递归 CTE 生成组内的所有组合子组

Question

我正在尝试为由一列 (package_id) 链接但另一列 (package_item_id) 不匹配的行的所有可能组合生成聚合结果。

我有一个派生表包：

package_id	package_item_id	数量	variant_id	零售价	销售价格	有库存
1000001	28	1	3399	489.00	342.30	26
1000001	28	1	3400	489.00	342.30	23
1000001	28	1	3409	489.00	342.30	81
1000001	29	1	821	44.95	35.95	741
1000001	30	1	807	39.95	31.95	162
1000002	31	1	6369	519.00	363.30	93
1000002	31	1	6371	519.00	363.30	60
1000002	31	1	6372	519.00	363.30	40
1000002	31	1	19289	519.00	363.30	24
1000002	31	1	19290	519.00	363.30	20
1000002	31	1	21774	519.00	363.30	14
1000002	31	1	23045	519.00	363.30	0
1000002	37	1	821	44.95	35.95	741
1000002	38	1	807	39.95	31.95	162

此结构支持零售网站的打包交易系统：

package_id 与一组可以作为包购买的元素相关。

package_item_id 分解了子元素（根据包的不同，数量也不同），用户可以从中选择variant_ids。

即包装：高级山地自行车，包装物品：车架，款式：蓝色、红色、黄色，包装物品：车轮，款式：窄、宽，包装物品：车座，款式：标准

我正在尝试生成一个聚合 package_id 的每个潜在组合的结果。将 package_id 和variants_ids 连接到查询字符串中并计算分组项目的总价格。数量列决定了为 package_item_id 提供的数量，因此是价格的乘数。此外，我想检查每个变体是否有足够的库存来组成包装 true (1) 或 false (0)

我试图仅针对第一个 package_id 实现的结果示例：

package_id	查询字符串	零售价	销售价格	库存
1000001	p_id=1000001&v_id=3399&v_id=819&v_id=807	598.9	430.2	1
1000001	p_id=1000001&v_id=3400&v_id=819&v_id=807	598.9	430.2	1
1000001	p_id=1000001&v_id=3409&v_id=819&v_id=807	598.9	430.2	1
1000001	p_id=1000001&v_id=3399&v_id=821&v_id=807	573.9	410.2	1
1000001	p_id=1000001&v_id=3400&v_id=821&v_id=807	573.9	410.2	1
1000001	p_id=1000001&v_id=3409&v_id=821&v_id=807	573.9	410.2	1

我一直在尝试在组合递归 CTE 中使用窗口函数来在组合 ID 下创建所有不同的组合，我希望它们能够获得我需要的聚合结果，但到目前为止我无法实现我的目标。

WITH RECURSIVE cte_packages AS (
SELECT 
  package_id,
  DENSE_RANK() OVER (ORDER BY package_id) AS p_group,
  package_item_id,
  DENSE_RANK() OVER (PARTITION BY package_id ORDER BY package_item_id) AS pi_group,
  ROW_NUMBER() OVER (PARTITION BY package_item_id ORDER BY variant_id) AS pi_group_row,
  COUNT(*) OVER (PARTITION BY package_item_id) AS pi_group_row_count,
  quantity,
  variant_id,
  retail_price,
  sell_price,
  stock_available
FROM packages
),
cte_combinations AS
(
  SELECT 
        0 AS combination_id,
        p_group,
        p_group AS next_p_group,
        pi_group,
        pi_group_row,
        CASE 
          WHEN pi_group_row < pi_group_row_count
          THEN pi_group_row + 1
          ELSE 1
        END
        AS next_pi_group_row,
        pi_group_row_count,
        package_id,
        package_item_id,
        quantity,
        variant_id,
        retail_price,
        sell_price,
        stock_available     
    FROM 
        cte_packages
    WHERE 
        p_group = 1 AND pi_group = 1 AND pi_group_row = 1
    UNION
    SELECT
        cc.combination_id + 1,
        cpi.p_group,
        CASE 
          WHEN cpi.pi_group_row = cpi.pi_group_row_count
          THEN cpi.p_group + 1
          ELSE cpi.p_group
        END
        AS next_p_group,
        cpi.pi_group,
        cpi.pi_group_row,
        CASE 
          WHEN cpi.pi_group_row < cpi.pi_group_row_count
          THEN cpi.pi_group_row + 1
          ELSE 1
        END
        AS next_pi_group_row,
        cpi.pi_group_row_count,
        cpi.package_id,
        cpi.package_item_id,
        cpi.quantity,
        cpi.variant_id,
        cpi.retail_price,
        cpi.sell_price,
        cpi.stock_available
    FROM
        cte_combinations cc
    INNER JOIN
        cte_packages cpi
        ON cpi.p_group = cc.next_p_group AND cpi.pi_group = 1 AND cpi.pi_group_row = cc.next_pi_group_row
    LEFT JOIN
        cte_packages opi
        ON opi.p_group = cpi.p_group AND opi.pi_group <> 1
)
SELECT * FROM cte_combinations

我不确定我是否正确地实现了这个目标，或者我想要实现的目标是否可能。

DB Fiddle 链接如下：

https://dbfiddle.uk/hEXZa674

Answer 1

AFAICT 我们需要为每个

variant_id

生成

package_id

的所有可能组合。由于

package_item_id

的数量不同，这将涉及每个

package_item_id

的自连接。然而，对于不同数量的项目，通常需要动态数量的联接，而 SQL 本身并不支持这一点。

此代码片段可能会实现您正在寻找的结果，但仅每个

package_item_id

有三个

package_id

值

SELECT 
  p1.package_id,
  CONCAT('p_id=', p1.package_id, '&v_id=', p1.variant_id, 
         '&v_id=', p2.variant_id, '&v_id=', p3.variant_id) AS querystring,
  (p1.quantity * p1.retail_price + p2.quantity * p2.retail_price + p3.quantity * p3.retail_price) AS total_retail_price,
  (p1.quantity * p1.sell_price + p2.quantity * p2.sell_price + p3.quantity * p3.sell_price) AS total_sell_price,
  CASE 
    WHEN p1.stock_available >= p1.quantity AND p2.stock_available >= p2.quantity AND p3.stock_available >= p3.quantity 
    THEN 1 
    ELSE 0 
  END AS stock_available
FROM packages p1
JOIN packages p2 ON p1.package_id = p2.package_id AND p1.package_item_id <> p2.package_item_id
JOIN packages p3 ON p1.package_id = p3.package_id AND p1.package_item_id <> p3.package_item_id AND p2.package_item_id <> p3.package_item_id
GROUP BY p1.package_id, p1.variant_id, p2.variant_id, p3.variant_id;