Thanks LGTM. +1 if travis is green.

If you squash them they will be easier to cherry pick.

Now the four commits have been squashed into one.

thanks :)

Repository health decreased by 0.04% when pulling 956f88a on bryandeng:immutable_defaults into c4df19d on scikit-learn:master.

• 56 new problems were found (including 0 errors and 15 code smells).
• 8 problems were fixed (including 5 errors and 1 code smell).

Repository health decreased by 0.04% when pulling 956f88a on bryandeng:immutable_defaults into c4df19d on scikit-learn:master.

• 56 new problems were found (including 0 errors and 15 code smells).
• 8 problems were fixed (including 5 errors and 1 code smell).

@carlio again, not sure what is happening here :-/

Travis failed.

Travis shows that problems are in my modified rfe.py. It's not compatible with set_params().

TypeError: set_params() argument after ** must be a mapping, not NoneType

I think I can't simply modify init that way.
I'll do local tests before the next commit.

Thanks, amueller!

Don't mutate the default argument in __init__. That won't work. You need to do it in fit. You don't need to run local tests, you should have to watch travis. Running tests locally is usually faster, though.

Repository health decreased by 0.07% when pulling bf41ad5 on bryandeng:immutable_defaults into 81acb94 on scikit-learn:master.

• 56 new problems were found (including 0 errors and 15 code smells).
• 6 problems were fixed (including 3 errors and 1 code smell).

Repository health decreased by 0.07% when pulling bf41ad5 on bryandeng:immutable_defaults into 81acb94 on scikit-learn:master.

• 56 new problems were found (including 0 errors and 15 code smells).
• 6 problems were fixed (including 3 errors and 1 code smell).

@amueller Can you restart the build later on?
I got strange error information about cross validation. But I didn't touch that part of code.
And #4368 shows that @xuewei4d was working on rfe.py so I reverted it to avoid merge conflict.

I built my fork repo on Travis and it passed.

master was broken, sorry, my fault :-/

I just restarted, should be fine now.

Also, don't worry much about merge conflicts. We have 200 open PRs. It will be impossible not to conflict with any ;)

Do I need to add some tests/documentation? Maybe I can help on #4379.

#4379 will be mergable once everything is fixed.

I don't think additional tests are needed. It looks good to me.

OK. I'll now squash the commits into one for easier cherry-picking and start to work on other issues.

thanks :)

Repository health decreased by 0.07% when pulling 2638fe5 on bryandeng:immutable_defaults into bc5acea on scikit-learn:master.

• 56 new problems were found (including 0 errors and 15 code smells).
• 6 problems were fixed (including 3 errors and 1 code smell).

Coverage increased (+0.0%) to 95.09% when pulling 2638fe5 on bryandeng:immutable_defaults into bc5acea on scikit-learn:master.

LGTM. Thanks @bryandeng!