Write a Requests adapter about hyper

[Lukasa]

By the way, @michaelhelmick, you may end up wanting to keep an eye on this issue, I plan to play with Twython a bit here.

[michaelhelmick]

👍 I'll help be a Guinea Pig :) just let me know if I can so anything. I need to get back to some open source stuff. Been keeping busy with work D:

[Lukasa]

So mostly this is fine, except for the fact that Twython relies on urllib3's transparent content decoding, which hyper doesn't have. Awkward.

[Lukasa]

Tracking that issue in #20.

[Lukasa]

Current draft adapter is on the requests_adapter branch. Not sure how best to test it yet.

[sigmavirus24]

@Lukasa with the integration test fixtures?

[Lukasa]

That might be the way to go.

[Lukasa]

Alright, the adapter got checked in as part of 799e37e. It's been enhanced a few times since as well. @michaelhelmick, if you want to start messing about with it go ahead. I've confirmed in a quick local test file that it works in the basic case, as you can see below. Note that you'll need to install hyper from GitHub, not from PyPI, as I haven't done a new release yet.

from hyper.contrib import HTTP20Adapter
from twython import Twython

API_KEY = ...
API_SECRET = ...
ACCESS_TOKEN = ..
ACCESS_SECRET = ...

t = Twython(API_KEY, API_SECRET, ACCESS_TOKEN, ACCESS_SECRET)
a = HTTP20Adapter()
t.client.mount('https://www.twitter.com', a)
t.client.mount('https://twitter.com', a)
t.client.mount('https://api.twitter.com', a)

print(t.search(q='python'))

[Lukasa]

You can now get the Transport Adapter in v0.0.2 from PyPI. =)

[michaelhelmick]

I replied via email, but it never sent I guess.. haha.

👌 I'm not free until Saturday to test, but I'll be sure to make time to test then!

[michaelhelmick]

Just an update, I got sick from moving our office this past week (with rain and dust and everything) and didn't feel up to testing this. I'll try to test some stuff by mid-net week! Sorry!!

[Lukasa]

That's not a problem at all @michaelhelmick, there's no rush here. I'm doing some basic performance analysis today anyway. =)

[Lukasa]

With that said, I'm getting some insane numbers that don't make any sense to me. It seems that, in HTTP/1.1, we're receiving ten times as much data as we are in HTTP/2.0. I have a question: does Twython turn off the Accept-Encoding headers in Requests?