When FTPHost.walk is used to examine a filesystem tree which somewhere
contains a non-7-bit-ASCII character in a name and the argument passed
in is any unicode string, the walk method implicitly will raise a
UnicodeEncodeError.
Imagine this directory structure:
some_dir
some_stränge_file
then
ftp_host.walk(u"some_dir")
will cause a UnicodeEncodeError in posixpath like:
File "/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/posixpath.py", line 70, in join
path += '/' + b
UnicodeDecodeError: 'ascii' codec can't decode byte 0xdf in position 41: ordinal not in range(128)
(original report by Henning Hraban Ramm - thanks!)
As pointed out in this mail, FTP has no concept of encodings. As the encoding of the directories and files on the remote side is unknown, there's no convenient solution.
At the moment, I think the most appropriate approach is to have a method fail as early as possible if it accepts remote paths and gets a unicode string for them.
A solution might be something like:
def walk(self, root): # If `root` isn't ASCII, fail now instead of later. # Otherwise, continue with a byte string. root = str(root) ...
I put this off until ftputil 2.5.1. I'll have to go through all the methods and see how they're affected, and I don't want to delay the release of ftputil 2.5 final even more after the apparently long beta phase.
Fixed in commits 927259c707a43adfcf4bf45b9e5939b3d83db40d and 8f6ce57b858f888c1279ca1310e663773b2200c7.