Message9282

The problem with returning encoded strings, vs unicode, is that we cannot simply use the results of Python os operations, including ones that build on it, like glob, with Java IO, since they do not know about encoded strings.

For once, this may be one advantage of the conflation of strings and Unicode strings in Python 2.x. With Jython 3.x, there will also be a seamless transition, since it will be unicode anyway unless bytes are asked for specifically.

The other advantage is that this is then an easy fix, per the diff in my earlier message. Getting similar Java IO seamlessness with returning encoded strings from os.listdir is involved (of unknown scope), and doesn't correspond to anything else we have done. But Jython has always tried to make for a seamless experience between Python and Java; is it why people use Jython.

Boundaries always present challenges like these. Providing unicode simplifies what we need to do.