RFE: allow to selectively disable loggers from command-line

[ankostis]

A common debugging strategy is to study the logs.
But sometimes the program issues excessive logging messages,
necessitating the selective disabling of babbler loggers.

This SO captures the crux & solution of this Request For Enhancement:
https://stackoverflow.com/a/57002853/548792

Although the proposed SO solution of a new --disable-log option works perfectly fine,
it is annoying to have to patch every new project,
plus, it does not support auto-completion e.g. on bash.

• Would it make sense to include such capability into core code?

In any case, I'm suggesting the new option to be called --logger-disabled,
to fit with the existing option names starting with --log-....

[RonnyPfannschmidt]

i believe this could play into #7417

[symonk]

Back after some time out :)

What do we think about a --suppress-logger= appendable parsearg option here, which takes a list of logger names (convert to set to avoid duplicates) and doing the following:

• add NullHandler() to the logger to avoid warnings from last resort writing to stderr
• mark propagate = False to avoid events being passed to ancestor handlers

Let me know, thanks :) its where I'm thinking to 'start' with this

[ankostis]

As i suggested in the opening-post, i would choose an option name starting from --log-..., to fit with existing options related to the logging subsystem (e.g. in the --help message).

[ankostis]

Additionally, i would allow a separate option for each logger suppressed, to specify as value whether it should propagate=false/true.

[dluiscosta]

@symonk, are you still working on this and on #7417?

I've been looking through the issues labeled with status: easy in hopes of making a first contribution. This issue stood out to me since I've had problems with excessive logs during tests myself. Also, a possible solution doesn't seem to involve any subjects I'm too unfamiliar with.

[silvarthur]

Hello, @ankostis. Could you assign me to this issue?