Don't install requests with the use_chardet_on_py3 extra

[spakin]

This extra seems no longer to exist. I receive

WARNING: requests 2.25.1 does not provide the extra 'use_chardet_on_py3'

and later

Traceback (most recent call last):
  File "/home/pakin/.anaconda3/lib/python3.8/site-packages/pkg_resources/__init__.py", line 2739, in requires
    deps.extend(dm[safe_extra(ext)])
KeyError: 'use_chardet_on_py3'

As a result, I was unable to use, in particular, the D-Wave Problem Inspector.

[codecov-commenter]

Codecov Report

Merging #476 (f624e58) into master (502d68e) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #476   +/-   ##
=======================================
  Coverage   83.60%   83.60%           
=======================================
  Files          16       16           
  Lines        2641     2641           
=======================================
  Hits         2208     2208           
  Misses        433      433           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 502d68e...f624e58. Read the comment docs.

[randomir]

https://docs.python-requests.org/en/master/community/updates/#id1

It looks like we should bump the requests lower bound instead.

Also, see #471.

[Ousret]

Hi,

I am aware of what caused #471 and was fixed recently. You should be able to remove Chardet as a dependency. In case there are any concerns left, do not hesitate to reach.

Also, for this matter at hand, don't use the [use_chardet_on_py3] but add explicitly chardet instead, will do the same without breaking anything for people having requests < 2.26.

Regards,

[randomir]

Thanks @Ousret for adding more context.

In summary:

• requests<2.26 uses chardet to detect character encoding in a response (even for JSON decoding)
• requests>=2.26 uses charset-normalizer~=2.0.0 which fails to provide a meaningful encoding of some JSON strings, like {}, and in turn, response .json() decoding fails.
• requests>=2.26 will prefer chardet over charset-normalizer, if chardet is installed.
• You've fixed [BUG] Wrong encoding detected for empty JSON response Ousret/charset_normalizer#58 in charset-normalizer>=2.0.2, some 12 days after the 2.0.0 release, and just one day after requests==2.26 was released.
• A client that expects specific encoding (e.g. UTF-8 for JSON), can specify the encoding explicitly, like the original bug reporter did in tests: unb0rk tests broken by requests 2.26.0. googleapis/python-cloud-core#118.
• The complete charset guessing game might stop in the next major requests release, as said in Make chardet/charset_normalizer optional? psf/requests#5871 (comment)

Given all that, we can safely merge this change. Thanks @spakin for bringing it up.

In a follow-up PR I'll explicitly specify the encoding, as that should make response decoding not just resilient to future charset guessing issues, but also faster.

[spakin]

Thanks for merging the pull request. I didn't realize until I read #476 (comment) what a mess requests has produced.