Support directories with unicode chars

[canepan]

Hi, I stumbled upon this issue while using duviz on my disk with an accentuated letter ("é").
The error was:

   File "/usr/local/bin/duviz.py", line 72, in from_path_size_pairs
    for path, size in pairs:
  File "/usr/local/bin/duviz.py", line 130, in pairs
    progress_report(path)
  File "/usr/local/bin/duviz.py", line 453, in progress
    write(info.ljust(terminal_width)[:terminal_width] + '\r')
UnicodeEncodeError: 'ascii' codec can't encode character '\u2014' in position 138: ordinal not in range(128)

and according to this article it can be resolved by adding (the one which worked for me):

    os.environ["PYTHONIOENCODING"] = "utf-8"

at the beginning of main()

Nicola

[soxofaan]

I don't completely understand: the exception talks about character '\u2014' which is "EM DASH", not accented "é"

I'm also not sure about setting os.environ["PYTHONIOENCODING"] from within the script. It doesn't seem good style to blindly set an env variable in a script, I think this should be set properly in the shell. Moreover, if I understand the docs https://docs.python.org/3/using/cmdline.html#envvar-PYTHONIOENCODING correctly, setting it inside the script (so not "before running the interpreter") will not be effective in the first place.

[canepan]

It has been effective for me, but it may change with newer versions.
To stay on the safe side, but still avoid failing, I see 2 options:

• wrap the duviz.py invocation in a shell script to be added to the repo (and a .cmd if needed)
• do a try/except and possibly (since it's just dealing with the transient progress display) ignore the line