Microsoft Word - Porting from Oracle to Postgres V2.doc

Porting from Oracle to PostgreSQL

by Paulo Merson February/2002

If you are starting to use PostgreSQL or you will migrate from Oracle database server, I hope this document helps. If you have Java

applications and use JDBC, the “Data types and JDBC” section will be particularly useful.

Oracle and PostgreSQL both conform to standard SQL. However, they contain several extensions and implementation details that

differentiate one from the other. The most important differences are listed in this document.

If you have comments about this document, please email me: [email protected].

1. SQL Syntax, Functions, Sequences, Etc.

Oracle PostgreSQL

select sysdate from dual select ‘now’::datetime

There is no “dual” table

Unlike other RDBMS, PostgreSQL allows a “select” without the

”from” clause. This use does not affect portability because

the syntax to get current time is already DBMS specific.

CREATE SEQUENCE seqname [ INCREMENT BY integer ]

[ MINVALUE integer ] [ MAXVALUE integer ]

[ START WITH integer ] [ CACHE integer ] [ CYCLE | NOCYCLE ]

Oracle’s “create sequence” has other arguments not listed here and not

supported by PostgreSQL, but the main difference is the need of ‘by’ and

“with” after “increment” and “start”.

If you don’t specify ”MAXVALUE” or if you use the parameter

“NOMAXVALUE”, then the actual limit is 10

CREATE SEQUENCE seqname [ INCREMENT increment ]

[ MINVALUE minvalue ] [ MAXVALUE maxvalue ]

[ START start ] [ CACHE cache ] [ CYCLE ]

If you don’t specify MAXVALUE, then the maximum value is

2147483647 for ascending sequences.

Oracle PostgreSQL

To return the current value and increment the counter:

sequence_name.nextval;

Possible usage in a select statement:

select sequence_name.nextval from dual;

To return the current value and increment the counter:

nextval(‘sequence_name’);

Possible usage in a select statement

select nextval(‘sequence_name’);

Note that unlike other RDBMS, PostgreSQL allows a select without the

‘from’ clause. This use does not affect portability because the sequence

syntax is already DBMS specific.

SELECT product_id,

DECODE (warehouse_id, 1, ’Southlake’,

2, ’San Francisco’,

3, ’New Jersey’,

4, ’Seattle’,

’Non-domestic’)

quantity_on_hand

FROM inventories

SELECT a,

CASE WHEN a=1 THEN 'one'

WHEN a=2 THEN 'two'

ELSE 'other'

END

FROM test

select employeeid,

NVL(hire_date, sysdate)

from employee

where employeeid = 10;

Oracle also has a “coalesce” function that is a generalization of the

commonly used NVL function.

select employeeid,

coalesce(hire_date, 'now'::datetime)

from employee

where employeeid = 10;

Outer join – “(+)” Doesn’t support outer join. The workaround is to use a union.

Hierarchical queries – “CONNECT BY” Nothing similar

SELECT product_id FROM inventories

MINUS

SELECT product_id FROM order_items;

SELECT product_id FROM inventories

EXCEPT

SELECT product_id FROM order_items;

select unique col1, col2 from table1

In Oracle “distinct” and “unique” keywords are synonymous in the select

statement.

select distinct col1, col2 from table1

PostgreSQL doesn’t allow “select unique”.

Oracle PostgreSQL

Oracle relational operators may have a space between the characters. For

example, the following select will work:

select id,name from employee where id > = 10;

// There are spaces and tabs between “>” and “=”

PostgreSQL relational operators doesn’t allow spaces, the characters that

compound an operator must be consecutive when the command is

parsed:

select id,name from employee where id >= 10;

To get the remainder of the division of 10 by 4 (modulo) use the “mod”

function:

select mod(10,4) from dual;

To get the remainder of the division of 10 by 4 (modulo)

use the “%” operator. (And PostgreSQL has many other arithmetic

operators.)

select 10 % 4;

The “ROWNUM” pseudo-column returns a number indicating the order

in which Oracle selects the row.

ROWNUM can be used to limit the number of rows returned by a query,

for example:

select * from employees where rownum < 10 order by name;

ROWNUM can be used in the projection as one of the values returned by

the query (first line has value 1, second line value 2, and so on):

select rownum, name from employees order by name;

There isn’t anything equivalent to Oracle ROWNUM.

However, you can limit the number of rows returned by a query using the

“LIMIT” clause:

select * from employees order by name limit 10;

In some cases, it’s possible that the pseudo-column OID may substitute

ROWNUM, although they have different behavior. OID is a unique

identifier of each line per table, while ROWNUM is always 1, 2, …, N

for each different query.

select oid, name from employees order by name;

And the query that uses ROWNUM can have join tables. If your select is

a join you’ll have a different OID for each table, because each one has an

OID column.

2. Database Server General Characteristics

Oracle PostgreSQL

A view can be “updatable” if some conditions are satisfied. Views are read only.

Oracle PostgreSQL

Transactions are initiated implicitly.

By default, the “auto-commit” behavior is disabled.

BEGIN – initiates a transaction disabling the default behavior,

which is “auto-commit” enabled, i.e., a commit is performed after

each user statement.

In Java, we need to write: con.setAutoCommit(false)

Tables, views and other objects can be logically grouped in

“schemas”. A schema usually maps to the user name of the user that

created the objects (owner). Thus, a table can be referenced in a

statement as “schemaName.tableName”. For example:

select * from mySchema.myTable;

There is no schema support, but it’s planned for a future version.

The alternative is to use separate databases: You have to connect to

the specific database and use that connection to execute your SQL

command.

However, if you have a SQL statement that uses tables in different

(Oracle) schemas, you cannot use separate (PostgreSQL) databases;

there is no direct workaround, you’ll need to rewrite the code.

Interactive command prompt tool:

SQL*Plus

Interactive command prompt tool:

psql

Oracle permissions are granted/revoked to/from users or roles. You

can create roles and grant/revoke roles to/from users. But you can

“create” and “alter” groups to insert and remove users.

PostgreSQL permissions are granted/revoked to/from users or

groups. You can create groups and then alter the groups

inserting/removing users.

By default, a password is always required to connect to the database. By default, you can connect to the database simply by specifying the

database name, no user ID and password are required.

You should follow the instructions in the Administrator’s Guide to

configure the pg_hba.conf file in order to use password

authentication.

3. Data Types and JDBC

Oracle

JDBC

PostgreSQL JDBC

SMALLINT - 2 bytes getShort

INTEGER - 4 bytes getInt

NUMBER(p) where p is the precision, i.e., the

number of digits

getByte

getShort

getInt

getLong

BIGINT - 8 bytes getLong

NUMERIC(p,s) getBigDecimal

REAL – 4 bytes, 6 decimal places getDouble

NUMBER(p,s) where p is the total number of

digits and s is the number of digits to the

right of the decimal point

getDouble

getBigDecimal

DOUBLE PRECISION – 8 bytes, 15 decimal places getDouble

Nothing similar SERIAL - 0 to +2147483647, typically used to

create unique identifiers. Generates an

implicit sequence that is incremented when a

line is inserted in the table.

getInt

VARCHAR2(size) where maximum size is 4000 getString CHARACTER VARYING(n) where maximum n is 1 GB

VARCHAR(n) is an alias

getString

CHAR(size) where maximum size is 2000 getString CHARACTER(n) where maximum n is 1 GB

CHAR(n) is an alias

It’s suggested that you use TEXT if n > 10 MB

getString

LONG - Character data of variable length up to

2 GB

getString TEXT – variable length up to 1 GB

It’s suggested that you use TEXT if n > 10 MB

getString

DATE – holds date and time

getDate

getTime

getTimestamp

TIMESTAMP oracle.sql.

getTIMESTAMP

TIMESTAMP

You still can use getDate to read a TIMESTAMP

column, but you will loose the time portion of

the data.

getTimestamp

Nothing similar DATE – holds only the date (resolution is one

day)

getDate

Nothing similar TIME – holds only the time (00:00:00.00 –

23:59:59.99)

getTime

Oracle

JDBC

PostgreSQL JDBC

RAW(size) – binary data of length size bytes

(max 2000)

LONG RAW – binary data or variable length up

to 2GB

getBytes BYTEA getBytes

Nothing similar BIT(n) – fixed length string of 1’s and 0’s

BIT VARYING(n) – variable length string of 1’s

and 0’s

(?)

CLOB – character large object (max 4GB) getClob TEXT (max 1GB) getString

BLOB – binary large object (max 4GB) getBlob BYTEA (max 1GB)

BYTEA is not documented in PostgreSQL 7.1 but

it’s fully implemented; Jdbc 7.2-1.2 is

required though in order to use getBytes and

setBytes.

Besides TEXT and BYTEA, PostgreSQL supports

large objects as separate files. They are

stored in a separate table in a special

format, and are referred to from regular

tables by an OID value. More information:

http://www.postgresql.org/idocs/index.php?larg

eobjects.html

http://www.postgresql.org/idocs/index.php?jdbc

-lo.html

getBytes

ROWID oracle.sql.

getROWID

Nothing similar

Typically, char(1) is used to store a value

that is translated to Boolean in the

application logic.

If you store '0' and '1' in a varchar2(1) or

char column, then the jdbc driver can

correctly interpret these values as boolean

false an true respectively using

ResultSet.getBoolean and

PreparedStatement.setBoolean

BOOLEAN – can have the value TRUE, FALSE or

NULL

If you store '0' and '1' in a varchar(1) or

char(1) column, then the jdbc driver can

correctly interpret these values as boolean

false an true respectively using

ResultSet.getBoolean. However,

PreparedStatement.setBoolean simply does not

work.

If you use PostgreSQL BOOLEAN, then your Java

code can use getBoolean and setBoolean.

getBoolean

Oracle Spatial features ? Geometric data types: POINT, LINE, CIRCLE,

etc.

org.

postgresql.

geometric.*

Oracle

JDBC

PostgreSQL JDBC

Nothing similar Network address data types: INET, MACADDR,

CIDR.

(?)

JDBC note: Typically PreparedStatement.setXxx is used to set the value of arguments (or “host” variables) inside SQL statements.

And there is a correspondent ResultSet.getXxx method to read the value returned by a query into java variables. Each get/setXXX

method has a specific Java data type or class associated to it (e.g. set/getInt deals with int; set/getDate deals with

java.sql.Date, etc.). Further, you can use different methods to read the same database data type, but usually there is a recommended

method. For example, a BIGINT column can be read with getShort, getInt, getLong, getDouble, etc., but the recommended method

is getLong.

So, to indicate the proper way to use each data type in Java I simply listed the recommended getXxx JDBC method.

4. Other Considerations:

• The set of operators and SQL functions is very similar though Oracle has a richer set. For example, both DBMS have the concatenation

operator “||”, as well as substr, upper, to_char and other functions with the same syntax. However, any Oracle function that is being

used must have its syntax compared to the equivalent function in PostgreSQL, if such exists.

• PostgreSQL lacks the ability to query across multiple databases.

• PostgreSQL’s PL/pgSQL is similar to Oracle PL/SQL and can be used to write stored functions. PostgreSQL doesn’t have packages or

procedures (only functions). More about this: http://www.postgresql.org/idocs/index.php?plpgsql-porting.html

• Both DBMS have triggers and the create trigger statement is similar, but the code executed by the trigger for PostgreSQL must be

in a stored function written by the user, while in Oracle you have the option of writing the code in a PL/SQL block in the create

trigger

statement. PostgreSQL has yet an additional resource called the “rule system” that allows the definition of business logic that

is executed upon an event.

• The create table statement is similar in both DBMS. One noticeable difference is that PostgreSQL doesn’t have pctfree,

pctused, inittrans, and maxtrans clauses. They also differ in the create database statement, mainly in the arguments and

clauses that specify storage details.

5. References:

• Oracle 9i documentation - http://download-east.oracle.com/otndoc/oracle9i/901_doc/nav/docindex.htm

• PostgreSQL documentation - http://www.postgresql.org/idocs/

• “Oracle to Postgre Conversion” - http://openacs.org/doc/openacs/html/oracle-to-pg-porting.html

• “PostgreSQL JDBC 2.0 compliance” - http://lab.applinet.nl/postgresql-jdbc/

• An important source of information is the PostgreSQL mailing lists: http://archives.postgresql.org/