Question

我写了下面的查询，里面有很多＆＃39; AND＆＃39;运营商，我想知道如何优化以下查询的性能[我可以删除一些＆＃39; AND＆＃39;算]

    SELECT I.date,
              K.somcolumn,
              L.somcolumn,
              D.somcolumn
         FROM Table1 I,
              Table2 K,
              Table3 L,
              Table4 D
        WHERE I._ID = K._ID
              AND K.ID = L._ID
              AND L._ID = I._ID
              AND I._CODE = L._CODE
              AND K.ID = D._ID(+)
              AND L._ID IN ( SELECT _id
                                       FROM I
                                      WHERE UPPER (someflag) = 'TRUE'
                                   GROUP BY _id
                                     HAVING COUNT (*) > 1)
              AND L._ID IN ( SELECT _id
                                       FROM I
                                      WHERE UPPER (code) = 'OPEN'
                                   GROUP BY _id
                                     HAVING COUNT (*) > 1)
     ORDER BY I._ID, I._CODE;

Answer 1

根据我的判断，您无法合并任何条件，但您可以使用标准AND语法改进查询并减少JOIN运算符的数量：

SELECT I.date,
  K.somcolumn,
  L.somcolumn,
  D.somcolumn
FROM Table1 I
INNER JOIN Table2 K ON I._ID = K._ID
INNER JOIN Table3 L ON K.ID = L._ID
LEFT JOIN Table4 D ON K.ID = D._ID
WHERE L._ID IN ( SELECT _id
   FROM I
   WHERE UPPER (someflag) = 'TRUE'
   GROUP BY _id
   HAVING COUNT (*) > 1)
 AND L._ID IN ( SELECT _id
   FROM I
   WHERE UPPER (code) = 'OPEN'
   GROUP BY _id
   HAVING COUNT (*) > 1)
ORDER BY I._ID, I._CODE;

以此为基础，如果您加入子查询条件而不是使用相关子查询，可能会获得优化提升。没有保证，但这样的事情可能会有所帮助：

SELECT I.date,
  K.somcolumn,
  L.somcolumn,
  D.somcolumn
FROM Table1 I
INNER JOIN Table2 K ON I._ID = K._ID
INNER JOIN Table3 L ON K.ID = L._ID
LEFT JOIN Table4 D ON K.ID = D._ID
INNER JOIN (
    SELECT _id
    FROM I
    WHERE UPPER (someflag) = 'TRUE'
    GROUP BY _id
    HAVING COUNT (*) > 1
  ) someflagtrue ON L._ID = someflagtrue._id
INNER JOIN (
    SELECT _id
     FROM I
     WHERE UPPER (code) = 'OPEN'
     GROUP BY _id
     HAVING COUNT (*) > 1
  ) codeopen ON L._ID = codeopen._id
ORDER BY I._ID, I._CODE;

Answer 2

您可以将两个子查询替换为一个。

旧子查询：

  SELECT _id
    FROM I
   WHERE UPPER (someflag) = 'TRUE'
GROUP BY _id
  HAVING COUNT (*) > 1)

  SELECT _id
    FROM I
   WHERE UPPER (code) = 'OPEN'
GROUP BY _id
  HAVING COUNT (*) > 1)

新子查询：

SELECT _ID
  FROM I
 GROUP BY _ID
HAVING COUNT(CASE WHEN UPPER(SOMEFLAG) = 'TRUE' THEN 1 ELSE 0 END) > 0
   AND COUNT(CASE WHEN UPPER(CODE) = 'OPEN' THEN 1 ELSE 0 END) > 0

在大多数情况下，这应该至少快一点，因为它可能会减少全表扫描和连接的数量。但是很难说你的系统是否会更快，因为优化器有很多可能的选择。

在清理查询后，性能调整的下一步是生成解释计划。运行explain plan for select ...;，然后运行select * from table(dbms_xplan.display);这将显示查询是如何执行的，这可能会提示您什么是缓慢的，哪些可以改进。如果您需要更多帮助，请将解释计划的完整输出添加到您的问题中。它还可以帮助添加有关相关表中的行数，返回的行数以及索引的信息。

ORACLE QUERY - 优化

2 个答案: