ORACLE 中NUMBER类型默认的精度和Scale问题

在ORACLE数据库中，NUMBER(P,S)是最常见的数字类型，可以存放数据范围为10^-130~10^126（不包含此值)，需要1~22字节(BYTE)不等的存储空间。P 是Precison的英文缩写，即精度缩写，表示有效数字的位数，最多不能超过38个有效数字。S是Scale的英文缩写，表示从小数点到最低有效数字的位数，它为负数时，表示从最大有效数字到小数点的位数。有时候，我们在创建表的时候，NUMBER往往没有指定P，S的值，那么默认情况下，NUMBER的P、S的值分别是多少呢？相信这个问题能问倒一大片DBA。 在之前，我遇到了一个问题，总结整理在“ORACLE NUMBER类型Scale为0引发的问题”这篇博客当中，当时武断的判断“如果不指定p和s，NUMBER类型，它的默认精度值为38, 默认的scale值为0”，因为当时参考了官方文档https://docs.oracle.com/cd/B28359_01/server.111/b28318/datatype.htm#CNCPT1832

当然文档没有错误，文档应该是指在定义字段数据类型为NUMBER时，指定了NUMBER类型的P值，但是没有指定S的值，那么Scale默认就是0，如下测试所示，当时应该是我自己没有完全理解文档意思，当然文档也有误导的嫌疑。

SQL> drop table test;

Table dropped.

SQL> create table test(id number(38));

Table created.

SQL> insert into test

  2  select 123 from dual union all

  3  select 123.123 from dual;

2 rows created.

SQL> commit;

Commit complete.

SQL> select * from test;

        ID

----------

       123

       123

SQL> 

当在指定字段类型为NUMBER时，如果P和S都不指定，那么P和S又是什么值呢？今天特意实验验证了一下，具体实验过程如下：

SQL> drop table test;

Table dropped.

SQL> create table test(id  number, id1 number(38,4));

Table created.

SQL> insert into test                                      

  2  select 12, 12 from dual union all

  3  select 12.123456789, 12.123456789 from dual;

2 rows created.

SQL> commit;

Commit complete.

SQL> col id  for 999999999999.999999999999999999999999999999999999;

SQL> col id1 for 99999999999.9999999999999999999999999999999999999;

SQL> select * from test;

                                                ID                                                ID1

-------------------------------------------------- --------------------------------------------------

           12.000000000000000000000000000000000000           12.0000000000000000000000000000000000000

           12.123456789000000000000000000000000000           12.1235000000000000000000000000000000000

SQL> 

如上所示，当我插入上面两条记录后，发现如果不指定p和s，NUMBER类型，此时的Scale至少是9，我们继续测试，插入下面数据

SQL> insert into test

  2  select 12.123456789123456789123456789123456,

  3         12.123456789123456789123456789123456

  4  from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> select * from test;

                                                ID                                                ID1

-------------------------------------------------- --------------------------------------------------

           12.000000000000000000000000000000000000           12.0000000000000000000000000000000000000

           12.123456789000000000000000000000000000           12.1235000000000000000000000000000000000

           12.123456789123456789123456789123456000           12.1235000000000000000000000000000000000

如下所示，此时可以看到Scale的值33了，那么Scale的值是否可以继续变大呢？

SQL> insert into test

  2  select 12.123456789123456789123456789123456789123,

  3         12.123456789123456789123456789123456789123

  4  from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> select * from test;

                                                ID                                                ID1

-------------------------------------------------- --------------------------------------------------

           12.000000000000000000000000000000000000           12.0000000000000000000000000000000000000

           12.123456789000000000000000000000000000           12.1235000000000000000000000000000000000

           12.123456789123456789123456789123456000           12.1235000000000000000000000000000000000

           12.123456789123456789123456789123456789           12.1235000000000000000000000000000000000

如下截图所示，插入的记录为12.123456789123456789123456789123456789123，但是显示的值为12.123456789123456789123456789123456789，总共为38位，由于格式化列的缘故，可能导致部分小数位没有显示，

我们继续测试，调整格式化列，我们发现值变为了12.12345678912345678912345678912345678912，总共40位了，Scale的值为38了。这个是为什么呢？不是数字精度为38，意味着最多是38位吗？

SQL> col id  for 999999999999.99999999999999999999999999999999999999999

SQL> col id1 for 99999999999.999999999999999999999999999999999999999999

SQL> select * from test;

                                                     ID                                                     ID1

------------------------------------------------------- -------------------------------------------------------

           12.00000000000000000000000000000000000000000           12.000000000000000000000000000000000000000000

           12.12345678900000000000000000000000000000000           12.123500000000000000000000000000000000000000

           12.12345678912345678912345678912345600000000           12.123500000000000000000000000000000000000000

           12.12345678912345678912345678912345678912000           12.123500000000000000000000000000000000000000

继续其它测试，我们发现Sacle的值会随着小数点前面数字的变化而变化，如下所示：

SQL> insert into test

  2  select 123456789.123456789123456789123456789123456,

  3         123456789.123456789123456789123456789123456

  4  from dual;

1 row created.

SQL> commit;

Commit complete

SQL> insert into test

  2  select 123456789123.123456789123456789123456789123456,

  3         123456789123.123456789123456789123456789123456

  4  from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> select * from test;

                                                     ID                                                     ID1

------------------------------------------------------- -------------------------------------------------------

           12.00000000000000000000000000000000000000000           12.000000000000000000000000000000000000000000

           12.12345678900000000000000000000000000000000           12.123500000000000000000000000000000000000000

           12.12345678912345678912345678912345600000000           12.123500000000000000000000000000000000000000

           12.12345678912345678912345678912345678912000           12.123500000000000000000000000000000000000000

    123456789.12345678912345678912345678912300000000000    123456789.123500000000000000000000000000000000000000

 123456789123.12345678912345678912345678910000000000000 #######################################################

6 rows selected.

从上面测试可以看出，Scale的值是变化的，跟数据值有关系，目前看来，小数点前的数字位数和小数点后的数字位数相加为40（有时候又是39），为了测试是否这个规律，我特意用下面案例测试一下

SQL> create table test2(id number);

Table created.

SQL> insert into test2

  2  select 0.123456789123456789123456789123456789123456789 from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> col id for 9999999999.9999999999999999999999999999999999999999999999;

SQL> select * from test2;

                                                        ID

----------------------------------------------------------

           .1234567891234567891234567891234567891235000000

SQL> insert into test2

  2  select 123456789.123456789123456789123456789123456789 from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> select * from test2;

                                                        ID

----------------------------------------------------------

           .1234567891234567891234567891234567891235000000

  123456789.1234567891234567891234567891230000000000000000

SQL> insert into test2

  2  select 123456789123.123456789123456789123456789123456789 from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> select * from test2;

                                                        ID

----------------------------------------------------------

           .1234567891234567891234567891234567891235000000

  123456789.1234567891234567891234567891230000000000000000

##########################################################

SQL>  col id for 9999999999999.9999999999999999999999999999999999999999999999;

SQL> select * from test2;

                                                           ID

-------------------------------------------------------------

              .1234567891234567891234567891234567891235000000

     123456789.1234567891234567891234567891230000000000000000

  123456789123.1234567891234567891234567891000000000000000000

SQL> insert into test2

  2  select 12345678912345.12345678912345678912345678912345 from dual;

1 row created.

SQL> commit;

Commit complete.

SQL> select * from test2;

                                                           ID

-------------------------------------------------------------

              .1234567891234567891234567891234567891235000000

     123456789.1234567891234567891234567891230000000000000000

  123456789123.1234567891234567891234567891000000000000000000

#############################################################

SQL> col id for 9999999999999999999.99999999999999999999999999999999999999999999;

SP2-0246: Illegal FORMAT string "9999999999999999999.99999999999999999999999999999999999999999999"

SQL> col id for 9999999999999999999.9999999999999999999999999999999999999999

SQL> select * from test2;

                                                           ID

-------------------------------------------------------------

                    .1234567891234567891234567891234567891235

           123456789.1234567891234567891234567891230000000000

        123456789123.1234567891234567891234567891000000000000

      12345678912345.1234567891234567891234567900000000000000

SQL> 

这个问题纠结了很久，不明白为什么是39或40，后面在Oracle Database SQL Reference 10g Release 2终于找到解释了，如下所示：

p is the precision, or the total number of significant decimal digits, where the most

significant digit is the left-most nonzero digit, and the least significant digit is the

right-most known digit. Oracle guarantees the portability of numbers with

precision of up to 20 base-100 digits, which is equivalent to 39 or 40 decimal digits

depending on the position of the decimal point.

p是精度，或是有效十进制数的总位数。最大的有效数字是最左边的非零数字，而最小有效位是最右边的数字。 Oracle保证数字的可移植性

精度高达20 base-100 digits，相当于39位或40位十进制数字，取决于小数点的位置。