From: James Gray <james@...>
Date: 2008-03-15T04:58:36+09:00
Subject: Re: How to separate stdout & stderr from executed shell comm

On Mar 14, 2008, at 2:49 PM, jqwoods@gmail.com wrote:

> On Mar 14, 12:16 pm, Rodrigo Bermejo <rodrigo.berm...@ps.ge.com>
> wrote:
>> unknown wrote:
>>> On Mar 14, 11:40 am, jqwo...@gmail.com wrote:
>>
>>>>  irb> my_exec_command("ls NON_existant_file")
>>>>  => ["", "ls: NON_existant_file: No such file or directory\n", 2]
>>
>>>> Any pointers on how to implement my_exec_command?
>>
>>> PS: Unless there's a simpler way, if someone could just point me  
>>> to an
>>> API that provides stdout and stderr of sub-processes as IO  
>>> objects, I
>>> could probably figure it out from there. Thanks!
>>
>> Take a look at the IO class.
>> The Pickaxe book has some good examples.
>>
>> -r.
>> --
>> Posted viahttp://www.ruby-forum.com/.
>
> The closest API I could find is IO::popen. But as the doc for
> IO::popen says "...the subprocess's standard input and output will be
> connected..." So far I haven't found any way to get stdout and stderr
> of a sub-process *separately*.
>
> Any other pointers?

Yes, Ruby comes with a standard library for this, called open3.  Here  
are the docs for the method you want:

$ qri Open3#popen3
----------------------------------------------------------- Open3#popen3
      popen3(*cmd) {|| ...}
------------------------------------------------------------------------
      Open stdin, stdout, and stderr streams and start external
      executable. Non-block form:

        require 'open3'

        [stdin, stdout, stderr] = Open3.popen3(cmd)

      Block form:

        require 'open3'

        Open3.popen3(cmd) { |stdin, stdout, stderr| ... }

      The parameter cmd is passed directly to Kernel#exec.

Hope that helps.

James Edward Gray II