PostgreSQL：为所有行创建具有唯一时间戳的表

Posted 2023-03-31

【中文标题】PostgreSQL：为所有行创建具有唯一时间戳的表

【英文标题】：PostgreSQL: create table with unique timestamp for all rows

【发布时间】：2020-08-08 16:13:40

【问题描述】：

我在如下表格中记录了用户的行程，包括开始/结束位置和时间：

CREATE TABLE trips(id integer, start_timestamp timestamp with time zone, 
                   session_id integer, start_lat double precision,
                   start_lon double precision, end_lat double precision, 
                   end_lon double precision, mode integer);

INSERT INTO trips (id, start_timestamp, session_id, start_lat,start_lon,end_lat,end_lon,mode)
VALUES (563097015,'2017-05-20 17:47:12+01', 128618, 41.1783308,-8.5949878, 41.1784478, -8.5948463, 0),
 (563097013, '2017-05-20 17:45:29+01', 128618, 41.1781344, -8.5951169, 41.1782919, -8.5950689,  0),
 (563097011, '2017-05-20 17:43:41+01', 128618, 41.1781196, -8.5954075, 41.1782139, -8.5950689,  0),
 (563097009, '2017-05-20 17:41:48+01', 128618, 41.1782497,  -8.595197, 41.1781101, -8.5954124,  0),
 (563097003, '2017-05-20 17:10:29+01', 128618, 41.1832512, -8.6081606, 41.1782561, -8.5950259,  0)

第二个表是所有行程的原始 gps 跟踪记录，类似于：

CREATE TABLE gps_traces (session_id integer, seconds integer, lat double precision,
                         lon double precision, speed double precision);

INSERT INTO gps_traces (session_id, seconds , lat , lon , speed )
VALUES (128618,1495296443,41.1844471,-8.6065158,1.35148),
     (128618,1495296444,41.1844482,-8.6065303,1.28004),
     (128618,1495296445,41.1844572,-8.6065503,1.46086),
     (128618,1495296446,41.1844541,-8.6065691,1.23),
     (128618,1495296446,41.1844589,-8.6065861, 1.22919),
     (128618,1495296447,41.1844587, -8.6066043, 1.30188),
     (128618, 1495296448, 41.1844604, -8.6066261, 1.43126),
     (128618, 1495296449, 41.184471, -8.6066412, 1.55003),
     (128618,1495296450, 41.1844715, -8.6066572, 1.29062),
     (128618,1495296450, 41.1844707, -8.6066736,  1.3618)

据此，我想创建一个新表 mytable，其中包含 GPS 在 session_id 上加入这些表，如下所示：

CREATE TABLE mytable AS SELECT id, seconds, lat, lon, speed, mode
FROM trips t
JOIN gps_traces g
ON t.session_id=g.session_id

但是，在新表中，我想确保对于在同一行程中以相同 unix 时间戳记录两次的行，仅将其选择到我的新表中。例如在这种情况下：

SELECT * FROM mytable WHERE id = 563097003; 
+-----------+------------+------------+------------+---------+------+
|    id     |  seconds   |    lat     |    lon     |  speed  | mode |
+-----------+------------+------------+------------+---------+------+
| 563097003 | 1495296443 | 41.1844471 | -8.6065158 | 1.35148 |    0 |
| 563097003 | 1495296444 | 41.1844482 | -8.6065303 | 1.28004 |    0 |
| 563097003 | 1495296445 | 41.1844572 | -8.6065503 | 1.46086 |    0 |
| 563097003 | 1495296446 | 41.1844541 | -8.6065691 | 1.23    |    0 |
| 563097003 | 1495296446 | 41.1844589 | -8.6065861 | 1.22919 |    0 |
| 563097003 | 1495296447 | 41.1844587 | -8.6066043 | 1.30188 |    0 |
| 563097003 | 1495296448 | 41.1844604 | -8.6066261 | 1.43126 |    0 |
| 563097003 | 1495296449 | 41.184471  | -8.6066412 | 1.55003 |    0 |
| 563097003 | 1495296450 | 41.1844715 | -8.6066572 | 1.29062 |    0 |
| 563097003 | 1495296450 | 41.1844707 | -8.6066736 | 1.3618  |    0 |
| 10 rows   |            |            |            |         |      |
+-----------+------------+------------+------------+---------+------+

seconds 列是 Unix 时间戳。如图所示，我们可以看到在 1495296446 和 1495296450 处具有超过 1 个唯一时间戳计数的行。我想确保对于每次旅行，将记录选择到具有唯一时间戳的新表中（因此在上述情况下，应该只将一条记录选择到新表中）。我在这个db<>fiddle 中说明了这一点。

编辑

预期输出：

    +-----------+------------+------------+------------+---------+------+
    |    id     |  seconds   |    lat     |    lon     |  speed  | mode |
    +-----------+------------+------------+------------+---------+------+
    | 563097003 | 1495296443 | 41.1844471 | -8.6065158 | 1.35148 |    0 |
    | 563097003 | 1495296444 | 41.1844482 | -8.6065303 | 1.28004 |    0 |
    | 563097003 | 1495296445 | 41.1844572 | -8.6065503 | 1.46086 |    0 |
    | 563097003 | 1495296446 | 41.1844541 | -8.6065691 | 1.23    |    0 |
    | 563097003 | 1495296447 | 41.1844587 | -8.6066043 | 1.30188 |    0 |
    | 563097003 | 1495296448 | 41.1844604 | -8.6066261 | 1.43126 |    0 |
    | 563097003 | 1495296449 | 41.184471  | -8.6066412 | 1.55003 |    0 |
    | 563097003 | 1495296450 | 41.1844715 | -8.6066572 | 1.29062 |    0 |
    | 8 rows    |            |            |            |         |      |
    +-----------+------------+------------+------------+---------+------+

【问题讨论】：

"只应选择一个记录" 这里的主要问题：到底是哪一个？

@Abelisto 例如在预期答案表中，选择了时间戳1495296446 和1495296450 的第一条记录。在新表中，id= 563097003 现在有 8 行，而不是 10 行。

我的意思是：那些具有重复的seconds 值的行仅相差lat/lon/speed。您如何决定最终数据中应该包含哪一个？

有固定戈登的回答：dbfiddle.uk/…

祝你好运！ ) PS：请注意，没有任何重要排序的“遇到的第一行”表示“随机”。因此，您可能需要将g.ctid desc 添加到order by 子句以获取最后插入的行。

【参考方案1】：

使用DISTINCT ON:

CREATE TABLE mytable AS
    SELECT DISTINCT ON (t.session_id, seconds) id, seconds, lat, lon, speed, mode
    FROM trips t JOIN
         gps_traces g
         ON t.session_id = g.session_id
    ORDER BY t.session_id, seconds;

注意：我希望您也将session_id 包含在新表中。

感谢@Abelisto，事实证明，对这个答案的以下修改按预期工作。

CREATE TABLE mytable AS SELECT DISTINCT ON (id, seconds)id, 
 seconds, lat, lon, speed, mode
FROM trips t
JOIN gps_traces g
ON t.session_id=g.session_id
ORDER BY id, seconds

Here 是一个 dbfiddle。

【讨论】：

错误：column reference "session_id" is ambiguous LINE 2: SELECT DISTINCT ON (session_id, seconds) id, seconds, la..

@super_ask 。 . .那应该是合格的。

但这仅给出两行而不是问题中的所有行，而seconds中没有重复

奇怪，它适用于SELECT DISTINCT ON (..)，但不适用于CREATE TABLE AS SELECT DISTINCT ON (..) 所以例如，id=563097003 只返回了两行。这个 id 应该有八行。