WebAPI + EF5, Entity framework does not recover from a DBUpdateException and requires a full restart of the service - asp.net-web-api

I have two table in my database. Table USER and table USER_ADDRESSES
Table USER has UserID , UserName , HumanName
Table USER_ADDRESSES has UserAddresseID, UserID, Address
The two tables are related through the UserID foreign key.
I am working code first. The POCOs are as one would expect I guess:
Public class User
{
Public guid UserID {get;set;}
Public string UserName {get;set;}
Public string HumanName {get;set;}
}
Public class UserAddress
{
Public guid UserAddressID {get;set;}
Public guid UserID {get;set;}
Public string Address {get;set;}
Public virtual User User {get;set;}
}
If I POST a user, it flows through EF and into the physical database - cool
If I POST a UserAddress with the UserID foreign key properly assigned as the UserID of the above user, the userAddress flows through EF into the physical database - perfect.
Now as I am learning this subject, I like to test scenarios that I know will cause errors… Understanding how something fails often leads to greater understanding (IMHO).
To this end, I repeat the above steps.POST a user, then POST an Address but this time with an empty GUID in the Address.UserID property, something I know will cause SQL server to error on,this time EF gets an error back from the DB and surfaces that error in my controller class as a DbUpdateException, which I then act upon to send back a BadRequest ErrorResponse, this would be fine except now I can post nothing to the DB, If I subsequently try to post new Users EF throws a DbUpdateException on every POST. The only way out of this condition is to Stop debugging the WebAPI project, kill IISExpress and restart everything.
My suspicion is that it is something to do with the tracked entities in dbset.Local it appears this collection is happy to store entities which violate database constraints, and even after the DB has thrown an error EF does not remove the faulted entity from its cache, and I guess that every time I call DataContext.SaveChanges(), EF keeps trying to submit the Faulty address entity that it has in it's bdset.Local property and thus prevents me from submitting anything further to the DB until the whole project is restarted and the various dbset.Local collections are flushed.
Ok I understand that, In my controller class I can prevent some of these issues by evaluating the various properties of the object before submitting them to my repository/UnitOfWork and thus prevent them from getting into dbset.Local in the first place, perhaps there is some decoration I can add to my POCOs such that ModelState.IsValid returns false if UserID == Guid.Empty in the UserAddress Entity??? . But surely EF should be able to recover from DB originated errors, such that if an INSERT statement fails when run against SQL server, EF should remove the faulting Entity from its Local cache of tracked objects. Thus preventing the whole system from requiring a complete restart. The upshot of this is that if I had unique constraints on other non key fields, before I attempt to submit an entity I will have to query the DB to ensure that there are no records in the DB which would result in a DBUdateException from being thrown, that approach seems inefficient to me. I would rather take the error and throw it back to the client for them to consider.
Now perhaps it's my responsibility that on the occurrence of a DbUpdateException to do something to remove the faulted entity from the dbset.Local collection or similar. If that's the case how should I go about this?
As I have been writing this I am left wondering: Should EF not be able to recover from DBUdateExceptions in and of its self!? Without the need for an IIS restart. Or should I build my DB with no constraints what so ever and allow clients to put any old rubbish in there. It's my believe that EF Entity tracking should honour db constraints and prevent faulted entities from getting into the dbset in the first place.
Please guide me on the correct way to recover from a DBUpdateException?

Related

asp.net core Identity user customization

Detail
I am developing web application in asp.net core with Identity. now in my application I have two kind of user. Customer and Partner both have different profile information and login scenario.customer can login from simple signup from web page but partner can signup from different view with different mandatory fields.
Problem
How can I design Schema.
what are the good practices in this case.
What are the drawback.
Code
This is what I have done so far
public class ApplicationUser : IdentityUser
{
public CustomerProfile CustomerProfile { get; set; }
}
Use inheritance:
public class ApplicationUser : IdentityUser {}
public class Customer : ApplicationUser
{
// Customer-specific properties
}
public class Partner : ApplicationUser
{
// Partner-specific properties
}
By default, this will be implemented via STI (single-table inheritance). That means you'll have just your standard AspNetUsers table containing columns for the properties on ApplicationUser and all derived types. A discriminator column will be added to indicate which type was actually saved, which will then be used to instantiate the right type when queried.
For the most part, this works just fine. The one downside is that properties on derived classes must be nullable. The reason is simple: it would be impossible to provide values for Customer columns while saving a Partner and vice versa. However, the properties only need be nullable at the database-level. You can still require that they be set in forms and such via a view model.
The alternative is to use TPT (table-per-type). With this approach, you'll get AspNetUsers, but also Customers and Partners tables as well. However, the tables for the derived types will have columns corresponding only to the properties specific to that type and a foreign key back to AspNetUsers. All common properties are stored there. With this, you can now enforce columns have values at the database-level, but querying users will require a join. To use TPT, you simply add the Table attribute to your class, i.e. [Table("Customers")] and [Table("Partners")], respectively.
The one important thing to keep in mind with using inheritance, though, is that you need to work with the type you actually want to be persisted. If you save an ApplicationUser instance, it will be an ApplicationUser, not a Customer or Partner. In this regard, you need to be careful with using the correct types with things like UserManager which generically reference the user type. Even if you create an instance of Customer, if you save it via an instance of UserManager<ApplicationUser>, it will upcast to ApplicationUser and that is what will be persisted. To create a new Customer, you'll need an instance of UserManager<Customer>. Likewise, for partners.
However, this also works to your benefit, as if you attempt to look up a user from an instance of UserManager<Customer> for example, you will only find them if they are in fact a Customer. In this way, it makes it trivially simple to have separate portals where only one or the other can log in, as you've indicated that you want.

ASP.NET MVC User.Identity.GetUserId() returning wrong GUID

I am calling this method from within a Controller method while logged in as a user, with the standard DefaultMembershipProvider code in MVC5 that comes completely out of the box. Like this:
string LoggedInUsersId = User.Identity.GetUserId();
When debugging, this method call returns a GUID that is completely different to the "Id" field GUID stored in the AspNetUsers table. As I only have one user in this table, I have no idea where this other GUID is coming from?
I thought that the Id field value from the AspNetUsers table would be returned for the currently logged in user when you call User.Identity.GetUserId()?
I am definitely checking the same database/environment and looking at the right place.
Solved!
Thanks to this blog post (see point #2):
https://digitaltoolfactory.net/blog/2013/10/five-things-should-should-know-about-the-new-asp-net-identity-membership-system/
The new MVC5 membership stuff had automatically created a new local database on my laptop called "DefaultConnection" where it had generated and placed all the membership tables and data. So GetUserId() was returning a GUID that lived locally in the AspNetUsers table from that local database which did not exist in my other main database that the application was using! Gah.
So the fix was changing this line inside of IdentityModels.cs:
public ApplicationDbContext() : base("DefaultConnection") { }
to:
public ApplicationDbContext() : base("MyConnectionString") { }
Now it works and everything is saved centrally into my application's main database.

Entity Framework - ObjectContext or DbContext when AuditLog and CreatedOn/ModifiedOn and DB Modeling Team

Which one to choose either ObjectContext or DbContext, when the requirements are,
Data Modeling done by Modeler and provides the DEV team a sql
script. Due to this we have opted Model First. Is this a correct
choice ?
Existing denormalized db will be migrated to new db created by
modeler
Need to maintain audit log for all the updates, at the field level,
from the UI
Each table has CreatedBy, CreatedOn, ModifiedBy, ModifiedOn. These
fields should be automatically filled by during
context.SaveChanges().
If you're starting a new app, just use DbContext. You can always drill down to ObjectContext if you need to.
If you prefer no designer, you can use Code First with Migrations and create a SQL Script via update-database -script as well.
Sounds like a task for the DBAs?
field by field changes..If this is a disconnected app, you'll be better off handling that outside of EF (IMHO)
you can easily override SaveChanges for this. You said in a tweet that you have the dbcontext book. There's an example of this where we do this using a base class. However if you are going to use model first, be sure to avoid this problem : http://msdn.microsoft.com/en-us/magazine/jj553510.aspx
Thanks a lot Julie for your super quick response. You are The-EF-Angel.
I have read your MSDN article on Logging in EF.
To your reponse:
1. As a mandate, We need to use sql scripts provided by our Modeler to create our db. Also these scripts will be keep changing(With addition of new tables & update of exising schema) for each sprints. Hope DataFirst Model is fine. Whenever new we get new sql scripts, we plan to recreate the DB and update our EDMX. Do you see any issues with this approach ?
2. Ya we have a migration specialist for this task. I justed pointed that in question as an FYI.
3. We use MVC app and for field by field changes in audit log table, we planned to let EF decide what fields have changed(using change tracking ideas from your book) and capture that info into a DTO(we borrow your repository ideas from the course "EF in enterprise course" you authored in PS). And push that DTO into our messaging infrastructure and that will insert the audit logs to the DB.
Is this fine ? Do you foresee any issues ?
4. As you pointed out, we could change our interface for our needs by referring to your MSDN article and there "Figure 3 Setting the Interface’s DateCreated Property During SaveChanges"
I plan to use,
public interface ITheBaseType
{
DateTime DateCreated { get; set; }
DateTime DateModified { get; set; }
string CreatedBy { get; set; }
string ModifiedBy { get; set; }
}

Updating object and relations with Entity Framework Code First and ASP.Net MVC

I'm using Entity Framework Code First and whilst I have working code, I'm having to make what are strictly unnecessary database calls in order to process the following update.
I have a simple POCO class for an album with a collection of related tags:
public class Album
{
public int Id { get; set; }
public string Title { get; set; }
public decimal Price { get; set; }
public virtual IList<Tag> Tags { get; private set; }
}
public class Tag
{
public int Id { get; set; }
public string Name { get; set; }
}
This is updated via an MVC form - with the tags represented by a series of check-boxes.
So when I get to my Update method in the respository, I have an album class populated with a list of tags - with in theory all I need to make the update.
However the only way I could find to get the list of tags to update (to delete any that were previously set but are now unchecked, and to add any that are currently checked) was to retrieve the original Album from the context and update it.
And secondly because in my implementation the Name field of the Tag is marked with [Required], and that in my Album object populated from the form I only have the IDs of the tags, I also have to retrieve each tag before updating.
Here's my code:
public void Update(Album album)
{
var albumToUpdate = GetById(album.Id); // - need to retrieve album with tags in order to update tags
albumToUpdate.Title = album.Title;
albumToUpdate.Price = album.Price;
albumToUpdate.Tags.Clear();
if (album.Tags != null)
{
foreach (var tag in album.Tags)
{
var tagToAdd = context.Tags.Find(tag.Id); // - need to retrieve full details of tag so doesn't fail validation
albumToUpdate.AddTag(tagToAdd);
}
}
}
Appreciate any thoughts as to how I could accomodate this with fewer database hits. It's not a major deal for this particular function (part of a site admin tool) but would like to know I'm doing things the best way.
Your approach - reloading the entity graph from the database and merge the changes manually into it - is correct in my opinion and the best you can do.
Forget for a moment that you use Entity Framework. What would you do if you had to write SQL statements manually? (EF is a wrapper around a SQL statement generator.) You get posted back an object graph - an Album with a list of Tags. How would you decide now which tags you have to write an INSERT, which tags a DELETE and which tags an UPDATE statement for? (I assume that your relationship between Album and Tag is many-to-many, so you write into a join table.) If you don't know the original state in the database you can't decide. Does the tag relation exist in the database or not? You have to query the database to find the answer, no matter if you use EF or direct SQL.
I see only two alternatives:
Track the entity changes yourself. For you MVC web application it would mean that you have to store the original state with the former GET request somewhere, for example in a session state or in hidden input fields in the page. With the POST request you can retrieve then the original state, build and attach the orginal graph and merge changes into it.
Write a Stored Procedure which takes the album and tag collection and let the SP do the work to create the appropriate SQL statements.
The first way is complicated and has its costs in HTTP payload (hidden input fields) or is depending on a fragile session state. And the second conflicts with why you are using an ORM. Unless you have really serious performance problems or are a SQL master I would not consider a Stored Procedure.
Firstly, I think that this pattern of updates is wrong somehow in that instead of passing in an Album which I assume is a replica or partial replica of the one you want to update (same ID at least), why don't you load the actual one first and apply your changes to it?
If you cannot do that, it might be less confusing to not pass in the same entity (Album) but instead use a data transfer object (DTO) or other message with just the fields you need and then apply that to the loaded Album.
As to the main problem of how to avoid loading each tag, EF should do that for you, but I don't know that it does. For example, NHibernate will not load a lazy entity if you are only setting a relationship because you have not touched any properties of Tag, so it only needs the Id to use it. Hopefully, EF does the same but maybe not (I'm assuming you've profiled it).
If EF does not behave like that you could try two things: firstly, so long as there is no cascade update on Tag, use a skeleton one with just the ID (that is, create the object yourself and just set the Id); this won't work if EF cascade updates the Tag. Secondly, you could implement your own cache for Tags and get them from memory.

MVC3 - Error setting up Controller with Entity framework

The steps I go through...
Add new ADO.NET Entity Data Model > Generate from DB > Setup new connection string to adventureworks db > Next > Select table "DatabaseLog" > Finish. Verify DatabaseLog is visible in the edmx view.
Right click controller > Add controller
TemplateController with read/write actions and views, using Entity
Model class
AdventureWorksDWEntities
Context
New data Context > Accept default name
View
Razor
Click Add.
Produce Error:
"Unable to retrieve metadata for 'DatabaseDocumentor.models.AdventureWorksDWEntities'.
System.Data.Edm.EdmEntityeType: EntityType 'AdventureWorksDWEntities' has no key defined. Define the key for this entitytype.
System.Data.Edm.EdmEntitySet: EntityType: EntitySet 'AdventureWorksDWEntities' is based on type 'AdventureWorksDWEntities' that has no keys defined.
I tried again using AdventureWorks (not AdventureWorksDW) and this time it worked. But, I still don't understand what to pick when generating a controller. I have 3 options:
Template
Here I picked Controller with read/write actions and views, using Entity. This is easy enough to understand. I want to have my tables generated for me so I pick this option.
Model
This is what I want to model. In this case I want to model the "Department" table. So I choose Department.
Context
This one is real fuzzy to me. I chose *Name*Entities. This is the value in the web.config connection strings area. Why do I need to choose my connection string as the context? I only know context as "an object that I use to get to my entities" in C#. So, here it's hard for me to visualize. Do I need to always choose my connection string for the context?
This issue can occur when the Context is not correctly chosen from the dropdown. The context should be the value stored in the web.config
<add name="NamedEntitiesCs1"
that also contains the Model you want to generate.
I found what the issue is...
I have a 3 tiered architiecture I'm using with each of the below projects in one solution.
1.YeagerTech
2.YeagerTechWcfService
3.YeagerTechModel
No matter what, even though my wcf service references my model, the startup project (1) is not "smart" enough to recognize the metadata to create the Controller. In this case, you must include a reference to your project that includes your edmx model.
You must also ensure that the connectionstring also resides in the startup project (1) via the web.config file, in order to get the connection for the metadata.
I found the answer, the model class should have a key, that is an ID property i.e
public int ID { get; set;}
save the changes and the build or rebuild the solution.
That should be able to work out.
your property in your Model for the ID must be declared as public. rebuild and try again, it should work

Resources