LINQ
Slides from Gang Luo, Xuting Zhao and
Damien Guard
What is LINQ?

Language Integrated Query
 Make query a part of the language
 Component of .NET Framework 3.5
Query without LINQ

Objects using loops and conditions
foreach (Customer c in customers)
if (c.Region == "UK") ...

Databases using SQL
SELECT * FROM Customers WHERE Region='UK'

XML using XPath/XQuery
//Customers/Customer[@Region='UK']
ADO without LINQ
SqlConnection con = new SqlConnection(...);
con.Open();
SqlCommand cmd = new SqlCommand(
@"SELECT * FROM Customers
WHERE c.Region = @Region", con
);
cmd.Parameters.AddWithValue("@Region", "UK");
DataReader dr = cmd.ExecuteReader();
while (dr.Read()) {
string name = dr.GetString(dr.GetOrdinal("Name"));
string phone = dr.GetString(dr.GetOrdinal("Phone"));
DateTime date = dr.GetDateTime(3);
}
dr.Close();
con.Close();
Query with LINQ
C#
var myCustomers = from c in customers
where c.Region == "UK"
select c;
More LINQ queries
var goodCusts = (from c in db.Customers
where c.PostCode.StartsWith("GY")
orderby c.Sales descending
select c).Skip(10).Take(10);
Local variable type inference

Compiler can infer types from initializer
assignments
 var keyword indicates compiler inferred type
 Still strong-typed
 This is not like JavaScript var
var a = 10; // Simple types
var x = new {
Blog = “attardi”, Created = DateTime.Now };
// Anonymous types

Essential when using anonymous types
Anonymous Types

Object of new type generated on the fly without first defining
it.
Useful for projection to select one or more fields of another
structure.

The type will be dynamically generated with setters and
getters to corresponding members.
Some common methods are also provided.

No other methods will be added to this type.
But that is already enough!

The object is created and initialized by Anonymous Object
Initializer.
Advantages

Unified data access
Single syntax to learn and remember

Strongly typed
Catch errors during compilation

IntelliSense
Prompt for syntax and attributes

Bindable result sets
Architecture
LINQ to Objects
int[] nums = new int[] {0,4,2,6,3,8,3,1};
double average = nums.Take(6).Average();
var above = from n in nums
where n > average
select n;
LINQ to Objects

Query any IEnumerable<T> source
Includes arrays, List<T>, Dictionary...

Many useful operators available
Sum, Max, Min, Distinct, Intersect, Union

Expose your own data with
IEnumerable<T> or IQueryable<T>

Create operators using extension methods
LINQ operators
Aggregate Conversion
Aggregate
Average
Count
Max
Min
Sum
Ordering Partitioning
Cast
OrderBy
OfType
ThenBy
ToArray
Descending
ToDictionary Reverse
ToList
ToLookup
ToSequence
and others …
Skip
SkipWhile
Take
TakeWhile
Sets
Concat
Distinct
Except
Intersect
Union
Query Expression

SQL-like:
from s in names
where s.Length == 5 orderby select s.ToUpper();

OO-style:
names.Where(s => s.Length==5)
.OrderBy(s => s)
.Select(s => s.ToUpper());

Where, OrderBy, and Select are operators.
The arguments to these operators are
Lambda Expression.
Lambda Expressions

Examples:
 s => s.Length == 5

Executable function
 Anonymous functional. Can be assigned to a delegate
variable.
 No need to indicate the types
 Can be passed to methods as parameters.

Expression Tree
 Efficient in-memory data representations of lambda
expressions
 Changing the behaviors of the expressions
 Applying your own optimization
Function Types

Func<int, bool> is a shorthand for
public delegate bool Func(int a0);
// Initialized with anonymous method
Func<int, bool> even=
delegate (int x) { return x % 2 == 0; };
// Initialized with lambda expression
Func<int, bool> even2 = x => x % 2 == 0;
Methods Extension

Control not only by Lambda Expression, but
also by methods extension
public static class Enumerable {
public static IEnumerable<T> Where<T>(
this IEnumerable<T> source,
Func<T, bool> predicate) {
foreach (T item in source)
if (predicate(item))
yield return item;
}
}
LINQ Operations

Join
 When there is relationship (e.g. foreign key) between two tables,
no explicit join operation is needed
 Using dot notation to access the relationship properties, and
navigate all the matching objects.
var q = from o in db.Orders where o.Customer.City == “London”
select o;
 To join any two data sources on any attribute, you need an
explicit join operation.
var query = names.Join(people, n => n, p => p.Name, (n, p) =>
p);
The lambda expression for shaping (n, p) => p will be applied on
each matching pairs.
LINQ Operations (cont.)

Group Join
 The lambda expression for shaping is applied on the
outer element and the set of all the inner elements that
matches the outer one.
 Shape the result at a set level
var query = names.GroupJoin(people,
n => n,
p => p.Name,
(n, matching) => new { Name = n, Count =
matching.Count() }
)
LINQ Operations (cont.)

Select Many
 Each object in the result set may contain a collection or array
 Select many help decompose the structure and flatten the result
var query = names.SelectMany(n =>
people.Where(p => n == p.Name))
 All the elements could be traversed in one foreach loop.

Aggregation
 Standard aggregation operators: Min, Max, Sum, Average.
int totalLength=names.Sum(s => s.Length);
 General purpose (generic) operator:
static U Aggregate<T, U>(this IEnumerable<T> source,
U seed, Func<U, T, U> func)
LINQ Deferred

A LINQ data source can actually implement
one of two interfaces:
 IEnumerable<T>
 IQueryable<T>
public interface IQueryable<T> :
IEnumerable<T>, IQueryable, IEnumerable {
IQueryable<S> CreateQuery<S>(Expression
exp);
S Execute<S>(Expression exp);
}

Create deferred query execution plan
IQueryable

IQueryable<T> interface will defer the
evaluation of the query.

An expression tree will represent all the
deferred queries as a whole.

Several operations could be “merged”, only
one SQL query will be generated and sent to
database.

Multi-level defer
Lambda Expressions Revisited

Lambda expressionscan represent either IL
code or data
 Expression<T> makes the difference
Func<int, bool> lambdaIL =
n => n % 2 == 0;
Expression<Func<int, bool>> lambdaTree =
n => n % 2 == 0;

Compiler handles Expression<T> types
differently
 Emits code to generate expression tree instead of
usual IL for delegate
Expression Trees

Expression tree are hierarchical trees of
instructions that compose an expression
 Add value of expression trees
 Actual creating of IL is deferred until execution of
query
 Implementation of IL creation can vary

Trees can even be remoted for parallel
processing
Creating IL from Expression Trees

Right before execution tree is compiled into IL
Expression<Func<Posting,bool>> predicate =
p => p.Posted < DateTime.Now.AddDays(-5);
Func<Posting,bool> d = predicate.Compile();

Implementation of IL generation differs very
much for each Linq flavor.
 Linq to SQL generates IL that runs SQL
commands
 Linq to Objects builds IL with Sequence
extensions methods
Nested defer

Nested defer
var q = from c in db.Customers
where c.City == “London”
select new { c.ContactName, c.Phone }
var q2 = from c in q.AsEnumerable()
select new {
Name = DoNameProcessing(c.ContactName),
Phone = DoPhoneProcessing(C.Phone)
};

What if you want the intermediate result?
string lastName = “Simpson”
var persons = from p in personList where p.LastName = lastName select p;
lastName = “Flanders”
foreach (Person p in persons)
Console.WriteLine(“{0} {1}”, p.FirstName, p.LastName);
Deferred Execution
 Advantages
 Performance!
 Query dependency!
 Disadvantages
 Divide one query into multiple ones
 If you iterate over the result set 100
times, the query will be executed 100
times.
 Users have to be very careful
LINQ to SQL

Data Model
[Table(Name=“Customers”)]
public class Customer {
[Column(Id=true)]
public string CustomerID;
…
private EntitySet<Order> _Orders;
[Association(Storage=“_Orders”, OtherKey=“CustomerID”)]
public EntitySet<Order> Orders {
get { return _Orders; }
set { _Orders.Assign(value); }
}
}

LINQ to SQL helps connect to relational and manipulate the
relational data as objects in memory.

It achieves this by translating the operations into SQL statements.
LINQ to SQL





Object-relational mapping
Records become strongly-typed objects
Data context is the controller mechanism
Facilitates update, delete & insert
Translates LINQ queries behind the scenes
Type, parameter and injection safe
Database mapping





Map tables & fields to classes & properties
Generates partial classes with attributes
Each record becomes an object
Data context represents the database
Utilize tables, views or stored procedures
Modifying objects

Update
Set object properties
 Delete
context.Table.DeleteOnSubmit(object)

Insert
context.Table.InsertOnSubmit(object)

Commit changes back
context.SubmitChanges()
Transactional - all or nothing
Consistency

Every object will be tracked by LINQ the
moment it is loaded from database.

The tracking mechanism monitor the
manipulation on relationship properties.
Once you modify one side of the relationship,
LINQ will modify the other to keep it
consistent.

When an object is deleted, it could still exist in
memory, but it will not cause inconsistency.
Concurrency





Optimistic concurrency
Conflict checking when SubmitChanges() is
called
By default, transaction will abort and an
exception will be thrown when a conflict is
detected.
User can handle the conflict in the exception
catch block.
User can set whether or not to detect the
conflict when one column get updated.
Transaction/Update

When update, first check whether new object
is added (by tracking mechanism) if yes,
insert statement will be generated.
What does Django do here?

Modification will not hit the database until the
SubmitChanges() method is called

All operations will be translated into SQL
statements

All modifications will be encapsulated into a
transaction.
Transaction/Update (cont.)

If an exception is throw during the
update, all the changes will be rolled
back

One SubmitChanges() is actually one
transaction. (pros and cons?)

Users can also explicitly indicate a new
transaction scope.
LINQ to XML Class Hierarchy
http://msdn.microsoft.com/en-us/library/bb308960.aspx
LINQ to XML

LINQ to XML
var query = from p in people
where p.CanCode
select new XElement(“Person”,
new XAttribute(“Age”, p.Age),
p.Name);

XML to LINQ
var x = new XElement(“People”,
from p in people
where p.CanCode
select
new XElement(“Person”,
new XAttribute(“Age”, p.Age),
p.Name);
Performance

LINQ has more control and efficiency in O/R
Mapping than NHibernate
 LINQ: Externl Mapping or Attribute Mapping
 NHibernate: Externl Mapping

Because of mapping, LINQ is slower than
database tools such as SqlDataReader or
SqlDataAdapter
 In large dataset, their performance are more and
more similar