[python] Allow to register custom logger in Python-package

[StrikerRUS]

Centralize all logging logic in one place.
Fixed #3805.
Related #3156.

[guolinke]

LGTM

[StrikerRUS]

@jameslamb

Can these two logger calls (this is the all logs we have from Dask so far) be replaced with new _log() function?

LightGBM/python-package/lightgbm/dask.py

Line 28 in ac706e1

logger = logging.getLogger(__name__)

LightGBM/python-package/lightgbm/dask.py

Line 260 in ac706e1

logger.warning('Parameter tree_learner not set. Using "data" as default')

LightGBM/python-package/lightgbm/dask.py

Line 263 in ac706e1

logger.warning('Parameter tree_learner set to %s, which is not allowed. Using "data" as default' % tree_learner)

Please check this: dmlc/xgboost#6609 (comment).

The logging module should be preferred over the simple print, since Dask has a good integration with the logging module. See related discussion at rapidsai/cuml#3350.

[jameslamb]

I really like this! To answer your question...yes I think this can and should be used by the code in dask.py as well.

I left a suggestion for the a change to the API, though.

[jameslamb]

instead of using a flag for the log level, I'd prefer functions _log_warn() and _log_info(). I think that makes the code easier to understand, and is easier to extend to other log levels (error, fatal, and debug). That is how most loggers are structured.

[StrikerRUS]

I don't think that adding new arguments like is_debug, is_fatal is harder and less clear than writing new functions. I believe it's more a matter of taste. But I don't have any strong preferences, so I'm going to implement your suggestion.

[StrikerRUS]

@jameslamb

yes I think this can and should be used by the code in dask.py as well.

Done in 48c4058.

instead of using a flag for the log level, I'd prefer functions _log_warn() and _log_info()

Done in 7247790.

[jameslamb]

looks great! Really nice addition

[github-actions]

This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.