从lambda表达式生成sql server查询的备选方案

本文关键字：查询 方案 server sql lambda 表达式 | 更新日期: 2023-09-27 18:15:56

我写了一个lambda表达式，它产生了预期的结果，但是它产生了一个绝对庞大的sql查询，它的性能很差。查看最底部的io/time统计信息。

是否有其他方法来实现下面的查询?

select distinct(searchterms) as SearchTerms, max(totalresults) FROM cmsSearchLog 
where totalresults != 0 and searchterms like 'de%' group by searchterms 
order by max(totalresults) desc
c#代码片段: 
// current lamda expression; has bad performance compared to above query
List<SearchTerm> existingSearchTerms1 = context.cmsSearchLogs.Where(oq =>
context.cmsSearchLogs.Where(q =>
q.SearchTerms.ToLower().Contains(terms.ToLower()) && q.TotalResults != 0)
.Select(s => s.SearchTerms)
.Distinct()
.Contains(oq.SearchTerms))
.Select(a => new { a.SearchTerms, a.TotalResults })
.GroupBy(gb => gb.SearchTerms)
.OrderByDescending(ob => ob.Max(m => m.TotalResults))
.Select(s => new SearchTerm()
    {
        SearchTerms = s.FirstOrDefault().SearchTerms,
        TotalResults = s.FirstOrDefault().TotalResults
    }
)
.ToList();
// get the suggestions back as a list of strings
List<string> suggestions = Enumerable.Range(0, 
  existingSearchTerms1.Count())
  .Select(x => existingSearchTerms1.ElementAt(x).SearchTerms).ToList();

这是保存查询

结果的私有类
private class SearchTerm
{
    public string SearchTerms { get; set; }
    public int TotalResults { get; set; }
}

lambda表达式生成的sql是巨大的:

SELECT 
[Project13].[C2] AS [C1], 
[Project13].[C3] AS [C2], 
[Project13].[C4] AS [C3]
FROM ( SELECT 
    [Project12].[C1] AS [C1], 
    1 AS [C2], 
    [Project12].[C2] AS [C3], 
    [Project12].[C3] AS [C4]
    FROM ( SELECT 
        [Project8].[C1] AS [C1], 
        [Project8].[C2] AS [C2], 
        (SELECT TOP (1) 
            [Extent5].[TotalResults] AS [TotalResults]
            FROM [dbo].[cmsSearchLog] AS [Extent5]
            WHERE ( EXISTS (SELECT 1 AS [C1]                    
               FROM ( SELECT DISTINCT 
            [Extent6].[SearchTerms] AS [SearchTerms]
            FROM [dbo].[cmsSearchLog] AS [Extent6]
            WHERE (( CAST(CHARINDEX(LOWER('dew'), 
                             LOWER([Extent6].[SearchTerms])) AS int)) > 0) 
                             AND (0 <> [Extent6].[TotalResults])
                )  AS [Distinct3]
            WHERE [Distinct3].[SearchTerms] = [Extent5].[SearchTerms]
            )) AND ([Project8].[SearchTerms] = [Extent5].[SearchTerms])) 
                                AS [C3]
        FROM ( SELECT 
           [Project7].[C1] AS [C1], 
           [Project7].[SearchTerms] AS [SearchTerms], 
           [Project7].[C2] AS [C2]
           FROM ( SELECT 
              [Project3].[C1] AS [C1], 
              [Project3].[SearchTerms] AS [SearchTerms], 
              (SELECT TOP (1) 
              [Extent3].[SearchTerms] AS [SearchTerms]
              FROM [dbo].[cmsSearchLog] AS [Extent3]
              WHERE ( EXISTS (SELECT 1 AS [C1] FROM ( SELECT DISTINCT 
            [Extent4].[SearchTerms] AS [SearchTerms]
            FROM [dbo].[cmsSearchLog] AS [Extent4]
            WHERE (( CAST(CHARINDEX(LOWER('dew'), 
                             LOWER([Extent4].[SearchTerms])) AS int)) > 0) 
                             AND (0 <> [Extent4].[TotalResults]))  AS [Distinct2] 
           WHERE [Distinct2].[SearchTerms] = [Extent3].[SearchTerms]
               )) AND ([Project3].[SearchTerms] = [Extent3].[SearchTerms])) AS [C2]
                FROM ( SELECT 
                  [GroupBy1].[A1] AS [C1], 
                  [GroupBy1].[K1] AS [SearchTerms]
                  FROM ( SELECT 
                   [Extent1].[SearchTerms] AS [K1], 
                   MAX([Extent1].[TotalResults]) AS [A1]
                   FROM [dbo].[cmsSearchLog] AS [Extent1]
                   WHERE EXISTS (SELECT 1 AS [C1]
                FROM ( SELECT DISTINCT [Extent2].[SearchTerms]
                  AS [SearchTerms] FROM [dbo].[cmsSearchLog] AS [Extent2]
                        WHERE (( CAST(CHARINDEX(LOWER('dew'),
                                      LOWER([Extent2].[SearchTerms])) AS int)) > 0)
                                       AND (0 <> [Extent2].[TotalResults]))  AS [Distinct1]
                                       WHERE [Distinct1].[SearchTerms] = [Extent1].[SearchTerms])
                 GROUP BY [Extent1].[SearchTerms])  AS [GroupBy1]
                )  AS [Project3]
            )  AS [Project7]
        )  AS [Project8]
    )  AS [Project12]
)  AS [Project13]
ORDER BY [Project13].[C1] ASC

我在io和时间统计打开的情况下执行了两个查询，结果如下:(注意:lambda生成的查询是第一，我手写的查询是第二)所以这证实了我的怀疑，即生成的查询与我实际想要的查询相比，执行得很糟糕。

(8 row(s) affected)
Table 'cmsSearchLog'. Scan count 6, logical reads 106, physical reads 0, 
read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 1 ms.
(7 row(s) affected)
Table 'cmsSearchLog'. Scan count 1, logical reads 5, physical reads 0, 
read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 0 ms.

从lambda表达式生成sql server查询的备选方案

尝试用这个查询代替当前的LINQ查询:

var query = from x in context.cmsSearchLog
            where totalresults != 0 &&
                  searchterms.BeginsWith("de")
            group x by x.searchterms into terms
            select new {
                           SearchTerms = terms.Key(),
                           TotalResults = terms.Max(t => t.totalresults)
                       };

我还没有测试过，但我认为它会生成一个相当有效的查询并返回所需的结果。

LINQ转换(无论是LINQ到SQL，实体框架等)是关于高效开发的。它允许(在理论上)更具可读性和可维护性的代码，并且减少了由于fat-fingering等导致的运行时数据库错误的可能性。LINQ是关于性能的而不是。LINQ通常提供"足够好"的性能，但它永远无法击败一些更接近金属的东西，比如手工编码的查询或存储过程。

也就是说，您的查询返回不同的行计数，因此其中一个(或两个)是错误的;第一个查询产生8行，而第二个查询产生7行。您无法很好地比较提供不同结果的查询!

对于复杂或性能密集型查询，不要觉得不能创建视图或用户定义函数并映射到它们。在这种情况下，您甚至可以使用存储过程并映射到该过程。

为什么不让数据库处理此查询的工作并将结果直接转储到SearchTerm类中呢?如果需要查找特定的术语，可以对过程进行参数化。在您提供的示例中，您可以通过索引searchterms列进一步提高性能，因为where子句中的通配符引用列值文本的末尾部分。此外，由于是根据搜索词进行分组，因此不需要在该列上调用distinct(这可能会提高性能，也可能不会提高性能，这取决于系统选择执行的查询计划)。

首先，您需要知道lambda表达式方法不适用于这种查询。但是，如果您不介意这样做，可以创建一个使用:

的视图。
select distinct searchTerm, max(totalresults) 
from cmsSearchLog 
group by searchterms 
order by max(totalresults) desc

然后使用lambda表达式执行过滤部分

相关文章：