Provide permissions, owner, and group for files on UNIX platforms about pyftpdlib HOT 21 CLOSED

Checked in a preliminary version of this into my branch in SVN and created a testbed
script called ls.py to play with this. Tested briefly on Linux and OS X platforms.
Will review and test on Windows platform also and check into main branch once
everything looks good.

- rename "stat" variable to "stat_result"
- replace stat[int] with stat.value
- add number of links to inode as well instead of just using '1'

Turns out tarfile.filemode() returns sane values for Windows too so we don't need to
set default values for Win platforms. removed unneeded code and use UID/GID for
Windows platforms where pwd/grp modules are not available

Fixed typo, remove UID/GID and returned to default values since Windows always
returns 0 for those stat values. Added 10 space padding to the file size again to
restore some type of formatting also.

Still need to iron out a better way to do the formatting like ls does it but
otherwise should be most of the way there.

- Moved tarfile, pwd and grp module imports outside format_list function.
- (posix) Used specific KeyError exception for catching pwd.getpwuid() and
grp.getgrgid() exceptions.
- (posix) If can't get user and group names use raw uid/gid instead.
- Where posix security model is nonsense use bogus "owner" and "group" values instead.
- (non-posix) Have nlinks = '1' if can't have a reliable number of links to the inode.
- Removed dir_perms variable which was useless.
- Added other space paddings to restore the output formatting as it was before as
long as we don't get a better way to emulate ls formatting.
- Can now use an optional cmd line argument to specify the directory to list.

Updated version incorporates some of the changes billiejoex submitted, but also adds
support for formatting the output in the same manner as ls. We will need to discuss
if the overhead is worth bothering with this and do some benchmarking on Linux,
Windows, and OS X systems to see what the actual performance is in real-life testing.

This version uses a similar method to the coreutils ls C code, storing meta
information about a directory listing in a class and then creating the output using
the meta info.

A new problem came out. Please take a look at Issue 49 .

Returned to fixed formatting after discussion, also fixed Issue 49 and a few other
bugs relating to the arguments passed to ls.py

- updated comments and docstrings
- removed unnecessary str() call 
- re-ordered code blocks for formatting order

- added parens around except parameters

- comment / docstring width to 79 chars
- added note about permissions on Windows platforms

- used os.readlink(file) instead of os.path.basename(os.path.realpath(file)) to get
the real path to which the symbolic link points.

As far as I can tell, it should be definitively ok for inclusion now.

Implemented in SVN revision 159

format_list doesn't format symlinks correctly.


The current implementation of format_list in AbstractedFS class always follows 
symlinks.
This should NEVER be done. The linux ls command never does that.

For that reason os.lstat should be used instead of os.stat.

Using os.stat, the l flag in filemode rights is never displayed.
The link name will never be resolved as well [lines 1059-1060].

The solution is to replace [lines 1020-1023]:
            try:
                stat_result = os.stat(file)
            except (OSError,AttributeError):
                stat_result = os.lstat(file)

with a simple:
            stat_result = os.lstat(file)

Agreed.
My only concern is if it is available on all platforms.
As far as I know where symlinks are not supported os.lstat should be an alias for
os.stat but I'm not 100% sure.

I decided to use hasattr() to make sure that os.lstat is available,
otherwise os.stat() will be used in replacement.
I also noticed that some module in the stdlib do the same.
This shouldn't be really necessary but, who knows...?
...As we say here in Italy: "'sure' is dead". ;)

Change applied in revision #179 .

Fixed in version 0.3.0.