Deprecate dataset errors with module __getattr__

[deepyaman]

NOTE: Kedro datasets are moving from kedro.extras.datasets to a separate kedro-datasets package in
kedro-plugins repository. Any changes to the dataset implementations
should be done by opening a pull request in that repository.

Description

Deprecating using DeprecatedClassMeta only deprecates the name when a deprecated object is instantiated. In the case of errors, they are often used without instantiation (e.g. in except blocks), so DeprecatedClassMeta doesn't work.

Development notes

Use __getattr__ approach for deprecation. This is a relatively-newer method (added in 3.7, see https://peps.python.org/pep-0562/).

Checklist

• Read the contributing guidelines
• Opened this PR as a 'Draft Pull Request' if it is work-in-progress
• Updated the documentation to reflect the code changes
• Added a description of this change in the RELEASE.md file
• Added tests to cover my changes

[astrojuanlu]

Thanks a lot @deepyaman! Left a few suggestions for improvement.

[noklam]

This looks good. Can we do a quick manual test it works for it will works for import and importlib?

[noklam]

some minor comments on styling, I have done manual testing and it's working, but I have some questions.

Questions:
We will get different a error when we do from xxx import versus importlib.import_module, is this fine? I think it should be fine, as this behavior is consistent when I test a 3-rd party library like pandas, we didn't change anything, but just want to check to make sure.

ref: the code I used for manual test

# Expected

from kedro.io import DataSetNotFoundError

# Expected
import importlib
io = importlib.import_module("kedro.io")
io.DataSetNotFoundError

# ImportError
from kedro.io import DUMMY

# AttributeError
io.DUMMY

[astrojuanlu]

To your comment, I get the exact same error using import or importlib:

In [1]: from kedro.io import DataSetNotFoundError
/Users/juan_cano/Projects/QuantumBlack Labs/kedro/kedro/io/__init__.py:37: DeprecationWarning: DataSetNotFoundError has been renamed to DatasetNotFoundError, and the alias will be removed in Kedro 0.19.0
  return getattr(kedro.io.core, name)

In [1]: import importlib

In [2]: kedro_io = importlib.import_module("kedro.io")

In [3]: kedro_io.DataSetNotFoundError
/Users/juan_cano/Projects/QuantumBlack Labs/kedro/kedro/io/__init__.py:37: DeprecationWarning: DataSetNotFoundError has been renamed to DatasetNotFoundError, and the alias will be removed in Kedro 0.19.0
  return getattr(kedro.io.core, name)
Out[3]: kedro.io.core.DatasetNotFoundError

In [4]: kedro_io.DatasetNotFoundError
Out[4]: kedro.io.core.DatasetNotFoundError

In [5]: getattr(kedro_io, "DatasetNotFoundError")
Out[5]: kedro.io.core.DatasetNotFoundError

In [6]: getattr(kedro_io, "DataSetNotFoundError")
/Users/juan_cano/Projects/QuantumBlack Labs/kedro/kedro/io/__init__.py:37: DeprecationWarning: DataSetNotFoundError has been renamed to DatasetNotFoundError, and the alias will be removed in Kedro 0.19.0
  return getattr(kedro.io.core, name)
Out[6]: kedro.io.core.DatasetNotFoundError

[noklam]

@astrojuanlu Yes I have done the same and verify it works, they are only different when you imported a invalid module/class, but the behavior is consistent with any library so I think it's correct.

[astrojuanlu]

💯

[astrojuanlu]

I don't understand why DCO was happy until now, but then when I updated the branch it started complaining about all the previous commits, and even after I removed my merge commit, it keeps complaining. @deepyaman do you want to do the honors and signoff everything?