script/korapxml2conllu

#!/usr/bin/env perl
use strict;
use warnings;
use POSIX;
use Getopt::Std;
use Encode;
use List::Util qw[min max];

my $MAX_SENTENCE_LENGTH=10000;
my $COMMENT_START="#";

my $test=0;
my $text_no=0;
my %opts;
my %plain_texts;
my $usage=<<EOF;
Usage: $0 [options] ZIPFILE [ZIPFILE...]

Options:
 -p pattern

Description:
 Convert KorAP-XML morpho zip to CoNLL(-U) format with all information necessary
 for reconstruction in comment lines.

Examples:
 $0 /vol/corpora/DeReKo/current/KorAP/zip/zca15.tree_tagger.zip

 ZIPSIGLEPATTERN='-x "*15/FEB*" "*15/MAR*"' $0 /vol/corpora/DeReKo/current/KorAP/zip/zca15.tree_tagger.zip

 Results will be written to stdout
EOF

getopts('dhp:', \%opts);
die $usage if($opts{h} || @ARGV == 0);
my $debug=($opts{d}? 1 : 0);

my $docid="";
my ($current_id, $current_from, $current_to, $token);
my $current;
my ($unknown, $known) = (0, 0);
my @current_lines;
my %processedFilenames;
my $zipsiglepattern = (defined($ENV{ZIPSIGLEPATTERN})? $ENV{ZIPSIGLEPATTERN} : "");

my ($ID_idx, $FORM_idx, $LEMMA_idx, $UPOS_idx, $XPOS_idx, $FEATS_idx, $HEAD_idx, $DEPREC_idx, $DEPS_idx, $MISC_idx) = (0..9);

my $UNZIP = `sh -c 'command -v unzip'`;
chomp $UNZIP;


if ($UNZIP eq '') {
  warn('No unzip executable found in PATH.');
  return 0;
};

foreach my $morpho_zip (@ARGV) {
  die "cannot open $morpho_zip" if(! -r $morpho_zip);
  my $data_zip = $morpho_zip;
  if ($data_zip !~ /\.zip/ && $data_zip =~ /\.conllu?/i) {
    open(CONLL, "<$data_zip") or die "cannot open $data_zip";
    while(<CONLL>) {
      print;
    }
    close(CONLL);
    next;
  }
  $data_zip =~ s/\.([^.]+)\.zip$/.zip/;
  my $foundry = $1;
  die "cannot open data file $data_zip corresponding to $morpho_zip" if(! -r $data_zip);
  
  my $first=1;
  my $pattern = (defined($opts{p})? $opts{p} : '');
  my @conll = ("_") x 10;
  my $filename;

  my $morphocommand = "$UNZIP -c $morpho_zip '*/${pattern}*/*/*/morpho.xml' $zipsiglepattern |";
  # print STDERR $morphocommand, "\n";
  open (MORPHOPIPE, $morphocommand) or die "cannot unzip $morpho_zip";
  open (PLAINTEXTPIPE, "$UNZIP -c $data_zip '*/${pattern}*/*/data.xml' $zipsiglepattern |") or die "cannot unzip $data_zip";
  print "$COMMENT_START foundry = $foundry\n";
  while (<MORPHOPIPE>) {
    if (/^  inflating: (.*)/) {
      $filename=$1;
      while($processedFilenames{$filename} && !eof(MORPHOPIPE)) {
        print STDERR "WARNING: $filename already processed\n";
        while (<MORPHOPIPE>) {
          last if(/\s+inflating:\s+(.*)/);
        }
        $filename=$1 if(!eof(MORPHOPIPE) && /\s+inflating:\s+(.*)/);
      }
    } elsif(m@^\s*<layer\s+.*docid="([^"]+)"@) {
      last if($test && $text_no++ > 3);
      if(!$first) {
        closeDoc(0);
      }
      $processedFilenames{$filename}=1;
      $docid=$1;
      @current_lines=();
      $known=$unknown=0;
      $current="";
      if ($first) {
          $first = 0;
      }
      if(!fetch_plaintext($docid)) { # skip this text
        while (<MORPHOPIPE>) {
          last if(m@</layer>@);
        }
      }
      print STDOUT "$COMMENT_START filename = $filename\n$COMMENT_START text_id = $docid\n";
      print STDERR "Analyzing $docid\n" if ($debug);
    } elsif (m@^\s*<f\s+.*name="([^"]+)">([^<]+)</f>@) {
      if ($1 eq "lemma") {
        $conll[$LEMMA_idx] = $2;
        $conll[$LEMMA_idx] =~ s/[\t\n\r]//g; # make sure that lemmas never contain tabs or newlines
        if($conll[$LEMMA_idx] eq 'UNKNOWN') {
          $conll[$LEMMA_idx] = "--";
          $unknown++;
        } else {
          $known++;
        }
      } elsif ($1 eq 'pos' || $1 eq "ctag") {
        $unknown++;
        $conll[$XPOS_idx] = $conll[$UPOS_idx] = $2;
      } elsif ($1 eq 'msd') {
        $conll[$FEATS_idx] = $2;
      } elsif ($1 eq 'certainty') {
        $conll[$MISC_idx] = $2;
      }
    } elsif (/<span /) {
      ($current_id) = /id="[^0-9]*([^\"]*)"/;
      ($current_from) = /from="([^\"]*)"/;
      ($current_to) = /to="([^\"]*)"/;
      print STDERR "found span: $current_id $current_from $current_to\n" if($debug);
      $token = substr($plain_texts{$docid}, $current_from, $current_to - $current_from);
      if (!defined $token) {
        print STDERR "WARNING: could not retrieve token for $docid at $current_from-$current_to/", length($plain_texts{$docid}), " - ending with: ", substr($plain_texts{$docid},length($plain_texts{$docid})-10), "\n";
        $token = "_";
      }
      $token=~s/[\t\n\r]//g; # make sure that tokens never contain tabs or newlines
      @conll = ("_") x 10;
      $conll[$FORM_idx] = encode("utf-8", $token);
    } elsif (m@^\s*</fs>@) {
      my @vals = ($current_from, $current_to);
      print STDERR "joining : ", join(" ", @vals), "\n" if($debug);
      push @current_lines, \@vals;
      # convert gathered information to CONLL
      $conll[$ID_idx] = $#current_lines+1;
      $current .= join("\t", @conll) . "\n"; # conll columns
      if($conll[$XPOS_idx] eq '$.' || ($conll[$XPOS_idx] eq 'SENT' && $token eq '.') || $known + $unknown >= $MAX_SENTENCE_LENGTH) {
        $current .= "\n";
        if($known + $unknown > 0) { # only print sentence if it contains some words
          printTokenRanges();
          print STDOUT $current;
        }
        $current=""; $known=0; $unknown=0;
        @current_lines = ();
      }
      while (<MORPHOPIPE>) {
        last if (m@</span>@);    # only consider first interpretation
      }
    }
  }
  $current .= "\n";
  closeDoc(1);
  close(MORPHOPIPE);
  close(PLAINTEXTPIPE);
}
exit;

sub printTokenRanges {
  print "$COMMENT_START start_offsets = ", $current_lines[0]->[0];
  foreach my $t (@current_lines) {
    print STDOUT " $t->[0]";
  }
  print "\n$COMMENT_START end_offsets = ", $current_lines[$#current_lines]->[1];
  foreach my $t (@current_lines) {
    print STDOUT " $t->[1]";
  }
  print "\n";
}

sub closeDoc {
  my ($end) = @_;
  print STDERR "closing doc\n" if($debug);
  if($known + $unknown > 0) { # only parse a sentence if it has some words
    chomp $current;
    chomp $current;
    chomp $current;
    $current .= "\n\n";
    printTokenRanges();
    print STDOUT $current;
  }
}

# read data.xml to figure out the tokens
# (ideally tokens should also be in in morpho.xml, but they are not)
sub fetch_plaintext {
  my ($target_id) = @_;
  my $docid;
  my $text_started=0;

  if($plain_texts{$target_id}) {
#    print STDERR "already got $target_id\n";
    return;
  }
  while(<PLAINTEXTPIPE>) {
    if(/<raw_text[^>]+docid="([^"]*)/) {
      $docid=$1;
      $text_started=0;
    } elsif (m@<text>(.*)</text>@) {
      $_= decode("utf-8", $1, Encode::FB_DEFAULT);
      s/&lt;/</go;
      s/&gt;/>/go;
      s/&amp;/&/go;
      tr/…•⋅»«ˮ“”„›‹ʼ‘’‚′‐‑‒–—―⁓⁻₋−﹣－/...""""""'''''''-/;
      $plain_texts{$docid} = $_;
      last if($docid eq $target_id);
    } elsif (m@<text>(.*)@) {
      $_= decode("utf-8", $1, Encode::FB_DEFAULT);
      s/&lt;/</go;
      s/&gt;/>/go;
      s/&amp;/&/go;
      tr/…•⋅»«ˮ“”„›‹ʼ‘’‚′‐‑‒–—―⁓⁻₋−﹣－/...""""""'''''''-/;
      $plain_texts{$docid} = "$_ ";
      $text_started=1;
    } elsif ($text_started && m@(.*)</text>@) {
      $_= decode("utf-8", $1, Encode::FB_DEFAULT);
      s/&lt;/</go;
      s/&gt;/>/go;
      s/&amp;/&/go;
      tr/…•⋅»«ˮ“”„›‹ʼ‘’‚′‐‑‒–—―⁓⁻₋−﹣－/...""""""'''''''-/;
      $plain_texts{$docid} .= $_;
      $text_started=0;
      last if($docid eq $target_id);
    } elsif ($text_started) {
      chomp;
      $_ = decode("utf-8", $_, Encode::FB_DEFAULT) . ' ';
      s/&lt;/</go;
      s/&gt;/>/go;
      s/&amp;/&/go;
      tr/…•⋅»«ˮ“”„›‹ʼ‘’‚′‐‑‒–—―⁓⁻₋−﹣－/...""""""'''''''-/;
      $plain_texts{$docid} .= $_;
    }
  }
  if(defined($ENV{PLAINTEXTFILTER})) {
    if ($plain_texts{$docid} !~ $ENV{PLAINTEXTFILTER}) {
      $plain_texts{$docid} = undef;
      print STDERR "Skipping $docid\n";
      return(undef);
    } else {
      print STDERR "Using $docid\n";
    }
  } 
  return(1);
}