iris-pgwire

PostgreSQL wire protocol server for InterSystems IRIS. Connects BI tools, Python frameworks, data pipelines, and any PostgreSQL-compatible client to IRIS databases — no IRIS-specific drivers needed.

Connection string: postgresql://user:pass@localhost:5432/USER

Quick Start

Docker (recommended)

git clone https://github.com/intersystems-community/iris-pgwire.git
cd iris-pgwire
docker compose up -d

# Test it (PGWire runs on port 5432 inside the IRIS container)
psql -h localhost -p 5432 -U _SYSTEM -d USER -c "SELECT 'Hello from IRIS!'"

Python package

pip install iris-pgwire psycopg[binary]

export IRIS_HOST=localhost IRIS_PORT=1972 \
       IRIS_USERNAME=_SYSTEM IRIS_PASSWORD=SYS IRIS_NAMESPACE=USER

python -m iris_pgwire.server

First query

import psycopg

with psycopg.connect("host=localhost port=5432 dbname=USER user=_SYSTEM password=SYS") as conn:
    with conn.cursor() as cur:
        cur.execute("SELECT COUNT(*) FROM MyTable")
        print(cur.fetchone()[0])

Verified Client Compatibility

Tested against real IRIS instances via the wire protocol:

Language	Clients
Python	psycopg3, asyncpg, SQLAlchemy (sync + async)
Node.js	pg (node-postgres)
Java	PostgreSQL JDBC
.NET	Npgsql
Go	pgx v5
Ruby	pg gem
Rust	tokio-postgres
PHP	PDO PostgreSQL

ORMs: SQLAlchemy, Prisma, Drizzle, Sequelize, Hibernate
BI tools: Apache Superset, Metabase, Grafana (standard PostgreSQL driver)

See Client Compatibility Guide for setup examples.

Key Features

pgvector operators — <=> (cosine), <#> (dot product), <-> (L2) auto-translate to IRIS VECTOR_COSINE/VECTOR_DOT_PRODUCT. HNSW indexes give 5× speedup at 100K+ vectors. See Vector Operations.

DDL compatibility — Automatic public ↔ SQLUser schema mapping; strips fillfactor, GENERATED columns, USING btree, IF NOT EXISTS guards, and other PostgreSQL-specific DDL so ORM migrations run cleanly. See DDL Compatibility.

SQL translation — RETURNING emulation, ON CONFLICT, boolean literals, pg_catalog → INFORMATION_SCHEMA rewrites, JSON operators (-> / ->> → JSON_EXTRACT), parameterized queries.

Authentication — SCRAM-SHA-256, OAuth 2.0 (RFC 6749), IRIS Wallet credentials.

Dual backend — Embedded Python (irispython, lowest latency) or external DBAPI (standard TCP connection). Selectable via IRIS_BACKEND env var.

COPY protocol — Bulk load via COPY … FROM STDIN (~600 rows/sec on DBAPI path).

Usage Examples

Parameterized queries (psycopg3)

import psycopg

with psycopg.connect("host=localhost port=5432 dbname=USER user=_SYSTEM password=SYS") as conn:
    with conn.cursor() as cur:
        cur.execute("SELECT * FROM MyTable WHERE id = %s", (42,))
        row = cur.fetchone()

Vector similarity search

query_vector = [0.1, 0.2, 0.3]
with conn.cursor() as cur:
    cur.execute("""
        SELECT id, embedding <=> %s::vector AS score
        FROM vectors
        ORDER BY score
        LIMIT 5
    """, (query_vector,))
    results = cur.fetchall()

Async SQLAlchemy

from sqlalchemy.ext.asyncio import create_async_engine, AsyncSession, async_sessionmaker
from sqlalchemy import text

engine = create_async_engine("postgresql+psycopg://localhost:5432/USER")
SessionLocal = async_sessionmaker(engine, class_=AsyncSession)

async def query():
    async with SessionLocal() as session:
        result = await session.execute(text("SELECT * FROM MyTable"))
        return result.fetchall()

COPY bulk load

with conn.cursor() as cur:
    with cur.copy("COPY MyTable (col1, col2) FROM STDIN") as copy:
        for row in data:
            copy.write_row(row)

Documentation

Guide	Description
Installation	Docker, PyPI, Embedded Python deployment
Architecture	System design, dual backend, request flow
DDL Compatibility	PostgreSQL DDL transformations
Vector Operations	pgvector syntax, HNSW indexes
Client Compatibility	Per-language setup and caveats
Deployment	Production setup, SSL/TLS, auth
Performance	Benchmarks, tuning
Developer Guide	Development setup, contribution guidelines

Development

# Install dependencies
uv sync --frozen

# Run unit + contract tests (no IRIS needed)
pytest tests/unit/ tests/contract/ -v

# Run with live IRIS (container must be up)
docker compose up -d
PGWIRE_BACKEND_TYPE=dbapi PGWIRE_POOL_SIZE=1 pytest tests/ -v

# Code quality check
python -m iris_pgwire.quality

Test coverage: 92% (5349 tests)
Code quality: black (formatter), ruff (linter), bandit (security)

Known Limitations

See KNOWN_LIMITATIONS.md for the full list. Key items:

IRIS Community Edition: 5-user connection limit — use PGWIRE_POOL_SIZE=1 for dev
No native SSL termination — use nginx/HAProxy in front for TLS
Kerberos/GSSAPI auth wiring deferred (OAuth 2.0 is the recommended enterprise auth)
INFORMATION_SCHEMA only, no pg_catalog tables (translated automatically)

Contributing

git clone https://github.com/intersystems-community/iris-pgwire.git
cd iris-pgwire
uv sync --frozen
docker compose up -d
pytest tests/unit/ tests/contract/ -v

Open an issue or PR on GitHub.

Links

MIT License — see LICENSE

Metadata

Release files for iris-pgwire 1.7.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for iris-pgwire 1.7.3
File	Size	Uploaded
iris_pgwire-1.7.3.tar.gz	395.3 kB	Aug 26, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for iris-pgwire 1.7.3
File	Interpreter	ABI	Platform	Reset
iris_pgwire-1.7.3-py3-none-any.whl 462.9 kB Aug 26, 2026	Python 3	none	any	Details

Total release size: 858.2 kB

Release files / iris_pgwire-1.7.3.tar.gz

Download URL	iris_pgwire-1.7.3.tar.gz
Size	395.3 kB
Tags	Source
SHA-256 checksum How to use checksums	`9d631bc362e5042fc7f0b67dc27fe55d6ccaa201a1c456b81d4c26b36321c9ea`
BLAKE2b-256 checksum How to use checksums	`919cf5f6040f9982bcd6e1947dfb2f82c157ac9e3babd2bf3130a6af4a5c071a`
Upload date	Aug 26, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.12`

Release files / iris_pgwire-1.7.3-py3-none-any.whl

Download URL	iris_pgwire-1.7.3-py3-none-any.whl
Size	462.9 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`3f6dfd9c43164fa91fea4f6ee90e05cfc25ae319273d76232e2c6b3d5a975941`
BLAKE2b-256 checksum How to use checksums	`1ac7685faf954a60a12a361af66f2c6bc75bb8a0f2a624460001edd122c58ff2`
Upload date	Aug 26, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.12`