Setup
Using MVC 3 + Code First
Here are my classes
public class Member
{
[Key]
public Guid ID { get; set; }
[Required]
public String Email { get; set; }
[Required]
public String FirstName { get; set; }
[Required]
public String LastName { get; set; }
public String Sex { get; set; }
public String Password { get; set; }
public String PasswordSalt { get; set; }
public DateTime RegisterDate { get; set; }
public DateTime LastOnline { get; set; }
public String SecurityQuestion { get; set; }
public String SecurityAnswer { get; set; }
public virtual ICollection<FamilyMember> Families { get; set; }
public virtual ICollection<Relationship> Relationships { get; set; }
}
public class Relationship
{
[Key]
public Guid ID { get; set; }
[ForeignKey("Member1")]
public Guid Member1ID { get; set; }
[ForeignKey("Member2")]
public Guid Member2ID { get; set; }
public Guid RelationshipTypeID { get; set; }
public virtual RelationshipType RelationshipType { get; set; }
public virtual Member Member1 { get; set; }
public virtual Member Member2 { get; set; }
}
Here is the problem
The database table "Relationship" is being created with the following columns:
ID, Member1ID, Member2ID, RelationshipTypeID, Member_ID
Why is it creating the Member_ID column?
I've seen this post in which the user has the same type of setup, but I am unsure of how to define the InverseProperty correctly. I tried using fluent API calls but from what I can tell they will not work here since I have two foreign keys referring to the same table.
Any help would be appreciated!
Member_ID is the foreign key column which EF created for the navigation property Member.Relationships. It belongs to a third association from Member.Relationships refering to an end endpoint which is not exposed in your Relationship entity. This relationship has nothing to do with the other two relationships from Relationship.Member1 and Relationship.Member2 which also both have an endpoint not exposed in Member.
I guess, this is not what you want. You need always pairs of endpoints in two entities to create an association. One endpoint is always a navigation property. The second endpoint can also be a navigation property but it is not required, you can omit the second navigation property.
Now, what is not possible, is to associate two navigation properties (Member1 and Member2) in one entity with one navigation property (Relationships) in the other entity. That is what you are trying to do apparently.
I assume that your Member.Relationships property is supposed to express that the member is either Member1 or Member2 in the relationship, or that it participates in the relationship, no matter if as Member1 or Member2.
Unfortunately you cannot express this in the model appropriately. You have to introduce something like RelationsshipsAsMember1 and RelationsshipsAsMember2 and for these two collection you can use the InverseProperty attribute as shown in the other question. In addition you can add a helper property which concats the two collections. But this is not a mapped property but readonly:
public class Member
{
// ...
[InverseProperty("Member1")]
public virtual ICollection<Relationship> RelationshipsAsMember1 { get; set; }
[InverseProperty("Member2")]
public virtual ICollection<Relationship> RelationshipsAsMember2 { get; set; }
public IEnumerable<Relationship> AllRelationships
{
get { return RelationshipsAsMember1.Concat(RelationshipsAsMember2); }
}
}
Accessing AllRelationships will cause two queries and roundtrips to the database (with lazy loading) to load both collections first before they get concatenated in memory.
With this mapping the Member_ID column will disappear and you will only get the two expected foreign key columns Member1ID, Member2ID because now you have only two associations and not three anymore.
You could also think about if you need the Relationships collection in the Member entity at all. As said, navigation properties on both sides are not required. If you rarely need to navigate from a member to its relationships you could fetch the relationships also with queries on the Relationship set, like so:
var relationships = context.Relationships
.Where(r => r.Member1ID == givenMemberID || r.Member2ID == givenMemberID)
.ToList();
...or...
var relationships = context.Relationships
.Where(r => r.Member1ID == givenMemberID)
.Concat(context.Relationships
.Where(r => r.Member2ID == givenMemberID)
.ToList();
This would give you all relationships the member with ID = givenMemberID participates in without the need of a navigation collection on the Member entity.
Related
Note: These classes are related, but not part of the same Aggregate (like PurchaseOrder and OrderLine) - so I do not have a navigation property from "One" to "Many".
=== Entities ===
public class One
{
public Guid Id { get; set; }
public string Name { get; set; }
}
public class Many
{
public Guid Id { get; set; }
public string Name { get; set; }
public Guid One { get; set; }
}
=== Contracts ===
public class OneWithMany
{
public Guid Id { get; set; }
public string Name { get; set; }
public IEnumerable<Many>? ManyRelatedObjects { get; set; }
}
I want to select all One objects and any related Many objects from DbSet/DbSet into OneWithMany.
To ensure I don't miss properties added in future I am using ProjectTo in AutoMapper - but I can't work out how to fit it into the equation.
Unfortunately, it seems Entity Framework does not support GroupJoin.
The solution is to do the projection and as much filtering as possible as two separate queries, and then combine them into a result in memory.
If you find EF related answers on the web related to GroupJoin make sure you check the example code to see if they are actually showing code working on arrays instead of DbSet.
I have added computed fields(Active and CreditsLeft) directly into my CodeFirst entity class. Is it good idea to add computed field logic inside CF Entity class?
public class User : Entity
{
public User()
{
Id = Helper.GetRandomInt(9);
DateStamp = DateTime.UtcNow;
TimeZone = TimeZoneInfo.Utc.Id;
}
[DatabaseGenerated(DatabaseGeneratedOption.None)]
public int Id { get; set; }
[Required]
[MaxLength(50)]
public string Email { get; set; }
[Required]
[MaxLength(50)]
public string Password { get; set; }
[MaxLength(50)]
public string FirstName { get; set; }
[MaxLength(50)]
public string LastName { get; set; }
[Required]
public DateTime DateStamp { get; set; }
public virtual ICollection<Order> Orders { get; set; }
public virtual ICollection<Statistic> Statistics { get; set; }
public virtual ICollection<Notification> Notifications { get; set; }
[DatabaseGenerated(DatabaseGeneratedOption.Computed)]
public bool Active
{
get
{
return Orders.Any(c => c.Active && (c.TransactionType == TransactionType.Order || c.TransactionType == TransactionType.Subscription));
}
}
[DatabaseGenerated(DatabaseGeneratedOption.Computed)]
public int CreditsLeft
{
get
{
return Orders.Sum(p => p.Credits != null ? p.Credits.Value : 0);
}
}
}
Is it good idea to add computed field logic inside CF Entity class?
Sure, you can do this, but there are a few things you must take care of.
First, the attribute for a property that is computed by business logic is not [DatabaseGenerated(DatabaseGeneratedOption.Computed)], because this indicates that the value is computed in the database (as in a computed column). You should mark the property by the [NotMapped] attribute. This tells Entity Framework to ignore the property in database mapping.
Second, since both properties use Orders, you must make sure that the orders are loaded or can be lazy loaded when either property is accessed. So you may want to load Users with an Include statement (Include(user => user.Orders)). Or else you must ensure that the context is still alive when Active or CreditsLeft is accessed.
Third, you can't address the properties directly in an EF LINQ query, as in
db.Users.Select(u => u.Active);
because EF will throw an exception that it doesn't know Active. You can address the properties only on materialized user objects in memory.
Given the following:
public class Department
{
public int DepartmentID { get; set; }
public string Name { get; set; }
public virtual ICollection<Course> Courses { get; set; }
}
public class Course
{
public int CourseID { get; set; }
public string Title { get; set; }
public int Credits { get; set; }
public int DepartmentID { get; set; }
public virtual Department Department { get; set; }
}
If I turn lazy loading off and issue the following:
var departments = _DepartmentRepository.GetAll()
.Include(c => c.courses);
Then I get the answers with a Department object inside of them.
Is there a way I can just include the courses and not get back the Department object. For example can I just include one level (courses).
You are just including one level. The department object inside the course is there because EF has done some relationship fixup so that you can navigate to the department from the course.
If you don't want departments then just get the courses directly. That is context.Courses.ToList(); or via a courses repo if you have one.
When fetching entities EF will automatically populate navigation properties where it is already tracking the target object. This means if you do say:
// Load the department with a PK of 1
_DepartmentRepository.Find(1);
and then, using the same context, for example:
// Load a course with PK of 17
_CourseRepository.Find(17);
If this courses department id is 1, then EF will have automatically populated it's Department navigation property even though you didn't specify the include. You could stop this behavior by not making the Department navigation property virtual.
So I decided to go with the code first/DbContext approach, but already have an existing database file. Nothing complex, so I am thinking I can just create the DbContext derived container class with DbSets for the respective POCO's, create the connection string to my database and I should be set. However I believe I am having difficulties properly declaring the properties in my entity classes since I am getting errors when trying access an object through the navigational properties. Usually telling me Object reference not set to an instance of an object when I try context.Products.Find(1).Category.CATNAME; etc. Also tried declaring the collection properties with virtual keyword to no avail.
Some specifics of the database schema are:
In Categories table the PCATID is a foreign key to the CategoryID in
the same Categories table and can be null.
Both CategoryID and RootCategoryID in Products table can be null and
are both foreign keys to CategoryID in the Categories table.
I am testing things at the moment but will be setting a lot of the fields to non null types eventually.
Here are my entity POCO's and the entity Dbset container class:
public class Category
{
[Key]
public int CategoryID { get; set; }
public string CATNAME { get; set; }
public int PCATID { get; set; }
public ICollection<Category> Categories { get; set; }
public ICollection<Product> Products { get; set; }
}
public class Product
{
[Key]
public int ProductID { get; set; }
public int CategoryID { get; set; }
public int RootCategoryID { get; set; }
public string Name { get; set; }
public string ShortDescription { get; set; }
public string LongDescription { get; set; }
public string Keywords { get; set; }
public decimal ListPrice { get; set; }
public Category Category { get; set; }
}
public class EFDbContext: DbContext
{
public DbSet<Product> Products { get; set; }
public DbSet<Category> Categories { get; set; }
}
You need to make PCATID a nullable property as you have said it can be null. Make all those navigation properties and collection properties virtual. EF will not be able to detect the category hierarchy so you have use either attributes or fluent API to configure that.
public class Category
{
[Key]
public int CategoryID { get; set; }
public string CATNAME { get; set; }
[ForeignKey("ParentCategory")]
public int? PCATID { get; set; }
[InverseProperty("Categories")]
public virtual Category ParentCategory { get; set; }
[InverseProperty("ParentCategory")]
public virtual ICollection<Category> Categories { get; set; }
public virtual ICollection<Product> Products { get; set; }
}
Requirements for Creating POCO Proxies
Everything looks ready for POCO but Lazy Loading isn't sorted out at this point. By default LL is on, but in order to enable lazy loading, the Category property must be Virtual (a proxy is created that catches the reference and loads the data). If you don't want lazy loading then disable it in your EFDbContext constructor.
So your options are:
public virtual Category Category { get; set; }
or
public class EFDbContext: DbContext
{
public static EFDbContext()
{
LazyLoadingEnabled = false
}
...
}
You'd probably want to do the first one...
Are you certain you really want to use Code First? Or do you just want to use DbContext and DbSet? You can get the same benefits with Database First, using DbContext and DbSet. Since you already have a database, it's generally a lot simpler.
See: http://blogs.msdn.com/b/adonet/archive/2011/03/15/ef-4-1-model-amp-database-first-walkthrough.aspx
The only difference between Code First and Database First with DbContext is that Code first uses the fluent mapping model, while Database First uses an .edmx file. Maintaining the .edmx is much easier with an existing database.
If you're bound and determined to use Code First, then I suggest getting the Entity Framework Power Tools CTP1 and reverse engineering your database to Code First.
I agree with #Eranga about class Category (+1 to #Eranga).
public class Category {
[Key]
public int CategoryID { get; set; }
public string CATNAME { get; set; }
[ForeignKey("ParentCategory")]
public int? PCATID { get; set; }
[InverseProperty("Categories")]
public virtual Category ParentCategory { get; set; }
[InverseProperty("ParentCategory")]
public virtual ICollection<Category> Categories { get; set; }
public virtual ICollection<Product> Products { get; set; }
}
And you also have problem with your Linq query:
context.Products.Find(1).Category.CATNAME;
EF return data only from tables that you request with Include or you use this tables in functions.
With this code all work:
db.Products
.Include(p => p.Category) // here I demand to load data from Category table
.First(p => p.ProductID == 3)
.Category
.CATNAME;
I'm using the Entity Framework 4 code first approach to design my database in ASP MVC 3 and I ran into a bit of a hitch. I have a POCO class as below:
public class User
{
[Key]
public Guid UserID { get; set; }
public string UserName { get; set; }
public DateTime CreationDate { get; set; }
public string Email { get; set; }
public int Points { get; set; }
public Session ActiveSession { get; set; }
public ICollection<Session> InSessions { get; set; }
}
with Session as another one of my model classes defined elsewhere, with ICollection<User> as one of its properties. If I remove the public Session ActiveSession { get; set; } property from the User class, then the many-to-many mapping and UserSessions intermediate table are constructed correctly, but when I add the ActiveSession one-to-one mapping back in, it breaks the many-to-many mapping and the intermediate table is not constructed. Instead the Users table has a single foreign key to the Sessions table each for both the ActiveSession and InSessions properties. Any ideas why this is happening?
In your case EF thinks that ICollection<User> in Session class is the Many end of the one to many relationship created by ActiveSession.
So you need to configure mannualy
modelBuilder.Entity<User>().HasOptional(u => u.ActiveSession)
.WithMany()
.Map(u => u.MapKey("ForeignKeyColumn"));
modelBuilder.Entity<User>()
.HasMany(u => u.InSessions)
.WithMany(s => s.Users)
.Map(m =>
{
m.ToTable("UserSessions");
m.MapLeftKey("UserID");
m.MapRightKey("SessionID");
});