PostgreSQL错误：子查询必须只返回一列

Question

使用PostgreSQL-9.1和PostGIS 2.0.1时，在执行包含返回多列的子查询的SELECT查询时，出现错误subquery must return only one column。

如何修改查询/子查询以返回多个列？

查询

SELECT l.id, l.lat, l.lng, l.geom, 
     (SELECT g.id, g.lat, g.lng, ST_Distance(l.geom, g.geom) 
     FROM stage.dogs as g 
     LIMIT 1) 

FROM stage.users As l 

---

完整的查询

SELECT l.id, l.lat, l.lng, l.geom, 
    g.id, g.lat, g.lng, ST_Distance(l.geom, g.geom) 
FROM stage.users As l 
CROSS JOIN (SELECT * 
    FROM stage.dogs as g 
    ORDER BY g.geom <-> l.geom 
    LIMIT 1) as g 

错误

ERROR: invalid reference to FROM-clause entry for table "l" 
SQL state: 42P01 
Hint: There is an entry for table "l", but it cannot be referenced from this part of the query. 

Answer 1

SELECT l.id, l.lat, l.lng, l.geom, 
     g.id, g.lat, g.lng, ST_Distance(l.geom, g.geom) 
FROM stage.users As l 
CROSS JOIN (SELECT * FROM stage.dogs LIMIT 1) as g 

这实际上是你所拥有的（假设stage.dogs）不是空的。不知道是否应该有users和dogs之间的相关性。

---

要找到离用户最近的狗，可以使用此查询。标量子查询查找狗的ID，并将其返回到表中以检索其他列。

SELECT l.id, l.lat, l.lng, l.geom, 
     g.id, g.lat, g.lng, ST_Distance(l.geom, g.geom) 
FROM (
    SELECT l1.*, (SELECT g1.id 
        FROM stage.dogs as g 
        ORDER BY g.geom <-> l.geom 
        LIMIT 1) g_id 
    FROM stage.users As l1 
) l 
JOIN stage.dogs as g ON g.id = l.g_id; 

公平的警告，这不会是一个快速查询。

---

在执行更慢的风险，请参见下面的查询多个表

SELECT l.id, l.lat, l.lng, l.geom, 
     g.id, g.lat, g.lng, ST_Distance(l.geom, g.geom) dog_distance, 
     c.id, c.lat, c.lng, ST_Distance(l.geom, c.geom) cat_distance, 
     b.id, b.lat, b.lng, ST_Distance(l.geom, b.geom) bird_distance 
FROM (
    SELECT l1.*, (SELECT g1.id 
        FROM stage.dogs as g1 
        ORDER BY g1.geom <-> l.geom 
        LIMIT 1) dog_id, 
       (SELECT c1.id 
        FROM stage.cats as c1 
        ORDER BY c1.geom <-> l.geom 
        LIMIT 1) cat_id, 
       (SELECT b1.id 
        FROM stage.cats as b1 
        ORDER BY b1.geom <-> l.geom 
        LIMIT 1) bird_id 
    FROM stage.users As l1 
) l 
LEFT JOIN stage.dogs as g ON g.id = l.dog_id 
LEFT JOIN stage.dogs as c ON c.id = l.cat_id 
LEFT JOIN stage.dogs as b ON b.id = l.bird_id; 

Answer 2

这给你每个用户一行与最接近的狗：上

SELECT DISTINCT ON (l.id) 
     l.id, l.lat, l.lng, l.geom 
     ,g.id, g.lat, g.lng, ST_Distance(l.geom, g.geom) 
FROM stage.users  l 
CROSS JOIN stage.dogs g 
ORDER BY l.id, (l.geom <-> g.geom) 

更多信息该技巧与DISTINCT ON在此有关的答案：

• Select first row in each GROUP BY group?

我想如果你有GiST index上g.geom，规划可能是足够聪明，只是从中挑选关闭项目。不确定，没有测试。否则，这种CROSS JOIN会导致O（N²），并且性能可能会随着更大的表格快速失控。

我引用POSTGIS手动here：

Index only kicks in if one of the geometries is a constant (not in a subquery/cte). e.g. 'SRID=3005;POINT(1011102 450541)'::geometry instead of a.geom

所以，你可能是出于运气在这里。

根据手册，您可能需要订购ST_Distance()才能获得精确的排序顺序，但是您不应该得到最远的一个。这是没有意义的。