Automapper performance

Question

I'm using Automapper to map my business model to a ViewModel.

It works, but it's very slow.

I have a collection with 6893 objects with 23 properties (test environment, production should have much more).

With a loop it takes 00:02:32.8118534 to map everything.

var objects = // get all items (returns a collection of MyObj)
List<MyViewModel> collection = new List<MyViewModel>();
foreach (MyObj obj in objects)
{
     MyViewModel vm = Mapper.Map<MyObj, MyViewModel>(obj);
     collection.Add(vm);
}

I tried to improve it like this:

var objects = // get all items (returns a collection of MyObj)
IEnumerable<MyViewModel> collection = mapper.Map<IEnumerable<MyObj>, IEnumerable<MyViewModel>>(objects);

And it took 00:02:25.4527961 to map everything.

So it didn't help that much.

None of my object's properties can be null.

This is how I configured the mapper:

var config = new MapperConfiguration(cfg =>
        {
            cfg.CreateMap<MyObj, MyViewModel>();
            cfg.CreateMap<MyObjOtherObj, MyViewModelOtherObj>();
        });
mapper = config.CreateMapper();

MyObj:

public partial class MyObj
{
    public MyObj()
    {
        this.MyObjOtherObj= new HashSet<MyObjOtherObj>();
    }

    public long a{ get; set; }
    public short b{ get; set; }
    public string c{ get; set; }
    public string d{ get; set; }
    public string e{ get; set; }
    public string f{ get; set; }
    public string g{ get; set; }
    public string h{ get; set; }
    public string i{ get; set; }
    public string j{ get; set; }
    public string k{ get; set; }
    public string l{ get; set; }
    public string m{ get; set; }
    public bool n{ get; set; }
    public bool o{ get; set; }
    public bool p{ get; set; }
    public bool q{ get; set; }

    public virtual ICollection<MyObjOtherObj> MyObjOtherObj{ get; set; }
    public virtual Types Types { get; set; }
}

MyViewModel:

public class MyViewModel
{
    public long a{ get; set; }
    public short b{ get; set; }
    public string c{ get; set; }
    public string d{ get; set; }
    public string e{ get; set; }
    public string f{ get; set; }
    public string g{ get; set; }
    public string h{ get; set; }
    public string i{ get; set; }
    public string j{ get; set; }
    public string k{ get; set; }
    public string l{ get; set; }
    public string m{ get; set; }
    public bool n{ get; set; }
    public bool o{ get; set; }
    public bool p{ get; set; }
    public bool q{ get; set; }
    public string TypesDescription { get; set; }

    public List<MyViewModelOtherObj> MyObjOtherObj { get; set; }
}

MyObjOtherObj:

public partial class MyObjOtherObj
{
    public long id{ get; set; }
    public long MyObjId { get; set; }
    public short x{ get; set; }
    public string z{ get; set; }

    public virtual MyObj MyObj{ get; set; }
    public virtual SourceTypes SourceTypes { get; set; }
}

MyViewModelOtherObj:

public class MyViewModelOtherObj
{
    public long Id { get; set; }
    public long MyObjId { get; set; }
    public short x{ get; set; }
    public string z{ get; set; }
    public string SourceTypesDescription { get; set; }
}

EDIT:

SourceTypes:

public partial class SourceTypes
{
    public SourceTypes()
    {
        this.MyObjOtherObj = new HashSet<MyObjOtherObj>();
    }

    public short SourceTypeId { get; set; }
    public string Description { get; set; }

    public virtual ICollection<MyObjOtherObj> MyObjOtherObj { get; set; }
}

Types:

public partial class Types
{
    public Types()
    {
        this.MyObj = new HashSet<MyObj>();
    }

    public short TypeId { get; set; }
    public string Description { get; set; }

    public virtual ICollection<MyObj> MyObj{ get; set; }
}

Answer 1

The 5.0 version of AutoMapper has significant performance increases. In our benchmarks, using a very similar type that you've shown here, we can map one million items in a little over a second. In the upcoming 5.1 version, that shrinks even more that we're only about 3x slower than hand-mapping, mainly due to null checking that hand mapping won't do.

I'd upgrade.