PostgreSQL Adapter-nexj/schema-test

From CDOT Wiki
Revision as of 16:42, 13 December 2010 by Minooz (talk | contribs) (CREATE INDEX)
Jump to: navigation, search

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. So the line in the file is removed for now. The file itself may be deleted later, if there was no need for it. Name of the script file is being returned in 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
/* 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 </code>

if it's needed for PostgreSQL as well as MySQL