beeprint interacts badly with wrapped sys.stdout

[jhermann]

Concrete example is bpython, this is what happens:

$ bpython
bpython version 0.16 on top of Python 2.7.11+
>>> import sys
>>> sys.stdout
<bpython.curtsiesfrontend.coderunner.FakeOutput object at 0x7f6a156c0b50>
>>> import beeprint
>>> sys.stdout<open file '<stdout>', mode 'w' at 0x7f6a1a2a6150>

I don't know why sys.stdout is changed at all by a simple import (it shouldn't), but if you do, chain the existing value.

[jhermann]

And I found the culprit:

beeprint/beeprint/printer.py

Lines 15 to 21 in 2debfda

	if pyv == 2:
	# avoid throw [UnicodeEncodeError: 'ascii' codec can't encode characters]
	# exceptions, without these lines, the sys.getdefaultencoding() returns ascii
	from imp import reload
	reload(sys)
	sys.setdefaultencoding('utf-8')

What you do there is just horrible.

[gimbo]

I've just come across this issue too (with jupyter not bpython, but I guess it's the same thing); python 2.7.13 and beeprint 2.4.7.