LEFT OUTER JOIN in LINQ

Question

How to perform left outer join in C# LINQ to objects without using join-on-equals-into clauses? Is there any way to do that with where clause? Correct problem: For inner join is easy and I have a solution like this

List<JoinPair> innerFinal = (from l in lefts from r in rights where l.Key == r.Key
                             select new JoinPair { LeftId = l.Id, RightId = r.Id})

but for left outer join I need a solution. Mine is something like this but it's not working

List< JoinPair> leftFinal = (from l in lefts from r in rights
                             select new JoinPair { 
                                            LeftId = l.Id, 
                                            RightId = ((l.Key==r.Key) ? r.Id : 0
                                        })

where JoinPair is a class:

public class JoinPair { long leftId; long rightId; }

Answer 1

Using lambda expression

db.Categories    
  .GroupJoin(
      db.Products,
      Category => Category.CategoryId,
      Product => Product.CategoryId,
      (x, y) => new { Category = x, Products = y })
  .SelectMany(
      xy => xy.Products.DefaultIfEmpty(),
      (x, y) => new { Category = x.Category, Product = y })
  .Select(s => new
  {
      CategoryName = s.Category.Name,     
      ProductName = s.Product.Name   
  })

Answer 2

Necromancing.
If a database driven LINQ provider is used, a significantly more readable left outer join can be written as such:

from maintable in Repo.T_Whatever 
from xxx in Repo.T_ANY_TABLE.Where(join condition).DefaultIfEmpty()

If you omit the DefaultIfEmpty() you will have an inner join.

Take the accepted answer:

  from c in categories
    join p in products on c equals p.Category into ps
    from p in ps.DefaultIfEmpty()

This syntax is very confusing, and it's not clear how it works when you want to left join MULTIPLE tables.

Note
It should be noted that from alias in Repo.whatever.Where(condition).DefaultIfEmpty() is the same as an outer-apply/left-join-lateral, which any (decent) database-optimizer is perfectly capable of translating into a left join, as long as you don't introduce per-row-values (aka an actual outer apply). Don't do this in Linq-2-Objects (because there's no DB-optimizer when you use Linq-to-Objects).

Detailed Example

var query2 = (
    from users in Repo.T_User
    from mappings in Repo.T_User_Group
         .Where(mapping => mapping.USRGRP_USR == users.USR_ID)
         .DefaultIfEmpty() // <== makes join left join
    from groups in Repo.T_Group
         .Where(gruppe => gruppe.GRP_ID == mappings.USRGRP_GRP)
         .DefaultIfEmpty() // <== makes join left join

    // where users.USR_Name.Contains(keyword)
    // || mappings.USRGRP_USR.Equals(666)  
    // || mappings.USRGRP_USR == 666 
    // || groups.Name.Contains(keyword)

    select new
    {
         UserId = users.USR_ID
        ,UserName = users.USR_User
        ,UserGroupId = groups.ID
        ,GroupName = groups.Name
    }

);


var xy = (query2).ToList();

When used with LINQ 2 SQL it will translate nicely to the following very legible SQL query:

SELECT 
     users.USR_ID AS UserId 
    ,users.USR_User AS UserName 
    ,groups.ID AS UserGroupId 
    ,groups.Name AS GroupName 
FROM T_User AS users

LEFT JOIN T_User_Group AS mappings
   ON mappings.USRGRP_USR = users.USR_ID

LEFT JOIN T_Group AS groups
    ON groups.GRP_ID == mappings.USRGRP_GRP

Edit:

See also " Convert SQL Server query to Linq query " for a more complex example.

Also, If you're doing it in Linq-2-Objects (instead of Linq-2-SQL), you should do it the old-fashioned way (because LINQ to SQL translates this correctly to join operations, but over objects this method forces a full scan, and doesn't take advantage of index searches, whyever...):

    var query2 = (
    from users in Repo.T_Benutzer
    join mappings in Repo.T_Benutzer_Benutzergruppen on mappings.BEBG_BE equals users.BE_ID into tmpMapp
    join groups in Repo.T_Benutzergruppen on groups.ID equals mappings.BEBG_BG into tmpGroups
    from mappings in tmpMapp.DefaultIfEmpty()
    from groups in tmpGroups.DefaultIfEmpty()
    select new
    {
         UserId = users.BE_ID
        ,UserName = users.BE_User
        ,UserGroupId = mappings.BEBG_BG
        ,GroupName = groups.Name
    }

);

Answer 3

As stated on:

101 LINQ Samples - Left outer join

var q =
    from c in categories
    join p in products on c.Category equals p.Category into ps
    from p in ps.DefaultIfEmpty()
    select new { Category = c, ProductName = p == null ? "(No products)" : p.ProductName };

Answer 4

I would like to add that if you get the MoreLinq extension there is now support for both homogenous and heterogeneous left joins now

http://morelinq.github.io/2.8/ref/api/html/Overload_MoreLinq_MoreEnumerable_LeftJoin.htm

example:

//Pretend a ClientCompany object and an Employee object both have a ClientCompanyID key on them

return DataContext.ClientCompany
    .LeftJoin(DataContext.Employees,                         //Table being joined
        company => company.ClientCompanyID,                  //First key
        employee => employee.ClientCompanyID,                //Second Key
        company => new {company, employee = (Employee)null}, //Result selector when there isn't a match
        (company, employee) => new { company, employee });   //Result selector when there is a match

EDIT:

In retrospect this may work, but it converts the IQueryable to an IEnumerable as morelinq does not convert the query to SQL.

You can instead use a GroupJoin as described here: https://stackoverflow.com/a/24273804/4251433

This will ensure that it stays as an IQueryable in case you need to do further logical operations on it later.

Answer 5

An implementation of left outer join by extension methods could look like

public static IEnumerable<Result> LeftJoin<TOuter, TInner, TKey, Result>(
  this IEnumerable<TOuter> outer, IEnumerable<TInner> inner
  , Func<TOuter, TKey> outerKeySelector, Func<TInner, TKey> innerKeySelector
  , Func<TOuter, TInner, Result> resultSelector, IEqualityComparer<TKey> comparer)
  {
    if (outer == null)
      throw new ArgumentException("outer");

    if (inner == null)
      throw new ArgumentException("inner");

    if (outerKeySelector == null)
      throw new ArgumentException("outerKeySelector");

    if (innerKeySelector == null)
      throw new ArgumentException("innerKeySelector");

    if (resultSelector == null)
      throw new ArgumentException("resultSelector");

    return LeftJoinImpl(outer, inner, outerKeySelector, innerKeySelector, resultSelector, comparer ?? EqualityComparer<TKey>.Default);
  }

  static IEnumerable<Result> LeftJoinImpl<TOuter, TInner, TKey, Result>(
      IEnumerable<TOuter> outer, IEnumerable<TInner> inner
      , Func<TOuter, TKey> outerKeySelector, Func<TInner, TKey> innerKeySelector
      , Func<TOuter, TInner, Result> resultSelector, IEqualityComparer<TKey> comparer)
  {
    var innerLookup = inner.ToLookup(innerKeySelector, comparer);

    foreach (var outerElment in outer)
    {
      var outerKey = outerKeySelector(outerElment);
      var innerElements = innerLookup[outerKey];

      if (innerElements.Any())
        foreach (var innerElement in innerElements)
          yield return resultSelector(outerElment, innerElement);
      else
        yield return resultSelector(outerElment, default(TInner));
     }
   }

The resultselector then has to take care of the null elements. Fx.

   static void Main(string[] args)
   {
     var inner = new[] { Tuple.Create(1, "1"), Tuple.Create(2, "2"), Tuple.Create(3, "3") };
     var outer = new[] { Tuple.Create(1, "11"), Tuple.Create(2, "22") };

     var res = outer.LeftJoin(inner, item => item.Item1, item => item.Item1, (it1, it2) =>
     new { Key = it1.Item1, V1 = it1.Item2, V2 = it2 != null ? it2.Item2 : default(string) });

     foreach (var item in res)
       Console.WriteLine(string.Format("{0}, {1}, {2}", item.Key, item.V1, item.V2));
   }

Answer 6

Take a look at this example. This query should work:

var leftFinal = from left in lefts
                join right in rights on left equals right.Left into leftRights
                from leftRight in leftRights.DefaultIfEmpty()
                select new { LeftId = left.Id, RightId = left.Key==leftRight.Key ? leftRight.Id : 0 };