Overview of C# 3.0

 

March 2007

Anders Hejlsberg, Mads Torgersen

Applies to:
   Visual C# 3.0

Summary: Technical overview of C# 3.0 ("C# Orcas"), which introduces several language extensions that build on C# 2.0 to support the creation and use of higher order, functional style class libraries. (38 printed pages)

Contents

Introduction
26.1 Implicitly Typed Local Variables
26.2 Extension Methods
   26.2.1 Declaring Extension Methods
   26.2.2 Available Extension Methods
   26.2.3 Extension Method Invocations
26.3 Lambda Expressions
   26.3.1 Anonymous Method and Lambda Expression Conversions
   26.3.2 Delegate Creation Expressions
   26.3.3 Type Inference
      26.3.3.1 The first phase
      26.3.3.2 The second phase
      26.3.3.3 Input types
      26.3.3.4 Output types
      26.3.3.5 Dependence
      26.3.3.6 Output type inferences
      26.3.3.7 Explicit argument type inferences
      26.3.3.8 Exact inferences
      26.3.3.9 Lower-bound inferences
      26.3.3.10 Fixing
      26.3.3.11 Inferred return type
      26.3.3.12 Type inference for conversion of method groups
      26.3.3.13 Finding the best common type of a set of expressions
   26.3.4 Overload Resolution
26.4 Object and Collection Initializers
   26.4.1 Object Initializers
   26.4.2 Collection Initializers
26.5 Anonymous Types
26.6 Implicitly Typed Arrays
26.7 Query Expressions
   26.7.1 Query Expression Translation
      26.7.1.1 Select and groupby clauses with continuations
      26.7.1.2 Explicit range variable types
      26.7.1.3 Degenerate query expressions
      26.7.1.4 From, let, where, join and orderby clauses
      26.7.1.5 Select clauses
      26.7.1.6 Groupby clauses
      26.7.1.7 Transparent identifiers
   26.7.2 The Query Expression Pattern
26.8 Expression Trees
   26.8.1 Overload Resolution
26.9 Automatically Implemented Properties

Introduction

This article contains some updates that apply to Visual C# 3.0. A comprehensive specification will accompany the release of the language.

C# 3.0 ("C# Orcas") introduces several language extensions that build on C# 2.0 to support the creation and use of higher order, functional style class libraries. The extensions enable construction of compositional APIs that have equal expressive power of query languages in domains such as relational databases and XML. The extensions include:

  • Implicitly typed local variables, which permit the type of local variables to be inferred from the expressions used to initialize them.
  • Extension methods, which make it possible to extend existing types and constructed types with additional methods.
  • Lambda expressions, an evolution of anonymous methods that provides improved type inference and conversions to both delegate types and expression trees.
  • Object initializers, which ease construction and initialization of objects.
  • Anonymous types, which are tuple types automatically inferred and created from object initializers.
  • Implicitly typed arrays, a form of array creation and initialization that infers the element type of the array from an array initializer.
  • Query expressions, which provide a language integrated syntax for queries that is similar to relational and hierarchical query languages such as SQL and XQuery.
  • Expression trees, which permit lambda expressions to be represented as data (expression trees) instead of as code (delegates).

This document is a technical overview of those features. The document makes reference to the C# Language Specification Version 1.2 (§1 through §18) and the C# Language Specification Version 2.0 (§19 through §25), both of which are available on the C# Language Home Page (https://msdn.microsoft.com/vcsharp/aa336809.aspx).

26.1 Implicitly Typed Local Variables

In an implicitly typed local variable declaration, the type of the local variable being declared is inferred from the expression used to initialize the variable. When a local variable declaration specifies var as the type and no type named var is in scope, the declaration is an implicitly typed local variable declaration. For example:

var i = 5;
var s = "Hello";
var d = 1.0;
var numbers = new int[] {1, 2, 3};
var orders = new Dictionary<int,Order>();

The implicitly typed local variable declarations above are precisely equivalent to the following explicitly typed declarations:

int i = 5;
string s = "Hello";
double d = 1.0;
int[] numbers = new int[] {1, 2, 3};
Dictionary<int,Order> orders = new Dictionary<int,Order>();

A local variable declarator in an implicitly typed local variable declaration is subject to the following restrictions:

  • The declarator must include an initializer.
  • The initializer must be an expression.
  • The initializer expression must have a compile-time type which cannot be the null type.
  • The local variable declaration cannot include multiple declarators.
  • The initializer cannot refer to the declared variable itself

The following are examples of incorrect implicitly typed local variable declarations:

var x;               // Error, no initializer to infer type from
var y = {1, 2, 3};   // Error, collection initializer not permitted
var z = null;        // Error, null type not permitted
var u = x => x + 1;  // Error, lambda expressions do not have a type
var v = v++;         // Error, initializer cannot refer to variable itself

For reasons of backward compatibility, when a local variable declaration specifies var as the type and a type named var is in scope, the declaration refers to that type. Since a type named var violates the established convention of starting type names with an upper case letter, this situation is unlikely to occur.

The for-initializer of a for statement (§8.8.3) and the resource-acquisition of a using statement (§8.13) can be an implicitly typed local variable declaration. Likewise, the iteration variable of a foreach statement (§8.8.4) may be declared as an implicitly typed local variable, in which case the type of the iteration variable is inferred to be the element type of the collection being enumerated. In the example

int[] numbers = { 1, 3, 5, 7, 9 };
foreach (var n in numbers) Console.WriteLine(n);

the type of n is inferred to be int, the element type of numbers.

Only local-variable-declaration, for-initializer, resource-acquisition and foreach-statement can contain implicitly typed local variable declarations.

26.2 Extension Methods

Extension methods are static methods that can be invoked using instance method syntax. In effect, extension methods make it possible to extend existing types and constructed types with additional methods.

Note   Extension methods are less discoverable and more limited in functionality than instance methods. For those reasons, it is recommended that extension methods be used sparingly and only in situations where instance methods are not feasible or possible. Extension members of other kinds, such as properties, events, and operators, are being considered but are currently not supported.

26.2.1 Declaring Extension Methods

Extension methods are declared by specifying the keyword this as a modifier on the first parameter of the methods. Extension methods can only be declared in non-generic, non-nested static classes. The following is an example of a static class that declares two extension methods:

namespace Acme.Utilities
{
   public static class Extensions
   {
      public static int ToInt32(this string s) {
         return Int32.Parse(s);
      }
      public static T[] Slice<T>(this T[] source, int index, int count) {
         if (index < 0 || count < 0 || source.Length – index < count)
            throw new ArgumentException();
         T[] result = new T[count];
         Array.Copy(source, index, result, 0, count);
         return result;
      }
   }
}

The first parameter of an extension method can have no modifiers other than this, and the parameter type cannot be a pointer type.

Extension methods have all the capabilities of regular static methods. In addition, once imported, extension methods can be invoked using instance method syntax.

26.2.2 Available Extension Methods

Extension methods are available in a namespace if declared in a static class or imported through using-namespace-directives (§9.3.2) in that namespace. In addition to importing the types contained in an imported namespace, a using-namespace-directive thus imports all extension methods in all static classes in the imported namespace.

In effect, available extension methods appear as additional methods on the types that are given by their first parameter and have lower precedence than regular instance methods. For example, when the Acme.Utilities namespace from the example above is imported with the using-namespace-directive

using Acme.Utilities;

it becomes possible to invoke the extension methods in the static class Extensions using instance method syntax:

string s = "1234";
int i = s.ToInt32();               // Same as Extensions.ToInt32(s)
int[] digits = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
int[] a = digits.Slice(4, 3);      // Same as Extensions.Slice(digits, 4, 3)

26.2.3 Extension Method Invocations

The detailed rules for extension method invocation are described in the following. In a method invocation (§7.5.5.1) of one of the forms

expr . identifier ( )
expr . identifier ( args )
expr . identifier < typeargs > ( )
expr . identifier < typeargs > ( args )

if the normal processing of the invocation finds no applicable instance methods (specifically, if the set of candidate methods for the invocation is empty), an attempt is made to process the construct as an extension method invocation. The method invocation is first rewritten to one of the following, respectively:

identifier ( expr )
identifier ( expr , args )
identifier < typeargs > ( expr )
identifier < typeargs > ( expr , args )

The rewritten form is then processed as a static method invocation, except for the way in which identifier is resolved: Starting with the closest enclosing namespace declaration, continuing with each enclosing namespace declaration, and ending with the containing compilation unit, successive attempts are made to process the rewritten method invocation with a method group consisting of all available and accessible extension methods in the namespace with the name given by identifier. From this set remove all the methods that are not applicable (§7.4.2.1) and the ones where no implicit identity, reference or boxing conversion exists from the first argument to the first parameter. The first method group that yields a non-empty such set of candidate methods is the one chosen for the rewritten method invocation, and normal overload resolution (§7.4.2) is applied to select the best extension method from the set of candidates. If all attempts yield empty sets of candidate methods, a compile-time error occurs.

The preceding rules mean that instance methods take precedence over extension methods, and extension methods available in inner namespace declarations take precedence over extension methods available in outer namespace declarations. For example:

public static class E
{
   public static void F(this object obj, int i) { }
   public static void F(this object obj, string s) { }
}
class A { }
class B
{
   public void F(int i) { }
}
class C
{
   public void F(object obj) { }
}
class X
{
   static void Test(A a, B b, C c) {
      a.F(1);            // E.F(object, int)
      a.F("hello");      // E.F(object, string)
      b.F(1);            // B.F(int)
      b.F("hello");      // E.F(object, string)
      c.F(1);            // C.F(object)
      c.F("hello");      // C.F(object)
   }
}

In the example, the B method takes precedence over the first extension method, and the C method takes precedence over both extension methods.

26.3 Lambda Expressions

C# 2.0 introduces anonymous methods, which allow code blocks to be written "in-line" where delegate values are expected. While anonymous methods provide much of the expressive power of functional programming languages, the anonymous method syntax is rather verbose and imperative in nature. Lambda expressions provide a more concise, functional syntax for writing anonymous methods.

A lambda expression is written as a parameter list, followed by the => token, followed by an expression or a statement block.

expression:
assignment
non-assignment-expression
non-assignment-expression:
conditional-expression
lambda-expression
query-expression
lambda-expression:
(   lambda-parameter-listopt)=>   lambda-expression-body
implicitly-typed-lambda-parameter   =>   lambda-expression-body
lambda-parameter-list:
explicitly-typed-lambda-parameter-list
implicitly-typed-lambda-parameter-list
explicitly-typed-lambda-parameter-list
explicitly-typed-lambda-parameter
explicitly-typed-lambda-parameter-list   ,   explicitly-typed-lambda-parameter
explicitly-typed-lambda-parameter:
parameter-modifieropt   type   identifier
implicitly-typed-lambda-parameter-list
implicitly-typed-lambda-parameter
implicitly-typed-lambda-parameter-list   ,   implicitly-typed-lambda-parameter
implicitly-typed-lambda-parameter:
identifier
lambda-expression-body:
expression
block

The => operator has the same precedence as assignment (=) and is right-associative.

The parameters of a lambda expression can be explicitly or implicitly typed. In an explicitly typed parameter list, the type of each parameter is explicitly stated. In an implicitly typed parameter list, the types of the parameters are inferred from the context in which the lambda expression occurs—specifically, when the lambda expression is converted to a compatible delegate type, that delegate type provides the parameter types (§26.3.1).

In a lambda expression with a single, implicitly typed parameter, the parentheses may be omitted from the parameter list. In other words, a lambda expression of the form

( param ) => expr

can be abbreviated to

param => expr

Some examples of lambda expressions follow below:

x => x + 1                     // Implicitly typed, expression body
x => { return x + 1; }         // Implicitly typed, statement body
(int x) => x + 1               // Explicitly typed, expression body
(int x) => { return x + 1; }   // Explicitly typed, statement body
(x, y) => x * y               // Multiple parameters
() => Console.WriteLine()      // No parameters

In general, the specification of anonymous methods, provided in §21 of the C# 2.0 Specification, also applies to lambda expressions. Lambda expressions are functionally similar to anonymous methods, except for the following points:

  • Anonymous methods permit the parameter list to be omitted entirely, yielding convertibility to delegate types of any list of parameters.
  • Lambda expressions permit parameter types to be omitted and inferred whereas anonymous methods require parameter types to be explicitly stated.
  • The body of a lambda expression can be an expression or a statement block whereas the body of an anonymous method can only be a statement block.
  • Lambda expressions with an expression body can be converted to expression trees (§26.8).

26.3.1 Anonymous Method and Lambda Expression Conversions

Note   This section replaces §21.3.

An anonymous-method-expression and a lambda-expression is classified as a value with special conversion rules. The value does not have a type but can be implicitly converted to a compatible delegate type. Specifically, a delegate type D is compatible with an anonymous method or lambda-expression L provided:

  • D and L have the same number of parameters.
  • If L is an anonymous method that does not contain an anonymous-method-signature, then D may have zero or more parameters of any type, as long as no parameter of D has the out parameter modifier.
  • If L has an explicitly typed parameter list, each parameter in D has the same type and modifiers as the corresponding parameter in L.
  • If L is a lambda expression that has an implicitly typed parameter list, D has no ref or out parameters.
  • If D has a void return type and the body of L is an expression, when each parameter of L is given the type of the corresponding parameter in D, the body of L is a valid expression (wrt §7) that would be permitted as a statement-expression (§8.6).
  • If D has a void return type and the body of L is a statement block, when each parameter of L is given the type of the corresponding parameter in D, the body of L is a valid statement block (wrt §8.2) in which no return statement specifies an expression.
  • If D has a non-void return type and the body of L is an expression, when each parameter of L is given the type of the corresponding parameter in D, the body of L is a valid expression (wrt §7) that is implicitly convertible to the return type of D.
  • If D has a non-void return type and the body of L is a statement block, when each parameter of L is given the type of the corresponding parameter in D, the body of L is a valid statement block (wrt §8.2) with a non-reachable end point in which each return statement specifies an expression that is implicitly convertible to the return type of D.

The examples that follow use a generic delegate type Func<A,R> which represents a function taking an argument of type A and returning a value of type R:

delegate R Func<A,R>(A arg);

In the assignments

Func<int,int> f1 = x => x + 1;         // Ok
Func<int,double> f2 = x => x + 1;      // Ok
Func<double,int> f3 = x => x + 1;      // Error

the parameter and return types of each lambda expression are determined from the type of the variable to which the lambda expression is assigned. The first assignment successfully converts the lambda expression to the delegate type Func<int,int> because, when x is given type int, x + 1 is a valid expression that is implicitly convertible to type int. Likewise, the second assignment successfully converts the lambda expression to the delegate type Func<int,double> because the result of x + 1 (of type int) is implicitly convertible to type double. However, the third assignment is a compile-time error because, when x is given type double, the result of x + 1 (of type double) is not implicitly convertible to type int.

26.3.2 Delegate Creation Expressions

Note   This section replaces §21.10.

Delegate creation expressions (§7.5.10.3) are extended to permit the argument to be an expression classified as a method group, an expression classified as an anonymous method or lambda expression, or a value of a delegate type.

The compile-time processing of a delegate-creation-expression of the form new D(E), where D is a delegate-type and E is an expression, consists of the following steps:

  • If E is a method group, a method group conversion (§21.9) must exist from E to D, and the delegate creation expression is processed in the same way as that conversion.
  • If E is an anonymous method or lambda expression, an anonymous method or lambda expression conversion (§ 26.3.1) must exist from E to D, and the delegate creation expression is processed in the same way as that conversion.
  • If E is a value of a delegate type, the method signature of E must be consistent (§21.9) with D, and the result is a reference to a newly created delegate of type D that refers to the same invocation list as E. If E is not consistent with D, a compile-time error occurs.

26.3.3 Type Inference

Note   This section replaces §20.6.4.

When a generic method is called without specifying type arguments, a type inference process attempts to infer type arguments for the call. The presence of type inference allows a more convenient syntax to be used for calling a generic method, and allows the programmer to avoid specifying redundant type information. For example, given the method declaration:

class Chooser
{
   static Random rand = new Random();
   public static T Choose<T>(T first, T second) {
      return (rand.Next(2) == 0)? first: second;
   }
}

it is possible to invoke the Choose method without explicitly specifying a type argument:

int i = Chooser.Choose(5, 213);               // Calls Choose<int>
string s = Chooser.Choose("foo", "bar");      // Calls Choose<string>

Through type inference, the type arguments int and string are determined from the arguments to the method.

Type inference occurs as part of the compile-time processing of a method invocation (§20.9.7) and takes place before the overload resolution step of the invocation. When a particular method group is specified in a method invocation, and no type arguments are specified as part of the method invocation, type inference is applied to each generic method in the method group. If type inference succeeds, then the inferred type arguments are used to determine the types of arguments for subsequent overload resolution. If overload resolution chooses a generic method as the one to invoke, then the inferred type arguments are used as the actual type arguments for the invocation. If type inference for a particular method fails, that method does not participate in overload resolution. The failure of type inference, in and of itself, does not cause a compile-time error. However, it often leads to a compile-time error when overload resolution then fails to find any applicable methods.

If the supplied number of arguments is different than the number of parameters in the method, then inference immediately fails. Otherwise, assume that the generic method has the following signature:

Tr M<X1...Xn>(T1 x1 ... Tm xm)

With a method call of the form M(e1...em) the task of type inference is to find unique type arguments S1...Sn for each of the type parameters X1...Xn so that the call M<S1...Sn>(e1...em) becomes valid.

During the process of inference each type parameter Xi is either fixed to a particular type Si or unfixed with an associated set of bounds. Each of the bounds is some type T. Initially each type variable Xi is unfixed with an empty set of bounds.

Type inference takes place in phases. Each phase will try to infer type arguments for more type variables based on the findings of the previous phase. The first phase makes some initial inferences of bounds, whereas the second phase fixes type variables to specific types and infers further bounds. The second phase may have to be repeated a number of times.

Note   When we refer to delegate types throughout the following, this should be taken to include also types of the form Expression<D> where D is a delegate type. The argument and return types of Expression<D> are those of D.

Note   Type inference takes place not only when a generic method is called. Type inference for conversion of method groups is described in §26.3.3.12 and finding the best common type of a set of expressions is described in §26.3.3.13.

26.3.3.1 The first phase

For each of the method arguments ei:

  • An explicit argument type inference (§26.3.3.7) is made from ei with type Ti if ei is a lambda expression, an anonymous method, or a method group.
  • An output type inference (§26.3.3.6) is made from ei with type Ti if ei is not a lambda expression, an anonymous method, or a method group.

26.3.3.2 The second phase

All unfixed type variables Xi that depend on (§26.3.3.5) no Xj are fixed (§26.3.3.10).

If no such type variables exist, all unfixed type variables Xi are fixed for which all of the following hold:

  • There is at least one type variable **Xj **that depends on Xi.
  • Xi has a non-empty set of bounds.

If no such type variables exist and there are still unfixed type variables, type inference fails. If no further unfixed type variables exist, type inference succeeds. Otherwise, for all arguments ei with corresponding argument type Ti where the output types (§26.3.3.4) contain unfixed type variables Xj but the input types (§26.3.3.3) do not, an output type inference (§26.3.3.6) is made for ei with type Ti. Then the second phase is repeated.

26.3.3.3 Input types

If e is a method group or implicitly typed lambda expression and T is a delegate type then all the argument types of T are input types of e with type T.

26.3.3.4 Output types

If e is a method group, an anonymous method, a statement lambda or an expression lambda and T is a delegate type then the return type of T is an output type of e with type T.

26.3.3.5 Dependence

An unfixed type variable Xi depends directly on an unfixed type variable Xj if for some argument ek with type Tk Xj occurs in an input type of ek with type Tk and Xi occurs in an output type of ek with type Tk.

Xj depends on Xi if Xj depends directly on Xi or if Xi depends directly on Xk and Xk depends on Xj. Thus "depends on" is the transitive but not reflexive closure of "depends directly on".

26.3.3.6 Output type inferences

An output type inference is made from an expression e with type T in the following way:

  • If e is a lambda or anonymous method with inferred return type U (§26.3.3.11) and T is a delegate type with return type Tb, then a lower-bound inference (§26.3.3.9) is made from U for Tb.
  • Otherwise, if e is a method group and T is a delegate type with parameter types T1...Tk and overload resolution of e with the types T1...Tk yields a single method with return type U, then a lower-bound inference is made from U for Tb.
  • Otherwise, if e is an expression with type U, then a lower-bound inference is made from U for T.
  • Otherwise, no inferences are made.

26.3.3.7 Explicit argument type inferences

An explicit argument type inference is made from an expression e with type T in the following way:

  • If e is an explicitly typed lambda expression or anonymous method with argument types U1...Uk and T is a delegate type with parameter types V1...Vk then for each Ui an exact inference (§26.3.3.8) is made from Ui for the corresponding Vi.

26.3.3.8 Exact inferences

An exact inference from a type U for a type V is made as follows:

  • If V is one of the unfixed Xi then U is added to the set of bounds for Xi.
  • Otherwise, if U is an array type Ue[...] and V is an array type Ve[...] of the same rank then an exact inference from Ue to Ve is made.
  • Otherwise, if V is a constructed type C<V1...Vk> and U is a constructed type C<U1...Uk> then an exact inference is made from each Ui to the corresponding Vi.
  • Otherwise, no inferences are made.

26.3.3.9 Lower-bound inferences

A lower-bound inference from a type U for a type V is made as follows:

  • If V is one of the unfixed Xi then U is added to the set of bounds for Xi.
  • Otherwise if U is an array type Ue[...] and V is either an array type Ve[...] of the same rank, or if U is a one-dimensional array type **Ue[]**and V is one of IEnumerable<Ve>, ICollection<Ve> or IList<Ve> then:
    • If Ue is known to be a reference type then a lower-bound inference from Ue to Ve is made.
    • Otherwise, an exact inference from Ue to Ve is made.
  • Otherwise if V is a constructed type C<V1...Vk> and there is a unique set of types U1...Uk such that a standard implicit conversion exists from U to C<U1...Uk> then an exact inference is made from each Ui for the corresponding Vi.
  • Otherwise, no inferences are made.

26.3.3.10 Fixing

An unfixed type variable Xi with a set of bounds is fixed as follows.

  • The set of candidate types Uj starts out as the set of all types in the set of bounds for Xi.
  • We then examine each bound for Xi in turn. For each bound U of X all types Uj to which there is not a standard implicit conversion from U are removed from the candidate set.
  • If among the remaining candidate types Uj there is a unique type V from which there is a standard implicit conversion to all the other candidate types, then Xi is fixed to V.
  • Otherwise, type inference fails.

26.3.3.11 Inferred return type

For purposes of type inference and overload resolution, the inferred return type of a lambda expression or anonymous method e is determined as follows:

  • If the body of e is an expression, the type of that expression is the inferred return type of e.
  • If the body of e is a statement block, if the set of expressions in the block's return statements has a best common type, and if that type is not the null type, then that type is the inferred return type of L.
  • Otherwise, a return type cannot be inferred for L.

As an example of type inference involving lambda expressions, consider the Select extension method declared in the System.Linq.Enumerable class:

namespace System.Linq
{
   public static class Enumerable
   {
      public static IEnumerable<TResult> Select<TSource,TResult>(
         this IEnumerable<TSource> source,
         Func<TSource,TResult> selector)
      {
         foreach (TSource element in source) yield return selector(element);
      }
   }
}

Assuming the System.Linq namespace was imported with a using clause, and given a class Customer with a Name property of type string, the Select method can be used to select the names of a list of customers:

List<Customer> customers = GetCustomerList();
IEnumerable<string> names = customers.Select(c => c.Name);

The extension method invocation (§26.2.3) of Select is processed by rewriting the invocation to a static method invocation:

IEnumerable<string> names = Enumerable.Select(customers, c => c.Name);

Since type arguments were not explicitly specified, type inference is used to infer the type arguments. First, the customers argument is related to the source parameter, inferring T to be Customer. Then, using the lambda expression type inference process described above, c is given type Customer, and the expression c.Name is related to the return type of the selector parameter, inferring S to be string. Thus, the invocation is equivalent to:

Sequence.Select<Customer,string>(customers, (Customer c) => c.Name)

and the result is of type IEnumerable<string>.

The following example demonstrates how lambda expression type inference allows type information to "flow" between arguments in a generic method invocation. Given the method:

static Z F<X,Y,Z>(X value, Func<X,Y> f1, Func<Y,Z> f2) {
   return f2(f1(value));
}

type inference for the invocation

double seconds = F("1:15:30", s => TimeSpan.Parse(s), t => t.TotalSeconds);

proceeds as follows: First, the argument "1:15:30" is related to the value parameter, inferring X to be string. Then, the parameter of the first lambda expression, s, is given the inferred type string, and the expression TimeSpan.Parse(s) is related to the return type of f1, inferring Y to be System.TimeSpan. Finally, the parameter of the second lambda expression, t, is given the inferred type System.TimeSpan, and the expression t.TotalSeconds is related to the return type of f2, inferring Z to be double. Thus, the result of the invocation is of type double.

26.3.3.12 Type inference for conversion of method groups

Similar to calls of generic methods, type inference must also be applied when a method group M containing a generic method is assigned to a given delegate type D. Given a method

Tr M<X1...Xn>(T1 x1 ... Tm xm)

and the method group M being assigned to the delegate type D the task of type inference is to find type arguments S1...Sn so that the expression:

M<S1...Sn>

becomes assignable to D.

Unlike the type inference algorithm for generic method calls, in this case there are only argument types, no argument expressions. In particular, there are no lambda expressions and hence no need for multiple phases of inference.

Instead, all Xi are considered unfixed, and a lower-bound inference is made from each argument type Uj of D to the corresponding parameter type Tj of M. If for any of the Xi no bounds were found, type inference fails. Otherwise, all Xi are fixed to corresponding Si, which are the result of type inference.

26.3.3.13 Finding the best common type of a set of expressions

In some cases, a common type needs to be inferred for a set of expressions. In particular, the element types of implicitly typed arrays and the return types of anonymous methods and statement lambdas are found in this way.

Intuitively, given a set of expressions e1...em this inference should be equivalent to calling a method

Tr M<X>(X x1 ... X xm)

with the ei as arguments.

More precisely, the inference starts out with an unfixed type variable X. Output type inferences are then made from each ei with type X. Finally, X is fixed and the resulting type S is the resulting common type for the expressions.

26.3.4 Overload Resolution

Lambda expressions in an argument list affect overload resolution in certain situations. Please refer to §7.4.2.3 for the exact rules.

The following example illustrates the effect of lambdas on overload resolution.

class ItemList<T>: List<T>
{
   public int Sum(Func<T,int> selector) {
      int sum = 0;
      foreach (T item in this) sum += selector(item);
      return sum;
   }
   public double Sum(Func<T,double> selector) {
      double sum = 0;
      foreach (T item in this) sum += selector(item);
      return sum;
   }
}

The ItemList<T> class has two Sum methods. Each takes a selector argument, which extracts the value to sum over from a list item. The extracted value can be either an int or a double and the resulting sum is likewise either an int or a double.

The Sum methods could for example be used to compute sums from a list of detail lines in an order.

class Detail
{
   public int UnitCount;
   public double UnitPrice;
   ...
}
void ComputeSums() {
   ItemList<Detail> orderDetails = GetOrderDetails(...);
   int totalUnits = orderDetails.Sum(d => d.UnitCount);
   double orderTotal = orderDetails.Sum(d => d.UnitPrice * d.UnitCount);
   ...
}

In the first invocation of orderDetails.Sum, both Sum methods are applicable because the lambda expression d => d.UnitCount is compatible with both Func<Detail,int> and Func<Detail,double>. However, overload resolution picks the first Sum method because the conversion to Func<Detail,int> is better than the conversion to Func<Detail,double>.

In the second invocation of orderDetails.Sum, only the second Sum method is applicable because the lambda expression d => d.UnitPrice * d.UnitCount produces a value of type double. Thus, overload resolution picks the second Sum method for that invocation.

26.4 Object and Collection Initializers

An object creation expression (§7.5.10.1) may include an object or collection initializer which initializes the members of the newly created object or the elements of the newly created collection.

object-creation-expression:
new   type   (   argument-listopt)   object-or-collection-initializeroptnew   type   object-or-collection-initializer
object-or-collection-initializer:
object-initializer
collection-initializer

An object creation expression can omit the constructor argument list and enclosing parentheses provided it includes an object or collection initializer. Omitting the constructor argument list and enclosing parentheses is equivalent to specifying an empty argument list.

Execution of an object creation expression that includes an object or collection initializer consists of first invoking the instance constructor and then performing the member or element initializations specified by the object or collection initializer.

It is not possible for an object or collection initializer to refer to the object instance being initialized.

In order to correctly parse object and collection initializers with generics, the disambiguating list of tokens in §20.6.5 must be augmented with the } token.

26.4.1 Object Initializers

An object initializer specifies values for one or more fields or properties of an object.

object-initializer:
{   member-initializer-listopt}{   member-initializer-list   ,}
member-initializer-list:
member-initializer
member-initializer-list   ,   member-initializer
member-initializer:
identifier   =   initializer-value
initializer-value:
expression
object-or-collection-initializer

An object initializer consists of a sequence of member initializers, enclosed by { and } tokens and separated by commas. Each member initializer must name an accessible field or property of the object being initialized, followed by an equals sign and an expression or an object or collection initializer. It is an error for an object initializer to include more than one member initializer for the same field or property. It is not possible for the object initializer to refer to the newly created object it is initializing.

A member initializer that specifies an expression after the equals sign is processed in the same way as an assignment (§7.13.1) to the field or property.

A member initializer that specifies an object initializer after the equals sign is a nested object initializer, i.e., an initialization of an embedded object. Instead of assigning a new value to the field or property, the assignments in the nested object initializer are treated as assignments to members of the field or property. Nested object initializers cannot be applied to properties with a value type, or to read-only fields with a value type.

A member initializer that specifies a collection initializer after the equals sign is an initialization of an embedded collection. Instead of assigning a new collection to the field or property, the elements given in the initializer are added to the collection referenced by the field or property. The field or property must be of a collection type that satisfies the requirements specified in §26.4.2.

The following class represents a point with two coordinates:

public class Point
{
   int x, y;
   public int X { get { return x; } set { x = value; } }
   public int Y { get { return y; } set { y = value; } }
}

An instance of Point can be created and initialized as follows:

var a = new Point { X = 0, Y = 1 };

which has the same effect as

var __a = new Point();
__a.X = 0;
__a.Y = 1; 
var a = __a;

where __a is an otherwise invisible and inaccessible temporary variable. The following class represents a rectangle created from two points:

public class Rectangle
{
   Point p1, p2;
   public Point P1 { get { return p1; } set { p1 = value; } }
   public Point P2 { get { return p2; } set { p2 = value; } }
}

An instance of Rectangle can be created and initialized as follows:

var r = new Rectangle {
   P1 = new Point { X = 0, Y = 1 },
   P2 = new Point { X = 2, Y = 3 }
};

which has the same effect as

var __r = new Rectangle();
var __p1 = new Point();
__p1.X = 0;
__p1.Y = 1;
__r.P1 = __p1;
var __p2 = new Point();
__p2.X = 2;
__p2.Y = 3;
__r.P2 = __p2; 
var r = __r;

where __r, __p1 and __p2 are temporary variables that are otherwise invisible and inaccessible.

If the Rectangle constructor allocates the two embedded Point instances

public class Rectangle
{
   Point p1 = new Point();
   Point p2 = new Point();
   public Point P1 { get { return p1; } }
   public Point P2 { get { return p2; } }
}

the following construct can be used to initialize the embedded Point instances instead of assigning new instances:

var r = new Rectangle {
   P1 = { X = 0, Y = 1 },
   P2 = { X = 2, Y = 3 }
};

which has the same effect as

var __r = new Rectangle();
__r.P1.X = 0;
__r.P1.Y = 1;
__r.P2.X = 2;
__r.P2.Y = 3;
var r = __r;

26.4.2 Collection Initializers

A collection initializer specifies the elements of a collection.

collection-initializer:
{   element-initializer-list   }{   element-initializer-list   ,}
element-initializer-list:
element-initializer
element-initializer-list   ,   element-initializer
element-initializer:
non-assignment-expression
{   expression-list   }

A collection initializer consists of a sequence of element initializers, enclosed by { and } tokens and separated by commas. Each element initializer specifies an element to be added to the collection object being initialized, and consists of a list of expressions enclosed by { and } tokens and separated by commas. A single-expression element initializer can be written without braces, but cannot then be an assignment expression, to avoid ambiguity with member initializers. The non-assignment-expression production is defined in §26.3.

The following is an example of an object creation expression that includes a collection initializer:

List<int> digits = new List<int> { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 };

The collection object to which a collection initializer is applied must be of a type that implements System.Collections.IEnumerable or a compile-time error occurs. For each specified element in order, the collection initializer invokes the Add method on the target object with the expression list of the element initializer, applying normal overload resolution for each invocation.

The following class represents a contact with a name and a list of phone numbers:

public class Contact
{
   string name;
   List<string> phoneNumbers = new List<string>();
   public string Name { get { return name; } set { name = value; } }
   public List<string> PhoneNumbers { get { return phoneNumbers; } }
}

A List<Contact> can be created and initialized as follows:

var contacts = new List<Contact> {
   new Contact {
      Name = "Chris Smith",
      PhoneNumbers = { "206-555-0101", "425-882-8080" }
   },
   new Contact {
      Name = "Bob Harris",
      PhoneNumbers = { "650-555-0199" }
   }
};

which has the same effect as

var contacts = new List<Contact>();
var __c1 = new Contact();
__c1.Name = "Chris Smith";
__c1.PhoneNumbers.Add("206-555-0101");
__c1.PhoneNumbers.Add("425-882-8080");
contacts.Add(__c1);
var __c2 = new Contact();
__c2.Name = "Bob Harris";
__c2.PhoneNumbers.Add("650-555-0199");
contacts.Add(__c2);

where __c1 and __c2 are temporary variables that are otherwise invisible and inaccessible.

26.5 Anonymous Types

C# 3.0 permits the new operator to be used with an anonymous object initializer to create an object of an anonymous type.

primary-no-array-creation-expression:
...
anonymous-object-creation-expression
anonymous-object-creation-expression:
new   anonymous-object-initializer
anonymous-object-initializer:
{   member-declarator-listopt}{   member-declarator-list   ,}
member-declarator-list:
member-declarator
member-declarator-list   ,   member-declarator
member-declarator:
simple-name
member-access
identifier   =   expression

An anonymous object initializer declares an anonymous type and returns an instance of that type. An anonymous type is a nameless class type that inherits directly from object. The members of an anonymous type are a sequence of read/write properties inferred from the object initializer(s) used to create instances of the type. Specifically, an anonymous object initializer of the form

new { p1 = e1 , p2 = e2 , ...pn = en }

declares an anonymous type of the form

class __Anonymous1
{
   private T1f1 ;
   private T2f2 ;
   ...
   private Tnfn ;
   public T1p1 { get { return f1 ; } set { f1 = value ; } }
   public T2p2 { get { return f2 ; } set { f2 = value ; } }
   ...
   public T1p1 { get { return f1 ; } set { f1 = value ; } }
}

where each Tx is the type of the corresponding expression ex. It is a compile-time error for an expression in an anonymous object initializer to be of the null type or an unsafe type.

The name of an anonymous type is automatically generated by the compiler and cannot be referenced in program text.

Within the same program, two anonymous object initializers that specify a sequence of properties of the same names and compile-time types in the same order will produce instances of the same anonymous type. (This definition includes the order of the properties because it is observable and material in certain circumstances, such as reflection).

In the example

var p1 = new { Name = "Lawnmower", Price = 495.00 };
var p2 = new { Name = "Shovel", Price = 26.95 };
p1 = p2;

the assignment on the last line is permitted because p1 and p2 are of the same anonymous type.

The Equals and GetHashcode methods on anonymous types are defined in terms of the Equals and GetHashcode of the properties, so that two instances of the same anonymous type are equal if and only if all their properties are equal.

A member declarator can be abbreviated to a simple name (§7.5.2) or a member access (§7.5.4). This is called a projection initializer and is shorthand for a declaration of and assignment to a property with the same name. Specifically, member declarators of the forms

identifier                        expr . identifier

are precisely equivalent to the following, respectively:

identifer = identifieridentifier = expr . identifier

Thus, in a projection initializer the identifier selects both the value and the field or property to which the value is assigned. Intuitively, a projection initializer projects not just a value, but also the name of the value.

26.6 Implicitly Typed Arrays

The syntax of array creation expressions (§7.5.10.2) is extended to support implicitly typed array creation expressions:

array-creation-expression:
...
new[]   array-initializer

In an implicitly typed array creation expression, the type of the array instance is inferred from the elements specified in the array initializer. Specifically, the set formed by the types of the expressions in the array initializer must contain exactly one type to which each type in the set is implicitly convertible, and if that type is not the null type, an array of that type is created. If exactly one type cannot be inferred, or if the inferred type is the null type, a compile-time error occurs.

The following are examples of implicitly typed array creation expressions:

var a = new[] { 1, 10, 100, 1000 };            // int[]
var b = new[] { 1, 1.5, 2, 2.5 };            // double[]
var c = new[] { "hello", null, "world” };      // string[]
var d = new[] { 1, "one", 2, "two" };         // Error

The last expression causes a compile-time error because neither int nor string is implicitly convertible to the other. An explicitly typed array creation expression must be used in this case, for example specifying the type to be object[]. Alternatively, one of the elements can be cast to a common base type, which would then become the inferred element type.

Implicitly typed array creation expressions can be combined with anonymous object initializers to create anonymously typed data structures. For example:

var contacts = new[] {
   new {
      Name = "Chris Smith",
      PhoneNumbers = new[] { "206-555-0101", "425-882-8080" }
   },
   new {
      Name = "Bob Harris",
      PhoneNumbers = new[] { "650-555-0199" }
   }
};

26.7 Query Expressions

Query expressions provide a language integrated syntax for queries that is similar to relational and hierarchical query languages such as SQL and XQuery.

query-expression:
from-clause   query-body
from-clause:
from   typeopt   identifier   in   expression
query-body:
query-body-clausesopt   select-or-group-clause   query-continuationopt
query-body-clauses:
query-body-clause
query-body-clauses   query-body-clause
query-body-clause:
from-clause
let-clause
where-clause
join-clause
join-into-clause
orderby-clause
let-clause:
let   identifier   =   expression
where-clause:
where   boolean-expression
join-clause:
join   typeopt   identifier   in   expression   on   expression   equals   expression 
join-into-clause:
join   typeopt   identifier   in   expression   on   expression   equals   expression   into   identifier
orderby-clause:
orderby   orderings
orderings:
ordering
orderings   ,   ordering
ordering:
expression    ordering-directionopt
ordering-direction:
ascending
descending
select-or-group-clause:
select-clause
group-clause
select-clause:
select   expression
group-clause:
group   expression   by   expression
query-continuation:
into   identifier   query-body

A query-expression is classified as a non-assignment-expression, the definition of which occurs in §26.3.

A query expression begins with a from clause and ends with either a select or group clause. The initial from clause can be followed by zero or more from, let, where or join clauses. Each from clause is a generator introducing a range variable ranging over a sequence. Each let clause computes a value and introduces an identifier representing that value. Each where clause is a filter that excludes items from the result. Each join clause compares specified keys of the source sequence with keys of another sequence, yielding matching pairs. Each orderby clause reorders items according to specified criteria. The final select or group clause specifies the shape of the result in terms of the range variable(s). Finally, an into clause can be used to "splice" queries by treating the results of one query as a generator in a subsequent query.

Ambiguities in query expressions

Query expressions contain a number of new contextual keywords, i.e., identifiers that have special meaning in a given context. Specifically these are: from, join, on, equals, into, let, orderby, ascending, descending, select, group and by. In order to avoid ambiguities caused by mixed use of these identifiers as keywords or simple names in query expressions, they are considered keywords anywhere within a query expression.

For this purpose, a query expression is any expression starting with "from identifier" followed by any token except ";", "=" or ",".

In order to use these words as identifiers within a query expression, they can be prefixed with "@" (§2.4.2).

26.7.1 Query Expression Translation

The C# 3.0 language does not specify the exact execution semantics of query expressions. Rather, C# 3.0 translates query expressions into invocations of methods that adhere to the query expression pattern. Specifically, query expressions are translated into invocations of methods named Where, Select, SelectMany, Join, GroupJoin, OrderBy, OrderByDescending, ThenBy, ThenByDescending, GroupBy, and Cast that are expected to have particular signatures and result types, as described in §26.7.2. These methods can be instance methods of the object being queried or extension methods that are external to the object, and they implement the actual execution of the query.

The translation from query expressions to method invocations is a syntactic mapping that occurs before any type binding or overload resolution has been performed. The translation is guaranteed to be syntactically correct, but it is not guaranteed to produce semantically correct C# code. Following translation of query expressions, the resulting method invocations are processed as regular method invocations, and this may in turn uncover errors, for example if the methods do not exist, if arguments have wrong types, or if the methods are generic and type inference fails.

A query expression is processed by repeatedly applying the following translations until no further reductions are possible. The translations are listed in order of precedence: each section assumes that the translations in the preceding sections have been performed exhaustively.

Certain translations inject range variables with transparent identifiers denoted by *. The special properties of transparent identifiers are discussed further in §26.7.1.7.

26.7.1.1 Select and groupby clauses with continuations

A query expression with a continuation

from ... into x...

is translated into

from x in ( from ... ) ...

The translations in the following sections assume that queries have no into continuations.

The example

from c in customers
group c by c.Country into g
select new { Country = g.Key, CustCount = g.Count() }

is translated into

from g in
   from c in customers
   group c by c.Country
select new { Country = g.Key, CustCount = g.Count() }

the final translation of which is

customers.
GroupBy(c => c.Country).
Select(g => new { Country = g.Key, CustCount = g.Count() })

26.7.1.2 Explicit range variable types

A from clause that explicitly specifies a range variable type

from Tx in e

is translated into

from x in ( e ) . Cast < T > ( )

A join clause that explicitly specifies a range variable type

join Tx in e on k1 equals k2

is translated into

join x in ( e ) . Cast < T > ( ) on k1 equals k2

The translations in the following sections assume that queries have no explicit range variable types.

The example

from Customer c in customers
where c.City == "London"
select c

is translated into

from c in customers.Cast<Customer>()
where c.City == "London"
select c

the final translation of which is

customers.
Cast<Customer>().
Where(c => c.City == "London")

Explicit range variable types are useful for querying collections that implement the non-generic IEnumerable interface, but not the generic IEnumerable<T> interface. In the example above, this would be the case if customers were of type ArrayList.

26.7.1.3 Degenerate query expressions

A query expression of the form

from x in e select x

is translated into

( e ) . Select ( x => x )

The example

from c in customers
select c

Is translated into

customers.Select(c => c)

A degenerate query expression is one that trivially selects the elements of the source. A later phase of the translation removes degenerate queries introduced by other translation steps by replacing them with their source. It is important however to ensure that the result of a query expression is never the source object itself, as that would reveal the type and identity of the source to the client of the query. Therefore this step protects degenerate queries written directly in source code by explicitly calling Select on the source. It is then up to the implementers of Select and other query operators to ensure that these methods never return the source object itself.

26.7.1.4 From, let, where, join and orderby clauses

A query expression with a second from clause followed by a select clause

from x1 in e1
from x2 in e2
select v

is translated into

( e1 ) . SelectMany( x1 => e2 , ( x1 , x2 ) => v )

A query expression with a second from clause followed by something other than a select clause:

from x1 in e1
from x2 in e2
...

is translated into

from * in ( e1 ) . SelectMany( x1 => e2 , ( x1 , x2 ) => new { x1 , x2 } )
...

A query expression with a let clause

from x in e
let y = f...

is translated into

from * in ( e ) . Select ( x => new { x , y = f } )
...

A query expression with a where clause

from x in e
where f...

is translated into

from x in ( e ) . Where ( x => f )
...

A query expression with a join clause without an into followed by a select clause

from x1 in e1
join x2 in e2 on k1 equals k2
select v

is translated into

( e1 ) . Join( e2 , x1 => k1 , x2 => k2 , ( x1 , x2 ) => v )

A query expression with a join clause without an into followed by something other than a select clause

from x1 in e1
join x2 in e2 on k1 equals k2 
...

is translated into

from * in ( e1 ) . Join(
   e2 , x1 => k1 , x2 => k2 , ( x1 , x2 ) => new { x1 , x2 })
...

A query expression with a join clause with an into followed by a select clause

from x1 in e1
join x2 in e2 on k1 equals k2 into g
select v

is translated into

( e1 ) . GroupJoin( e2 , x1 => k1 , x2 => k2 , ( x1 , g ) => v )

A query expression with a join clause with an into followed by something other than a select clause

from x1 in e1
join x2 in e2 on k1 equals k2 into g
...

is translated into

from * in ( e1 ) . GroupJoin(
   e2 , x1 => k1 , x2 => k2 , ( x1 , g ) => new { x1 , g })
...

A query expression with an orderby clause

from x in e
orderby k1 , k2 , ... ,kn...

is translated into

from x in ( e ) . 
OrderBy ( x => k1 ) . 
ThenBy ( x => k2 ) .
 ... . 
ThenBy ( x => kn )
...

If an ordering clause specifies a descending direction indicator, an invocation of OrderByDescending or ThenByDescending is produced instead.

The following translations assume that there are no let, where, join or orderby clauses, and no more than the one initial from clause in each query expression.

The example

from c in customers
from o in c.Orders
select new { c.Name, o.OrderID, o.Total }

is translated into

customers.
SelectMany(c => c.Orders,
    (c,o) => new { c.Name, o.OrderID, o.Total }
)

The example

from c in customers
from o in c.Orders
orderby o.Total descending
select new { c.Name, o.OrderID, o.Total }

is translated into

from * in customers.
   SelectMany(c => c.Orders, (c,o) => new { c, o })
orderby o.Total descending
select new { c.Name, o.OrderID, o.Total }

the final translation of which is

customers.
SelectMany(c => c.Orders, (c,o) => new { c, o }).
OrderByDescending(x => x.o.Total).
Select(x => new { x.c.Name, x.o.OrderID, x.o.Total })

where x is a compiler generated identifier that is otherwise invisible and inaccessible.

The example

from o in orders
let t = o.Details.Sum(d => d.UnitPrice * d.Quantity)
where t >= 1000
select new { o.OrderID, Total = t }

is translated into

from * in orders.
   Select(o => new { o, t = o.Details.Sum(d => d.UnitPrice * d.Quantity) })
where t >= 1000 
select new { o.OrderID, Total = t }

the final translation of which is

orders.
Select(o => new { o, t = o.Details.Sum(d => d.UnitPrice * d.Quantity) }).
Where(x => x.t >= 1000).
Select(x => new { x.o.OrderID, Total = x.t })

where x is a compiler generated identifier that is otherwise invisible and inaccessible.

The example

from c in customers
join o in orders on c.CustomerID equals o.CustomerID
select new { c.Name, o.OrderDate, o.Total }

is translated into

customers.Join(orders, c => c.CustomerID, o => o.CustomerID,
   (c, o) => new { c.Name, o.OrderDate, o.Total })

The example

from c in customers
join o in orders on c.CustomerID equals o.CustomerID into co
let n = co.Count()
where n >= 10
select new { c.Name, OrderCount = n }

is translated into

from * in customers.
   GroupJoin(orders, c => c.CustomerID, o => o.CustomerID,
      (c, co) => new { c, co })
let n = co.Count()
where n >= 10 
select new { c.Name, OrderCount = n }

the final translation of which is

customers.
GroupJoin(orders, c => c.CustomerID, o => o.CustomerID,
   (c, co) => new { c, co }).
Select(x => new { x, n = x.co.Count() }).
Where(y => y.n >= 10).
Select(y => new { y.x.c.Name, OrderCount = y.n)

where x and y are compiler generated identifiers that are otherwise invisible and inaccessible.

The example

from o in orders
orderby o.Customer.Name, o.Total descending
select o

has the final translation

orders.
OrderBy(o => o.Customer.Name).
ThenByDescending(o => o.Total)

26.7.1.5 Select clauses

A query expression of the form

from x in e select v

is translated into

( e ) . Select ( x => v )

except when v is the identifier x, the translation is simply

( e )

For example

from c in customers.Where(c => c.City == "London")
select c

is simply translated into

customers.Where(c => c.City == "London")

26.7.1.6 Groupby clauses

A query expression of the form

from x in e group v by k

is translated into

( e ) . GroupBy ( x => k , x => v )

except when v is the identifier x, the translation is

( e ) . GroupBy ( x => k )

The example

from c in customers
group c.Name by c.Country

is translated into

customers.
GroupBy(c => c.Country, c => c.Name)

26.7.1.7 Transparent identifiers

Certain translations inject range variables with transparent identifiers denoted by *. Transparent identifiers are not a proper language feature; they exist only as an intermediate step in the query expression translation process.

When a query translation injects a transparent identifier, further translation steps propagate the transparent identifier into lambda expressions and anonymous object initializers. In those contexts, transparent identifiers have the following behavior:

  • When a transparent identifier occurs as a parameter in a lambda expression, the members of the associated anonymous type are automatically in scope in the body of the lambda expression.
  • When a member with a transparent identifier is in scope, its members are in scope as well.
  • When a transparent identifier occurs as a member declarator in an anonymous object initializer, it introduces a member with a transparent identifier.

The example

from c in customers
from o in c.Orders
orderby o.Total descending
select new { c.Name, o.Total }

is translated into

from * in
   from c in customers
   from o in c.Orders
   select new { c, o }
orderby o.Total descending
select new { c.Name, o.Total }

which is further translated into

customers.
SelectMany(c => c.Orders.Select(o => new { c, o })).
OrderByDescending(* => o.Total).
Select(* => new { c.Name, o.Total })

which, when transparent identifiers are erased, is equivalent to

customers.
SelectMany(c => c.Orders.Select(o => new { c, o })).
OrderByDescending(x => x.o.Total).
Select(x => new { x.c.Name, x.o.Total })

where x is a compiler generated identifier that is otherwise invisible and inaccessible.

The example

from c in customers
join o in orders on c.CustomerID equals o.CustomerID
join d in details on o.OrderID equals d.OrderID
join p in products on d.ProductID equals p.ProductID
select new { c.Name, o.OrderDate, p.ProductName }

is translated into

from * in
   from * in
      from * in
         from c in customers
         join o in orders o c.CustomerID equals o.CustomerID
         select new { c, o }
      join d in details on o.OrderID equals d.OrderID
      select new { *, d }
   join p in products on d.ProductID equals p.ProductID
   select new { *, p }
select new { c.Name, o.OrderDate, p.ProductName }

which is further reduced to

customers.
Join(orders, c => c.CustomerID, o => o.CustomerID,
   (c, o) => new { c, o }).
Join(details, * => o.OrderID, d => d.OrderID,
   (*, d) => new { *, d }).
Join(products, * => d.ProductID, p => p.ProductID,
   (*, p) => new { *, p }).
Select(* => new { c.Name, o.OrderDate, p.ProductName })

the final translation of which is

customers.
Join(orders, c => c.CustomerID, o => o.CustomerID,
   (c, o) => new { c, o }).
Join(details, x => x.o.OrderID, d => d.OrderID,
   (x, d) => new { x, d }).
Join(products, y => y.d.ProductID, p => p.ProductID,
   (y, p) => new { y, p }).
Select(z => new { z.y.x.c.Name, z.y.x.o.OrderDate, z.p.ProductName })

where x, y, and z are compiler generated identifiers that are otherwise invisible and inaccessible.

26.7.2 The Query Expression Pattern

The Query Expression Pattern establishes a pattern of methods that types can implement to support query expressions. Because query expressions are translated to method invocations by means of a syntactic mapping, types have considerable flexibility in how they implement the query expression pattern. For example, the methods of the pattern can be implemented as instance methods or as extension methods because the two have the same invocation syntax, and the methods can request delegates or expression trees because lambda expressions are convertible to both.

The recommended shape of a generic type C<T> that supports the query expression pattern is shown below. A generic type is used in order to illustrate the proper relationships between parameter and result types, but it is possible to implement the pattern for non-generic types as well.

delegate R Func<T1,R>(T1 arg1);
delegate R Func<T1,T2,R>(T1 arg1, T2 arg2);
class C
{
   public C<T> Cast<T>();
}
class C<T>
{
   public C<T> Where(Func<T,bool> predicate);
   public C<U> Select<U>(Func<T,U> selector);
   public C<U> SelectMany<U,V>(Func<T,C<U>> selector,
      Func<T,U,V> resultSelector);
   public C<V> Join<U,K,V>(C<U> inner, Func<T,K> outerKeySelector,
      Func<U,K> innerKeySelector, Func<T,U,V> resultSelector);
   public C<V> GroupJoin<U,K,V>(C<U> inner, Func<T,K> outerKeySelector,
      Func<U,K> innerKeySelector, Func<T,C<U>,V> resultSelector);
   public O<T> OrderBy<K>(Func<T,K> keySelector);
   public O<T> OrderByDescending<K>(Func<T,K> keySelector);
   public C<G<K,T>> GroupBy<K>(Func<T,K> keySelector);
   public C<G<K,E>> GroupBy<K,E>(Func<T,K> keySelector,
      Func<T,E> elementSelector);
}
class O<T> : C<T>
{
   public O<T> ThenBy<K>(Func<T,K> keySelector);
   public O<T> ThenByDescending<K>(Func<T,K> keySelector);
}
class G<K,T> : C<T>
{
   public K Key { get; }
}

The methods above use the generic delegate types Func<T1, R> and Func<T1, T2, R>, but they could equally well have used other delegate or expression tree types with the same relationships in parameter and result types.

Notice the recommended relationship between C<T> and O<T> which ensures that the ThenBy and ThenByDescending methods are available only on the result of an OrderBy or OrderByDescending. Also notice the recommended shape of the result of GroupBy—a sequence of sequences, where each inner sequence has an additional Key property.

The Standard Query Operators (described in a separate specification) provide an implementation of the query operator pattern for any type that implements the System.Collections.Generic.IEnumerable<T> interface.

26.8 Expression Trees

Expression trees permit lambda expressions to be represented as data structures instead of executable code. A lambda expression that is convertible to a delegate type D is also convertible to an expression tree of type System.Query.Expression<D>. Whereas the conversion of a lambda expression to a delegate type causes executable code to be generated and referenced by a delegate, conversion to an expression tree type causes code that creates an expression tree instance to be emitted. Expression trees are efficient in-memory data representations of lambda expressions and make the structure of the expression transparent and explicit.

The following example represents a lambda expression both as executable code and as an expression tree. Because a conversion exists to Func<int,int>, a conversion also exists to Expression<Func<int,int>>.

Func<int,int> f = x => x + 1;                  // Code
Expression<Func<int,int>> e = x => x + 1;      // Data

Following these assignments, the delegate f references a method that returns x + 1, and the expression tree e references a data structure that describes the expression x + 1.

26.8.1 Overload Resolution

For the purpose of overload resolution there are special rules regarding the Expression<D> types. Specifically the following rule is added to the definition of betterness:

  • Expression<D1> is better than Expression<D2> if and only if D1 is better than D2

Note   that there is no betterness rule between Expression<D> and delegate types.

26.9 Automatically Implemented Properties

Oftentimes properties are implemented by trivial use of a backing field, as in the following example:

public Class Point {
   private int x;
   private int y;
   public int X { get { return x; } set { x = value; } }
   public int Y { get { return y; } set { y = value; } }
}

Automatically implemented (auto-implemented) properties automate this pattern. More specifically, non-abstract property declarations are allowed to have semicolon accessor bodies. Both accessors must be present and both must have semicolon bodies, but they can have different accessibility modifiers. When a property is specified like this, a backing field will automatically be generated for the property, and the accessors will be implemented to read from and write to that backing field. The name of the backing field is compiler generated and inaccessible to the user.

The following declaration is equivalent to the example above:

 public Class Point {
   public int X { get; set; }
   public int Y { get; set; }
}

Because the backing field is inaccessible, it can be read and written only through the property accessors. This means that auto-implemented read-only or write-only properties do not make sense, and are disallowed. It is however possible to set the access level of each accessor differently. Thus, the effect of a read-only property with a private backing field can be mimicked like this:

Public class ReadOnlyPoint {
   public int X { get; private set; }
   public int Y { get; private set; }
   public ReadOnlyPoint(int x, int y) { X = x; Y = y; }
}

This restriction also means that definite assignment of struct types with auto-implemented properties can only be achieved using the standard constructor of the struct, since assigning to the property itself requires the struct to be definitely assigned.