PostgreSQL vs Oracle

Quick verdict

PostgreSQL is an open source object-relational database developed by the PostgreSQL Global Development Group. It is released under the PostgreSQL Licence, which the project describes as a liberal open source licence similar to the BSD or MIT licences, and the project states it has no plans to change it. The current release series is PostgreSQL 18 (18.6 at the time of checking), with 19 in beta.

Oracle AI Database is Oracle's commercial relational database, previously sold as Oracle Database. The current release is 26ai, which Oracle announced as the replacement for Oracle Database 23ai, delivered by applying a release update rather than a database upgrade. Oracle's 26ai licensing manual lists Enterprise Edition (on Linux x86-64), Enterprise Edition on engineered systems, Standard Edition 2 on Oracle Database Appliance, a family of cloud offerings, and Oracle AI Database Free, a no-cost edition with fixed resource limits.

Moving from Oracle to PostgreSQL is a common project, and PostgreSQL's own documentation has a chapter on porting from PL/SQL. The sections below focus on the differences that matter when choosing between them or moving code across.

Side by side

Key differences

Licensing, editions and the free edition

PostgreSQL has one edition and no licence fee; commercial support is available from many independent companies. Oracle licenses by edition, and its 26ai Licensing Information User Manual lists many capabilities as extra-cost options on Enterprise Edition, including Real Application Clusters, Partitioning, Database In-Memory, Active Data Guard, Advanced Compression, Advanced Security, Multitenant (beyond 3 user-created PDBs per container) and the Diagnostics and Tuning packs. Several of these are included in Oracle AI Database Free and in higher cloud tiers.

Oracle AI Database Free runs on Linux x86-64, ARM Linux and Windows. Oracle documents that it limits itself to two cores, 2 GB of RAM and 12 GB of user data (exceeding the data limit raises ORA-12954), and allows one installation per logical environment, whether a VM, a container or a physical host. It is a practical way to learn Oracle or develop against it; read the Free Use Terms before relying on it for anything else.

PL/SQL vs PL/pgSQL

The languages look alike, but PostgreSQL's porting guide lists real differences: function bodies are string literals (usually dollar-quoted), RETURN in the signature becomes RETURNS, IS becomes AS, a LANGUAGE clause is required, varchar2 and number become varchar/text and numeric, and reverse FOR loops take their bounds in the opposite order.

-- Oracle (PL/SQL)
CREATE OR REPLACE FUNCTION add_tax(p_amount NUMBER) RETURN NUMBER IS
BEGIN
  RETURN p_amount * 1.2;
END;
/

-- PostgreSQL (PL/pgSQL)
CREATE OR REPLACE FUNCTION add_tax(p_amount numeric) RETURNS numeric AS $$
BEGIN
  RETURN p_amount * 1.2;
END;
$$ LANGUAGE plpgsql;

The largest structural gap is packages. PostgreSQL has none; its documentation suggests grouping functions in schemas and, because there are no package-level variables, keeping per-session state in temporary tables. Code that relies on package state or initialisation blocks needs redesign, not just translation. PostgreSQL also rolls back all changes made since a block's BEGIN when an exception is caught in that block.

NULL and the empty string

Oracle's SQL Language Reference states that the database currently treats a character value with a length of zero as null, and recommends not relying on that in future. PostgreSQL follows the SQL standard: an empty string is a value, not NULL. Queries and constraints that test for missing values can return different rows after a migration.

-- Same statements, different results
INSERT INTO people (middle_name) VALUES ('');
SELECT COUNT(*) FROM people WHERE middle_name IS NULL;
-- Oracle: the row is counted ('' was stored as NULL)
-- PostgreSQL: the row is not counted

Oracle's DATE type also holds a time of day, while PostgreSQL's date holds only the date; Oracle DATE columns usually map to timestamp in PostgreSQL.

Row limiting, identity, sequences and upserts

Oracle has supported the standard FETCH FIRST clause for several releases, and older code often uses the ROWNUM pseudocolumn, which PostgreSQL does not have. PostgreSQL accepts both LIMIT and FETCH FIRST. From release 23, Oracle also allows SELECT without a FROM clause and GROUP BY by alias or position, which removes some old porting friction (FROM dual still works).

-- Oracle (legacy and current)
SELECT * FROM (SELECT * FROM orders ORDER BY created_at DESC) WHERE ROWNUM <= 10;
SELECT * FROM orders ORDER BY created_at DESC FETCH FIRST 10 ROWS ONLY;

-- PostgreSQL
SELECT * FROM orders ORDER BY created_at DESC LIMIT 10;

Both support identity columns and sequences, with different call syntax:

-- Oracle
CREATE TABLE orders (id NUMBER GENERATED ALWAYS AS IDENTITY PRIMARY KEY, total NUMBER(10,2));
CREATE SEQUENCE invoice_seq START WITH 1000;
SELECT invoice_seq.NEXTVAL FROM dual;

-- PostgreSQL
CREATE TABLE orders (id bigint GENERATED ALWAYS AS IDENTITY PRIMARY KEY, total numeric(10,2));
CREATE SEQUENCE invoice_seq START WITH 1000;
SELECT nextval('invoice_seq');

For upserts Oracle uses MERGE; PostgreSQL has INSERT ... ON CONFLICT and also supports MERGE. String aggregation is LISTAGG in Oracle and string_agg in PostgreSQL.

-- Oracle
MERGE INTO stock t
USING (SELECT 'A1' AS sku, 5 AS qty FROM dual) s
ON (t.sku = s.sku)
WHEN MATCHED THEN UPDATE SET t.qty = s.qty
WHEN NOT MATCHED THEN INSERT (sku, qty) VALUES (s.sku, s.qty);
SELECT LISTAGG(name, ', ') WITHIN GROUP (ORDER BY name) FROM staff;

-- PostgreSQL
INSERT INTO stock (sku, qty) VALUES ('A1', 5)
ON CONFLICT (sku) DO UPDATE SET qty = EXCLUDED.qty;
SELECT string_agg(name, ', ' ORDER BY name) FROM staff;

Data types, JSON and AI features

PostgreSQL offers jsonb, a binary JSON type that its documentation recommends for most applications and that supports GIN indexes for containment and key queries, along with arrays, range types and user-defined types. Oracle 26ai lists a native JSON type (up to 32 MB per value), a SQL BOOLEAN type (previously PL/SQL only) and a VECTOR type for AI Vector Search. In PostgreSQL, vector search comes from extensions such as pgvector rather than the core server. Oracle's VARCHAR2 is limited to 4,000 bytes unless MAX_STRING_SIZE is set to EXTENDED (32,767 bytes).

High availability, cloud and tools

PostgreSQL ships streaming replication (synchronous or asynchronous, whole server) and logical replication (per table), and its documentation states that it does not offer synchronous multi-primary replication natively. Failover management usually comes from external tools or a managed service. Oracle offers Data Guard standbys, with Active Data Guard and RAC as extra-cost Enterprise Edition options, or included in the higher OCI tiers and Exadata services.

PostgreSQL is offered as a managed service by all the large clouds. Oracle runs on OCI (including Always Free Autonomous Database), inside Azure, Google Cloud and AWS data centres through Oracle Database@ services, and on Amazon RDS for Oracle, which offers License Included and Bring Your Own License models. For tools, pgAdmin is the PostgreSQL project's own GUI; Oracle provides the free SQL Developer extension for VS Code; DBeaver, DataGrip and Toad cover Oracle as well.

Pricing and licensing

PostgreSQL. Free under the PostgreSQL Licence for any use, including commercial. Paid support and managed hosting are sold by third parties and cloud providers at their own rates.

Oracle. Oracle AI Database Free costs nothing, within the limits above. Enterprise Edition is licensed separately from its extra-cost options and management packs, as set out in Oracle's Licensing Information User Manual. Oracle publishes list prices in its Technology Global Price List, but that document could not be retrieved from oracle.com when we checked, so we do not quote figures; ask Oracle or a reseller, and include options, packs and annual support in any estimate. On AWS, Amazon RDS for Oracle offers hourly License Included pricing or Bring Your Own License.

Pricing checked on the vendors' official pages on 7 October 2026. Prices change; confirm before buying.

Where each one leads

Limitations

When to choose each

When neither is right

• Your organisation standardises on Microsoft tooling and T-SQL: see SQL Server vs Oracle and SQL Server vs PostgreSQL.
• You need a simple open source database for a typical web stack: see MySQL vs PostgreSQL and MySQL vs Oracle.
• You want an embedded database inside an application: see SQLite vs PostgreSQL.

Final recommendation

Frequently asked questions

Sources

• PostgreSQL Licence
• PostgreSQL 18 documentation: Porting from Oracle PL/SQL
• PostgreSQL 18 documentation: JSON types
• PostgreSQL 18 documentation: Comparison of replication solutions
• pgvector extension (GitHub)
• Oracle AI Database 26ai documentation
• Oracle AI Database 26ai Licensing Information User Manual
• Oracle AI Database Free: licensing restrictions
• Oracle SQL Language Reference 26ai: Nulls
• Oracle SQL Language Reference 26ai: Data types
• Oracle Database blog: Oracle announces Oracle AI Database 26ai
• Amazon RDS for Oracle
• Oracle SQL Developer for VS Code documentation

Checked October 2026.

How we research comparisons: our editorial method.