Open main menu

CDOT Wiki β

PostgreSQL Adapter-nexj/schema-test

Revision as of 11:51, 13 December 2010 by Minooz (talk | contribs) (2- Modifying Syntax)

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
Also the same line should be commented out in postgresql_create.sql script, which is being used to create tables in the database 'test':
/* 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
/* 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).
  • CREATE INDEX
  • CREATE Trigger // if it's needed for PostgreSQL as well as MySQL