Skema
CREATE TABLE log
( user_id int, request_timestamp datetime);
INSERT INTO log
VALUES
(1, '2014-10-26 10:51:18'), (1, '2014-10-26 10:52:20'), (1, '2014-10-26 11:15:03'), (1, '2014-10-26 11:39:18'), (1, '2014-10-26 15:01:18'), (1, '2014-10-26 15:01:21'), (1, '2014-10-27 21:22:19'),
(2, '2014-10-15 12:19:01'), (2, '2014-10-15 12:19:12'), (2, '2014-10-15 12:19:45'), (2, '2014-10-15 12:20:03'), (2, '2014-10-17 14:55:13'), (2, '2014-10-17 14:55:19'),(2, '2014-10-17 14:55:22');
Først vil vi give følgende et navn bare for at visualisere det:
Bemærk under 1800 betyder 30 min * 60 sek/minut
Specimen A
-----
select l.user_id,l.request_timestamp,
@sessionnum :=
if((@curuser = user_id and TIME_TO_SEC(TIMEDIFF(request_timestamp,@theDt))>1800),@sessionnum + 1,
if(@curuser <> user_id,1,@sessionnum)) as sessionnum,
@curuser := user_id as v_curuser,
@theDt:=request_timestamp as v_theDt
from log l cross join
(select @curuser := '', @sessionnum := 0,@theDt:='') gibberish
order by l.user_id,l.request_timestamp
+---------+---------------------+------------+-----------+---------------------+
| user_id | request_timestamp | sessionnum | v_curuser | v_theDt |
+---------+---------------------+------------+-----------+---------------------+
| 1 | 2014-10-26 10:51:18 | 1 | 1 | 2014-10-26 10:51:18 |
| 1 | 2014-10-26 10:52:20 | 1 | 1 | 2014-10-26 10:52:20 |
| 1 | 2014-10-26 11:15:03 | 1 | 1 | 2014-10-26 11:15:03 |
| 1 | 2014-10-26 11:39:18 | 1 | 1 | 2014-10-26 11:39:18 |
| 1 | 2014-10-26 15:01:18 | 2 | 1 | 2014-10-26 15:01:18 |
| 1 | 2014-10-26 15:01:21 | 2 | 1 | 2014-10-26 15:01:21 |
| 1 | 2014-10-27 21:22:19 | 3 | 1 | 2014-10-27 21:22:19 |
| 2 | 2014-10-15 12:19:01 | 1 | 2 | 2014-10-15 12:19:01 |
| 2 | 2014-10-15 12:19:12 | 1 | 2 | 2014-10-15 12:19:12 |
| 2 | 2014-10-15 12:19:45 | 1 | 2 | 2014-10-15 12:19:45 |
| 2 | 2014-10-15 12:20:03 | 1 | 2 | 2014-10-15 12:20:03 |
| 2 | 2014-10-17 14:55:13 | 2 | 2 | 2014-10-17 14:55:13 |
| 2 | 2014-10-17 14:55:19 | 2 | 2 | 2014-10-17 14:55:19 |
| 2 | 2014-10-17 14:55:22 | 2 | 2 | 2014-10-17 14:55:22 |
+---------+---------------------+------------+-----------+---------------------+
Så er vi færdige, hvis du vil. Men til smukke print , kan ombryde Specimen A inde i en anden:
select user_id,request_timestamp,sessionnum
from
( select l.user_id,l.request_timestamp,
@sessionnum :=
if((@curuser = user_id and TIME_TO_SEC(TIMEDIFF(request_timestamp,@theDt))>1800),@sessionnum + 1,
if(@curuser <> user_id,1,@sessionnum)) as sessionnum,
@curuser := user_id as v_curuser,
@theDt:=request_timestamp as v_theDt
from log l cross join
(select @curuser := '', @sessionnum := 0,@theDt:='') gibberish
order by l.user_id,l.request_timestamp
) SpecimenA
order by user_id,sessionnum
+---------+---------------------+------------+
| user_id | request_timestamp | sessionnum |
+---------+---------------------+------------+
| 1 | 2014-10-26 10:51:18 | 1 |
| 1 | 2014-10-26 10:52:20 | 1 |
| 1 | 2014-10-26 11:15:03 | 1 |
| 1 | 2014-10-26 11:39:18 | 1 |
| 1 | 2014-10-26 15:01:18 | 2 |
| 1 | 2014-10-26 15:01:21 | 2 |
| 1 | 2014-10-27 21:22:19 | 3 |
| 2 | 2014-10-15 12:19:01 | 1 |
| 2 | 2014-10-15 12:19:12 | 1 |
| 2 | 2014-10-15 12:19:45 | 1 |
| 2 | 2014-10-15 12:20:03 | 1 |
| 2 | 2014-10-17 14:55:13 | 2 |
| 2 | 2014-10-17 14:55:19 | 2 |
| 2 | 2014-10-17 14:55:22 | 2 |
+---------+---------------------+------------+
14 rows in set (0.02 sec)
Bemærk OP's definition af en session. Det er en af inaktivitet , ikke varighed.