From: Brian Candler <b.candler@...>
Date: 2008-09-29T21:28:09+09:00
Subject: Re: [working directory to be changed (in terminal) by the Ru

Max Dev wrote:
> I created a script. At some point I should change from whatever
> directory I'm currently in (in Terminal) to a certain directory from the
> ruby script.
> 
> So far I find other solutions that do not satisfy me. The one with which
> I managed to change directories in the terminal needs another .sh file.
> 
> When launching the .sh script from the terminal, actually changes the
> directory (pwd). But if I launch from the ruby script, does not change
> the output terminal (pwd).
> 
> Then, I want the working directory to be changed (in terminal) by the
> Ruby program after it exits.
> 
> Any suggestions?

If this is under Unix, then basically it can't happen.

When you start a ruby interpreter, you are spawning (fork+exec) a new 
process. This new process has its own environment, including its own 
concept of "current directory". Any change to current directory in this 
new process does not, indeed cannot, affect the current directory in the 
parent.

This isn't a Ruby limitation, but of any spawned process, including 
another shell. See what happens when you do this:

  cd /tmp
  pwd
  sh                    # start a new shell
  cd /usr
  pwd
  exit                  # exit the new shell, returning control to 
parent
  pwd                   # parent working directory unchanged

The 'cd' command only works because it is a shell builtin, not an 
external command like /bin/cd.

Another test: create a file "chdir.sh" containing the following:

  #!/bin/sh
  cd /tmp

Make it executable (chmod +x chdir.sh) and run it (./chdir.sh). Nothing 
happens; the directory change was in the subshell.

However if you run it as ". chdir.sh" (note the space after the dot) it 
is read as a series of instructions into the current shell. The first 
line is ignored as a comment, and the second is run as if you had typed 
it into the current shell, so it invokes the cd builtin as before.

I'm not sure if you can get a shell to accept commands from its child on 
a pipe. I tried

  . <(ruby -e 'puts "cd /tmp"')

but that doesn't seem to work, even though

  cat <(ruby -e 'puts "cd /tmp"')

works as expected.
-- 
Posted via http://www.ruby-forum.com/.