liblinear bias/intercept handling

[paolo-losi]

As discussed on the mailing list

PS: pls tell me if the pull request is correct. My git fu is embarassing :-)

[fabianp]

I'm +1.

Let's hear Alexandre's opinion and we're ready to merge.

[agramfort]

+1 for merge

[fabianp]

Great, I'll do this today.

[fabianp]

It's pushed. Note that I've squashed it into a single commit:

a7047ca

so you'll probably have conflicts if you try to pull directly. You can solve this by resetting to the master branch (which will destroy changes in your current branch). Supposing origin is the main scikit-learn repo:

git fetch origin
git reset --hard origin/master

will do the work.

By the way, Paolo, ping me if you would like to continue working on the scikit-learn so that I can grant you write access.

Fabian.

[fabianp]

Oh, the wiki has cut the line. It's actually two separate commands:

git fetch origin

git reset --hard origin/master

[paolo-losi]

Well, this time (at the contrary of the first merge) I had no conflict at all.

On my pc I did

git fetch upstream
git merge upstream/master
git push

with no errors

A couple of questions as a git newbie.

1. I merged upstream/master on my master branch in order to test your changes to the binding.
   Was it correct? Was there a better way to do it (e.g. cherry picking, rebasing)?
   Is something to avoid in the general case and let "you" do the merge?

2. How did you squashed or the commits into one? Which command did you use?

I will keep working on scikit-learn for sure. But before asking for write access let me
familiarize with git :-)

[fabianp]

I don't really understand how you managed to merge without conflicts, but all seems fine :-)

The commits where squashed with git rebase -i (look for "git interactive rebase" on the web). However, rebasing is a somewhat advanced concept and is not needed to be able to do regular work with git (in fact most commiters use little beyond commit/push/pull).

Best,

Fabian