Pushing Results from Exec into an Array

[bivouac]

Anyone know how to push something from 'exec' into an array?

user="foo"
 host="bar@barfoo"
 r_dir="/home/foo/test"
 
 files = exec "ssh #{user}@#{host} ls #{r_dir}"

This prints to STDOUT the contents of 'r_dir' - I'd like to put this into a array so I could iterate over the actual files that are in 'r_dir.'

Do I have to use the I/O module?

[Luke Redpath]

I'm not sure how to capture the results of exec, however there is probably a better way of sshing into a server and geting a list of a directory and I'd imagine you'll find net/ssh is the answer. Look here:

http://net-ssh.rubyforge.org/chapter-4.html

[stereofrog]

Unlike php, exec doesn't return. Use `` (backquotes) to run a subshell and capture output.

[bivouac]

I played around with Net-SSH last week, I think it's a bit too much with what I want to do. Basically, I just need to log in to some remote machines and scp the files to a local machine.

files = `ssh #{user}@#{host} ls {r_dir}

So that line pushes what I need into something I can iterate over.

files.each { | file |
       # remove the trailing new line
       file.chomp!
     
       # issue scp commands
       `scp #{user}@#{host}:#{r_dir}/#{file} #{l_dir}`
     }

I'm have another quandry now though. How in the heck do I capture the output from that shell command to make sure the SCP actually went through? I saw this link about IO and was thinking that this would be the way to control for that - not sure though.

[Luke Redpath]

How in the heck do I capture the output from that shell command to make sure the SCP actually went through?

Again, I think you are pushing against what you can do with the backtick operator/exec. Perhaps you should revisit net/ssh and if you can accomplish what you want that way - at least then you have the advantage of a full API and the ability to capture errors/exceptions as usual.

[bivouac]

Luke, the more I think about it the more I think you're right.

Double drat.

Was trying to accomplish this in an easy way and Net::SSH with it's channels and sessions seems confusing to me. I guess I'll need to master the module and see how it can fit into my script. But you're right - I can't reasonably tell with the backtick what's happening. With the Net::SSH module, I can monitor what's going on and then react appropriately.

[stereofrog]

IHow in the heck do I capture the output from that shell command to make sure the SCP actually went through?

You should be able to check $? (command return code)

out = `command`
if $? != 0 
   something went wrong

[ezmobius1]

There is another easier way to drive shell commands from ruby and capture sdout, sterr and send stdin. My favorite is a gem called session:

$ sudo gem install session

Then its really easy to drive a shell in a subprocess. here's an example that will make a bash shell in ruby with prompts and everything:

#!/usr/bin/env ruby
require 'tempfile'
require 'readline'
include Readline

require 'rubygems'
require 'session'

shell = Session::Shell.new

require 'tempfile'
require 'readline'
include Readline

n = 0
loop do
  command = readline("#{ n }: SHELL > ", true) 
  if command =~ /^\s*\!*history\!*\s*$/
    open('shell.history','w'){|f| f.puts shell.history}
    next
  end
  exit if command =~ /^\s*(?:exit|quit)\s*$/io

  out, err = shell.execute command
  out ||= ''
  err ||= ''
  printf "STDOUT:\n%s\nSTDERR:\n%s\n", out.gsub(%r/^/,"\t"), err.gsub(%r/^/,"\t")
  n += 1
end

Here is a little simpler example of just sending an ls command and capturing stout and sterr:

require 'tempfile'
require 'rubygems'
require 'session'

shell = Session::Shell.new :history => false

shell.execute('ls -ltar') do |out, err|
  if out
    puts "OUT:\n#{ out }"
  elsif err
    puts "ERR:\n#{ err }"
  end
end
puts shell.history

Hopefully that help some. But Net::SSH is a great tool to drive ssh commands. And Net::SFTP might be worth l;ooking at too if you are trying to copy files from a remote server securely.

Cheers

-Ezra