With the release of PyGObject 3.50.0, GBulb is no longer required. PyGObject now contains native integration with the Python event loop. If you require asyncio support in your GTK app, we advise upgrading your project to use the native asyncio API of PyGObject 3.50.0+.
Example usage of PyGObject with native asyncio support:
import asyncio import gi gi.require_version("Gtk", "3.0") from gi.events import GLibEventLoopPolicy from gi.repository import Gtk asyncio.set_event_loop_policy(GlibEventLoopPolicy()) app = Gtk.Application(...) app.run()
The last published version of this project (v0.6.6) is compatible with versions of PyGObject prior to 3.50.0, and CPython 3.8-3.13.
Gbulb is a Python library that implements a PEP 3156 interface for the GLib main event loop under UNIX-like systems.
As much as possible, except where noted below, it mimics asyncio's interface. If you notice any differences, please report them.
- python 3.8+
- pygobject
- glib
- gtk+3 (optional)
Example usage:
import asyncio, gbulb gbulb.install() asyncio.get_event_loop().run_forever()
Example usage:
import asyncio, gbulb gbulb.install(gtk=True) asyncio.get_event_loop().run_forever()
Example usage:
import asyncio, gbulb gbulb.install(gtk=True) # only necessary if you're using GtkApplication loop = asyncio.get_event_loop() loop.run_forever(application=my_gapplication_object)
See examples/wait_signal.py
- Windows is not supported, sorry. If you are interested in this, please help me get it working! I don't have Windows so I can't test it.
In GLib, the concept of event loop is split in two classes: GLib.MainContext and GLib.MainLoop.
The event loop is mostly implemented by MainContext. MainLoop is just a wrapper that implements the run() and quit() functions. MainLoop.run() atomically acquires a MainContext and repeatedly calls MainContext.iteration() until MainLoop.quit() is called.
A MainContext is not bound to a particular thread, however it cannot be used by multiple threads concurrently. If the context is owned by another thread, then MainLoop.run() will block until the context is released by the other thread.
MainLoop.run() may be called recursively by the same thread (this is mainly used for implementing modal dialogs in Gtk).
The issue: given a context, GLib provides no ways to know if there is an existing event loop running for that context. It implies the following divergences with PEP 3156:
.run_forever()
and.run_until_complete()
are not guaranteed to run immediately. If the context is owned by another thread, then they will block until the context is released by the other thread..stop()
is relevant only when the currently running Glib.MainLoop object was created by this asyncio object (i.e. by calling.run_forever()
or.run_until_complete()
). The event loop will quit only when it regains control of the context. This can happen in two cases:- when multiple event loop are enclosed (by creating new
MainLoop
objects and calling.run()
recursively) - when the event loop has not even yet started because it is still trying to acquire the context
- when multiple event loop are enclosed (by creating new
It would be wiser not to use any recursion at all. GLibEventLoop
will
actually prevent you from doing that (in accordance with PEP 3156), however
GtkEventLoop
will allow you to call run()
recursively. You should also keep
in mind that enclosed loops may be started at any time by third-party code
calling GLib's primitives.
Testing GBulb requires a Linux environment that has GLib and GTK development libraries available.
The tests folder contains a Dockerfile that defines a complete testing environment. To use the Docker environment, run the following from the root of the git checkout:
$ docker buildx build --tag beeware/gbulb:latest --file ./tests/Dockerfile . $ docker run --rm --volume $(PWD):/home/brutus/gbulb:z -it beeware/gbulb:latest
This will drop you into an Ubuntu 24.04 shell that has Python 3.8-3.13 installed, mounting the current working directory as /home/brutus/gbulb. You can use this to create virtual environments for each Python version.
Once you have an active virtual environment, run:
(venv) $ pip install -e .[dev] (venv) $ pytest
to run the test suite. Alternatively, you can install tox, and then run:
# To test a single Python version (venv) $ tox -e py
# To test Python 3.10 specifically (venv) $ tox -e py310
# To test all versions (venv) $ tox
gbulb is part of the BeeWare suite. You can talk to the community through:
We foster a welcoming and respectful community as described in our BeeWare Community Code of Conduct.
If you experience problems with gbulb, log them on GitHub. If you want to contribute code, please fork the code and submit a pull request.