• musql 单表查询


    一 介绍

    本节内容:

    查询语法

    关键字的执行优先级

    简单查询

    单条件查询:WHERE

    分组查询:GROUP BY

    HAVING

    查询排序:ORDER BY

    限制查询的记录数:LIMIT

    使用聚合函数查询

    使用正则表达式查询

    复制代码
    company.employee
        员工id      id                  int             
        姓名        emp_name            varchar
        性别        sex                 enum
        年龄        age                 int
        入职日期     hire_date           date
        岗位        post                varchar
        职位描述     post_comment        varchar
        薪水        salary              double
        办公室       office              int
        部门编号     depart_id           int
    
    
    
    #创建表
    create table employee(
    id int not null unique auto_increment,
    name varchar(20) not null,
    sex enum('male','female') not null default 'male', #大部分是男的
    age int(3) unsigned not null default 28,
    hire_date date not null,
    post varchar(50),
    post_comment varchar(100),
    salary double(15,2),
    office int, #一个部门一个屋子
    depart_id int
    );
    
    
    #查看表结构
    mysql> desc employee;
    +--------------+-----------------------+------+-----+---------+----------------+
    | Field        | Type                  | Null | Key | Default | Extra          |
    +--------------+-----------------------+------+-----+---------+----------------+
    | id           | int(11)               | NO   | PRI | NULL    | auto_increment |
    | name         | varchar(20)           | NO   |     | NULL    |                |
    | sex          | enum('male','female') | NO   |     | male    |                |
    | age          | int(3) unsigned       | NO   |     | 28      |                |
    | hire_date    | date                  | NO   |     | NULL    |                |
    | post         | varchar(50)           | YES  |     | NULL    |                |
    | post_comment | varchar(100)          | YES  |     | NULL    |                |
    | salary       | double(15,2)          | YES  |     | NULL    |                |
    | office       | int(11)               | YES  |     | NULL    |                |
    | depart_id    | int(11)               | YES  |     | NULL    |                |
    +--------------+-----------------------+------+-----+---------+----------------+
    
    #插入记录
    #三个部门:教学,销售,运营
    insert into employee(name,sex,age,hire_date,post,salary,office,depart_id) values
    ('egon','male',18,'20170301','老男孩驻沙河办事处外交大使',7300.33,401,1), #以下是教学部
    ('alex','male',78,'20150302','teacher',1000000.31,401,1),
    ('wupeiqi','male',81,'20130305','teacher',8300,401,1),
    ('yuanhao','male',73,'20140701','teacher',3500,401,1),
    ('liwenzhou','male',28,'20121101','teacher',2100,401,1),
    ('jingliyang','female',18,'20110211','teacher',9000,401,1),
    ('jinxin','male',18,'19000301','teacher',30000,401,1),
    ('成龙','male',48,'20101111','teacher',10000,401,1),
    
    ('歪歪','female',48,'20150311','sale',3000.13,402,2),#以下是销售部门
    ('丫丫','female',38,'20101101','sale',2000.35,402,2),
    ('丁丁','female',18,'20110312','sale',1000.37,402,2),
    ('星星','female',18,'20160513','sale',3000.29,402,2),
    ('格格','female',28,'20170127','sale',4000.33,402,2),
    
    ('张野','male',28,'20160311','operation',10000.13,403,3), #以下是运营部门
    ('程咬金','male',18,'19970312','operation',20000,403,3),
    ('程咬银','female',18,'20130311','operation',19000,403,3),
    ('程咬铜','male',18,'20150411','operation',18000,403,3),
    ('程咬铁','female',18,'20140512','operation',17000,403,3)
    ;
    复制代码

    二 查询语法

    复制代码
    SELECT 字段1,字段2... FROM 表名
                      WHERE 条件
                      GROUP BY field
                      HAVING 筛选
                      ORDER BY field
                      LIMIT 限制条数
    复制代码

    三 关键字的执行优先级(重点)

    复制代码
    重点中的重点:关键字的执行优先级
    from
    where
    group by
    having
    select
    distinct
    order by
    limit
    复制代码

    1.找到表:from

    2.拿着where指定的约束条件,去文件/表中取出一条条记录

    3.将取出的一条条记录进行分组group by,如果没有group by,则整体作为一组

    4.按照select后的字段得到一张新的虚拟表,如果有聚合函数,则将组内数据进行聚合

    5.将4的结果过滤:having,如果有聚合函数也是先执行聚合再having过滤

    6.查出结果:select

    7.去重

    8.将结果按条件排序:order by

    9.限制结果的显示条数

    详细见:http://www.cnblogs.com/linhaifeng/articles/7372774.html

    四 简单查询

    复制代码
    #简单查询
        SELECT id,name,sex,age,hire_date,post,post_comment,salary,office,depart_id 
        FROM employee;
    
        SELECT * FROM employee;
    
        SELECT name,salary FROM employee;
    
    #避免重复DISTINCT
        SELECT DISTINCT post FROM employee;    
    
    #通过四则运算查询
        SELECT name, salary*12 FROM employee;
        SELECT name, salary*12 AS Annual_salary FROM employee;
        SELECT name, salary*12 Annual_salary FROM employee;
    
    #定义显示格式
       CONCAT() 函数用于连接字符串
       SELECT CONCAT('姓名: ',name,'  年薪: ', salary*12)  AS Annual_salary 
       FROM employee;
       
       CONCAT_WS() 第一个参数为分隔符
       SELECT CONCAT_WS(':',name,salary*12)  AS Annual_salary 
       FROM employee;
    复制代码

    小练习:

    1 查出所有员工的名字,薪资,格式为
        <名字:egon>    <薪资:3000>
    2 查出所有的岗位(去掉重复)
    3 查出所有员工名字,以及他们的年薪,年薪的字段名为annual_year
     
    select concat('<名字:',name,'>    ','<薪资:',salary,'>') from employee;
    select distinct depart_id from employee;
    select name,salary*12 annual_salary from employee;

    五 WHERE约束

    强调:where是一种约束条件,mysql会拿着where指定的条件去表中取数据,而having则是在取出数据后进行过滤

    where字句中可以使用:

    1. 比较运算符:> < >= <= <> !=
    2. between 80 and 100 值在10到20之间
    3. in(80,90,100) 值是10或20或30
    4. like 'egon%'
        pattern可以是%或_,
        %表示任意多字符
        _表示一个字符 
    5. 逻辑运算符:在多个条件直接可以使用逻辑运算符 and or not

    复制代码
    #1:单条件查询
        SELECT name FROM employee
            WHERE post='sale';
            
    #2:多条件查询
        SELECT name,salary FROM employee
            WHERE post='teacher' AND salary>10000;
    
    #3:关键字BETWEEN AND
        SELECT name,salary FROM employee 
            WHERE salary BETWEEN 10000 AND 20000;
    
        SELECT name,salary FROM employee 
            WHERE salary NOT BETWEEN 10000 AND 20000;
        
    #4:关键字IS NULL(判断某个字段是否为NULL不能用等号,需要用IS)
        SELECT name,post_comment FROM employee 
            WHERE post_comment IS NULL;
    
        SELECT name,post_comment FROM employee 
            WHERE post_comment IS NOT NULL;
            
        SELECT name,post_comment FROM employee 
            WHERE post_comment=''; 注意''是空字符串,不是null
        ps:
            执行
            update employee set post_comment='' where id=2;
            再用上条查看,就会有结果了
    
    #5:关键字IN集合查询
        SELECT name,salary FROM employee 
            WHERE salary=3000 OR salary=3500 OR salary=4000 OR salary=9000 ;
        
        SELECT name,salary FROM employee 
            WHERE salary IN (3000,3500,4000,9000) ;
    
        SELECT name,salary FROM employee 
            WHERE salary NOT IN (3000,3500,4000,9000) ;
    
    #6:关键字LIKE模糊查询
        通配符’%’
        SELECT * FROM employee 
                WHERE name LIKE 'eg%';
    
        通配符’_’
        SELECT * FROM employee 
                WHERE name LIKE 'al__';
    复制代码
    复制代码
    1. 查看岗位是teacher的员工姓名、年龄
    2. 查看岗位是teacher且年龄大于30岁的员工姓名、年龄
    3. 查看岗位是teacher且薪资在9000-1000范围内的员工姓名、年龄、薪资
    4. 查看岗位描述不为NULL的员工信息
    5. 查看岗位是teacher且薪资是10000或9000或30000的员工姓名、年龄、薪资
    6. 查看岗位是teacher且薪资不是10000或9000或30000的员工姓名、年龄、薪资
    7. 查看岗位是teacher且名字是jin开头的员工姓名、年薪
    复制代码
     
    select name,age from employee where post = 'teacher';
    select name,age from employee where post='teacher' and age > 30; 
    select name,age,salary from employee where post='teacher' and salary between 9000 and 10000;
    select * from employee where post_comment is not null;
    select name,age,salary from employee where post='teacher' and salary in (10000,9000,30000);
    select name,age,salary from employee where post='teacher' and salary not in (10000,9000,30000);
    select name,salary*12 from employee where post='teacher' and name like 'jin%';

    六 分组查询:GROUP BY

    大前提:可以按照任意字段分组,但分完组后,只能查看分组的那个字段,要想取的组内的其他字段信息,需要借助函数

    复制代码
    单独使用GROUP BY关键字分组
        SELECT post FROM employee GROUP BY post;
        注意:我们按照post字段分组,那么select查询的字段只能是post,想要获取组内的其他相关信息,需要借助函数
    
    GROUP BY关键字和GROUP_CONCAT()函数一起使用
        SELECT post,GROUP_CONCAT(name) FROM employee GROUP BY post;#按照岗位分组,并查看组内成员名
        SELECT post,GROUP_CONCAT(name) as emp_members FROM employee GROUP BY post;
    
    GROUP BY与聚合函数一起使用
        select post,count(id) as count from employee group by post;#按照岗位分组,并查看每个组有多少人
    复制代码

    强调:

    如果我们用unique的字段作为分组的依据,则每一条记录自成一组,这种分组没有意义
    多条记录之间的某个字段值相同,该字段通常用来作为分组的依据
    复制代码
    #!!!MySQL 5.7默认ONLY_FULL_GROUP_BY语义介绍
    #参考链接:http://www.ywnds.com/?p=8184 #分组查询的常见问题: mysql> select id,count from tt group by id; ERROR 1055 (42000): Expression #2 of SELECT list is not in GROUP BY clause and contains nonaggregated column 'test.tt.count' which is not functionally dependent on columns in GROUP BY clause; this is incompatible with sql_mode=only_full_group_by #查看MySQL 5.7默认的sql_mode如下: mysql> select @@global.sql_mode; ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION #去掉ONLY_FULL_GROUP_BY模式,如下操作: mysql> set global sql_mode='STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION'; #!!!注意 ONLY_FULL_GROUP_BY的语义就是确定select target list中的所有列的值都是明确语义,简单的说来,在ONLY_FULL_GROUP_BY模式下,target list中的值要么是来自于聚集函数的结果,要么是来自于group by list中的表达式的值。
    复制代码
    复制代码
    1. 查询岗位名以及岗位包含的所有员工名字
    2. 查询岗位名以及各岗位内包含的员工个数
    3. 查询公司内男员工和女员工的个数
    4. 查询岗位名以及各岗位的平均薪资
    5. 查询岗位名以及各岗位的最高薪资
    6. 查询岗位名以及各岗位的最低薪资
    7. 查询男员工与男员工的平均薪资,女员工与女员工的平均薪资
    复制代码
     
    #题1:分组
    mysql> select post,group_concat(name) from employee group by post;
    +-----------------------------------------+---------------------------------------------------------+
    | post                                    | group_concat(name)                                      |
    +-----------------------------------------+---------------------------------------------------------+
    | operation                               | 张野,程咬金,程咬银,程咬铜,程咬铁                        |
    | sale                                    | 歪歪,丫丫,丁丁,星星,格格                                |
    | teacher                                 | alex,wupeiqi,yuanhao,liwenzhou,jingliyang,jinxin,成龙   |
    | 老男孩驻沙河办事处外交大使              | egon                                                    |
    +-----------------------------------------+---------------------------------------------------------+
    
    
    #题目2:
    mysql> select post,count(id) from employee group by post;
    +-----------------------------------------+-----------+
    | post                                    | count(id) |
    +-----------------------------------------+-----------+
    | operation                               |         5 |
    | sale                                    |         5 |
    | teacher                                 |         7 |
    | 老男孩驻沙河办事处外交大使              |         1 |
    +-----------------------------------------+-----------+
    
    
    #题目3:
    mysql> select sex,count(id) from employee group by sex;
    +--------+-----------+
    | sex    | count(id) |
    +--------+-----------+
    | male   |        10 |
    | female |         8 |
    +--------+-----------+
    
    #题目4:
    mysql> select post,avg(salary) from employee group by post;
    +-----------------------------------------+---------------+
    | post                                    | avg(salary)   |
    +-----------------------------------------+---------------+
    | operation                               |  16800.026000 |
    | sale                                    |   2600.294000 |
    | teacher                                 | 151842.901429 |
    | 老男孩驻沙河办事处外交大使              |   7300.330000 |
    +-----------------------------------------+---------------+
    
    #题目5
    mysql> select post,max(salary) from employee group by post;
    +-----------------------------------------+-------------+
    | post                                    | max(salary) |
    +-----------------------------------------+-------------+
    | operation                               |    20000.00 |
    | sale                                    |     4000.33 |
    | teacher                                 |  1000000.31 |
    | 老男孩驻沙河办事处外交大使              |     7300.33 |
    +-----------------------------------------+-------------+
    
    #题目6
    mysql> select post,min(salary) from employee group by post;
    +-----------------------------------------+-------------+
    | post                                    | min(salary) |
    +-----------------------------------------+-------------+
    | operation                               |    10000.13 |
    | sale                                    |     1000.37 |
    | teacher                                 |     2100.00 |
    | 老男孩驻沙河办事处外交大使              |     7300.33 |
    +-----------------------------------------+-------------+
    
    #题目七
    mysql> select sex,avg(salary) from employee group by sex;
    +--------+---------------+
    | sex    | avg(salary)   |
    +--------+---------------+
    | male   | 110920.077000 |
    | female |   7250.183750 |
    +--------+---------------+

    七 使用聚合函数查询

    先from找到表

    再用where的条件约束去表中取出记录

    然后进行分组group by,没有分组则默认一组

    然后进行聚合

    最后select出结果

    复制代码
    示例:
        SELECT COUNT(*) FROM employee;
        SELECT COUNT(*) FROM employee WHERE depart_id=1;
        SELECT MAX(salary) FROM employee;
        SELECT MIN(salary) FROM employee;
        SELECT AVG(salary) FROM employee;
        SELECT SUM(salary) FROM employee;
        SELECT SUM(salary) FROM employee WHERE depart_id=3;
    复制代码

    八 HAVING过滤

    HAVING与WHERE在语法上是一样的

    select * from employee where salary > 10000;
    select * from employee having salary > 10000;

    HAVING与WHERE不一样的地方在于!!!!!!

    复制代码
    #!!!执行优先级从高到低:where > group by > 聚合函数 > having 
    #1. Where 是一个约束声明,使用Where约束来自数据库的数据,Where是在结果返回之前起作用的(先找到表,按照where的约束条件,从表(文件)中取出数据),Where中不能使用聚合函数。
    
    #2. Having是一个过滤声明,是在查询返回结果集以后对查询结果进行的过滤操作(先找到表,按照where的约束条件,从表(文件)中取出数据,然后group by分组,如果没有group by则所有记录整体为一组,然后执行聚合函数,然后使用having对聚合的结果进行过滤),在Having中可以使用聚合函数。
    
    #3. having可以放到group by之后,而where只能放到group by之前
    
    #4. 在查询过程中聚合语句(sum,min,max,avg,count)要比having子句优先执行。而where子句在查询过程中执行优先级高于聚合语句。
    复制代码

    验证不同之处

    复制代码
    #验证之前再次强调:执行优先级从高到低:where > group by > 聚合函数 > having 
    select count(id) from employee where salary > 10000; #正确,分析:where先执行,后执行聚合count(id),然后select出结果
    select count(id) from employee having salary > 10000;#错误,分析:先执行聚合count(id),后执行having过滤,无法对id进行salary>10000的过滤
    
    #以上两条sql的顺序是
    1:找到表employee--->用where过滤---->没有分组则默认一组执行聚合count(id)--->select执行查看组内id数目
    2:找到表employee--->没有分组则默认一组执行聚合count(id)---->having 基于上一步聚合的结果(此时只有count(id)字段了)进行salary>10000的过滤,很明显,根本无法获取到salary字段
    复制代码

    其他需要注意的问题

    select post,group_concat(name) from employee group by post having salary > 10000;#错误,分组后无法直接取到salary字段
    select post,group_concat(name) from employee group by post having avg(salary) > 10000;

    小练习:

    1. 查询各岗位内包含的员工个数小于2的岗位名、岗位内包含员工名字、个数
    3. 查询各岗位平均薪资大于10000的岗位名、平均工资
    4. 查询各岗位平均薪资大于10000且小于20000的岗位名、平均工资
    复制代码
    #题1:
    mysql> select post,group_concat(name),count(id) from employee group by post having count(id) < 2;
    +-----------------------------------------+--------------------+-----------+
    | post                                    | group_concat(name) | count(id) |
    +-----------------------------------------+--------------------+-----------+
    | 老男孩驻沙河办事处外交大使              | egon               |         1 |
    +-----------------------------------------+--------------------+-----------+
    
    #题目2:
    mysql> select post,avg(salary) from employee group by post having avg(salary) > 10000;
    +-----------+---------------+
    | post      | avg(salary)   |
    +-----------+---------------+
    | operation |  16800.026000 |
    | teacher   | 151842.901429 |
    +-----------+---------------+
    
    #题目3:
    mysql> select post,avg(salary) from employee group by post having avg(salary) > 10000 and avg(salary) <20000;
    +-----------+--------------+
    | post      | avg(salary)  |
    +-----------+--------------+
    | operation | 16800.026000 |
    +-----------+--------------+
    复制代码

    九 查询排序:ORDER BY

    复制代码
    按单列排序
        SELECT * FROM employee ORDER BY salary;
        SELECT * FROM employee ORDER BY salary ASC;
        SELECT * FROM employee ORDER BY salary DESC;
    
    按多列排序:先按照age排序,如果年纪相同,则按照薪资排序
        SELECT * from employee
            ORDER BY age,
            salary DESC;
    复制代码

    小练习:

    1. 查询所有员工信息,先按照age升序排序,如果age相同则按照hire_date降序排序
    2. 查询各岗位平均薪资大于10000的岗位名、平均工资,结果按平均薪资升序排列
    3. 查询各岗位平均薪资大于10000的岗位名、平均工资,结果按平均薪资降序排列
     
    #题目1
    mysql> select * from employee ORDER BY age asc,hire_date desc;
    
    #题目2
    mysql> select post,avg(salary) from employee group by post having avg(salary) > 10000 order by avg(salary) asc;
    +-----------+---------------+
    | post      | avg(salary)   |
    +-----------+---------------+
    | operation |  16800.026000 |
    | teacher   | 151842.901429 |
    +-----------+---------------+
    
    #题目3
    mysql> select post,avg(salary) from employee group by post having avg(salary) > 10000 order by avg(salary) desc;
    +-----------+---------------+
    | post      | avg(salary)   |
    +-----------+---------------+
    | teacher   | 151842.901429 |
    | operation |  16800.026000 |
    +-----------+---------------+

    十 限制查询的记录数:LIMIT

    复制代码
    示例:
        SELECT * FROM employee ORDER BY salary DESC 
            LIMIT 3;                    #默认初始位置为0 
        
        SELECT * FROM employee ORDER BY salary DESC
            LIMIT 0,5; #从第0开始,即先查询出第一条,然后包含这一条在内往后查5条
    
        SELECT * FROM employee ORDER BY salary DESC
            LIMIT 5,5; #从第5开始,即先查询出第6条,然后包含这一条在内往后查5条
    复制代码

    小练习:

    1. 分页显示,每页5条
    mysql> select * from  employee limit 0,5;
    +----+-----------+------+-----+------------+-----------------------------------------+--------------+------------+--------+-----------+
    | id | name      | sex  | age | hire_date  | post                                    | post_comment | salary     | office | depart_id |
    +----+-----------+------+-----+------------+-----------------------------------------+--------------+------------+--------+-----------+
    |  1 | egon      | male |  18 | 2017-03-01 | 老男孩驻沙河办事处外交大使              | NULL         |    7300.33 |    401 |         1 |
    |  2 | alex      | male |  78 | 2015-03-02 | teacher                                 |              | 1000000.31 |    401 |         1 |
    |  3 | wupeiqi   | male |  81 | 2013-03-05 | teacher                                 | NULL         |    8300.00 |    401 |         1 |
    |  4 | yuanhao   | male |  73 | 2014-07-01 | teacher                                 | NULL         |    3500.00 |    401 |         1 |
    |  5 | liwenzhou | male |  28 | 2012-11-01 | teacher                                 | NULL         |    2100.00 |    401 |         1 |
    +----+-----------+------+-----+------------+-----------------------------------------+--------------+------------+--------+-----------+
    rows in set (0.00 sec)
    
    mysql> select * from  employee limit 5,5;
    +----+------------+--------+-----+------------+---------+--------------+----------+--------+-----------+
    | id | name       | sex    | age | hire_date  | post    | post_comment | salary   | office | depart_id |
    +----+------------+--------+-----+------------+---------+--------------+----------+--------+-----------+
    |  6 | jingliyang | female |  18 | 2011-02-11 | teacher | NULL         |  9000.00 |    401 |         1 |
    |  7 | jinxin     | male   |  18 | 1900-03-01 | teacher | NULL         | 30000.00 |    401 |         1 |
    |  8 | 成龙       | male   |  48 | 2010-11-11 | teacher | NULL         | 10000.00 |    401 |         1 |
    |  9 | 歪歪       | female |  48 | 2015-03-11 | sale    | NULL         |  3000.13 |    402 |         2 |
    | 10 | 丫丫       | female |  38 | 2010-11-01 | sale    | NULL         |  2000.35 |    402 |         2 |
    +----+------------+--------+-----+------------+---------+--------------+----------+--------+-----------+
    rows in set (0.00 sec)
    
    mysql> select * from  employee limit 10,5;
    +----+-----------+--------+-----+------------+-----------+--------------+----------+--------+-----------+
    | id | name      | sex    | age | hire_date  | post      | post_comment | salary   | office | depart_id |
    +----+-----------+--------+-----+------------+-----------+--------------+----------+--------+-----------+
    | 11 | 丁丁      | female |  18 | 2011-03-12 | sale      | NULL         |  1000.37 |    402 |         2 |
    | 12 | 星星      | female |  18 | 2016-05-13 | sale      | NULL         |  3000.29 |    402 |         2 |
    | 13 | 格格      | female |  28 | 2017-01-27 | sale      | NULL         |  4000.33 |    402 |         2 |
    | 14 | 张野      | male   |  28 | 2016-03-11 | operation | NULL         | 10000.13 |    403 |         3 |
    | 15 | 程咬金    | male   |  18 | 1997-03-12 | operation | NULL         | 20000.00 |    403 |         3 |
    +----+-----------+--------+-----+------------+-----------+--------------+----------+--------+-----------+
    rows in set (0.00 sec)

    十一 使用正则表达式查询

    复制代码
    SELECT * FROM employee WHERE name REGEXP '^ale';
    
    SELECT * FROM employee WHERE name REGEXP 'on$';
    
    SELECT * FROM employee WHERE name REGEXP 'm{2}';
    
    
    小结:对字符串匹配的方式
    WHERE name = 'egon';
    WHERE name LIKE 'yua%';
    WHERE name REGEXP 'on$';
    复制代码

    小练习:

    查看所有员工中名字是jin开头,n或者g结果的员工信息
     
    select * from employee where name regexp '^jin.*[gn]$';
  • 相关阅读:
    Running APP 使用说明
    Android 控件八 WebView 控件
    Android 控件七 ImageView 控件
    Android 控件六 CheckBox 控件
    Android 控件五 RadioButton 控件
    Android 控件四 EditText 控件
    Android 控件三 TextView 控件实现 Button
    Android 控件二 Button
    Android 基础控件演示实例
    Android 控件一 TextView
  • 原文地址:https://www.cnblogs.com/Xanderzyl/p/10957694.html
Copyright © 2020-2023  润新知