[ruby-talk:23693] Re: Path class (was: Re: FileSpec vs Dir/File class methods)

On Sun, 28 Oct 2001 09:16:48 +0900, Kevin Smith <[e-mail]>
wrote:

> > The Path class inherits from String and implements:
> 
> I'm curious why you chose to extend String, rather than
> just contain a String. Do you really want/expect to be able
> to perform every String operation on a Path object? 

Not anymore. The next version won't inherits from String and, as
suggested by Matz, implement #to_str. It will also provide some String
methods (like +, ==, empty?, gsub[!], ...)

> What if you're on a Mac, where (as I probably misunderstand
> it), it's difficult to represent an entire pathspec in a
> single string?
> 

I don't know anything about Mac. How do their paths differ?

> >    - #/(other)
> >        like #join(other)
> 
> Not sure if I like this or if it scares me (or both).
 
I personally like it. There is mainly two type of concatenation of
paths: adding a string at the end of the last entry, and adding a
subname. I don't think I use the first one more, so an operator for the
second one sounds good for me. The '/' operator is never used except in
Numerics and I don't think someone will come with a nice generic '/'
behavior that might apply to many classes and make Path's one
disturbing.
(ROOT / 'bin' / binary) looks better than ("#{ROOT}/bin/#{binary}") or
(ROOT.join 'bin', binary) for me.

> > A new method String#path converts a String to a Path.
> 
> I would rather just do Path.new(string) than clutter up
> String itself.

I'm going to remove String#path. Path[str] will also be availiable.

> Perhaps you could have:
> #delete (deletes files, no effect for directories)
> #rmdir (deletes directories, no effect on files)
> #delete! (deletes file OR directory)
> 
> I can't think of a better name than #delete! but there
> probably is one.

#delete! makes me think that it will try everything availiable to ensure
the path is removed. It sound good, and might even replace #rmtree.

> > - Should #join (and #/) remove duplicate slashes?
> >   ie: '/bar/'.path.join('foo') => '/bar//foo' or
> > '/bar/foo'?
> >   File::join does not.
> 
> Wow. I would have thought that would be the main reason
> join would exist. Yes, I think it should remove duplicate
> slashes (except if it handles things like
> "file://c:/config.sys" or "\\MyWinBox\MyShare\file"

So maybe should it only check for slashes at the end of the path and at
the beginning of the subpath.
For example
  'a'  / 'b',
  'a/' / 'b',
  'a'  / '/b' and
  'a/' / '/b'
would all return 'a/b', but '//a/' / 'b' would give '//a/b'

As I wrote these replies, I worked on version 0.2.0. It's availiable
there:
http://www.lepton.fr/ruby/path

As soon as I write some documentation, I post it on the RAA.

Mike.