planning release 0.8

[josef...@gmail.com]

(I'm back in Montreal although the kids are still on vacation.)

I think we should plan for a final 0.8.0 release in 1.5 to 2 weeks.

I still need to go over the issues and PRs that accumulated in the last month.
Several PRs have been merged already into master but still need to be
backported.
We have some new open bugs that could be fixed for 0.8.

However, the only serious release stopper is the inconsistent behavior
in predict with patsy and missing/nan interaction
https://github.com/statsmodels/statsmodels/issues/3087
patsy has been dropping rows with nans while the numpy interface
doesn't. This causes currently an exception with adding a pandas index
to the prediction which was added in 0.8.

Also, we have a cointegration test that is not in the official API and
documentation which has bugs. However, this function has now been
advertised in blog posts and should be fixed to avoid incorrect
results when users use it.

Josef

[josef...@gmail.com]

spot checking the MacPython wheel build on
https://travis-ci.org/MacPython/statsmodels-wheels

https://travis-ci.org/MacPython/statsmodels-wheels/jobs/142252453
shows green for build success, but terminates in the test run with
"On entry to DGESV parameter number 4 had an illegal value"
which usually means there is a nan or inf in a linalg call.

Can anyone provide more information about this? I don't have a Mac.

Josef

>
> Josef

[Big Stone]

hi,

Any updated schedule for the rc2 or final release ? 

[josef...@gmail.com]

On Tue, Aug 16, 2016 at 1:38 PM, Big Stone <stone...@gmail.com> wrote:
> hi,
> Any updated schedule for the rc2 or final release ?

Most likely next week. I'll try to finish this week. It is planned as
final release, no rc2. (*)

I'm down to 3 PRs and 2 issues with the backport label that are still
open. (plus another check on other PRs that could maybe go in.)
I just wrote the list of PRs that need backporting
https://github.com/statsmodels/statsmodels/issues/3075


(I lost two weeks because I couldn't stop thinking, and reading, about
VAR, VECM and cointegration.)

(*) If anything shows up, then we can make a 0.8.1 release later. But
except for one PR related to pandas, I'm pretty sure everything is
fine.

Josef

[josef...@gmail.com]

I backported all merges from master to maintenance/0.8.x
master and maintenance 0.8 should be identical except for
version/release number.

A few small things are missing (*), and the doc build needs another check.
(*) including two PRs for statespace problems in edge/unusual cases.

I expect the release on monday or tuesday next week.

If possible, please check that we (or I) didn't mess up or miss something.


Josef

[Big Stone]

hum, any re-forecast ? 

[Tom Augspurger]

I'll be at PyData Chicago Thursday through Sunday. I won't be available immediately,

but I should be able to build and upload the documentation at some point.

I did a build over the weekend and the skim I gave it looked fine.

Tom