Re: Please backport Python scikit learn, pandas and seaborn

[Andres Cimmarusti <acimmarusti@gmail.com>]

Hello

> With the Stretch freeze and release approaching, I don't see why
> we should spend effort based on a request for a package with
> a popcon of ~350 installations.

This is a fair point. It might be just easier to go ahead and upgrade
to Stretch. But let me point out the effort was made for Wheezy, where
pandas was backported. What made that any special?

> you might want start by making sure the packages will be part of Stretch, trying
> to do the no-change backports required, and testing them.
>
> E.g. right now seaborn, pandas will be kicked out of Stretch in 18 days, because of
> statsmodels [1]

This issue is supposedly fixed with the statsmodels version currently
in Sid which is due to migrate in 5 days. So this point is no longer
valid.

> scikit-learn is FTBFS on ppc64el, so a backport is impossible
>
I tried building the packages on Jessie from the Stretch source, but
both Pandas and sklearn failed. I did not try seaborn. Joblib worked
flawlessly.
I will elaborate later, I didn't have time to try much, as I'm in a hurry now.

Thanks,
Andres