Add support for unicode literals to

[pytestbot]

Originally reported by: Jason R. Coombs (BitBucket: jaraco, GitHub: jaraco)

---

I like to use doctests for some use cases, and pytest makes inclusion of them in tests suites quite elegant.

However, doctests have a particularly hard problem with Python2/3 compatibility (because unicode strings render as u'' in Python 2 and '' in Python 3).

The NLTK project has built a nose plugin to better support the disparity.

I'd like to see pytest support this functionality as well.

---

• Bitbucket: https://bitbucket.org/pytest-dev/pytest/issue/710

[pytestbot]

Original comment by holger krekel (BitBucket: hpk42, GitHub: hpk42):

---

If i am not mistaken, most current core committers do not use doctests much so you might consider adding necessary support yourself. Or maybe @charles or @HolgerPeters who did some relatively recent PR to the pytest doctest code can help.

[pytestbot]

Original comment by Jason R. Coombs (BitBucket: jaraco, GitHub: jaraco):

---

I hear that this project exists as a standalone package supporting the behavior. Perhaps it could be extended to be a pytest plugin.

[pytestbot]

Original comment by Jason R. Coombs (BitBucket: jaraco, GitHub: jaraco):

---

I'm willing to offer a bounty for this feature. Would that pique any interest?

[RonnyPfannschmidt]

@jaraco this one got lost under the rug, is it still needed, it seems easy enough to hand to a new contributor at first glance (i hope im right at least)

[jaraco]

It is still needed. I regularly have to massage my doctests to satisfy Python 2 or just let them fail on Python 2. This one issue creates a great deal of impedance for doctest authors, and if it could be addressed in the test runner, that would be such a delight.

[dimazest]

I'm using pytest to execute NLTK tests.
I'm I right, that the core part is the _UnicodeOutputChecker (see NLTK nose plugin referenced earlier). What would be the best way to replace the default one with this?

[RonnyPfannschmidt]

as far as i can tell, all core developers would have to investigate this first

[dimazest]

It seems so. Just rising the awareness about this issue!

[nicoddemus]

Opened a PR at #938.

The use case I used was:

>>> b'12'.decode('ascii')
'12'

You guys feel that is enough?

[jaraco]

@nicoddemus So awesome you're working this! I believe the use case should be sufficient.

[RonnyPfannschmidt]

im under the impression the solution is to the spot