The bot does not response to some of the creation ways

[tetelevm]

Issue I am facing

Hello! I found a strange thing for me. (code below)

I usually run the bot the way app_2 runs (sorry if that's clumsy :) ). But I decided to put the non-asynchronous things in a separate function like in app_1, and the bot stopped responding. In the example below, I moved the building of the bot to a special function, and the bot stopped working.

I found out that the problem happens in Application._update_fetcher in the line update = await self.update_queue.get(). The problem is that update is always successfully queued but never read.

However, the first time it is read normally as long as the queue is not empty (for example, if you timeout before the loop and have time to fill the queue). But when update_queue.qsize() == 0, await just freezes and is no longer executed, even if the queue fills again.

Since I found the bug and fixed it locally, I don't need any help with that, I'm just interested in why this might be happening? Maybe it has something to do with the asynchronous event loop or something like that?

Traceback to the issue

No response

Related part of your code

import asyncio
from telegram.ext.filters import ChatType
from telegram.ext import Application, MessageHandler

# vvv

def create_app_1(token: str):
    return Application.builder().token(token).build()

async def run_app_1(app_1: Application):
    app_1.add_handler(MessageHandler(ChatType.PRIVATE, send, block=False))
    await app_1.initialize()
    await app_1.updater.start_polling()
    await app_1.start()
    print("Bot 1 is running!")

async def run_app_2(token: str):
    app_2 = Application.builder().token(token).build()
    app_2.add_handler(MessageHandler(ChatType.PRIVATE, send, block=False))
    await app_2.initialize()
    await app_2.updater.start_polling()
    await app_2.start()
    print("Bot 2 is running!")

# ^^^

async def send(update, context):
    await update.effective_chat.send_message("TEXT")

async def pulling(*coros):
    await asyncio.gather(*coros)
    while True:
        await asyncio.sleep(1)

TOKEN_1 = "123:TOKEN_1"
TOKEN_2 = "321:TOKEN_2"

app_1 = create_app_1(TOKEN_1)
run_coro_1 = run_app_1(app_1)

run_coro_2 = run_app_2(TOKEN_2)

asyncio.run(pulling(run_coro_1, run_coro_2))

Operating System

Ubuntu 20.04

Version of Python, python-telegram-bot & dependencies

python3.8
python-telegram-bot==20.2

nothing else :)

[Bibo-Joshi]

Hi. I tested your example locally and it actually works prefectly fine for me. Both bots are repesonding as expected. However you meantioned "non-asynchronous things" in your description, which I don't really see in the example, so I'm wondering if your actual use case is different. Please double check that you show us an mwe that actually includes the problem :)

[tetelevm]

Hmm, that's strange. I checked again on a whole new project, and everything works just as I described. I did:

mkdir test
cd test
python3.8 -m venv env
. ./env/bin/activate
pip3 install python-telegram-bot==20.2
> bot.py
nano bot.py
# here I pasted code from the question and replaced the tokens
python3 bot.py

And the bot with token TOKEN_1 still did not answer me, but with token TOKEN_2 answered.

Well, the code, which I posted, I first tested locally, but on existing environment, currently it worked the same way on an new environment.

About non-synchronous - I meant "not coroutines", i.e. what doesn't require await (in first version I also set handler in function, but it turned out that even remote initialization breaks the bot). At the same time, both bots work fine if they are initialized through run_app_2.

[Bibo-Joshi]

If it everything works fine in a fresh venv, then tbh it sounds like something is just going wrong in your existing venv …

[tetelevm]

But no, on the contrary, the new venv still does not work correctly.

Although, to be honest, given that the problem with the queue work, most likely the problem lies deeper, somewhere in the work of asyncio.

I thought that such a problem has already occurred to someone, but if not - that's okay, I was just curious to know why it is so :)

[Bibo-Joshi]

Mh, okay. Would you mind testing on python 3.10 or 3.11 again? That's what I have been using …

[tetelevm]

Yes, I was going to do that myself, but I've been busy.
I tested on versions Python3.8-12, and it doesn't work on versions 3.8 and 3.9, but it works fine on 3.10, 3.11 and 3.12.

[Bibo-Joshi]

I double checked as well. The hint can be seen when stopping the script with CLTR+C: on py 3.9- you'll see a line saying "… got Future attached to a different loop".

TBH I'm not entirely sure on all the details, but from what I understand, the problem is the following: run_app_2 creates and runs the app_2, but the creation happens only once you actually await the coroutine. That happens inside pulling, i.e. inside asyncio.run such that app_2 is created only once the event loop used by asyncio.run is already in place.
On the other hand, app_1 is created outside of asyncio.run_polling, where the event loop used by asyncio.run is not yet set.

This behavior has been improved in python 3.10 and hence your code works in py 3.10+.
Unfortunately, I don't think that PTB can reasonably improve this for the user …

[tetelevm]

Thank you so much, got it!