MySQL 中多列的嵌套聚合和分组

Question

考虑下product表（高度精简）：

`id` int AUTO_INCREMENT
`category_id` int
`subcategory_id` int
`vendor_id` int
`price` decimal(6,2)
`inserted_at` timestamp

对于给定的类别 ID，我试图检索包含每个子类别最新价格最低的供应商的列表。“最新”是指供应商可能对给定的类别 ID/子类别 ID 组合有多个价格，因此只应使用该类别 ID/子类别 ID/供应商 ID 的最近插入的价格。如果 2 个或更多供应商的价格之间存在平局，则应使用最低的 ID 作为决胜局。

例如，使用此数据：

id | category_id | subcategory_id | vendor_id | price | inserted_at
---------------------------------------------------------------------------
1  | 1           | 2              | 3         | 16.00 | 2015-07-23 04:00:00  
2  | 1           | 1              | 2         |  9.00 | 2015-07-26 08:00:00 
3  | 1           | 2              | 4         | 16.00 | 2015-08-02 10:00:00 
4  | 1           | 1              | 1         |  7.00 | 2015-08-04 11:00:00
5  | 1           | 1              | 1         | 11.00 | 2015-08-09 16:00:00

因此，首先找到每个子类别/供应商组合的最新价格（行将price=7.00被删除，因为它不是该子类别中该供应商的最新价格）。然后，对于子类别 1，最低价格将是 9（因此 vendor_id = 2），而对于子类别 2，最低价格是 16（两个供应商的 ()ids 3 和 4），因此我们选择最低 vendor_id = 3 的供应商。

我希望得到以下结果category_id = 1：

subcategory_id | vendor_id | price
----------------------------------
1              | 2         |  9.00
2              | 3         | 16.00

这是我到目前为止所拥有的。我觉得它已经开始失控了，这甚至没有考虑到 2 个或更多供应商的价格之间的联系。

SELECT c.subcategory_id, c.vendor_id, c.price
FROM products AS c
JOIN
(
    SELECT MIN(a.price) AS min_price,
           a.subcategory_id
    FROM products AS a
    JOIN
    (
        SELECT  MAX(`inserted_at`) AS latest_price_time,
                vendor_id,
                subcategory_id
        FROM    products
        WHERE   category_id = 1
        GROUP BY vendor_id, subcategory_id
    ) AS b
    ON a.inserted_at = b.latest_price_time AND a.vendor_id = b.vendor_id AND a.subcategory_id = b.subcategory_id
    WHERE a.category_id = 1
    GROUP BY a.subcategory_id
) AS d
ON c.price = d.min_price AND c.subcategory_id = d.subcategory_id
WHERE c.category_id = 1

在我继续之前，我想看看是否有更简单的方法。当涉及到额外分组/聚合的分组/聚合结果时，是否有一种方法可以为我提供最佳性能（最重要）和/或更易于阅读（不太重要）？

Answer 1

这应该有效：

SELECT
    d.subcategory_id,
    d.vendor_id,
    MIN(d.price) AS price,
    d.inserted_at
FROM product AS d
JOIN (SELECT
        b.category_id,
        b.subcategory_id,
        b.vendor_id,
        a.last_iat
    FROM product AS b 
    JOIN (SELECT
            a.category_id,
            a.subcategory_id,
            a.vendor_id,
            a.price,
            MAX(a.inserted_at) AS last_iat
        FROM product AS a
        GROUP BY a.category_id,a.subcategory_id,a.vendor_id
        ) AS a
        ON (a.category_id=b.category_id AND a.subcategory_id=b.subcategory_id AND a.vendor_id=b.vendor_id)
    GROUP BY b.category_id,b.subcategory_id,b.vendor_id) AS c
    ON (c.category_id=d.category_id AND c.subcategory_id=d.subcategory_id AND c.last_iat=d.inserted_at)
WHERE d.category_id=1
GROUP BY d.category_id,d.subcategory_id;

测试：

mysql> SELECT
    ->     d.subcategory_id,
    ->     d.vendor_id,
    ->     MIN(d.price) AS price,
    ->     d.inserted_at
    -> FROM product AS d
    -> JOIN (SELECT
    ->         b.category_id,
    ->         b.subcategory_id,
    ->         b.vendor_id,
    ->         a.last_iat
    ->     FROM product AS b 
    ->     JOIN (SELECT
    ->             a.category_id,
    ->             a.subcategory_id,
    ->             a.vendor_id,
    ->             a.price,
    ->             MAX(a.inserted_at) AS last_iat
    ->         FROM product AS a
    ->         GROUP BY a.category_id,a.subcategory_id,a.vendor_id
    ->         ) AS a
    ->         ON (a.category_id=b.category_id AND a.subcategory_id=b.subcategory_id AND a.vendor_id=b.vendor_id)
    ->     GROUP BY b.category_id,b.subcategory_id,b.vendor_id) AS c
    ->     ON (c.category_id=d.category_id AND c.subcategory_id=d.subcategory_id AND c.last_iat=d.inserted_at)
    -> WHERE d.category_id=1
    -> GROUP BY d.category_id,d.subcategory_id;
+----------------+-----------+-------+---------------------+
| subcategory_id | vendor_id | price | inserted_at         |
+----------------+-----------+-------+---------------------+
|              1 |         2 |  9.00 | 2015-07-26 08:00:00 |
|              2 |         3 | 16.00 | 2015-07-23 04:00:00 |
+----------------+-----------+-------+---------------------+
2 rows in set (0.00 sec)

mysql> 

解释：

我使用了@ypercube 的索引推荐。

mysql> EXPLAIN SELECT d.subcategory_id, d.vendor_id, MIN(d.price) AS price, d.inserted_at FROM product AS d JOIN (SELECT b.category_id, b.subcategory_id, b.vendor_id, a.last_iat FROM product AS b  JOIN (SELECT a.category_id, a.subcategory_id, a.vendor_id, a.price, MAX(a.inserted_at) AS last_iat FROM product AS a GROUP BY a.category_id,a.subcategory_id,a.vendor_id ) AS a ON (a.category_id=b.category_id AND a.subcategory_id=b.subcategory_id AND a.vendor_id=b.vendor_id) GROUP BY b.category_id,b.subcategory_id,b.vendor_id) AS c ON (c.category_id=d.category_id AND c.subcategory_id=d.subcategory_id AND c.last_iat=d.inserted_at) WHERE d.category_id=1 GROUP BY d.category_id,d.subcategory_id;
+----+-------------+------------+-------+---------------+------+---------+--------------------------------------------+------+----------------------------------------------+
| id | select_type | table      | type  | possible_keys | key  | key_len | ref                                        | rows | Extra                                        |
+----+-------------+------------+-------+---------------+------+---------+--------------------------------------------+------+----------------------------------------------+
|  1 | PRIMARY     | <derived2> | ALL   | NULL          | NULL | NULL    | NULL                                       |    4 | Using where; Using temporary; Using filesort |
|  1 | PRIMARY     | d          | ALL   | q_ix          | NULL | NULL    | NULL                                       |    5 | Using where; Using join buffer               |
|  2 | DERIVED     | <derived3> | ALL   | NULL          | NULL | NULL    | NULL                                       |    4 | Using temporary; Using filesort              |
|  2 | DERIVED     | b          | ref   | q_ix          | q_ix | 15      | a.category_id,a.subcategory_id,a.vendor_id |    1 | Using where; Using index                     |
|  3 | DERIVED     | a          | index | NULL          | q_ix | 19      | NULL                                       |    5 |                                              |
+----+-------------+------------+-------+---------------+------+---------+--------------------------------------------+------+----------------------------------------------+
5 rows in set (0.00 sec)

mysql> 

Answer 2

这是一个“greatest-n-per-group”查询，在 MySQL 中编写非常复杂 - 首先是因为缺少窗口函数，其次是因为您有 2 个最大的每组规范，首先是每个最新的日期供应商，每个子类别的最低价格位居第二。

这是一种相当复杂的写法：

SELECT 
    ps.subcategory_id,  ps.vendor_id,  ps.price   -- , p.inserted_at
FROM
    ( SELECT DISTINCT subcategory_id
      FROM product
      WHERE category_id = 1
    ) AS s
  JOIN
    product AS ps
      ON  ps.category_id = 1
      AND ps.subcategory_id = s.subcategory_id
      AND ps.id =
          ( SELECT psv.id
            FROM
                ( SELECT DISTINCT subcategory_id, vendor_id
                  FROM product
                  WHERE category_id = 1
                ) AS sv
              JOIN
                product AS psv
                ON  psv.category_id = 1
                AND psv.subcategory_id = sv.subcategory_id
                AND psv.vendor_id = sv.vendor_id
                AND psv.inserted_at =
                    ( SELECT pi.inserted_at
                      FROM product AS pi
                      WHERE pi.category_id = 1
                        AND pi.subcategory_id = sv.subcategory_id
                        AND pi.vendor_id = sv.vendor_id
                      ORDER BY pi.inserted_at DESC
                      LIMIT 1
                    )
            WHERE sv.subcategory_id = s.subcategory_id
            ORDER BY psv.price,
                     psv.vendor_id
            LIMIT 1
          ) ;

在SQLfiddle-2 中测试。计划也不错，在(category_id, subcategory_id, vendor_id, inserted_at).

它可能不是最有效的，我肯定会尝试使用索引（请参阅 Fiddle，我还有一个索引。它可能不是很有用，但在更大的表上测试 t。）

（SQLfidle-1 中查询的第一个版本）