MySQL时间序列操作

Question

我在 5.6 版中使用 MySQL。 （提到是因为不存在可用的mysql函数）

给定一个带有“运动检测”时间戳的表格。每一行都检测到运动。没有动作就意味着没有进入。

Id | Date Time
---------------
1  | 2018-01-01 15:00:01  // Start of activity phase 1
2  | 2018-01-01 15:00:03
3  | 2018-01-01 15:00:06  // Stop of activity phase 1
// Non-Activity phase
4  | 2018-01-01 17:01:06  // Start of activity phase 2
5  | 2018-01-01 17:02:06
6  | 2018-01-01 17:02:09  // Stop of activity phase 2
// Non-Activity phase, big one because of holiday
7  | 2018-01-10 19:40:06  // Start of activity phase 3
8  | 2018-01-10 19:41:06  // Stop of activity phase 3

我很难找到一个 SQL 查询来概述“活动阶段”。

我试图得到的是这样的：

Id | Activity starts     | Activity ends
---------------------------------------------
1  | 2018-01-01 15:00:01 | 2018-01-01 15:00:06
2  | 2018-01-01 17:01:06 | 2018-01-01 17:02:09
3  | 2018-01-10 19:40:06 | 2018-01-10 19:41:06

我想查看“活动阶段”的开始和结束时间戳。 “活动阶段”的定义：“活动阶段”位于（例如）至少两个“非活动阶段”之间30 分钟。

提前谢谢你。

Answer 1

我认为在 MySQL 5.6 中实现这一点的唯一方法是使用存储过程（尽管我很乐意看到有人证明我错了）。这个会做你想做的。请注意，它返回许多单行结果集，因此您需要在应用程序框架中处理这些结果集。或者，您可以修改过程以将中间结果存储到临时表中，然后在过程结束时将临时表中的所有内容都存储在 SELECT 中（见下文）。

DELIMITER //
DROP PROCEDURE IF EXISTS get_activity //
CREATE PROCEDURE get_activity()
BEGIN
  DECLARE start, thistime, lasttime DATETIME;
  DECLARE activity_count INT DEFAULT 1;
  DECLARE finished INT DEFAULT 0;
  DECLARE activity_cursor CURSOR FOR SELECT atime FROM activity;
  DECLARE CONTINUE HANDLER FOR NOT FOUND SET finished = 1;
  OPEN activity_cursor;
  FETCH activity_cursor INTO start;
  SET lasttime = start;
  act_loop: LOOP
    FETCH activity_cursor INTO thistime;
    IF finished = 1 THEN
      SELECT activity_count, start, lasttime AS end;
      LEAVE act_loop;
    END IF;
    IF thistime > lasttime + INTERVAL 30 MINUTE THEN
      SELECT activity_count, start, lasttime AS end;
      SET start = thistime;
      SET activity_count = activity_count + 1;
    END IF;
    SET lasttime = thistime;
  END LOOP;
END //

对于您的示例数据，此过程返回：

activity_count  start                   end     
1               2018-01-01 15:00:01     2018-01-01 15:00:06
activity_count  start                   end     
2               2018-01-01 17:01:06     2018-01-01 17:02:09
activity_count  start                   end     
3               2018-01-10 19:40:06     2018-01-10 19:41:06

这是使用临时表的过程：

DELIMITER //
DROP PROCEDURE IF EXISTS get_activity //
CREATE PROCEDURE get_activity()
BEGIN
  DECLARE start, thistime, lasttime DATETIME;
  DECLARE activity_count INT DEFAULT 1;
  DECLARE finished INT DEFAULT 0;
  DECLARE activity_cursor CURSOR FOR SELECT atime FROM activity;
  DECLARE CONTINUE HANDLER FOR NOT FOUND SET finished = 1;
  CREATE TEMPORARY TABLE activity_temp (id INT, start DATETIME, end DATETIME);
  OPEN activity_cursor;
  FETCH activity_cursor INTO start;
  SET lasttime = start;
  act_loop: LOOP
    FETCH activity_cursor INTO thistime;
    IF finished = 1 THEN
      INSERT INTO activity_temp VALUES (activity_count, start, lasttime);
      LEAVE act_loop;
    END IF;
    IF thistime > lasttime + INTERVAL 30 MINUTE THEN
      INSERT INTO activity_temp VALUES (activity_count, start, lasttime);
      SET start = thistime;
      SET activity_count = activity_count + 1;
    END IF;
    SET lasttime = thistime;
  END LOOP;
  SELECT * FROM activity_temp;
  DROP TABLE activity_temp;
END //

输出（来自CALL get_activity()）：

id  start                   end     
1   2018-01-01 15:00:01     2018-01-01 15:00:06
2   2018-01-01 17:01:06     2018-01-01 17:02:09
3   2018-01-10 19:40:06     2018-01-10 19:41:06

Answer 2

尽管@Nick 给出了有效的答案，但我还是花了几个小时才找到以下解决方案：

SET @row_number = 0;
SET @row_number2 = 0;
SET @gap_time = "00:30:00";
Select
  "",
  DAYNAME(prep1.datetime2) "Weekday",
  prep1.datetime2 "Activity starts",
  prep2.datetime1 "Activity ends"

FROM

(SELECT
 (@row_number:=@row_number + 1) AS num
 ,detection1.id id1
 ,detection1.insert_datetime datetime1
 ,detection2.id id2
 ,detection2.insert_datetime datetime2
 ,timediff(detection2.insert_datetime,detection1.insert_datetime) as diff
FROM
 MOVEMENT_TRACKING detection1,
 MOVEMENT_TRACKING detection2
WHERE
 detection1.id + 1 = detection2.id
 and timediff(detection2.insert_datetime,detection1.insert_datetime)  > @gap_time
order by detection1.id) as prep1,

(SELECT
 (@row_number2:=@row_number2 + 1) AS num
 ,detection1.id id1
 ,detection1.insert_datetime datetime1
 ,detection2.id id2
 ,detection2.insert_datetime datetime2
 ,timediff(detection2.insert_datetime,detection1.insert_datetime) as diff
FROM
 MOVEMENT_TRACKING detection1,
 MOVEMENT_TRACKING detection2
WHERE
 detection1.id + 1 = detection2.id
 and timediff(detection2.insert_datetime,detection1.insert_datetime)  > @gap_time
order by detection1.id) as prep2

WHERE
 prep1.num + 1 = prep2.num
ORDER BY
 prep1.datetime2 DESC

一些减少仍然是可能的，由于调试原因，我让它们进入。需要一些技巧，例如

Select "" 

因为数据库托管安全警察规则。