Add optional JSON logging

[drdavella]

Overview

Implement JSON logging per the logging spec

{"timestamp": "2023-10-18 14:36:18,122", "level": "INFO", "message": "\n[startup]", "file": "logging.py", "line": 46}
{"timestamp": "2023-10-18 14:36:18,122", "level": "INFO", "message": "codemodder: python/0.57.0", "file": "codemodder.py", "line": 135}
{"timestamp": "2023-10-18 14:36:18,122", "level": "INFO", "message": "\n[setup]", "file": "logging.py", "line": 46}
{"timestamp": "2023-10-18 14:36:18,122", "level": "INFO", "message": "running:", "file": "logging.py", "line": 53}

[codecov]

Codecov Report

Merging #83 (fd95622) into main (9aa46b3) will decrease coverage by 0.38%.
The diff coverage is 74.19%.

@@            Coverage Diff             @@
##             main      #83      +/-   ##
==========================================
- Coverage   96.31%   95.94%   -0.38%     
==========================================
  Files          48       48              
  Lines        1874     1900      +26     
==========================================
+ Hits         1805     1823      +18     
- Misses         69       77       +8     

Files	Coverage Δ
src/codemodder/cli.py	100.00% <100.00%> (ø)
src/codemodder/codemodder.py	98.11% <100.00%> (ø)
src/codemodder/logging.py	82.60% <72.41%> (-17.40%)	⬇️

[drdavella]

I don't really understand what's happening here, when I run this locally it's telling me logging.py is even more covered than it was before:

src/codemodder/logging.py                                             46      8    83%

[drdavella]

I'm tempted to ignore this block for the purposes of coverage. I tried to write some tests here but it's proving fairly difficult to get the mocks right.

[clavedeluna]

I'm super surprised codecov says most of this isn't tested. Curious, if you put a debugger in here and run codemodder as pixeebot would, it does it it, right? That would tell me integration tests would test this but right now we only calc coverage based on unit tests. This makes sense bc the way we run integration tests with subprocess.run creates a separate process and pytest / cov can't calculate coverage there.

Is there any help from the python-json-logger docs in regards to testing? Does chatgpt help in any way here? If you've tried these avenues and felt like you've spent sufficient time, I would be ok with adding no-cover here though it does break my heart a little.

[drdavella]

This codepath is only covered if we use --log-format=json which probably doesn't happen in any of the integration tests right now. This basically just boils down to level of effort vs reward right now and I can't afford to spend any more time trying to get this covered.

[clavedeluna]

Totally fine, ticket and move on. Could be a good starter task for someone who may join in the upcoming months or just an easy "end of day" task.

[clavedeluna]

<3