Anonymous types

The Basics

Anonymous types are those types which are not declared before they are used. Say while you are doing your program, you want a temporary storage of your data, what you need to do, you need either declare a concrete class for the storage of the same, or you can use any collection like ArrayList, HashTable etc to store the key value collection. C# 3.5 and onwards allows you to dynamically create a type and use it directly in your program. You can use ‘var’ or implicit type declaration to ensure you could use the properties just like normal types.

var myruntimeObject = new { FirstProperty = 10, SecondProperty = new DateTime(), ThirdProperty = "string type" };

Console.WriteLine("Type of myruntimeObject is {0}", myruntimeObject.GetType().Name);

Console.WriteLine("Type of FirstProperty is {0}",       myruntimeObject.FirstProperty.GetType().Name);

Console.WriteLine("Type of SecondProperty is {0}", myruntimeObject.SecondProperty.GetType().Name);

Console.WriteLine("Type of ThirdProperty is {0}",        myruntimeObject.ThirdProperty.GetType().Name);

Console.Read();

In the above example I have created one anonymous type using var. The var will give you a chance to get intellesense in your code. You can see the variable holds an object of an anonymous type generated runtime. The power of var in code can also be tested. Just hover over the var keyword and you will see that the type is defined as anonymous.

In the previous code after the declaration of anonymous object, I have tried to see what exactly the type of the object and the members look like that I have declared. Lets see the output below :

So the actual type created by CLR at runtime looks like <>f_AnnonymousType0`3. We will discuss about it later in this post, but the thing that you might wonder, is the compiler is smart enough to declare the type of the members exactly. The FirstProperty is declared as Int32, second as DateTime and third as String. The type is self evaluated in the code and declared accordingly.

Now if I change the value just a bit like :

var myruntimeObject = new

{

   FirstProperty = 10d,

   SecondProperty = new DateTime(),

   ThirdProperty = "string type"

};

The code says the FirstProperty is double. The 10d is evaluated as double value.

You are also free to declare anonymous type for its members.

var myruntimeObject = new { FirstProperty = 10, SecondProperty = new DateTime(), ThirdProperty = "string type", MoreProperty = new { N1 = "" } };

Here the MoreProperty declares another anonymous type implicitly. But there is restriction the usage of anonymous types. You cannot declare a method inside any object declaration, even there is no way to give accessibility specifier to its members. All the members inside an anonymous type are readonly properties, hence if you try to do :

myruntimeObject.FirstProperty = 20;

It will prompt you an error.

Another important thing about anonymous type is that the compiler is smart enough to use anonymous type gracefully. It does not create a new anonymous type if the next instruction is exactly tries to create the same class. Hence if you use :

var myruntimeObject = new { FirstProperty = 10, SecondProperty = new DateTime(), ThirdProperty = "string type", MoreProperty = new { N1 = "" } };

var myruntimeObject2 = new { FirstProperty = 30, SecondProperty = new DateTime(), ThirdProperty = "string type2", MoreProperty = new { N1 = "blah blah" } };

Both internally represent the same type.

The Internals

Now as you know the basics of anonymous Types, lets look deep into its internal structure. In terms of IL, both var and anonymous type is nothing. Var is actually represented as actual type while anonymous type is mapped to an actual type.

The object creation looks similar to

<>f__AnonymousType0<<N1>j__TPar> moreproperty = new <>f__AnonymousType0<<N1>j__TPar>("");

<>f__AnonymousType1<<FirstProperty>j__TPar, <SecondProperty>j__TPar, <ThirdProperty>j__TPar, <MoreProperty>j__TPar>

                myruntimeObject = new

                <>f__AnonymousType1<<FirstProperty>j__TPar, <SecondProperty>j__TPar, <ThirdProperty>j__TPar, <MoreProperty>j__TPar>(10, new DateTime(), "string type", moreproperty);

Hence it seems to be quite big name for the type and also includes a generic argument named <N1>j__TPar. This lets the object to refer the same type when argument differs. As Reflection API is not good enough to show the actual generic types, it shows a numeric representation for the same.

Now as you can see, the Compiler generated type for MoreProperty looks like one above. The first thing that you should notice is the DebuggerDisplayAttribute just above class. With this attribute, Visual Studio debugger shows only <anonymous type> rather than the whole type name.

The Type takes a Generic type argument <N1>j__TPar in its constructor and builds the class. You should notice the parameter is marked as Readonly, so it is not writable. Also the type overrides ToString, GetHashCode and Equals by using the generic EqualityComparer. This looks a nice implementation.

Conclusion

Finally, to conclude, I must say anonymous types is an interesting feature which is widely used and eventually changed the way of coding. It is very much useful while working with LINQ. We will discuss internals of LINQ in later part of the series. But there should be a provision to define methods inside an anonymous type, which I think will be added sooner than later.

------------------------------------------------------------------------------------------------------------------------------

LINQ means Language Integrated Query is one of the major step forward to .NET framework to support queries to work on objects. LINQ allows you to write custom query statements on .NET objects or more specifically any IEnumerables to filter or fetch data from it.



Simple Projection

To start with Internals, let me create a List of strings on a Type. You should already know, that you can easily replace .NET initializer to allow initialize array into any IEnumerable.

public static List<string> myList = new List<string>

{

"Abhishek",

"Amit",

"Abhijit",

"Kunal",

"Krishna",

"Dhananjay"

};

Yes it will compile, and by your previous knowledge, you think it is initialized the same way if you use an array. But it isnt.

C# compiler automatically replaces the code with proper List.Add statements to add each individual statements. If you see the same code in reflector, you will see a new backup list variable is added to eventually add each strings internally.

var simpleprojection = from i in Program.myList

                                   select i;

Now starting what we call the most basic LINQ query represents to project each individual elements in the list directly into another IEnumerable. Now how does the IL looks like :

Now here you wonder how both the sections as I have discussed looks like. Yes, the initializer introduces a Static constructor internally (created by compiler itself) which adds up each of the strings into the List.  On the other hand, simpleprojection gets an Enumerable from Enumerable.Select which takes an IEnumerable and a delegate.

Internally the Enumerable puts each element from an enumerator and process it using the delegate we pass. Just as our delegate actually returns i. So nothing will be changed. Now let me change the code slightly.

var simpleprojection = from i in Program.myList

                       select new { x = i.StartsWith("A"), y = i.Reverse() };

Hmm, the code looks similar, but here we have yield a new object. Now if you see the IL, it produces a concrete implementation of the class to hold X and Y and eventually creates object of each.

Click to see the Image

Well, yes,  as you can see, the IEnumerable returns an object of Concrete Type. Know more about anonymous types from "Internals of Anonymous Type".

Now coming to our scenarios, the code is same as the other one, just replacing the select statement inside the delegate.

Lets add up a bit more with Let :

var simpleprojection = from i in Program.myList

                        let isStartWithA = i.StartsWith("A")

                        let reverseString = i.Reverse()

                        select new { x = isStartWithA , y = reverseString };

Well, the code works the same with two let statements which assigns the value of each on separate variable isStartWithA and reverseString. But internally everything is now changed.

Hmm, the statement is now replaced with three Enumeration.Select calls nested one within another. Lets break these lines manually to understand what is happening.

var  firstEnumerable = Enumerable.Select(myList, delegate (string i) {

                                                    return new { i = i, isStartWithA = i.StartsWith("A") };

          });

          //Replacing <>f__AnonymousType0<string, bool> fType<string, bool>

          var secondEnumerable = Enumerable.Select(

                                            firstEnumerable,

                                            delegate (fType<string, bool> obj) {

                                                return new { Fobj = obj,

                                                                reverseString = obj.i.Reverse<char>() };

                                            });

        //Replacing <>f__AnonymousType1<<>f__AnonymousType0<string, bool> to fType2<fType<string,bool>, IEnumerable<char>>

         var simpleprojection = Enumerable.Select(

                                   secondEnumerable,

                                    delegate (fType2<fType<string,bool>, IEnumerable<char>> obj) {

                                    return new { x = obj.Fobj.isStartWithA,

                                                y = obj.reverseString };

                                });

I have intentionally replaced few compiler generated typenames to some other to make you understand code better.

In the first call we pass the list to a delegate which creates a object with variable isStartWith which is the boolean representation of our first let statement.

In the second call, we pass the result of first call. The second delegate generates the reverseString wrapping around the first call object.

Finally, in the outermost call, it will generate the end product.

Now why should you need three nested calls ? Yes there is a reason. Actually when you define the first let statement, you would internally think that from the next statement we should make the variable x available. So the scope of the First let, isStartWith lies on both of the next two lines reverseString and select. Hence we pass on the value in a anonymous type. Thus we can easily use isStartWith on both of these lines.

Similarly on the next successive calls we need to wrap around all the let statements internally into its respective anonymous types until finally we reach out the select. The compiler is smart enough to go on create anonymous types with first property being the parent object and second being the actual let statement.

Hence you can infer, for every let statement in LINQ query, compiler will generate 1 anonymous type at least.

Now lets alter our query to introduce Order By.

var simpleprojection = from i in Program.myList

                        orderby i

                        select i;

If we look into the Reflector for the code we see the code writes the same thing as the first code, the only difference is it is using Enumerable.OrderBy instead of Select and returns back IOrderedEnumerable. The OrderBy actually processes each element using the delegate that we pass, and puts it into a call to OrderedEnumerable, which internally uses EnumerableSorter to get each object from the List.

If you ask me does it internally stores the elements in sorted order after it is being processed, then I think you need to read more about IEnumerable may be from C# iterators. Actually when you enumerate from IOrderedEnumerable, it will internally get the next element from the list which represents the ordering. As IEnumerable internally maintains a state machine, it will save its state until further yield.

Next, lets put a GroupBy.

var simpleGroup = from i in Program.myList

                        group i by i.Count() into g

                        select new { Count = g.Count(), Elements = g };

Well, here we grouped the list based on character count. You should note each Group statement gives you an object (g) which represents the group and you can use g in your select statement.

Now lets see it into reflector:

So you can see here in to code, the C# compiler wraps around a Enumerable.GroupBy inside the Enumerable.Select where the former returns an IEnumerable of IGrouping members. The select statement then uses this IGrouping object to generate its result.

Similar to OrderBy the GroupBy uses GroupedEnumerable object to generate each Group object on each yield. You should note, the object g is internally managed IGrouping object, hence groups does not generate additional Types.

Finally lets see how Join statement looks like. For simplicity I used inner join statement.

var customers = new List<Customer>{ new Customer { Id = 1, Name="Abhishek"},

                                new Customer { Id = 2, Name = "Amit"},

                                new Customer { Id = 3, Name = "Abhijit"},

                                new Customer { Id= 4, Name="Kunal"},

                                new Customer { Id = 5, Name= "Krishna"},

                                new Customer { Id = 6, Name = "Dhananjay"}

};

var orders = new List<Orders>

{

    new Orders { Id = 1, CustomerId = 4, NosOrder=20},

    new Orders { Id = 2, CustomerId = 1, NosOrder=30},

    new Orders { Id = 3, CustomerId = 2, NosOrder=50},

    new Orders { Id = 4, CustomerId = 1, NosOrder=10},

    new Orders { Id = 5, CustomerId = 3, NosOrder=60},

    new Orders { Id = 6, CustomerId = 4, NosOrder=10},

};

var simpleJoin = from c in customers

                join o in orders on c.Id equals o.CustomerId

                select new { c.Name, o.NosOrder };

Well, here I have put one of the most simplest customer and order Join which puts only the equal elements on the result. If you see this on reflector

Hmm, it eventually calls upon the Enumerable.Join statement with delegates which filters the first argument, the  second argument and the result. In case of Joins, for each yield, the first and second delegate is fetched once, and 3rd delegate is called upon for each successful match in the lists. The framework uses JoinIterator internally to iterate between two lists.

For OuterJoins :

var simpleJoin = from c in customers

                join o in orders on c.Id equals o.CustomerId into outer

                from l in outer.DefaultIfEmpty()

                select new { c.Name, l.NosOrder };

It introduces Enumerable.GroupJoin nested inside Enumerable.SelectMany.

The GroupJoin uses GroupJoinIterator to group each individual first delegate with second delegate and each group is selected.


by abhishek sur