PostgreSQL Adapter-nexj/schema-test
PostgreSQL Adapter for NexJ - Working Inside Schema 'test'
1- Initializing the SQL environment
- PostgreSQLAdapter.java
- When a connection is first established, this initial SQL statement should execute in MySQL, which is not the case for PostgreSQL, so the implementation was removed from PostgreSQLAdapter.java:
/* MySQLAdapter.java */
public String getInitialSQL()
{
StringBuffer buf = new StringBuffer();
buf.append("set sql_mode = concat(@@sql_mode, ',ANSI_QUOTES')"); // allow using doublequote when quoting column names in "CREATE TABLE" statements
buf.append(";set optimizer_search_depth = 0"); // let DB automatically decide on how long it takes to examine plans, improves long planning sessions
buf.append(";set max_sort_length = ").append
(Math.max(MAX_VARCHAR_PRECISION, MAX_VARBINARY_PRECISION)); // set TEXT/BLOB minimum sorting length to be same as cutoff between varchar/text
return buf.toString();
}
- postgresql_create.sql
- This file was just created as a copy of mysql_create.sql, which was not the case for PostgreSQL again and was removed from getCreateEtcScriptName()' in PostgreSQLSchemaManager.java
/* nexj/core/persistence/sql/etc/postgresql_create.sql */
set sql_mode = concat(@@sql_mode, ',ANSI_QUOTES');
- After activating the connection, the database is locked.
2- Modifying Syntax
-
CREATE TABLE
- File to configure: PostgreSQLSchemaManager.java
- 1- DATA TYPE: data types are changed in
appendColumnType()
based on each jdbc type's equivalent in PostgreSQL. - 2- STORAGE ENGINE: In MySQL, the storage engine is set to a transactional safe engine such as InnoDB, whereas PostgreSQL has a single built in engine. So, implementation of
appendTableSuffix()
in PostgreSQLManager.java was removed - 3- CHARACTER SET: In MySQL, the character set should be defined when creating a table. Whereas in PostgreSQL when a database is created, the character set is set to 'UTF8' by default.
- 4- AUTO INCREMENT: PostgreSQL doesn't support 'auto_increment' as some other databases for a unique identifier column. There are two ways around this, 1- To create a 'sequence', 2- Use SERIAL, which is a macro around 'sequence'. SERIAL is an 'integer' and a 'sequence, with the column default to the sequences next value. As of postgreSQL 7.3, to make a serial column unique, it should be specified as a unique constraint or a primary key. More info @ PostgreSQL Resources
- 1- DATA TYPE: data types are changed in
/* MySQL version of creating table script */
create table test.RangeTest(
id int auto_increment not null, s varchar(1) character set utf8 null, bin varbinary(1) null,
n int null, n1 tinyint unsigned null, l bigint null, "DEC" decimal(10,5) null,
f float null, d double null, tm datetime null, b boolean null,
constraint RangeTest_PK primary key(id)
)engine=InnoDB character set = utf8
/* PostgreSQL version of creating table script */
CREATE TABLE test.RangeTest(
id serial not null, s text null, bin bytea null,
n integer null, n1 smallint null, l bigint null, "DEC" decimal(10,5) null,
f double precision null, d double precision null, tm timestamp null, b boolean null,
constraint RangeTest_PK primary key(id)
);
// Using SERIAL is another way to work around 'sequence'
//CREATE SEQUENCE test.RangeTest_id_seq;
//ALTER TABLE test.RangeTest ALTER COLUMN id SET DEFAULT NEXTVAL('test.RangeTest_id_seq');
CREATE TEXT TABLE
- The implementation in
crateTextTable
is removed for now (base class's version is being called).
- The implementation in
-
CREATE INDEX
-
CREATE Trigger // if it's needed for PostgreSQL as well as MySQL
Files to Configure |
---|
|