将多个自连接转换为 window 函数或子查询 (SQL)

Question

我有一个包含多个自连接的查询：

SELECT t2.Name, t2.Date as date, t2.Value as value1, t3.Value as value2, t4.Value as value3
FROM table_t t1
    RIGHT JOIN #tempPredefinedIndicators t2
        ON t1.Name = t2.Name
    JOIN #tempPredefinedIndicators t3 
        ON t2.Name = t3.Name AND t2.Date = t3.Date
    JOIN #tempPredefinedIndicators t4 
        ON t2.Name = t4.Name AND t2.Date = t4.Date    
WHERE t1.ValueName = 'one' AND DATEDIFF(year, ISDATE(t1.Value), GETDATE()) >=10
AND (t2.ValueName = 'two' AND TRY_CONVERT(FLOAT, t2.Value) > 0.15)
AND (t3.ValueName = 'three' AND TRY_CONVERT(FLOAT, t3.Value) BETWEEN 0 AND 0.5)
AND (t4.ValueName = 'four' AND TRY_CONVERT(FLOAT, t4.Value) > 0.15)

这是一个示例 table，我用黄色标记了 where 条件边界内的记录：

这是预期的结果视图：

基本上，我用几个连续的条件过滤一个 table，并将结果显示为 table。

有没有办法，使用 window 函数或子查询将其转换得更快？

Answer 1

您可以通过删除类型转换、向经常查询的适当列添加索引，或者将 4-table 连接替换为 PIVOT 来提高性能。

SELECT * FROM (
  SELECT
    [name],
    [valuename],
    [value]
  FROM table_t
  WHERE 
    (valuename = 'one' and datediff(year, isdate(value), getdate()) > 10)
     or 
    (valuename = 'two' and TRY_CONVERT(FLOAT, Value) > 0.15)
     or 
    (ValueName = 'three' AND TRY_CONVERT(FLOAT, Value) BETWEEN 0 AND 0.5)
     or
    (ValueName = 'four' AND TRY_CONVERT(FLOAT, Value) > 0.15)
) a
PIVOT (
  MAX([value])
  FOR [valuename]
  IN (
    [one], [two], [three], [four]
  )
) AS PivotTable

通过此处的示例自学：https://www.sqlshack.com/dynamic-pivot-tables-in-sql-server/

您的查询在 sql-fiddle 中用了 30 毫秒到运行；运行 WHERE 子句中的枢轴 w/o 类型转换只用了 17 毫秒，所有 WHERE 子句转换运行只用了 20-22 毫秒。

将多个自连接转换为 window 函数或子查询 (SQL)

Convert multiple self JOINs to window function or subqueries (SQL)

sql

tsql

subquery

self-join

window-functions