Four tests for inspectdb fail under Oracle

[Aymeric Augustin]

​http://ci.djangoproject.com/job/Django%20Oracle/88/database=oracle,python=python2.7/testReport/inspectdb.tests/

Roughly, the failure happens because the module enables unicode_literals, but Oracle returns table names as bytestrings.

[Shai Berger]

Beh. cx_Oracle returns a str as the column name, whether we are on python2 or python3. Arguably a cx_Oracle bug.

​https://github.com/django/django/pull/1075 fixes (or, if you like, works around it).

[Aymeric Augustin]

Thanks for finding the right spot for fixing this.

I'd write the patch a bit differently:

from django.utils.encoding import force_text

...

name = force_text(desc[0])

[Shai Berger]

I'm not sure it's an improvement -- force_text deals with all sorts of non-strings and lazy objects, and seems a bit overkill for this case where we know exactly what the object is. However, I won't argue too much about it.

[Aymeric Augustin]

You're right, I've considered introducing another API just to cover the utf-8 str / unicode case.

But we're already using force_text throughout the code base and it does the job just fine. I think the conditional branches for str / unicode are early in the function so it's fast.

[Shai Berger]

As I said, I don't feel very attached to the three lines. force_text will definitely do the job.

[Shai Berger <shai.berger@…>]

In 215647c0f7614abe4fe0cd76bc70dd7b02f829b4:

Fixed #20386 - Introspection problem on Oracle

Made introspection always return a unicode as column name on Oracle.
Thanks aaugustin for review and suggestion to use force_text().