@Bibo-Joshi Looks good in general. :-)

I thought that it would be better if python-telegram-bot would add python-telegram-bot-raw as a dependency (instead of vendoring the same files) but maybe your approach is simpler.

@Bibo-Joshi Oh, and we'd need to update the wiki on how to release a new version.

@Bibo-Joshi Following up on my first comment:
What would happen if someone would try to pip install both packages?

@tsnoam Seems like I didn't test thoroughly enough earlier; tg/ext is still included in ptb-raw.whl. Will have to fix that before testing what happens when installing both packages

@Bibo-Joshi Make sure to rm -rf build/ dist/
Might be left overs

@Bibo-Joshi Following up on my first comment:
What would happen if someone would try to pip install both packages?

Tested and it seems like the union of the files, i.e. installing first ptb-raw and then ptb gives ptb and so does first installing ptb and then ptb-raw.

@Bibo-Joshi And what happens if you just uninstall one of them afterwards?

@Bibo-Joshi And what happens if you just uninstall one of them afterwards?

uninstalling only ptb-raw leaves tg/ext in place and deletes everthing else. uninstalling only ptb removes everything (except the dist-info). that's not perfect but to be expected given that both modules share the same name. Would defintely put a big warning "install only one!" in the readme

Big warning should be fine in this situation imo

Check if we can drop more dependencies for ptb-raw.

Can we drop the decorator module? It was added since inspect.getfullargspec couldn't preserve the signatures of @log methods. However, if we switch back to functools.wraps() and use inspect.signature and inspect.Parameter which seem to get the right signatures of decorated functions, we could drop the decorator module, yes?

Check if we can drop more dependencies for ptb-raw.

Can we drop the decorator module? It was added since inspect.getfullargspec couldn't preserve the signatures of @log methods. However, if we switch back to functools.wraps() and use inspect.signature and inspect.Parameter which seem to get the right signatures of decorated functions, we could drop the decorator module, yes?

you're absolutely right 👍 Already did that locally, just didn.t pubsh the changes yet :D

Edit: Switching to functools.partial has the side effect that for unexpected arguments the TypeError happens inside the decorator, not inside the actual function, leading to tracbacks like

Traceback (most recent call last):
  File "foo.py", line 77, in <module>
    main()
  File "foo.py", line 66, in main
    message = updater.bot.send_dice(1145108092, foo=123)
  File "…\python-telegram-bot\telegram\bot.py", line 118, in decorator
    result = func(*args, **kwargs)
TypeError: send_dice() got an unexpected keyword argument 'foo'

However, as TypeError: send_dice() … still clearly states the method, I think that's something we can live with …

A first test-release can be viewed at https://test.pypi.org/project/python-telegram-bot-raw/

Oh and I extended the kwargs for setup a bit, i.e. added some links and most importantly set python_requires='>=3.6'

All right, the latest commits remove the ptb-raw pytz dependency and move the implementation of utils.{webhookhandler, promise} to a new ext.utils. However, I created shortcuts in utils.{webhookhandler, promise} in order to maintain backwards compatibility. Importing them in ptb-raw will ofc fail, but at least utils.Promise is something I can image users importing for checks like

out = bot.send_message(…)
if isinstance(out, Promise):
   ...
else:
  ...

in the context of MessageQueue. I added a deprecation notice and we can remove the shortcuts in one of the next major releases.

While typing I see that test fail timezone related, will have to have another look …

For some reason codecov results don't appear anymore, but on the codecov page it looks fine.