Skip to main content

Easy multithreading with Tkinter

Project description

tkthread

Easy multithreading with Tkinter on CPython 2.7/3.x and PyPy 2.7/3.x.

import tkthread; tkthread.patch()   # do this before importing tkinter

Background

The Tcl/Tk language that comes with Python follows a different threading model than Python itself which can raise obtuse errors when mixing Python threads with Tkinter, such as:

RuntimeError: main thread is not in main loop
RuntimeError: Calling Tcl from different apartment
NotImplementedError: Call from another thread

Tcl can have many isolated interpreters, and each are tagged to the its particular OS thread when created. Python's _tkinter module checks if the calling Python thread is different than the Tcl/Tk thread, and if so, waits one second for the Tcl/Tk main loop to begin dispatching. If there is a timeout, a RuntimeError is raised. On PyPy, a NotImplementedError is raised.

For non-Tk calls into Tcl, Python will raise an apartment RuntimeError when calling a Tcl interpreter from a different thread.

A common approach to avoid these errors involves using .after to set up periodic polling of a message queue from the Tcl/Tk main loop, which can slow the responsiveness of the GUI.

The initial approach used in tkthread is to use the Tcl/Tk thread::send messaging to notify the Tcl/Tk main loop of a call for execution. This interrupt-style architecture has lower latency and better CPU utilization than periodic polling. This works with CPython and PyPy.

The newer approach used in tkthread is to use tkthread.tkinstall() to patch Tkinter when make calls into Tcl/Tk. This only works on CPython and it does not require the Thread package in Tcl.

Usage on CPython (simplest)

For CPython 2.7/3.x, tkthread.patch() (same as tkthread.tkinstall()) can be called first, and will patch Tkinter to re-route threaded calls to the Tcl interpreter using the willdispatch internal API call.

import tkthread; tkthread.patch()
import tkinter as tk

root = tk.Tk()

import threading
def thread_run(func): threading.Thread(target=func).start()

@thread_run
def func():
    root.wm_title(threading.current_thread())

    @tkthread.main(root)
    @tkthread.current(root)
    def testfunc():
        tk.Label(text=threading.current_thread()).pack()

root.mainloop()

Usage on CPython/PyPy (compatibility/legacy)

The tkthread module provides the TkThread class, which can synchronously interact with the main thread.

from tkthread import tk, TkThread

root = tk.Tk()        # create the root window
tkt = TkThread(root)  # make the thread-safe callable

import threading, time
def run(func):
    threading.Thread(target=func).start()

run(lambda:     root.wm_title('FAILURE'))
run(lambda: tkt(root.wm_title,'SUCCESS'))

root.update()
time.sleep(2)  # _tkinter.c:WaitForMainloop fails
root.mainloop()

The tkt instance is callable, and will wait for the Tcl/Tk main loop to execute and compute a result which is then passed back for return in the calling thread. A non-synchronous version also exists that does not block:

tkt.nosync(root.wm_title, 'ALSO SUCCESS')

There is an optional tkt.install() method which intercepts Python-to-Tk calls. This must be called on the default root, before the creation of child widgets. If installed, then wrapping Tk widget calls in threaded code with tkt is not necessary. There is, however, a slight performance penalty for Tkinter widgets that operate only on the main thread because of the thread-checking indirection.

The root Tcl/Tk interpreter must be the primary interpreter on the main thread. If it is not, then you will receive a TclError of the form:

_tkinter.TclError: invalid command name "140520536224520_call_from"

For example, creating several Tk() instances and then using TkThread on those will cause this error.

A good practice is to create a root window and then call root.withdraw() to keep the primary Tcl/Tk interpreter active. Future Toplevel windows use root as the master.

Install

pip install tkthread

API

  • tkthread.patch()
    • patch Tkinter to support multi-threading.
  • tkthread.tkinstall()
    • same as .patch()
  • tkthread.call(func, *args, **kw)
    • call func on the main thread, with arguments and keyword arguments.
    • waits for return value (or raises error)
  • tkthread.call_nosync(func, *args, **kw)
    • call func on the main thread, with arguments and keyword arguments.
    • returns immediately, ignore return func return or error.
  • @tkthread.called_on_main
    • decorator to dispatch the function call on the main thread from the calling thread.
  • @tkthread.main()
    • decorator to call a function immediately on the main thread.
  • @tkthread.current()
    • decorator to call a function immediately on the current thread.
  • TkThread(root)
    • class to dispatch thread calls to Tk using thread::send

Known (and solved) Error Messages

You may receive this error when using tkthread.TkThread(root):

_tkinter.TclError: can't find package Thread

This means that Python's Tcl/Tk libraries do not include the Thread package, which is needed by TkThread.

On Debian/Ubuntu:

apt install tcl-thread

On Windows, you'll need to manually update your Tcl installation to include the Thread package.

The simpler solution is to use tkthread.patch() instead.

When using Matplotlib, you may receive a warning message that can be ignored:

UserWarning: Starting a Matplotlib GUI outside of the main thread will likely fail.

The demo/mpl_plot.py script shows an example with this message.

License

Licensed under the Apache License, Version 2.0 (the "License")

See Also

These libraries offer similar functionality, using periodic polling:

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

tkthread-0.5.1.zip (26.8 kB view details)

Uploaded Source

File details

Details for the file tkthread-0.5.1.zip.

File metadata

  • Download URL: tkthread-0.5.1.zip
  • Upload date:
  • Size: 26.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.7.1 importlib_metadata/4.8.1 pkginfo/1.8.2 requests/2.27.1 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.10.5

File hashes

Hashes for tkthread-0.5.1.zip
Algorithm Hash digest
SHA256 74aceeffec63327c7c57e17b90187256f90240e45d552b699a583acfb3000dff
MD5 39ca69f9a5bdd7afc0e5b4c5bb413690
BLAKE2b-256 079e5c0c3e84ce4f34302695b4bbdbd9b0716c9fa0cdb01854d9bea209a9e9a2

See more details on using hashes here.

Supported by

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