如何为每个类别选择前 n 条记录

Question

我正在尝试编写一个连接 2 个表的查询，并将给出前 5 个名称以及他们在某个日期范围（即 2016 年 1 月 1 日至 2017 年 12 月 31 日）之间的某个位置销售的商品数量.

根据我的研究，这是我想出的：

SELECT
    EmployeeName, 
    COUNT(ID) AS 'Number of Deals', 
    CompanyNumber
FROM
(
    SELECT
        EmployeeName, 
        DealID, 
        CompanyNumber,
        ROW_NUMBER() OVER (PARTITION BY CompanyNumber ORDER BY DealID) AS rn
    FROM Deal
    JOIN DealEmployee
        ON Deal.DealID. =DealEmployee.DealID AS T
    WHERE
        Deal.Status = 2 AND
        Date BETWEEN '2016-01-01' AND '2017-12-31' AND
        EmployeeName != '' AND T.rn <=5 

我希望得到以下结果：

我是个新手，我知道我的语法不正确。给予任何帮助将不胜感激

Answer 1

我想你想要这个：

SELECT EmployeeName, num_deals, CompanyNumber
FROM (SELECT EmployeeName, CompanyNumber, COUNT(*) as num_deals,
             ROW_NUMBER() OVER (PARTITION BY CompanyNumber ORDER BY COUNT(*) DESC) AS seqnum
      FROM Deal d JOIN
           DealEmployee de
           ON Deal.DealID. = de.DealID 
      WHERE d.Status = 2 AND
            Date BETWEEN '2016-01-01' AND '2017-12-31' AND
            EmployeeName <> ''
      GROUP BY employee, CompanyNumber
     ) t
WHERE seqnum <= 5;

注意事项：

• 您需要在执行row_number() 之前进行聚合，因为您想根据计数进行枚举。 （此版本将聚合和 row_number() 组合到一个子查询中；聚合先进行。）
• 交易数量按员工和公司汇总。
• 如果您有多个表，请限定所有列名。
• 您的查询存在许多其他错误，例如表别名放错位置和缺少某些行。

Answer 2

在计算/实现后，您需要过滤子查询之外的行号。我想这就是你想要的：

SELECT EmployeeName, CompanyNumber, cnt
FROM
(
    SELECT EmployeeName, CompanyNumber, COUNT(*) AS cnt,
        ROW_NUMBER() OVER (PARTITION BY CompanyNumber ORDER BY COUNT(*) DESC) rn
    FROM Deal d
    INNER JOIN DealEmployee de
        ON d.DealID = de.DealID
    WHERE d.Status = 2 AND Date BETWEEN '2016-01-01' AND '2017-12-31' AND
          EmployeeName != ''
    GROUP BY EmployeeName, CompanyNumber
) t
WHERE rn <= 5
ORDER BY
    CompanyNumber,
    cnt DESC;

请注意，我们在子查询中执行GROUP BY 聚合，我们还计算行号。

Answer 3

未测试，但我会尝试类似：

with
basedata as (    
select EmployeeName
     , CompanyNumber
     , COUNT(ID) as Number_of_Deals
  from Deal
  join DealEmployee
    on Deal.DealID = DealEmployee.DealID    
 where Deal.Status = 2
   and Date between '2016-01-01' and '2017-12-31'
   and EmployeeName !=''
 group by EmployeeName
        , CompanyNumber 
)
,
basedata_with_rank as (
select t.*
     , row_number() over (partition by CompanyNumber order by Number_of_Deals desc) rn
  from basedata
)
select *
  from basedata_with_rank
 where rn <= 5
 order by CompanyNumber
        , Number_of_Deals desc

使用 CTE 可使查询通常更具可读性。顺便说一句：我会避免给列命名为“日期”（保留字），并且我总是为我的列使用限定名称。也许使用分析函数 rank 或dense rank 会更合适，但 row_number 也应该可以。