A problem with using GNU's ls -v is that the -v option isn't portable to BSD or OS X. In BSD there is no -v so you get an illegal option -- V. In OS X there is a -v option but instead of natural version sorting it Force(s) unedited printing of non-graphic characters.

I wonder if we could get away with falling back to normal ls? Although the difference in functionality would confuse OSX/Linux users.

I think that doing a check for proper ls -v support and then falling back to the current fileglob might be the only way to go. I've tried piping through sort to try to sort properly without using sort -V (also GNU only) and it is problematic. A custom sorting shell function would need to be written to cover all platforms and I'm not sure the issue warrants that kind of development time. I'll make another commit that includes a check for ls function and a fallback.

Can we get away with sort -t. -k 1,1n -k 2,2n -k 3,3n -k 4,4n and trust that 5 hierarchies of versioning won't happen here?

@aprescott Yeah, @mpapis suggested about the same and I think it may be viable. We do need to sort after the last slash because ~/.rubies matches the . field separator and borks the order. Worth exploring I think.

Yes, it is more complicated than using sort. Witness:

% \ls -d /tmp/rubies/* | sort -t. -k 1,1n -k 2,2n -k 3,3n -k 4,4n
/tmp/rubies/ruby-2.1.0
/tmp/rubies/rbx-2.2.7
/tmp/rubies/rbx-2.2.10
/tmp/rubies/ruby-1.9.1
/tmp/rubies/ruby-1.9.3-p152

Sorting the first field numerically doesn't order these example rubies correctly. We can change the first sort to numeric:

% \ls -d /tmp/rubies/* | sort -t. -k 1,1 -k 2,2n -k 3,3n -k 4,4n
/tmp/rubies/rbx-2.2.7
/tmp/rubies/rbx-2.2.10
/tmp/rubies/ruby-1.9.1
/tmp/rubies/ruby-1.9.3-p152
/tmp/rubies/ruby-2.1.0

Looks much better, right? Until a ruby goes double digits:

% \ls -d /tmp/rubies/* | sort -t. -k 1,1 -k 2,2n -k 3,3n -k 4,4n
/tmp/rubies/rbx-10.0.5
/tmp/rubies/rbx-2.2.7
/tmp/rubies/rbx-2.2.10
/tmp/rubies/ruby-1.9.1
/tmp/rubies/ruby-1.9.3-p152
/tmp/rubies/ruby-2.1.0

The sort needs to be by ruby type (rbi vs ruby, etc) and then sorted on just the isolated version. Without GNU coreutils or perl or something, this is difficult :)

$ ls -1d ~/.rvm/rubies/* | 
  sed 'h; s/-/./g ;; s/.*\///g ;; G ; s/\n/ /' | 
  LC_ALL=C sort -t. -k 1,1 -k 2,2n -k 3,3n -k 4,4n | 
  awk '{print $2}'
/home/mpapis/.rvm/rubies/current
/home/mpapis/.rvm/rubies/default
/home/mpapis/.rvm/rubies/rbx-1.3.3
/home/mpapis/.rvm/rubies/rbx-2.0.0
/home/mpapis/.rvm/rubies/rbx-10.0.0
/home/mpapis/.rvm/rubies/ruby-1.9.3-p545
/home/mpapis/.rvm/rubies/ruby-1.9.3-p547
/home/mpapis/.rvm/rubies/ruby-2.0.0-p481
/home/mpapis/.rvm/rubies/ruby-2.1.2

or

$ ls -1d ~/.rvm/rubies/* | 
  sed 's/.*\///g ;; h; s/-/./g ;; G ; s/\n/ /' | 
  LC_ALL=C sort -t. -k 1,1 -k 2,2n -k 3,3n -k 4,4n | 
  awk '{print $2}'
current
default
rbx-1.3.3
rbx-2.0.0
rbx-10.0.0
ruby-1.9.3-p545
ruby-1.9.3-p547
ruby-2.0.0-p481
ruby-2.1.2

the whole magic is in sed 'h; s/-/./ ;; s/.*\///g ;; G ; s/\n/ /':

• h - store current line in "hold space"
• s/-/./g - replace - with . for better sorting (g all occurrences)
• ;; - double required to fix a bug on OSX for undetermined results
• s/.*\///g - remove path from "pattern space"
• G - add the unchanged path from "hold space" to "pattern space"
• s/\n/ / - replace the new line that is between strings
• sed will automatically print "pattern space" after processing all commands

this works because:

$ ls -1d ~/.rvm/rubies/* | 
  sed -n 'h; s/-/./g ;; s/.*\///g ;; G ; s/\n/ / ;; p ' | 
  LC_ALL=C sort -t. -k 1,1 -k 2,2n -k 3,3n -k 4,4n
current /home/mpapis/.rvm/rubies/current
default /home/mpapis/.rvm/rubies/default
rbx.1.3.3 /home/mpapis/.rvm/rubies/rbx-1.3.3
rbx.2.0.0 /home/mpapis/.rvm/rubies/rbx-2.0.0
rbx.10.0.0 /home/mpapis/.rvm/rubies/rbx-10.0.0
ruby.1.9.3.p545 /home/mpapis/.rvm/rubies/ruby-1.9.3-p545
ruby.1.9.3.p547 /home/mpapis/.rvm/rubies/ruby-1.9.3-p547
ruby.2.0.0.p481 /home/mpapis/.rvm/rubies/ruby-2.0.0-p481
ruby.2.1.2 /home/mpapis/.rvm/rubies/ruby-2.1.2

and we use awk '{print $2}' to get the full path from second place (could be done with sed too for consistency).

Brilliant! Should have know there was some sed or awk magic that could save the day. I tried this on Linux, darwin, and FreeBSD with success.

sed magic sort now incorporated.

@grimm26 I also opened #278, which is an alternative variant of this that creates the sorted RUBIES array by piping find command output to the sed magic sort.

Is this or #278 sufficient to be merged?

closing in deference to #278 since more development has been done.