ClientResponse.get_encoding() throws TypeError if it hasn't been read

[Transfusion]

Long story short

ClientResponse.get_encoding() throws TypeError: object of type 'NoneType' has no len() if we don't first call text() or json() to read from it, i.e. _body is None

Related: #4112

Expected behaviour

Does not throw an error, returns some default value

Actual behaviour

Traceback (most recent call last):
  File "/home/transfusion/aiohttp/aiohttp_issue_4112.py", line 14, in <module>
    loop.run_until_complete(test())
  File "/usr/lib/python3.7/asyncio/base_events.py", line 579, in run_until_complete
    return future.result()
  File "/home/transfusion/aiohttp/aiohttp_issue_4112.py", line 10, in test
    print(r.get_encoding())
  File "/home/transfusion/aiohttp/aiohttp/client_reqrep.py", line 942, in get_encoding
    encoding = chardet.detect(self._body)['encoding']
  File "/home/transfusion/python37/lib/python3.7/site-packages/cchardet/__init__.py", line 15, in detect
    encoding, confidence = _cchardet.detect_with_confidence(msg)
  File "src/cchardet/_cchardet.pyx", line 15, in cchardet._cchardet.detect_with_confidence
TypeError: object of type 'NoneType' has no len()

Steps to reproduce

Find any webpage that doesn't include a charset parameter in its Content-Type response header; you may check headers only with curl -v -s http://example.com 1> /dev/null

import aiohttp
import asyncio


async def test():

    async with aiohttp.ClientSession() as session:
        r = await session.get('http://www.africau.edu', ssl=None)
        # print(await r.text())
        print(r.get_encoding())

if __name__ == "__main__":
    loop = asyncio.get_event_loop()
    loop.run_until_complete(test())

Your environment

4.0.0a1, Commit f356eeb

[asvetlov]

I would raise an explicit error here instead of returning a default.
raise RuntimeError(...) with an explanation text message would be good.
Can you propose a pull request?

[Purusah]

Hello, does this issue still actual? I'd like to propose PR

[asvetlov]

please do