Fix SQLAlchemy enum generation

[sheinbergon]

Fixes #361

[coveralls]

coverage: 96.957% (-0.07%) from 97.025%
when pulling 84b9ecb on sheinbergon:master
into 82251f3 on agronholm:master.

[agronholm]

This needs a regression test and a changelog entry.

[sheinbergon]

This needs a regression test and a changelog entry.

OK, I'll add the changes to the Changes.rst

As for a regression test - there's already a test that generates an enum (an intiial trial attempt caused it to fail).
However, like all of your tests, it doesn't really generate code from a live db, but rather from an SQLalchemy definition set,

so it was passing all along, while differing from real world behavior which caused a differen enum statement to be generated. How would you like to tackle that?

I can see if maybe a specific postgresql.ENUM dialect would trigger the error

[agronholm]

This needs a regression test and a changelog entry.

OK, I'll add the changes to the Changes.rst

As for a regression test - there's already a test that generates an enum (an intiial trial attempt caused it to fail). However, like all of your tests, it doesn't really generate code from a live db, but rather from an SQLalchemy definition set,

so it was passing all along, while differing from real world behavior which caused a differen enum statement to be generated. How would you like to tackle that?

I can see if maybe a specific postgresql.ENUM dialect would trigger the error

Yes, that would probably do the trick. IIRC I've used this technique elsewhere in the test suite when the generic types weren't reproducing the issues I wanted to fix.