在聚合函数之前排序分组的行

Question

我有一个带点几何的postgis表。

点数表：

id | uid | date | geom

具有相同uid的点数是相同的目标。我正在尝试使用ST_MakeLine GROUP BY uid来创建完整的目标LineString。

SELECT uid, ST_MakeLine(geom)
FROM points
GROUP BY uid

这样可行，但我想确保这些点的顺序正确。 我尝试在分组之前添加ORDER BY date来实现此目的。

SELECT uid, ST_MakeLine(geom)
FROM points
ORDER BY date <-- does not work
GROUP BY uid

ERROR: syntax error at or near "GROUP"

有没有办法在分组行输入聚合函数之前对其进行排序？

Answer 1

ORDER BY子句可以放在聚合参数的末尾。

SELECT uid, ST_MakeLine(geom ORDER BY date)
FROM points
GROUP BY uid

http://www.postgresql.org/docs/9.1/static/sql-expressions.html#SYNTAX-AGGREGATES

Answer 2

也许你想在临时表中订购数据，然后通过uid进行外部分组？

 SELECT uid, ST_MakeLine(geom)
 FROM
 (
    SELECT uid, geom
    FROM points
    ORDER BY date
 ) AS temp
 GROUP BY uid