恋爱虽易，相处不易：当EntityFramework爱上AutoMapper

剧情开始

• 为何相爱？
• 相处的问题？
• 女人的伟大？
• 剧情收尾？

　　有时候相识即是一种缘分，相爱也不需要太多的理由，一个眼神足矣，当EntityFramework遇上AutoMapper，就是如此，恋爱虽易，相处不易。

　　在DDD（领域驱动设计）中，使用AutoMapper一般场景是（Domain Layer）领域层与Presentation Layer（表现层）之间数据对象的转换，也就是DTO与Domin Model之间的相互转换，但是如果对AutoMapper有深入了解之后，就会发现她所涉及的领域不仅仅局限如此，应该包含所有对象之间的转换。另一边，当EntityFramework还在为单身苦恼时，不经意的一瞬间相识了AutoMapper，从此就深深的爱上了她。

　　AutoMapper是一个强大的Object-Object Mapping工具，关于AutoMapper请参照：

• 【AutoMapper官方文档】DTO与Domin Model相互转换（上）
• 【AutoMapper官方文档】DTO与Domin Model相互转换（中）
• 【AutoMapper官方文档】DTO与Domin Model相互转换（下）

为何相爱？

　　上面是AutoMapper对象转换示意图，可以看出AutoMapper的主要用途是用在对象映射转换上，她不管是什么对象，只是负责转换，就像一个女人在家只负责相夫教子一样。看下AutoMapper的基本用法：

       // 配置 AutoMapper
       Mapper.CreateMap<Order, OrderDto>();
       // 执行 mapping
       OrderDto dto = Mapper.Map<Order, OrderDto>(order);

　　EntityFramework是什么？他是微软开发的基于ADO.NET的ORM(Object/Relational Mapping)框架，是个大人物，是有身份和地位的人，就像一个“王子”一样，而AutoMapper准确的来说只是一个小角色，就像“灰姑娘”一样，况且他们也不是一个世界的人，那为什么EntityFramework会看上AutoMapper呢？这里面必定有内情，我们来探查一番。

　　假如存在这样一个业务场景，Order表中存在百万条订单数据，而且Order表有几百列，根据业务场景要求，我们要对订单进行分离，比如：客户信息订单、产品订单等等，可能只是用到订单表中的某些字段，如果我们去做这样的一个操作，可以想象这样查询出的数据是怎样的，某些我们并不需要的字段会查询出来，而且数据并没有得到过滤，所以我们要在数据访问层做下面这样一个操作：

         using (var context = new OrderContext())
         {
             var orderConsignee = from order in context.Orders
                                  select new OrderConsignee
                                  {
                                      OrderConsigneeId = order.OrderId,
                                      //OrderItems = order.OrderItems,
                                      OrderItemCount = order.OrderItemCount,
                                      ConsigneeName = order.ConsigneeName,
                                      ConsigneeRealName = order.ConsigneeRealName,
                                      ConsigneePhone = order.ConsigneePhone,
                                      ConsigneeProvince = order.ConsigneeProvince,
                                      ConsigneeAddress = order.ConsigneeAddress,
                                      ConsigneeZip = order.ConsigneeZip,
                                      ConsigneeTel = order.ConsigneeTel,
                                      ConsigneeFax = order.ConsigneeFax,
                                      ConsigneeEmail = order.ConsigneeEmail
                                  };
             Console.ReadKey();
         }

　　orderConsignee表示订单客户，这只是订单信息分离的一种子集，如果有多种分离的子集，并且子集中的字段并不比订单表少多少，你就会发现在数据访问层填充这些子集要做的工作量有多少了，虽然它是高效的，从生成的SQL代码中就可以看出：

 SELECT
     [Extent1].[OrderItemCount] AS [OrderItemCount],
     [Extent1].[OrderId] AS [OrderId],
     [Extent1].[ConsigneeName] AS [ConsigneeName],
     [Extent1].[ConsigneeRealName] AS [ConsigneeRealName],
     [Extent1].[ConsigneePhone] AS [ConsigneePhone],
     [Extent1].[ConsigneeProvince] AS [ConsigneeProvince],
     [Extent1].[ConsigneeAddress] AS [ConsigneeAddress],
     [Extent1].[ConsigneeZip] AS [ConsigneeZip],
     [Extent1].[ConsigneeTel] AS [ConsigneeTel],
     [Extent1].[ConsigneeFax] AS [ConsigneeFax],
     [Extent1].[ConsigneeEmail] AS [ConsigneeEmail]
     FROM [dbo].[Orders] AS [Extent1]

　　但是这种效果并不能让EntityFramework满意，于是他就盯上了人家AutoMapper，为什么？因为AutoMapper的一段代码就可以搞定上面的问题：

     OrderDto dto = Mapper.Map<Order, OrderDto>(order);

相处的问题？

　　因为EntityFramework的疯狂追求，再加上他显赫的地位，让AutoMapper不得不接受了他，于是他们就交往了，但好像就是后羿和嫦娥的故事一样，不是一个世界的人，相处起来总会出现一些问题。虽然AutoMapper在对象转换方面很强大，而且大部分应用场景是Domain与ViewModel之间的映射转换，当涉及到数据访问时，AutoMapper就不是那么有用了。换句话说，AutoMapper工作在内存中的对象转换，而不是应用在数据访问中IQueryable的接口，在数据访问层我们使用EntityFramework把要查询的对象转化为SQL命令，如果在数据访问层使用AutoMapper，那么查询数据一定会发生在映射转换之后，而且查询出的数据一定会比转换的数据多，从而产生性能问题。

　　上面的示例我们修改下：

     Mapper.CreateMap<Order, OrderConsignee>();
     var details = Mapper.Map<IEnumerable<Order>, IEnumerable<OrderConsignee>>(context.Orders).ToList();

　　其实这就是EntityFramework看上AutoMapper的原因，也是EntityFramework想要的效果，看下生成的SQL语句：

 SELECT
     [Extent1].[OrderId] AS [OrderId],
     [Extent1].[OrderItemCount] AS [OrderItemCount],
     [Extent1].[UserId] AS [UserId],
     [Extent1].[ReceiverId] AS [ReceiverId],
     [Extent1].[ShopDate] AS [ShopDate],
     [Extent1].[OrderDate] AS [OrderDate],
     [Extent1].[ConsigneeRealName] AS [ConsigneeRealName],
     [Extent1].[ConsigneeName] AS [ConsigneeName],
     [Extent1].[ConsigneePhone] AS [ConsigneePhone],
     [Extent1].[ConsigneeProvince] AS [ConsigneeProvince],
     [Extent1].[ConsigneeAddress] AS [ConsigneeAddress],
     [Extent1].[ConsigneeZip] AS [ConsigneeZip],
     [Extent1].[ConsigneeTel] AS [ConsigneeTel],
     [Extent1].[ConsigneeFax] AS [ConsigneeFax],
     [Extent1].[ConsigneeEmail] AS [ConsigneeEmail],
     [Extent1].[WhetherCouAndinte] AS [WhetherCouAndinte],
     [Extent1].[ParvalueAndInte] AS [ParvalueAndInte],
     [Extent1].[PaymentType] AS [PaymentType],
     [Extent1].[Payment] AS [Payment],
     [Extent1].[Courier] AS [Courier],
     [Extent1].[TotalPrice] AS [TotalPrice],
     [Extent1].[FactPrice] AS [FactPrice],
     [Extent1].[Invoice] AS [Invoice],
     [Extent1].[Remark] AS [Remark],
     [Extent1].[OrderStatus] AS [OrderStatus],
     [Extent1].[SaleUserID] AS [SaleUserID],
     [Extent1].[SaleUserType] AS [SaleUserType],
     [Extent1].[BusinessmanID] AS [BusinessmanID],
     [Extent1].[Carriage] AS [Carriage],
     [Extent1].[PaymentStatus] AS [PaymentStatus],
     [Extent1].[OgisticsStatus] AS [OgisticsStatus],
     [Extent1].[OrderType] AS [OrderType],
     [Extent1].[IsOrderNormal] AS [IsOrderNormal]
     FROM [dbo].[Orders] AS [Extent1]

　　通过上面的SQL语句，会发现，虽然数据访问层代码写的简单了，但是查询的字段并不是我们想要的，也就是说查询发生在映射之前，可以想象如果存在上百万的数据或是上百行，使用AutoMapper进行映射转换是多么的不靠谱，难道EntityFramework和AutoMapper就没有缘分？或者只是EntityFramework的一厢情愿？请看下面。

女人的伟大？

　　在EntityFramework和AutoMapper的相处过程中，虽然出现了某些问题，但其实也并不是EntityFramework的错，错就错在他们生不逢地，通过相处AutoMapper也发现EntityFramework是真心对她好，于是AutoMapper决定要做些改变，为了EntityFramework，也为了他们的将来。

　　EntityFramework和AutoMapper不在一个世界的原因，前面我们也分析过，一个存在于内存中，一个存在于数据访问中，AutoMapper要做的就是去扩展IQueryable表达式（有点嫦娥下凡的意思哈），从而使他们可以存在于一个世界，于是她为了EntityFramework就做了以下工作：

     public static class QueryableExtensions
     {
         public static ProjectionExpression<TSource> Project<TSource>(this IQueryable<TSource> source)
         {
             return new ProjectionExpression<TSource>(source);
         }
     }

     public class ProjectionExpression<TSource>
     {
         private static readonly Dictionary<string, Expression> ExpressionCache = new Dictionary<string, Expression>();

         private readonly IQueryable<TSource> _source;

         public ProjectionExpression(IQueryable<TSource> source)
         {
             _source = source;
         }

         public IQueryable<TDest> To<TDest>()
         {
              var queryExpression = GetCachedExpression<TDest>() ?? BuildExpression<TDest>();

             return _source.Select(queryExpression);
         }        

         private static Expression<Func<TSource, TDest>> GetCachedExpression<TDest>()
         {
             var key = GetCacheKey<TDest>();

             return ExpressionCache.ContainsKey(key) ? ExpressionCache[key] as Expression<Func<TSource, TDest>> : null;
         }

         private static Expression<Func<TSource, TDest>> BuildExpression<TDest>()
         {
             var sourceProperties = typeof(TSource).GetProperties();
             var destinationProperties = typeof(TDest).GetProperties().Where(dest => dest.CanWrite);
             var parameterExpression = Expression.Parameter(typeof(TSource), "src");

             var bindings = destinationProperties
                                 .Select(destinationProperty => BuildBinding(parameterExpression, destinationProperty, sourceProperties))
                                 .Where(binding => binding != null);

             var expression = Expression.Lambda<Func<TSource, TDest>>(Expression.MemberInit(Expression.New(typeof(TDest)), bindings), parameterExpression);

             var key = GetCacheKey<TDest>();

             ExpressionCache.Add(key, expression);

             return expression;
         }        

         private static MemberAssignment BuildBinding(Expression parameterExpression, MemberInfo destinationProperty, IEnumerable<PropertyInfo> sourceProperties)
         {
             var sourceProperty = sourceProperties.FirstOrDefault(src => src.Name == destinationProperty.Name);

             if (sourceProperty != null)
             {
                 return Expression.Bind(destinationProperty, Expression.Property(parameterExpression, sourceProperty));
             }

             var propertyNames = SplitCamelCase(destinationProperty.Name);

             if (propertyNames.Length == )
             {
                 sourceProperty = sourceProperties.FirstOrDefault(src => src.Name == propertyNames[]);

                 if (sourceProperty != null)
                 {
                     var sourceChildProperty = sourceProperty.PropertyType.GetProperties().FirstOrDefault(src => src.Name == propertyNames[]);

                     if (sourceChildProperty != null)
                     {
                         return Expression.Bind(destinationProperty, Expression.Property(Expression.Property(parameterExpression, sourceProperty), sourceChildProperty));
                     }
                 }
             }

             return null;
         }

         private static string GetCacheKey<TDest>()
         {
             return string.Concat(typeof(TSource).FullName, typeof(TDest).FullName);
         }

         private static string[] SplitCamelCase(string input)
         {
             return Regex.Replace(input, "([A-Z])", " $1", RegexOptions.Compiled).Trim().Split(' ');
         }
     }    

　　修改示例代码：

       Mapper.CreateMap<Order, OrderConsignee>();
       var details = context.Orders.Project().To<OrderConsignee>();

　　通过AutoMapper所做的努力，使得代码更加简化，只要配置一个类型映射，传递目标类型，就可以得到我们想要的转换对象，代码如此简洁，我们再来看下生成SQL代码：

 SELECT
     [Project1].[OrderId] AS [OrderId],
     [Project1].[OrderItemCount] AS [OrderItemCount],
     [Project1].[ConsigneeRealName] AS [ConsigneeRealName],
     [Project1].[ConsigneeName] AS [ConsigneeName],
     [Project1].[ConsigneePhone] AS [ConsigneePhone],
     [Project1].[ConsigneeProvince] AS [ConsigneeProvince],
     [Project1].[ConsigneeAddress] AS [ConsigneeAddress],
     [Project1].[ConsigneeZip] AS [ConsigneeZip],
     [Project1].[ConsigneeTel] AS [ConsigneeTel],
     [Project1].[ConsigneeFax] AS [ConsigneeFax],
     [Project1].[ConsigneeEmail] AS [ConsigneeEmail],
     [Project1].[C1] AS [C1],
     [Project1].[OrderItemId] AS [OrderItemId],
     [Project1].[ProName] AS [ProName],
     [Project1].[ProImg] AS [ProImg],
     [Project1].[ProPrice] AS [ProPrice],
     [Project1].[ProNum] AS [ProNum],
     [Project1].[AddTime] AS [AddTime],
     [Project1].[ProOtherPara] AS [ProOtherPara],
     [Project1].[Order_OrderId] AS [Order_OrderId]
     FROM ( SELECT
         [Extent1].[OrderId] AS [OrderId],
         [Extent1].[OrderItemCount] AS [OrderItemCount],
         [Extent1].[ConsigneeRealName] AS [ConsigneeRealName],
         [Extent1].[ConsigneeName] AS [ConsigneeName],
         [Extent1].[ConsigneePhone] AS [ConsigneePhone],
         [Extent1].[ConsigneeProvince] AS [ConsigneeProvince],
         [Extent1].[ConsigneeAddress] AS [ConsigneeAddress],
         [Extent1].[ConsigneeZip] AS [ConsigneeZip],
         [Extent1].[ConsigneeTel] AS [ConsigneeTel],
         [Extent1].[ConsigneeFax] AS [ConsigneeFax],
         [Extent1].[ConsigneeEmail] AS [ConsigneeEmail],
         [Extent2].[OrderItemId] AS [OrderItemId],
         [Extent2].[ProName] AS [ProName],
         [Extent2].[ProImg] AS [ProImg],
         [Extent2].[ProPrice] AS [ProPrice],
         [Extent2].[ProNum] AS [ProNum],
         [Extent2].[AddTime] AS [AddTime],
         [Extent2].[ProOtherPara] AS [ProOtherPara],
         [Extent2].[Order_OrderId] AS [Order_OrderId],
         CASE WHEN ([Extent2].[OrderItemId] IS NULL) THEN CAST(NULL AS int) ELSE 1 END AS [C1]
         FROM  [dbo].[Orders] AS [Extent1]
         LEFT OUTER JOIN [dbo].[OrderItems] AS [Extent2] ON [Extent1].[OrderId] = [Extent2].[Order_OrderId]
     )  AS [Project1]
     ORDER BY [Project1].[OrderId] ASC, [Project1].[C1] ASC

　　可以看出因为Order和OrderConsignee包含对OrderItems子集的映射关系：

         /// <summary>
         /// 订单项
         /// </summary>
         public virtual ICollection<OrderItem> OrderItems { get; set; }

　　所以AutoMapper会自动匹配关联子集进行查询，当然也可以在创建映射关系的时候对OrderItems进行忽略：Mapper.CreateMap<Order, OrderConsignee>().ForMember(dest => dest.OrderItems, opt => opt.Ignore()); 排除OrderItems关联因素，从SQL代码可以看出并没有查询多余的字段，也就是我们想要的效果，这所以的一切都归功于AutoMapper，也许如果没有AutoMapper的努力，她和EntityFramework说不准还真不能在一起，女人真是伟大啊。

剧情收尾？

　　示例代码下载：http://pan.baidu.com/s/1c0h9TNM

　　经过一切风风雨雨，EntityFramework终于和AutoMapper过上了幸福美满的日子，但是看似幸福，但是问题还是不断，有人又提出疑问：

• http://rogeralsing.com/2013/12/01/why-mapping-dtos-to-entities-using-automapper-and-entityframework-is-horrible/

　　文章的标题用了“horrible”这个单词，翻译为可怕的，难道说EntityFramework和AutoMapper在一起有那么可怕吗？当然这只是针对EntityFramework使用AutoMapper进行CURD操作，但是我相信EntityFramework和AutoMapper会克服重重困难，生死不渝的。我们也会一直关注他们的婚后生活，未完待续。。。

　　如果你也祝福EntityFramework和AutoMapper会永远在一起，那就疯狂的“戳”右下角的“推荐”吧。^_^