Re: Proposal : Path

On Fri, 14 Nov 2003, Nicolas Cannasse wrote:

> Hi list,
> I came up very recently with a typical problem.
> I had a file A and its relative path, I had a file B and its absolute path,
> I wanted to get the file B path relative to A.... not so easy actually. I
> wrote this small module Path :
> 
> Path.make "my/../path/./" build a relative path
> Path.absolute give the absolute path
> Path.relative a b give the path of a relative to b
> Path.up goes up one directory
> 
> I did some tests under windows , it looks like it works well. If people can
> test it under *nix...
> 
> Note : this does not actually do any system checking ( the path may not
> exists or contains invalid characters ). Only special paths . .. and / ( and
> windows drive letters ) are recognized as special.

Hi,

Such a module seems very important to me too.
I've got some suggestions too.

It seems important to me to be able to choose input/output 
style: Unix, Windows, MacOS, and others. Maybe as an option 
passed to functions "make" and "to_string".

type style = Unix | Windows | Cygwin | ...

let default_style = 
  match Sys.os_type with
    "Win32" -> Windows
  | "Unix" -> Unix
  | ...

let of_string ?(style = default_style) s = ...

let to_string ?(style = default_style) x = ...



This is less clear to me, but I feel that it would be convenient to split 
the type t into a root part and a virtual path part:

type root = Windows of string | Unix | Current_dir | Path of t

and t = { root : root;
          virtual_path : string list }

... and implement functions that change either the root or the 
virtual path.



Some problems concerning the current implementation under Linux:
1. //tmp///toto is valid and equivalent to /tmp/toto (because any filename 
must have at least one character)
2. /../../.. is equivalent to / (which is dangerous anyway)
3. toto/ is more restrictive than toto (trailing slash indicates that toto 
must be a directory), so don't output trailing slashes except for the root 
directory, or introduce a distinction between files and directories.

For solving points 1 and 2, maybe it would be nice to introduce 2 input 
styles: Unix and Abstract, where Unix conforms to Unix (Linux?) standards 
and Abstract is something more abstract, that makes "/.." invalid and 
allows 0-length filenames (would it be useful?).


Martin


-------------------------------------------------------
This SF.Net email sponsored by: ApacheCon 2003,
16-19 November in Las Vegas. Learn firsthand the latest
developments in Apache, PHP, Perl, XML, Java, MySQL,
WebDAV, and more! http://www.apachecon.com/