Support connection pooling via requests.Session

[frostbtn]

This addresses #100.

Slacker() now optionally receives requests.Session object to run requests through. If not provided it uses requests.get or requests.post as before (these create a separate session for each api call resulting in https handshake every time).

Implementation is straightforward. requests.get/requests.post are effectively inlined to utilize user provided session object.

I'm getting 30% off from mean time to request Slack (a chat archiving crawler) and like 60% off from traffic volume (https overhead).

[alexwlchan]

One nice side effect of this change is that it becomes easier to test code that uses slacker, because you can use something like betamax to cache API calls. I now have a nice little pytest fixture for creating a cached Slacker instance:

# -*- encoding: utf-8 -*-

import os

import betamax
import pytest
import requests
import slacker

SLACK_TOKEN = os.environ.get('SLACK_TOKEN', 'TEST_TOKEN')

with betamax.Betamax.configure() as config:
    config.cassette_library_dir = 'tests/cassettes'
    config.define_cassette_placeholder('<SLACK_TOKEN>', SLACK_TOKEN)

@pytest.fixture
def slack():
    session = requests.Session()
    with betamax.Betamax(session).use_cassette('slack'):
        yield slacker.Slacker(
            token=SLACK_TOKEN,
            session=session
        )

I’m installing from Git for now, but would love to see this in a public release soon. 😄

(I’ve just started a new project using slacker, and wanted to try betamax, so this is super convenient timing for me.)

[simonsolnes]

@alexwlchan Issued a version to os just now.