Postgresql/psycopg2 does not like str enum values

Storm uses byte strings to represent binary data and unicode strings to represent text. Enumeration labels are text, so if you are using byte strings to represent them in your code that sounds like a misuse of the API.

Unless I'm misunderstanding what you're trying to do, this bug is probably invalid.

I'm assuming that if I'm allowed to define

class E(Storm):
   x = Enum(map={'A':'A'})

then it follows that I'm also allowed to assign

   E().x = 'A'

and expect the flush to work. If that's not the case, Enum should be coercing the labels to Unicode, and EnumVariable should be checking the values; like, e.g.,. UnicodeVariable does.

You should be using a unicode string for the values in the map dictionary. Something like:

    x = Enum(map={'A': u'A'})

As I said before, the values you want sent to the database are text rather than binary data. Therefore, you need to use a unicode string.

Looks like your suggestion works, up to a point. In the course of verifying it, I ran into another bug that indicates that the EnumVariable implementation is flakey in other ways. Unless I'm completely misunderstanding the intent of _get_map and _set_map, of course. See the attached updated test case, and I'll submit a separate bug report about that.

Meh, belay that ... silly bug in the test script. Here's a better one.

This example highlights another bug related to enums and unicode, this time apparently in the expression compiler. I'm at a loss to understand how I can get this error:

ProgrammingError: operator does not exist: enumtype = bytea
LINE 1: DELETE FROM stormenum WHERE stormenum.ident = E'\\x42'::byte...

if both the keys and the values in the map are unicode.