我有一个名为RPT的表，其中包含以下列：
CODENUMBER AXSUBGROUP ITEMSTRUCTURE UNITSSOLD DOLLARS 我需要根据TOP 20 AXSUBGROUPS以及UNITSSOLD和UNITSSOLD值返回DOLLARS用户。另外，我需要返回属于前20名ITEMSTRUCTURES的{{1}}（由codenumber加入），以及他们的AXSUBGROUPS和UNITSSOLD。
由于我需要按DOLLARS进行分组，然后AXSUBGROUP进行分组，但ITEMSTRUCTURE与ITEMSTRUCTURE aXSUBGROUP相关联，我不确定要查询此方法。 {1}}。
我很感激您能给予的任何帮助。
谢谢。
EDIT 为了更好地说明我的需求：
...
这是表格中的示例数据（抱歉 - 无法弄清楚如何格式化表格......）
CODENUMBER... 如果它有帮助，最初这是2个表 - RPT（CodeNumber，UnitsSold，Dollars）和FLEX（CodeNumber，AxSubGroup，ItemStructure）。
答案 0 :(得分：3)
看起来好像是ROW_NUMBER救援！
SELECT CODENUMBER
, AXSUBGROUP
, ITEMSTRUCTURE
, UNITSSOLD
, DOLLARS
FROM
(
SELECT CODENUMBER
, AXSUBGROUP
, ITEMSTRUCTURE
, UNITSSOLD
, DOLLARS
, ROW_NUMBER() OVER (PARTITION BY AXSUBGROUP ORDER BY UNITSSOLD DESC) AS r
FROM RPT
) a
WHERE r <= 20
答案 1 :(得分：2)
SELECT
RPT.CodeNumber
, RPT.AxSubgroup
, RPT.ItemStructure
, RPT.UnitsSold
, RPT.Dollars
, grp.TotalUnitsSold
FROM
RPT
JOIN
( SELECT TOP 20
AxSubgroup
, SUM(UnitsSold) AS TotalUnitsSold
FROM
RPT
GROUP BY
AxSubgroup
ORDER BY
TotalUnitsSold DESC
) AS grp
ON
grp.AxSubgroup = RPT.AxSubgroup
ORDER BY
RPT.AxSubgroup
答案 2 :(得分：1)
需要一个与样本源数据匹配的预期输出的更好示例。这个问题仍然存在很多含糊之处。但这是另一个刺：
将样本源数据与4个“AxSubGroup”项目（“Bar”，“Tin”，Truffle“，”Box“）一起使用，此解决方案限制在前3个AxSubGroup（而不是前20个）。
“top”子查询为每个“AxSubGroup”求和“UnitsSold”，按降序排序，为它们编号，并抓住前3个。这是最后一个条目被削掉之前的结果：
axsubgroup | totalsold | rownum
------------+-----------+--------
Bar | 27 | 1
Box | 19 | 2
Truffle | 10 | 3
Tin | 5 | 4
(4 rows)
然后将此集合连接回原始RPT表，以获取每个这些顶级axsubgroup项目的详细信息。
CREATE TABLE rpt(CodeNumber varchar(6), UnitsSold int, Dollars int, AxSubGroup varchar(32), ItemStructure varchar(32));
INSERT INTO rpt VALUES('ABC123', 9, 500, 'Bar', '1 oz');
INSERT INTO rpt VALUES('ABC456', 9, 800, 'Bar', '2 oz');
INSERT INTO rpt VALUES('ABC789', 9, 500, 'Bar', '3 oz');
INSERT INTO rpt VALUES('DEF123', 3, 200, 'Tin', 'Round');
INSERT INTO rpt VALUES('DEF456', 2, 200, 'Tin', 'Rectangular');
INSERT INTO rpt VALUES('GHI123', 1, 200, 'Truffle', 'Luxe');
INSERT INTO rpt VALUES('GHI456', 5, 100, 'Truffle', 'Executive');
INSERT INTO rpt VALUES('GHI789', 4, 200, 'Truffle', 'Eco');
INSERT INTO rpt VALUES('JKL123', 9, 500, 'Box', '2 pc');
INSERT INTO rpt VALUES('JKL456', 5, 100, 'Box', '4 pc');
INSERT INTO rpt VALUES('JKL789', 5, 100, 'Box', '8 pc');
WITH top AS (
-- Top AxSubGroups based on UnitsSold with order number assigned, restricted to top 20
SELECT AxSubGroup, totalSold, rowNum
FROM (
-- Top AxSubGroups based on UnitsSold with order number assigned
SELECT AxSubGroup, totalSold
,ROW_NUMBER() OVER (ORDER BY totalSold DESC) AS rowNum
FROM (
-- Top AxSubGroups based on UnitsSold
-- (do you really want this 'UnitsSold' or rather on 'gross': 'UnitsSold' * 'Dollars'?)
SELECT AxSubGroup, SUM(UnitsSold) AS totalSold
FROM rpt r
GROUP BY AxSubGroup
ORDER by totalSold DESC
) x
ORDER BY totalSold DESC
) y
WHERE rowNum <= 3
)
SELECT t.AxSubGroup, t.totalSold, r.ItemStructure, r.UnitsSold, r.Dollars
FROM top t
,rpt r
WHERE r.AxSubGroup = t.AxSubGroup;
axsubgroup | totalsold | itemstructure | unitssold | dollars
------------+-----------+---------------+-----------+---------
Bar | 27 | 1 oz | 9 | 500
Bar | 27 | 2 oz | 9 | 800
Bar | 27 | 3 oz | 9 | 500
Truffle | 10 | Luxe | 1 | 200
Truffle | 10 | Executive | 5 | 100
Truffle | 10 | Eco | 4 | 200
Box | 19 | 2 pc | 9 | 500
Box | 19 | 4 pc | 5 | 100
Box | 19 | 8 pc | 5 | 100
(9 rows)
结果的前两列是对每个成员重复的摘要/总列，而最后三列是构成总和的详细信息。
答案 3 :(得分：1)
所以看起来你需要：
计算小组总数。
按总数对组进行排名。
获取前20名UNION中排名前20位的组的详细信息。
所以也许是这样的：
;
WITH RPT (CodeNumber, UnitsSold, Dollars, AxSubGroup, ItemStructure) AS (
SELECT 'ABC123', 9, 500, 'Bar' , '1 oz' UNION ALL
SELECT 'ABC456', 9, 800, 'Bar' , '2 oz' UNION ALL
SELECT 'ABC789', 9, 500, 'Bar' , '3 oz' UNION ALL
SELECT 'DEF123', 3, 200, 'Tin' , 'Round' UNION ALL
SELECT 'DEF456', 2, 200, 'Tin' , 'Rectangular' UNION ALL
SELECT 'GHI123', 1, 200, 'Truffle', 'Luxe' UNION ALL
SELECT 'GHI456', 5, 100, 'Truffle', 'Executive' UNION ALL
SELECT 'GHI789', 4, 200, 'Truffle', 'Eco' UNION ALL
SELECT 'JKL123', 9, 500, 'Box' , '2 pc' UNION ALL
SELECT 'JKL456', 5, 100, 'Box' , '4 pc' UNION ALL
SELECT 'JKL789', 5, 100, 'Box' , '8 pc'
),
totals AS (
SELECT
*,
TotalUnitsSold = SUM(UnitsSold) OVER (PARTITION BY AxSubGroup),
TotalDollars = SUM(Dollars ) OVER (PARTITION BY AxSubGroup)
FROM RPT
),
ranked AS (
SELECT
*,
rnk = DENSE_RANK() OVER (ORDER BY TotalDollars DESC, AxSubGroup)
FROM totals
)
SELECT
rnk,
AxSubGroup,
ItemStructure,
UnitsSold,
Dollars
FROM ranked
WHERE rnk <= 3
UNION ALL
SELECT DISTINCT
rnk,
AxSubGroup,
NULL,
TotalUnitsSold,
TotalDollars
FROM ranked
WHERE rnk <= 3
ORDER BY rnk, AxSubGroup, ItemStructure
结果是：
rnk AxSubGroup ItemStructure UnitsSold Dollars
--- ---------- ------------- ----------- -----------
1 Bar NULL 27 1800
1 Bar 1 oz 9 500
1 Bar 2 oz 9 800
1 Bar 3 oz 9 500
2 Box NULL 19 700
2 Box 2 pc 9 500
2 Box 4 pc 5 100
2 Box 8 pc 5 100
3 Truffle NULL 10 500
3 Truffle Eco 4 200
3 Truffle Executive 5 100
3 Truffle Luxe 1 200