查找mysql数据库中不同行中时间戳之间的重叠，按天和ID分组

Question

I have a table, where events of certain objects are listed.

There are two events: "movement" and "load". They can start and end, and these events are listed with timestamps when they happened.

With the following request, i can get the following values:

sum of time, when movement took place (value in the fiddle: 421)
sum of time, when load took place (value in the fiddle: 520)
sum of time, when movement and load took place (value in the fiddle: 391)

The current version calculates these numbers for each DAY, this works fine.

http://sqlfiddle.com/#!2/518dd/1

But i will also have different ID's. I want now to calculate these numbers for each DAY and for each ID. When I add this data, it crashes, see this fiddle:

http://sqlfiddle.com/#!2/b43fa/6

This is not a duplicate to this question: Find intersections between rows and timestamps in a mysql db , because the solution to this question is not suitable/adaptable in this case.

Answer 1

The problem is the query that you are starting with. I simply don't understand what the third column is. But, the first two are much more easily calculated using:

select (sum(case when event = 'movement end' then timestamp end) -
        sum(case when event = 'movement start' then timestamp end)
       ),
       (sum(case when event = 'load end' then timestamp end) -
        sum(case when event = 'load start' then timestamp end)
       )
from table1

From this, it should be trivial to add the group by:

select id, date(timestamp),
        (sum(case when event = 'movement end' then timestamp end) -
        sum(case when event = 'movement start' then timestamp end)
       ),
       (sum(case when event = 'load end' then timestamp end) -
        sum(case when event = 'load start' then timestamp end)
       )
from table1
group by id, date(timestamp)

Answer 2

As shown by Gordon, grouping is made possible by moving the filtering into the outermost SELECT . The search for the loading events that need to enclose the movement events is done in the same way as in the earlier answer :

SELECT id,
       date(timestamp, 'unixepoch') AS date,
       (SUM(CASE WHEN event = 'movement end'   THEN timestamp END) -
        SUM(CASE WHEN event = 'movement start' THEN timestamp END)
       ) AS all_movement,
       (SUM(CASE WHEN event = 'load end'   THEN timestamp END) -
        SUM(CASE WHEN event = 'load start' THEN timestamp END)
       ) AS all_load,
       (SUM(CASE WHEN event = 'movement end' AND
                      (SELECT event
                       FROM Table1 b
                       WHERE timestamp = (SELECT MIN(timestamp)
                                          FROM Table1 c
                                          WHERE c.timestamp >= a.timestamp
                                            AND c.id = a.id
                                            AND c.event LIKE 'load %')
                         AND b.id = a.id
                         AND b.event LIKE 'load %'
                      ) = 'load end'
            THEN timestamp END) -
        SUM(CASE WHEN event = 'movement start' AND
                      (SELECT event
                       FROM Table1 b
                       WHERE timestamp = (SELECT MAX(timestamp)
                                          FROM Table1 c
                                          WHERE c.timestamp <= a.timestamp
                                            AND c.id = a.id
                                            AND c.event LIKE 'load %')
                         AND b.id = a.id
                         AND b.event LIKE 'load %'
                      ) = 'load start'
            THEN timestamp END)
       ) AS load_movement
FROM Table1 a
GROUP BY id,
         date(timestamp, 'unixepoch')