3 perlfaq8 - System Interaction
7 This section of the Perl FAQ covers questions involving operating
8 system interaction. Topics include interprocess communication (IPC),
9 control over the user-interface (keyboard, screen and pointing
10 devices), and most anything else not related to data manipulation.
12 Read the FAQs and documentation specific to the port of perl to your
13 operating system (eg, L<perlvms>, L<perlplan9>, ...). These should
14 contain more detailed information on the vagaries of your perl.
16 =head2 How do I find out which operating system I'm running under?
18 The $^O variable ($OSNAME if you use English) contains an indication of
19 the name of the operating system (not its release number) that your perl
22 =head2 How come exec() doesn't return?
23 X<exec> X<system> X<fork> X<open> X<pipe>
25 (contributed by brian d foy)
27 The C<exec> function's job is to turn your process into another
28 command and never to return. If that's not what you want to do, don't
31 If you want to run an external command and still keep your Perl process
32 going, look at a piped C<open>, C<fork>, or C<system>.
34 =head2 How do I do fancy stuff with the keyboard/screen/mouse?
36 How you access/control keyboards, screens, and pointing devices
37 ("mice") is system-dependent. Try the following modules:
43 Term::Cap Standard perl distribution
45 Term::ReadLine::Gnu CPAN
46 Term::ReadLine::Perl CPAN
51 Term::Cap Standard perl distribution
61 Some of these specific cases are shown as examples in other answers
62 in this section of the perlfaq.
64 =head2 How do I print something out in color?
66 In general, you don't, because you don't know whether
67 the recipient has a color-aware display device. If you
68 know that they have an ANSI terminal that understands
69 color, you can use the Term::ANSIColor module from CPAN:
72 print color("red"), "Stop!\n", color("reset");
73 print color("green"), "Go!\n", color("reset");
77 use Term::ANSIColor qw(:constants);
78 print RED, "Stop!\n", RESET;
79 print GREEN, "Go!\n", RESET;
81 =head2 How do I read just one key without waiting for a return key?
83 Controlling input buffering is a remarkably system-dependent matter.
84 On many systems, you can just use the B<stty> command as shown in
85 L<perlfunc/getc>, but as you see, that's already getting you into
88 open(TTY, "+</dev/tty") or die "no tty: $!";
89 system "stty cbreak </dev/tty >/dev/tty 2>&1";
90 $key = getc(TTY); # perhaps this works
92 sysread(TTY, $key, 1); # probably this does
93 system "stty -cbreak </dev/tty >/dev/tty 2>&1";
95 The Term::ReadKey module from CPAN offers an easy-to-use interface that
96 should be more efficient than shelling out to B<stty> for each key.
97 It even includes limited support for Windows.
104 However, using the code requires that you have a working C compiler
105 and can use it to build and install a CPAN module. Here's a solution
106 using the standard POSIX module, which is already on your systems
107 (assuming your system supports POSIX).
112 And here's the HotKey module, which hides the somewhat mystifying calls
113 to manipulate the POSIX termios structures.
119 @EXPORT = qw(cbreak cooked readkey);
122 use POSIX qw(:termios_h);
123 my ($term, $oterm, $echo, $noecho, $fd_stdin);
125 $fd_stdin = fileno(STDIN);
126 $term = POSIX::Termios->new();
127 $term->getattr($fd_stdin);
128 $oterm = $term->getlflag();
130 $echo = ECHO | ECHOK | ICANON;
131 $noecho = $oterm & ~$echo;
134 $term->setlflag($noecho); # ok, so i don't want echo either
135 $term->setcc(VTIME, 1);
136 $term->setattr($fd_stdin, TCSANOW);
140 $term->setlflag($oterm);
141 $term->setcc(VTIME, 0);
142 $term->setattr($fd_stdin, TCSANOW);
148 sysread(STDIN, $key, 1);
157 =head2 How do I check whether input is ready on the keyboard?
159 The easiest way to do this is to read a key in nonblocking mode with the
160 Term::ReadKey module from CPAN, passing it an argument of -1 to indicate
167 if (defined ($char = ReadKey(-1)) ) {
168 # input was waiting and it was $char
170 # no input was waiting
173 ReadMode('normal'); # restore normal tty settings
175 =head2 How do I clear the screen?
177 (contributed by brian d foy)
179 To clear the screen, you just have to print the special sequence
180 that tells the terminal to clear the screen. Once you have that
181 sequence, output it when you want to clear the screen.
183 You can use the C<Term::ANSIScreen> module to get the special
184 sequence. Import the C<cls> function (or the C<:screen> tag):
186 use Term::ANSIScreen qw(cls);
187 my $clear_screen = cls();
191 The C<Term::Cap> module can also get the special sequence if you want
192 to deal with the low-level details of terminal control. The C<Tputs>
193 method returns the string for the given capability:
197 $terminal = Term::Cap->Tgetent( { OSPEED => 9600 } );
198 $clear_string = $terminal->Tputs('cl');
202 On Windows, you can use the C<Win32::Console> module. After creating
203 an object for the output filehandle you want to affect, call the
208 $OUT = Win32::Console->new(STD_OUTPUT_HANDLE);
209 my $clear_string = $OUT->Cls;
213 If you have a command-line program that does the job, you can call
214 it in backticks to capture whatever it outputs so you can use it
217 $clear_string = `clear`;
221 =head2 How do I get the screen size?
223 If you have Term::ReadKey module installed from CPAN,
224 you can use it to fetch the width and height in characters
228 ($wchar, $hchar, $wpixels, $hpixels) = GetTerminalSize();
230 This is more portable than the raw C<ioctl>, but not as
233 require 'sys/ioctl.ph';
234 die "no TIOCGWINSZ " unless defined &TIOCGWINSZ;
235 open(TTY, "+</dev/tty") or die "No tty: $!";
236 unless (ioctl(TTY, &TIOCGWINSZ, $winsize='')) {
237 die sprintf "$0: ioctl TIOCGWINSZ (%08x: $!)\n", &TIOCGWINSZ;
239 ($row, $col, $xpixel, $ypixel) = unpack('S4', $winsize);
240 print "(row,col) = ($row,$col)";
241 print " (xpixel,ypixel) = ($xpixel,$ypixel)" if $xpixel || $ypixel;
244 =head2 How do I ask the user for a password?
246 (This question has nothing to do with the web. See a different
249 There's an example of this in L<perlfunc/crypt>). First, you put the
250 terminal into "no echo" mode, then just read the password normally.
251 You may do this with an old-style ioctl() function, POSIX terminal
252 control (see L<POSIX> or its documentation the Camel Book), or a call
253 to the B<stty> program, with varying degrees of portability.
255 You can also do this for most systems using the Term::ReadKey module
256 from CPAN, which is easier to use and in theory more portable.
261 $password = ReadLine(0);
263 =head2 How do I read and write the serial port?
265 This depends on which operating system your program is running on. In
266 the case of Unix, the serial ports will be accessible through files in
267 /dev; on other systems, device names will doubtless differ.
268 Several problem areas common to all device interaction are the
275 Your system may use lockfiles to control multiple access. Make sure
276 you follow the correct protocol. Unpredictable behavior can result
277 from multiple processes reading from one device.
281 If you expect to use both read and write operations on the device,
282 you'll have to open it for update (see L<perlfunc/"open"> for
283 details). You may wish to open it without running the risk of
284 blocking by using sysopen() and C<O_RDWR|O_NDELAY|O_NOCTTY> from the
285 Fcntl module (part of the standard perl distribution). See
286 L<perlfunc/"sysopen"> for more on this approach.
290 Some devices will be expecting a "\r" at the end of each line rather
291 than a "\n". In some ports of perl, "\r" and "\n" are different from
292 their usual (Unix) ASCII values of "\012" and "\015". You may have to
293 give the numeric values you want directly, using octal ("\015"), hex
294 ("0x0D"), or as a control-character specification ("\cM").
296 print DEV "atv1\012"; # wrong, for some devices
297 print DEV "atv1\015"; # right, for some devices
299 Even though with normal text files a "\n" will do the trick, there is
300 still no unified scheme for terminating a line that is portable
301 between Unix, DOS/Win, and Macintosh, except to terminate I<ALL> line
302 ends with "\015\012", and strip what you don't need from the output.
303 This applies especially to socket I/O and autoflushing, discussed
306 =item flushing output
308 If you expect characters to get to your device when you print() them,
309 you'll want to autoflush that filehandle. You can use select()
310 and the C<$|> variable to control autoflushing (see L<perlvar/$E<verbar>>
311 and L<perlfunc/select>, or L<perlfaq5>, "How do I flush/unbuffer an
312 output filehandle? Why must I do this?"):
318 You'll also see code that does this without a temporary variable, as in
320 select((select(DEV), $| = 1)[0]);
322 Or if you don't mind pulling in a few thousand lines
323 of code just because you're afraid of a little $| variable:
328 As mentioned in the previous item, this still doesn't work when using
329 socket I/O between Unix and Macintosh. You'll need to hard code your
330 line terminators, in that case.
332 =item non-blocking input
334 If you are doing a blocking read() or sysread(), you'll have to
335 arrange for an alarm handler to provide a timeout (see
336 L<perlfunc/alarm>). If you have a non-blocking open, you'll likely
337 have a non-blocking read, which means you may have to use a 4-arg
338 select() to determine whether I/O is ready on that device (see
339 L<perlfunc/"select">.
343 While trying to read from his caller-id box, the notorious Jamie Zawinski
344 C<< <jwz@netscape.com> >>, after much gnashing of teeth and fighting with sysread,
345 sysopen, POSIX's tcgetattr business, and various other functions that
346 go bump in the night, finally came up with this:
350 my $stty = `/bin/stty -g`;
351 open2( \*MODEM_IN, \*MODEM_OUT, "cu -l$modem_device -s2400 2>&1");
352 # starting cu hoses /dev/tty's stty settings, even when it has
353 # been opened on a pipe...
354 system("/bin/stty $stty");
357 if ( !m/^Connected/ ) {
358 print STDERR "$0: cu printed `$_' instead of `Connected'\n";
362 =head2 How do I decode encrypted password files?
364 You spend lots and lots of money on dedicated hardware, but this is
365 bound to get you talked about.
367 Seriously, you can't if they are Unix password files--the Unix
368 password system employs one-way encryption. It's more like hashing
369 than encryption. The best you can do is check whether something else
370 hashes to the same string. You can't turn a hash back into the
371 original string. Programs like Crack can forcibly (and intelligently)
372 try to guess passwords, but don't (can't) guarantee quick success.
374 If you're worried about users selecting bad passwords, you should
375 proactively check when they try to change their password (by modifying
376 passwd(1), for example).
378 =head2 How do I start a process in the background?
380 (contributed by brian d foy)
382 There's not a single way to run code in the background so you don't
383 have to wait for it to finish before your program moves on to other
384 tasks. Process management depends on your particular operating system,
385 and many of the techniques are in L<perlipc>.
387 Several CPAN modules may be able to help, including IPC::Open2 or
388 IPC::Open3, IPC::Run, Parallel::Jobs, Parallel::ForkManager, POE,
389 Proc::Background, and Win32::Process. There are many other modules you
390 might use, so check those namespaces for other options too.
392 If you are on a unix-like system, you might be able to get away with a
393 system call where you put an C<&> on the end of the command:
397 You can also try using C<fork>, as described in L<perlfunc> (although
398 this is the same thing that many of the modules will do for you).
402 =item STDIN, STDOUT, and STDERR are shared
404 Both the main process and the backgrounded one (the "child" process)
405 share the same STDIN, STDOUT and STDERR filehandles. If both try to
406 access them at once, strange things can happen. You may want to close
407 or reopen these for the child. You can get around this with
408 C<open>ing a pipe (see L<perlfunc/"open">) but on some systems this
409 means that the child process cannot outlive the parent.
413 You'll have to catch the SIGCHLD signal, and possibly SIGPIPE too.
414 SIGCHLD is sent when the backgrounded process finishes. SIGPIPE is
415 sent when you write to a filehandle whose child process has closed (an
416 untrapped SIGPIPE can cause your program to silently die). This is
417 not an issue with C<system("cmd&")>.
421 You have to be prepared to "reap" the child process when it finishes.
423 $SIG{CHLD} = sub { wait };
425 $SIG{CHLD} = 'IGNORE';
427 You can also use a double fork. You immediately wait() for your
428 first child, and the init daemon will wait() for your grandchild once
431 unless ($pid = fork) {
433 exec "what you really wanna do";
440 See L<perlipc/"Signals"> for other examples of code to do this.
441 Zombies are not an issue with C<system("prog &")>.
445 =head2 How do I trap control characters/signals?
447 You don't actually "trap" a control character. Instead, that character
448 generates a signal which is sent to your terminal's currently
449 foregrounded process group, which you then trap in your process.
450 Signals are documented in L<perlipc/"Signals"> and the
451 section on "Signals" in the Camel.
453 You can set the values of the %SIG hash to be the functions you want
454 to handle the signal. After perl catches the signal, it looks in %SIG
455 for a key with the same name as the signal, then calls the subroutine
458 # as an anonymous subroutine
460 $SIG{INT} = sub { syswrite(STDERR, "ouch\n", 5 ) };
462 # or a reference to a function
466 # or the name of the function as a string
470 Perl versions before 5.8 had in its C source code signal handlers which
471 would catch the signal and possibly run a Perl function that you had set
472 in %SIG. This violated the rules of signal handling at that level
473 causing perl to dump core. Since version 5.8.0, perl looks at %SIG
474 *after* the signal has been caught, rather than while it is being caught.
475 Previous versions of this answer were incorrect.
477 =head2 How do I modify the shadow password file on a Unix system?
479 If perl was installed correctly and your shadow library was written
480 properly, the getpw*() functions described in L<perlfunc> should in
481 theory provide (read-only) access to entries in the shadow password
482 file. To change the file, make a new shadow password file (the format
483 varies from system to system--see L<passwd> for specifics) and use
484 pwd_mkdb(8) to install it (see L<pwd_mkdb> for more details).
486 =head2 How do I set the time and date?
488 Assuming you're running under sufficient permissions, you should be
489 able to set the system-wide date and time by running the date(1)
490 program. (There is no way to set the time and date on a per-process
491 basis.) This mechanism will work for Unix, MS-DOS, Windows, and NT;
492 the VMS equivalent is C<set time>.
494 However, if all you want to do is change your time zone, you can
495 probably get away with setting an environment variable:
497 $ENV{TZ} = "MST7MDT"; # unixish
498 $ENV{'SYS$TIMEZONE_DIFFERENTIAL'}="-5" # vms
499 system "trn comp.lang.perl.misc";
501 =head2 How can I sleep() or alarm() for under a second?
502 X<Time::HiRes> X<BSD::Itimer> X<sleep> X<select>
504 If you want finer granularity than the 1 second that the C<sleep()>
505 function provides, the easiest way is to use the C<select()> function as
506 documented in L<perlfunc/"select">. Try the C<Time::HiRes> and
507 the C<BSD::Itimer> modules (available from CPAN, and starting from
508 Perl 5.8 C<Time::HiRes> is part of the standard distribution).
510 =head2 How can I measure time under a second?
511 X<Time::HiRes> X<BSD::Itimer> X<sleep> X<select>
513 (contributed by brian d foy)
515 The C<Time::HiRes> module (part of the standard distribution as of
516 Perl 5.8) measures time with the C<gettimeofday()> system call, which
517 returns the time in microseconds since the epoch. If you can't install
518 C<Time::HiRes> for older Perls and you are on a Unixish system, you
519 may be able to call C<gettimeofday(2)> directly. See
522 =head2 How can I do an atexit() or setjmp()/longjmp()? (Exception handling)
524 Release 5 of Perl added the END block, which can be used to simulate
525 atexit(). Each package's END block is called when the program or
526 thread ends (see L<perlmod> manpage for more details).
528 For example, you can use this to make sure your filter program
529 managed to finish its output without filling up the disk:
532 close(STDOUT) || die "stdout close failed: $!";
535 The END block isn't called when untrapped signals kill the program,
536 though, so if you use END blocks you should also use
538 use sigtrap qw(die normal-signals);
540 Perl's exception-handling mechanism is its eval() operator. You can
541 use eval() as setjmp and die() as longjmp. For details of this, see
542 the section on signals, especially the time-out handler for a blocking
543 flock() in L<perlipc/"Signals"> or the section on "Signals" in
546 If exception handling is all you're interested in, try the
547 exceptions.pl library (part of the standard perl distribution).
549 If you want the atexit() syntax (and an rmexit() as well), try the
550 AtExit module available from CPAN.
552 =head2 Why doesn't my sockets program work under System V (Solaris)? What does the error message "Protocol not supported" mean?
554 Some Sys-V based systems, notably Solaris 2.X, redefined some of the
555 standard socket constants. Since these were constant across all
556 architectures, they were often hardwired into perl code. The proper
557 way to deal with this is to "use Socket" to get the correct values.
559 Note that even though SunOS and Solaris are binary compatible, these
560 values are different. Go figure.
562 =head2 How can I call my system's unique C functions from Perl?
564 In most cases, you write an external module to do it--see the answer
565 to "Where can I learn about linking C with Perl? [h2xs, xsubpp]".
566 However, if the function is a system call, and your system supports
567 syscall(), you can use the syscall function (documented in
570 Remember to check the modules that came with your distribution, and
571 CPAN as well--someone may already have written a module to do it. On
572 Windows, try Win32::API. On Macs, try Mac::Carbon. If no module
573 has an interface to the C function, you can inline a bit of C in your
574 Perl source with Inline::C.
576 =head2 Where do I get the include files to do ioctl() or syscall()?
578 Historically, these would be generated by the h2ph tool, part of the
579 standard perl distribution. This program converts cpp(1) directives
580 in C header files to files containing subroutine definitions, like
581 &SYS_getitimer, which you can use as arguments to your functions.
582 It doesn't work perfectly, but it usually gets most of the job done.
583 Simple files like F<errno.h>, F<syscall.h>, and F<socket.h> were fine,
584 but the hard ones like F<ioctl.h> nearly always need to hand-edited.
585 Here's how to install the *.ph files:
591 If your system supports dynamic loading, for reasons of portability and
592 sanity you probably ought to use h2xs (also part of the standard perl
593 distribution). This tool converts C header files to Perl extensions.
594 See L<perlxstut> for how to get started with h2xs.
596 If your system doesn't support dynamic loading, you still probably
597 ought to use h2xs. See L<perlxstut> and L<ExtUtils::MakeMaker> for
598 more information (in brief, just use B<make perl> instead of a plain
599 B<make> to rebuild perl with a new static extension).
601 =head2 Why do setuid perl scripts complain about kernel problems?
603 Some operating systems have bugs in the kernel that make setuid
604 scripts inherently insecure. Perl gives you a number of options
605 (described in L<perlsec>) to work around such systems.
607 =head2 How can I open a pipe both to and from a command?
609 The IPC::Open2 module (part of the standard perl distribution) is an
610 easy-to-use approach that internally uses pipe(), fork(), and exec() to do
611 the job. Make sure you read the deadlock warnings in its documentation,
612 though (see L<IPC::Open2>). See
613 L<perlipc/"Bidirectional Communication with Another Process"> and
614 L<perlipc/"Bidirectional Communication with Yourself">
616 You may also use the IPC::Open3 module (part of the standard perl
617 distribution), but be warned that it has a different order of
618 arguments from IPC::Open2 (see L<IPC::Open3>).
620 =head2 Why can't I get the output of a command with system()?
622 You're confusing the purpose of system() and backticks (``). system()
623 runs a command and returns exit status information (as a 16 bit value:
624 the low 7 bits are the signal the process died from, if any, and
625 the high 8 bits are the actual exit value). Backticks (``) run a
626 command and return what it sent to STDOUT.
628 $exit_status = system("mail-users");
629 $output_string = `ls`;
631 =head2 How can I capture STDERR from an external command?
633 There are three basic ways of running external commands:
635 system $cmd; # using system()
636 $output = `$cmd`; # using backticks (``)
637 open (PIPE, "cmd |"); # using open()
639 With system(), both STDOUT and STDERR will go the same place as the
640 script's STDOUT and STDERR, unless the system() command redirects them.
641 Backticks and open() read B<only> the STDOUT of your command.
643 You can also use the open3() function from IPC::Open3. Benjamin
644 Goldberg provides some sample code:
646 To capture a program's STDOUT, but discard its STDERR:
650 use Symbol qw(gensym);
651 open(NULL, ">", File::Spec->devnull);
652 my $pid = open3(gensym, \*PH, ">&NULL", "cmd");
656 To capture a program's STDERR, but discard its STDOUT:
660 use Symbol qw(gensym);
661 open(NULL, ">", File::Spec->devnull);
662 my $pid = open3(gensym, ">&NULL", \*PH, "cmd");
666 To capture a program's STDERR, and let its STDOUT go to our own STDERR:
669 use Symbol qw(gensym);
670 my $pid = open3(gensym, ">&STDERR", \*PH, "cmd");
674 To read both a command's STDOUT and its STDERR separately, you can
675 redirect them to temp files, let the command run, then read the temp
679 use Symbol qw(gensym);
681 local *CATCHOUT = IO::File->new_tmpfile;
682 local *CATCHERR = IO::File->new_tmpfile;
683 my $pid = open3(gensym, ">&CATCHOUT", ">&CATCHERR", "cmd");
685 seek $_, 0, 0 for \*CATCHOUT, \*CATCHERR;
686 while( <CATCHOUT> ) {}
687 while( <CATCHERR> ) {}
689 But there's no real need for *both* to be tempfiles... the following
690 should work just as well, without deadlocking:
693 use Symbol qw(gensym);
695 local *CATCHERR = IO::File->new_tmpfile;
696 my $pid = open3(gensym, \*CATCHOUT, ">&CATCHERR", "cmd");
697 while( <CATCHOUT> ) {}
700 while( <CATCHERR> ) {}
702 And it'll be faster, too, since we can begin processing the program's
703 stdout immediately, rather than waiting for the program to finish.
705 With any of these, you can change file descriptors before the call:
707 open(STDOUT, ">logfile");
710 or you can use Bourne shell file-descriptor redirection:
712 $output = `$cmd 2>some_file`;
713 open (PIPE, "cmd 2>some_file |");
715 You can also use file-descriptor redirection to make STDERR a
718 $output = `$cmd 2>&1`;
719 open (PIPE, "cmd 2>&1 |");
721 Note that you I<cannot> simply open STDERR to be a dup of STDOUT
722 in your Perl program and avoid calling the shell to do the redirection.
725 open(STDERR, ">&STDOUT");
726 $alloutput = `cmd args`; # stderr still escapes
728 This fails because the open() makes STDERR go to where STDOUT was
729 going at the time of the open(). The backticks then make STDOUT go to
730 a string, but don't change STDERR (which still goes to the old
733 Note that you I<must> use Bourne shell (sh(1)) redirection syntax in
734 backticks, not csh(1)! Details on why Perl's system() and backtick
735 and pipe opens all use the Bourne shell are in the
736 F<versus/csh.whynot> article in the "Far More Than You Ever Wanted To
737 Know" collection in http://www.cpan.org/misc/olddoc/FMTEYEWTK.tgz . To
738 capture a command's STDERR and STDOUT together:
740 $output = `cmd 2>&1`; # either with backticks
741 $pid = open(PH, "cmd 2>&1 |"); # or with an open pipe
742 while (<PH>) { } # plus a read
744 To capture a command's STDOUT but discard its STDERR:
746 $output = `cmd 2>/dev/null`; # either with backticks
747 $pid = open(PH, "cmd 2>/dev/null |"); # or with an open pipe
748 while (<PH>) { } # plus a read
750 To capture a command's STDERR but discard its STDOUT:
752 $output = `cmd 2>&1 1>/dev/null`; # either with backticks
753 $pid = open(PH, "cmd 2>&1 1>/dev/null |"); # or with an open pipe
754 while (<PH>) { } # plus a read
756 To exchange a command's STDOUT and STDERR in order to capture the STDERR
757 but leave its STDOUT to come out our old STDERR:
759 $output = `cmd 3>&1 1>&2 2>&3 3>&-`; # either with backticks
760 $pid = open(PH, "cmd 3>&1 1>&2 2>&3 3>&-|");# or with an open pipe
761 while (<PH>) { } # plus a read
763 To read both a command's STDOUT and its STDERR separately, it's easiest
764 to redirect them separately to files, and then read from those files
765 when the program is done:
767 system("program args 1>program.stdout 2>program.stderr");
769 Ordering is important in all these examples. That's because the shell
770 processes file descriptor redirections in strictly left to right order.
772 system("prog args 1>tmpfile 2>&1");
773 system("prog args 2>&1 1>tmpfile");
775 The first command sends both standard out and standard error to the
776 temporary file. The second command sends only the old standard output
777 there, and the old standard error shows up on the old standard out.
779 =head2 Why doesn't open() return an error when a pipe open fails?
781 If the second argument to a piped open() contains shell
782 metacharacters, perl fork()s, then exec()s a shell to decode the
783 metacharacters and eventually run the desired program. If the program
784 couldn't be run, it's the shell that gets the message, not Perl. All
785 your Perl program can find out is whether the shell itself could be
786 successfully started. You can still capture the shell's STDERR and
787 check it for error messages. See L<"How can I capture STDERR from an
788 external command?"> elsewhere in this document, or use the
791 If there are no shell metacharacters in the argument of open(), Perl
792 runs the command directly, without using the shell, and can correctly
793 report whether the command started.
795 =head2 What's wrong with using backticks in a void context?
797 Strictly speaking, nothing. Stylistically speaking, it's not a good
798 way to write maintainable code. Perl has several operators for
799 running external commands. Backticks are one; they collect the output
800 from the command for use in your program. The C<system> function is
801 another; it doesn't do this.
803 Writing backticks in your program sends a clear message to the readers
804 of your code that you wanted to collect the output of the command.
805 Why send a clear message that isn't true?
811 You forgot to check C<$?> to see whether the program even ran
812 correctly. Even if you wrote
814 print `cat /etc/termcap`;
816 this code could and probably should be written as
818 system("cat /etc/termcap") == 0
819 or die "cat program failed!";
821 which will echo the cat command's output as it is generated, instead
822 of waiting until the program has completed to print it out. It also
823 checks the return value.
825 C<system> also provides direct control over whether shell wildcard
826 processing may take place, whereas backticks do not.
828 =head2 How can I call backticks without shell processing?
830 This is a bit tricky. You can't simply write the command
833 @ok = `grep @opts '$search_string' @filenames`;
835 As of Perl 5.8.0, you can use C<open()> with multiple arguments.
836 Just like the list forms of C<system()> and C<exec()>, no shell
839 open( GREP, "-|", 'grep', @opts, $search_string, @filenames );
846 if (open(GREP, "-|")) {
853 exec 'grep', @opts, $search_string, @filenames;
856 Just as with C<system()>, no shell escapes happen when you C<exec()> a
857 list. Further examples of this can be found in L<perlipc/"Safe Pipe
860 Note that if you're using Windows, no solution to this vexing issue is
861 even possible. Even if Perl were to emulate C<fork()>, you'd still be
862 stuck, because Windows does not have an argc/argv-style API.
864 =head2 Why can't my script read from STDIN after I gave it EOF (^D on Unix, ^Z on MS-DOS)?
866 This happens only if your perl is compiled to use stdio instead of
867 perlio, which is the default. Some (maybe all?) stdio's set error and
868 eof flags that you may need to clear. The POSIX module defines
869 clearerr() that you can use. That is the technically correct way to
870 do it. Here are some less reliable workarounds:
876 Try keeping around the seekpointer and go there, like this:
879 seek(LOG, $where, 0);
883 If that doesn't work, try seeking to a different part of the file and
888 If that doesn't work, try seeking to a different part of
889 the file, reading something, and then seeking back.
893 If that doesn't work, give up on your stdio package and use sysread.
897 =head2 How can I convert my shell script to perl?
899 Learn Perl and rewrite it. Seriously, there's no simple converter.
900 Things that are awkward to do in the shell are easy to do in Perl, and
901 this very awkwardness is what would make a shell->perl converter
902 nigh-on impossible to write. By rewriting it, you'll think about what
903 you're really trying to do, and hopefully will escape the shell's
904 pipeline datastream paradigm, which while convenient for some matters,
905 causes many inefficiencies.
907 =head2 Can I use perl to run a telnet or ftp session?
909 Try the Net::FTP, TCP::Client, and Net::Telnet modules (available from
910 CPAN). http://www.cpan.org/scripts/netstuff/telnet.emul.shar
911 will also help for emulating the telnet protocol, but Net::Telnet is
912 quite probably easier to use..
914 If all you want to do is pretend to be telnet but don't need
915 the initial telnet handshaking, then the standard dual-process
916 approach will suffice:
918 use IO::Socket; # new in 5.004
919 $handle = IO::Socket::INET->new('www.perl.com:80')
920 or die "can't connect to port 80 on www.perl.com: $!";
921 $handle->autoflush(1);
922 if (fork()) { # XXX: undef means failure
924 print while <STDIN>; # everything from stdin to socket
926 print while <$handle>; # everything from socket to stdout
931 =head2 How can I write expect in Perl?
933 Once upon a time, there was a library called chat2.pl (part of the
934 standard perl distribution), which never really got finished. If you
935 find it somewhere, I<don't use it>. These days, your best bet is to
936 look at the Expect module available from CPAN, which also requires two
937 other modules from CPAN, IO::Pty and IO::Stty.
939 =head2 Is there a way to hide perl's command line from programs such as "ps"?
941 First of all note that if you're doing this for security reasons (to
942 avoid people seeing passwords, for example) then you should rewrite
943 your program so that critical information is never given as an
944 argument. Hiding the arguments won't make your program completely
947 To actually alter the visible command line, you can assign to the
948 variable $0 as documented in L<perlvar>. This won't work on all
949 operating systems, though. Daemon programs like sendmail place their
952 $0 = "orcus [accepting connections]";
954 =head2 I {changed directory, modified my environment} in a perl script. How come the change disappeared when I exited the script? How do I get my changes to be visible?
960 In the strictest sense, it can't be done--the script executes as a
961 different process from the shell it was started from. Changes to a
962 process are not reflected in its parent--only in any children
963 created after the change. There is shell magic that may allow you to
964 fake it by eval()ing the script's output in your shell; check out the
965 comp.unix.questions FAQ for details.
969 =head2 How do I close a process's filehandle without waiting for it to complete?
971 Assuming your system supports such things, just send an appropriate signal
972 to the process (see L<perlfunc/"kill">). It's common to first send a TERM
973 signal, wait a little bit, and then send a KILL signal to finish it off.
975 =head2 How do I fork a daemon process?
977 If by daemon process you mean one that's detached (disassociated from
978 its tty), then the following process is reported to work on most
979 Unixish systems. Non-Unix users should check their Your_OS::Process
980 module for other solutions.
986 Open /dev/tty and use the TIOCNOTTY ioctl on it. See L<tty>
987 for details. Or better yet, you can just use the POSIX::setsid()
988 function, so you don't have to worry about process groups.
992 Change directory to /
996 Reopen STDIN, STDOUT, and STDERR so they're not connected to the old
1001 Background yourself like this:
1007 The Proc::Daemon module, available from CPAN, provides a function to
1008 perform these actions for you.
1010 =head2 How do I find out if I'm running interactively or not?
1012 (contributed by brian d foy)
1014 This is a difficult question to answer, and the best answer is
1015 only a guess. What do you really want to know? If you merely
1016 want to know if one of your filehandles is connected to a terminal,
1017 you can try the C<-t> file test:
1020 print "I'm connected to a terminal!\n";
1023 However, you might be out of luck if you expect that means there is a
1024 real person on the other side. With the C<Expect> module, another
1025 program can pretend to be a person. The program might even come close
1026 to passing the Turing test.
1028 The C<IO::Interactive> module does the best it can to give you an
1029 answer. Its C<is_interactive> function returns an output filehandle;
1030 that filehandle points to standard output if the module thinks the
1031 session is interactive. Otherwise, the filehandle is a null handle
1032 that simply discards the output:
1034 use IO::Interactive;
1036 print { is_interactive } "I might go to standard output!\n";
1038 This still doesn't guarantee that a real person is answering your
1039 prompts or reading your output.
1041 If you want to know how to handle automated testing for your
1042 distribution, you can check the environment. The CPAN
1043 Testers, for instance, set the value of C<AUTOMATED_TESTING>:
1045 unless( $ENV{AUTOMATED_TESTING} ) {
1046 print "Hello interactive tester!\n";
1049 =head2 How do I timeout a slow event?
1051 Use the alarm() function, probably in conjunction with a signal
1052 handler, as documented in L<perlipc/"Signals"> and the section on
1053 "Signals" in the Camel. You may instead use the more flexible
1054 Sys::AlarmCall module available from CPAN.
1056 The alarm() function is not implemented on all versions of Windows.
1057 Check the documentation for your specific version of Perl.
1059 =head2 How do I set CPU limits?
1060 X<BSD::Resource> X<limit> X<CPU>
1062 (contributed by Xho)
1064 Use the C<BSD::Resource> module from CPAN. As an example:
1067 setrlimit(RLIMIT_CPU,10,20) or die $!;
1069 This sets the soft and hard limits to 10 and 20 seconds, respectively.
1070 After 10 seconds of time spent running on the CPU (not "wall" time),
1071 the process will be sent a signal (XCPU on some systems) which, if not
1072 trapped, will cause the process to terminate. If that signal is
1073 trapped, then after 10 more seconds (20 seconds in total) the process
1074 will be killed with a non-trappable signal.
1076 See the C<BSD::Resource> and your systems documentation for the gory
1079 =head2 How do I avoid zombies on a Unix system?
1081 Use the reaper code from L<perlipc/"Signals"> to call wait() when a
1082 SIGCHLD is received, or else use the double-fork technique described
1083 in L<perlfaq8/"How do I start a process in the background?">.
1085 =head2 How do I use an SQL database?
1087 The DBI module provides an abstract interface to most database
1088 servers and types, including Oracle, DB2, Sybase, mysql, Postgresql,
1089 ODBC, and flat files. The DBI module accesses each database type
1090 through a database driver, or DBD. You can see a complete list of
1091 available drivers on CPAN: http://www.cpan.org/modules/by-module/DBD/ .
1092 You can read more about DBI on http://dbi.perl.org .
1094 Other modules provide more specific access: Win32::ODBC, Alzabo, iodbc,
1095 and others found on CPAN Search: http://search.cpan.org .
1097 =head2 How do I make a system() exit on control-C?
1099 You can't. You need to imitate the system() call (see L<perlipc> for
1100 sample code) and then have a signal handler for the INT signal that
1101 passes the signal on to the subprocess. Or you can check for it:
1104 if ($rc & 127) { die "signal death" }
1106 =head2 How do I open a file without blocking?
1108 If you're lucky enough to be using a system that supports
1109 non-blocking reads (most Unixish systems do), you need only to use the
1110 O_NDELAY or O_NONBLOCK flag from the Fcntl module in conjunction with
1114 sysopen(FH, "/foo/somefile", O_WRONLY|O_NDELAY|O_CREAT, 0644)
1115 or die "can't open /foo/somefile: $!":
1117 =head2 How do I tell the difference between errors from the shell and perl?
1119 (answer contributed by brian d foy)
1121 When you run a Perl script, something else is running the script for you,
1122 and that something else may output error messages. The script might
1123 emit its own warnings and error messages. Most of the time you cannot
1126 You probably cannot fix the thing that runs perl, but you can change how
1127 perl outputs its warnings by defining a custom warning and die functions.
1129 Consider this script, which has an error you may not notice immediately.
1131 #!/usr/locl/bin/perl
1133 print "Hello World\n";
1135 I get an error when I run this from my shell (which happens to be
1136 bash). That may look like perl forgot it has a print() function,
1137 but my shebang line is not the path to perl, so the shell runs the
1138 script, and I get the error.
1141 ./test: line 3: print: command not found
1143 A quick and dirty fix involves a little bit of code, but this may be all
1144 you need to figure out the problem.
1149 $SIG{__WARN__} = sub{ print STDERR "Perl: ", @_; };
1150 $SIG{__DIE__} = sub{ print STDERR "Perl: ", @_; exit 1};
1157 The perl message comes out with "Perl" in front. The BEGIN block
1158 works at compile time so all of the compilation errors and warnings
1159 get the "Perl:" prefix too.
1161 Perl: Useless use of division (/) in void context at ./test line 9.
1162 Perl: Name "main::a" used only once: possible typo at ./test line 8.
1163 Perl: Name "main::x" used only once: possible typo at ./test line 9.
1164 Perl: Use of uninitialized value in addition (+) at ./test line 8.
1165 Perl: Use of uninitialized value in division (/) at ./test line 9.
1166 Perl: Illegal division by zero at ./test line 9.
1167 Perl: Illegal division by zero at -e line 3.
1169 If I don't see that "Perl:", it's not from perl.
1171 You could also just know all the perl errors, and although there are
1172 some people who may know all of them, you probably don't. However, they
1173 all should be in the perldiag manpage. If you don't find the error in
1174 there, it probably isn't a perl error.
1176 Looking up every message is not the easiest way, so let perl to do it
1177 for you. Use the diagnostics pragma with turns perl's normal messages
1178 into longer discussions on the topic.
1182 If you don't get a paragraph or two of expanded discussion, it
1183 might not be perl's message.
1185 =head2 How do I install a module from CPAN?
1187 (contributed by brian d foy)
1189 The easiest way is to have a module also named CPAN do it for you by using
1190 the C<cpan> command the comes with Perl. You can give it a list of modules
1193 $ cpan IO::Interactive Getopt::Whatever
1195 If you prefer C<CPANPLUS>, it's just as easy:
1197 $ cpanp i IO::Interactive Getopt::Whatever
1199 If you want to install a distribution from the current directory, you can
1200 tell C<CPAN.pm> to install C<.> (the full stop):
1204 See the documentation for either of those commands to see what else
1207 If you want to try to install a distribution by yourself, resolving
1208 all dependencies on your own, you follow one of two possible build
1211 For distributions that use I<Makefile.PL>:
1216 For distributions that use I<Build.PL>:
1222 Some distributions may need to link to libraries or other third-party
1223 code and their build and installation sequences may be more complicated.
1224 Check any I<README> or I<INSTALL> files that you may find.
1226 =head2 What's the difference between require and use?
1228 (contributed by brian d foy)
1230 Perl runs C<require> statement at run-time. Once Perl loads, compiles,
1231 and runs the file, it doesn't do anything else. The C<use> statement
1232 is the same as a C<require> run at compile-time, but Perl also calls the
1233 C<import> method for the loaded package. These two are the same:
1235 use MODULE qw(import list);
1239 MODULE->import(import list);
1242 However, you can suppress the C<import> by using an explicit, empty
1243 import list. Both of these still happen at compile-time:
1251 Since C<use> will also call the C<import> method, the actual value
1252 for C<MODULE> must be a bareword. That is, C<use> cannot load files
1253 by name, although C<require> can:
1255 require "$ENV{HOME}/lib/Foo.pm"; # no @INC searching!
1257 See the entry for C<use> in L<perlfunc> for more details.
1259 =head2 How do I keep my own module/library directory?
1261 When you build modules, tell Perl where to install the modules.
1263 For C<Makefile.PL>-based distributions, use the INSTALL_BASE option
1264 when generating Makefiles:
1266 perl Makefile.PL INSTALL_BASE=/mydir/perl
1268 You can set this in your CPAN.pm configuration so modules automatically install
1269 in your private library directory when you use the CPAN.pm shell:
1272 cpan> o conf makepl_arg INSTALL_BASE=/mydir/perl
1275 For C<Build.PL>-based distributions, use the --install_base option:
1277 perl Build.PL --install_base /mydir/perl
1279 You can configure CPAN.pm to automatically use this option too:
1282 cpan> o conf mbuild_arg --install_base /mydir/perl
1285 INSTALL_BASE tells these tools to put your modules into
1286 F</mydir/perl/lib/perl5>. See L<How do I add a directory to my
1287 include path (@INC) at runtime?> for details on how to run your newly
1290 There is one caveat with INSTALL_BASE, though, since it acts
1291 differently than the PREFIX and LIB settings that older versions of
1292 ExtUtils::MakeMaker advocated. INSTALL_BASE does not support
1293 installing modules for multiple versions of Perl or different
1294 architectures under the same directory. You should consider if you
1295 really want that , and if you do, use the older PREFIX and LIB
1296 settings. See the ExtUtils::Makemaker documentation for more details.
1298 =head2 How do I add the directory my program lives in to the module/library search path?
1300 (contributed by brian d foy)
1302 If you know the directory already, you can add it to C<@INC> as you would
1303 for any other directory. You might <use lib> if you know the directory
1308 The trick in this task is to find the directory. Before your script does
1309 anything else (such as a C<chdir>), you can get the current working
1310 directory with the C<Cwd> module, which comes with Perl:
1314 our $directory = cwd;
1319 You can do a similar thing with the value of C<$0>, which holds the
1320 script name. That might hold a relative path, but C<rel2abs> can turn
1321 it into an absolute path. Once you have the
1324 use File::Spec::Functions qw(rel2abs);
1325 use File::Basename qw(dirname);
1327 my $path = rel2abs( $0 );
1328 our $directory = dirname( $path );
1333 The C<FindBin> module, which comes with Perl, might work. It finds the
1334 directory of the currently running script and puts it in C<$Bin>, which
1335 you can then use to construct the right library path:
1337 use FindBin qw($Bin);
1339 =head2 How do I add a directory to my include path (@INC) at runtime?
1341 Here are the suggested ways of modifying your include path, including
1342 environment variables, run-time switches, and in-code statements:
1346 =item the PERLLIB environment variable
1348 $ export PERLLIB=/path/to/my/dir
1351 =item the PERL5LIB environment variable
1353 $ export PERL5LIB=/path/to/my/dir
1356 =item the perl -Idir command line flag
1358 $ perl -I/path/to/my/dir program.pl
1360 =item the use lib pragma:
1362 use lib "$ENV{HOME}/myown_perllib";
1366 The last is particularly useful because it knows about machine
1367 dependent architectures. The lib.pm pragmatic module was first
1368 included with the 5.002 release of Perl.
1370 =head2 What is socket.ph and where do I get it?
1372 It's a Perl 4 style file defining values for system networking
1373 constants. Sometimes it is built using h2ph when Perl is installed,
1374 but other times it is not. Modern programs C<use Socket;> instead.
1378 Revision: $Revision$
1382 See L<perlfaq> for source control details and availability.
1384 =head1 AUTHOR AND COPYRIGHT
1386 Copyright (c) 1997-2009 Tom Christiansen, Nathan Torkington, and
1387 other authors as noted. All rights reserved.
1389 This documentation is free; you can redistribute it and/or modify it
1390 under the same terms as Perl itself.
1392 Irrespective of its distribution, all code examples in this file
1393 are hereby placed into the public domain. You are permitted and
1394 encouraged to use this code in your own programs for fun
1395 or for profit as you see fit. A simple comment in the code giving
1396 credit would be courteous but is not required.