Oracle SQL Count 或 Sum 以防万一查询？

Question

我在 StackOverflow 上研究了五个不同的线程，我觉得我已经接近了，但仍然没有雪茄。我的原始表格有几个部分，每个部分有几行，每行有几个位置。我希望能够查询一个结果，该结果列出了部分、行、位置、有多少位置高于该行（位置编号较低）以及有多少位置低于该行（位置编号更大）。

这是我所拥有的：

SELECT section, 
       row, 
       position AS P, 
       SUM(case when position < P then 1 else 0 end) AS Above, 
       SUM(case when position > P then 1 else 0 end) AS Below 
FROM OriginalTable 
GROUP BY section, row, position 
ORDER BY section, row, position

求和的情况没有将 P 识别为位置，当我将 P 更改为数字时，总和就差了。

Answer 1

一种可能的解决方案是相关子查询获取部分和行的较低/较高位置的count()。

SELECT ot1.section,
       ot1.row,
       ot1.position,
       (SELECT count(*)
               FROM originaltable ot2
               WHERE ot2.section = ot1.section
                     AND ot2.row = ot2.row
                     AND ot2.position > ot1.position) above,
       (SELECT count(*)
               FROM originaltable ot2
               WHERE ot2.section = ot1.section
                     AND ot2.row = ot2.row
                     AND ot2.position < ot1.position) below
      FROM originaltable ot1
      ORDER BY ot1.section,
               ot1.row,
               ot1.position;

另一个正在使用count()的窗口版本。

SELECT ot1.section,
       ot1.row,
       ot1.position,
       count(*) OVER (PARTITION BY ot1.section,
                                   ot1.row
                      ORDER BY ot1.position
                      RANGE BETWEEN CURRENT ROW
                                    AND UNBOUNDED FOLLOWING) - 1 above
       count(*) OVER (PARTITION BY ot1.section,
                                   ot1.row
                      ORDER BY ot1.position
                      RANGE BETWEEN UNBOUNDED PRECEDING
                                    AND CURRENT ROW) - 1 below
      FROM originaltable ot1
      ORDER BY ot1.section,
               ot1.row,
               ot1.position;

Answer 2

我只会使用row_number():

SELECT section, row, position AS P, 
       ROW_NUMBER() OVER (PARTITION BY section, row ORDER BY position ASC) - 1 as Above,
       ROW_NUMBER() OVER (PARTITION BY section, row ORDER BY position DESC) - 1 as Below
FROM OriginalTable
ORDER BY section, row, position;

如果你有重复，那么问题就有点棘手了（尽管GROUP BY 会解决这个问题）。您可能需要DENSE_RANK() 而不是ROW_NUMBER()。