NHibernate 多对多映射的数据更新

最近在用 NHibernate 作多对多更新时忽然发现 NHibernate 更新的策略不好， 对多对多关系的更新竟然是先所有删除再插入所有数据， 感受很是奇怪， 如今还原以下：

原来的实体类关系以下：

public class User { public virtual int Id { get; set; } public virtual string Name { get; set; } public virtual ICollection<Role> Roles { get; set; } public User() {
        Roles = new HashSet<Role>();
    }
} public class Role { public virtual int Id { get; set; } public virtual string Name { get; set; } public virtual ICollection<User> Users { get; set; } public Role() {
        Users = new HashSet<User>();
    }

}

即一个用户能够有多个角色， 一个角色也能够有多我的， 典型的多对多关系， 对应的映射代码以下：

public class UserMapping : ClassMapping<User> { public UserMapping() {
        Table("[User]");

        Id(m => m.Id, map => { map.Column("[Id]"); map.Type(NHibernateUtil.Int32); map.Generator(Generators.Identity);
        });

        Property(m => m.Name, map => { map.Column("[Name]"); map.Type(NHibernateUtil.String);
        });

        Bag(
            m => m.Roles, map => { map.Table("[User_Role]"); map.Key(k => { k.Column("[UserId]"); });
            },
            rel => {
                rel.ManyToMany(map => { map.Class(typeof(Role)); map.Column("[RoleId]");
                });
            }
        );
    }
} public class RoleMapping : ClassMapping<Role> { public RoleMapping() {
        Table("[Role]");

        Id(m => m.Id, map => { map.Column("[Id]"); map.Type(NHibernateUtil.Int32); map.Generator(Generators.Identity);
        });

        Property(m => m.Name, map => { map.Column("[Name]"); map.Type(NHibernateUtil.String);
        });


        Bag(
            m => m.Users, map => { map.Table("[User_Role]"); map.Key(k => { k.Column("[RoleId]"); }); map.Inverse(true);
            },
            rel => {
                rel.ManyToMany(map => { map.Class(typeof(User)); map.Column("[UserId]");
                });
            }
        );

    }
}

数据库关系图以下：

当向用户添加或删除角色是， 发现更新的效率特别低， 代码以下：

using (var session = sessionFactory.OpenSession()) { var user = session.Query<User>().First(); var firstRole = user.Roles.First();
    user.Roles.Remove(firstRole);
    session.Update(user); var roleCount = session.Query<Role>().Count(); var role = new Role { Name = "Role " + (roleCount + 1) };
    session.Save(role);

    user.Roles.Add(role);
    session.Update(user);

    session.Update(user);
    session.Flush();
}

上面的代码是将用户的第一个角色删除， 再添加一个新的角色， NHibernate 生成的 SQL 语句以下（仅包含对关系表User_Role的操做）：

DELETE FROM [User_Role] WHERE [UserId] = @p0;@p0 = 1 [Type: Int32 (0)] INSERT INTO [User_Role] ([UserId], [RoleId]) VALUES (@p0, @p1);@p0 = 1 [Type: Int32 (0)], @p1 = 2 [Type: Int32 (0)] INSERT INTO [User_Role] ([UserId], [RoleId]) VALUES (@p0, @p1);@p0 = 1 [Type: Int32 (0)], @p1 = 7 [Type: Int32 (0)] INSERT INTO [User_Role] ([UserId], [RoleId]) VALUES (@p0, @p1);@p0 = 1 [Type: Int32 (0)], @p1 = 6 [Type: Int32 (0)] INSERT INTO [User_Role] ([UserId], [RoleId]) VALUES (@p0, @p1);@p0 = 1 [Type: Int32 (0)], @p1 = 10 [Type: Int32 (0)]

竟然是先将属于该用户的所有角色删除， 再添加一份新的进来， 彻底没法接受， 反过来思考以为确定是本身的问题， 通过一番搜索 (Google)， 发现 StackOverflow 上也有人问相似的问题， 而且最终在 NHibernate Tip: Use set for many-to-many associations 发现了解决方案， 将多对多的映射的bag改成用set， 问题终于获得了解决， 改事后的映射以下：

Set(
    m => m.Roles, map => { map.Table("[User_Role]"); map.Key(k => { k.Column("[UserId]"); });
    },
    rel => {
        rel.ManyToMany(map => { map.Class(typeof(Role)); map.Column("[RoleId]");
        });
    }
);

将UserMapping和RoleMapping中多对多映射所有改成Set以后， 上面的测试代码生成的 SQL 以下：

DELETE FROM [User_Role] WHERE [UserId] = @p0 AND [RoleId] = @p1;@p0 = 1 [Type: Int32 (0)], @p1 = 8 [Type: Int32 (0)] INSERT INTO [User_Role] ([UserId], [RoleId]) VALUES (@p0, @p1);@p0 = 1 [Type: Int32 (0)], @p1 = 9 [Type: Int32 (0)]

在 NHibernate 参考文档的 19.5. Understanding Collection performance 中这样描述：

Bags are the worst case. Since a bag permits duplicate element values and has no index column, no primary key may be defined. NHibernate has no way of distinguishing between duplicate rows. NHibernate resolves this problem by completely removing (in a single DELETE) and recreating the collection whenever it changes. This might be very inefficient.

不仅是多对多， 若是你的集合须要更新， NHibernate 推荐的是：

19.5.2. Lists, maps, idbags and sets are the most efficient collections to update

然而 bags 也不是一无可取：

19.5.3. Bags and lists are the most efficient inverse collections

Just before you ditch bags forever, there is a particular case in which bags (and also lists) are much more performant than sets. For a collection with inverse="true" (the standard bidirectional one-to-many relationship idiom, for example) we can add elements to a bag or list without needing to initialize (fetch) the bag elements! This is because IList.Add() must always succeed for a bag or IList (unlike an ISet). This can make the following common code much faster.

Parent p = sess.Load(id); Child c = new Child(); c.Parent = p;
p.Children.Add(c); //no need to fetch the collection! sess.Flush();

因而可知，bag在多对多映射更新时性能较差， 若是不须要更新，则能够放心使用， 在须要更新时则set是更好的选择。