+2 votes
If I want to try new features from a Python package (e.g. jupyter or scikit-learn), what is the best way to update the package in DSS?

Is there any list of "safe package updates" so I can see which Python packages I can safely update without impacting DSS and which are not?

edited by

1 Answer

+2 votes
Best answer

Experimentally, updates to the "core" packages used by DSS (pandas, scikit-learn, jupyter, ...) are likely to break something.

At the moment, DSS 2.3 is known not to work properly with newer Pandas (behavior changes on dates), and scikit-learn (API change). For Jupyter, note that DSS uses a patched version of the "notebook" package, installing an updated pristine API change will make the integration at least partly non functional.

The more stable packages like requests or numpy are less dangerous (but there is also less incentive to update)

We'll probably update pandas and scikit learn in the next DSS update end of March
selected by
1,195 questions
1,226 answers
11,754 users

©Dataiku 2012-2018 - Privacy Policy