Logging in tokenizer.py ‘absorbs’ downstream logging

[lfashby]

Importing the segments package leads to logging messages being absorbed by global logging.basicConfig() in tokenizer.py.

For example:

import logging

logging.basicConfig(level="INFO")
logging.info('Logging to console') # Will log to console.

While:

import logging
import segments

logging.basicConfig(level="INFO")
logging.info('Logging to console') # Will NOT log to console.

[xrotwang]

If you want to use segments as library within a package, you may want to import it in function scope:

>>> import logging
>>> logging.basicConfig(level="INFO")
>>> def f():
...     import segments
...     logging.info('Logging to console')
... 
>>> f()
INFO:root:Logging to console

[kylebgorman]

Just wanted to put a +1 on @lfashby's issue---importing a third-party module inside of a function call to avoid an unnecessary side effect is very ugly!

I've created a PR that removes the basicConfig call. As far as I can tell this has no negative consequences for the larger library. But perhaps there's some (untested) desirable side effect to calling basicConfig at module scope in tokenizer.py. If so you can ignore the PR.