From: Daniel Berger <Daniel.Berger@...>
Date: 2005-09-14T03:31:14+09:00
Subject: Re: File.split edge cases

Austin Ziegler wrote:
> On 9/13/05, Daniel Berger <Daniel.Berger@qwest.com> wrote:
> 
>>Yukihiro Matsumoto wrote:
>>
>>>Hi,
>>>In message "Re: File.split edge cases"
>>>    on Wed, 14 Sep 2005 02:11:28 +0900, Daniel Berger
>>>    <Daniel.Berger@qwest.com> writes:
>>>| I would argue that this should always be true:
>>>|
>>>| path == File.split(File.join(path))
>>>This sounds good, but it's kinda hard to define exact behavior for
>>>related methods. I will examine posixpath.split in Python.
>>
>>Another interesting difference is the way Python interprets a trailing
>>slash:
> 
> 
> Assuming that File.split should be implemented as [ File.dirname,
> File.basename ]:
> 
>   # Linux RedHat 9 2.4.20-8
>   # coreutils 4.5.3
>   % dirname /foo/bar/
>   /foo
>   % basename /foo/bar/
>   bar
>   % dirname /
>   /
>   % basename /
>   /
> 
>   # SunOS 5.8 Generic_117350-15
>   % dirname /foo/bar/
>   /foo
>   % basename /foo/bar/
>   bar
>   % dirname /
>   /
>   % basename /
>   /

Yeah, the closer I look at posixpath, the more I realize it's....wrong. 
  Well, damn, with a name like "posixpath" you would *think* it would be 
correct, eh?

> Ruby has it right in both cases. On Windows, File.split() should be a
> bit smarter:
> 
>   File.split("C:\\")                # => [ "C:", "\\" ]
>   File.split("\\\\server\\path\\")  # => [ "\\\\server\\path", "\\" ]
> 
> -austin

Yes, I think really all that's needed are explicit checks for empty and 
root paths:

# pseudo-code
if path.root? or path.empty?
    return path, ""
else
    return File.dirname(path), File.basename(path)
end

Regards,

Dan