Business system compatible database oracle/postgresql (opengauss) /mysql Trivia

I remember before 《 The business system starts from Oracle Migrate to openGauss Simple records of the database 》：）
But it's not over yet ……

（ One ） Continue to move the database

It's endless ……
PostgreSQL -> Huawei openGauss
MySQL -> ZTE GoldenDB
……

（1.1） About GoldenDB

Official website

Mature and stable , Commercial leading financial level distributed database .
The product is as solid as a rock , Leading commercial market , Lead the formulation of industry standards , Build a new industrial ecology

According to the information given by ZTE , It is completely compatible in development MySQL Of , even JDBC Packages and drivers are MySQL.
And it seems not open source , Nor is it free software , So it can only be regarded as MySQL got .

（1.2） About openGauss Compatibility

Anyone who PostgreSQL Things that are openGauss All compatible .
and openGauss Extended some syntax similar Oracle Of , In turn, PostgreSQL No use .
So in order to unify and simplify , All use PostgreSQL Standards for .

（ Two ） Database differences

（2.1）Oracle/PostgreSQL/MySQL Basic differences

In order to unify the code, whoever is rigorous will follow whoever .
There are many differences , I don't know how to list them more appropriately .

1. Default object name Oracle Capitalize and PostgreSQL/MySQL A lowercase letter . You can use quotation marks （ Even the quotation marks are different ） Forced small / Capitalization . Under normal circumstances, do not add quotation marks .
2. Force case changing fields Oracle stay SQL The statement must also be quoted to indicate ,MySQL It seems that you can ignore case .
3. For fields with different types Oracle You can directly compare CharField1 = NumField2（ Seems to be openGauss It's also ）, and PostgreSQL/MySQL The data type should be consistent .
4. The binding value in the program is also the same as SQL In the same ,Oracle Can mix String and Int use , and PostgreSQL/MySQL The data type should be consistent .
5. Many sentences are slightly different , Don't list them one by one , such as Oracle Of CONSTRAINT The word , There are other databases everywhere ……

（2.2） Create table statement, etc （DDL）

The main difference is the data type , such as varchar Is there any 2,number,numberic,decimal……
This part is relatively simple , Just modify the sentence , Or use Navicat direct Data Transfer.
PS： Find out Navicat It's convenient to synchronize table structure and data across databases （ But I don't want to change the original article ）, You can first use the mutual import table structure , Then export from the corresponding library DDL sentence , Avoid changing yourself one by one ,yeah！

（2.2.1） Common data types

• Oracle :VARCHAR2,NUMBER,DATE,……
• PostgreSQL :varchar,numeric,timestamp,……
• MySQL :varchar,numeric,datetime,……

（2.2.2） Notes to table

The writing is different , But the difference is small ：

• Oracle : First Create Table TName (Field1,Field2,...), Again COMMENT ON Table TName IS ' notes '
• PostgreSQL : First Create Table TName (Field1,Field2,...), Again COMMENT ON Table TName IS ' notes ', ditto
• MySQL : First Create Table TName (Field1,Field2,...), Again ALTER Table TName COMMENT = ' notes '

（2.2.3） Field comments

Field comments are different , It involves some code logic ：

• Oracle : First Create Table TName (Field1,Field2,...)
  Recycle comments for each field ：COMMENT ON COLUMN TName.Field1 IS ' notes '
• PostgreSQL : First Create Table TName (Field1,Field2,...)
  Recycle comments for each field ：COMMENT ON COLUMN TName.Field1 IS ' notes ', ditto
• MySQL : direct Create Table TName (Field1 COMMENT ' notes 1‘,Field2 COMMENT ' notes 2‘,...)
  In fact, you can also create tables first and then annotate fields , But with Alter Table sentence , And a little complicated （ Write everything down ）, Therefore, it is relatively simple to annotate the code at one time .

（2.3） Delete table statement

Under normal circumstances , need CASCADE Delete other objects associated with the table （ Indexes , constraint ）.
about Oracle Our data sheet is too big for us to flash back , So add PURGE.
and MySQL Of drop … if exists Interesting as it is , But try to unify or not .

• Oracle : DROP TABLE TName CASCADE CONSTRAINTS PURGE'
• PostgreSQL : DROP TABLE TName CASCADE
• MySQL : DROP TABLE TName CASCADE

（2.4） Sequence

It is not recommended to make MySQL Realize similar sequence functions , Because the target database may have permission problems .
Because the sequence takes out the value first and then uses , The autoincrement field is Insert Then get , So the logic of some codes involved is different .

（2.4.1） establish

• Oracle : Use sequence objects CREATE SEQUENCE SEQXXX START WITH 1000 MAXVALUE 99999999 MINVALUE 1000
• PostgreSQL : Use sequence objects CREATE SEQUENCE SEQXXX START WITH 1000 MAXVALUE 99999999 MINVALUE 1000, Basically the same as above
• MySQL : Set table field self increment Create Table TName (Field1 int NOT NULL AUTO_INCREMENT ,Field2,...),

（2.4.2） Use

• Oracle : Select SEQXXX.nextval from dual
• PostgreSQL : select nextval('SEQXXX')
• MySQL : First Insert into TName Record , Auto increment field value assignment 0 or Null, Again select last_insert_id() Select the value after self increment .
  We need to pay attention to ： Insert statement and select from value-added statement need to be carried out in the same connection .
  Even in the same connection , Due to the development language / The drive is different , occasionally select last_insert_id() Will also choose 0 It's worth it （ Well, I met ）.

（2.5） Check the user object

（2.5.1） surface

Due to database differences , The statement here is not very accurate , Change it according to the actual situation .
Is it a choice (User/Owner) User's table , still (Schema) Table of schemes ？

• Oracle : select TABLE_NAME from user_tables where TABLE_NAME like ?
  —— Users are solutions
• PostgreSQL : select tablename from pg_tables where schemaname = ? and upper(tablename) like ?
  —— User owned solutions , Can also be combined with MySQL equally , as follows ：
• MySQL : select TABLE_NAME from information_schema.tables where TABLE_SCHEMA=? and upper(TABLE_NAME) like ?
  —— Users and Libraries （ programme ）

（2.5.2） Field

• Oracle : select * from USER_TAB_COLUMNS where TABLE_NAME = ? and COLUMN_NAME = ?
• PostgreSQL : select * from information_schema.columns where TABLE_SCHEMA= ? and upper(TABLE_NAME) = ? and upper(COLUMN_NAME) = ? , Uh, no pg_columns.
• MySQL : select * from information_schema.columns where TABLE_SCHEMA= ? and upper(TABLE_NAME) = ? and upper(COLUMN_NAME) = ? , ditto .

（2.6） Table data import

except insert outside , A large number of data import databases have their own methods .
This place here openGauss Acting strange , It's too slow .

• Oracle : Use SQL Loader, That is to say sqlldr... command , This command file and related information files are required under the client , Skip here .
• PostgreSQL : Java Use org.postgresql.copy.CopyManager,COPY ... FROM STDIN..., There is no time to study other languages .
• MySQL : perform SQL sentence ：LOAD DATA LOCAL INFILE ... INTO TABLE ... , Both client and server are required to allow this Load How to file （ Understand specific safety risks by yourself ）.
  Otherwise, the report will be wrong ：Loading local data is disabled; this must be enabled on both the client and server sides
  Need server settings （Windows Next is ini）：my.ini file [mysqld] Lower join local-infile=1

// Client side Settings ：
Property.setProperty("allowLoadLocalInfile","true");
Connection conn = DriverManager.getConnection(jdbcUrl, Property);
......

Compared with ：

（2.7） Common functions

（2.7.1） Date to string

• Oracle : To_Char(DATE,'YYYY-MM-DD')
• PostgreSQL : To_Char(DATE,'YYYY-MM-DD'), ditto
• MySQL : date_format(DATE,'%Y-%m-%d')

It is not recommended to make MySQL The implementation is similar to To_Char function , Because the target database may have permission problems .
attach ： Not recommended MySQL Custom function ：

CREATE FUNCTION to_char(d datetime, format varchar(40)) RETURNS varchar(40) 
    DETERMINISTIC
begin
	declare str varchar(40) DEFAULT '';
	set str = replace(format, 'YYYY', '%Y');	
	set str = replace(str, 'yyyy', '%Y');	
	set str = replace(str, 'MM', '%m');	
	set str = replace(str, 'mm', '%m');	
	set str = replace(str, 'DD', '%d');	
	set str = replace(str, 'dd', '%d');	
	set str = replace(str, 'HH24', '%H');	
	set str = replace(str, 'hh24', '%H');	
	set str = replace(str, 'HH', '%h');	
	set str = replace(str, 'hh', '%h');	
	set str = replace(str, 'MI', '%i');	
	set str = replace(str, 'mi', '%i');	
	set str = replace(str, 'SS', '%s');	
	set str = replace(str, 'ss', '%s');	
	return date_format(d, str);
end

（2.7.2） Date calculation

Add or subtract months as an example ：

• Oracle : select add_months(sysdate,xxx) from dual
• PostgreSQL : select now()::timestamp + 'yyy month'
• MySQL : select date_add(now(),interval zzz month)

（2.7.3） String to date

Be careful date and timestamp stay Oracle The difference between .

• Oracle : select to_date('2022-06-01 23:45:59','yyyy-mm-dd hh24:mi:ss') from dual
• PostgreSQL : select to_timestamp('2022-06-01 23:45:59','yyyy-mm-dd hh24:mi:ss')
• MySQL : select str_to_date('2022-06-01 23:45:59','%Y-%m-%d %T')

（2.7.4） String splicing

• Oracle : Field1 || '_' || Field2 || '_' ||..., It can also be used. CONCAT()
• PostgreSQL : Field1 || '_' || Field2 || '_' ||..., It can also be used. CONCAT(), ditto
• MySQL : CONCAT(Field1,'_',Field2,'_',...)

（2.7.5） Substring function

️️️ ️ hole ️️️
Everyone supports substr, The problem is substr It's from 1 Start counting Index, but Oracle from 0 from 1 Are all the same .
So if it was written like substr(field1,0,x) This kind of sentence ……
that Oracle It's okay , Other databases cannot get the corresponding results （ The mistakes are different ）.

• Oracle : substr(name,0,2) = substr(name,1,2) ： obtain 2 Characters .
• PostgreSQL : substr(name,0,2) = substr(name,1,1) ： obtain 1 Characters .
• MySQL : substr(name,0,2) ： obtain 0 Characters （ Can't get ）.

（2.7.6） Return non null function

It is also not recommended to implement custom functions nvl name （ Prefer in the program SQL Add variables ）.

• Oracle : nvl(field,-1)
• PostgreSQL : coalesce(field,-1)
• MySQL : ifnull(field,-1)

（2.7.7） length （Blob type ）

• Oracle : Length(Blobfield)
• PostgreSQL : pg_column_size(Blobfield)
• MySQL : Length(Blobfield)

（2.8） Dataset operations （DML）

Fortunately 【 Additions and deletions 】 It's the same , After all, are SQL, Set operation ：
First union（ Filter repeat ）, union all（ Do not filter duplicates ） It also means the same in different databases , however ……

• Oracle : select * from T1 MINUS select * from T2
• PostgreSQL : select * from T1 EXCEPT select * from T2
• MySQL : No, MINUS or EXCEPT, Use other SQL Instead of （ What is other ？ Uh ：join, not in）.
  If you use not in Be careful about the efficiency problems caused by the amount of data （ Probably originally 1 Seconds of operation becomes dozens of minutes ）.

（2.9） Join syntax of table （ Left connection ）

It was used Oracle Of (+) All methods need to be changed into left join, Ah ……
If you don't use it at the beginning (+) Of Oracle Grammar is easy .

as for inner join and where The difference between , Find out for yourself . The result is the same , The size of the intermediate temporary set is different .

• Oracle : select x from T1,T2 where T1.a=T2.a(+) Equivalent select X from T1 left join T2 on T1.a=T2.a
• PostgreSQL : select X from T1 left join T2 on T1.a=T2.a
• MySQL : select X from T1 left join T2 on T1.a=T2.a

（2.10） UPDATE statement （ special ）

Something like this SQL（ Simplified , So it seems completely meaningless ）：

update tablename
set name = 'XXXXXX'
where id=0
	and id IN
	(
		select id from tablename
	);

• Oracle : Successful implementation ️
• PostgreSQL : Successful implementation ️
• MySQL : Execution failure ️ You can't specify target table 'XXXXX' for update in FROM clause

Need to nest another layer , You can perform ：

update tablename
set name = 'XXXXXX'
where id=0
	and id IN
	(
		select id from
			(select id from tablename) tn
	);

（2.11） Get database information

（2.11.1） essential information （ edition ）

• Oracle : select * from v$version
• PostgreSQL : select version()
• MySQL : select version() + show variables like '%storage_engine%'

（2.11.1） Database character set （ code ）

• Oracle : select value from nls_database_parameters where Parameter='NLS_CHARACTERSET'
• PostgreSQL : show server_encoding
• MySQL : show variables like '%character_set%'

---

…… At the end of the temporary article ……️