Rumah > Artikel > pangkalan data > Mysql中查找并删除重复数据的方法
(一)单个字段 1、查找表中多余的重复记录,根据(question_title)字段来判断 代码如下 select* from questions where question_title in (select question_title from peoplegroup byquestion_title having count(question_title) 1) 2、删除表中多余的重复记
(一)单个字段
1、查找表中多余的重复记录,根据(question_title)字段来判断
代码如下
select * from questions where question_title in (select question_title from peoplegroup by question_title having count(question_title) > 1)
2、删除表中多余的重复记录,根据(question_title)字段来判断,只留有一个记录
代码如下
delete from questions
where peopleId in (select peopleId from people group by peopleId having count(question_title) > 1)
and min(id) not in (select question_id from questions group by question_title having count(question_title)>1)
(二)多个字段
删除表中多余的重复记录(多个字段),只留有rowid最小的记录
代码如下
DELETE FROM questions WHERE (questions_title,questions_scope) IN (SELECT questions_title,questions_scope FROM questions GROUP BY questions_title,questions_scope HAVING COUNT(*) > 1) AND question_id NOT IN (SELECT MIN(question_id) FROM questions GROUP BY questions_scope,questions_title HAVING COUNT(*)>1)
用上述语句无法删除,,创建了临时表才删的,求各位达人解释一下。
代码如下
CREATE TABLE tmp AS SELECT question_id FROM questions WHERE (questions_title,questions_scope) IN (SELECT questions_title,questions_scope FROM questions GROUP BY questions_title,questions_scope HAVING COUNT(*) > 1) AND question_id NOT IN (SELECT MIN(question_id) FROM questions GROUP BY questions_scope,questions_title HAVING COUNT(*)>1);
DELETE FROM questions WHERE question_id IN (SELECT question_id FROM tmp);
DROP TABLE tmp;
(三) 存储过程
代码如下
declare @max integer,@id integer
declare cur_rows cursor local for select 主字段,count(*) from 表名 group by 主字段 having count(*) >; 1
open cur_rows
fetch cur_rows into @id,@max
while @@fetch_status=0
begin
select @max = @max -1
set rowcount @max
delete from 表名 where 主字段 = @id
fetch cur_rows into @id,@max
end
close cur_rows
set rowcount 0
例,
数据库版本 Server version: 5.1.41-community-log MySQL Community Server (GPL)
例1,表中有主键(可唯一标识的字段),且该字段为数字类型
例1测试数据
代码如下
/* 表结构 */
DROP TABLE IF EXISTS `t1`;
CREATE TABLE IF NOT EXISTS `t1`(
`id` INT(1) NOT NULL AUTO_INCREMENT,
`name` VARCHAR(20) NOT NULL,
`add` VARCHAR(20) NOT NULL,
PRIMARY KEY(`id`)
)Engine=InnoDB;
/* 插入测试数据 */
INSERT INTO `t1`(`name`,`add`) VALUES
('abc',"123"),
('abc',"123"),
('abc',"321"),
('abc',"123"),
('xzy',"123"),
('xzy',"456"),
('xzy',"456"),
('xzy',"456"),
('xzy',"789"),
('xzy',"987"),
('xzy',"789"),
('ijk',"147"),
('ijk',"147"),
('ijk',"852"),
('opq',"852"),
('opq',"963"),
('opq',"741"),
('tpk',"741"),
('tpk',"963"),
('tpk',"963"),
('wer',"546"),
('wer',"546"),
('once',"546");
SELECT * FROM `t1`;
+----+------+-----+
| id | name | add |
+----+------+-----+
| 1 | abc | 123 |
| 2 | abc | 123 |
| 3 | abc | 321 |
| 4 | abc | 123 |
| 5 | xzy | 123 |
| 6 | xzy | 456 |
| 7 | xzy | 456 |
| 8 | xzy | 456 |
| 9 | xzy | 789 |
| 10 | xzy | 987 |
| 11 | xzy | 789 |
| 12 | ijk | 147 |
| 13 | ijk | 147 |
| 14 | ijk | 852 |
| 15 | opq | 852 |
| 16 | opq | 963 |
| 17 | opq | 741 |
| 18 | tpk | 741 |
| 19 | tpk | 963 |
| 20 | tpk | 963 |
| 21 | wer | 546 |
| 22 | wer | 546 |
| 23 | once | 546 |
+----+------+-----+
rows in set (0.00 sec)
查找id最小的重复数据(只查找id字段)
代码如下
/* 查找id最小的重复数据(只查找id字段) */
SELECT DISTINCT MIN(`id`) AS `id`
FROM `t1`
GROUP BY `name`,`add`
HAVING COUNT(1) > 1;
+------+
| id |
+------+
| 1 |
| 12 |
| 19 |
| 21 |
| 6 |
| 9 |
+------+
rows in set (0.00 sec)
查找所有重复数据
代码如下
/* 查找所有重复数据 */
SELECT `t1`.*
FROM `t1`,(
SELECT `name`,`add`
FROM `t1`
GROUP BY `name`,`add`
HAVING COUNT(1) > 1
) AS `t2`
WHERE `t1`.`name` = `t2`.`name`
AND `t1`.`add` = `t2`.`add`;
+----+------+-----+
| id | name | add |
+----+------+-----+
| 1 | abc | 123 |
| 2 | abc | 123 |
| 4 | abc | |
| 6 | xzy | 456 |
| 7 | xzy | 456 |
| 8 | xzy | 456 |
| 9 | xzy | 789 |
| 11 | xzy | 789 |
| 12 | ijk | 147 |
| 13 | ijk | 147 |
| 19 | tpk | 963 |
| 20 | tpk | 963 |
| 21 | wer | 546 |
| 22 | wer | 546 |
+----+------+-----+
rows in set (0.00 sec)