LINQ to Entities对带有嵌套对象的Union的null引用

Question

Given the following class structure:

public class User  // DB model
{
    public Guid Id { get; set; }
    public Address Address { get; set; }
    // And other propeties
}

public class Invitation  // DB model
{
    public Guid Id { get; set; }
    // And other propeties
}

public class Address  // DB model
{
    public string Zip { get; set; }
    // And other properties
}

public class ResponseModel
{
    public Guid Id { get; set; }
    public ResponseAddress Address { get; set; }
}

public class ResponseAddress
{
    public string Zip { get; set; }
    // And other properties
}

And the following queries which return Users and Invitations, respectively, with the intent of getting a union of the two queries:

var users = db.Users.Select(x => new ResponseModel() 
{
    Id = x.Id,
    Address = new ResponseAddress()
    {
        Zip = x.Address.Zip
    }
});
var invitations = db.Invitations.Select(x => new ResponseModel() 
{
    Id = x.Id,
    Address = new ResponseAddress()
    {
        Zip = String.Empty
    }
});
var union = users.Union(invitations).ToList();

When I attempt to do the union DB-side, I get a null reference exception deep down in System.Data.Entity.CoreQuery.PlanCompiler. If I call ToList() on each part individually, it works; and if I call ToList() on each part and then union those, it works.

users.ToList();
invitations.ToList();
users.ToList().Union(invitations.ToList());

It also appears that if I union them BEFORE creating the ResponseAddress part, then create the ResponseAddress part in a later call to Select, it works:

var users = db.Users.Select(x => new  
{
    Id = x.Id,
    Zip = x.Address.Zip
});
var invitations = db.Invitations.Select(x => new  
{
    Id = x.Id,
    Zip = String.Empty
});
var union = users.Union(invitations).Select(x=>new ResponseModel() {
    Id = x.Id,
    Address = new ResponseAddress() {
        Zip = x.Zip
    }
}).ToList();

Any thoughts as to why the call to Union in the first set of queries would return a null reference exception, while the call in the last query does not? Both are executed DB-side, and both should produce similar queries (theoretically virtually identical, save for the way LINQ does query nesting.)

Answer 1

In the ResponseAddress part, you create a new instance of ResponseAddress class. This has no meaning for the relational database. When you turn your query results into list, then it's runtime's job to handle the union and it knows about objects whereas database server has no knowledge of ResponseAddress because that's not how it represents data.

Answer 2

DbSet and IDbSet implements IQueryable meaning a query is executed against the database only when:

1. It is enumerated by a foreach (C#) or For Each (Visual Basic) statement.
2. It is enumerated by a collection operation such as ToArray, ToDictionary or ToList.
3. LINQ operators such as First or Any are specified in the outermost part of the query.

4. The following methods are called: the Load extension method on a DbSet, DbEntityEntry.Reload and Database.ExecuteSqlCommand

    var users = db.Users.Select(x => new   \n{ \nId = x.Id, \nZip = x.Address.Zip \n});  

As it is IQueryable above query has not yet executed against the database and users is an empty collection rather than null

var users = db.Users.Select(x => new  
{
    Id = x.Id,
    Zip = x.Address.Zip
}).ToList();

Now calling the ToList() has executed the query to the db.