[笔记]将系统的数据库从MySQL 5.5迁移到PostgreSQL 9.1

环境
Windows Server 2003 x64 简体中文, MySQL 5.5 (UTF8编码), PostgreSQL 9.1.4-1 (UTF8编码)
Spring 3.0.7, Struts 2.3.4, Hibernate 3.5.5

从MySQL迁移到PostgreSQL

-----------------------------分隔线-----------------------------
* 分页写法的区别

	PostgreSQL	MySQL
LIMIT 数量	支持	支持
LIMIT 下标, 数量	不支持	支持
LIMIT 数量 OFFSET 下标	支持	支持

e.g.

SELECT * FROM user LIMIT 10; -- PostgreSQL与MySQL均支持

SELECT * FROM user LIMIT 10, 10; -- PostgreSQL不支持，MySQL支持

SELECT * FROM user LIMIT 10 OFFSET 10; -- PostgreSQL与MySQL均支持

-----------------------------分隔线-----------------------------

* DDL定义中的各种写法区别

** 前者可用符号"`"(不含引号，在键盘上是Esc键下面的那个键对应的字符)来包裹住表名、列名等，后者不可以。

* 转义字符
MySQL \

PostgreSQL 默认不支持惯用的\，而是用的英文单引号'

* 使用DB

MySQL中可以 use DB名;

PostgreSQL中不支持 use 关键字!

* 主键约束

二者没什么区别

* 自增长主键
MySQL中写成这样

CREATE TABLE users (     id INT(11) NOT NULL AUTO_INCREMENT,     name VARCHAR(50) NOT NULL,     PRIMARY KEY (id) );

PostgreSQL中写成这样

CREATE TABLE users(     id serial NOT NULL,     name VARCHAR(50) NOT NULL,     PRIMARY KEY (id ) );

注: PostgreSQL会对 serial数据类型的列隐式生成名为 表名_PK名_seq 的SEQUENCE，此例的SEQUENCE名为 users_id_seq 。

* 改变已有表的自增长序列

-- 更安全有效的解决方案请参考http://stackoverflow.com/questions/244243/how-to-reset-postgres-primary-key-sequence-when-it-falls-out-of-sync/* 假如有表a从MySQL迁移过来，其中有数据100条，若在建立此表时没有指定序列，PG会默认给此表的主键列一个
sequence——它是增长的，幅度为1。此时若用Hibernate来给表a添加数据会报错，说主键1已经存在！
所以可以在迁移表之后，将表的sequence作少许修改，让其从当前表的主键的最大值再加1来开始！即可解决
Hibernate添加数据时报错的问题

ALTER SEQUENCE "public"."表名_主键名_seq" RESTART WITH (PK的最大值 + 1);
或
ALTER SEQUENCE 表名_主键名_seq RESTART WITH (PK的最大值 + 1);

e.g.
ALTER SEQUENCE file_types_id_seq" RESTART WITH 10;

从上面stackoverflow.com网站上得到的更加简单有效的一句SQL语句如下：
SELECT pg_catalog.setval(pg_get_serial_sequence('table_name', 'id'), (SELECT MAX(id) FROM table_name)+1);

*/

* 唯一键约束

MySQL UNIQUE KEY name (name)

PostgreSQL UNIQUE (name)

* 内置SQL函数区别

MySQL格式化日期 DATE_FORMAT(CURRENT_TIMESTAMP,'%Y-%m-%d %H:%i:%s')

PostgreSQL格式化日期 to_char(CURRENT_TIMESTAMP,'yyyy-mm-dd hh24:mi:ss'), to_date(text, text), to_timestamp(text, text)

* 日期类型

MySQL date time datetime

PostgreSQL timestamp

* 布尔类型
MySQL 木有，可用int(1)、枚举或者字符串的方式来模拟。

PostgreSQL boolean

* 外键约束

在MySQL中创建带有外键的表写法如下:

DROP TABLE IF EXISTS recipient_recipientgroup; 

CREATE TABLE IF NOT EXISTS recipient_recipientgroup ( 
  id serial NOT NULL, 
  recipient_id INTEGER DEFAULT NULL, 
  recipient_group_id INTEGER DEFAULT NULL, 
  PRIMARY KEY (id), 
  KEY FK_recipient_recipientgroup_recipient (recipient_id), 
  KEY FK_recipient_recipientgroup_recipient_group (recipient_group_id), 
  CONSTRAINT FK_recipient_recipientgroup_recipient FOREIGN KEY (recipient_id) REFERENCES recipient (id), 
  CONSTRAINT FK_recipient_recipientgroup_recipient_group FOREIGN KEY (recipient_group_id) REFERENCES recipient_group (id) 
);

在 PostgreSQL 中创建带有外键的表写法如下: (即在MySQL的写法上去掉 KEY ... 这句!)

DROP TABLE IF EXISTS recipient_recipientgroup; 

CREATE TABLE IF NOT EXISTS recipient_recipientgroup ( 
  id serial NOT NULL, 
  recipient_id INTEGER DEFAULT NULL, 
  recipient_group_id INTEGER DEFAULT NULL, 
  PRIMARY KEY (id), 
--  KEY FK_recipient_recipientgroup_recipient (recipient_id), 
--  KEY FK_recipient_recipientgroup_recipient_group (recipient_group_id), 
  CONSTRAINT FK_recipient_recipientgroup_recipient FOREIGN KEY (recipient_id) REFERENCES recipient (id), 
  CONSTRAINT FK_recipient_recipientgroup_recipient_group FOREIGN KEY (recipient_group_id) REFERENCES recipient_group (id) 
);

代码修改
----------
* 将SpringSide 3.3.4中提供的IdEntity类修改如下

/**
 * 参考SpringSide3，统一定义id的entity基类.
 * 
 * 基类统一定义id的属性名称、数据类型、列名映射及生成策略.
 * 子类可重载getId()函数重定义id的列名映射和生成策略.
 */
//JPA 基类的标识
@MappedSuperclass
public abstract class IdEntity {

    protected Long id;

    @Id
//    @GeneratedValue(strategy = GenerationType.AUTO)
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    @Column(unique = true, nullable = false)
    public Long getId() {
        return this.id;
    }

    public void setId(Long id) {
        this.id = id;
    }

}

* 代码中的SQL/HQL 更改

public List<LeakageDetail> findExceptLeakageDetailList(String ids) {
        String queryString = "SELECT * FROM leakage_detail "
                + "WHERE " // -- DATE_FORMAT(find_date, '%Y%m')<(DATE_FORMAT(NOW(), '%Y%m')-1) AND 
                + "CONCAT(find_date, find_process) IN ( "
                + "SELECT CONCAT(find_date, find_process) AS xx "
                + "FROM leakage_detail WHERE id IN(" + ids + ")"
                + "GROUP BY find_date, find_process "
//                + "HAVING COUNT(xx)>5)"; // 这种写法MySQL支持，PostgreSQL不支持！
                + "HAVING COUNT(CONCAT(find_date, find_process))>5) AND id IN(" + ids + ") ORDER BY find_date, find_process";
        logger.info("Leakage模块数据导入时发送漏液异常邮件的查询sql->"+queryString);
        Query query = getSession().createSQLQuery(queryString).addEntity(LeakageDetail.class);
        return query.list();
    }

public List<StatisticalAnalysisVo> getStatisticalAnalysisList() {
//        String hql = "select workshop as name, count(id) as num from DataModel where date_format(create_at, '%Y-%m')=date_format(now(), '%Y-%m') group by workshop";
        String sql = "select workshop as name, count(id) as num "
                + "from data_models "
//                + "where date_format(create_at, '%Y-%m')=date_format(now(), '%Y-%m') " // date_format函数是MySQL专用的
                + "where to_char(create_at, 'yyyy-MM')=to_char(now(), 'yyyy-MM') " // PostgreSQL中的日期格式化函数是to_char
                + "group by workshop";
//        Query query = getSession().createSQLQuery(hql);
        Query query = getSession().createSQLQuery(sql).addScalar("name", Hibernate.STRING).addScalar("num", Hibernate.LONG);
        query.setResultTransformer(Transformers.ALIAS_TO_ENTITY_MAP)
            .setResultTransformer(Transformers.aliasToBean(StatisticalAnalysisVo.class));
        return query.list();
    }

* 存储过程更改
MySQL

DROP PROCEDURE IF EXISTS `calcUlclp`;
DELIMITER //
CREATE DEFINER=`root`@`localhost` PROCEDURE `calcUlclp`()
    COMMENT '计算Hipot不良率的上下限的存储过程'
BEGIN

SELECT (@rownum := @rownum + 1) AS `id`, DATE_FORMAT(lot_no_to_date, '%Y%m') AS year_and_month,
 `model_no`, group_no,
 SUM(liquid_injected_input_num) AS total_input,
 SUM(short_circuit_num) AS total_short,
 COUNT(DISTINCT(lot_no)) AS month_num_of_product_days
 ,ROUND(SUM(liquid_injected_input_num) / COUNT(DISTINCT(lot_no))) AS sample_size_n
 ,ROUND(SUM(short_circuit_num) / SUM(liquid_injected_input_num), 4) AS nonconforming_rate_mean_p
    FROM hipot, (SELECT @rownum := 0) AS r WHERE  liquid_injected_input_num!=0
	 GROUP BY `model_no`, group_no, year_and_month
;

END//
DELIMITER ;

PostgreSQL

DROP FUNCTION IF EXISTS calcUlclp();

CREATE OR REPLACE FUNCTION calcUlclp()
    RETURNS SETOF record AS
$BODY$
 declare
--   sql varchar;
   rownum int;
   v_rc record;
BEGIN
  for v_rc in
SELECT (rownum = rownum + 1) AS id, to_char(lot_no_to_date, 'yyyyMM') AS year_and_month,
 model_no, group_no,
 SUM(liquid_injected_input_num) AS total_input,
 SUM(short_circuit_num) AS total_short,
 COUNT(DISTINCT(lot_no)) AS month_num_of_product_days
 ,ROUND(SUM(liquid_injected_input_num) / COUNT(DISTINCT(lot_no))) AS sample_size_n
 ,ROUND(SUM(short_circuit_num) / SUM(liquid_injected_input_num), 4) AS nonconforming_rate_mean_p
    FROM hipot, (SELECT rownum = 0) AS r WHERE  liquid_injected_input_num!=0
	 GROUP BY model_no, group_no, year_and_month
  loop
  return next v_rc;
  end loop;

END;
$BODY$
  LANGUAGE 'plpgsql' VOLATILE;

-- 调用存储过程
/*
 SELECT * from calcUlclp() as
  t(id_ boolean, year_and_month text, model_no varchar, group_no varchar,
   total_input bigint, total_short numeric, month_num_of_product_days bigint,
    sample_size_n double precision, nonconforming_rate_mean_p numeric);
*/

来源：oschina

链接：https://my.oschina.net/u/115036/blog/68427

标签

postgresql

mysql

数据库迁移