Alvis::Treetagger - Perl module providing FIFO interface to Treetagger

     $tagginglines = &Alvis::Treetagger::tag($linetotag);

    &Alvis::Treetagger::reopen();

    &Alvis::Treetagger::shut();

    $tagginglines = &Alvis::Treetagger::tag($linetotag);

#!/usr/bin/perl -w
package Alvis::Treetagger;

#  must hardwire the root directory for Treetagger below
#  error handling somewhat haphazard, and no auto restart
#  of treetagger if it shutsdown

$Alvis::Treetagger::VERSION = '0.1';

use strict;
use warnings;
use encoding 'utf8';
use open ':utf8';
binmode STDERR, ":utf8";

our $commandLine;
our $taggerRoot;
our $FIFO;
our $parFile;
our $errorFile;
our $verbose;

my $tagpid;

INIT {

  #   note that to get tagger to work without buffer hangs, 
  #   we need to stuff an extra two ".\n" at the end of each input;
  #   this requires we can eliminate the extra output as it comes;
  #   so don't fiddle with this format unless you
  #   fix up the output filtering too
  $commandLine = " -token -lemma -sgml";

  $taggerRoot = "/usr/local/treetagger";
  $FIFO = "/tmp/treetagger_$$";
  $parFile = "english.par";
  $verbose = 0;
  
}   #INIT

sub reopen() {
  
  #  paranoid step
  &Alvis::Treetagger::shut();

  if ( system('mkfifo',"$FIFO.in") || system('mkfifo',"$FIFO.out")  ) { 
    print STDERR "Cannot create pipes $FIFO.in or $FIFO.in\n";
    exit(1);
  }
  
  $tagpid = fork();
  if ( !$tagpid ) {
    #child
    if ( $verbose ) { print STDERR "Starting treetagger\n"; }
    my $execline = 
	"$taggerRoot/bin/tree-tagger $taggerRoot/lib/$parFile $commandLine"
         . " < $FIFO.in > $FIFO.out";
    if ( defined($errorFile) ) {
	$execline .= " 2>$errorFile ";
    }
    exec($execline);
  }
  sleep(1);
  
  if ( $verbose ) { print STDERR "Starting fifo input $FIFO.in\n"; }
  open(FI,">$FIFO.in");
  select((select(FI), $| = 1)[0]);

  #  send some stuff down the pipe to warm it up;
  #  note we add to extra "\n." two evey message
  #  to force TreeTagger to run the buffer
  print FI "Start\nup\nnow\n.\n.\n.\n";
  
  if ( $verbose ) { print STDERR "Starting fifo output $FIFO.out\n"; }
  open(FO,"<$FIFO.out");
  
  $_=<FO>; $_.=<FO>; $_.=<FO>; $_.=<FO>;
}

sub shut() {
  #  closing the input file for tagger will
  #   also shut it down gracefully
  close(FI);
  close(FO);
  unlink("$FIFO.in");
  unlink("$FIFO.out");
  $tagpid = undef;
  if ( $verbose ) { print STDERR "Shutdown fifos\n"; }
}

#  returns undef on a failure and a full reopen()
#  should be done
#  input will be carriage return delimited, as
#  Treetagger requires;
#  should be properly tokenised prior to use
sub tag() {
  if ( ! $tagpid ) {
	 print STDERR "Treetagger not opened\n";
	exit(1);
  }

  $_ = join("\n",split(/\s+/,shift())) . "\n.\n.\n.\n";

  if ( $verbose>1 ) { print "Writing " . join("#",split(/\n/,$_)) . "\n"; }
  
  if ( ! print(FI $_) ) {
    Alvis::Treetagger::shut();
    return undef;
  }

  #  first skip over initial "." from previous output forcing
  while ( defined($_=<FO>) && $_ =~ /^\.\t/ ) {
    ;
  }
  if ( ! defined($_) ) {
    Alvis::Treetagger::shut();
    return undef;
  }

  #  now pick up the results
  my $result = $_;  
  while ( defined($_=<FO>) && $_ !~ /^\.\t/ ) {
    $result .= $_;
  }
  if ( ! defined($_) ) {
    Alvis::Treetagger::shut();
    return undef;
  }  
  if ( $verbose>1 ) { print "Received: $result\n"; }
  return $result;
}

1;

__END__