Revise the logger instanciation/initial handlers

[nmaynes]

I added the logging functionality described in the proposal. I also took care to make sure the explain argument would operate the same way. I left the behavior in api.py where if explain is not set, the logger will still log messages at the WARNING level. That behavior is really up to you as the package maintainer. It is as easy as removing that branch from the if statement and adding documentation to the repository that describes how a logger must be set via the handler if an application developer so desires.

I also added two simple tests that check whether the set_stream_handler function does what it should. Apologies if the tests are not in the correct style. Let me know if anything is in need of attention or you have changed your mind about the behavior change for logging. Thanks for the awesome library.

Close #134

[Ousret]

Thanks for the submission. I have assembled some comments/remarks around the proposed changes.

[codecov-commenter]

Codecov Report

Merging #135 (c3970d4) into master (6480728) will decrease coverage by 0.03%.
The diff coverage is 91.30%.

@@            Coverage Diff             @@
##           master     #135      +/-   ##
==========================================
- Coverage   89.66%   89.62%   -0.04%     
==========================================
  Files          11       11              
  Lines        1171     1186      +15     
==========================================
+ Hits         1050     1063      +13     
- Misses        121      123       +2     

Impacted Files	Coverage Δ
charset_normalizer/api.py	88.09% <84.61%> (-0.69%)	⬇️
charset_normalizer/__init__.py	100.00% <100.00%> (ø)
charset_normalizer/utils.py	85.02% <100.00%> (+0.58%)	⬆️

---

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 6480728...c3970d4. Read the comment docs.

[Ousret]

I can see that you marked some comments as resolved but did not see any commits to refer to. Do you need any help to pursue this PR? Or is it just the lack of free time?

[nmaynes]

It is the time that is missing. I do have a commit that addresses all but one of the comments. I think the default behavior of the logger, when not set and explain is set to False should be no stream logger set. Using the NullHandler will allow messages of WARNING level or higher to be pushed to stderr on most systems. One thing I would like to add is a note in the documentation that describes the loggers behavior. How do you feel about the default behavior being set in that manner? And sorry for the slow movement on my part.

[Ousret]

No problems about the delay.

Using the NullHandler will allow messages of WARNING level or higher to be pushed to stderr on most systems.

I don't see how this can happen. Maybe I misunderstood? https://docs.python.org/3/library/logging.handlers.html#logging.NullHandler

when not set and explain is set to False should be no stream logger set

I think that we should give up on using set_stream_handler in the function from_bytes(...), we should rather have a unique handler instantiated once and added/removed from handlers. That does not mean we remove the set_stream_handler.

Lastly, the documentation should be updated to reflect those changes and explain a bit more what we are doing from the begining.

[Ousret]

LGTM now

[Ousret]

Hopefully, the principal concern is now answered. I am going forward and merging this, do not hesitate to review it and notify me in case I missed something.

[nmaynes]

Thank you for merging these changes!