使用 JOIN 或关联子查询与 TEMPORARY TABLE 进行 MySQL 查找

Question

我有两个主表，providers' 和depots`。每个提供商都有一个强制性的第一位置。通过添加额外的“仓库”来提供额外的位置。 “etc”代表与提供者帐户相关的大量附加变量。为了简化起见，此处省略了额外的表信息和过滤器。

providers
+--------+------------+-----+
| id     | location   | etc |
+--------+------------+-----+
| 1      | POINT(1,1) | ... |
| 2      | POINT(1,2) | ... |
| 3      | POINT(1,3) | ... |
+--------+------------+-----+

depots
+---------+------------+------------+
| depotId | providerId | location   |
+---------+------------+------------+
| 1       | 1          | POINT(2,1) |
| 2       | 1          | POINT(2,2) |
| 3       | 1          | POINT(2,3) |
| 4       | 2          | POINT(2,4) |
| 5       | 2          | POINT(2,5) |
+---------+------------+------------+

提供商可能有零个或多个附加仓库。这些“位置”用于计算来自每个提供商的传入“工作”的距离。传统上，我使用 UNION 将 providers 和 depots 表连接起来形成一个表，我将其称为 provDeps。

从提供商中选择 id、位置、0 AS depotId UNION SELECT p.id, d.location d.id AS depotId FROM 提供商 p, depots d

我们假设这是一个 View ，暂时放弃效率和索引。它有望降低查询的视觉复杂性。

provDeps
+--------+------------+---------+-----+
| id     | location   | depotId | etc |
+--------+------------+---------+-----+
| 1      | POINT(1,1) | 0       | ... |
| 1      | POINT(2,1) | 1       | ... |
| 1      | POINT(2,2) | 2       | ... |
| 1      | POINT(2,3) | 3       | ... |
| 2      | POINT(1,2) | 0       | ... |
| 2      | POINT(2,4) | 4       | ... |
| 2      | POINT(2,5) | 5       | ... |
| 3      | POINT(1,3) | 0       | ... |
+--------+------------+---------+-----+

然后，我使用 provDeps 执行额外的查找。这里的想法是计算工作到每个仓库的距离。这是通过存储过程执行的。

SELECT loc.*, degToMeter(st_distance(jobLocation, location)) AS distanceToJob FROM provDeps;

+--------+------------+---------+---------------+-----+
| id     | location   | depotId | distanceToJob | etc |
+--------+------------+---------+---------------+-----+
| 1      | POINT(1,1) | 0       | 8234          | ... |
| 1      | POINT(2,1) | 1       | 7334          | ... |
| 1      | POINT(2,2) | 2       | 6434          | ... |
| 1      | POINT(2,3) | 3       | 5534          | ... |
| 2      | POINT(1,2) | 0       | 4634          | ... |
| 2      | POINT(2,4) | 4       | 3734          | ... |
| 2      | POINT(2,5) | 5       | 2834          | ... |
| 3      | POINT(1,3) | 0       | 1934          | ... |
+--------+------------+---------+---------------+-----+

我现在需要减少这个列表，只保留最近的仓库，按提供商 ID 分组。结果将包括每个提供商，但每个提供商只有一个仓库 - 要么是主要位置，仓库“0”，要么是最近的仓库的 ID。这是期望的结果:-

+--------+------------+---------+---------------+-----+
| id     | location   | depotId | distanceToJob | etc |
+--------+------------+---------+---------------+-----+
| 1      | POINT(2,3) | 3       | 5534          | ... |
| 2      | POINT(2,5) | 5       | 2834          | ... |
| 3      | POINT(1,3) | 0       | 1934          | ... |
+--------+------------+---------+---------------+-----+

我尝试了多种方法，但每种方法我都遇到了不同的问题。我最接近成功的是使用临时表:-

DROP TEMPORARY TABLE IF EXISTS locTemp;
CREATE TEMPORARY TABLE locTemp AS
    SELECT depots.*, st_distance(jobLocation, location) AS distanceToJob
    FROM provDeps

然后我尝试使用相关子查询，但这会产生有关在一次查找中尝试两次访问临时表的错误:-

SELECT * FROM locTemp
WHERE distanceToJob = (SELECT MIN(distanceToJob) FROM locTemp AS lt WHERE lt.id = locTemp.id);

这会导致错误“无法重新打开表:'locTemp'”。我也尝试过执行联接，但随后出现分组错误，或者无法从子查询中访问临时表本身:-

SELECT * FROM
(
    SELECT id, depotId, MIN(distanceToJob) as minDist
    FROM locTemp GROUP BY id
) AS res
INNER JOIN locTemp AS lt on lt.id = res.id and lt.minDist = res.distanceToJob;

任何指示或更好的解决方案，将不胜感激! :)

Answer 1

我认为您不一定需要这里的临时表或 View 。您的最终查询在这里看起来完全符合预期，我已在下面对其进行了修改。我看到的唯一问题是您使用 GROUP BY 选择非聚合列，并且使用临时表可能存在问题。

SELECT
    t1.id, t1.location, t1.depotId,
    degToMeter(st_distance(t1.jobLocation, t1.location)) AS distanceToJob
FROM provDeps t1
INNER JOIN
(
    SELECT
        id,
        MIN(degToMeter(st_distance(t1.jobLocation, t1.location))) AS minDistanceToJob
    FROM provDeps
    GROUP BY id
) t2
    ON t1.id = t2.id AND
       degToMeter(st_distance(t1.jobLocation, t1.location)) = t2.minDistanceToJob;