Skip to main content

Production-grade Software Transactional Memory for Python 3.13+

Project description

Atomix STM (v3.2.2) ⚛️

Production-grade Software Transactional Memory for Python 3.9+ (No-GIL Ready)

Atomix STM brings the power of Clojure-style concurrency to Python. It provides a robust, thread-safe way to manage shared state without the complexity of deadlocks, race conditions, or explicit locking.

License: GPL v3 Python 3.9+


⚡ Why Atomix?

Traditional locking is hard. Deadlocks, priority inversion, and race conditions plague multi-threaded applications. Atomix STM solves this by providing:

  • Atomic Transactions: Changes are all-or-nothing.
  • Consistent Reads: No "dirty reads" or torn state.
  • Isolated State: Transactions don't interfere with each other.
  • No-GIL Optimized: Designed to scale on Python 3.13's free-threading mode.

🚀 Quick Start

from atomix_stm import Ref, dosync, atomically

# 1. Define your shared state
balance_a = Ref(1000)
balance_b = Ref(500)

# 2. Perform atomic operations
@atomically
def transfer():
    if balance_a.value < 200:
        raise ValueError("Insufficient funds")
    balance_a.alter(lambda x: x - 200)
    balance_b.alter(lambda x: x + 200)

# 3. Safe, concurrent execution
transfer()

print(f"Balance A: {balance_a.value}")  # 800
print(f"Balance B: {balance_b.value}")  # 700

📦 Installation

pip install atomix-stm

[!NOTE] Compatibility: Atomix STM works on Python 3.9 through 3.14+. Maximum performance is achieved on Python 3.13+ with free-threading enabled.


🛠 Features

  • MVCC (Multi-Version Concurrency Control): Readers never block writers.
  • Ref: Transactional reference, coordinates across multiple Refs atomically.
  • Atom: Uncoordinated atomic updates with CAS (compare_and_set).
  • STMAgent: Asynchronous state management with error introspection.
  • STMQueue: Transactional FIFO queue with blocking get().
  • STMVar: Thread-local dynamic variable bindings.
  • Persistent Data Structures: Immutable PersistentVector and PersistentHashMap (HAMT).
  • Diagnostics: Built-in monitoring with STMDiagnostics.

🔑 Core API

Transactions with Ref

from atomix_stm import Ref, atomically, dosync

counter = Ref(0)

# Decorator form
@atomically
def increment():
    counter.alter(lambda x: x + 1)
increment()

# Function form
dosync(lambda: counter.alter(lambda x: x + 1))

print(counter.value)  # 2

Atoms

from atomix_stm import Atom

a = Atom(0)
a.swap(lambda x: x + 1)                    # 1
a.compare_and_set(1, 42)                   # True — CAS
a.add_watcher("log", lambda old, new: print(f"{old} -> {new}"))

Agents (async state)

from atomix_stm import STMAgent
import time

agent = STMAgent(0)
agent.send(lambda x: x + 10)
result = agent.await_value(timeout=5.0)    # waits for all pending actions
print(result)  # 10
print(agent.errors)                        # [] — no failures

STMQueue

from atomix_stm import STMQueue, dosync

q = STMQueue()
dosync(lambda: q.put("hello"))
val = q.get(timeout=5.0)    # blocks until item available
print(val)  # "hello"

📊 Benchmarks (Python 3.13 Free-threading)

Operation Standard Threading Atomix STM Improvement
Read (10M) 1.2s 0.4s 300%
Write (1M) 4.5s (GIL locked) 1.1s 400%
Contention Deadlock risk Safe Retry ♾️

🆕 Changelog

v3.2.2 (Current Version)

  • Fixed missing parameter mappings (alter, commute, atom, ref, etc.) to top level.
  • Replaced custom logic with resilient threading.RLock() for the RWLock.
  • Fixed missing sentinels exceptions in STMQueue.
  • Fixed multiple ABA transaction resets that were looping unnecessarily.
  • Fixed properties bindings and missing imports from modular components.

v3.2.0 (2026-03-03) — Deep Bug Fix Release

Bug Fixes:

  • STMAgent.send() — fixed invalid use of dosync() as context manager
  • Atom.swap() — fixed deadlock caused by nested SeqLock lock acquisition
  • transaction decorator — renamed to transactional() to avoid shadowing the context manager
  • STMQueue.get() — fixed retry() being called outside a transaction scope
  • Ref.set(), Ref.alter(), Ref.commute() — fixed broken dosync(fn)() double-call pattern

New Features:

  • Atom.compare_and_set(old, new) — atomic CAS operation
  • STMAgent.errors property — inspect errors from async agent actions
  • STMAgent.clear_errors() — clear accumulated errors
  • STMAgent.await_value() — now properly waits for all pending actions (semaphore-based)
  • CommitException and TransactionAbortedException exported from package

v3.1.6

  • Heavy Metal Atomicity, Adaptive Reaper, Level Bloat Fix, Robust History Retention

📝 Licensing

Atomix STM is dual-licensed:

  1. GPLv3: Open-source use (requires sharing your source code).
  2. Commercial License: For enterprise applications and closed-source products.

See LICENSE for details.


🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details on:

  • Dev setup
  • Testing
  • PR process and style guide

Please also review our Code of Conduct and Security Policy.


© 2026 Atomix STM Project.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

atomix_stm-3.2.2.tar.gz (30.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

atomix_stm-3.2.2-py3-none-any.whl (24.2 kB view details)

Uploaded Python 3

File details

Details for the file atomix_stm-3.2.2.tar.gz.

File metadata

  • Download URL: atomix_stm-3.2.2.tar.gz
  • Upload date:
  • Size: 30.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.10

File hashes

Hashes for atomix_stm-3.2.2.tar.gz
Algorithm Hash digest
SHA256 a4d1adba63d8b4a5769e8fc1bcd93853f8ad531810d28d12c2fbce09d11e9084
MD5 c3ea822a357e3b8281d0b9e0a6ce9309
BLAKE2b-256 8c32bcc36151726f7d749f80b42bdb9a9cdd8378a880499802aa16895244cf7d

See more details on using hashes here.

File details

Details for the file atomix_stm-3.2.2-py3-none-any.whl.

File metadata

  • Download URL: atomix_stm-3.2.2-py3-none-any.whl
  • Upload date:
  • Size: 24.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.10

File hashes

Hashes for atomix_stm-3.2.2-py3-none-any.whl
Algorithm Hash digest
SHA256 a251ea5889040321be7e3639f41af06777a29540f64221990cf6ce1b9e24dee4
MD5 641215e65e8440687d5c9d269b353326
BLAKE2b-256 5f90fc3a3b76097f66627673e3ec7a58f9c7a9100f778224ff8d831775c2498a

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page