我有下表。对于2.5分钟（150秒），两个工厂有一个记录。我希望得到一个结果，显示工厂的相应状态的最大值（日期时间）和最小值（日期时间）。但是对于状态的每次变化都应该有一条新线。
plant state unix_timest datetime
1 1 1452639750 2016-01-13 00:02:30
2 1 1452639750 2016-01-13 00:02:30
1 1 1452639900 2016-01-13 00:05:00
2 1 1452639900 2016-01-13 00:05:00
1 1 1452640050 2016-01-13 00:07:30
2 1 1452640050 2016-01-13 00:07:30
1 1 1452640200 2016-01-13 00:10:00
2 1 1452640200 2016-01-13 00:10:00
1 1 1452640350 2016-01-13 00:12:30
2 0 1452640350 2016-01-13 00:12:30
1 1 1452640500 2016-01-13 00:15:00
2 0 1452640500 2016-01-13 00:15:00
1 1 1452640650 2016-01-13 00:17:30
2 1 1452640650 2016-01-13 00:17:30
1 1 1452640800 2016-01-13 00:20:00
2 1 1452640800 2016-01-13 00:20:00
我的SQL语句是：
SELECT a.line, a.current_state, min( a.`datetime` ) AS mindate, max( a.`datetime` ) AS maxdate
FROM `bde_production`
LEFT JOIN bde_production a ON a.unix_timestamp = bde_production.unix_timestamp
LEFT JOIN bde_production b ON b.unix_timestamp = bde_production.unix_timestamp -150
GROUP BY a.line, a.current_state ORDER BY a.line, mindate
结果：
plant current_state mindate maxdate
1 1 2016-01-13 00:02:30 2016-01-13 00:20:00
2 1 2016-01-13 00:02:30 2016-01-13 00:20:00
2 0 2016-01-13 00:12:30 2016-01-13 00:15:00
我希望得到什么：
plant current_state mindate maxdate
1 1 2016-01-13 00:02:30 2016-01-13 00:20:00
2 1 2016-01-13 00:02:30 2016-01-13 00:10:00
2 0 2016-01-13 00:12:30 2016-01-13 00:15:00
2 1 2016-01-13 00:17:30 2016-01-13 00:20:00
由于
答案 0 :(得分：0)
同样地，我没有得到正确的结果，但在这里留下我的划痕，以防它有用。我的基本想法是找到所有更改，为它们分配一个id（change_num），并按转换分组。我稍后会修改
CREATE TABLE bde_production (
`id` INT(11) UNSIGNED AUTO_INCREMENT PRIMARY KEY,
`plant` INT(3) UNSIGNED NOT NULL,
`state` TINYINT(1) UNSIGNED NOT NULL,
`unix_timest` INT(10) UNSIGNED NOT NULL,
`datetime` DATETIME NOT NULL
);
INSERT INTO bde_production (plant, state, unix_timest, datetime) VALUES
(1, 1, 1452639750, "2016-01-13 00:02:30"),
(2, 1, 1452639750, "2016-01-13 00:02:30"),
(1, 1, 1452639900, "2016-01-13 00:05:00"),
(2, 1, 1452639900, "2016-01-13 00:05:00"),
(1, 1, 1452640050, "2016-01-13 00:07:30"),
(2, 1, 1452640050, "2016-01-13 00:07:30"),
(1, 1, 1452640200, "2016-01-13 00:10:00"),
(2, 1, 1452640200, "2016-01-13 00:10:00"),
(1, 1, 1452640350, "2016-01-13 00:12:30"),
(2, 0, 1452640350, "2016-01-13 00:12:30"),
(1, 1, 1452640500, "2016-01-13 00:15:00"),
(2, 0, 1452640500, "2016-01-13 00:15:00"),
(1, 1, 1452640650, "2016-01-13 00:17:30"),
(2, 1, 1452640650, "2016-01-13 00:17:30"),
(1, 1, 1452640800, "2016-01-13 00:20:00"),
(2, 1, 1452640800, "2016-01-13 00:20:00");
SET @line_num_a = 0;
SET @line_num_b = 0;
SET @change_num = 0;
SELECT plant,
IFNULL(`changes_max_state`, `bde_max_state`) AS `current_state`,
`bde_min_date`,
IFNULL(`changes_max_date`, `bde_max_date`),
`changes_max_date`,
`bde_max_date`
FROM (
SELECT
bde.plant,
change_num,
MIN(bde.datetime) `bde_min_date`,
MAX(bde.datetime) `bde_max_date`,
MAX(bde.state) `bde_max_state`,
MIN(changes.datetime) `changes_min_date`,
MAX(changes.state) `changes_max_state`,
MAX(changes.datetime) `changes_max_date`-- MIN(bde.datetime), MAX(bde.datetime)
, MIN(changes.datetime_b),
MAX(changes.datetime_b)
FROM
bde_production bde LEFT JOIN (
SELECT
# a.datetime is the max reading for plant in a.state
b.*, a.datetime AS datetime_b, @change_num := @change_num + 1 `change_num`
FROM
(
SELECT
*,
@line_num_a := @line_num_a + 1 AS `line`
FROM (
SELECT *
FROM bde_production
ORDER BY plant DESC
) a1) a
JOIN (
SELECT
*,
@line_num_b := @line_num_b + 1 AS `line`
FROM (
SELECT *
FROM bde_production
ORDER BY plant DESC
) b1) b
ON
a.plant = b.plant AND a.line = b.line - 1 AND a.state <> b.state
) changes
ON changes.plant=bde.plant AND bde.id > changes.id
GROUP BY change_num, plant
) q
;
答案 1 :(得分：-1)
我几乎得到了结果：
SELECT
d.plant, d.current_state, d.unix_timestamp, d.`datetime`
FROM
bde_production d
left join
bde_production d_prev
on
d_prev.line = d.line and
d_prev.unix_timestamp< (d.unix_timestamp)
left join
bde_production d_inter
on
d_prev.plant= d.plant and
d_inter.unix_timestamp< (d.unix_timestamp) and
d_prev.unix_timestamp< (d_inter.unix_timestamp)
WHERE
d_inter.unix_timestamp is null AND
(d_prev.current_state is null OR d_prev.current_state <> d.current_state
)
结果：
line current_state unix_timestamp datetime
1 1 1452639750 2016-01-13 00:02:30
2 1 1452639750 2016-01-13 00:02:30
2 0 1452640350 2016-01-13 00:12:30
2 1 1452640650 2016-01-13 00:17:30
我只想添加一个列，其中包含current_state更改的下一行的datetime