I need to run a shell command with system() in Perl. For example,
system('ls')
The system call will print to STDOUT, but I want to capture the output into a variable so that I can do future processing with my Perl code.
That's what backticks are for. From perldoc perlfaq8:
Why can't I get the output of a command with system()?
You're confusing the purpose of system() and backticks (``). system()
runs a command and returns exit status information (as a 16 bit value:
the low 7 bits are the signal the process died from, if any, and the
high 8 bits are the actual exit value). Backticks (``) run a command
and return what it sent to STDOUT.
my $exit_status = system("mail-users");
my $output_string = `ls`;
See perldoc perlop for more details.
IPC::Run is my favourite module for this kind of task. Very powerful and flexible, and also trivially simple for small cases.
use IPC::Run 'run';
run [ "command", "arguments", "here" ], ">", \my $stdout;
# Now $stdout contains output
Simply use similar to the Bash example:
$variable=`some_command some args`;
That's all. Notice, you will not see any printings to STDOUT on the output because this is redirected to a variable.
This example is unusable for a command that interact with the user, except when you have prepared answers. For that, you can use something like this using a stack of shell commands:
$variable=`cat answers.txt|some_command some args`;
Inside the answers.txt file you should prepare all answers for some_command to work properly.
I know this isn't the best way for programming :) But this is the simplest way how to achieve the goal, specially for Bash programmers.
Of course, if the output is bigger (ls with subdirectory), you shouldn't get all output at once. Read the command by the same way as you read a regular file:
open CMD,'-|','your_command some args' or die $#;
my $line;
while (defined($line=<CMD>)) {
print $line; # Or push #table,$line or do whatever what you want processing line by line
}
close CMD;
An additional extended solution for processing a long command output without extra Bash calling:
my #CommandCall=qw(find / -type d); # Some example single command
my $commandSTDOUT; # File handler
my $pid=open($commandSTDOUT),'-|'); # There will be an implicit fork!
if ($pid) {
#parent side
my $singleLine;
while(defined($singleline=<$commandSTDOUT>)) {
chomp $line; # Typically we don't need EOL
do_some_processing_with($line);
};
close $commandSTDOUT; # In this place $? will be set for capture
$exitcode=$? >> 8;
do_something_with_exit_code($exitcode);
} else {
# Child side, there you really calls a command
open STDERR, '>>&', 'STDOUT'; # Redirect stderr to stdout if needed. It works only for child - remember about fork
exec(#CommandCall); # At this point the child code is overloaded by an external command with parameters
die "Cannot call #CommandCall"; # Error procedure if the call will fail
}
If you use a procedure like that, you will capture all procedure output, and you can do everything processing line by line. Good luck :)
I wanted to run system() instead of backticks because I wanted to see the output of rsync --progress. However, I also wanted to capture the output in case something goes wrong depending on the return value. (This is for a backup script). This is what I am using now:
use File::Temp qw(tempfile);
use Term::ANSIColor qw(colored colorstrip);
sub mysystem {
my $cmd = shift; # "rsync -avz --progress -h $fullfile $copyfile";
my ($fh, $filename) = tempfile();
# http://stackoverflow.com/a/6872163/2923406
# I want to have rsync progress output on the terminal AND capture it in case of error.
# Need to use pipefail because 'tee' would be the last cmd otherwise and hence $? would be wrong.
my #cmd = ("bash", "-c", "set -o pipefail && $cmd 2>&1 | tee $filename");
my $ret = system(#cmd);
my $outerr = join('', <$fh>);
if ($ret != 0) {
logit(colored("ERROR: Could not execute command: $cmd", "red"));
logit(colored("ERROR: stdout+stderr = $outerr", "red"));
logit(colored("ERROR: \$? = $?, \$! = $!", "red"));
}
close $fh;
unlink($filename);
return $ret;
}
# And logit() is something like:
sub logit {
my $s = shift;
my ($logsec, $logmin, $loghour, $logmday, $logmon, $logyear, $logwday, $logyday, $logisdst) = localtime(time);
$logyear += 1900;
my $logtimestamp = sprintf("%4d-%02d-%02d %02d:%02d:%02d", $logyear, $logmon+1, $logmday, $loghour, $logmin, $logsec);
my $msg = "$logtimestamp $s\n";
print $msg;
open LOG, ">>$LOGFILE";
print LOG colorstrip($msg);
close LOG;
}
Related
I am new to Perl and trying to write a code to keep executing an action until the match is found and else give an error.
I am trying to execute a command ps -ef and check if it has got any process running in the name of "box", if there is no process named "box" found, I want to repeat ps -ef command execution until it gets the "box" process and then proceed to next action.
#!/usr/bin/perl -w
open (FH, "ps -ef |") or die "Cannot run the command:$!\n";
$line = "box";
while (<FH>) {
if (/$line/i) { next; }
else {
print ("ps -ef |") or die "Cannot run the command:$!\n");
}
}
close (FH);
You need to use an infinite loop and an exit-condition. Your condition is that the ps -ef command contains the word box. There is no need to open a pipe to that command explicitly, you can just run it as a system call with the qx operator (same as backticks).
use strict;
use warnings;
my $ps;
PS: while (1) {
$ps = qx/ps -ef/;
last PS if $ps =~ m/box/i;
print '.'; # do something in every run
}
print $ps;
As this has come up in the comments as well as in in AdrianHHH's answer: it might make sense to sleep after every run to make sure you don't hog the CPU. Depending on the nature of the process you are looking for, either the sleep builtin or usleep from Time::HiRes might be appropriate. The latter let's your program rest for milliseconds, while the builtin only works with full seconds. These might be too long if the target box process is very quick.
Explanation of your code:
Note that you have some issues in your implementation. I'll explain what your code does. This is taken from the question, comments are mine.
#!/usr/bin/perl -w
# open a filehandle to the ps command
open (FH, "ps -ef |") or die "Cannot run the command:$!\n";
$line = "box";
# read the output of one run line by line, for each line execute
# the block
while (<FH>) {
# if there is 'box' case-insensitive, skip the line
if (/$line/i) { next; }
else {
# else output (not run!) the command
print ("ps -ef |") or die "Cannot run the command:$!\n");
}
}
close (FH);
After it went through all the lines of the output of your command once it will stop.
I would recommend using pgrep(1) instead of ps because it lets you do a more granular search. With ps -ef, you potentially have to deal with cases like:
boxford 6254 6211 0 08:23 pts/1 00:00:00 /home/boxford/box --bounding-box=123
It's hard to tell if you're matching a process being run by a user with box in their username, a process that has box somewhere in its path, a process named box, or a process with box somewhere in its argument list.
pgrep, on the other hand, lets you match against just the process name or the full path, a specific user or users, and more. The following prints a message when a process named box appears (this looks for an exact match, so it will not match processes named dropbox, for example):
use strict;
use warnings;
use 5.010;
use String::ShellQuote qw(shell_quote);
sub is_running {
my ($proc) = #_;
my $cmd = 'pgrep -x ' . shell_quote($proc) . ' >/dev/null 2>&1';
system($cmd);
if ($? == -1) {
die "failed to execute pgrep: $!";
}
elsif ($? & 127) {
die "pgrep died with signal ", $? & 127;
}
else {
my $status = $? >> 8;
die "pgrep exited with error: exit status $status" if $status > 1;
return $status == 0;
}
}
my $proc = 'box';
until ( is_running($proc) ) {
sleep 1;
}
say "Process '$proc' is running";
Note that pgrep doesn't have a case-insensitive flag, probably because process names in *nix are almost always lowercase. If you really need to do a case-insensitive match, you can pass [Bb][Oo][Xx] to the is_running function.
The ps command outputs the current list of processes, then it completes. The code in the question reads that output. Suppose that the first ps command that is executed does not contain the wanted line, then there is nothing in the code in the question to run the ps command again.
The next statement in the question makes the script move on to the next line in the output from ps, not to rerun the command. The else print ... after the next will probably be executed for the first line of the output from ps. The outcome is that the print is run for each line in the ps output that does not have the wanted text and that the next command has no significant effect. In the code print ... or die "..." the or die "..." part is not very useful, the print is unlikely to fail and even if it did the die message would be wrong.
Perhaps you should write some code in the following style. Here the ps is run repeatedly until the wanted text is found. Note the sleep call, without that the script will keep running without pause, possibly preventing real work or at least slowing it down.
# This code is not tested.
use strict;
use warnings;
my $found_wanted_line = 0; # Boolean, set to false
my $line = "box";
while ( ! $found_wanted_line ) {
open (my $FH, "ps -ef |") or die "Cannot run the command:$!\n";
while (<$FH>) {
if (/$line/i) {
$found_wanted_line = 1; # Boolean, set to true
last;
}
}
close ($FH);
if ( ! $found_wanted_line )
sleep 2; # Pause for 2 seconds, to prevent this script hogging the CPU.
}
}
From my Perl program, I am trying to run another command written in C using system. The command requires several arguments: a string, a float, two integers, a pair of floats, and another string. I am running this as
my $arg1="electron";
my $arg2=0.511;
# more definitions
system("./fermions $arg1 $arg2 $arg3 $arg4 " .
"$arg5 $arg6 \"string\" > outfile.out");
I need to vary $arg5 to be several different float values. I made it work by setting $arg5="1.0e5" and then running an if-statement in the for-loop to change the value as a string. I would like to be able to do this as floats, and tried
system("./fermions $arg1 $arg2 $arg3 $arg4 " .
"%e $arg6 \"string\" >outfile.out",
$arg5);
but that did not work. Is there another alternative, or is my if-statement option the only one?
If you want to use printf notation (like "%e"), you need to use the Perl sprintf builtin. Otherwise you just end up passing "%e" as a literal argument.
Important: You should always check the return value from Perl’s system function to determine whether the command failed.
Use Perl’s sprintf to format the float values as in the code below. Yes, you may be able to get away with using the command as your format specifier, but you’re likely to get surprising results if the command has stray % characters elsewhere. Using two steps is safer.
#! /usr/bin/env perl
use strict;
use warnings;
my #float_values = (1.0e5, 3.14159, 2.71828);
for my $f (#float_values) {
my $arg5 = sprintf "%e", $f;
system(qq[./fermions $arg5 "string" >> outfile.out]) == 0
or warn "$0: fermions failed";
}
In case you aren’t familiar with the syntax, qq[...] works like a double-quoted string, but the different delimiter means you don’t have to escape double-quotes in your command.
Note that I elided the other arguments for typographical purposes, but you can interpolate them along with the value of $arg5. Another subtle change is the switch to >> for appending rather than > for clobbering.
Using a stand-in for fermions
#! /usr/bin/env perl
$" = "][";
warn "[#ARGV]\n";
the output of the two programs running together is
[1.000000e+05][string]
[3.141590e+00][string]
[2.718280e+00][string]
With respect to terminology, a system call refers to a low-level request for service from the operating system, e.g., open, close, unlink, and so on. Although Perl’s system function makes use of system calls, the two concepts are distinct.
To be really safe about the shell not fudging your command-line arguments, use the techniques described in the “Safe Pipe Opens” section of perlipc. Perl’s system and exec functions bypass the shell when given a list of arguments rather than a single string containing the entire command.
Your situation is a little trickier because you want to redirect the standard output. The code below forks a child, sets the child’s STDOUT to append to outfile.out, and then runs fermion with exec in the child. The parent waits for the child to exit and reports any failure.
#! /usr/bin/env perl
use strict;
use warnings;
my #float_values = (1.0e5, 3.14159, 2.71828);
for my $f (#float_values) {
my $arg5 = sprintf "%e", $f;
my $pid = fork;
if (defined $pid) {
if ($pid == 0) {
my $path = "outfile.out";
open STDOUT, ">>", $path or die "$0: open $path: $!";
exec "./fermions", $arg5, "string" or die "$0: exec: $!";
}
else {
local $!;
my $pid = waitpid $pid, 0;
warn "$0: waitpid: $!" if $pid == -1 && $!;
warn "$0: fermion exited " . ($? >> 8) if $?;
}
}
else {
die "$0: fork: $!";
}
}
I want to call "env.sh " from "my_perl.pl" without forking a subshell. I tried with backtics and system like this --> system (. env.sh) [dot space env.sh] , however wont work.
Child environments cannot change parent environments. Your best bet is to parse env.sh from inside the Perl code and set the variables in %ENV:
#!/usr/bin/perl
use strict;
use warnings;
sub source {
my $name = shift;
open my $fh, "<", $name
or die "could not open $name: $!";
while (<$fh>) {
chomp;
my ($k, $v) = split /=/, $_, 2;
$v =~ s/^(['"])(.*)\1/$2/; #' fix highlighter
$v =~ s/\$([a-zA-Z]\w*)/$ENV{$1}/g;
$v =~ s/`(.*?)`/`$1`/ge; #dangerous
$ENV{$k} = $v;
}
}
source "env.sh";
for my $k (qw/foo bar baz quux/) {
print "$k => $ENV{$k}\n";
}
Given
foo=5
bar=10
baz="$foo$bar"
quux=`date +%Y%m%d`
it prints
foo => 5
bar => 10
baz => 510
quux => 20110726
The code can only handle simple files (for instance, it doesn't handle if statements or foo=$(date)). If you need something more complex, then writing a wrapper for your Perl script that sources env.sh first is the right way to go (it is also probably the right way to go in the first place).
Another reason to source env.sh before executing the Perl script is that setting the environment variables in Perl may happen too late for modules that are expecting to see them.
In the file foo:
#!/bin/bash
source env.sh
exec foo.real
where foo.real is your Perl script.
You can use arbitrarily complex shell scripts by executing them with the relevant shell, dumping their environment to standard output in the same process, and parsing that in perl. Feeding the output into something other than %ENV or filtering for specific values of interest is prudent so you don't change things like PATH that may have interesting side effects elsewhere. I've discarded standard output and error from the spawned shell script although they could be redirected to temporary files and used for diagnostic output in the perl script.
foo.pl:
#!/usr/bin/perl
open SOURCE, "bash -c '. foo.sh >& /dev/null; env'|" or
die "Can't fork: $!";
while(<SOURCE>) {
if (/^(BAR|BAZ)=(.*)/) {
$ENV{$1} = ${2} ;
}
}
close SOURCE;
print $ENV{'BAR'} . "\n";
foo.sh:
export BAR=baz
Try this (unix code sample):
cd /tmp
vi s
#!/bin/bash
export blah=test
vi t
#!/usr/bin/perl
if ($ARGV[0]) {
print "ENV second call is : $ENV{blah}\n";
} else {
print "ENV first call is : $ENV{blah}\n";
exec(". /tmp/s; /tmp/t 1");
}
chmod 777 s t
./t
ENV first call is :
ENV second call is : test
The trick is using the exec to source your bash script first and then calling your perl script again with an argument so u know that you are being called for a second time.
I'm writing a Perl script which uses an external script. The external script must run from a specific directory so I found the following useful:
use IPC::System::Simple qw(capture);
my #args = ('external script path...', 'arg1', ...);
my $out = capture( [0], "cd $dir ; #args" );
Sometimes the external script writes stuff to STDERR but still returns 0. I wish to capture these times and confess (or die). Since I don't control the return value of the external script, I thought maybe I could capture its STDERR so I'll have something like this:
my ($out, $err) = cool_capture( [0], "cd $dir ; #args" );
say "Output was: $out";
if ($err) {
die "Error: this was written to STDERR: $err";
}
What can I do?
This is covered in the Perl FAQ.
Presuming test_app is a program that outputs one line to stdout and one line to stderr:
use IPC::Open3;
use Symbol 'gensym';
my($wtr, $rdr, $err);
$err = gensym;
my $pid = open3($wtr, $rdr, $err, 'test_app');
waitpid($pid, 0);
my $status = $? >> 8;
my $stdout = <$rdr>;
my $stderr = <$err>;
print "out output: $stdout\n";
print "err output: $stderr\n";
print "Exit code: $status\n";
EDIT: Per the request updated to include capturing the exit code. You could also have asked perldoc IPC::Open3 which says
waitpid( $pid, 0 );
my $child_exit_status = $? >> 8;
And which you should read anyway for its cautions and caveats.
If significant output is being written to stdout and/or stderr or you're both reading and writing to the process. You need to be a lot more careful with your I/O handling to avoid various blocking problems.
my ($wtr, $rdr, $err) ;
my $pid = IPC::Open3::open3($wtr, $rdr, $err, #_);
close($wtr);
my $stdout = '';
my $stderr = '';
my $s = IO::Select->new;
$s->add($rdr) if $rdr;
$s->add($err) if $err;
while (my #ready = $s->can_read) {
foreach my $ioh (#ready) {
my $bytes_read = sysread($ioh, my $chunk = '', 1024);
die "read error: $!" unless $bytes_read >= 0;
if ($bytes_read) {
($ioh eq $rdr? $stdout: $stderr) .= $chunk;
}
else {
$s->remove($ioh);
}
}
}
my $pid1;
for (;;) {
last if kill(0, $pid);
$pid1 = wait();
#
# Wait until we see the process or -1 (no active processes);
#
last if ($pid1 == $pid || $pid1 <= 0);
}
Finish reading before you shutdown the process. If you're writing to the process's stdin, you'd also need to add $wtr and syswrite to the above select loop.
EDIT
Rationale:
The above is probably overkill for simple cases. This advanced handling of input and output comes into play when you're likely to move more than a few K of data.
You wouldn't need it if you were executing a 'df' command for example.
However, it's when system buffers for any of stdin, stdout or stderr fill up that blocking becomes likely and things can get more involved.
If the child process fills up the stderr and/or stdout buffers, it'll likely block and wait for you to clear them. But if you're waiting for the process finish before you read from stdout or stderr; thats a deadlock. You'll likely to see that the system call never finishes and the child process never completes.
There's a similar possibility of deadlock if stdin is being written to, but the child process is unable to consume the input. This is particularly likely in a 'pipe' situation where the child process is consuming input and writing to stdout.
The select loop is about progressively clearing the buffers to avoid blocking. Both stdout and stderr are monitored concurrently.
If you're writing to stdin and reading from stdout (a pipe), you'll want to keep stdout and stderr clear and only write to stdin when its ready to receive input.
Simply waiting for the process to finish, then reading stdout/stderr probably works 90% of the time. This reply is just to give you somewhere to go if things get more complicated and processes start to block or go into deadlock.
EDIT2
As for which to use, I'd say start simple, test hard.
Go with Sorpigal's approach, but try to stress test with higher data volumes and under more difficult loads and conditionals that you'd ever expect in a live system.
In Perl, you can execute system commands using system() or `` (backticks). You can even capture the output of the command into a variable. However, this hides the program execution in the background so that the person executing your script can't see it.
Normally this is useful but sometimes I want to see what is going on behind the scenes. How do you make it so the commands executed are printed to the terminal, and those programs' output printed to the terminal? This would be the .bat equivalent of "#echo on".
I don't know of any default way to do this, but you can define a subroutine to do it for you:
sub execute {
my $cmd = shift;
print "$cmd\n";
system($cmd);
}
my $cmd = $ARGV[0];
execute($cmd);
And then see it in action:
pbook:~/foo rudd$ perl foo.pl ls
ls
file1 file2 foo.pl
As I understand, system() will print the result of the command, but not assign it. Eg.
[daniel#tux /]$ perl -e '$ls = system("ls"); print "Result: $ls\n"'
bin dev home lost+found misc net proc sbin srv System tools var
boot etc lib media mnt opt root selinux sys tmp usr
Result: 0
Backticks will capture the output of the command and not print it:
[daniel#tux /]$ perl -e '$ls = `ls`; print "Result: $ls\n"'
Result: bin
boot
dev
etc
home
lib
etc...
Update: If you want to print the name of the command being system() 'd as well, I think Rudd's approach is good. Repeated here for consolidation:
sub execute {
my $cmd = shift;
print "$cmd\n";
system($cmd);
}
my $cmd = $ARGV[0];
execute($cmd);
Use open instead. Then you can capture the output of the command.
open(LS,"|ls");
print LS;
Here's an updated execute that will print the results and return them:
sub execute {
my $cmd = shift;
print "$cmd\n";
my $ret = `$cmd`;
print $ret;
return $ret;
}
Hmm, interesting how different people are answering this different ways. It looks to me like mk and Daniel Fone interpreted it as wanting to see/manipulate the stdout of the command (neither of their solutions capture stderr fwiw). I think Rudd got closer. One twist you could make on Rudd's response is to overwite the built in system() command with your own version so that you wouldn't have to rewrite existing code to use his execute() command.
using his execute() sub from Rudd's post, you could have something like this at the top of your code:
if ($DEBUG) {
*{"CORE::GLOBAL::system"} = \&{"main::execute"};
}
I think that will work but I have to admit this is voodoo and it's been a while since I wrote this code. Here's the code I wrote years ago to intercept system calls on a local (calling namespace) or global level at module load time:
# importing into either the calling or global namespace _must_ be
# done from import(). Doing it elsewhere will not have desired results.
delete($opts{handle_system});
if ($do_system) {
if ($do_system eq 'local') {
*{"$callpkg\::system"} = \&{"$_package\::system"};
} else {
*{"CORE::GLOBAL::system"} = \&{"$_package\::system"};
}
}
Another technique to combine with the others mentioned in the answers is to use the tee command. For example:
open(F, "ls | tee /dev/tty |");
while (<F>) {
print length($_), "\n";
}
close(F);
This will both print out the files in the current directory (as a consequence of tee /dev/tty) and also print out the length of each filename read.