glob.glob unicode bug or feature

[Peter Otten]

Elbert Lev wrote:

> #Here is the script
> #Python 2.3 on W2K
> 
> import glob
> 
> name = glob.glob(u"./*.mp3")[0]
> print type(name)
> name = glob.glob(u"*.mp3")[0]
> print type(name)
> 
> ##OUTPUT##
> #<type 'unicode'>
> #<type 'str'>
> 
> #Is this a bug, or a feature? I beleve it's a bug.

The information whether glob() was passed a unicode or str is lost in the
following line in glob.glob1():

if not dirname: dirname = os.curdir

A quick fix would be

if not dirname:
    dirname = type(dirname)(os.curdir)

This will work if os.curdir consists entirely of characters in the ASCII
range, which I believe to be the case for all available platforms. Is there
a portable way to convert a str filename to unicode or should the os module
grow a unicode curdiru (uuugly name, but there already is os.getcwdu())?

Peter