KingbaseES避免表的重写与数据类型二进制兼容
一、关于KingbaseES变更表结构表的重写：
1.修改表结构可能会导致表进行重写（表OID发生变化）。
2.修改表结构带有索引或者字段类型长度或者精度操作时，会触发索引重建。
3.不修改列内容且旧类型与新类型二进制兼容（binary coercible），不需要重写表。
4.数据表字段类型长度或者精度由小变大的操作，不需要rewrite重写表。
5.数字类型int4到int8这种更改，需要重写数据表，主要是由于底层存储不一样。
详细实践过程参考：KingbaseES变更表结构表重写问题
二、避免表重写修改字段类型方法：
2.1 通过修改系统表避免表发生重写：
2.1.1 准备环境：
test=# create table t01(id numeric(6));
CREATE TABLE
test=# select sys_relation_filenode('t01');
SYS_RELATION_FILENODE
-----------------------
206188
(1 row)
test=# alter table t01 alter COLUMN id type numeric(9,3);
ALTER TABLE
test=# select sys_relation_filenode('t01');
SYS_RELATION_FILENODE
-----------------------
206191
(1 row)
2.1.2 由于numeric的类型在sys_attribute里面没有具体精度：
计算numeric类型精度：
test=# \d sys_attribute
View "sys_catalog.sys_attribute"
Column | Type | Collation | Nullable | Default
---------------+-----------+-----------+----------+---------
attrelid | oid | | |
attname | name | | |
atttypid | oid | | |
attstattarget | integer | | |
attlen | smallint | | |
attnum | smallint | | |
attndims | integer | | |
attcacheoff | integer | | |
atttypmod | integer | | |
attbyval | boolean | | |
attstorage | "char" | | |
attalign | "char" | | |
attnotnull | boolean | | |
atthasdef | boolean | | |
atthasmissing | boolean | | |
attidentity | "char" | | |
attgenerated | "char" | | |
attisdropped | boolean | | |
attislocal | boolean | | |
attinhcount | integer | | |
attcollation | oid | | |
attacl | aclitem[] | | |
attoptions | text[] | c | |
attfdwoptions | text[] | c | |
attmissingval | anyarray | | |
test=# select attname,atttypmod from sys_attribute where attrelid='t01'::regclass and attname='id';
ATTNAME | ATTTYPMOD
---------+-----------
id | 589831
(1 row)
# 使用以下SQL查看numeric类型的精度：
SELECT
CASE atttypid
WHEN 21 /*int2*/ THEN 16
WHEN 23 /*int4*/ THEN 32
WHEN 20 /*int8*/ THEN 64
WHEN 1700 /*numeric*/ THEN
CASE WHEN atttypmod = -1
THEN null
ELSE ((atttypmod - 4) >> 16) & 65535 -- 计算精度
END
WHEN 700 /*float4*/ THEN 24 /*FLT_MANT_DIG*/
WHEN 701 /*float8*/ THEN 53 /*DBL_MANT_DIG*/
ELSE null
END AS numeric_precision,
CASE
WHEN atttypid IN (21, 23, 20) THEN 0
WHEN atttypid IN (1700) THEN
CASE
WHEN atttypmod = -1 THEN null
ELSE (atttypmod - 4) & 65535 -- 计算标度
END
ELSE null
END AS numeric_scale
FROM sys_attribute
where attrelid = 't01'::regclass and attname = 'id';
test=# where attrelid = 't01'::regclass and attname = 'id';
NUMERIC_PRECISION | NUMERIC_SCALE
-------------------+---------------
9 | 3
attname：列名
atttypmod：记录了在表创建时提供的类型相关数据（例如一 个 varchar列的最大长度）。它会被传递给类型相关的输入函数和长度强制函数。对于那些不需要 atttypmod的类型，这个值通常总是为-1。
2.1.3 通过修改系统表数据实现字段类型的修改：
在kingbase.cong添加allow_system_table_dml=on参数，重启数据库
test=# select attname,atttypmod from sys_attribute where attrelid='t01'::regclass;
ATTNAME | ATTTYPMOD
----------+-----------
tableoid | -1
cmax | -1
xmax | -1
cmin | -1
xmin | -1
ctid | -1
id | 589831
(7 rows)
test=# select attname,atttypmod from sys_attribute where attrelid='t01'::regclass and attname='id';
ATTNAME | ATTTYPMOD
---------+-----------
id | 589831
(1 row)
# 计算字段ID的atttypmod值
589831 = 9*65,536+3+4
atttypmod值计算公式：atttypmod = precision * 65,536 + scale + 4
如果要修改为numeric(10,3)，计算一下：atttypmod = precision * 65,536 + scale + 4 = 655360 + 3 + 4 = 655367
2.1.3.1 无数据场景的修改：
# 修改t01表id字段数据类型为numeric(10,3)
test=# \d t01
Table "public.t01"
Column | Type | Collation | Nullable | Default
--------+--------------+-----------+----------+---------
id | numeric(9,3) | | |
test=# UPDATE sys_attribute SET atttypmod = 655367 WHERE attrelid = 't01'::regclass AND attname = 'id';
UPDATE 1
test=# \d t01
Table "public.t01"
Column | Type | Collation | Nullable | Default
--------+---------------+-----------+----------+---------
id | numeric(10,3) | | |
test=# select sys_relation_filenode('t01');
SYS_RELATION_FILENODE
-----------------------
206191
(1 row)
# 修改t01表id字段数据类型为numeric(6,5)
test=# \d t01
Table "public.t01"
Column | Type | Collation | Nullable | Default
--------+---------------+-----------+----------+---------
id | numeric(10,3) | | |
test=# UPDATE sys_attribute SET atttypmod = 393225 WHERE attrelid = 't01'::regclass AND attname = 'id';
UPDATE 1
test=# \d t01
Table "public.t01"
Column | Type | Collation | Nullable | Default
--------+--------------+-----------+----------+---------
id | numeric(6,5) | | |
test=# select sys_relation_filenode('t01');
SYS_RELATION_FILENODE
-----------------------
206191
(1 row)
通过测试可以发现，修改系统表更改数据类型，表oid不会发生改变，也就是表不会发生重写。
2.1.3.2 表有数据场景的修改：
test=# \d t01
Table "public.t01"
Column | Type | Collation | Nullable | Default
--------+--------------+-----------+----------+---------
id | numeric(6,5) | | |
test=# insert into t01 select generate_series(1,5) from dual;
INSERT 0 5
test=# select * from t01;
ID
---------
1.00000
2.00000
3.00000
4.00000
5.00000
(5 rows)
# 修改t01表id字段数据类型为numeric(8,3)
test=# UPDATE sys_attribute SET atttypmod = 524295 WHERE attrelid = 't01'::regclass AND attname = 'id';
UPDATE 1
test=# \d t01
Table "public.t01"
Column | Type | Collation | Nullable | Default
--------+--------------+-----------+----------+---------
id | numeric(8,3) | | |
test=# select * from t01;
ID
---------
1.00000
2.00000
3.00000
4.00000
5.00000
(5 rows)
test=# insert into t01 values(6);
INSERT 0 1
test=# select * from t01;
ID
---------
1.00000
2.00000
3.00000
4.00000
5.00000
6.000
(6 rows)
test=# select sys_relation_filenode('t01');
SYS_RELATION_FILENODE
-----------------------
206191
(1 row)
通过测试可以发现，修改系统表更改数据类型，表oid不会发生改变并且已存在的数据数据类型不会发生变化，只有新插入的数据才可以使用修改后的数据类型。
注意：
不推荐在生产系统使用这种方式进行表结构的修改，本文只做研究学习使用。
三、数据类型的二进制兼容：
二进制可兼容表示该转换可以被“免费”执行而不用调用任何函数。相应的数据类型值使用同样的内部表示。
KingbaseES内置转换函数，Function列有binary coercible信息说明是二进制兼容的：
test=# \dC varchar
List of casts
Source type | Target type | Function | Implicit?
-------------+-----------------------------+--------------------+---------------
boolean | varchar | text | in assignment
bpchar | varchar | text | yes
bpcharbyte | varchar | (binary coercible) | yes
"char" | varchar | text | in assignment
cidr | varchar | text | in assignment
dsinterval | varchar | dsinterval_text | in assignment
inet | varchar | text | in assignment
name | varchar | varchar | in assignment
text | varchar | (binary coercible) | yes
tinyint | varchar | text | in assignment
varchar | bigint | int8 | yes
varchar | bpchar | (binary coercible) | in assignment
varchar | bpcharbyte | (binary coercible) | in assignment
varchar | bytea | (with inout) | yes
varchar | "char" | char | in assignment
varchar | date | text_date | yes
varchar | double precision | float8 | yes
varchar | dsinterval | to_dsinterval | in assignment
varchar | integer | int4 | yes
varchar | name | name | yes
varchar | numeric | text_numeric | yes
varchar | real | float4 | yes
varchar | regclass | regclass | yes
varchar | smallint | int2 | yes
varchar | text | (binary coercible) | yes
varchar | timestamp without time zone | text_timestamp | yes
varchar | timestamp with time zone | timestamptz | yes
varchar | time without time zone | text_time | yes
varchar | time with time zone | text_timetz | yes
varchar | tinyint | tinyint | yes
varchar | varchar | varchar | yes
varchar | varcharbyte | (binary coercible) | yes
varchar | xml | xml | in assignment
varchar | yminterval | to_yminterval | in assignment
varcharbyte | varchar | (binary coercible) | in assignment
xml | varchar | (binary coercible) | in assignment
yminterval | varchar | yminterval_text | in assignment
(37 rows)
# 也可以使用以下语句查询
SELECT format_type(castsource, NULL) AS "Source type",
format_type(casttarget, NULL) AS "Target type",
CASE WHEN c.castmethod = 'b' THEN '(binary coercible)'
WHEN c.castmethod = 'i' THEN '(with inout)'
ELSE p.proname
END AS "Function",
CASE WHEN c.castcontext = 'e' THEN 'no'
WHEN c.castcontext = 'a' THEN 'in assignment'
ELSE 'yes'
END AS "Implicit?",
d.description AS "Description"
FROM sys_cast c LEFT JOIN sys_proc p
ON c.castfunc = p.oid
LEFT JOIN sys_type ts
ON c.castsource = ts.oid
LEFT JOIN sys_namespace ns
ON ns.oid = ts.typnamespace
LEFT JOIN sys_type tt
ON c.casttarget = tt.oid
LEFT JOIN sys_namespace nt
ON nt.oid = tt.typnamespace
LEFT JOIN sys_description d
ON d.classoid = c.oid AND d.objoid = c.oid AND d.objsubid = 0
WHERE ( (true AND sys_type_is_visible(ts.oid)
) OR (true AND sys_type_is_visible(tt.oid)
) ) ORDER BY 1, 2;
列Source type：表示源数据类型。
列Target type：表示目标数据类型。
列Function（根据 sys_cast表castmethod字段得到）：表示如何进行类型转换。
sys_cast.castmethod字段说明：
| 值 | 说明 |
|---|---|
| f | 表明使用castfunc中指定的函数 |
| i | 表明使用输入/输出函数 |
| b | 表明该类型是二进制兼容的，因此不需要转换 |
列Implicit（根据sys_cast表castcontext字段得到）：有以下三种情况
sys_cast.castcontext 字段说明：
| 值 | 说明 |
|---|---|
| assignment | castcontext = a，表示在赋值给目标列时隐式调用， 和显式调用一样。也就是说在赋值时自动对类型进行转换 |
| no | pg_cast.castcontext = e，表示仅能作为一个显式转换（使用CAST或::语法） |
| yes | pg_cast.castcontext = i，表示在表达式中隐式调用，和其他转换一样 |
2.1 测试过程：
2.1.1 使用cast强制转换数据类型：
"char"转化为integer，castcontext为e，表示不会隐式调用，不能对类型进行自动转换。
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from sys_cast where castsource= '"char"'::regtype and casttarget = 'integer'::regtype;
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
------------+------------+-----------------+-------------+------------
"char" | integer | pg_catalog.int4 | e | f
(1 row)
test=#
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from sys_cast where CASTCONTEXT='e';
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
---------------------+------------------+----------------------+-------------+------------
"char" | integer | pg_catalog.int4 | e | f
integer | "char" | pg_catalog."char" | e | f
lseg | point | pg_catalog.point | e | f
path | point | pg_catalog.point | e | f
box | point | pg_catalog.point | e | f
box | lseg | pg_catalog.lseg | e | f
box | circle | pg_catalog.circle | e | f
polygon | point | pg_catalog.point | e | f
polygon | box | pg_catalog.box | e | f
polygon | circle | pg_catalog.circle | e | f
circle | point | pg_catalog.point | e | f
circle | box | pg_catalog.box | e | f
circle | polygon | pg_catalog.polygon | e | f
bigint | bit | pg_catalog."bit" | e | f
integer | bit | pg_catalog."bit" | e | f
bit | bigint | pg_catalog.int8 | e | f
bit | integer | pg_catalog.int4 | e | f
jsonb | boolean | pg_catalog.bool | e | f
jsonb | numeric | pg_catalog."numeric" | e | f
jsonb | smallint | pg_catalog.int2 | e | f
jsonb | integer | pg_catalog.int4 | e | f
jsonb | bigint | pg_catalog.int8 | e | f
jsonb | real | pg_catalog.float4 | e | f
jsonb | double precision | pg_catalog.float8 | e | f
pg_catalog.interval | dsinterval | interval_dsinterval | e | f
(25 rows)
# 创建测试表并插入数据
test=# create table t01(id "char",i_id integer );
CREATE TABLE
test=# insert into t01 values('1',"char" '1');
ERROR: column "i_id" is of type integer but expression is of type "char"
LINE 1: insert into t01 values('1',"char" '1');
^
HINT: You will need to rewrite or cast the expression.
# 使用cast强制转换
test=# insert into t01 values('1',cast(("char" '1') as integer));
INSERT 0 1
2.1.2 修改系统表自动转换数据类型：
"char"转换为integer，修改castcontext为a，表示在赋值给目标列时隐式调用，自动对类型进行转换。
test=# begin;
BEGIN
test=# update sys_cast set castcontext = 'a' where castsource= '"char"'::regtype and casttarget = 'integer'::regtype;
UPDATE 1
test=# insert into t01 values('1',"char" '1');
INSERT 0 1
test=# rollback;
ROLLBACK
test=# insert into t01 values('1',"char" '1');
ERROR: column "i_id" is of type integer but expression is of type "char"
LINE 1: insert into t01 values('1',"char" '1');
^
HINT: You will need to rewrite or cast the expression.
test=#
修改系统表之后，就可以正常插入数据不报错。
2.1.3 不在系统表显示的内置转换：
需要注意的是，sys_cast并不表示系统知道如何执行的所有类型转换，它只包括哪些不能从某些普通规则推导出的转换。
例如，一个域及其基类型之间的转换并未显式地在pg_cast中展示。
另一个重要的例外是"自动 I/O转换造型"，它们通过数据类型自己的I/O函数来转换成（或者转换自）text或其他字符串类型，这些转换也没有显式地在sys_cast中表示。
例如：date、timestamp转换为varchar
test=# \dC date
List of casts
Source type | Target type | Function | Implicit?
-----------------------------+-------------+---------------------+-----------
bpchar | date | text_date | yes
text | date | text_date | yes
timestamp without time zone | date | timestamp_to_date | yes
timestamp with time zone | date | timestamptz_to_date | yes
varchar | date | text_date | yes
(5 rows)
test=# \dC timestamp
List of casts
Source type | Target type | Function | Implicit?
-----------------------------+-----------------------------+-----------------------+---------------
bpchar | timestamp without time zone | text_timestamp | yes
bpcharbyte | timestamp without time zone | text_timestamp | yes
integer | timestamp without time zone | cast_int_to_timestamp | in assignment
pg_catalog.date | timestamp without time zone | timestamp | yes
text | timestamp without time zone | text_timestamp | yes
timestamp without time zone | date | timestamp_to_date | yes
timestamp without time zone | pg_catalog.date | date | in assignment
timestamp without time zone | text | text | yes
timestamp without time zone | timestamp without time zone | timestamp | yes
timestamp without time zone | timestamp with time zone | timestamptz | yes
timestamp without time zone | time without time zone | time | in assignment
timestamp with time zone | timestamp without time zone | timestamp | in assignment
varchar | timestamp without time zone | text_timestamp | yes
varcharbyte | timestamp without time zone | text_timestamp | yes
(14 rows)
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from sys_cast where castsource= 'date'::regtype and casttarget ='varchar'::regtype;
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
------------+------------+----------+-------------+------------
(0 rows)
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from sys_cast where castsource= 'timestamp(0)'::regtype and casttarget ='varchar'::regtype;
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
------------+------------+----------+-------------+------------
(0 rows)
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from sys_cast where castsource= 'timestamp'::regtype and casttarget ='varchar'::regtype;
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
------------+------------+----------+-------------+------------
(0 rows)
test=# create table t01(dat varchar(20));
CREATE TABLE
test=# insert into t01 values(date '2023-04-11');
INSERT 0 1
test=# insert into t01 values(timestamp(0) '2023-04-11');
INSERT 0 1
test=# insert into t01 values(timestamp '2023-04-11');
INSERT 0 1
虽然date、timestamp转换为varchar在sys_cast系统表查询不到，但是测试过程种并不需要使用cast进行数据类型的强制转换。数据类型通过自己的I/O函数来进行转换。
2.1.4 使用内置的IO函数自定义数据类型转换：
语法：
test=# \h create cast
Command: CREATE CAST
Description: define a new cast
Syntax:
CREATE CAST (source_type AS target_type)
WITH FUNCTION function_name [ (argument_type [, ...]) ]
[ AS ASSIGNMENT | AS IMPLICIT ]
CREATE CAST (source_type AS target_type)
WITHOUT FUNCTION
[ AS ASSIGNMENT | AS IMPLICIT ]
CREATE CAST (source_type AS target_type)
WITH INOUT
[ AS ASSIGNMENT | AS IMPLICIT ]
创建自定义类型转换函数：
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from pg_cast where castsource= '"char"'::regtype and casttarget = 'bigint'::regtype;
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
------------+------------+----------+-------------+------------
(0 rows)
test=# create cast ("char" as bigint) with inout as assignment;
CREATE CAST
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from pg_cast where castsource= '"char"'::regtype and casttarget = 'bigint'::regtype;
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
------------+------------+----------+-------------+------------
"char" | bigint | - | a | i
(1 row)
test=# create table t01(id bigint);
CREATE TABLE
test=# insert into t01 values ("char" '123');
INSERT 0 1
test=# insert into t01 values ("char" '1234');
INSERT 0 1
test=# drop cast("char" as bigint);
DROP CAST
test=# insert into t01 values ("char" '12345');
ERROR: column "id" is of type bigint but expression is of type "char"
LINE 1: insert into t01 values ("char" '12345');
^
HINT: You will need to rewrite or cast the expression.
通过使用WITH INOUT语法来定义一个I/O转换映射。一个I/O转换是通过调用源数据类型的输出函数，并将结果字符串传递给目标数据类型的输入函数来执行的。在常用的场景下，这个功能避免了为数据类型转换编写一个单独的转换函数。I/O转换的行为与基于常规函数的转换相同，只是实现方式不同。
每个数据类型都有一个内置的I/O转换函数，如varchar、text..
test=# \df textin
List of functions
Schema | Name | Result data type | Argument data types | Type
------------+--------+------------------+---------------------+------
pg_catalog | textin | text | cstring | func
(1 row)
test=# \df textout
List of functions
Schema | Name | Result data type | Argument data types | Type
------------+---------+------------------+---------------------+------
pg_catalog | textout | cstring | text | func
(1 row)
test=# \df varcharin
List of functions
Schema | Name | Result data type | Argument data types | Type
------------+-----------+------------------+-----------------------+------
pg_catalog | varcharin | varchar | cstring, oid, integer | func
(1 row)
test=# \df varcharout
List of functions
Schema | Name | Result data type | Argument data types | Type
------------+------------+------------------+---------------------+------
pg_catalog | varcharout | cstring | varchar | func
(1 row)
varcharin表示用于转化为内部的表示格式，varcharout表示用于将内部格式转化为想要的格式。
2.1.5 二进制兼容数据类型简单整理：
数据类型转换重写与不重写：
- varchar(x) 转换到 varchar(y) 当 y>=x，不需要重写。
- numeric(x,z) 转换到 numeric(y,z) 当 y>=x，或者不指定精度类型，不需要重写。
- numeric(x,c) 转换到 numeric(y,z) 当 y=x c>z，当numeric数据类型标度不一致时，需要重写。
- varbit(x) 转换到 varbit(y) 当 y>=x，不需要重写。
- timestamp(x) 转换到 timestamp(y) 当 y>=x，或者转换为timestamp，不需要重写。
- timestamptz(x) 转换到 timestamptz(y) 当 y>=x，或者转换为timestamptz，不需要重写。
- interval(x) 转换到 interval(y) 当 y>=x ，或者转换为interval，不需要重写。
- timestamp 转换到 text、varchar、varchar(n)，char(n)，需要重写。
- timestamp(x)转换到 text、varchar、varchar(n)、char(n)，n>=x，需要重写。
- text 转换到 char、char(x)、varchar(n)，需要重写。
- text 转换到 varchar，不需要重写。
- numeric(x) 转换到 numeric(y)，y>=x，不需要重写。
- numeric(x) 转换到 numeric，不需要重写。
- numeric(x,y) 转换到 numeric，不需要重写。
KingbaseES全部的二进制兼容类型：
test=# select castsource::regtype,casttarget::regtype,castfunc::regproc,castcontext,castmethod from sys_cast where castmethod='b';
CASTSOURCE | CASTTARGET | CASTFUNC | CASTCONTEXT | CASTMETHOD
-----------------+---------------------+----------+-------------+------------
integer | oid | - | i | b
integer | regconfig | - | i | b
integer | regoperator | - | i | b
integer | regtype | - | i | b
integer | regclass | - | i | b
integer | regprocedure | - | i | b
integer | regproc | - | i | b
integer | regoper | - | i | b
integer | regnamespace | - | i | b
integer | regrole | - | i | b
integer | regdictionary | - | i | b
regproc | oid | - | i | b
regproc | regprocedure | - | i | b
regproc | integer | - | a | b
text | bpchar | - | i | b
text | varcharbyte | - | i | b
text | varchar | - | i | b
text | bpcharbyte | - | i | b
oid | regoperator | - | i | b
oid | integer | - | a | b
oid | regproc | - | i | b
oid | regprocedure | - | i | b
oid | regoper | - | i | b
oid | regclass | - | i | b
oid | regtype | - | i | b
oid | regconfig | - | i | b
oid | regdictionary | - | i | b
oid | regrole | - | i | b
oid | regnamespace | - | i | b
xml | text | - | a | b
xml | varchar | - | a | b
xml | bpchar | - | a | b
pg_node_tree | text | - | i | b
cidr | inet | - | i | b
bpchar | text | - | i | b
bpchar | varcharbyte | - | i | b
bpchar | bpcharbyte | - | i | b
varchar | text | - | i | b
varchar | bpcharbyte | - | a | b
varchar | varcharbyte | - | i | b
varchar | bpchar | - | a | b
bit | bit varying | - | i | b
bit varying | bit | - | i | b
regprocedure | regproc | - | i | b
regprocedure | integer | - | a | b
regprocedure | oid | - | i | b
regoper | integer | - | a | b
regoper | regoperator | - | i | b
regoper | oid | - | i | b
regoperator | integer | - | a | b
regoperator | oid | - | i | b
regoperator | regoper | - | i | b
regclass | oid | - | i | b
regclass | integer | - | a | b
regtype | integer | - | a | b
regtype | oid | - | i | b
pg_ndistinct | bytea | - | i | b
pg_dependencies | bytea | - | i | b
regconfig | oid | - | i | b
regconfig | integer | - | a | b
regdictionary | oid | - | i | b
regdictionary | integer | - | a | b
regnamespace | integer | - | a | b
regnamespace | oid | - | i | b
regrole | integer | - | a | b
regrole | oid | - | i | b
pg_mcv_list | bytea | - | i | b
yminterval | pg_catalog.interval | - | a | b
bpcharbyte | bpchar | - | a | b
bpcharbyte | varchar | - | i | b
bpcharbyte | text | - | i | b
bpcharbyte | varcharbyte | - | i | b
varcharbyte | bpchar | - | a | b
varcharbyte | bpcharbyte | - | a | b
varcharbyte | varchar | - | a | b
varcharbyte | text | - | i | b
(76 rows)
在KingbaseES变更表结构可以参考以上内容。