sqlalchemy-session-proxy 0.3.0

A lightweight proxy for SQLAlchemy sessions that seamlessly supports both synchronous (`Session`) and asynchronous (`AsyncSession`) usage. This library provides a unified interface to interact with SQLAlchemy sessions, making it easy to write code that works in both sync and async environments.

sqlalchemy-session-proxy

A lightweight proxy for SQLAlchemy sessions that provides a unified interface over both synchronous (Session) and asynchronous (AsyncSession) usage.

This library allows you to write database-access code that works in both sync and async environments with minimal branching, while still respecting SQLAlchemy’s execution model.

---

Features

• Unified API for Session and AsyncSession
• Automatic async detection via is_async
• Explicit dispatching to sync or async implementations
• Async-compatible method signatures for mixed environments
• Fully type-annotated for IDEs and static analysis
• No hidden magic beyond SQLAlchemy’s own async design

---

Installation

pip install sqlalchemy-session-proxy

---

Basic Usage

from sqlalchemy.orm import Session
from sqlalchemy.ext.asyncio import AsyncSession
from sqlalchemy_session_proxy.session_proxy import SqlalchemySessionProxy

Synchronous Session

session = Session(...)
proxy = SqlalchemySessionProxy(session)

proxy.add(obj)
proxy.commit()

result = proxy.execute(statement)

Asynchronous Session

async def main():
    async_session = AsyncSession(...)
    proxy = SqlalchemySessionProxy(async_session)

    proxy.add(obj)          # NOT awaitable (by SQLAlchemy design)
    await proxy.commit()    # awaitable

    result = await proxy.execute(statement)

⚠️ Important

Some methods (such as add, add_all, expire) are synchronous by design even when used with AsyncSession. This behavior follows SQLAlchemy’s official API.

---

Unified Execution Pattern

The following methods automatically dispatch based on the session type:

result = await proxy.execute(stmt)
rows = await proxy.scalars(stmt)
value = await proxy.scalar(stmt)
obj = await proxy.get(User, user_id)

• In sync mode, these methods do not require await
• In async mode, they must be awaited

---

API Overview

Core Properties

• SqlalchemySessionProxy(session)
• .session — underlying Session or AsyncSession
• .is_async — True if using AsyncSession

---

Method Compatibility Matrix

Method	Sync	Async	Notes
add	✔️	✔️	Not awaitable
add_all	✔️	✔️	Not awaitable
commit	✔️	✔️	Awaitable in async
rollback	✔️	✔️	Awaitable in async
close	✔️	✔️	Awaitable in async
flush	✔️	✔️	Awaitable in async
merge	✔️	✔️	Awaitable in async
delete	✔️	✔️	Awaitable in async
get	✔️	✔️	Awaitable in async
get_one	✔️	✔️	Awaitable in async
execute	✔️	✔️	Awaitable in async
scalars	✔️	✔️	Awaitable in async
scalar	✔️	✔️	Awaitable in async
refresh	✔️	✔️	Awaitable in async
expire	✔️	✔️	Not awaitable
expire_all	✔️	✔️	Not awaitable
expunge	✔️	✔️	Not awaitable
expunge_all	✔️	✔️	Not awaitable
is_modified	✔️	✔️	Not awaitable
in_transaction	✔️	✔️	Not awaitable
in_nested_transaction	✔️	✔️	Not awaitable
query	✔️	❌	Sync-only (legacy API)
stream	❌	✔️	Async-only
stream_scalars	❌	✔️	Async-only
run_sync	❌	✔️	Async-only

---

Notes on query()

• query() is sync-only

• Calling it on an AsyncSession proxy raises NotImplementedError

• This mirrors SQLAlchemy’s own guidance:

  • Query is legacy
  • Prefer select() for new code

---

run_sync

def legacy_fn(session: Session, value: str) -> str:
    session.add(MyModel(name=value))
    session.flush()
    return "ok"

async def main():
    async with AsyncSession(engine) as session:
        proxy = SqlalchemySessionProxy(session)
        result = await proxy.run_sync(legacy_fn, "test")

• Runs synchronous ORM logic inside async code
• Uses SQLAlchemy’s greenlet bridge
• Available only for AsyncSession

---

License

Apache-2.0

---

Author

Tercel ([e-mail])

---

Links

• GitHub: https://github.com/aipartnerup/sqlalchemy-session-proxy
• PyPI: https://pypi.org/project/sqlalchemy-session-proxy/
• Issues: https://github.com/aipartnerup/sqlalchemy-session-proxy/issues
• Discussions: https://github.com/aipartnerup/sqlalchemy-session-proxy/discussions