Install Error: "SyntaxError: invalid character in identifier" about imbalanced-learn

[glemaitre]

mmm that's a good one. Could you open the file setup.py and check if you have some unrecognized characters.

Check also which encoding are you using:

http://stackoverflow.com/questions/21236910/using-exec-function-to-run-script-in-the-interpreter-shell

[ErnstDinkelmann]

Interesting.
OK, I forced the encoding as follows, just in the open (as suggested in the link):

with open(os.path.join('imblearn', 'version.py'), encoding='utf-8-sig') as fp:
        exec(fp.read(), globals_dict)

and voila, it works.

Weirdly though, in my PyCharm IDE, in the right-bottom corner, it says "UTF-8", which I'm assuming is the encoding for the context, but it seems the Win current encoding was still somehow overriding that.

I also used a "chchp" command in Win cmd, and it displayed "437", which is apparently a "US encoding"

I don't know a lot, but the above might give you insight to change the code to avoid this bug on other Win systems.

Cheers

[glemaitre]

I got the same bug with Appveyor so I will investigate a bit more.
Thanks for the solution.

[glemaitre]

So actually there is a need for

# *- encoding: utf-8 -*-

in version.py to enforce the encoding. open function do not allow encoding argument in Python 2.
This is a Windows thing probably, I don't know to much about these stuff.

[ErnstDinkelmann]

I tested that for you and it made no difference, exact same problem.
But found this:

[http://stackoverflow.com/questions/10971033/backporting-python-3-openencoding-utf-8-to-python-2]

So, the following worked and apparently will do on Python 2 as well:

import codecs
with codecs.open(os.path.join('imblearn', 'version.py'), encoding='utf-8-sig') as fp:
        exec(fp.read(), globals_dict)

[glemaitre]

thanks