date = t.Date,
                Text = t.Text,
                UserID = t.UserID,
                Details = t.Details.Select(x=>x.FromDataObject()).ToList()
            };
    }

    internal static AssetHistoryDetailPO FromDataObject(this AssetHistoryDetail t)
    {
        return t == null ? null :
            new AssetHistoryDetailPO()
            {
                Id = t.Id,
                ChangedDetail = t.ChangedDetail,
                OldValue = t.OldValue,
                NewValue = t.NewValue
            };
    }
}

and that's pretty much it. All dependencies are in one place. Then, when calling a data object from the business logic layer I'd let LINQ do the rest...

var da = new AssetDA();
var entry =  da.GetHistoryRecord(1234);
var domainModelEntry = entry.FromDataObject();

Note that you can define the same to map Domain Model objects to Data Entities.

4
votes

With entity framework, across all layers IMHO, it is generally a bad idea to convert from entity models to other form of models (domain models, value objects, view models, etc.) and vice versa except only in the application layer because you will lose a lot of EF capabilities that you can only achieve through the entity objects, like loss of change tracking and loss of LINQ queryable.

It's better to do the mapping between your repository layer and the application layer. Keep the entity models in the repository layer.

3
votes

I like using custom-built Extension methods to do the mapping between Entity and Domain objects.

  • You can easily call extension methods for other entities when they are within a containing entity.
  • You can easily deal with collections by creating IEnumerable<> extensions.

A simple example:

public static class LevelTypeItemMapping
{
    public static LevelTypeModel ToModel(this LevelTypeItem entity)
    {
        if (entity == null) return new LevelTypeModel();

        return new LevelTypeModel
        { 
            Id = entity.Id;
            IsDataCaptureRequired = entity.IsDataCaptureRequired;
            IsSetupRequired = entity.IsSetupRequired;
            IsApprover = entity.IsApprover;
            Name = entity.Name;
        }
    }
    public static IEnumerable<LevelTypeModel> ToModel(this IEnumerable<LevelTypeItem> entities)
    {
        if (entities== null) return new List<LevelTypeModel>();

        return (from e in entities
                select e.ToModel());
    }

}

...and you use them like this.....

 using (IUnitOfWork uow = new NHUnitOfWork())
        {
        IReadOnlyRepository<LevelTypeItem> repository = new NHRepository<LevelTypeItem>(uow);
        record = repository.Get(id);

        return record.ToModel();

        records = repository.GetAll(); // Return a collection from the DB
        return records.ToModel(); // Convert a collection of entities to a collection of models
        }

Not perfect, but very easy to follow and very easy to reuse.

2
votes

Like previous posts have said. It's probably best to wait untill after the repositories to do actual mapping.. BUT I like working with auto-mapper. It provides a very easy way to map objects to other objects. For some separation of concerns, you can also define the mappings in a separate project. These mappings are also generic / type-based:

  1. You specify the base type in the mapping and define how it fills the destination type
  2. Where you need the mapping, you just call Mapper.Map(baseTypeObject, DestinationTypeObject)
  3. Automapper should handle the rest

This could do the trick, if I understood your question correctly.

0
votes

I dont like to map manually so for mapping i am using http://valueinjecter.codeplex.com/