Updated date and unicode handling code.

[obi1kenobi]

Partially addresses #129, in that it shows how to avoid the error (see the updated test case) but does not provide a comprehensive solution.

[obi1kenobi]

The CI build failed on a transient connection error, not my fault this time :)

[obi1kenobi]

To be clear, here's what this PR suggests as a partial solution.

The binary protocol decoder, on python 2, implicitly decodes received strings to UTF-8. This breaks the "what you put in is what you get out" expectation, because a user could put in a unicode object but get it back as a UTF-8 encoded str. The proposed (clunky but working) fix, for python 2, is that whenever the user expects a particular string object to be of unicode type rather than str, they should transform it using UTF-8 manually upon receiving it back from pyorient. This is exactly how the updated Unicode test works.

[obi1kenobi]

Here's an alternative which is slightly more drastic, but is perhaps a cleaner solution from a technical standpoint -- and I'd love to hear your feedback on this.

All strings that pyorient receives are considered to be of type unicode. Any str objects are always implicitly converted to unicode using UTF-8. pyorient always returns unicode objects for fields of type String. This would be consistent across python 2 and 3, and not require that the programmer using the library remember to cast to unicode upon receiving the data back.

[TropicalPenguin]

I like the maintainability this would add, abstracting away differences between python 2 and 3 usage.

[obi1kenobi]

The additional upside by moving to all-unicode when using pyorient is that it would let pyorient only use unicode on the inside, thereby fixing a few more problems I discovered. If you'd like, I can prepare a PR that converts all string literals that pyorient uses to unicode. I feel like that would make it much easier to reason about correctness and much less likely to accidentally have different behavior on python 2 vs 3.

[obi1kenobi]

Now also fixes #146, and subsumes #141.

[obi1kenobi]

Moving all of pyorient to use unicode internally proved to be difficult, as I don't understand much of the code. I decided to leave that to you @TropicalPenguin and @Ostico.

This PR is consistent with that goal, but still requires python 2 users to manually convert any received string data to unicode if it was passed in as unicode. Please review and let me know what you think!

[obi1kenobi]

@mogui @Ostico thoughts on merging this soon?

[obi1kenobi]

Sorry to keep poking at this, but I'd really appreciate it if we could fix this set of bugs before the holidays start -- they are major concerns on the project I'm working on.

[mogui]

sorry @obi1kenobi for the delay :P

[obi1kenobi]

No worries! Thank you for getting to it :)