[BUG] Strange import behaviour? about darts

[madtoinou]

Hi @elbal,

Thank you for reporting this. I think that the modules need to be linked to the top __init__.py in order to get the expected behavior when using import darts. Just need to make sure that it does not cause other issues.

However, it is expected to be able to import the models from different levels of the library: darts.models for the sake of simplicity (thanks to the logic in the corresponding __init__.py) as visible in the documentation code example but also their "original" file, for example; darts/models/forecasting/baselines.py for NaiveMean (as displayed in the name of the class in the documentation). Not sure to understand why this could be a problem.

[elbal]

@madtoinou Thank you for your reply. I missed the example in the documentation as I went straight for the API.
I think it would make sense just to link the main modules in the main __init__.py.

A slightly different issue (?) arises for the submodules. For example:

import pandas as pd
import darts.models
import matplotlib.pyplot as plt

df = pd.DataFrame({"demand": [
    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,
    14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
    ]})
ts = darts.TimeSeries.from_dataframe(df)

darts.models.forecasting.baselines.NaiveMean(ts)
darts.utils.statistics.plot_acf(ts)

plt.show()

Works, but Pycharm does not recognize the reference baselines and throws a waring.
The reference forecasting, instead, is recognized even if it is not direcly linked in the darts.models __init__.py:
https://github.com/unit8co/darts/blob/master/darts/models/__init__.py

I suspect the reason is that in the darts.models __init__.py there are references like darts.models.forecasting.exponential_smoothing that imply darts.models.forecasting but not darts.models.forecasting.baselines.

Is it something that should be fixed or the user is suppoosed to call the classes directly form darts.models?