当前位置: 代码迷 >> SQL >> lin to sql grouping 再有糖衣炮弹
  详细解决方案

lin to sql grouping 再有糖衣炮弹

热度:227   发布时间:2016-05-05 11:01:01.0
lin to sql grouping 还有糖衣炮弹

在linq to sql中grouping很简单、很灵活,但是如果不注意则会中了Microsoft的糖衣炮弹。

本文使用的数据模型如下图:

1. 用linq to sql写一个最简单的group语句:
查询Students表并按ClassID分组,输出每个班级的学生数。下面的代码是糖衣版。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
static void Main(string[] args)
{
    using (var writer = new StreamWriter(WatchSqlPath, false, Encoding.UTF8))
    {
        using (DbAppDataContext db = new DbAppDataContext())
        {
            //打印sql
            db.Log = writer;
 
            //最简单的group by ,按ClassID对Students进行分组
            var query = from s in db.Students
                        group s by s.ClassID;
 
            foreach (var item in query)
            {
                //输出班级编号和班级人数
                Console.WriteLine("class id = {0} student count = {1}",item.Key,item.Count());
            }
        }
    }
    Console.ReadLine();
}

Linq to sql的写法很灵活,我们随时可以调用Count等一大堆方法,上例中我在输出时调用Count方法,导致了linq to sql真正执行的sql有一大坨,如下:

这个不能怪linq to sql,我们这样写它就必须得那样执行;只能我们自己使用时处处小心。上例中正确的写法应该是:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
static void Main(string[] args)
{
    using (var writer = new StreamWriter(WatchSqlPath, false, Encoding.UTF8))
    {
        using (DbAppDataContext db = new DbAppDataContext())
        {
            //打印sql
            db.Log = writer;
 
            //最简单的group by ,按ClassID对Students进行分组
            //var query = from s in db.Students
            //            group s by s.ClassID;
 
            var query = from s in db.Students
                        group s by s.ClassID into sg
                        select new {
                            ClassID = sg.Key,
                            Count = sg.Count()
                        };
 
            foreach (var item in query)
            {
                //输出班级编号和班级人数
                Console.WriteLine("class id = {0} student count = {1}", item.ClassID, item.Count);
            }
        }
    }
    Console.ReadLine();
}

这样执行时才是那个我们以前写t-sql时经常看到的那个group by语句。

2. 对分组聚合进行排序输出,我忘记了用let
请看下面代码,输出每个班级体重最大的同学,并要求最大体重得大于39,并按照体重大下,对分组结果进行排序。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
static void Main(string[] args)
{
    using (var writer = new StreamWriter(WatchSqlPath, false, Encoding.UTF8))
    {
        using (DbAppDataContext db = new DbAppDataContext())
        {
            //打印sql
            db.Log = writer;
 
            var query = from s in db.Students
                        group s by s.ClassID into gS
                        where gS.Max<Student>(s => s.WeightInKg) > 39
                        orderby gS.Max<Student>(s => s.WeightInKg) descending
                        select new
                        {
                            ClassID = gS.Key,
                            MaxWeight = gS.Max<Student>(s => s.WeightInKg)
                        };
 
            foreach (var item in query)
            {
                Console.WriteLine("class id = {0} student max weight = {1}", item.ClassID, item.MaxWeight);
            }
        }
    }
    Console.ReadLine();
}

上例中,在query变量声明时我用了三次gS.Max<Student>(s => s.WeightInKg),我们知道在t-sql中如果使用聚合值,必须得用表达式,在linq to sql中用会不会有问题呢,看下执行上述代码生成的sql吧:

1
2
3
4
5
6
7
8
9
10
SELECT [t1].[ClassID], [t1].[value3] AS [MaxWeight]
FROM (
    SELECT MAX([t0].[WeightInKg]) AS [value], MAX([t0].[WeightInKg]) AS [value2], MAX([t0].[WeightInKg]) AS [value3], [t0].[ClassID]
    FROM [dbo].[Student] AS [t0]
    GROUP BY [t0].[ClassID]
    ) AS [t1]
WHERE [t1].[value] > @p0
ORDER BY [t1].[value2] DESC
-- @p0: Input Float (Size = -1; Prec = 0; Scale = 0) [39]
-- Context: SqlProvider(Sql2005) Model: AttributedMetaModel Build: 4.0.30319.1

Linq to sql忠实的执行了我写的代码,我的query变量声明中用了三次Max,在t-sql的嵌套表中也有三个MAX对应,这个不是我想要的,我又错了,呼叫let,看看使用let之后linq to sql会怎么执行吧:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
static void Main(string[] args)
{
    using (var writer = new StreamWriter(WatchSqlPath, false, Encoding.UTF8))
    {
        using (DbAppDataContext db = new DbAppDataContext())
        {
            //打印sql
            db.Log = writer;
 
            // group by with having
            var query = from s in db.Students
                        group s by s.ClassID into gS
                        let mw = gS.Max<Student>(s => s.WeightInKg)
                        where mw > 39
                        select new
                        {
                            ClassID = gS.Key,
                            MaxWeight = mw
                        };
            foreach (var item in query)
            {
                Console.WriteLine("class id = {0} student max weight = {1}", item.ClassID, item.MaxWeight);
            }
        }
    }
    Console.ReadLine();
}

这次的sql语句比上一个好多了,但是还没有足够好:

1
2
3
4
5
6
7
8
9
SELECT [t1].[ClassID], [t1].[value] AS [MaxWeight]
FROM (
    SELECT MAX([t0].[WeightInKg]) AS [value], [t0].[ClassID]
    FROM [dbo].[Student] AS [t0]
    GROUP BY [t0].[ClassID]
    ) AS [t1]
WHERE [t1].[value] > @p0
-- @p0: Input Float (Size = -1; Prec = 0; Scale = 0) [39]
-- Context: SqlProvider(Sql2005) Model: AttributedMetaModel Build: 4.0.30319.1

还是多嵌套了一层没有必要的select。如何才能不嵌套这一层呢?如果你知道,请告诉我。

3. 使用linq to sql Group by多个字段:
这次我们先写一个标准的t-sql

1
2
3
4
SELECT ClassID,Hometown,count(*) Cn
FROM Student
GROUP BY ClassID,Hometown
ORDER BY ClassID,Hometown;

这个t-sql对Student表按照ClassID和Hometown两个字段进行分组,并输出每个班级中某个地方的学生数

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
static void Main(string[] args)
{
    using (var writer = new StreamWriter(WatchSqlPath, false, Encoding.UTF8))
    {
        using (DbAppDataContext db = new DbAppDataContext())
        {
            //打印sql
            db.Log = writer;
 
            // group by with having
            var query = from s in db.Students
                        group s by new { s.ClassID, s.Hometown } into gS
                        let cn = gS.Count<Student>()
                        select new
                        {
                            ClassID = gS.Key.ClassID,
                            Hometown = gS.Key.Hometown,
                            Count = cn
                        };
            foreach (var item in query)
            {
                Console.WriteLine("class id = {0} hometown {1} student count = {2}", item.ClassID, item.Hometown,item.Count);
            }
        }
    }
    Console.ReadLine();
}

这一次linq to sql给了我一个惊喜,它用的t-sql和我开始写的一样。

4. 在3的基础上加一点点需求,要求分组后的结果按照count排序,query的声明代码如下:

1
2
3
4
5
6
7
8
9
10
11
// group by with having
var query = from s in db.Students
            group s by new { s.ClassID, s.Hometown } into gS
            let cn = gS.Count<Student>()
            orderby cn descending
            select new
            {
                ClassID = gS.Key.ClassID,
                Hometown = gS.Key.Hometown,
                Count = cn
            };

这次link to sql使用的sql语句多了一层没有必要的select嵌套

1
2
3
4
5
6
7
8
SELECT [t1].[ClassID], [t1].[Hometown], [t1].[value] AS [Count]
FROM (
    SELECT COUNT(*) AS [value], [t0].[ClassID], [t0].[Hometown]
    FROM [dbo].[Student] AS [t0]
    GROUP BY [t0].[ClassID], [t0].[Hometown]
    ) AS [t1]
ORDER BY [t1].[value] DESC
-- Context: SqlProvider(Sql2005) Model: AttributedMetaModel Build: 4.0.30319.1

看来使用linq to sql少多总要付出点代价的,linq to sql给了我们查询的方便、灵活,也给了我们性能的陷阱。