dropping support for numpy recarrays - any objections?

30 views
Skip to first unread message

josef...@gmail.com

unread,
Dec 24, 2016, 11:46:49 PM12/24/16
to pystatsmodels
We still have some code including in the datasets that use and support recarrays. This dates mostly back to the time before pandas became the de facto standard for data frames.


We have a FutureWarning from numpy about recarray incompatible use, that I didn't look into yet. I haven't seen anyone using recarrays in our neighborhood, Kevin was arguing for droppings support already some time ago.

Given our lack of maintainer time, I think now also it's not worth keeping it maintained.

Is anyone using recarrays with statsmodels?

Josef

Kevin Sheppard

unread,
Dec 30, 2016, 8:01:14 AM12/30/16
to pystatsmodels
This should be dropped -- not proactively -- but not fixed if things break and tests removed.

Chad Fulton

unread,
Dec 30, 2016, 8:40:59 AM12/30/16
to Statsmodels Mailing List
On Fri, Dec 30, 2016 at 8:01 AM, Kevin Sheppard <kevin.k....@gmail.com> wrote:
This should be dropped -- not proactively -- but not fixed if things break and tests removed.

I'll add that I've never actively supported recarrays, so I have no idea if they work in the state space models.

josef...@gmail.com

unread,
Dec 30, 2016, 9:11:32 AM12/30/16
to pystat...@googlegroups.com
Recarrays are supported in the data handling for the models. The models itself will only have numpy arrays after the super.__init__ call. So if i understand correctly then all exog could. Still be recarrays.

Josef
Getting (the kids) ready for the slopes.
Reply all
Reply to author
Forward
0 new messages