Mysql_Learning_Notes_系统结构_1

数据类型

整型

1.tinyint 1Bytes -128~127(255)
2.smallint 2Bytes -32768~32676(65535)
3.mdeiumint 3Bytes -8388608~8388607(16777215)
4.int 4Bytes -2147483648~2147483647(4294967295,42亿)
5.bigint 8Bytes -9223372036854775808~9223372036854775807(18446744073709551615)
int列如果是主键不能online DDL 转成bigint.

onlineDDL
:pt-osc(更容易导致主从数据延时,执行过程报错,无法从上一个位置开始,只能重头开始,pt 创建trigger 或者删除trigger 的时候有坑。 )和gh-ost(推荐,因基于binlog,可以随时停止或继续).推荐使用gh-st

int(11),11是修饰符,不是长度限制,int(8)zerofill zerofill也是修饰符,左侧补零.
溢出:
cast(9223372036854775807 as unsigned) 改为不符号

IPv4地址可以用INT存储:
select length('255.255.255.255')
+---------------------------+
| length('255.255.255.255') |
+---------------------------+
| 15 |
+---------------------------+
root@localhost [(none)]>select inet_aton('255.255.255.255'); //IPv4最大值,正好是int的无符号数最大值.
+------------------------------+
| inet_aton('255.255.255.255') |
+------------------------------+
| 4294967295 |
+------------------------------+
root@localhost [(none)]>select inet_ntoa(4294967295);
+-----------------------+
| inet_ntoa(4294967295) |
+-----------------------+
| 255.255.255.255 |
+-----------------------+
IPv6和IPv4共用的方法(数据类型为VARBINARY(16)，而不是BINARY(16)。唯一的原因是MySQL函数同时适用于IPv6和IPv4地址。 BINARY(16)只适用于存储IPv6地址，并保存一个字节。在处理IPv6和IPv4地址时应使用VARBINARY(16)):

root@localhost [(none)]>SELECT HEX(INET6_ATON('fdfe::5a55:caff:fefa:9089'));
+----------------------------------------------+
| HEX(INET6_ATON('fdfe::5a55:caff:fefa:9089')) |
+----------------------------------------------+
| FDFE0000000000005A55CAFFFEFA9089 |
+----------------------------------------------+
1 row in set (0.00 sec)
root@localhost [(none)]>SELECT HEX(INET6_ATON('192.168.9.1'));
+--------------------------------+
| HEX(INET6_ATON('192.168.9.1')) |
+--------------------------------+
| C0A80901 |
+--------------------------------+
1 row in set (0.00 sec)

mysql> SELECT INET6_NTOA(INET6_ATON('fdfe::5a55:caff:fefa:9089'));
-> 'fdfe::5a55:caff:fefa:9089'
mysql> SELECT INET6_NTOA(INET6_ATON('192.168.9.1'));
-> '192.168.9.1'

mysql> SELECT INET6_NTOA(UNHEX('FDFE0000000000005A55CAFFFEFA9089'));
-> 'fdfe::5a55:caff:fefa:9089'
mysql> SELECT INET6_NTOA(UNHEX('C0A80901'));
-> '192.168.9.1'

时间戳也是推荐使用无符号int也存储.
select unix_timestamp();
root@localhost [(none)]>select unix_timestamp();
+------------------+
| unix_timestamp() |
+------------------+
| 1539644023 |
+------------------+
1 row in set (0.00 sec)
root@localhost [(none)]>select from_unixtime(1539644023);
+---------------------------+
| from_unixtime(1539644023) |
+---------------------------+
| 2018-10-16 06:53:43 |
+---------------------------+
1 row in set (0.00 sec)

时间

优先使用:timestamp,其次datetime
timetamp\datetime 从5.6.6开始均支持自动更新为current_timestamp
日期转换:
CAST()datetime_col as DATE)
SELECT NOW()+0;

root@localhost [wenyz]>SELECT NOW()+0;
+----------------+
| NOW()+0 |
+----------------+
| 20181016075647 |
+----------------+
1 row in set (0.00 sec)

root@localhost [wenyz]>SELECT NOW();
+---------------------+
| NOW() |
+---------------------+
| 2018-10-16 07:56:54 |
+---------------------+
1 row in set (0.00 sec)

root@localhost [wenyz]>select CAST(now() as DATE)
-> ;
+---------------------+
| CAST(now() as DATE) |
+---------------------+
| 2018-10-16 |
+---------------------+
1 row in set (0.00 sec)

root@localhost [wenyz]>select CAST(now() as DATE);
+---------------------+
| CAST(now() as DATE) |
+---------------------+
| 2018-10-16 |
+---------------------+
1 row in set (0.00 sec)

root@localhost [wenyz]>select CAST(now() as time);
+---------------------+
| CAST(now() as time) |
+---------------------+
| 07:57:53 |
+---------------------+
1 row in set (0.00 sec)

root@localhost [wenyz]>select now(),now()+5;
+---------------------+----------------+
| now() | now()+5 |
+---------------------+----------------+
| 2018-10-16 07:59:36 | 20181016075941 |
+---------------------+----------------+
1 row in set (0.00 sec)

字符

1.char(M) Mw(字符集的单字符字节数) bytes,0<=M<=255 (utf8 2553,utf8.mb4:2554)
2.BINARY(M) M bytes,0<=M<=255
3.VARCHAR(M),VARBINARY(M) L+1 bytes if column values require 0-255 bytes if values may require more than 255 bytes,L+2bytes if values may require more than 255 bytes.(除Mw都需要额外1或2个节点存信息),实际存储长度超过255字节时,会被做TEXT处理.所有VARCHAR列的总可用长度是65535字节(实际上是65533)
字符集是utf8mb4时,实际可存储字符数是FLOOR(65533/4)=16384
4.mysql8.0默认字符集是utf8mb4

长文本\大对象

data type	storange required
tinyblob,tinytext	L+1bytes,where L<2^8
Blob,text	L+1bytes,where L<2^16
mediumblob,mediumtext	L+4bytes,where L<2^24
longlob,longtext	L+4bytes,where L<2^32

超长字段有可能会发生行溢出Off page (overflow)存储.一个page为16k,当一行超过8k时就会发现行溢出.
尽量不能,无法避免时则尽量独立表存放
对text列排序时,实际排序长是max_sort_length字节(默认1kb,排序时最多只考虑字段的前1kb)
实例:一个100G的表拆分成4个表后,总大小仅25G

浮点\小数位

FLOAT是单精度
用于存储一般精确度金额
事实上,更建议把金额转换成INT来存储,更精确.
FLOAT列类不指定精度时不能等值查询
- num FLOAT,where num=0.12,查询结果为空,需要用范围查询.
- num FLOAT(9,7),where num=0.12,可以查询到.
  DECIMAL
精确的十进制浮点型,高精度计算
M最大65(默认10),D最大30(默认0)
小数点后面的位数超限后,自动四舍五入(SQL_MODE=''时)

枚举,ENUM

ENMU(VALUE_LIST)

Value_list数量不超过255时,占用1个字节,超过时占2字节,最多65535个
表面上是用字符型表示,但实际底层采用INT来存储,显示时再转换为之前存储表示的值
所以,可以直接用TINYINT/SMALLINT代替(在单独的字典表里对应1表示xxx,2表示bbb,在需要枚举的列直接用数字来表示就可以了),
使用注意:千万不要用char或vchar来存储枚举.

类型选择建议:

最大限度减少IO请求
让数据表每行长度越小越好,可以避免行迁移或碎片
有TEXT/BLOB等大列时,尽量拆分到独立子表中,或者使用TokuDB引擎
经常更新的列和不经常更新的列分开不同的表存储.
更新数据也尽可能不要让长度变大.