Copyright © 2022 PawSQL
异构数据库的迁移(譬如从Oracle迁移到openGauss)工作主要包括三个方面,
目前对于数据库对象以及数据内容的迁移有很多成熟的工具,而对于应用迁移的工具却很少能够见到。原因是因为DML语句比DDL复杂的多,不同的数据库语法差异也比较大。目前市场上的迁移工具大多使用正则表达式来解析SQL语句,而DML语句的复杂性导致此类工具的解析成功率较低,难以作为一个成熟地商业产品进行推广。
PawSQL团队开发的DML语法转换工具Ora2ogSQL,通过PawSQL强大的SQLParser,能够解析几乎所有的Oracle语法,并将其转换为对应的openGauss语法,支持数据库应用的平滑迁移。
本手册介绍了Oracle和openGauss的语法区别,以及转换映射关系,可以作为迁移人员的SQL迁移参考手册。
本手册描述了PawSQL Ora2ogSQL内部的实现逻辑,PawSQL Ora2ogSQL能够帮助SQL迁移人员自动识别不兼容的语法,并完成语法转换。
Oracle获取一个常量需要通过一个dual,Opengauss不需要
编号 | Oracle | Opengauss |
---|---|---|
1 | select 2 from dual | select 2 |
对于查询返回的每行数据,rownum虚拟列会返回一个数字,第一行的ROWNUM为1,第二行为2,以此类推。
编号 | Oracle | Opengauss |
---|---|---|
1 | select rownum from customer; | select row_number() over () as rownum from customer |
2 | select tableoid from customer where rownum < 10 and rownum >= 2; | select tableoid from customer limit 9 OFFSET 2 |
3 | select c_name from customer where rownum < 10 and c_phone = ‘111’; | select customer.c_name from customer where customer.c_phone = ‘111’ limit 9 |
4 | select * from customer where rownum between 1 and 10; | select tableoid from customer limit 10 OFFSET 0 |
Oracle中的rowid虚拟列返回特定行的具体地址,在Opengauss中重写为tableoid || '#' || ctid
编号 | Oracle | Opengauss |
---|---|---|
1 | select rowid, c.* from customer c; | select tableoid || ‘#’ || ctid, c.* from customer as c |
Oracle中的nvl(col, value)用来设置默认值,col为空就设置为value;
在Opengauss中重写为coalesce
编号 | Oracle | Opengauss |
---|---|---|
1 | select nvl(c_phone, 1) from customer; | select coalesce(customer.c_phone, ‘1’) from customer |
nvl2对col的null值进行处理,如果col为null,则返回v1, 否则返回v2;
postgre中没有类似的函数,可以重写为case… when…
编号 | Oracle | Opengauss |
---|---|---|
1 | select nvl2(c_phone, 1, 2) from customer; | select case when c_phone is null then 1 else 2 end from customer |
Oracle中的decode(arg1, arg2, arg3, arg4)函数, 表示当 arg1 等于 arg2 时,取 arg3,否则取 arg4。
postgre中没有类似的函数,可以重写为case… when…
编号 | Oracle | Opengauss |
---|---|---|
1 | select decode(c_phone,‘110’, 1 , 2) from customer; | select case when c_phone = ‘110’ then 1 else 2 end from customer |
2 | select decode(c_phone,null, 1 , 2) from customer; | select case when c_phone is null then 1 else 2 end from customer |
Oracle中的substr用来取一个字符串的子串,Opengauss有同名的函数实现类似功能。不同的是Oracle中,第二、第三个参数可以为负数,代表从后面进行计数,Opengauss不允许其为负数,需对其进行转换。Oracle中是以0开始计数,Opengauss以1开始计数(需确认)。
编号 | Oracle | Opengauss |
---|---|---|
1 | select substr(c_phone, 1 , -2 ) from customer; | select substr(c_phone, 1, length(c_phone) - 2) from customer |
2 | select substr(c_phone, -3 , 1 ) from customer; | select substr(c_phone, length(c_phone) - 3, 1) from customer |
Oracle中的instr用来取一个字符串的子串位置,当其只有两个参数时,表示子串的第一次出现的位置,和Opengauss中对应的函数为strpos。当其有多个参数时,无对应函数。
编号 | Oracle | Opengauss |
---|---|---|
1 | select instr(‘123’, ‘23’) | select strpos(‘123’, ‘23’) |
在Oracle中,replace()函数用于替换字符串, replace(srcstr, oldsub[, newsub ] ),和Opengauss中的replace函数用法基本一致。只是需要注意在Oracle中无第三个参数时,代表删除此字符,在Opengauss可将第三个参数设置为’'。
编号 | Oracle | Opengauss |
---|---|---|
1 | select replace(‘123’,‘1’); | select replace(‘123’,‘1’,‘’); |
Oracle里的stragg函数实现在分组内对列值的拼接,它和listagg类似,但是不可以指定拼接的顺序。在Opengauss中,可以使用string_agg函数来替换。其第二个参数可选,默认值为’',在Opengauss需补充第二个参数。
编号 | Oracle | Opengauss |
---|---|---|
1 | select listagg(c_name,‘,’) as name from customer group by c_phone | select string_agg(c_name,‘,’) as name from customer group by c_phone |
2 | select listagg(c_name) as name from customer group by c_phone | select listagg(c_name,‘’) as name from customer group by c_phone |
Oracle里的listagg函数实现对列值的拼接,它可以在分组内以指定顺序对非分组列进行拼接。在Opengauss中,可以使用string_agg函数来实现,需注意语法方面也有区别. 另外,其第二个参数可选,默认值为’',在Opengauss需补充第二个参数。
当没有group by子句时,可以使用over(partiton by… order by…)进行替换
当指定group by子句时,它的重写算法比较复杂
编号 | Oracle | Opengauss |
---|---|---|
1 | select listagg(c_name,‘,’) within group(order by c_name) over (partition by c_phone) as name from customer; | sselect string_agg(customer.c_name, ‘,’) over (partition by customer.c_phone order by c_custkey) as name from customer |
2 | select listagg(c_name,‘,’) within group(order by c_name) as name from customer group by c_phone; | select max(paw_dt.name) as name from (select string_agg(customer.c_name, ‘,’) over (partition by customer.c_phone order by c_name) as name, customer.c_phone from customer) as paw_dt group by c_phone |
3 | select listagg(c_name,‘,’) within group(order by c_name) as name from customer group by c_phone | select string_agg(c_name,‘,’) as name from customer group by c_phone |
Oracle中的sysdate()/sysdate返回系统当前时间(日期+时分秒),在Opengauss中对应now()或是current_timestamp(日期+时分秒+毫秒)。
Oracle中的systimestamp返回系统当前时间戳(日期+时分秒+毫秒),在Opengauss中对应now()或是current_timestamp。
编号 | Oracle | Opengauss |
---|---|---|
1 | select sysdate | select current_timestamp |
2 | select sysdate() | select now() |
3 | select systimestamp | select current_timestamp |
Oracle中的to_date返回的是时间类型,而在Opengauss中to_date是日期类型,所以Oracle中的to_date在Opengauss中应该对应to_timestamp
编号 | Oracle | Opengauss |
---|---|---|
1 | select to_date( endTime ,‘yyyy-mm-dd hh24:mi:ss’) from t | select to_timestamp( endTime ,‘yyyy-mm-dd hh24:mi:ss’) from t |
在Oracle中trunc函数有两种用法
第一种是对数字进行截取, trunc(num,[int]); 是去掉数字num小数位以后的部分,并且不进行四舍五入。这种用法和在Opengauss的trunc用法一致,不需要转换
trunc函数的第二种用法是对日期进行提取,trunc(date,[fmt])。这种用法在Opengauss对应的函数是date_trunc(fmt, date),需注意在Opengauss中fmt是第一个参数,且不可省略。
编号 | Oracle | Opengauss |
---|---|---|
1 | select trunc( 111.23,2) | select trunc( 111.23,2) |
2 | select trunc(sysdate,‘year’) | select date_trunc(‘year’, current_timestamp) |
3 | select trunc(sysdate) | select date_trunc(‘dd’, current_timestamp) |
Oracle中的add_months 函数主要是对日期函数进行操作,对日期按月增加。在Opengauss没有对应的函数,需将其转化为基于日期和interval的运算。
编号 | Oracle | Opengauss |
---|---|---|
1 | select add_months(sysdate, 2) | select current_timestamp + 2 * interval ‘1 month’ |
Oracle中的last_day返回指定日期所在月份的最后一天; 在Opengauss没有对应的函数,需将其转化为基于日期和interval的运算。
编号 | Oracle | Opengauss |
---|---|---|
1 | select add_months(sysdate, 2) | select cast(date_trunc(‘MONTH’, current_timestamp) + interval ‘1 MONTH - 1 DAY’ as date) |
Oracle允许HAVING在GROUP BY子句之前或之后。在Opengauss中,HAVING子句必须出现在GROUP BY子句后面。
编号 | Oracle | Opengauss |
---|---|---|
1 | select c_name from customer having count(*) > 2 group by c_name | select c_name from customer group by c_name having count(*) > 2 |
Oracle中单表引用允许使用括号括起来,Opengauss不允许。
编号 | Oracle | Opengauss |
---|---|---|
1 | SELECT * FROM (CUSTOMER); | SELECT * FROM CUSTOMER; |
Oracle中允许使用UNIQUE进行去重,在Opengauss中迁移为DISTINCT关键字
编号 | Oracle | Opengauss |
---|---|---|
1 | select unique c_phone from customer | select distinct customer.c_phone from customer |
Oracle中可以使用minus关键字来取两个结果集的差,在Opengauss中需迁移为except.
编号 | Oracle | Opengauss |
---|---|---|
1 | select c_custkey from customer minus select o_custkey from orders | select c_custkey from customer except select o_custkey from orders |
Oracle的delete语句的FROM关键字可以省略,迁移至Opengauss需补充上。
编号 | Oracle | Opengauss |
---|---|---|
1 | delete customer where 1=0; | delete from customer where 1 = 0 |
Oracle在执行INSERT语句时,可以通过指定NOLOGGING关键字来减少日志记录,提升操作性能。Opengauss不支持此关键字。
编号 | Oracle | Opengauss |
---|---|---|
1 | insert into customer nologging select * from customer_bk; | insert into customer select * from customer_bk; |
INSERT INTO 后面不需要添加as关键字,insert into ... as select...
修改为insert into... select...
编号 | Oracle | Opengauss |
---|---|---|
1 | insert into t as select c1 from t1 | insert into t select c1 from t1 |
Oracle中在不引起歧义的情况下子查询可以不带别名,而在Opengauss中,所有的FROM子查询都必须带有别名
编号 | Oracle | Opengauss |
---|---|---|
1 | select * from (select * from CUSTOMER) | select * from (select * from CUSTOMER) as foo |
在Opengauss中,Update的时候,更新列不允许添加表名前缀。
编号 | Oracle | Opengauss |
---|---|---|
1 | update customer c set c.c_name = ‘xxx’ where c_custkey = 1; | update customer set c_name = ‘xxx’ where c_custkey = 1 |
在Oracle中,外连接可以通过在条件上添加(+)来定义, 连接符(+)跟在哪个条件后面就是哪张表被左连。在Opengauss中,需将其重写为标准的外连接语法。
编号 | Oracle | Opengauss |
---|---|---|
1 | select * from customer, orders where c_custkey = o_custkey(+) | select * from customer left outer join orders on c_custkey = o_custkey |
2 | select * from customer, orders where c_custkey(+) = o_custkey and c_name(+) = o.o_clerk and o_custkey>100 | select * fromcustomer right outer join orders on (c_custkey = o_custkey and c_name = o_clerk) where o_custkey > 100 |
Oracle中,CONNECT BY 用于存在上下级等层级关系的数据表进行递归查询。语法格式: START WITH condition1 CONNECT BY [ NOCYCLE ] condition2。在Opengauss通过Recursive Common Table Expression来实现此功能,主要是把START WITH… CONNECT BY Prior拆成两个部分,查询表一致,但条件不一致,用UNION ALL合并.
编号 | Oracle | Opengauss |
---|---|---|
1 | select id from city_branch start with id=rolebranchid connect by prior id=parent_id; | with RECURSIVE MIG_CTE as ( select id, 1 as level from city_branch where id = rolebranchid union all select id, level + 1 from city_branch, MIG_CTE where MIG_CTE.id = parent_id) select * from MIG_CTE |
2 | select t.branch_level, t.id from city_branch c where (c.branch_level = ‘1’ or t.branch_level = ‘2’) and (t.sign = ‘1’ or t.sign = ‘4’ or t.sign = ‘8’) and t.status = ‘1’ start with c.id = i_branch_id connect by c.id = prior c.parent_id order by c.branch_level desc | with RECURSIVE MIG_CTE as (select t.branch_level, t.id, 1 as level from city_branch as cwhere ((((branch_level = ‘1’ or t.branch_level = ‘2’)and ((t.sign = ‘1’ or t.sign = ‘4’) or t.sign = ‘8’)) and t.status = ‘1’) and c.id = i_branch_id) union all select t.branch_level, t.id, level + 1 from city_branch as c, MIG_CTE where ((((branch_level = ‘1’ or t.branch_level = ‘2’) and ((t.sign = ‘1’ or t.sign = ‘4’) or t.sign = ‘8’)) and t.status = ‘1’) and c.id = MIG_CTE.parent_id)) select * from MIG_CTE order by MIG_CTE.branch_level desc |
Oracle中不同类型进行基于操作符的运算,会自动转化类型,譬如select 1 + '1' from dual
。Opengauss是强类型,不同类型的运算会提示类型不匹配,执行select 1 + '1'
会报错,需要进行显式的类型转换。
涉及的操作符类型包括:
操作符 | 操作符名称 |
---|---|
+ | 加法 |
- | 减法 |
/ | 除法 |
% | 取余 |
* | 乘法 |
|| | 字符串拼接 |
编号 | Oracle | Opengauss |
---|---|---|
1 | select 1 + ‘1’ | select 1 + 1 |
2 | select 1 + charCol from tbl | select 1 + cast(charCol as numeric) from tbl |
3 | select ‘1’ - 1 | select 1- 1 |
4 | select 1 * charCol from tbl | select 1 * **cast(charCol as numeric) ** from tbl |
5 | select 1 / charCol from tbl | select 1 /cast(charCol as numeric) from tbl |
6 | select charCol % 2 from tbl | select cast(charCol as numeric) % 2 from tbl |
编号 | Oracle | Opengauss |
---|---|---|
1 | select sysdate - 1 | select current_timestamp - interval ‘1’ DAY |
2 | select 1 + sysdate() | select interval ‘1’ DAY + now() |
3 | select systimestamp +1 | select current_timestamp + interval ‘1’ DAY |
4 | select systimestamp - 1 | select current_timestamp - interval ‘1’ DAY |
编号 | Oracle | Opengauss |
---|---|---|
1 | select 1||1 | select ‘1’||‘1’ |
2 | select 1 || c_custkey | select 1 || cast(c_custkey as text) |
Oracle中在函数调用时,参数类型进行会自动转化类型,譬如 select substr(123.12,0,2)
是合法的,且返回123
。Opengauss是强类型, 执行select substr(123.12,0,2)
会报错,需要进行显式的类型转换。
编号 | Oracle | Opengauss |
---|---|---|
1 | select substr(1234.1, 0, 4) | select substr(‘1234.1’, 1, 4+1) |
2 | select substr(‘1234.1’, 0, ‘2’) | select substr(‘1234.1’, 0, 2) |
编号 | Oracle | Opengauss |
---|---|---|
1 | select sum(‘2’) | select sum(2) |
编号 | Oracle | Opengauss |
---|---|---|
1 | select avg(‘2’) | select avg(2) |
编号 | Oracle | Opengauss |
---|---|---|
1 | select round(‘2’) | select round(2) |
Oracle中在进行条件判断时,左右表达式的类型进行会自动转化,譬如 where c_phone = 110
是合法的。Opengauss是强类型, 执行where c_phone = 110
会报错,需要进行显式的类型转换。
转换原则,优先转换常量类型;当两个都为数据列时,优先转换左边的。
编号 | Oracle | Opengauss |
---|---|---|
1 | select * from customer where c_phone = 110 | select * from customer where c_phone = ‘110’ |
2 | select * from customer where ‘1’ = c_custkey | select * from customer where 1 = c_custkey |
3 | select * from customer where c_phone = c_custkey | select * from customer where cast(c_phone as int) = c_custkey |
转换原则,转换 var0 between var1 and var2 中的 var1, var2。
编号 | Oracle | Opengauss |
---|---|---|
1 | select * from customer where c_custkey between ‘100’ and ‘200’; | select * from customer where c_custkey between 100 and 200 |
转换原则,转换List中的变量。
编号 | Oracle | Opengauss |
---|---|---|
1 | select * from customer where c_phone in (110,120); | select * from customer where c_phone in (‘110’, ‘120’) |
Oracle中有部分函数存在默认参数,而在Opengauss其参数是必填项。
编号 | Oracle | Opengauss |
---|---|---|
1 | select to_char(c_custkey) from customer | select cast(c_custkey as text) from customer |
编号 | Oracle | Opengauss |
---|---|---|
1 | select to_number(‘100’) | select 100 |
2 | select to_number(c_phone) from customer; | select cast(c_phone as numeric) from customer |
PawSQL专注数据库性能优化的自动化和智能化,支持MySQL,PostgreSQL,openGauss,Oracle等,提供的SQL优化产品包括
Twitter: https://twitter.com/pawsql