Logging enhancements

[haim-cohen-moonactive]

The idea is not to set logging configuration in the code. As a library mljar should not set logging preferences, and let the application do it. In addition, verbose_print messages should go to logger so the application can get those messages if required.
Proposed changes:

• Remove calls to basicConfig on automl.py and exceptions.py
• Change verbose_print to use logger.

[pplonski]

Thank you @haim-cohen-moonactive. I would like to keep some default logging configuration for the library. Do you have any proposal on how to do this?

[haim-cohen-moonactive]

basicConfig should be called as part of a main or tests. Identify what are the entry points in your project and call the basicConfig there, and not in the library core files. If someone is calling *from supervised.automl import AutoML* they should not have the library configure logging for them. What scripts do you use to run your code? put the same basicConfig calls you had on automl.py there. Please note, calling *logger = logging.getLogger(__name__)* should stay in automl.py or exceptions.py, just dont call to basicConfig in these places. mljar-supervised should be maintained as a pypi library, avoiding setting default logging configuration. Only main/tests should.

…

On Wed, Jan 31, 2024 at 2:14 PM Piotr ***@***.***> wrote: Thank you @haim-cohen-moonactive <https://github.com/haim-cohen-moonactive>. I would like to keep some default logging configuration for the library. Do you have any proposal on how to do this? — Reply to this email directly, view it on GitHub <#701 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ATFKBZIUM7RAY3VV7M7HOLDYRIYHXAVCNFSM6AAAAABCS5U3G6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMJYHE4DMMJZHE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- *Haim Cohen* *Machine Learning Engineer* mobile: +972-54-4736117 email: ***@***.*** website: moonactive.com

[pplonski]

Thanks. I understand your point. Most of users are importing package in their script/notebook and just call constructor and fit(). I would leave default logging for them, and I prefer to make it configurable. The same with verbose_print() - it should print message to terminal by default and it should be configurable.

[DanielAvdar]

Thanks. I understand your point. Most of users are importing package in their script/notebook and just call constructor and fit(). I would leave default logging for them, and I prefer to make it configurable. The same with verbose_print() - it should print message to terminal by default and it should be configurable.

There are users who like to productionize their work. Not adhering to logging standards might lead to problems with productionizing applications dependent on MLjar.

[pplonski]

OK, so we need to update the code in that why to have:

• default logging for users that just call AutoML to build models,
• and configurable logging for users that integrate AutoML in their packages.