add contrib.discord: first Discord Bot!

[casperdcl]

• add contrib.discord (similar to contrib.telegram)
• document
• abstract common features with contrib.telegram
  • fix lazy large memory usage and discard unsent messages

Steps to test:

• pip install -U 'git+https://github.com/tqdm/tqdm@discord#egg=tqdm'
• create a discord bot (not public, no requirement of OAuth2 code grant, only send message permissions) & invite it to a channel
• copy the bot token & channel ID and paste below:

from tqdm.contrib.discord import tqdm, trange
for _ in trange(
	int(1e8),
	token='THIS1SSOMETOKEN0BTAINeDfrOmD1SC0rd',
    channel_id=0246813579,
):
    # worst-case empty for-loop
    # still does ~6M it/s, i.e. overhead ~170 ns/it
	pass

Result:

[codecov-commenter]

Codecov Report

Merging #976 into master will decrease coverage by 1.52%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master     #976      +/-   ##
==========================================
- Coverage   86.90%   85.37%   -1.53%     
==========================================
  Files          21       22       +1     
  Lines        1260     1279      +19     
  Branches      215      218       +3     
==========================================
- Hits         1095     1092       -3     
- Misses        144      166      +22     
  Partials       21       21              

[oohwooh]

Hi so I'm really glad I found this because it's exactly what I wanted for a project I'm working on, but I'm wondering if there's any reason you opted to use disco-py instead of discord.py? From my experience discord.py is much more reliable and better maintained, and it currently looks like there's an issue with Python 3.8 compatibility for disco-py
would you be opposed to me porting this to use discord.py instead? The libraries look to be structured similarly so it would probably be only a few code changes.

Thoughts?

[oohwooh]

I take back what I said about it being an easy fix, just realized disco is synchronous and discord.py isn't. I'm not sure what a clean solution would be to overcome that. Looking into it

[casperdcl]

Apart from the async issue and need to mess with mainloops, discord.py is also py3 only; whereas disco works perfectly fine with tqdm on a lot of old py2 code.

Also as a matter of personal taste, discord.py seems to have an extremely unpythonic design which goes against what I try to do with tqdm.

I think both discord libs are maintained - I don't particularly care about how much - but I'd be very interested in why you say one is more reliable than the other. If disco really is unreliable then I'd probably prefer switching to just plain requests like with https://github.com/tqdm/tqdm/blob/master/tqdm/contrib/telegram.py

[casperdcl]

About the py3.8 issue, is it b1naryth1ef/disco#178? It it similar to b1naryth1ef/disco#107 (i.e. a gevent issue)?

[oohwooh]

Thanks for the quick response! I appreciate the insight.
Reliable probably wasn't the best choice of words, in hindsight. What I was trying to say with that is that the current issue with installing the package from pip seems to have been present for a while. I just managed to resolve that one by using a constraints.txt file to ensure gevent was installed with an up-to-date version, and after doing that tqdm.contrib.discord works as expected.

Totally understand the matter of personal preference, and given that I think it's very reasonable for this code to just stay as is, and then when the gevent issue eventually gets fixed the package becomes a bit easier to use.

[casperdcl]

hmm yes just posted b1naryth1ef/disco#178 (comment). I might open another PR there. They merged in my last PR pretty quickly...

[casperdcl]

b1naryth1ef/disco#183 hopefully makes things more reliable :)