查询 SQL - 对具有空值的员工/列进行排名

Question

我需要修复以下 SQL 查询：

SELECT  sep.name AS company_name,
    atnp.name AS employee_name,
    (Case WHEN ip.id_item = 0001 THEN 'Product A'  
        ELSE 'Product B'
        END ) AS product_name,
    (Case WHEN ivcp.id_item = 0001 THEN SUM(ivcp.quantity) 
        ELSE NULL
        END ) AS quantityA,
    (Case WHEN ivcp.id_item = 0002 THEN SUM(ivcp.quantity) 
        ELSE NULL
        END ) AS quantityB
FROM    sale_item AS ivcp
    INNER JOIN stock AS mep ON (mep.id_stock = ivcp.id_stock)
    INNER JOIN sale AS vcp ON (vcp.id_sale = ivcp.id_sale)
    INNER JOIN terminal_sale AS mvtp ON (mvtp.id_terminal_sale = vcp.id_terminal_sale)
    INNER JOIN item AS ip ON (ivcp.id_item = ip.id_item)
    INNER JOIN item_category AS cip ON (ip.id_category_item = cip.id_category_item)
    LEFT OUTER JOIN employee AS atnp ON (atnp.id_employee = ivcp.id_employee)
    INNER JOIN sis_company AS sep ON (sep.id_company = mvtp.id_company)
WHERE   mvtp.id_company IN (1)
AND mvtp.date_sale BETWEEN :BeginDate AND  :EndDate
AND DATE(ivcp.date_sale) BETWEEN :BeginDate AND  :EndDate
AND DATE(vcp.data_rec) BETWEEN :BeginDate AND  :EndDate
AND mep.date_sale BETWEEN :BeginDate AND  :EndDate
AND (-1 = -1 OR ivcp.id_employee IN
           (SELECT id_employee FROM team_employee WHERE id_employee = -1))
AND ip.id_category_item = 100
AND ivcp.id_item IN (0002,0001)
AND (ivcp.cancel = 'N' AND vcp.cancel = 'N')
GROUP BY sep.name,
    atnp.name,
    ip.id_item,
    ivcp.id_item
ORDER BY sep.name

结果：

COMPANY   | EMPLOYEE   | PRODUCT NAME | QUANTITY A | QUANTITY B
Company A | Employee A | Product A    | 1.64       | NULL
Company A | Employee A | Product B    | NULL       | 4.70
Company A | Employee B | Product A    | 84.191     | NULL
Company A | Employee B | Product B    | NULL       | 91.885

但我想要这个：

COMPANY   | EMPLOYEE   | QUANTITY A | QUANTITY B
Company A | Employee A | 1.64       | 4.70
Company A | Employee B | 84.191     | 91.885

• 此查询目标是将某些员工的销售额排序到畅销书排名中。排名将按“产品 A”和“产品 B”销售额排序（ivcp.quantity、id_item = 0001 和 id_item = 0002）。

• 内连接用于匹配销售日期、商品类别、员工 ID、库存等。

我正在使用基于 Stimulsoft 的程序来设计报告布局，但我不太习惯。因此，如果可以在查询中解决问题，那就太好了！

Answer 1

这基本上是您的查询的简化。在此过程中，我还修复了其他一些问题：

SELECT sc.name AS company_name
     , em.name AS employee_name
     , SUM((CASE WHEN si.id_item = 0001 THEN si.quantity END) AS quantity_a
     , SUM((CASE WHEN si.id_item = 0002 THEN si.quantity END) AS quantity_b
FROM   team_employee  te
JOIN   sale_item      si USING (id_employee)
JOIN   stock          st ON st.id_stock = si.id_stock
JOIN   sale           sa ON sa.id_sale = si.id_sale
JOIN   terminal_sale  ts ON ts.id_terminal_sale = sa.id_terminal_sale
JOIN   sis_company    sc ON sc.id_company = ts.id_company
JOIN   item           ip ON ip.id_item = si.id_item
JOIN   item_category  ic ON ic.id_category_item = ip.id_category_item
LEFT   JOIN employee  em ON em.id_employee = si.id_employee
WHERE  te.id_employee = -1
AND    st.date_sale BETWEEN :BeginDate AND :EndDate
AND    ts.date_sale BETWEEN :BeginDate AND :EndDate
AND    si.date_sale >= :BeginDate      -- inlude lower border
AND    si.date_sale <  (:EndDate + 1)  -- exclude upper border
AND    sa.data_rec  >= :BeginDate
AND    sa.data_rec  <  (:EndDate + 1)
AND    ts.id_company = 1
AND    ip.id_category_item = 100
AND    si.id_item IN (0002,0001)
AND    si.cancel = 'N'
AND    sa.cancel = 'N'
GROUP  BY 1, 2
ORDER  BY 3 DESC NULLS LAST
        , 4 DESC NULLS LAST;

• 至于您的问题：使用类似演示的方法恢复方法：CASE 语句的总和，而不是相反。并相应地调整GROUP BY 和ORDER BY。
  在 Postgres 9.4 中，您将使用聚合 FILTER：

  • How can I simplify this game statistics query?

• -1 = -1 OR ... 只是噪音。我删除了它。

• 奇怪：

  AND   (ivcp.id_employee IN 
    (SELECT id_employee FROM team_employee WHERE id_employee = -1))
  

  假设-1 应该是一个参数。无论哪种方式，我都将其重写为另一个联接。

• 为什么ip.id_item = 0001 中的前导零？如果是数字，则删除零。如果是字符串，请添加单引号以使其成为字符串文字而不是数字文字。

• id_company IN (1) 只是说id_company = 1 的一种复杂方式。

• 为什么选择演员DATE(si.date_sale)？顾名思义，它已经是date。要么名称具有误导性，要么表达方式具有误导性。无论哪种方式，这都不利于性能，因为表达式不是sargable。
  假设它是timestamp，我将其重写为可搜索。 sa.data_rec 也是如此。 还假设:EndDate 是date，所以我们可以添加一个integer：:EndDate + 1

• si.cancel = 'N' 和 sa.cancel = 'N' 看起来像应该真正实现为 boolean 的列。

• 我根据你的描述改编了ORDER BY：

  排名将按“产品 A”和“产品 B”的销售额排序

  除非您将其设为DESC NULLS LAST，否则 NULL 值会按降序排列：

  • PostgreSQL sort by datetime asc, null first?