From: Gennady <bystr@...>
Date: 2003-05-28T08:10:41+09:00
Subject: Re: system calls

Recently I implemented a facility for such cases and if you are 
interested I can send it to you (I am not sure it is worth enough to be 
sent to RAA). It does not spawn threads, all is done with IO#select.

I called it "launch" and made it available through the Kernel. So in 
your case it can be used as:

require 'launch'
messages = launch "date"

It returns an array of message arrays, the first element -- messages 
read from the command's stdout, the second -- from the stderr.
If the command is killed or returns with error (return code is not 0), 
exception TerminateError is raised. If you rescue it, you can get the 
reason for termination (signal or non-zero return) and the return code 
(signal number if killed). Stderr messages are also available from the 
exception class.

In general, it was designed for launching a pipeline of commands with 
access to all error messages and return codes from all commands in the 
pipeline. Unfortunately, when you do it with system("command1 | 
command2 |command3"), you may not even know that command1 or  command2 
terminated abnormally as shell returns the code of the last command in 
a pipeline.

General syntax is:
   launch [<IO>, ] <command_1> [, <command_2> ... [, <command_N>] { | 
line_0, line_1, ... line_N | block } -> anArray
   <IO> - an optional IO object. If specified, the first command's stdin 
will be redirected from this IO object. Otherwise "/dev/null" is used.
   <command> - all commands are spawned (fork & exec) with stdout of the 
previous command redirected to stdin of the next command. Stdout of the 
last command is read out and messages are stored in the array available 
as the first element of the returned array. All other elements of the 
returned array store error messages from stderr's of all commands 
respectively.

If a block is specified, lines read from stdout of the last command and 
stderrs of all command are made available to the block immediately 
(messages are not stored in the returned array): line_0 (if not nil) is 
a line from stdout of the last command, line_1 .. line_N (if not nils) 
are lines read from stderrs of command_1 ... command_N respectively. 
Only one line passed to the block is not nil.

I use the block form when I need to show the progress of a long running 
Unix command (like tar).

Gennady.


On Tuesday, May 27, 2003, at 06:24 AM, mgarriss wrote:

> so I can get the standard output from a system call:
>
>   s = `date`
>
> and I can get the return value:
>
>   system( "date" )
>
> and I have no idea how to get the standard error:
>
>   ???
>
> but what I really want is all three, like:
>
>   out, err, ret = really_cool_system( "date" )
>
> Please Help!
>
> TIA
>    Michael Garriss
>