rt

#!/usr/bin/perl -w
# BEGIN BPS TAGGED BLOCK {{{
# 
# COPYRIGHT:
# 
# This software is Copyright (c) 1996-2009 Best Practical Solutions, LLC
#                                          <jesse@bestpractical.com>
# 
# (Except where explicitly superseded by other copyright notices)
# 
# 
# LICENSE:
# 
# This work is made available to you under the terms of Version 2 of
# the GNU General Public License. A copy of that license should have
# been provided with this software, but in any event can be snarfed
# from www.gnu.org.
# 
# This work is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.
# 
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
# 02110-1301 or visit their web page on the internet at
# http://www.gnu.org/licenses/old-licenses/gpl-2.0.html.
# 
# 
# CONTRIBUTION SUBMISSION POLICY:
# 
# (The following paragraph is not intended to limit the rights granted
# to you to modify and distribute this software under the terms of
# the GNU General Public License and is only of importance to you if
# you choose to contribute your changes and enhancements to the
# community by submitting them to Best Practical Solutions, LLC.)
# 
# By intentionally submitting any modifications, corrections or
# derivatives to this work, or any other work intended for use with
# Request Tracker, to Best Practical Solutions, LLC, you confirm that
# you are the copyright holder for those contributions and you grant
# Best Practical Solutions,  LLC a nonexclusive, worldwide, irrevocable,
# royalty-free, perpetual, license to use, copy, create derivative
# works based on those contributions, and sublicense and distribute
# those contributions and any derivatives thereof.
# 
# END BPS TAGGED BLOCK }}}
# Designed and implemented for Best Practical Solutions, LLC by
# Abhijit Menon-Sen <ams@wiw.org>

use strict;

# This program is intentionally written to have as few non-core module
# dependencies as possible. It should stay that way.

use Cwd;
use LWP;
use Text::ParseWords;
use HTTP::Request::Common;
use HTTP::Headers;
use Term::ReadLine;
use Time::Local; # used in prettyshow

# strong (GSSAPI based) authentication is supported if the server does provide
# it and the perl modules GSSAPI and LWP::Authen::Negotiate are installed
# it can be suppressed by setting externalauth=0 (default is undef)
eval { require GSSAPI };
my $no_strong_auth = 'missing perl module GSSAPI';
if ( ! $@ ) {
    eval {require LWP::Authen::Negotiate};
    $no_strong_auth = $@ ? 'missing perl module LWP::Authen::Negotiate' : 0;
}

# We derive configuration information from hardwired defaults, dotfiles,
# and the RT* environment variables (in increasing order of precedence).
# Session information is stored in ~/.rt_sessions.

my $VERSION = 0.02;
my $HOME = eval{(getpwuid($<))[7]}
           || $ENV{HOME} || $ENV{LOGDIR} || $ENV{HOMEPATH}
           || ".";
my %config = (
    (
        debug        => 0,
        user         => eval{(getpwuid($<))[0]} || $ENV{USER} || $ENV{USERNAME},
        passwd       => undef,
        server       => 'http://localhost/',
        query        => "Status!='resolved' and Status!='rejected'",
        orderby      => 'id',
        queue        => undef,
# to protect against unlimited searches a better choice would be
#       queue        => 'Unknown_Queue',
# setting externalauth => undef will try GSSAPI auth if the corresponding perl
# modules are installed, externalauth => 0 is the backward compatible choice 
        externalauth => 0,
    ),
    config_from_file($ENV{RTCONFIG} || ".rtrc"),
    config_from_env()
);
my $session = new Session("$HOME/.rt_sessions");
my $REST = "$config{server}/REST/1.0";
$no_strong_auth = 'switched off by externalauth=0'
    if defined $config{externalauth};


my $prompt = 'rt> ';

sub whine;
sub DEBUG { warn @_ if $config{debug} >= shift }

# These regexes are used by command handlers to parse arguments.
# (XXX: Ask Autrijus how i18n changes these definitions.)

my $name    = '[\w.-]+';
my $CF_name = '[\sa-z0-9_ :()/-]+';
my $field   = '(?i:[a-z][a-z0-9_-]*|C(?:ustom)?F(?:ield)?-'.$CF_name.'|CF\.\{'.$CF_name.'\})';
my $label   = '[a-zA-Z0-9@_.+-]+';
my $labels  = "(?:$label,)*$label";
my $idlist  = '(?:(?:\d+-)?\d+,)*(?:\d+-)?\d+';

# Our command line looks like this:
#
#     rt <action> [options] [arguments]
#
# We'll parse just enough of it to decide upon an action to perform, and
# leave the rest to per-action handlers to interpret appropriately.

my %handlers = (
#   handler     => [ ...aliases... ],
    version     => ["version", "ver"],
    shell       => ["shell"],
    logout      => ["logout"],
    help        => ["help", "man"],
    show        => ["show", "cat"],
    edit        => ["create", "edit", "new", "ed"],
    list        => ["search", "list", "ls"],
    comment     => ["comment", "correspond"],
    link        => ["link", "ln"],
    merge       => ["merge"],
    grant       => ["grant", "revoke"],
    take        => ["take", "steal", "untake"],
    quit        => ["quit", "exit"],
    setcommand  => ["del", "delete", "give", "res", "resolve",
                    "subject"],
);

my %actions;
foreach my $fn (keys %handlers) {
    foreach my $alias (@{ $handlers{$fn} }) {
        $actions{$alias} = \&{"$fn"};
    }
}

# Once we find and call an appropriate handler, we're done.

sub handler {
    my $action;

    push @ARGV, 'shell' if (!@ARGV);    # default to shell mode
    shift @ARGV if ($ARGV[0] eq 'rt');    # ignore a leading 'rt'
    if (@ARGV && exists $actions{$ARGV[0]}) {
        $action = shift @ARGV;
        return $actions{$action}->($action);
    }
    else {
        print STDERR "rt: Unknown command '@ARGV'.\n";
        print STDERR "rt: For help, run 'rt help'.\n";
        return 1;
    }
}

exit handler();

# Handler functions.
# ------------------
#
# The following subs are handlers for each entry in %actions.

sub shell {
    $|=1;
    my $term = new Term::ReadLine 'RT CLI';
    while ( defined ($_ = $term->readline($prompt)) ) {
        next if /^#/ || /^\s*$/;

        @ARGV = shellwords($_);
        handler();
    }
}

sub version {
    print "rt $VERSION\n";
    return 0;
}

sub logout {
    submit("$REST/logout") if defined $session->cookie;
    return 0;
}

sub quit {
    logout();
    exit;
}

my %help;
sub help {
    my ($action, $type, $rv) = @_;
    $rv = defined $rv ? $rv : 0;
    my $key;

    # What help topics do we know about?
    if (!%help) {
        local $/ = undef;
        foreach my $item (@{ Form::parse(<DATA>) }) {
            my $title = $item->[2]{Title};
            my @titles = ref $title eq 'ARRAY' ? @$title : $title;

            foreach $title (grep $_, @titles) {
                $help{$title} = $item->[2]{Text};
            }
        }
    }

    # What does the user want help with?
    undef $action if ($action && $actions{$action} eq \&help);
    unless ($action || $type) {
        # If we don't know, we'll look for clues in @ARGV.
        foreach (@ARGV) {
            if (exists $help{$_}) { $key = $_; last; }
        }
        unless ($key) {
            # Tolerate possibly plural words.
            foreach (@ARGV) {
                if ($_ =~ s/s$// && exists $help{$_}) { $key = $_; last; }
            }
        }
    }

    if ($type && $action) {
        $key = "$type.$action";
    }
    $key ||= $type || $action || "introduction";

    # Find a suitable topic to display.
    while (!exists $help{$key}) {
        if ($type && $action) {
            if ($key eq "$type.$action") { $key = $action;        }
            elsif ($key eq $action)      { $key = $type;          }
            else                         { $key = "introduction"; }
        }
        else {
            $key = "introduction";
        }
    }

    print STDERR $help{$key}, "\n\n";
    return $rv;
}

# Displays a list of objects that match some specified condition.

sub list {
    my ($q, $type, %data);
    my $orderby = $config{orderby};
    
    if ($config{orderby}) {
         $data{orderby} = $config{orderby};
    } 
    my $bad = 0;
    my $rawprint = 0;
    my $reverse_sort = 0;
    my $queue = $config{queue};

    while (@ARGV) {
        $_ = shift @ARGV;

        if (/^-t$/) {
            $bad = 1, last unless defined($type = get_type_argument());
        }
        elsif (/^-S$/) {
            $bad = 1, last unless get_var_argument(\%data);
        }
        elsif (/^-o$/) {
            $data{'orderby'} = shift @ARGV;
        }
        elsif (/^-([isl])$/) {
            $data{format} = $1;
            $rawprint = 1;
        }
        elsif (/^-q$/) {
            $queue = shift @ARGV;
        }
        elsif (/^-r$/) {
            $reverse_sort = 1;
        }
        elsif (/^-f$/) {
            if ($ARGV[0] !~ /^(?:(?:$field,)*$field)$/) {
                whine "No valid field list in '-f $ARGV[0]'.";
                $bad = 1; last;
            }
            $data{fields} = shift @ARGV;
            $data{format} = 's' if ! $data{format};
            $rawprint = 1;
        }
        elsif (!defined $q && !/^-/) {
            $q = $_;
        }
        else {
            my $datum = /^-/ ? "option" : "argument";
            whine "Unrecognised $datum '$_'.";
            $bad = 1; last;
        }
    }
    if ( ! $rawprint and ! exists $data{format} ) {
        $data{format} = 'l';
    }
    if ( $reverse_sort and $data{orderby} =~ /^-/ ) {
        $data{orderby} =~ s/^-/+/;
    } elsif ($reverse_sort) {
        $data{orderby} =~ s/^\+?(.*)/-$1/;
    }

    if (!defined $q) {
        $q = $config{query}; 
    }
    
    $q =~ s/^#//; # get rid of leading hash
    if ($q =~ /^\d+$/) {
        # only digits, must be an id, formulate a correct query
        $q = "id=$q" if $q =~ /^\d+$/;
    } else {
        # a string only, take it as an owner or requestor (quoting done later)
        $q = "(Owner=$q or Requestor like $q) and $config{query}"
             if $q =~ /^[\w\-]+$/;
        # always add a query for a specific queue or (comma separated) queues
        $queue =~ s/,/ or Queue=/g if $queue;
        $q .= " and (Queue=$queue)" if $queue and $q and $q !~ /Queue\s*=/i
            and $q !~ /id\s*=/i;
    }
    # correctly quote strings in a query
    $q =~ s/(=|like\s)\s*([^'\d\s]\S*)\b/$1\'$2\'/g;

    $type ||= "ticket";
    unless ($type && defined $q) {
        my $item = $type ? "query string" : "object type";
        whine "No $item specified.";
        $bad = 1;
    }
    #return help("list", $type) if $bad;
    return suggest_help("list", $type, $bad) if $bad;

    print "Query:$q\n" if ! $rawprint;
    my $r = submit("$REST/search/$type", { query => $q, %data });
    if ( $rawprint ) {
        print $r->content;
    } else {
        my $forms = Form::parse($r->content);
        prettylist ($forms);
    }
    return 0;
}

# Displays selected information about a single object.

sub show {
    my ($type, @objects, %data);
    my $slurped = 0;
    my $bad = 0;
    my $rawprint = 0;
    my $histspec;

    while (@ARGV) {
        $_ = shift @ARGV;
        s/^#// if /^#\d+/; # get rid of leading hash
        if (/^-t$/) {
            $bad = 1, last unless defined($type = get_type_argument());
        }
        elsif (/^-S$/) {
            $bad = 1, last unless get_var_argument(\%data);
        }
        elsif (/^-([isl])$/) {
            $data{format} = $1;
            $rawprint = 1;
        }
        elsif (/^-$/ && !$slurped) {
            chomp(my @lines = <STDIN>);
            foreach (@lines) {
                unless (is_object_spec($_, $type)) {
                    whine "Invalid object on STDIN: '$_'.";
                    $bad = 1; last;
                }
                push @objects, $_;
            }
            $slurped = 1;
        }
        elsif (/^-f$/) {
            if ($ARGV[0] !~ /^(?:(?:$field,)*$field)$/) {
                whine "No valid field list in '-f $ARGV[0]'.";
                $bad = 1; last;
            }
            $data{fields} = shift @ARGV;
            # option f requires short raw listing format
            $data{format} = 's';
            $rawprint = 1;
        }
        elsif (/^\d+$/ and my $spc2 = is_object_spec("ticket/$_", $type)) {
            push @objects, $spc2;
            $histspec = is_object_spec("ticket/$_/history", $type);
        }
        elsif (/^\d+\// and my $spc3 = is_object_spec("ticket/$_", $type)) {
            push @objects, $spc3;
            $rawprint = 1 if $_ =~ /\/content$/;
        }
        elsif (my $spec = is_object_spec($_, $type)) {
            push @objects, $spec;
            $rawprint = 1 if $_ =~ /\/content$/ or $_ !~ /^ticket/;
        }
        else {
            my $datum = /^-/ ? "option" : "argument";
            whine "Unrecognised $datum '$_'.";
            $bad = 1; last;
        }
    }
    if ( ! $rawprint ) {
        push @objects, $histspec if $histspec;
        $data{format} = 'l' if ! exists $data{format};
    }

    unless (@objects) {
        whine "No objects specified.";
        $bad = 1;
    }
    #return help("show", $type) if $bad;
    return suggest_help("show", $type, $bad) if $bad;

    my $r = submit("$REST/show", { id => \@objects, %data });
    my $c = $r->content;
    # if this isn't a text reply, remove the trailing newline so we
    # don't corrupt things like tarballs when people do
    # show ticket/id/attachments/id/content > foo.tar.gz
    if ($r->content_type !~ /^text\//) {
        chomp($c);
        $rawprint = 1;
    }
    if ( $rawprint ) {
        print $c;
    } else {
        # I do not know how to get more than one form correctly returned
        $c =~ s!^RT/[\d\.]+ 200 Ok$!--!mg;
        my $forms = Form::parse($c);
        prettyshow ($forms);
    }
    return 0;
}

# To create a new object, we ask the server for a form with the defaults
# filled in, allow the user to edit it, and send the form back.
#
# To edit an object, we must ask the server for a form representing that
# object, make changes requested by the user (either on the command line
# or interactively via $EDITOR), and send the form back.

sub edit {
    my ($action) = @_;
    my (%data, $type, @objects);
    my ($cl, $text, $edit, $input, $output);

    use vars qw(%set %add %del);
    %set = %add = %del = ();
    my $slurped = 0;
    my $bad = 0;
    
    while (@ARGV) {
        $_ = shift @ARGV;
        s/^#// if /^#\d+/; # get rid of leading hash

        if    (/^-e$/) { $edit = 1 }
        elsif (/^-i$/) { $input = 1 }
        elsif (/^-o$/) { $output = 1 }
        elsif (/^-t$/) {
            $bad = 1, last unless defined($type = get_type_argument());
        }
        elsif (/^-S$/) {
            $bad = 1, last unless get_var_argument(\%data);
        }
        elsif (/^-$/ && !($slurped || $input)) {
            chomp(my @lines = <STDIN>);
            foreach (@lines) {
                unless (is_object_spec($_, $type)) {
                    whine "Invalid object on STDIN: '$_'.";
                    $bad = 1; last;
                }
                push @objects, $_;
            }
            $slurped = 1;
        }
        elsif (/^set$/i) {
            my $vars = 0;

            while (@ARGV && $ARGV[0] =~ /^($field)([+-]?=)(.*)$/s) {
                my ($key, $op, $val) = ($1, $2, $3);
                my $hash = ($op eq '=') ? \%set : ($op =~ /^\+/) ? \%add : \%del;

                vpush($hash, lc $key, $val);
                shift @ARGV;
                $vars++;
            }
            unless ($vars) {
                whine "No variables to set.";
                $bad = 1; last;
            }
            $cl = $vars;
        }
        elsif (/^(?:add|del)$/i) {
            my $vars = 0;
            my $hash = ($_ eq "add") ? \%add : \%del;

            while (@ARGV && $ARGV[0] =~ /^($field)=(.*)$/s) {
                my ($key, $val) = ($1, $2);

                vpush($hash, lc $key, $val);
                shift @ARGV;
                $vars++;
            }
            unless ($vars) {
                whine "No variables to set.";
                $bad = 1; last;
            }
            $cl = $vars;
        }
        elsif (/^\d+$/ and my $spc2 = is_object_spec("ticket/$_", $type)) {
            push @objects, $spc2;
        }
        elsif (my $spec = is_object_spec($_, $type)) {
            push @objects, $spec;
        }
        else {
            my $datum = /^-/ ? "option" : "argument";
            whine "Unrecognised $datum '$_'.";
            $bad = 1; last;
        }
    }

    if ($action =~ /^ed(?:it)?$/) {
        unless (@objects) {
            whine "No objects specified.";
            $bad = 1;
        }
    }
    else {
        if (@objects) {
            whine "You shouldn't specify objects as arguments to $action.";
            $bad = 1;
        }
        unless ($type) {
            whine "What type of object do you want to create?";
            $bad = 1;
        }
        @objects = ("$type/new") if defined($type);
    }
    #return help($action, $type) if $bad;
    return suggest_help($action, $type, $bad) if $bad;

    # We need a form to make changes to. We usually ask the server for
    # one, but we can avoid that if we are fed one on STDIN, or if the
    # user doesn't want to edit the form by hand, and the command line
    # specifies only simple variable assignments.  We *should* get a
    # form if we're creating a new ticket, so that the default values
    # get filled in properly.

    my @new_objects = grep /\/new$/, @objects;

    if ($input) {
        local $/ = undef;
        $text = <STDIN>;
    }
    elsif ($edit || %add || %del || !$cl || @new_objects) {
        my $r = submit("$REST/show", { id => \@objects, format => 'l' });
        $text = $r->content;
    }

    # If any changes were specified on the command line, apply them.
    if ($cl) {
        if ($text) {
            # We're updating forms from the server.
            my $forms = Form::parse($text);

            foreach my $form (@$forms) {
                my ($c, $o, $k, $e) = @$form;
                my ($key, $val);

                next if ($e || !@$o);

                local %add = %add;
                local %del = %del;
                local %set = %set;

                # Make changes to existing fields.
                foreach $key (@$o) {
                    if (exists $add{lc $key}) {
                        $val = delete $add{lc $key};
                        vpush($k, $key, $val);
                        $k->{$key} = vsplit($k->{$key}) if $val =~ /[,\n]/;
                    }
                    if (exists $del{lc $key}) {
                        $val = delete $del{lc $key};
                        my %val = map {$_=>1} @{ vsplit($val) };
                        $k->{$key} = vsplit($k->{$key});
                        @{$k->{$key}} = grep {!exists $val{$_}} @{$k->{$key}};
                    }
                    if (exists $set{lc $key}) {
                        $k->{$key} = delete $set{lc $key};
                    }
                }
                
                # Then update the others.
                foreach $key (keys %set) { vpush($k, $key, $set{$key}) }
                foreach $key (keys %add) {
                    vpush($k, $key, $add{$key});
                    $k->{$key} = vsplit($k->{$key});
                }
                push @$o, (keys %add, keys %set);
            }

            $text = Form::compose($forms);
        }
        else {
            # We're rolling our own set of forms.
            my @forms;
            foreach (@objects) {
                my ($type, $ids, $args) =
                    m{^($name)/($idlist|$labels)(?:(/.*))?$}o;

                $args ||= "";
                foreach my $obj (expand_list($ids)) {
                    my %set = (%set, id => "$type/$obj$args");
                    push @forms, ["", [keys %set], \%set];
                }
            }
            $text = Form::compose(\@forms);
        }
    }

    if ($output) {
        print $text;
        return 0;
    }

    my $synerr = 0;

EDIT:
    # We'll let the user edit the form before sending it to the server,
    # unless we have enough information to submit it non-interactively.
    if ($edit || (!$input && !$cl)) {
        my $newtext = vi($text);
        # We won't resubmit a bad form unless it was changed.
        $text = ($synerr && $newtext eq $text) ? undef : $newtext;
    }

    if ($text) {
        my $r = submit("$REST/edit", {content => $text, %data});
        if ($r->code == 409) {
            # If we submitted a bad form, we'll give the user a chance
            # to correct it and resubmit.
            if ($edit || (!$input && !$cl)) {
                $text = $r->content;
                $synerr = 1;
                goto EDIT;
            }
            else {
                print $r->content;
                return 0;
            }
        }
        print $r->content;
    }
    return 0;
}

# handler for special edit commands. A valid edit command is constructed and
# further work is delegated to the edit handler

sub setcommand {
    my ($action) = @_;
    my ($id, $bad, $what);
    if ( @ARGV ) {
        $_ = shift @ARGV;
        $id = $1 if (m|^(?:ticket/)?($idlist)$|);
    }
    if ( ! $id ) {
        $bad = 1;
        whine "No ticket number specified.";
    }
    if ( @ARGV ) {
        if ($action eq 'subject') {
            my $subject = '"'.join (" ", @ARGV).'"';
            @ARGV = ();
            $what = "subject=$subject";
        } elsif ($action eq 'give') {
            my $owner = shift @ARGV;
            $what = "owner=$owner";
        }
    } else {
        if ( $action eq 'delete' or $action eq 'del' ) {
            $what = "status=deleted";
        } elsif ($action eq 'resolve' or $action eq 'res' ) {
            $what = "status=resolved";
        } elsif ($action eq 'take' ) {
            $what = "owner=$config{user}";
        } elsif ($action eq 'untake') {
            $what = "owner=Nobody";
        }
    }
    if (@ARGV) {
        $bad = 1;
        whine "Extraneous arguments for action $action: @ARGV.";
    }
    if ( ! $what ) {
        $bad = 1;
        whine "unrecognized action $action.";
    }
    return help("edit", undef, $bad) if $bad;
    @ARGV = ( $id, "set", $what );
    print "Executing: rt edit @ARGV\n";
    return edit("edit");
}

# We roll "comment" and "correspond" into the same handler.

sub comment {
    my ($action) = @_;
    my (%data, $id, @files, @bcc, @cc, $msg, $wtime, $edit);
    my $bad = 0;

    while (@ARGV) {
        $_ = shift @ARGV;

        if (/^-e$/) {
            $edit = 1;
        }
        elsif (/^-[abcmw]$/) {
            unless (@ARGV) {
                whine "No argument specified with $_.";
                $bad = 1; last;
            }

            if (/-a/) {
                unless (-f $ARGV[0] && -r $ARGV[0]) {
                    whine "Cannot read attachment: '$ARGV[0]'.";
                    return 0;
                }
                push @files, shift @ARGV;
            }
            elsif (/-([bc])/) {
                my $a = $_ eq "-b" ? \@bcc : \@cc;
                @$a = split /\s*,\s*/, shift @ARGV;
            }
            elsif (/-m/) {
                $msg = shift @ARGV;
                if ( $msg =~ /^-$/ ) {
                    undef $msg;
                    while (<STDIN>) { $msg .= $_ }
                }
            }

            elsif (/-w/) { $wtime = shift @ARGV }
        }
        elsif (!$id && m|^(?:ticket/)?($idlist)$|) {
            $id = $1;
        }
        else {
            my $datum = /^-/ ? "option" : "argument";
            whine "Unrecognised $datum '$_'.";
            $bad = 1; last;
        }
    }

    unless ($id) {
        whine "No object specified.";
        $bad = 1;
    }
    #return help($action, "ticket") if $bad;
    return suggest_help($action, "ticket") if $bad;

    my $form = [
        "",
        [ "Ticket", "Action", "Cc", "Bcc", "Attachment", "TimeWorked", "Text" ],
        {
            Ticket     => $id,
            Action     => $action,
            Cc         => [ @cc ],
            Bcc        => [ @bcc ],
            Attachment => [ @files ],
            TimeWorked => $wtime || '',
            Text       => $msg || '',
            Status => ''
        }
    ];

    my $text = Form::compose([ $form ]);

    if ($edit || !$msg) {
        my $error = 0;
        my ($c, $o, $k, $e);

        do {
            my $ntext = vi($text);
            return if ($error && $ntext eq $text);
            $text = $ntext;
            $form = Form::parse($text);
            $error = 0;

            ($c, $o, $k, $e) = @{ $form->[0] };
            if ($e) {
                $error = 1;
                $c = "# Syntax error.";
                goto NEXT;
            }
            elsif (!@$o) {
                return 0;
            }
            @files = @{ vsplit($k->{Attachment}) };

        NEXT:
            $text = Form::compose([[$c, $o, $k, $e]]);
        } while ($error);
    }

    my $i = 1;
    foreach my $file (@files) {
        $data{"attachment_$i"} = bless([ $file ], "Attachment");
        $i++;
    }
    $data{content} = $text;

    my $r = submit("$REST/ticket/$id/comment", \%data);
    print $r->content;
    return 0;
}

# Merge one ticket into another.

sub merge {
    my @id;
    my $bad = 0;

    while (@ARGV) {
        $_ = shift @ARGV;
        s/^#// if /^#\d+/; # get rid of leading hash

        if (/^\d+$/) {
            push @id, $_;
        }
        else {
            whine "Unrecognised argument: '$_'.";
            $bad = 1; last;
        }
    }

    unless (@id == 2) {
        my $evil = @id > 2 ? "many" : "few";
        whine "Too $evil arguments specified.";
        $bad = 1;
    }
    #return help("merge", "ticket") if $bad;
    return suggest_help("merge", "ticket", $bad) if $bad;

    my $r = submit("$REST/ticket/$id[0]/merge/$id[1]");
    print $r->content;
    return 0;
}

# Link one ticket to another.

sub link {
    my ($bad, $del, %data) = (0, 0, ());
    my $type;

    my %ltypes = map { lc $_ => $_ } qw(DependsOn DependedOnBy RefersTo
                                        ReferredToBy HasMember MemberOf);

    while (@ARGV && $ARGV[0] =~ /^-/) {
        $_ = shift @ARGV;

        if (/^-d$/) {
            $del = 1;
        }
        elsif (/^-t$/) {
            $bad = 1, last unless defined($type = get_type_argument());
        }
        else {
            whine "Unrecognised option: '$_'.";
            $bad = 1; last;
        }
    }
    
    $type = "ticket" unless $type; # default type to tickets
    
    if (@ARGV == 3) {
        my ($from, $rel, $to) = @ARGV;
        if ($from !~ /^\d+$/ || $to !~ /^\d+$/) {
            my $bad = $from =~ /^\d+$/ ? $to : $from;
            whine "Invalid $type ID '$bad' specified.";
            $bad = 1;
        }
        if (($type eq "ticket") && ( ! exists $ltypes{lc $rel})) {
            whine "Invalid link '$rel' for type $type specified.";
            $bad = 1;
        }
        %data = (id => $from, rel => $rel, to => $to, del => $del);
    }
    else {
        my $bad = @ARGV < 3 ? "few" : "many";
        whine "Too $bad arguments specified.";
        $bad = 1;
    }
    return suggest_help("link", $type, $bad) if $bad;
 
    my $r = submit("$REST/$type/link", \%data);
    print $r->content;
    return 0;
}

# Take/steal a ticket
sub take {
    my ($cmd) = @_;
    my ($bad, %data) = (0, ());

    my $id;

    # get the ticket id
    if (@ARGV == 1) {
        ($id) = @ARGV;
        unless ($id =~ /^\d+$/) {
            whine "Invalid ticket ID $id specified.";
            $bad = 1;
        }
        my $form = [
            "",
            [ "Ticket", "Action" ],
            {
                Ticket => $id,
                Action => $cmd,
                Status => '',
            }
        ];

        my $text = Form::compose([ $form ]);
        $data{content} = $text;
    }
    else {
        $bad = @ARGV < 1 ? "few" : "many";
        whine "Too $bad arguments specified.";
        $bad = 1;
    }
    return suggest_help("take", "ticket", $bad) if $bad;

    my $r = submit("$REST/ticket/$id/take", \%data);
    print $r->content;
    return 0;
}

# Grant/revoke a user's rights.

sub grant {
    my ($cmd) = @_;

    my $revoke = 0;
    while (@ARGV) {
    }

    $revoke = 1 if $cmd->{action} eq 'revoke';
    return 0;
}

# Client <-> Server communication.
# --------------------------------
#
# This function composes and sends an HTTP request to the RT server, and
# interprets the response. It takes a request URI, and optional request
# data (a string, or a reference to a set of key-value pairs).

sub submit {
    my ($uri, $content) = @_;
    my ($req, $data);
    my $ua = new LWP::UserAgent(agent => "RT/3.0b", env_proxy => 1);
    my $h = HTTP::Headers->new;

    # Did the caller specify any data to send with the request?
    $data = [];
    if (defined $content) {
        unless (ref $content) {
            # If it's just a string, make sure LWP handles it properly.
            # (By pretending that it's a file!)
            $content = [ content => [undef, "", Content => $content] ];
        }
        elsif (ref $content eq 'HASH') {
            my @data;
            foreach my $k (keys %$content) {
                if (ref $content->{$k} eq 'ARRAY') {
                    foreach my $v (@{ $content->{$k} }) {
                        push @data, $k, $v;
                    }
                }
                else { push @data, $k, $content->{$k} }
            }
            $content = \@data;
        }
        $data = $content;
    }

    # Should we send authentication information to start a new session?
    my $how = $config{server} =~ /^https/ ? 'over SSL' : 'unencrypted';
    (my $server = $config{server}) =~ s/^.*\/\/([^\/]+)\/?/$1/;
    if ($config{externalauth}) {
        $h->authorization_basic($config{user}, $config{passwd} || read_passwd() );
        print "   Password will be sent to $server $how\n",
              "   Press CTRL-C now if you do not want to continue\n"
            if ! $config{passwd};
    } elsif ( $no_strong_auth ) {
        if (!defined $session->cookie) {
            print "   Strong encryption not available, $no_strong_auth\n",
                  "   Password will be sent to $server $how\n",
                  "   Press CTRL-C now if you do not want to continue\n"
                if ! $config{passwd};
            push @$data, ( user => $config{user} );
            push @$data, ( pass => $config{passwd} || read_passwd() );
        }
    }

    # Now, we construct the request.
    if (@$data) {
        $req = POST($uri, $data, Content_Type => 'form-data');
    }
    else {
        $req = GET($uri);
    }
    $session->add_cookie_header($req);
    if ($config{externalauth}) {
        $req->header(%$h);
    }

    # Then we send the request and parse the response.
    DEBUG(3, $req->as_string);
    my $res = $ua->request($req);
    DEBUG(3, $res->as_string);

    if ($res->is_success) {
        # The content of the response we get from the RT server consists
        # of an HTTP-like status line followed by optional header lines,
        # a blank line, and arbitrary text.

        my ($head, $text) = split /\n\n/, $res->content, 2;
        my ($status, @headers) = split /\n/, $head;
        $text =~ s/\n*$/\n/ if ($text);

        # "RT/3.0.1 401 Credentials required"
        if ($status !~ m#^RT/\d+(?:\S+) (\d+) ([\w\s]+)$#) {
            warn "rt: Malformed RT response from $config{server}.\n";
            warn "(Rerun with RTDEBUG=3 for details.)\n" if $config{debug} < 3;
            exit -1;
        }

        # Our caller can pretend that the server returned a custom HTTP
        # response code and message. (Doing that directly is apparently
        # not sufficiently portable and uncomplicated.)
        $res->code($1);
        $res->message($2);
        $res->content($text);
        $session->update($res) if ($res->is_success || $res->code != 401);

        if (!$res->is_success) {
            # We can deal with authentication failures ourselves. Either
            # we sent invalid credentials, or our session has expired.
            if ($res->code == 401) {
                my %d = @$data;
                if (exists $d{user}) {
                    warn "rt: Incorrect username or password.\n";
                    exit -1;
                }
                elsif ($req->header("Cookie")) {
                    # We'll retry the request with credentials, unless
                    # we only wanted to logout in the first place.
                    $session->delete;
                    return submit(@_) unless $uri eq "$REST/logout";
                }
            }
            # Conflicts should be dealt with by the handler and user.
            # For anything else, we just die.
            elsif ($res->code != 409) {
                warn "rt: ", $res->content;
                #exit;
            }
        }
    }
    else {
        warn "rt: Server error: ", $res->message, " (", $res->code, ")\n";
        exit -1;
    }

    return $res;
}

# Session management.
# -------------------
#
# Maintains a list of active sessions in the ~/.rt_sessions file.
{
    package Session;
    my ($s, $u);

    # Initialises the session cache.
    sub new {
        my ($class, $file) = @_;
        my $self = {
            file => $file || "$HOME/.rt_sessions",
            sids => { }
        };
       
        # The current session is identified by the currently configured
        # server and user.
        ($s, $u) = @config{"server", "user"};

        bless $self, $class;
        $self->load();

        return $self;
    }

    # Returns the current session cookie.
    sub cookie {
        my ($self) = @_;
        my $cookie = $self->{sids}{$s}{$u};
        return defined $cookie ? "RT_SID_$cookie" : undef;
    }

    # Deletes the current session cookie.
    sub delete {
        my ($self) = @_;
        delete $self->{sids}{$s}{$u};
    }

    # Adds a Cookie header to an outgoing HTTP request.
    sub add_cookie_header {
        my ($self, $request) = @_;
        my $cookie = $self->cookie();

        $request->header(Cookie => $cookie) if defined $cookie;
    }

    # Extracts the Set-Cookie header from an HTTP response, and updates
    # session information accordingly.
    sub update {
        my ($self, $response) = @_;
        my $cookie = $response->header("Set-Cookie");

        if (defined $cookie && $cookie =~ /^RT_SID_(.[^;,\s]+=[0-9A-Fa-f]+);/) {
            $self->{sids}{$s}{$u} = $1;
        }
    }

    # Loads the session cache from the specified file.
    sub load {
        my ($self, $file) = @_;
        $file ||= $self->{file};
        local *F;

        open(F, $file) && do {
            $self->{file} = $file;
            my $sids = $self->{sids} = {};
            while (<F>) {
                chomp;
                next if /^$/ || /^#/;
                next unless m#^https?://[^ ]+ \w+ [^;,\s]+=[0-9A-Fa-f]+$#;
                my ($server, $user, $cookie) = split / /, $_;
                $sids->{$server}{$user} = $cookie;
            }
            return 1;
        };
        return 0;
    }

    # Writes the current session cache to the specified file.
    sub save {
        my ($self, $file) = shift;
        $file ||= $self->{file};
        local *F;

        open(F, ">$file") && do {
            my $sids = $self->{sids};
            foreach my $server (keys %$sids) {
                foreach my $user (keys %{ $sids->{$server} }) {
                    my $sid = $sids->{$server}{$user};
                    if (defined $sid) {
                        print F "$server $user $sid\n";
                    }
                }
            }
            close(F);
            chmod 0600, $file;
            return 1;
        };
        return 0;
    }

    sub DESTROY {
        my $self = shift;
        $self->save;
    }
}

# Form handling.
# --------------
#
# Forms are RFC822-style sets of (field, value) specifications with some
# initial comments and interspersed blank lines allowed for convenience.
# Sets of forms are separated by --\n (in a cheap parody of MIME).
#
# Each form is parsed into an array with four elements: commented text
# at the start of the form, an array with the order of keys, a hash with
# key/value pairs, and optional error text if the form syntax was wrong.

# Returns a reference to an array of parsed forms.
sub Form::parse {
    my $state = 0;
    my @forms = ();
    my @lines = split /\n/, $_[0] if $_[0];
    my ($c, $o, $k, $e) = ("", [], {}, "");

    LINE:
    while (@lines) {
        my $line = shift @lines;

        next LINE if $line eq '';

        if ($line eq '--') {
            # We reached the end of one form. We'll ignore it if it was
            # empty, and store it otherwise, errors and all.
            if ($e || $c || @$o) {
                push @forms, [ $c, $o, $k, $e ];
                $c = ""; $o = []; $k = {}; $e = "";
            }
            $state = 0;
        }
        elsif ($state != -1) {
            if ($state == 0 && $line =~ /^#/) {
                # Read an optional block of comments (only) at the start
                # of the form.
                $state = 1;
                $c = $line;
                while (@lines && $lines[0] =~ /^#/) {
                    $c .= "\n".shift @lines;
                }
                $c .= "\n";
            }
            elsif ($state <= 1 && $line =~ /^($field):(?:\s+(.*))?$/) {
                # Read a field: value specification.
                my $f  = $1;
                my @v  = ($2 || ());

                # Read continuation lines, if any.
                while (@lines && ($lines[0] eq '' || $lines[0] =~ /^\s+/)) {
                    push @v, shift @lines;
                }
                pop @v while (@v && $v[-1] eq '');

                # Strip longest common leading indent from text.
                my $ws = "";
                foreach my $ls (map {/^(\s+)/} @v[1..$#v]) {
                    $ws = $ls if (!$ws || length($ls) < length($ws));
                }
                s/^$ws// foreach @v;

                push(@$o, $f) unless exists $k->{$f};
                vpush($k, $f, join("\n", @v));

                $state = 1;
            }
            elsif ($line !~ /^#/) {
                # We've found a syntax error, so we'll reconstruct the
                # form parsed thus far, and add an error marker. (>>)
                $state = -1;
                $e = Form::compose([[ "", $o, $k, "" ]]);
                $e.= $line =~ /^>>/ ? "$line\n" : ">> $line\n";
            }
        }
        else {
            # We saw a syntax error earlier, so we'll accumulate the
            # contents of this form until the end.
            $e .= "$line\n";
        }
    }
    push(@forms, [ $c, $o, $k, $e ]) if ($e || $c || @$o);

    foreach my $l (keys %$k) {
        $k->{$l} = vsplit($k->{$l}) if (ref $k->{$l} eq 'ARRAY');
    }

    return \@forms;
}

# Returns text representing a set of forms.
sub Form::compose {
    my ($forms) = @_;
    my @text;

    foreach my $form (@$forms) {
        my ($c, $o, $k, $e) = @$form;
        my $text = "";

        if ($c) {
            $c =~ s/\n*$/\n/;
            $text = "$c\n";
        }
        if ($e) {
            $text .= $e;
        }
        elsif ($o) {
            my @lines;

            foreach my $key (@$o) {
                my ($line, $sp);
                my $v = $k->{$key};
                my @values = ref $v eq 'ARRAY' ? @$v : $v;

                $sp = " "x(length("$key: "));
                $sp = " "x4 if length($sp) > 16;

                foreach $v (@values) {
                    if ($v =~ /\n/) {
                        $v =~ s/^/$sp/gm;
                        $v =~ s/^$sp//;

                        if ($line) {
                            push @lines, "$line\n\n";
                            $line = "";
                        }
                        elsif (@lines && $lines[-1] !~ /\n\n$/) {
                            $lines[-1] .= "\n";
                        }
                        push @lines, "$key: $v\n\n";
                    }
                    elsif ($line &&
                           length($line)+length($v)-rindex($line, "\n") >= 70)
                    {
                        $line .= ",\n$sp$v";
                    }
                    else {
                        $line = $line ? "$line, $v" : "$key: $v";
                    }
                }

                $line = "$key:" unless @values;
                if ($line) {
                    if ($line =~ /\n/) {
                        if (@lines && $lines[-1] !~ /\n\n$/) {
                            $lines[-1] .= "\n";
                        }
                        $line .= "\n";
                    }
                    push @lines, "$line\n";
                }
            }

            $text .= join "", @lines;
        }
        else {
            chomp $text;
        }
        push @text, $text;
    }

    return join "\n--\n\n", @text;
}

# Configuration.
# --------------

# Returns configuration information from the environment.
sub config_from_env {
    my %env;

    foreach my $k (qw(EXTERNALAUTH DEBUG USER PASSWD SERVER QUERY ORDERBY)) {

        if (exists $ENV{"RT$k"}) {
            $env{lc $k} = $ENV{"RT$k"};
        }
    }

    return %env;
}

# Finds a suitable configuration file and returns information from it.
sub config_from_file {
    my ($rc) = @_;

    if ($rc =~ m#^/#) {
        # We'll use an absolute path if we were given one.
        return parse_config_file($rc);
    }
    else {
        # Otherwise we'll use the first file we can find in the current
        # directory, or in one of its (increasingly distant) ancestors.

        my @dirs = split /\//, cwd;
        while (@dirs) {
            my $file = join('/', @dirs, $rc);
            if (-r $file) {
                return parse_config_file($file);
            }

            # Remove the last directory component each time.
            pop @dirs;
        }

        # Still nothing? We'll fall back to some likely defaults.
        for ("$HOME/$rc", "/etc/rt.conf") {
            return parse_config_file($_) if (-r $_);
        }
    }

    return ();
}

# Makes a hash of the specified configuration file.
sub parse_config_file {
    my %cfg;
    my ($file) = @_;
    local $_; # $_ may be aliased to a constant, from line 1163

    open(CFG, $file) && do {
        while (<CFG>) {
            chomp;
            next if (/^#/ || /^\s*$/);

            if (/^(externalauth|user|passwd|server|query|orderby|queue)\s+(.*)\s?$/) {
                $cfg{$1} = $2;
            }
            else {
                die "rt: $file:$.: unknown configuration directive.\n";
            }
        }
    };

    return %cfg;
}

# Helper functions.
# -----------------

sub whine {
    my $sub = (caller(1))[3];
    $sub =~ s/^main:://;
    warn "rt: $sub: @_\n";
    return 0;
}

sub read_passwd {
    eval 'require Term::ReadKey';
    if ($@) {
        die "No password specified (and Term::ReadKey not installed).\n";
    }

    print "Password: ";
    Term::ReadKey::ReadMode('noecho');
    chomp(my $passwd = Term::ReadKey::ReadLine(0));
    Term::ReadKey::ReadMode('restore');
    print "\n";

    return $passwd;
}

sub vi {
    my ($text) = @_;
    my $file = "/tmp/rt.form.$$";
    my $editor = $ENV{EDITOR} || $ENV{VISUAL} || "vi";

    local *F;
    local $/ = undef;

    open(F, ">$file") || die "$file: $!\n"; print F $text; close(F);
    system($editor, $file) && die "Couldn't run $editor.\n";
    open(F, $file) || die "$file: $!\n"; $text = <F>; close(F);
    unlink($file);

    return $text;
}

# Add a value to a (possibly multi-valued) hash key.
sub vpush {
    my ($hash, $key, $val) = @_;
    my @val = ref $val eq 'ARRAY' ? @$val : $val;

    if (exists $hash->{$key}) {
        unless (ref $hash->{$key} eq 'ARRAY') {
            my @v = $hash->{$key} ne '' ? $hash->{$key} : ();
            $hash->{$key} = \@v;
        }
        push @{ $hash->{$key} }, @val;
    }
    else {
        $hash->{$key} = $val;
    }
}

# "Normalise" a hash key that's known to be multi-valued.
sub vsplit {
    my ($val) = @_;
    my ($word, @words);
    my @values = ref $val eq 'ARRAY' ? @$val : $val;

    foreach my $line (map {split /\n/} @values) {
        # XXX: This should become a real parser, à la Text::ParseWords.
        $line =~ s/^\s+//;
        $line =~ s/\s+$//;
        my ( $a, $b ) = split /,/, $line, 2;

        while ($a) {
            no warnings 'uninitialized';
            if ( $a =~ /^'/ ) {
                my $s = $a;
                while ( $a !~ /'$/ || (   $a !~ /(\\\\)+'$/
                            && $a =~ /(\\)+'$/ )) {
                    ( $a, $b ) = split /,/, $b, 2;
                    $s .= ',' . $a;
                }
                push @words, $s;
            }
            elsif ( $a =~ /^q{/ ) {
                my $s = $a;
                while ( $a !~ /}$/ ) {
                    ( $a, $b ) =
                      split /,/, $b, 2;
                    $s .= ',' . $a;
                }
                $s =~ s/^q{/'/;
                $s =~ s/}/'/;
                push @words, $s;
            }
            else {
                push @words, $a;
            }
            ( $a, $b ) = split /,/, $b, 2;
        }


    }

    return \@words;
}

# WARN: this code is duplicated in lib/RT/Interface/REST.pm
# change both functions at once
sub expand_list {
    my ($list) = @_;

    my @elts;
    foreach (split /,/, $list) {
        push @elts, /^(\d+)-(\d+)$/? ($1..$2): $_;
    }

    return map $_->[0], # schwartzian transform
        sort {
            defined $a->[1] && defined $b->[1]?
                # both numbers
                $a->[1] <=> $b->[1]
                :!defined $a->[1] && !defined $b->[1]?
                    # both letters
                    $a->[2] cmp $b->[2]
                    # mix, number must be first
                    :defined $a->[1]? -1: 1
        }
        map [ $_, (defined( /^(\d+)$/ )? $1: undef), lc($_) ],
        @elts;
}

sub get_type_argument {
    my $type;

    if (@ARGV) {
        $type = shift @ARGV;
        unless ($type =~ /^[A-Za-z0-9_.-]+$/) {
            # We want whine to mention our caller, not us.
            @_ = ("Invalid type '$type' specified.");
            goto &whine;
        }
    }
    else {
        @_ = ("No type argument specified with -t.");
        goto &whine;
    }

    $type =~ s/s$//; # "Plural". Ugh.
    return $type;
}

sub get_var_argument {
    my ($data) = @_;

    if (@ARGV) {
        my $kv = shift @ARGV;
        if (my ($k, $v) = $kv =~ /^($field)=(.*)$/) {
            push @{ $data->{$k} }, $v;
        }
        else {
            @_ = ("Invalid variable specification: '$kv'.");
            goto &whine;
        }
    }
    else {
        @_ = ("No variable argument specified with -S.");
        goto &whine;
    }
}

sub is_object_spec {
    my ($spec, $type) = @_;

    $spec =~ s|^(?:$type/)?|$type/| if defined $type;
    return $spec if ($spec =~ m{^$name/(?:$idlist|$labels)(?:/.*)?$}o);
    return 0;
}

sub suggest_help {
    my ($action, $type, $rv) = @_;

    print STDERR "rt: For help, run 'rt help $action'.\n" if defined $action;
    print STDERR "rt: For help, run 'rt help $type'.\n" if defined $type;
    return $rv;
}

sub str2time {
    # simplified procedure for parsing date, avoid loading Date::Parse
    my %month = (Jan => 0, Feb => 1, Mar => 2, Apr => 3, May =>  4, Jun =>  5,
                 Jul => 6, Aug => 7, Sep => 8, Oct => 9, Nov => 10, Dec => 11);
    $_ = shift;
    my ($mon, $day, $hr, $min, $sec, $yr, $monstr);
    if ( /(\w{3})\s+(\d\d?)\s+(\d\d):(\d\d):(\d\d)\s+(\d{4})/ ) {
        ($monstr, $day, $hr, $min, $sec, $yr) = ($1, $2, $3, $4, $5, $6);
        $mon = $month{$monstr} if exists $month{$monstr};
    } elsif ( /(\d{4})-(\d\d)-(\d\d)\s+(\d\d):(\d\d):(\d\d)/ ) {
        ($yr, $mon, $day, $hr, $min, $sec) = ($1, $2-1, $3, $4, $5, $6);
    }
    if ( $yr and defined $mon and $day and defined $hr and defined $sec ) {
        return timelocal($sec,$min,$hr,$day,$mon,$yr);
    } else {
        print "Unknown date format in parsedate: $_\n";
        return undef;
    }
}

sub date_diff {
    my ($old, $new) = @_;
    $new = time() if ! $new;
    $old = str2time($old) if $old !~ /^\d+$/;
    $new = str2time($new) if $new !~ /^\d+$/;
    return "???" if ! $old or ! $new;

    my %seconds = (min => 60,
                   hr  => 60*60,
                   day => 60*60*24,
                   wk  => 60*60*24*7,
                   mth => 60*60*24*30,
                   yr  => 60*60*24*365);

    my $diff = $new - $old;
    my $what = 'sec';
    my $howmuch = $diff;
    for ( sort {$seconds{$a} <=> $seconds{$b}} keys %seconds) {
        last if $diff < $seconds{$_};
        $what = $_;
        $howmuch = int($diff/$seconds{$_});
    }
    return "$howmuch $what";
}

sub prettyshow {
    my $forms = shift;
    my ($form) = grep { exists $_->[2]->{Queue} } @$forms;
    my $k = $form->[2];
    # dates are in local time zone
    if ( $k ) {
        print "Date: $k->{Created}\n";
        print "From: $k->{Requestors}\n";
        print "Cc: $k->{Cc}\n" if $k->{Cc};
        print "X-AdminCc: $k->{AdminCc}\n" if $k->{AdminCc};
        print "X-Queue: $k->{Queue}\n";
        print "Subject: [rt #$k->{id}] $k->{Subject}\n\n";
    }
    # dates in these attributes are in GMT and will be converted
    foreach my $form (@$forms) {
        my ($c, $o, $k, $e) = @$form;
        next if ! $k->{id} or exists $k->{Queue};
        if ( exists $k->{Created} ) {
            my ($y,$m,$d,$hh,$mm,$ss) = ($k->{Created} =~ /(\d\d\d\d)-(\d\d)-(\d\d) (\d\d):(\d\d):(\d\d)/);
            $m--;
            my $created = localtime(timegm($ss,$mm,$hh,$d,$m,$y));
            if ( exists $k->{Description} ) {
                print "===> $k->{Description} on $created\n";
            }
        }
        print "$k->{Content}\n" if exists $k->{Content} and
                                   $k->{Content} !~ /to have no content$/ and
                                   $k->{Type} ne 'EmailRecord';
        print "$k->{Attachments}\n" if exists $k->{Attachments} and
                                   $k->{Attachments};
    }
}

sub prettylist {
    my $forms = shift;
    my $heading = "Ticket Owner Queue    Age   Told Status Requestor Subject\n";
    $heading .= '-' x 80 . "\n";
    my (@open, @me);
    foreach my $form (@$forms) {
        my ($c, $o, $k, $e) = @$form;
        next if ! $k->{id};
        print $heading if $heading;
        $heading = '';
        my $id = $k->{id};
        $id =~ s!^ticket/!!;
        my $owner = $k->{Owner} eq 'Nobody' ? '' : $k->{Owner};
        $owner = substr($owner, 0, 5);
        my $queue = substr($k->{Queue}, 0, 5);
        my $subject = substr($k->{Subject}, 0, 30);
        my $age = date_diff($k->{Created});
        my $told = $k->{Told} eq 'Not set' ? '' : date_diff($k->{Told});
        my $status = substr($k->{Status}, 0, 6);
        my $requestor = substr($k->{Requestors}, 0, 9);
	$requestor =~ s#\@.*$##;
        my $line = sprintf "%6s %5s %5s %6s %6s %-6s %-15s %-50s\n",
            $id, $owner, $queue, $age, $told, $status, $requestor, $subject;
        if ( $k->{Owner} eq 'Nobody' ) {
            push @open, $line;
        } elsif ($k->{Owner} eq $config{user} ) {
            push @me, $line;
        } else {
            print $line;
        }
    }
    print "No matches found\n" if $heading;
    printf "========== my %2d open tickets ==========\n", scalar @me if @me;
    print @me if @me;
    printf "========== %2d unowned tickets ==========\n", scalar @open if @open;
    print @open if @open;
}

__DATA__

Title: intro
Title: introduction
Text:

    This is a command-line interface to RT 3.0 or newer.

    It allows you to interact with an RT server over HTTP, and offers an
    interface to RT's functionality that is better-suited to automation
    and integration with other tools.

    In general, each invocation of this program should specify an action
    to perform on one or more objects, and any other arguments required
    to complete the desired action.

    For more information:

        - rt help usage         (syntax information)
        - rt help objects       (how to specify objects)
        - rt help actions       (a list of possible actions)
        - rt help types         (a list of object types)

        - rt help config        (configuration details)
        - rt help examples      (a few useful examples)
        - rt help topics        (a list of help topics)

--

Title: usage
Title: syntax
Text:

    Syntax:

        rt <action> [options] [arguments]
      or
        rt shell

    Each invocation of this program must specify an action (e.g. "edit",
    "create"), options to modify behaviour, and other arguments required
    by the specified action. (For example, most actions expect a list of
    numeric object IDs to act upon.)

    The details of the syntax and arguments for each action are given by
    "rt help <action>". Some actions may be referred to by more than one
    name ("create" is the same as "new", for example).  

    You may also call "rt shell", which will give you an 'rt>' prompt at
    which you can issue commands of the form "<action> [options] 
    [arguments]".  See "rt help shell" for details.

    Objects are identified by a type and an ID (which can be a name or a
    number, depending on the type). For some actions, the object type is
    implied (you can only comment on tickets); for others, the user must
    specify it explicitly. See "rt help objects" for details.

    In syntax descriptions, mandatory arguments that must be replaced by
    appropriate value are enclosed in <>, and optional arguments are
    indicated by [] (for example, <action> and [options] above).

    For more information:

        - rt help objects       (how to specify objects)
        - rt help actions       (a list of actions)
        - rt help types         (a list of object types)
        - rt help shell         (how to use the shell)

--

Title: conf
Title: config
Title: configuration
Text:

    This program has two major sources of configuration information: its
    configuration files, and the environment.

    The program looks for configuration directives in a file named .rtrc
    (or $RTCONFIG; see below) in the current directory, and then in more
    distant ancestors, until it reaches /. If no suitable configuration
    files are found, it will also check for ~/.rtrc and /etc/rt.conf.

    Configuration directives:

        The following directives may occur, one per line:

        - server <URL>          URL to RT server.
        - user <username>       RT username.
        - passwd <passwd>       RT user's password.
        - query <RT Query>      Default RT Query for list action
        - orderby <order>       Default RT order for list action
        - queue <queuename>     Default RT Queue for list action
        - externalauth <0|1>    Use HTTP Basic authentication
         explicitely setting externalauth to 0 inhibits also GSSAPI based
         authentication, if LWP::Authen::Negotiate (and GSSAPI) is installed

        Blank and #-commented lines are ignored.

    Sample configuration file contents:

         server  https://rt.somewhere.com/
         # more than one queue can be given (by adding a query expression)
         queue helpdesk or queue=support
         query Status != resolved and Owner=myaccount


    Environment variables:

        The following environment variables override any corresponding
        values defined in configuration files:

        - RTUSER
        - RTPASSWD
        - RTEXTERNALAUTH
        - RTSERVER
        - RTDEBUG       Numeric debug level. (Set to 3 for full logs.)
        - RTCONFIG      Specifies a name other than ".rtrc" for the
                        configuration file.
        - RTQUERY       Default RT Query for rt list
        - RTORDERBY     Default order for rt list

--

Title: objects
Text:

    Syntax:

        <type>/<id>[/<attributes>]

    Every object in RT has a type (e.g. "ticket", "queue") and a numeric
    ID. Some types of objects can also be identified by name (like users
    and queues). Furthermore, objects may have named attributes (such as
    "ticket/1/history").

    An object specification is like a path in a virtual filesystem, with
    object types as top-level directories, object IDs as subdirectories,
    and named attributes as further subdirectories.

    A comma-separated list of names, numeric IDs, or numeric ranges can
    be used to specify more than one object of the same type. Note that
    the list must be a single argument (i.e., no spaces). For example,
    "user/root,1-3,5,7-10,ams" is a list of ten users; the same list
    can also be written as "user/ams,root,1,2,3,5,7,8-10".
    
    If just a number is given as object specification it will be
    interpreted as ticket/<number>

    Examples:

        1                   # the same as ticket/1
        ticket/1
        ticket/1/attachments
        ticket/1/attachments/3
        ticket/1/attachments/3/content
        ticket/1-3/links
        ticket/1-3,5-7/history

        user/ams
        user/ams/rights
        user/ams,rai,1/rights

    For more information:

        - rt help <action>      (action-specific details)
        - rt help <type>        (type-specific details)

--

Title: actions
Title: commands
Text:

    You can currently perform the following actions on all objects:

        - list          (list objects matching some condition)
        - show          (display object details)
        - edit          (edit object details)
        - create        (create a new object)

    Each type may define actions specific to itself; these are listed in
    the help item about that type.

    For more information:

        - rt help <action>      (action-specific details)
        - rt help types         (a list of possible types)

    The following actions on tickets are also possible:

        - comment       Add comments to a ticket
        - correspond    Add comments to a ticket
        - merge         Merge one ticket into another
        - link          Link one ticket to another
        - take          Take a ticket (steal and untake are possible as well)

    For several edit set subcommands that are frequently used abbreviations
    have been introduced. These abbreviations are:

        - delete or del  delete a ticket           (edit set status=deleted)
        - resolve or res resolve a ticket          (edit set status=resolved)
        - subject        change subject of ticket  (edit set subject=string)
        - give           give a ticket to somebody (edit set owner=user)

--

Title: types
Text:

    You can currently operate on the following types of objects:

        - tickets
        - users
        - groups
        - queues

    For more information:

        - rt help <type>        (type-specific details)
        - rt help objects       (how to specify objects)
        - rt help actions       (a list of possible actions)

--

Title: ticket
Text:

    Tickets are identified by a numeric ID.

    The following generic operations may be performed upon tickets:

        - list
        - show
        - edit
        - create

    In addition, the following ticket-specific actions exist:

        - link
        - merge
        - comment
        - correspond
        - take
        - steal
        - untake
        - give
        - resolve
        - delete
        - subject

    Attributes:

        The following attributes can be used with "rt show" or "rt edit"
        to retrieve or edit other information associated with tickets:

        links                      A ticket's relationships with others.
        history                    All of a ticket's transactions.
        history/type/<type>        Only a particular type of transaction.
        history/id/<id>            Only the transaction of the specified id.
        attachments                A list of attachments.
        attachments/<id>           The metadata for an individual attachment.
        attachments/<id>/content   The content of an individual attachment.

--

Title: user
Title: group
Text:

    Users and groups are identified by name or numeric ID.

    The following generic operations may be performed upon them:

        - list
        - show
        - edit
        - create

    In addition, the following type-specific actions exist:

        - grant
        - revoke

    Attributes:

        The following attributes can be used with "rt show" or "rt edit"
        to retrieve or edit other information associated with users and
        groups:

        rights                  Global rights granted to this user.
        rights/<queue>          Queue rights for this user.

--

Title: queue
Text:

    Queues are identified by name or numeric ID.

    Currently, they can be subjected to the following actions:

        - show
        - edit
        - create

--

Title: subject
Text:

    Syntax:

        rt subject <id> <new subject text>

    Change the subject of a ticket whose ticket id is given.

--

Title: give
Text:

    Syntax:

        rt give <id> <accountname>

    Give a ticket whose ticket id is given to another user.

--

Title: steal
Text:

        rt steal <id> 

    Steal a ticket whose ticket id is given, i.e. set the owner to myself.

--

Title: take
Text:

    Syntax:

        rt take <id>

    Take a ticket whose ticket id is given, i.e. set the owner to myself.

--

Title: untake
Text:

    Syntax:

        rt untake <id>

    Untake a ticket whose ticket id is given, i.e. set the owner to Nobody.

--

Title: resolve
Title: res
Text:

    Syntax:

        rt resolve <id>

    Resolves a ticket whose ticket id is given.

--

Title: delete
Title: del
Text:

    Syntax:

        rt delete <id>

    Deletes a ticket whose ticket id is given.

--

Title: logout
Text:

    Syntax:

        rt logout

    Terminates the currently established login session. You will need to
    provide authentication credentials before you can continue using the
    server. (See "rt help config" for details about authentication.)

--

Title: ls
Title: list
Title: search
Text:

    Syntax:

        rt <ls|list|search> [options] "query string"

    Displays a list of objects matching the specified conditions.
    ("ls", "list", and "search" are synonyms.)

    Conditions are expressed in the SQL-like syntax used internally by
    RT3. (For more information, see "rt help query".) The query string
    must be supplied as one argument.

    (Right now, the server doesn't support listing anything but tickets.
    Other types will be supported in future; this client will be able to
    take advantage of that support without any changes.)

    Options:

        The following options control how much information is displayed
        about each matching object:

        -i             Numeric IDs only. (Useful for |rt edit -; see examples.)
        -s             Short description.
        -l             Longer description.
        -f <field[s]   Display only the fields listed and the ticket id

        In addition,
        
        -o +/-<field>  Orders the returned list by the specified field.
        -r             reversed order (useful if a default was given)
        -q queue[s]    restricts the query to the queue[s] given
                       multiple queues are separated by comma
        -S var=val     Submits the specified variable with the request.
        -t type        Specifies the type of object to look for. (The
                       default is "ticket".)

    Examples:

        rt ls "Priority > 5 and Status=new"
        rt ls -o +Subject "Priority > 5 and Status=new"
        rt ls -o -Created "Priority > 5 and Status=new"
        rt ls -i "Priority > 5"|rt edit - set status=resolved
        rt ls -t ticket "Subject like '[PATCH]%'"
        rt ls -q systems
        rt ls -f owner,subject

--

Title: show
Text:

    Syntax:

        rt show [options] <object-ids>

    Displays details of the specified objects.

    For some types, object information is further classified into named
    attributes (for example, "1-3/links" is a valid ticket specification
    that refers to the links for tickets 1-3). Consult "rt help <type>"
    and "rt help objects" for further details.

    If only a number is given it will be interpreted as the objects
    ticket/number and ticket/number/history

    This command writes a set of forms representing the requested object
    data to STDOUT.

    Options:

        The following options control how much information is displayed
        about each matching object:

        Without any formatting options prettyprinted output is generated.
        Giving any of the two options below reverts to raw output.
        -s      Short description (history and attachments only).
        -l      Longer description (history and attachments only).

        In addition,
        -               Read IDs from STDIN instead of the command-line.
        -t type         Specifies object type.
        -f a,b,c        Restrict the display to the specified fields.
        -S var=val      Submits the specified variable with the request.

    Examples:

        rt show -t ticket -f id,subject,status 1-3
        rt show ticket/3/attachments/29
        rt show ticket/3/attachments/29/content
        rt show ticket/1-3/links
        rt show ticket/3/history
        rt show -l ticket/3/history
        rt show -t user 2
        rt show 2

--

Title: new
Title: edit
Title: create
Text:

    Syntax:

        rt edit [options] <object-ids> set field=value [field=value] ...
                                       add field=value [field=value] ...
                                       del field=value [field=value] ...

    Edits information corresponding to the specified objects.

    A purely numeric object id nnn is translated into ticket/nnn

    If, instead of "edit", an action of "new" or "create" is specified,
    then a new object is created. In this case, no numeric object IDs
    may be specified, but the syntax and behaviour remain otherwise
    unchanged.

    This command typically starts an editor to allow you to edit object
    data in a form for submission. If you specified enough information
    on the command-line, however, it will make the submission directly.

    The command line may specify field-values in three different ways.
    "set" sets the named field to the given value, "add" adds a value
    to a multi-valued field, and "del" deletes the corresponding value.
    Each "field=value" specification must be given as a single argument.

    For some types, object information is further classified into named
    attributes (for example, "1-3/links" is a valid ticket specification
    that refers to the links for tickets 1-3). These attributes may also
    be edited. Consult "rt help <type>" and "rt help object" for further
    details.

    Options:

        -       Read numeric IDs from STDIN instead of the command-line.
                (Useful with rt ls ... | rt edit -; see examples below.)
        -i      Read a completed form from STDIN before submitting.
        -o      Dump the completed form to STDOUT instead of submitting.
        -e      Allows you to edit the form even if the command-line has
                enough information to make a submission directly.
        -S var=val
                Submits the specified variable with the request.
        -t type Specifies object type.

    Examples:

        # Interactive (starts $EDITOR with a form).
        rt edit ticket/3
        rt create -t ticket

        # Non-interactive.
        rt edit ticket/1-3 add cc=foo@example.com set priority=3 due=tomorrow
        rt ls -t tickets -i 'Priority > 5' | rt edit - set status=resolved
        rt edit ticket/4 set priority=3 owner=bar@example.com \
                         add cc=foo@example.com bcc=quux@example.net
        rt create -t ticket set subject='new ticket' priority=10 \
                            add cc=foo@example.com

--

Title: comment
Title: correspond
Text:

    Syntax:

        rt <comment|correspond> [options] <ticket-id>

    Adds a comment (or correspondence) to the specified ticket (the only
    difference being that comments aren't sent to the requestors.)

    This command will typically start an editor and allow you to type a
    comment into a form. If, however, you specified all the necessary
    information on the command line, it submits the comment directly.

    (See "rt help forms" for more information about forms.)

    Options:

        -m <text>       Specify comment text.
        -a <file>       Attach a file to the comment. (May be used more
                        than once to attach multiple files.)
        -c <addrs>      A comma-separated list of Cc addresses.
        -b <addrs>      A comma-separated list of Bcc addresses.
        -w <time>       Specify the time spent working on this ticket.
        -e              Starts an editor before the submission, even if
                        arguments from the command line were sufficient.

    Examples:

        rt comment -m 'Not worth fixing.' -a stddisclaimer.h 23

--

Title: merge
Text:

    Syntax:

        rt merge <from-id> <to-id>

    Merges the first ticket specified into the second ticket specified.

--

Title: link
Text:

    Syntax:

        rt link [-d] <id-A> <link> <id-B>

    Creates (or, with -d, deletes) a link between the specified tickets.
    The link can (irrespective of case) be any of:

        DependsOn/DependedOnBy:     A depends upon B (or vice versa).
        RefersTo/ReferredToBy:      A refers to B (or vice versa).
        MemberOf/HasMember:         A is a member of B (or vice versa).

    To view a ticket's links, use "rt show ticket/3/links". (See
    "rt help ticket" and "rt help show".)

    Options:

        -d      Deletes the specified link.

    Examples:

        rt link 2 dependson 3
        rt link -d 4 referredtoby 6     # 6 no longer refers to 4

--

Title: grant
Title: revoke
Text:

--

Title: query
Text:

    RT3 uses an SQL-like syntax to specify object selection constraints.
    See the <RT:...> documentation for details.
    
    (XXX: I'm going to have to write it, aren't I?)

    Until it exists here a short description of important constructs:

    The two simple forms of query expressions are the constructs
    Attribute like Value and
    Attribute = Value or Attribute != Value

    Whether attributes can be matched using like or using = is built into RT.
    The attributes id, Queue, Owner Priority and Status require the = or !=
    tests.

    If Value is a string it must be quoted and may contain the wildcard
    character %. If the string does not contain white space, the quoting
    may however be omitted, it will be added automatically when parsing
    the input.

    Simple query expressions can be combined using and, or and parentheses
    can be used to group expressions.

    As a special case a standalone string (which would not form a correct
    query) is transformed into (Owner='string' or Requestor like 'string%')
    and added to the default query, i.e. the query is narrowed down.

    If no Queue=name clause is contained in the query, a default clause
    Queue=$config{queue} is added.

    Examples:
    Status!='resolved' and Status!='rejected'
    (Owner='myaccount' or Requestor like 'myaccount%') and Status!='resolved'

--

Title: form
Title: forms
Text:

    This program uses RFC822 header-style forms to represent object data
    in a form that's suitable for processing both by humans and scripts.

    A form is a set of (field, value) specifications, with some initial
    commented text and interspersed blank lines allowed for convenience.
    Field names may appear more than once in a form; a comma-separated
    list of multiple field values may also be specified directly.
    
    Field values can be wrapped as in RFC822, with leading whitespace.
    The longest sequence of leading whitespace common to all the lines
    is removed (preserving further indentation). There is no limit on
    the length of a value.

    Multiple forms are separated by a line containing only "--\n".

    (XXX: A more detailed specification will be provided soon. For now,
    the server-side syntax checking will suffice.)

--

Title: topics
Text:

    Syntax:

        rt help <topic>

    Get help on any of the following subjects:

        - tickets, users, groups, queues.
        - show, edit, ls/list/search, new/create.

        - query                                 (search query syntax)
        - forms                                 (form specification)

        - objects                               (how to specify objects)
        - types                                 (a list of object types)
        - actions/commands                      (a list of actions)
        - usage/syntax                          (syntax details)
        - conf/config/configuration             (configuration details)
        - examples                              (a few useful examples)

--

Title: example
Title: examples
Text:

    some useful examples

    All the following list requests will be restricted to the default queue.
    That can be changed by adding the option -q queuename

    List all tickets that are not rejected/resolved
        rt ls
    List all tickets that are new and do not have an owner
        rt ls "status=new and owner=nobody"
    List all tickets which I have sent or of which I am the owner
        rt ls myaccount
    List all attributes for the ticket 6977 (ls -l instead of ls)
        rt ls -l 6977
    Show the content of ticket 6977
        rt show 6977
    Show all attributes in the ticket and in the history of the ticket
        rt show -l 6977
    Comment a ticket (mail is sent to all queue watchers, i.e. AdminCc's)
        rt comment 6977
        This will open an editor and lets you add text (attribute Text:)
        Other attributes may be changed as well, but usually don't do that.
    Correspond a ticket (like comment, but mail is also sent to requestors)
        rt correspond 6977
    Edit a ticket (generic change, interactive using the editor)
        rt edit 6977
    Change the owner of a ticket non interactively
        rt edit 6977 set owner=myaccount
        or
        rt give 6977 account
        or
        rt take 6977
    Change the status of a ticket
        rt edit 6977 set status=resolved
        or
        rt resolve 6977
    Change the status of all tickets I own to resolved !!!
        rt ls -i owner=myaccount | rt edit - set status=resolved

--

Title: shell
Text:

    Syntax:

        rt shell

    Opens an interactive shell, at which you can issue commands of 
    the form "<action> [options] [arguments]".

    To exit the shell, type "quit" or "exit".

    Commands can be given at the shell in the same form as they would 
    be given at the command line without the leading 'rt' invocation.

    Example:
        $ rt shell
        rt> create -t ticket set subject='new' add cc=foo@example.com
        # Ticket 8 created.
        rt> quit
        $

--

Title: take
Title: untake
Title: steal
Text:

    Syntax:

        rt <take|untake|steal> <ticket-id>

    Sets the owner of the specified ticket to the current user, 
    assuming said user has the bits to do so, or releases the 
    ticket.  
    
    'Take' is used on tickets which are not currently owned 
    (Owner: Nobody), 'steal' is used on tickets which *are* 
    currently owned, and 'untake' is used to "release" a ticket 
    (reset its Owner to Nobody).  'Take' cannot be used on
    tickets which are currently owned.

    Example:
        alice$ rt create -t ticket set subject="New ticket"
        # Ticket 7 created.
        alice$ rt take 7
        # Owner changed from Nobody to alice
        alice$ su bob
        bob$ rt steal 7
        # Owner changed from alice to bob
        bob$ rt untake 7
        # Owner changed from bob to Nobody

--

Title: quit
Title: exit
Text:

    Use "quit" or "exit" to leave the shell.  Only valid within shell 
    mode.

    Example:
        $ rt shell
        rt> quit
        $