Question

我有以下查询，它对位置进行聚类，并返回一个聚类ID，以及一个包含该聚类中原始点的结构：

SELECT
    cluster_id,
    ST_AsGeoJSON(ST_Collect(origin)) AS points
FROM (
    SELECT
        origin,
        ST_ClusterDBSCAN(origin, eps := 0.01, minPoints := 5) OVER(ORDER BY id) AS cluster_id
    FROM
        ride
    WHERE
        region_id = 1 AND
        created_at > now() - interval '1 week'
) a
WHERE cluster_id IS NOT NULL
GROUP BY cluster_id

示例输出：

3   {"type":"MultiPoint","coordinates":[ ... ]}
4   {"type":"MultiPoint","coordinates":[ ... ]}

我不想返回原始点，而是要返回聚类的边界圆（中心和半径），可以使用ST_MinimumBoundingRadius完成此操作：

SELECT
    ST_AsGeoJSON(center),
    radius
FROM
    ST_MinimumBoundingRadius(
        ST_GeomFromGeoJSON('{"type":"MultiPoint","coordinates":[ ... ]}')
    )

示例输出：

{"type":"Point","coordinates":[ ... ]}  0.002677744742706528

但是，由于ST_MinimumBoundingRadius返回两列，因此以下内容会引发错误：

SELECT
    cluster_id,
    ST_MinimumBoundingRadius(ST_Collect(origin))
FROM (
    ...
) a
WHERE cluster_id IS NOT NULL
GROUP BY cluster_id;

我发现了this个问题，其中提到使用LATERAL，但是我无法获得有效的查询。

正确的方法是什么？

Answer 1

尝试另一个子选择：

SELECT q.cluster_id,
       (q.mbr).radius,
       (q.mbr).center
FROM (
   SELECT cluster_id,
          ST_MinimumBoundingRadius(
             ST_Collect(origin)
          ) AS mbr
   FROM (
      ...
   ) AS a
) AS q;

Answer 2

最后真的很简单，不知道我怎么没想到这一点。

SELECT
    cluster_id,
    (ST_MinimumBoundingRadius(ST_Collect(origin))).*
FROM (
    SELECT
        origin,
        ST_ClusterDBSCAN(origin, eps := 0.01, minPoints := 5) OVER(ORDER BY id) AS cluster_id
    FROM
        ride
    WHERE
        region_id = 1 AND
        created_at > now() - interval '1 week'
) a
WHERE cluster_id IS NOT NULL
GROUP BY cluster_id

从函数中返回多列的SELECT

2 个答案: