I'm trying to make a Jabber bot and I am having trouble keeping it running while waiting for messages. How do I get my script to continuously run? I have tried calling a subroutine that has a while loop that I, in theory, have set up to check for any messages and react accordingly but my script isn't behaving that way.
Here is my source: http://pastebin.com/03Habbvh
# set jabber bot callbacks
$jabberBot->SetMessageCallBacks(chat=>\&chat);
$jabberBot->SetPresenceCallBacks(available=>\&welcome,unavailable=>\&killBot);
$jabberBot->SetCallBacks(receive=>\&prnt,iq=>\&gotIQ);
$jabberBot->PresenceSend(type=>"available");
$jabberBot->Process(1);
sub welcome
{
print "Welcome!\n";
$jabberBot->MessageSend(to=>$jbrBoss->GetJID(),subject=>"",body=>"Hello There!",type=>"chat",priority=>10);
&keepItGoing;
}
sub prnt
{
print $_[1]."\n";
}
#$jabberBot->MessageSend(to=>$jbrBoss->GetJID(),subject=>"",body=>"Hello There! Global...",type=>"chat",priority=>10);
#$jabberBot->Process(5);
#&keepItGoing;
sub chat
{
my ($sessionID,$msg) = #_;
$dump->pl2xml($msg);
if($msg->GetType() ne 'get' && $msg->GetType() ne 'set' && $msg->GetType() ne '')
{
my $jbrCmd = &trimSpaces($msg->GetBody());
my $dbQry = $dbh->prepare("SELECT command,acknowledgement FROM commands WHERE message = '".lc($jbrCmd)."'");
$dbQry->execute();
if($dbQry->rows() > 0 && $jbrCmd !~ /^insert/si)
{
my $ref = $dbQry->fetchrow_hashref();
$dbQry->finish();
$jabberBot->MessageSend(to=>$msg->GetFrom(),subject=>"",body=>$ref->{'acknowledgement'},type=>"chat",priority=>10);
eval $ref->{'command'};
&keepItGoing;
}
else
{
$jabberBot->MessageSend(to=>$msg->GetFrom(),subject=>"",body=>"I didn't understand you!",type=>"chat",priority=>10);
$dbQry->finish();
&keepItGoing;
}
}
}
sub gotIQ
{
print "iq\n";
}
sub trimSpaces
{
my $string = $_[0];
$string =~ s/^\s+//; #remove leading spaces
$string =~ s/\s+$//; #remove trailing spaces
return $string;
}
sub keepItGoing
{
print "keepItGoing!\n";
my $proc = $jabberBot->Process(1);
while(defined($proc) && $proc != 1)
{
$proc = $jabberBot->Process(1);
}
}
sub killBot
{
print "killing\n";
$jabberBot->MessageSend(to=>$_[0]->GetFrom(),subject=>"",body=>"Logging Out!",type=>"chat",priority=>10);
$jabberBot->Process(1);
$jabberBot->Disconnect();
exit;
}
POE has some pretty good event frameworks. I don't know how good the one for Jabber (POE::Component::Jabber) is, but it's probably worth looking at.
AnyEvent::XMPP is ridiculously comprehensive, and, since it uses AnyEvent, can be run in any event driven application with a supported loop (AnyEvent's own, Event, EV, Tk, Glib/Gtk, even POE).
I think you can make your example work by doing this:
0 while $jabber->Process
Having said that, I would strongly recommend using a proper Event handling framework such as AnyEvent (my personal favorite) or POE (the traditional choice).
Related
I am using the following lib Net::SSH2
I can connect to my device and get the output OK in most case. Below is the relevant code:
sub logIntoDevice
{
my $self = shift;
my ($ssh2) = #_;
if(! $ssh2->connect($self->deviceIP))
{
say "Failed to connect to device:",$self->deviceIP;
$ssh2->disconnect();
exit 0;
}
if(! $ssh2->auth_password($self->usr, $self->pass))
{
say "Authentication Fail to ",$self->deviceIP;
$ssh2->disconnect();
exit 0;
}
my $channel = $ssh2->channel();
$channel->blocking(0);
$channel->shell();
return $channel;
}
sub sendCmd
{
my $self = shift;
my ($cmd,$channel) = #_;
my #cmdOutput;
print $channel "$cmd\n";
while (<$channel>)
{
chomp $_;
push(#cmdOutput, $_);
}
return #cmdOutput;
}
So below are the cmd i sent to the sub's. They work fine and the output is write to file OK.
$self->writeToFile($fileName,$self->sendCmd("show clock",$channel));
$self->writeToFile($fileName,$self->sendCmd("\n",$channel));
$self->writeToFile($fileName,$self->sendCmd("dir",$channel));
with the exception of when i sent the following cmd:
$self->writeToFile($fileName,$self->sendCmd("sh run",$channel));
the output of the cmd on the device using putty is:
sh run
Building configuration...
Current configuration : 16575 bytes
!
!
!Last configuration change at 16:37:19 CET+1 Sat Mar 15 2014
.....
but in the log file all you see is
sh run
Building configuration...
so the issue is the blank lines after the Building configuration output make the while (<$channel>) think its the end of the output.
My issue is I cant figure a way to loop through the data without using a While loop.
UPDATE
Ok come up with this solution but seems very clunky. must be a better way if doing this
sub sendCmd
{
my $self = shift;
my ($cmd,$channel) = #_;
my #cmdOutput;
my $currPrompt;
#get prompt. i am sure there is a better way!!! just cant figure it out
print $channel "\n";
while (<$channel>)
{
$currPrompt = $_;
}
print $channel "$cmd\n";
while(42)
{
my $inerOutput;
while (<$channel>)
{
chomp $_;
$inerOutput = $_;
push(#cmdOutput, $_);
}
if($inerOutput ne $currPrompt)
{
sleep(7);
}
else
{
last;
}
}
return #cmdOutput;
}
I don't think your issue is blank lines. Most likely, the issue is that you are using non-blocking mode, and it takes time for the device to perform the command. So you are getting an empty line (or an undef) after "Building configuration..." is read because there is no extra output produced yet.
I would use Net::SSH2's poll method with a timeout, which will let you know when there is something to read. If "sh run" takes substantially longer than other commands you are issuing, your sendCmd method needs to be aware of this, and allow more time to pass before it decides no more output is coming its way.
Alternatively, you can (as is the custom when using, for example, Net::Telnet) wait for more output until you see the prompt, whatever the prompt is for the device in question,
and then you will know that the command has finished its execution.
Net::SSH2->poll is deprecated as result of libss2_poll deprecation
In our company we were using this code (given at the end) for about 10 years and it worked fine.
Some days ago we faced some issues and we had to re-code the complete package, we decided to replace this code with Switch module by Damian (in order to improve the readability of code).
Everything is working fine for us.
Later I found on Perlmonks that Damian had put this module under
Damian modules you shouldn't use in production because their purpose
is to explore and prototype future core language features.
But it is working fine for us because we are not hitting the limitations of this module (I guess).
Now I ask you guys to please have a look at the both implementations (nested if else vs switch) and let me know whether using Switch in the newer implementation is fine or are we creating some future problems for us? Is using Switch in the code given below fine or are there any hidden bugs/problems?
I've already read the bugs and reviews of this module on CPAN and Perlmonks and I guess our code is far away from hitting those bugs (I think so).
We are using Perl 5.8.5.
PS: I know the alternatives of Switch, we have given/when in Perl 5.10, we can use dispatch table and other solutions which are specified here, but right now we just want to compare the new implementation which uses Switch.
Using nested if else
if ($command =~ /^enter$/) {
$self->show_main_frames();
}
elsif ($command =~ /^XYZ_MENU/i) {
$self->show_main_menu($manual, $dbot);
}
elsif ($command =~ /^DBOT/i) {
$dbot->process();
}
# XML is used for the reminders-history: Request 2666
elsif ($command =~ /^XML_DBOT/i) {
$dbot->process();
}
elsif ($command =~ /^UGS/i) {
$ugsui->process();
}
elsif ($command eq "kill") {
my $login = $self->{COMMON_HASH}{login} || "";
my $su_login = $self->{CONF}->get("start", "SU_LOGIN");
if ($login eq $su_login) {
# usually only certain user with certain permission will be
# able to do this.
$self->do_error("Daemon was killed by ".$login);
$self->db_connection->disconnect();
$self->{LOG}->write("User $login killed the daemon", 0);
exit; # this 'exit' actually kill the daemon
}
else {
$self->do_error("User $login tried to kill the daemon. ".
"This incident will be reported");
$self->{LOG}->write("User $login tried to kill the daemon", 2);
}
}
elsif ($command eq "logout") {
# check if we should delete the password cookie
my $forget_me = $self->{CGI}->param("forget_me") || 0;
if ($forget_me) {
$self->{DB_PASSWORD_COOKIE}->delete_cookie();
}
$ugsui->do_logout();
# Cliff edit remove id from logged_in
$session->remove_session($session->login());
# delete the session of the user
delete $self->{SESSIONS}{$session->id()};
if ($self->{CACHE_TO_FILE}) {
my $session_data_path =
XYZ_DIR
."/code/cache/session_data"
.$session->id();
unlink($session_data_path);
}
}
# if we just login we should create all the main frames
elsif ($command eq "login") {
# if extra_param holds "command*XXX" the XXX will be placed instead of
# the command. extra_param holds pairs that are astrics-separated
my $extra_param = $cgi->param("extra_param");
$extra_param = "" if (!defined($extra_param));
$extra_param =~ /command\*([^\*]+)/i;
my $other_command = defined($1) ? $1 : "";
if ($other_command =~ /^dbot/i) { # meanwhile - works only on dbot
# commands
$command = $other_command;
# now we will get the other parameters from the extra_param
# (actually including the command that is still in the
# $extra_param)
while ($extra_param =~ /^\*?([^\*]+)\*([^\*]+)(.*)/) {
$extra_param = $3;
my $name = $1;
my $value = $2;
$cgi->param(-name => $name,
-value => $value);
}#end while
}#end if
else{
$self->show_main_frames();
}
}#end elsif
else {
$self->show_main_frames();
}#end outer else
Using Switch
switch ($command)
{
case /^enter$/ { $self->show_main_frames() }
case /^XYZ_MENU/i { $self->show_main_menu($manual, $dbot) }
case /^DBOT/i { $dbot->process() }
case /^XML_DBOT/i { $dbot->process() }
case /^UGS/i { $ugsui->process() }
case "kill" {
my $login = $self->{COMMON_HASH}{login} || "";
my $su_login = $self->{CONF}->get("start", "SU_LOGIN");
if ($login eq $su_login) {
# usually only certain user with certain permission will be
# able to do this.
$self->do_error("Daemon was killed by ".$login);
$self->db_connection->disconnect();
$self->{LOG}->write("User $login killed the daemon", 0);
exit; # this 'exit' actually kill the daemon
}
else {
$self->do_error("User $login tried to kill the daemon. ".
"This incident will be reported");
$self->{LOG}->write("User $login tried to kill the daemon", 2);
}
}
case "logout" {
# check if we should delete the password cookie
my $forget_me = $self->{CGI}->param("forget_me") || 0;
if ($forget_me) {
$self->{DB_PASSWORD_COOKIE}->delete_cookie();
}
$ugsui->do_logout();
# Cliff edit remove id from logged_in
$session->remove_session($session->login());
# delete the session of the user
delete $self->{SESSIONS}{$session->id()};
if ($self->{CACHE_TO_FILE}) {
my $session_data_path =
XYZ_DIR
."/code/cache/session_data"
.$session->id();
unlink($session_data_path);
}
}
case "login" {
# if extra_param holds "command*XXX" the XXX will be placed instead of
# the command. extra_param holds pairs that are astrics-separated
my $extra_param = $cgi->param("extra_param");
$extra_param = "" if (!defined($extra_param));
$extra_param =~ /command\*([^\*]+)/i;
my $other_command = defined($1) ? $1 : "";
if ($other_command =~ /^dbot/i)
{ # meanwhile - works only on dbot
# commands
$command = $other_command;
# now we will get the other parameters from the extra_param
# (actually including the command that is still in the
# $extra_param)
while ($extra_param =~ /^\*?([^\*]+)\*([^\*]+)(.*)/) {
$extra_param = $3;
my $name = $1;
my $value = $2;
$cgi->param(-name => $name,
-value => $value);
}#end while
}#end if
else {$self->show_main_frames();}
}
else {$self->show_main_frames();}
} # end switch
Switch does its own parsing of the source code. This can lead to hard to diagnose errors in the code that directly uses it. The kind of problems Switch creates are not intermittent, so if your code works, you have nothing to worry about.
But really, it doesn't add much at all.
With Switch:
switch ($command) {
case /^enter$/ { $self->show_main_frames() }
case /^XYZ_MENU/i { $self->show_main_menu($manual, $dbot) }
case /^DBOT/i { $dbot->process() }
case /^XML_DBOT/i { $dbot->process() }
case /^UGS/i { $ugsui->process() }
case "kill" {
my $login = $self->{COMMON_HASH}{login} || "";
Without Switch:
for ($command) {
if (/^enter$/) { $self->show_main_frames() }
elsif (/^XYZ_MENU/i) { $self->show_main_menu($manual, $dbot) }
elsif (/^DBOT/i) { $dbot->process() }
elsif (/^XML_DBOT/i) { $dbot->process() }
elsif (/^UGS/i) { $ugsui->process() }
elsif ($_ eq "kill") {
my $login = $self->{COMMON_HASH}{login} || "";
(elsif (/^kill\z/) would also work.)
Actually Switch module does not provide you any "killer feature"; the same can be done with elsif statement which is secure, stable and does not have drawbacks that Switch does. Here is problems with Switch i got in my project (and i dont use it anymore):
Switch is made throgh Perl filters. This technique have following limits:
Your source code actually rewritten on-the-fly and replaces with
sequent elsif statements.
Some Perl error reports will refer wrong line; some of them showing code you dont have in your source (autogenerated code).
Not filter limit, but limit of module itself:
If the file(.pl or .pm) where you call use Swtich excess 1Mbyte size this can lead to "mysterious errors" (as written in doc). I can confirm these errors do not leading to Switch module and is completely unobivious, so you can have hard debug time after some weeks of coding/documentation.
I recommend to use elsif or given..when statements which is available since Perl 5.10. So if you using perl 5.8.x - use elsif.
Also you can read "Limitations" paragraph for Switch documentation.
Because Switch does own source code parsing, it does not work at all in certain circumstances. For example, it is impossible to use it with mod_perl.
However, if you have Perl 5.10 or later, there is much better replacement with effectively the same functionality: given/when
use v5.10;
given ($var) {
when (/^abc/) { $abc = 1 }
when (/^def/) { $def = 1 }
when (/^xyz/) { $xyz = 1 }
default { $nothing = 1 }
}
given is supported by Perl core (and works everywhere, including mod_perl) - you just use v5.10; and it is instantly available to you.
I need to establish this single task with Irssi Perl script. I have my own channel and I want to sent msg directly to that channel in certain scenarios.
My experience with Perl is quite limited so I haven't got this one. I am confused how to manage different chatnets and channels in Irssi Perl scripting. So how I can send message for example channel #testchan#Quakenet for example?
Test one:
server->command("^MSG $info{'#testchan'} $info{'Test message.'}");
Test two (tuto about scripting):
sub away_describe_pub_channels {
my($net, $channel) = #_;
my ($text) = #_;
my $c = Irssi::server_find_chatnet("QuakeNet")->channel_find("testchan");
$c->command("DESCRIBE $channel $text")
}
here is an example is used for a bot :)
#==========================BEGINNING OF PARMS======================================
#name of the channels where this feature will be used
my #channels = ("foo","bar");
#the public commands
#help
my $cmd_help = '!help';
#new ticket
my $cmd_newticket = "!stack";
my %url_newticket = ( 'foo'=>{url=>"http://stackoverflow.com/questions/ask"},
'bar'=>{url=>"http://https://github.com/repo/project/issues/new"}
sub bootstrap {
my ($server, $msg, $nick, $address, $target) = #_;
#lowercase of the channel name in case this one will be registered in camelCase ;)
$target = lc $target;
foreach my $channel (#channels) {
if ( $target eq "#".$channel) {
#split the line first peace the command second the rest
my ($cmd,$line) = split / /,$msg,2;
if ($cmd =~ $cmd_help) {
$server->command("MSG ". $nick ." Here are the available commands : !stack");
} elsif ($cmd eq $cmd_newticket) {
my $h = $url_newticket{$channel};
$server->command("MSG $target submit an issue/a ticket $h->{'url'}");
}
}
}
}
#let's add the sub as a signal and let's play
Irssi::signal_add_last('message public', 'bootstrap');
Hope this could help
Assuming a handle created with the following code:
use IO::File;
my $fh = IO::File->new;
my $pid = $fh->open('some_long_running_proc |') or die $!;
$fh->autoflush(1);
$fh->blocking(0);
and then read with a loop like this:
while (some_condition_here) {
my #lines = $fh->getlines;
...
sleep 1;
}
What do I put as some_condition_here that will return false if the process on the other end of the pipe has terminated?
Testing for $fh->eof will not work since the process could still be running without printing any new lines. Testing for $fh->opened doesn't seem to do anything useful.
Currently I am using $pid =! waitpid($pid, WNOHANG) which seems to work in POSIX compliant environments. Is this the best way? What about on Windows?
On using select,
use strict;
use warnings;
use IO::Select qw( );
sub process_msg {
my ($client, $msg) = #_;
chomp $msg;
print "$client->{id} said '$msg'\n";
return 1; # Return false to close the handle.
}
my $select = IO::Select->new();
my %clients;
for (...) {
my $fh = ...;
$clients{fileno($fh)} = {
id => '...'
buf => '',
# ...
};
$select->add($fh);
}
while (my #ready = $select->can_read) {
for my $fh (#ready) {
my $client = $clients{ fileno($fh) };
our $buf; local *buf = \( $client->{buf} );
my $rv = sysread($fh, $buf, 64*1024, length($buf));
if (!$rv) {
if (defined($rv)) {
print "[$client->{id} ended]\n";
} else {
print "[Error reading from $client->{id}: $!]\n";
}
print "[Incomplete message received from $client->{id}]\n"
if length($buf);
delete $clients{ fileno($fh) };
$select->remove($fh);
next;
}
while ($buf =~ s/^(.*\n)//) {
if (!process_msg($client, "$1")) {
print "[Dropping $client->{id}]\n";
delete $clients{ fileno($fh) };
$select->remove($fh);
last;
}
}
}
}
What's wrong with waiting for an actual EOF?
while (<$fh>) {
...
sleep 1;
}
You've set the handle for non-blocking reads, so it should just do the right thing. Indeed, given your example, you don't even need to set non-blocking and can get rid of the sleep.
Are there other things that you want to do while waiting on some_long_running_proc? If so, select is probably in your future.
There a number of options.
readline aka <$fh> will return false on eof (or error).
eof will return true on eof.
read (with block size > 0) will return defined and zero on eof.
sysread (with block size > 0) will return defined and zero on eof.
You can use select or make the handle non-blocking before any of the above to check without blocking.
You use select() to ascertain whether there is any data, or an exceptional condition such as a close.
Personally I prefer to use IO::Multiplex, especially where you're multiplexing input from several different descriptors, but that may not apply in this case.
I'm using the File::Find module to traverse a directory tree. Once I find a specific file, I want to stop searching. How can I do that?
find (\$processFile, $mydir);
sub processFile() {
if ($_ =~ /target/) {
# How can I return from find here?
}
}
Seems like you will have to die:
eval {
find (\$processFile, $mydir);
};
if ( $# ) {
if ( $# =~ m/^found it/ ) {
# be happy
}
else ( $# ) {
die $#;
}
}
else {
# be sad
}
sub processFile() {
if ($_ =~ /target/) {
die 'found it';
}
}
In addition to what everyone else said, you may wish to take a look at File-Find-Object, which is both iterative (and as such capable of being interrupted in the middle) and capable of instantiation (so you can initiate and use several at once, or instantiate an F-F-O object based while performing another scan, etc.)
The downside for it is that it isn't core, but it only has Class::Accessor as a dependency, and is pure-Perl so it shouldn't be hard to install.
I should warn you that I am its maintainer, so I may be a bit biased.
Can you throw custom exceptions in Perl?
You could use named blocks and jump to it if you find your result (with next, last, it depends from what you need).
I found this link:
http://www.perlmonks.org/index.pl?node_id=171367
I copied one of the scripts in that list of posts, and this seems to work:
#! /usr/bin/perl -w
use strict;
use File::Find;
my #hits = ();
my $hit_lim = shift || 20;
find(
sub {
if( scalar #hits >= $hit_lim ) {
$File::Find::prune = 1;
return;
}
elsif( -d $_ ) {
return;
}
push #hits, $File::Find::name;
},
shift || '.'
);
$, = "\n";
print #hits, "\n";
It appears that is actually causing find to not traverse any more by using $File::Find::prune.
The function processFile() should return true if it finds the file, and false otherwise. So, every time that processFile calls himself should check this return value. If it is true, some recursive call has found the file, so there's no need to call himself again, and it must also return true. If it's false, the file hasn't been found yet, and it should continue the search.