I have an application where you can create a new type of product and add to that product some ingredients. The product and the ingredients are both entities saved in a database. The product entity has a collection of ingredient entities.
(simplified version)
public class Product
Public Sub New()
Me.Ingredients = New List(Of Ingredient)()
End Sub
Property Ingredients as ICollection(Of Ingredient)
end class
When I save the product for the first time, all goes well: I just add it to the context and call SaveChanges.
myDataContext.Products.Add(product)
myDataContext.SaveChanges()
Both the product (parent) and the ingredients (children) are saved and linked to each other. All is well.
However when I add/remove an ingredient to an existing product, I start running into problems. I first clear the existing ingredients collection in the product entity and then add the updated list of ingredients again (I don't re-use ingredients add the moment). I then change the state of the product entity to modified and call savechanges. On the state changing I, however, get the exception "An object with the same key already exists in the ObjectStateManager".
myDataContext.Entry(product).State = EntityState.Modified
After "some" searching I figured out that the problem is that all the ingredients have a primary key of 0 (as they aren't added yet) and when you change the state of the parent entity (product), all child entities (ingredients) are attached to the context with the key of 0, which causes the problem as the keys are no longer unique.
I have been searching for a solution but can't figure out how to solve this problem. I tried adding the ingredients to the context before changing the state, but then the link between the product and ingredients is missing... How do I update an existing parent entity with new, not yet added child entities?
I use Entity Framework 4.1 and Code First.
Hope you can help me!
I first clear the existing ingredients collection in the product
entity and than add the updated list of ingredients again.
Well, this is kind of brute-force-attack to update the child collection. EF doesn't have any magic to update the children - which means: adding new children, deleting removed children, updating existing children - by only setting the state of the parent to Modified. Basically this procedure forces you to delete the old children also from the database and insert the new one, like so:
// product is the detached product with the detached new children collection
using (var context = new MyContext())
{
var productInDb = context.Products.Include(p => p.Ingredients)
.Single(p => p.Id == product.Id);
// Update scalar/complex properties of parent
context.Entry(productInDb).CurrentValues.SetValues(product);
foreach (var ingredient in productInDb.Ingredients.ToList())
context.Ingredients.Remove(ingredient);
productInDb.Ingredients.Clear(); // not necessary probably
foreach (var ingredient in product.Ingredients)
productInDb.Ingredients.Add(ingredient);
context.SaveChanges();
}
The better procedure is to update the children collection in memory without deleting all children in the database:
// product is the detached product with the detached new children collection
using (var context = new MyContext())
{
var productInDb = context.Products.Include(p => p.Ingredients)
.Single(p => p.Id == product.Id);
// Update scalar/complex properties of parent
context.Entry(productInDb).CurrentValues.SetValues(product);
var ingredientsInDb = productInDb.Ingredients.ToList();
foreach (var ingredientInDb in ingredientsInDb)
{
// Is the ingredient still there?
var ingredient = product.Ingredients
.SingleOrDefault(i => i.Id == ingredientInDb.Id);
if (ingredient != null)
// Yes: Update scalar/complex properties of child
context.Entry(ingredientInDb).CurrentValues.SetValues(ingredient);
else
// No: Delete it
context.Ingredients.Remove(ingredientInDb);
}
foreach (var ingredient in product.Ingredients)
{
// Is the child NOT in DB?
if (!ingredientsInDb.Any(i => i.Id == ingredient.Id))
// Yes: Add it as a new child
productInDb.Ingredients.Add(ingredient);
}
context.SaveChanges();
}
I found this recent article on the GraphDiff extension for DbContext.
Apparently it is a generic, reusable variant of Slauma's solution.
Example code:
using (var context = new TestDbContext())
{
// Update DBcompany and the collection the company and state that the company 'owns' the collection Contacts.
context.UpdateGraph(company, map => map.OwnedCollection(p => p.Contacts));
context.SaveChanges();
}
On a side note; I see the author has proposed to the EF team to use his code in issue #864 Provide better support for working with disconnected entities.
I reckon, This is more simpler solution.
public Individual
{
.....
public List<Address> Addresses{get;set;}
}
//where base.Update from Generic Repository
public virtual void Update(T entity)
{
_dbset.Attach(entity);
_dataContext.Entry(entity).State = EntityState.Modified;
}
//overridden update
public override void Update(Individual entity)
{
var entry = this.DataContext.Entry(entity);
var key = Helper.GetPrimaryKey(entry);
var dbEntry = this.DataContext.Set<Individual>().Find(key);
if (entry.State == EntityState.Detached)
{
if (dbEntry != null)
{
var attachedEntry = this.DataContext.Entry(dbEntry);
attachedEntry.CurrentValues.SetValues(entity);
}
else
{
base.Update(entity);
}
}
else
{
base.Update(entity);
}
if (entity.Addresses.Count > 0)
{
foreach (var address in entity.Addresses)
{
if (address != null)
{
this.DataContext.Set<Address>().Attach(address);
DataContext.Entry(address).State = EntityState.Modified;
}
}
}
}
after many many months struggling with understanding this entire crappy Entity Framework I hope this can help someone and not go through any of the frustration I have endured.
public void SaveOrder(SaleOrder order)
{
using (var ctx = new CompanyContext())
{
foreach (var orderDetail in order.SaleOrderDetails)
{
if(orderDetail.SaleOrderDetailId == default(int))
{
orderDetail.SaleOrderId = order.SaleOrderId;
ctx.SaleOrderDetails.Add(orderDetail);
}else
{
ctx.Entry(orderDetail).State = EntityState.Modified;
}
}
ctx.Entry(order).State = order.SaleOrderId == default(int) ? EntityState.Added : EntityState.Modified;
ctx.SaveChanges();
}
}
Related
I am trying to figure out a way to pass a collection of include statements into my repository so that I can have it include specific entities. Below is some sample code from my repository.
public TEntity GetById(Guid id)
{
return id != Guid.Empty ? GetSet().Find(id) : null;
}
private IDbSet<TEntity> GetSet()
{
return _unitOfWork.CreateSet<TEntity>();
}
The GetByID method calls the GetSet to return the entity set. I was thinking, if I could somehow pass in a collection of entities to include (via an expression) as part of my GetById, this way I wouldn't have to expose the GetSet to my services. So, something like this:
var entity = _repository.GetById(theId, e => {e.Prop1, e.Prop2, e.Prop3});
I could then pass that expression into my GetSet method and pass it into an include statement. Thoughts?
I have done something like this in my code recently. Would the following work for you?
public TEntity GetById(Guid id, params Expression<Func<TEntity, object>>[] includeProperties)
{
if (id == Guid.Empty) return null;
var set = _unitOfWork.CreateSet<TEntity>();
foreach(var includeProperty in includeProperties)
{
set.Include(includeProperty);
}
return set.First(i => i.Id == id);
}
Then you would call it like this...
var entity = _repository.GetById(theId, e => e.Prop1, e=> e.Prop2, e=> e.Prop3);
I know this doesn't exactly follow your pattern, but I think you could refactor it as required.
I don't think Paige Cook's code will work quite as shown.
I've included a modified version of the code that should work instead:
public TEntity GetById(Guid id, params Expression<Func<TEntity, object>>[] includeProperties)
{
if (id == Guid.Empty) return null;
IQueryable<TEntity> set = _unitOfWork.CreateSet<TEntity>();
foreach(var includeProperty in includeProperties)
{
set = set.Include(includeProperty);
}
return set.First(i => i.Id == id);
}
I only spotted this by tracing the SQL generated by Entity Framework, and realised the original code was only giving the illusion of working, by using lazy-loading to populate the entities specified for inclusion.
There's actually a more terse syntax for applying the Include statements using the LINQ Aggregate method, which is in the blog post linked to. My post also improves the method slightly by having a fall-back to the Find method, when no includes are needed and also shows an example of how to implement a "GetAll" method, using similar syntax.
It's bad idea to store context in non-local space, for many reasons.
I modify Steve's code and get this for my ASP.NET MVC projects:
public aspnet_User FirstElement(Func<aspnet_User, bool> predicate = null, params Expression<Func<aspnet_User, object>>[] includes)
{
aspnet_User result;
using (var context = new DataContext())
{
try
{
var set = context.Users.AsQueryable();
for (int i = 0; i < includes.Count(); i++ )
set = set.Include(includes[i]);
if (predicate != null)
result = set.ToList().FirstOrDefault(predicate);
else
result = set.ToList().FirstOrDefault();
}
catch
{
result = null;
}
}
return result;
}
The include method can be strung together in your linq query like so:
var result = (from i in dbContext.TableName.Include("RelationProperty")
.Include("RelationProperty")
.Include("RelationProperty")
select i);
Currently this is what I have:
public partial class LinqToSqlEntity {
public IQueryable<AnotherLinqToSqlEntity> AnotherLinqToSqlEntities {
using(DataContext context = new DataContext) {
return context.AnotherLinqToSqlEntities.Where(item => item.Property == SOME_VALUE);
}
}
}
Is there a way to get the DataContext of this so that I would not need to create a new DataContext?
Sorry, that is not possible. An entity or querable in that case keeps no direct reference of the context.
You can achieve that using the reflection by figuring out if PropertyChanging event was hooked up, but consider this a hack and maybe you can avoid using it with better design.
Our use case of this is on detach_EntityName delegate where we change the default Linq behaviour of only deleting the foreign key of a record (setting it to null), with the actual delete from DB.
public static DataContext GetDataContextFromEntityObject(object entity)
{
// Use a reflection to get the invocaiton list.
var o = (PropertyChangingEventHandler)entity.GetType().GetField("PropertyChanging", BindingFlags.Instance | BindingFlags.NonPublic).GetValue(entity);
var o = GetFieldValue(entity, "PropertyChanging");
if (o == null) return null;
var invocationList = o.GetInvocationList();
if (invocationList != null)
{
// DataContext changes are tracked through StandardChangeTracker
object changeTracker = (from i in invocationList where i.Target.GetType().FullName == "System.Data.Linq.ChangeTracker+StandardChangeTracker" select i.Target).FirstOrDefault();
if (changeTracker != null)
{
object services = GetFieldValue(changeTracker, "services");
return (DataContext)GetFieldValue(services, "context");
}
}
return null;
}
private static object GetFieldValue(object instance, string propertyName)
{
return instance.GetType().GetField(propertyName, BindingFlags.Instance | BindingFlags.NonPublic).GetValue(instance);
}
I've got a class T with an association to a cross reference table. I'd like to have a method on the class that adds an entry to the cross reference by receiving an entity representing the other side. For some reason, though, while I can add the item to the collection it doesn't get added to the change set for the data context.
So the class in question looks like:
class T
{
public EntitySet<T_U> t_users
{
get
{
if ((this.serializing && (this.t_user.HasLoadedOrAssignedValues == false)))
{
return null;
}
return this.t_user;
}
set
{
this.Users.Assign(value);
}
}
public AddUser(U user)
{
this.t_users.Add( new T_U() { TID = this.ID, UID = user.ID );
}
}
The client using this class would basically do something like:
var db = new DBDataContext();
var t = db.Ts.FirstOrDefault( t => t.ID = 100);
var u = db.Us.FirstOrDefault(u => u.ID == 3);
t.AddUser(u);
db.SubmitChanges();
Shouldn't that succesfully add a record to the cross reference table?
I think you want InsertOnSubmit instead of Add. Are you using an outdated version of Linq to sql?
You are adding new T_U to this.Users... Does that work? Isn't that the wrong type?
this.Users.Add( new T_U() { TID = this.ID, UID = user.ID );
I'm using LINQ To SQL to update a user address.
I'm trying to track what fields were updated.
The GetChangeSet() method just tells me I'm updating an entity, but doesn't tell me what fields.
What else do I need?
var item = context.Dc.Ecs_TblUserAddresses.Single(a => a.ID == updatedAddress.AddressId);
//ChangeSet tracking
item.Address1 = updatedAddress.AddressLine1;
item.Address2 = updatedAddress.AddressLine2;
item.Address3 = updatedAddress.AddressLine3;
item.City = updatedAddress.City;
item.StateID = updatedAddress.StateId;
item.Zip = updatedAddress.Zip;
item.Zip4 = updatedAddress.Zip4;
item.LastChangeUserID = request.UserMakingRequest;
item.LastChangeDateTime = DateTime.UtcNow;
ChangeSet set = context.Dc.GetChangeSet();
foreach (var update in set.Updates)
{
if (update is EberlDataContext.EberlsDC.Entities.Ecs_TblUserAddress)
{
}
}
Use ITable.GetModifiedMembers. It returns an array of ModifiedMemberInfo objects, one for each modified property on the entity. ModifiedMemberInfo contains a CurrentValue and OriginalValue, showing you exactly what has changed. It's a very handy LINQ to SQL feature.
Example:
ModifiedMemberInfo[] modifiedMembers = context.YourTable.GetModifiedMembers(yourEntityObject);
foreach (ModifiedMemberInfo mmi in modifiedMembers)
{
Console.WriteLine(string.Format("{0} --> {1}", mmi.OriginalValue, mmi.CurrentValue));
}
You can detect Updates by observing notifications of changes. Notifications are provided through the PropertyChanging or PropertyChanged events in property setters.
E.g. you can extend your generated Ecs_TblUserAddresses class like this:
public partial class Ecs_TblUserAddresses
{
partial void OnCreated()
{
this.PropertyChanged += new PropertyChangedEventHandler(User_PropertyChanged);
}
protected void User_PropertyChanged(object sender, PropertyChangedEventArgs e)
{
string propertyName = e.PropertyName;
// do what you want
}
}
Alternatively, if you want to track a special property changing, you could use one of those OnPropertyNameChanging partial methods, e.g. (for City in your example):
partial void OnCityChanging(string value)
{
// value parameter holds a new value
}
Let's say we have two tables with a many-to-many relationship:
public class Left{ /**/ }
public class Right{ /**/ }
public class LeftRight{ /**/ }
is the following sufficient to unhook these records (ignore the possibility of more than one relationship or no relationship defined)?
public void Unhook(Left left, Right right){
var relation = from x in Left.LeftRights where x.Right == right;
left.LeftRrights.Remove(relation.First());
Db.SubmitChanges();
}
Or do I have to do it on both parts? What's required here?
Here is a 'little' extension method I wrote to simplify this problem:
public static class EntitySetExtensions
{
public static void UpdateReferences<FK, FKV>(
this EntitySet<FK> refs,
Func<FK, FKV> fkvalue,
Func<FKV, FK> fkmaker,
Action<FK> fkdelete,
IEnumerable<FKV> values)
where FK : class
where FKV : class
{
var fks = refs.Select(fkvalue).ToList();
var added = values.Except(fks);
var removed = fks.Except(values);
foreach (var add in added)
{
refs.Add(fkmaker(add));
}
foreach (var r in removed)
{
var res = refs.Single(x => fkvalue(x) == r);
refs.Remove(res);
fkdelete(res);
}
}
}
It could probably be improved, but it has served me well :)
Example:
Left entity = ...;
IEnumerable<Right> rights = ...;
entity.LeftRights.UpdateReferences(
x => x.Right, // gets the value
x => new LeftRight { Right = x }, // make reference
x => { x.Right = null; }, // clear references
rights);
Algorithm description:
Suppose A and B is many-to-many relationship, where AB would be the intermediary table.
This will give you:
class A { EntitySet<B> Bs {get;} }
class B { EntitySet<A> As {get;} }
class AB { B B {get;} A A {get;} }
You now have an object of A, that reference many B's via AB.
Get all the B from A.Bs via 'fkvalue'.
Get what was added.
Get what was removed.
Add all the new ones, and construct AB via 'fkmaker'.
Delete all the removed ones.
Optionally, remove other referenced objects via 'fkdelete'.
I would like to improve this by using Expression instead, so I could 'template' the method better, but it would work the same.
Take two, using expressions:
public static class EntitySetExtensions
{
public static void UpdateReferences<FK, FKV>(
this EntitySet<FK> refs,
Expression<Func<FK, FKV>> fkexpr,
IEnumerable<FKV> values)
where FK : class
where FKV : class
{
Func<FK, FKV> fkvalue = fkexpr.Compile();
var fkmaker = MakeMaker(fkexpr);
var fkdelete = MakeDeleter(fkexpr);
var fks = refs.Select(fkvalue).ToList();
var added = values.Except(fks);
var removed = fks.Except(values);
foreach (var add in added)
{
refs.Add(fkmaker(add));
}
foreach (var r in removed)
{
var res = refs.Single(x => fkvalue(x) == r);
refs.Remove(res);
fkdelete(res);
}
}
static Func<FKV, FK> MakeMaker<FKV, FK>(Expression<Func<FK, FKV>> fkexpr)
{
var me = fkexpr.Body as MemberExpression;
var par = Expression.Parameter(typeof(FKV), "fkv");
var maker = Expression.Lambda(
Expression.MemberInit(Expression.New(typeof(FK)),
Expression.Bind(me.Member, par)), par);
var cmaker = maker.Compile() as Func<FKV, FK>;
return cmaker;
}
static Action<FK> MakeDeleter<FK, FKV>(Expression<Func<FK, FKV>> fkexpr)
{
var me = fkexpr.Body as MemberExpression;
var pi = me.Member as PropertyInfo;
var par = Expression.Parameter(typeof(FK), "fk");
var maker = Expression.Lambda(
Expression.Call(par, pi.GetSetMethod(),
Expression.Convert(Expression.Constant(null), typeof(FKV))), par);
var cmaker = maker.Compile() as Action<FK>;
return cmaker;
}
}
Now the usage is uber simple! :)
Left entity = ...;
IEnumerable<Right> rights = ...;
entity.LeftRights.UpdateReferences(x => x.Right, rights);
The first expression is now used to establish the 'relationship'. From there I can infer the 2 previously required delegates. Now no more :)
Important:
To get this to work properly in Linq2Sql, you need to mark the associations from intermediary table with 'DeleteOnNull="true"' in the dbml file. This will break the designer, but still works correctly with SqlMetal.
To unbreak the designer, you need to remove those additional attributes.
Personally, I'd replace
left.LeftRrights.Remove(relation.First());
with
Db.LeftRights.DeleteAllOnSubmit(relation)
because it seems more obvious what's going to happen. If you are wondering what the behaviour of ".Remove" is now, you'll be wondering anew when you look at this code in 6 months time.