C# by nature is a very strongly typed language which signifies that most of the verifications are executed statically during compile time. If not all, but most of features and structs exposed by C# language adheres to this principle. With the introduction of Dynamic in C#4.0, I found there is confusion in developer’s community around the difference between Var, Object and Dynamic. In this blog, I will try to highlight the difference from CLR and compiler’s perspective.
All the objects (value and referenced types) in managed environment are directly or indirectly inherit from System.Object. With this inheritance in place, a variable of System.Object can hold the reference of instance of any type.
In above example, o1 is holding the reference of XmlDocument and o2 is holding the reference of boxed double value. However in either case, compiler will allow to call only those methods /properties which exists in System.Object type. In other words compiler will not allow you to call ChildNodes property from o1 and similarly o2 cannot be used for any arithmetic operation. All of these restrictions are in place because from compiler’s perspective, o1 and o2 are holding the reference of System.Object type. The only way to access the ChildNodes property is to typecast it into XmlDocument type. In that case, compiler will emit the code to validate the type casting during runtime.
a)Though variable of System.Object type can hold the reference of any type, compiler will ensure the type safety. If type casting is happening, compiler will emit the code to validate the type safety during runtime.
b)Assigning the value type to a variable of System.Object will suffer from Boxing / Unboxing.
Var keyword was introduced in C# 3.0. It was primarily introduced to support the notion of anonymous types (which was the key requirement for LINQ). But important part is, var is resolved at the compile time, i.e. the actual type is resolved by the compiler during the compile time. Hence it provides same type safety as using normal typed variable.
In above example, var1 is resolved by compiler as double, var2 is resolved by compiler as int and var3 is resolved by compiler as string. From CLR perspective, there is nothing known as var.
a) Var is a compile time feature resolved to a valid underlying type. Due to this, compiler enforces you to assign a value during the declaration so it can infer the actual type.
b) Since var is resolved to a valid type during compilation, compiler will enforce all type safety and code will not suffer from boxing / unboxing.
In my own opinion, we should use “var” ONLY when actual type is not known during development i.e. type is anonymous as shown in following example.
If the actual type is known, then I will not recommend using var to save few keyboard strokes as it reduces the readability of code.
With C# 4.0, Microsoft introduces “dynamic”. When you define a variable of type dynamic, compiler internally performs two steps. The first thing compiler does is, it converts the type to System.Object and then it postpones all the type validation to the runtime. The following example uses the dynamic to declare an int type.
In above example, Double declares the multiplier as dynamic. During compilation, dynamic will be converted to System.Object and compiler will emit the code for type safety during runtime as shown in the following IL.
a) During compilation time, dynamic is converted to System.Object and compiler will emit the code for type safety during runtime.
b) As dynamic is treated as System.Object, it suffers from boxing / unboxing similar to System.Object.
c) Since compiler emits the code for all the type safety, application’s performance will suffer.
Dynamic was introduced only to simplify the access to the COM APIs. I strongly recommend that we should restrict the use of dynamic only when handling the COM APIs as it might cause runtime exceptions and can cause performance issues.