使用LINQ的IQueryable左外连接的扩展方法

N R*_*ing 21 .net c# linq

我试图用返回类型实现左外连接扩展方法IQueryable.

我写的功能如下

public static IQueryable<TResult> LeftOuterJoin2<TOuter, TInner, TKey, TResult>(
        this IQueryable<TOuter> outer,
        IQueryable<TInner> inner,
        Func<TOuter, TKey> outerKeySelector,
        Func<TInner, TKey> innerKeySelector,
        Func<TOuter, TInner, TResult> resultSelector)
{
        return
          from outerItem in outer
          join innerItem in inner on outerKeySelector(outerItem) 
            equals innerKeySelector(innerItem) into joinedData
          from r in joinedData.DefaultIfEmpty()
          select resultSelector(outerItem, r);
}
Run Code Online (Sandbox Code Playgroud)

它无法生成查询.原因可能是:我用过Func<>而不是Expression<>.我也尝试过Expression<>.它给我一个在线错误outerKeySelector(outerItem),这outerKeySelector是一个被用作方法的变量

我发现了一些关于SO(比如这里)和CodeProjects的讨论,但那些适用于IEnumerable不适合的类型IQueryable.

Ton*_*ony 33

介绍

这个问题非常有趣.问题是Funcs是委托,而表达式是,它们是完全不同的结构.当您使用当前的扩展实现时,它使用循环并在每个元素的每个步骤上执行选择器,并且它运行良好.但是当我们谈论实体框架和LINQ时,我们需要树遍历来将其转换为SQL查询.所以它比Funcs更"难"(但我还是喜欢Expressions)并且下面描述了一些问题.

当你想做左外连接时你可以使用这样的东西(取自这里:如何在JOIN扩展方法中实现左连接)

var leftJoin = p.Person.Where(n => n.FirstName.Contains("a"))
                   .GroupJoin(p.PersonInfo, 
                              n => n.PersonId,
                              m => m.PersonId,
                              (n, ms) => new { n, ms = ms.DefaultIfEmpty() })
                   .SelectMany(z => z.ms.Select(m => new { n = z.n, m ));
Run Code Online (Sandbox Code Playgroud)

这是好的,但它不是我们需要的扩展方法.我想你需要这样的东西:

using (var db = new Database1Entities("..."))
{
     var my = db.A.LeftOuterJoin2(db.B, a => a.Id, b => b.IdA, 
         (a, b) => new { a, b, hello = "Hello World!" });
     // other actions ...
}
Run Code Online (Sandbox Code Playgroud)

创建此类扩展有许多难点:

  • 手动创建复杂的树,编译器将无法帮助我们
  • 需要反思的类似的方法Where,Select
  • 匿名类型(!!我们需要codegen吗?我希望没有)

脚步

考虑2个简单的表:( A列:Id,文本)和B(列Id,IdA,文本).

外连接可以分3步实现:

// group join as usual + use DefaultIfEmpty
var q1 = Queryable.GroupJoin(db.A, db.B, a => a.Id, b => b.IdA, 
                              (a, b) => new { a, groupB = b.DefaultIfEmpty() });

// regroup data to associated list a -> b, it is usable already, but it's 
// impossible to use resultSelector on this stage, 
// beacuse of type difference (quite deep problem: some anonymous type != TOuter)
var q2 = Queryable.SelectMany(q1, x => x.groupB, (a, b) => new { a.a, b });

// second regroup to get the right types
var q3 = Queryable.SelectMany(db.A, 
                               a => q2.Where(x => x.a == a).Select(x => x.b), 
                               (a, b) => new {a, b});
Run Code Online (Sandbox Code Playgroud)

好吧,我不是一个好的出纳员,这里是我的代码(对不起,我无法更好地格式化它,但它有效!):

public static IQueryable<TResult> LeftOuterJoin2<TOuter, TInner, TKey, TResult>(
        this IQueryable<TOuter> outer,
        IQueryable<TInner> inner,
        Expression<Func<TOuter, TKey>> outerKeySelector,
        Expression<Func<TInner, TKey>> innerKeySelector,
        Expression<Func<TOuter, TInner, TResult>> resultSelector)
    {

        // generic methods
        var selectManies = typeof(Queryable).GetMethods()
            .Where(x => x.Name == "SelectMany" && x.GetParameters().Length == 3)
            .OrderBy(x=>x.ToString().Length)
            .ToList();
        var selectMany = selectManies.First();
        var select = typeof(Queryable).GetMethods().First(x => x.Name == "Select" && x.GetParameters().Length == 2);
        var where = typeof(Queryable).GetMethods().First(x => x.Name == "Where" && x.GetParameters().Length == 2);
        var groupJoin = typeof(Queryable).GetMethods().First(x => x.Name == "GroupJoin" && x.GetParameters().Length == 5);
        var defaultIfEmpty = typeof(Queryable).GetMethods().First(x => x.Name == "DefaultIfEmpty" && x.GetParameters().Length == 1);

        // need anonymous type here or let's use Tuple
        // prepares for:
        // var q2 = Queryable.GroupJoin(db.A, db.B, a => a.Id, b => b.IdA, (a, b) => new { a, groupB = b.DefaultIfEmpty() });
        var tuple = typeof(Tuple<,>).MakeGenericType(
            typeof(TOuter),
            typeof(IQueryable<>).MakeGenericType(
                typeof(TInner)
                )
            );
        var paramOuter = Expression.Parameter(typeof(TOuter));
        var paramInner = Expression.Parameter(typeof(IEnumerable<TInner>));
        var groupJoinExpression = Expression.Call(
            null,
            groupJoin.MakeGenericMethod(typeof (TOuter), typeof (TInner), typeof (TKey), tuple),
            new Expression[]
                {
                    Expression.Constant(outer),
                    Expression.Constant(inner),
                    outerKeySelector,
                    innerKeySelector,
                    Expression.Lambda(
                        Expression.New(
                            tuple.GetConstructor(tuple.GetGenericArguments()),
                            new Expression[]
                                {
                                    paramOuter,
                                    Expression.Call(
                                        null,
                                        defaultIfEmpty.MakeGenericMethod(typeof (TInner)),
                                        new Expression[]
                                            {
                                                Expression.Convert(paramInner, typeof (IQueryable<TInner>))
                                            }
                                )
                                },
                            tuple.GetProperties()
                            ),
                        new[] {paramOuter, paramInner}
                )
                }
            );

        // prepares for:
        // var q3 = Queryable.SelectMany(q2, x => x.groupB, (a, b) => new { a.a, b });
        var tuple2 = typeof (Tuple<,>).MakeGenericType(typeof (TOuter), typeof (TInner));
        var paramTuple2 = Expression.Parameter(tuple);
        var paramInner2 = Expression.Parameter(typeof(TInner));
        var paramGroup = Expression.Parameter(tuple);
        var selectMany1Result = Expression.Call(
            null,
            selectMany.MakeGenericMethod(tuple, typeof (TInner), tuple2),
            new Expression[]
                {
                    groupJoinExpression,
                    Expression.Lambda(
                        Expression.Convert(Expression.MakeMemberAccess(paramGroup, tuple.GetProperty("Item2")),
                                           typeof (IEnumerable<TInner>)),
                        paramGroup
                ),
                    Expression.Lambda(
                        Expression.New(
                            tuple2.GetConstructor(tuple2.GetGenericArguments()),
                            new Expression[]
                                {
                                    Expression.MakeMemberAccess(paramTuple2, paramTuple2.Type.GetProperty("Item1")),
                                    paramInner2
                                },
                            tuple2.GetProperties()
                            ),
                        new[]
                            {
                                paramTuple2,
                                paramInner2
                            }
                )
                }
            );

        // prepares for final step, combine all expressinos together and invoke:
        // var q4 = Queryable.SelectMany(db.A, a => q3.Where(x => x.a == a).Select(x => x.b), (a, b) => new { a, b });
        var paramTuple3 = Expression.Parameter(tuple2);
        var paramTuple4 = Expression.Parameter(tuple2);
        var paramOuter3 = Expression.Parameter(typeof (TOuter));
        var selectManyResult2 = selectMany
            .MakeGenericMethod(
                typeof(TOuter),
                typeof(TInner),
                typeof(TResult)
            )
            .Invoke(
                null,
                new object[]
                    {
                        outer,
                        Expression.Lambda(
                            Expression.Convert(
                                Expression.Call(
                                    null,
                                    select.MakeGenericMethod(tuple2, typeof(TInner)),
                                    new Expression[]
                                        {
                                            Expression.Call(
                                                null,
                                                where.MakeGenericMethod(tuple2),
                                                new Expression[]
                                                    {
                                                        selectMany1Result,
                                                        Expression.Lambda( 
                                                            Expression.Equal(
                                                                paramOuter3,
                                                                Expression.MakeMemberAccess(paramTuple4, paramTuple4.Type.GetProperty("Item1"))
                                                            ),
                                                            paramTuple4
                                                        )
                                                    }
                                            ),
                                            Expression.Lambda(
                                                Expression.MakeMemberAccess(paramTuple3, paramTuple3.Type.GetProperty("Item2")),
                                                paramTuple3
                                            )
                                        }
                                ), 
                                typeof(IEnumerable<TInner>)
                            ),
                            paramOuter3
                        ),
                        resultSelector
                    }
            );

        return (IQueryable<TResult>)selectManyResult2;
    }
Run Code Online (Sandbox Code Playgroud)

用法

再次使用:

db.A.LeftOuterJoin2(db.B, a => a.Id, b => b.IdA, 
       (a, b) => new { a, b, hello = "Hello World!" });
Run Code Online (Sandbox Code Playgroud)

看看这个你可以想一下这个sql查询是什么?这可能是巨大的.你猜怎么着?它很小:

SELECT 
1 AS [C1], 
[Extent1].[Id] AS [Id], 
[Extent1].[Text] AS [Text], 
[Join1].[Id1] AS [Id1], 
[Join1].[IdA] AS [IdA], 
[Join1].[Text2] AS [Text2], 
N'Hello World!' AS [C2]
FROM  [A] AS [Extent1]
INNER JOIN  (SELECT [Extent2].[Id] AS [Id2], [Extent2].[Text] AS [Text], [Extent3].[Id]    AS [Id1], [Extent3].[IdA] AS [IdA], [Extent3].[Text2] AS [Text2]
    FROM  [A] AS [Extent2]
    LEFT OUTER JOIN [B] AS [Extent3] ON [Extent2].[Id] = [Extent3].[IdA] ) AS [Join1] ON [Extent1].[Id] = [Join1].[Id2]
Run Code Online (Sandbox Code Playgroud)

希望能帮助到你.

  • 一段时间以来我看到的最英勇的答案。 (3认同)

Jan*_*gen 8

接受的答案是解释左外连接背后复杂性的一个很好的开始.

我发现了三个相当严重的问题,特别是在采用这种扩展方法并在更复杂的查询中使用它时(将多个左外连接与正常连接链接,然后汇总/ max/count/...)在将所选答案复制到您的生产环境,请继续阅读.

考虑链接的SO帖子中的原始示例,它代表LINQ中完成的任何左外连接:

var leftJoin = p.Person.Where(n => n.FirstName.Contains("a"))
                   .GroupJoin(p.PersonInfo, 
                              n => n.PersonId,
                              m => m.PersonId,
                              (n, ms) => new { n, ms = ms })
                   .SelectMany(z => z.ms.DefaultIfEmpty(), (n, m) => new { n = n, m ));
Run Code Online (Sandbox Code Playgroud)
  • 使用Tuple有效,但当这被用作更复杂查询的一部分时,EF失败(不能使用构造函数).要解决这个问题,您需要动态生成新的匿名类(搜索堆栈溢出)或使用无构造函数类型.我创造了这个

    internal class KeyValuePairHolder<T1, T2>
    {
        public T1 Item1 { get; set; }
        public T2 Item2 { get; set; }
    }
    
    Run Code Online (Sandbox Code Playgroud)
  • 使用"Queryable.DefaultIfEmpty"方法.在原始方法和GroupJoin方法中,编译器选择的正确方法是"Enumerable.DefaultIfEmpty"方法.这对简单查询没有影响,但请注意接受的答案如何有一堆转换(在IQueryable和IEnumerable之间).这些演员也会在更复杂的查询中引起问题. 可以在Expression中使用"Enumerable.DefaultIfEmpty"方法,EF知道不执行它,而是将其转换为连接.

  • 最后,这是一个更大的问题:有两个选择完成,而原始只做一个选择.你可以在代码注释中读取原因(因为类型不同(非常深的问题:一些匿名类型!= TOuter))并在SQL中看到它(从内连接中选择(左外连接b)) 这里的问题原始SelectMany方法采用在Join方法中创建的对象类型:TOuter的KeyValuePairHolder和Tinner的IEnumerable作为它的第一个参数,但传递的resultSelector表达式采用一个简单的TOUter作为它的第一个参数.您可以使用ExpressionVisitor重写传递给正确形式的表达式.

    internal class ResultSelectorRewriter<TOuter, TInner, TResult> : ExpressionVisitor
    {
        private Expression<Func<TOuter, TInner, TResult>> resultSelector;
        public Expression<Func<KeyValuePairHolder<TOuter, IEnumerable<TInner>>, TInner, TResult>> CombinedExpression { get; private set; }
    
        private ParameterExpression OldTOuterParamExpression;
        private ParameterExpression OldTInnerParamExpression;
        private ParameterExpression NewTOuterParamExpression;
        private ParameterExpression NewTInnerParamExpression;
    
    
        public ResultSelectorRewriter(Expression<Func<TOuter, TInner, TResult>> resultSelector)
        {
            this.resultSelector = resultSelector;
            this.OldTOuterParamExpression = resultSelector.Parameters[0];
            this.OldTInnerParamExpression = resultSelector.Parameters[1];
    
            this.NewTOuterParamExpression = Expression.Parameter(typeof(KeyValuePairHolder<TOuter, IEnumerable<TInner>>));
            this.NewTInnerParamExpression = Expression.Parameter(typeof(TInner));
    
            var newBody = this.Visit(this.resultSelector.Body);
            var combinedExpression = Expression.Lambda(newBody, new ParameterExpression[] { this.NewTOuterParamExpression, this.NewTInnerParamExpression });
            this.CombinedExpression = (Expression<Func<KeyValuePairHolder<TOuter, IEnumerable<TInner>>, TInner, TResult>>)combinedExpression;
        }
    
    
        protected override Expression VisitParameter(ParameterExpression node)
        {
            if (node == this.OldTInnerParamExpression)
                return this.NewTInnerParamExpression;
            else if (node == this.OldTOuterParamExpression)
                return Expression.PropertyOrField(this.NewTOuterParamExpression, "Item1");
            else
                throw new InvalidOperationException("What is this sorcery?", new InvalidOperationException("Did not expect a parameter: " + node));
    
        } 
    }
    
    Run Code Online (Sandbox Code Playgroud)

使用表达式visitor和KeyValuePairHolder来避免使用元组,我在下面选择的答案的更新版本修复了三个问题,更短,并产生更短的SQL:

 internal class QueryReflectionMethods
    {
        internal static System.Reflection.MethodInfo Enumerable_Select = typeof(Enumerable).GetMethods().First(x => x.Name == "Select" && x.GetParameters().Length == 2);
        internal static System.Reflection.MethodInfo Enumerable_DefaultIfEmpty = typeof(Enumerable).GetMethods().First(x => x.Name == "DefaultIfEmpty" && x.GetParameters().Length == 1);

        internal static System.Reflection.MethodInfo Queryable_SelectMany = typeof(Queryable).GetMethods().Where(x => x.Name == "SelectMany" && x.GetParameters().Length == 3).OrderBy(x => x.ToString().Length).First();
        internal static System.Reflection.MethodInfo Queryable_Where = typeof(Queryable).GetMethods().First(x => x.Name == "Where" && x.GetParameters().Length == 2);
        internal static System.Reflection.MethodInfo Queryable_GroupJoin = typeof(Queryable).GetMethods().First(x => x.Name == "GroupJoin" && x.GetParameters().Length == 5);
        internal static System.Reflection.MethodInfo Queryable_Join = typeof(Queryable).GetMethods(System.Reflection.BindingFlags.Static | System.Reflection.BindingFlags.Public).First(c => c.Name == "Join");
        internal static System.Reflection.MethodInfo Queryable_Select = typeof(Queryable).GetMethods().First(x => x.Name == "Select" && x.GetParameters().Length == 2);



        public static IQueryable<TResult> CreateLeftOuterJoin<TOuter, TInner, TKey, TResult>(
                   IQueryable<TOuter> outer,
                   IQueryable<TInner> inner,
                   Expression<Func<TOuter, TKey>> outerKeySelector,
                   Expression<Func<TInner, TKey>> innerKeySelector,
                   Expression<Func<TOuter, TInner, TResult>> resultSelector)
        { 

            var keyValuePairHolderWithGroup = typeof(KeyValuePairHolder<,>).MakeGenericType(
                typeof(TOuter),
                typeof(IEnumerable<>).MakeGenericType(
                    typeof(TInner)
                    )
                );
            var paramOuter = Expression.Parameter(typeof(TOuter));
            var paramInner = Expression.Parameter(typeof(IEnumerable<TInner>));
            var groupJoin =
                Queryable_GroupJoin.MakeGenericMethod(typeof(TOuter), typeof(TInner), typeof(TKey), keyValuePairHolderWithGroup)
                .Invoke(
                    "ThisArgumentIsIgnoredForStaticMethods",
                    new object[]{
                    outer,
                    inner,
                    outerKeySelector,
                    innerKeySelector,
                    Expression.Lambda(
                        Expression.MemberInit(
                            Expression.New(keyValuePairHolderWithGroup), 
                            Expression.Bind(
                                keyValuePairHolderWithGroup.GetMember("Item1").Single(),  
                                paramOuter
                                ), 
                            Expression.Bind(
                                keyValuePairHolderWithGroup.GetMember("Item2").Single(), 
                                paramInner
                                )
                            ),
                        paramOuter, 
                        paramInner
                        )
                    }
                );


            var paramGroup = Expression.Parameter(keyValuePairHolderWithGroup);
            Expression collectionSelector = Expression.Lambda(                    
                            Expression.Call(
                                    null,
                                    Enumerable_DefaultIfEmpty.MakeGenericMethod(typeof(TInner)),
                                    Expression.MakeMemberAccess(paramGroup, keyValuePairHolderWithGroup.GetProperty("Item2"))) 
                            ,
                            paramGroup
                        );

            Expression newResultSelector = new ResultSelectorRewriter<TOuter, TInner, TResult>(resultSelector).CombinedExpression;


            var selectMany1Result =
                Queryable_SelectMany.MakeGenericMethod(keyValuePairHolderWithGroup, typeof(TInner), typeof(TResult))
                .Invoke(
                    "ThisArgumentIsIgnoredForStaticMethods", new object[]{
                        groupJoin,
                        collectionSelector,
                        newResultSelector
                    }
                );
            return (IQueryable<TResult>)selectMany1Result;
        }
    }
Run Code Online (Sandbox Code Playgroud)


Rad*_*duV 5

如前面的答案所述,当您希望将IQueryable转换为SQL时,需要使用Expression而不是Func,因此必须走Expression Tree路线。

但是,这是一种无需您自己构建表达式树即可获得相同结果的方法。诀窍是,您需要引用LinqKit(可通过NuGet获取)并在查询上调用AsExpandable()。这将负责构建基础的表达式树(请参阅此处)。

下面的示例将GroupJoinSelectManyDefaultIfEmpty()方法一起使用:

    public static IQueryable<TResult> LeftOuterJoin<TOuter, TInner, TKey, TResult>(
        this IQueryable<TOuter> outer,
        IQueryable<TInner> inner,
        Expression<Func<TOuter, TKey>> outerKeySelector,
        Expression<Func<TInner, TKey>> innerKeySelector,
        Expression<Func<TOuter, TInner, TResult>> resultSelector)
    {
        return outer
            .AsExpandable()// Tell LinqKit to convert everything into an expression tree.
            .GroupJoin(
                inner,
                outerKeySelector,
                innerKeySelector,
                (outerItem, innerItems) => new { outerItem, innerItems })
            .SelectMany(
                joinResult => joinResult.innerItems.DefaultIfEmpty(),
                (joinResult, innerItem) => 
                    resultSelector.Invoke(joinResult.outerItem, innerItem));
    }
Run Code Online (Sandbox Code Playgroud)

样本数据

假设我们具有以下EF实体,并且usersaddress变量是对基础DbSet的访问:

public class User
{
    public int Id { get; set; }
    public string FirstName { get; set; }
    public string LastName { get; set; }
}

public class UserAddress
{
    public int UserId { get; set; }
    public string LastName { get; set; }
    public string Street { get; set; }
}

IQueryable<User> users;
IQueryable<UserAddress> addresses;
Run Code Online (Sandbox Code Playgroud)

用法1

让我们按用户ID加入:

var result = users.LeftOuterJoin(
            addresses,
            user => user.Id,
            address => address.UserId,
            (user, address) => new { user.Id, address.Street });
Run Code Online (Sandbox Code Playgroud)

转换为(使用LinqPad):

SELECT 
[Extent1].[Id] AS [Id],     
[Extent2].[Street] AS [Street]
FROM  [dbo].[Users] AS [Extent1]
LEFT OUTER JOIN [dbo].[UserAddresses] AS [Extent2] 
ON [Extent1].[Id] = [Extent2].[UserId]
Run Code Online (Sandbox Code Playgroud)

用法2

现在,让我们使用匿名类型作为键加入多个属性:

var result = users.LeftOuterJoin(
            addresses,
            user => new { user.Id, user.LastName },
            address => new { Id = address.UserId, address.LastName },
            (user, address) => new { user.Id, address.Street });
Run Code Online (Sandbox Code Playgroud)

请注意,匿名类型属性必须具有相同的名称,否则会出现语法错误。

这就是为什么我们拥有Id = address.UserId而不是address.UserId的原因

这将被翻译为:

SELECT 
[Extent1].[Id] AS [Id],     
[Extent2].[Street] AS [Street]
FROM  [dbo].[Users] AS [Extent1]
LEFT OUTER JOIN [dbo].[UserAddresses] AS [Extent2] 
ON ([Extent1].[Id] = [Extent2].[UserId]) AND ([Extent1].[LastName] = [Extent2].[LastName])
Run Code Online (Sandbox Code Playgroud)