Investigate why support for modern kedro datasets breaks for kedro<0.19.7

[Galileo-Galilei]

Description

#590 introduced support for modern kedro datasets, i.e. with load and save defined on user side instead of private _load and _save. This breaks my tests locally for kedro<0.19.7

FYI @deepyaman, I'll take a stab at it.

Context

I should ensure retrocompatiblity, and this change isn't

Steps to Reproduce

Install the last version of master , install kedro <0.19.7, run tests.

Expected Result

Actual Result

-- If you received an error, place it here.

-- Separate them if you have more than one.

Your Environment

• kedro and kedro-mlflow version used (pip show kedro and pip show kedro-mlflow):
• Python version used (python -V):
• Operating system and version:

Does the bug also happen with the last version on master?

[deepyaman]

@Galileo-Galilei Do you have a stack trace handy, by chance? The only thing I can imagine is if super().save and super().load aren't defined/raise attribute error, in which case may need the if condition to be if hasattr(super(), "save") and ..., etc.

[Galileo-Galilei]

There is an "obvious" (after 1 hour of debugging >< ) error in the code : his line misses a return statement and for the old dataset style, it leads to return None when loading.

I am a bit worried that is was not caught by this test in the first place 😕 (EDIT: Got it : when you test with kedro>=0.19.7, the MlflowArtifactDataset._load check against loadwrapped, and enters the first condition. Hard to test all possible combination of kedro x mlflow x python x os in the CI, it would be very expensive*)

[deepyaman]

There is an "obvious" (after 1 hour of debugging >< ) error in the code : his line misses a return statement and for the old dataset style, it leads to return None when loading.

Gah, sorry!!

[Galileo-Galilei]

No problem, I didn't see it either!