Table 14.1 summarizes the operations that you have for launching a process.
Operation | Standard Input | Standard Output | Standard Error | Waited for? |
---|---|---|---|---|
system() | Inherited from program | Inherited from program | Inherited from program | Yes |
Backquoted string | Inherited from program | Captured as string value | Inherited from program | Yes |
| Connected to filehandle | Inherited from program | Inherited from program | Only at time of |
| Inherited from program | Connected to filehandle | Inherited from program | Only at time of |
| User selected | User selected | User selected | User selected |
The simplest way to create a process is with the system
function. Standard input, output, and error are unaffected (they're inherited from the Perl process). A backquoted string creates a process, capturing the standard output of the process as a string value for the Perl program. Standard input and standard error are unaffected. Both these methods require that the process finish before any more code is executed.
A simple way to get an asynchronous process (one that allows the Perl program to continue before the process is complete) is to open a command as a filehandle, creating a pipe for the command's standard input or standard output. A command opened as a filehandle for reading inherits the standard input and standard error from the Perl program; a command opened as a filehandle for writing inherits the standard output and standard error from the Perl program.
The most flexible way of starting a process is to have your program invoke the fork
, exec
, and wait
or waitpid
functions, which map directly to their UNIX system call namesakes. Using these functions, you can select whether you are waiting or not, and configure the standard input, output, and error any way you choose.[5]
[5] Although it might also help to know about
open(STDERR,">&STDOUT")
forms for fine tuning the filehandles. See the open entry in Chapter 3 of Programming Perl, or in perlfunc (1).