Using async/await keywords with Tk.after() method of tkinter

Question

I am creating a cryptocurrency exchange API client using Python3.5 and Tkinter. I have several displays that I want to update asynchronously every 10 seconds. I am able to update the displays every 10 seconds using Tk.after() like in this example

def updateLoans():
    offers = dd.loanOffers()
    demands = dd.loanDemands()
    w.LoanOfferView.delete(1.0, END)
    w.LoanDemandView.delete(1.0, END)
    w.LoanOfferView.insert(END, offers)
    w.LoanDemandView.insert(END, demands)
    print('loans refreshed')

    root.after(10000, updateLoans)

In order for the after method to continue to update continuously every 10 seconds the function updateLoans() needs to be passed as a callable into after() inside of the function.

Now the part that is stumping me, when I make this function asynchronous with python's new async and await keywords

async def updateLoans():
    offers = await dd.loanOffers()
    demands = await dd.loanDemands()
    w.LoanOfferView.delete(1.0, END)
    w.LoanDemandView.delete(1.0, END)
    w.LoanOfferView.insert(END, offers)
    w.LoanDemandView.insert(END, demands)
    print('loans refreshed')

    root.after(10000, updateLoans)

The problem here is that I can not await a callable inside of the parameters for the after method. So I get a runtime warning. RuntimeWarning: coroutine 'updateLoans' was never awaited.

My initial function call IS placed inside of an event loop.

loop = asyncio.get_event_loop()
loop.run_until_complete(updateLoans())
loop.close()

The display populates just fine initially but never updates.

How can I use Tk.after to continuously update a tkinter display asynchronously?

Answer 1

tk.after accepts a normal function, not a coroutine. To run the coroutine to completion, you can use run_until_complete, just as you did the first time:

loop = asyncio.get_event_loop()
root.after(10000, lambda: loop.run_until_complete(updateLoans()))

Also, don't call loop.close(), since you'll need the loop again.

---

The above quick fix will work fine for many use cases. The fact is, however, that it will render the GUI completely unresponsive if updateLoans() takes a long time due to slow network or a problem with the remote service. A good GUI app will want to avoid this.

While Tkinter and asyncio cannot share an event loop yet, it is perfectly possible to run the asyncio event loop in a separate thread. The main thread then runs the GUI, while a dedicated asyncio thread runs all asyncio coroutines. When the event loop needs to notify the GUI to refresh something, it can use a queue as shown here. On the other hand, if the GUI needs to tell the event loop to do something, it can call call_soon_threadsafe or run_coroutine_threadsafe.

Example code (untested):

gui_queue = queue.Queue()

async def updateLoans():
    while True:
        offers = await dd.loanOffers()
        demands = await dd.loanDemands()
        print('loans obtained')
        gui_queue.put(lambda: updateLoansGui(offers, demands))
        await asyncio.sleep(10)

def updateLoansGui(offers, demands):
    w.LoanOfferView.delete(1.0, END)
    w.LoanDemandView.delete(1.0, END)
    w.LoanOfferView.insert(END, offers)
    w.LoanDemandView.insert(END, demands)
    print('loans GUI refreshed')

# http://effbot.org/zone/tkinter-threads.htm
def periodicGuiUpdate():
    while True:
        try:
            fn = gui_queue.get_nowait()
        except queue.Empty:
            break
        fn()
    root.after(100, periodicGuiUpdate)

# Run the asyncio event loop in a worker thread.
def start_loop():
    loop = asyncio.new_event_loop()
    asyncio.set_event_loop(loop)
    loop.create_task(updateLoans())
    loop.run_forever()
threading.Thread(target=start_loop).start()

# Run the GUI main loop in the main thread.
periodicGuiUpdate()
root.mainloop()

# To stop the event loop, call loop.call_soon_threadsafe(loop.stop).
# To start a coroutine from the GUI, call asyncio.run_coroutine_threadsafe.