什么更快：SUM over NULL 还是 over 0？

Question

我有这样的查询：

select sum(case when col1=@arg1 then value else null end) from t

在性能方面，使用 0 而不是 NULL 有区别吗？像这样：

select sum(case when col1=@arg1 then value else 0 end) from t

Answer 1

DECLARE @type CHAR(2) = 'U'

-- [Expr1042] = Scalar Operator(CASE WHEN [Expr1048]=(0) THEN NULL ELSE [Expr1049] END)

SELECT SUM(CASE WHEN [type] = @type THEN 1 END)
FROM sys.objects

-- [Expr1042] = Scalar Operator(CASE WHEN [Expr1048]=(0) THEN NULL ELSE [Expr1049] END)

SELECT SUM(CASE WHEN [type] = @type THEN 1 ELSE NULL END)
FROM sys.objects

-- [Expr1042] = Scalar Operator(CASE WHEN [Expr1048]=(0) THEN NULL ELSE [Expr1049] END)

SELECT SUM(CASE WHEN [type] = @type THEN 1 ELSE 0 END)
FROM sys.objects

结果：

Table 'sysschobjs'. Scan count 1, logical reads 1556, 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 = 7 ms.

Table 'sysschobjs'. Scan count 1, logical reads 1556, 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 = 6 ms.

Table 'sysschobjs'. Scan count 1, logical reads 1556, 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 = 7 ms.

所以...答案 - 相似（如果我们只谈论性能）

Answer 2

在下面的测试中，我始终发现 NULL 稍微快一些。

   SET STATISTICS TIME ON;
   DECLARE @i int = null; /*Or set to zero*/

   WITH 
    E1(N) AS 
    (
        SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL 
        SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL 
        SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1
    )                                       -- 1*10^1 or 10 rows
    , E2(N) AS (SELECT 1 FROM E1 a, E1 b)   -- 1*10^2 or 100 rows
    , E4(N) AS (SELECT 1 FROM E2 a, E2 b)   -- 1*10^4 or 10,000 rows
    , E8(N) AS (SELECT 1 FROM E4 a, E4 b)   -- 1*10^8 or 100,000,000 rows

    SELECT SUM(@i) FROM E8 
    OPTION (MAXDOP 1)

聚合 100,000,000 个值平均需要 608 毫秒。（即每次聚合 6 纳秒）。

NULL 人在

上花费的时间更多

sqllang.dll!CESRunTimeErrorSink::SetAggFnSkippedNull

大概设置了导致消息的标志

Warning: Null value is eliminated by an aggregate or other SET operation.

但总体上似乎更快（下面以毫秒为单位的经过时间）。

+---------+--------+------++-------+
|         |  NULL  |  0   || Diff  |
+---------+--------+------++-------+
| Trial 1 | 7027   | 7592 || 565   |
| Trial 2 | 6981   | 7743 || 762   |
| Trial 3 | 7451   | 8015 || 564   |
| Trial 4 | 6997   | 7591 || 594   |
| Trial 5 | 7018   | 7574 || 556   |
+---------+--------+------++-------+
| Avg     | 7094.8 | 7703 || 608.2 |
+---------+--------+------++-------+

当然，在这种情况下（所有输入都是 NULL），它们 return 不同的结果，如果您想互换处理这两者，则需要 ISNULL(SUM(@i),0)。

什么更快：SUM over NULL 还是 over 0？

What is faster: SUM over NULL or over 0?

sql-server

sql-server-2008-r2

query-performance

sql-server-2012