#!/usr/bin/perl -w
# xml2wsj: convert Alpino xml format to WSJ bracket format
# usage:   xml2wsj files
# 20021218 erikt@uia.ua.ac.be

use strict;

my ($i,$t,
    $file,$headlineSeen,$outFile,$line,$text,
    @tmp,@tokens);

# for each file
LOOP1: foreach $file (@ARGV) { if ($file =~ /xml$/i) {
   # open file, store it in buffer $text
   if (not open(INFILE,$file)) {
      print STDERR "cannot open $file for reading\n";
      next LOOP1 ;
   }
   $text = "";
   while (<INFILE>) { 
      $line = $_;
      chomp($line);
      $text .= $line;
   }
   close(INFILE);
   # divide file in tags/non-tags; store in @tokens
   @tmp = split(/([<>])/,$text);
   @tokens = ();
   $i = 0;
   foreach $t (@tmp) {
      if ($t eq "<") { 
         if (defined $tokens[$i]) {
            if ($tokens[$i] =~ /^</) { die "embedded tag: cannot happen\n"; }
            $i++; 
         }
         $tokens[$i] = $t; 
      } elsif (defined $tokens[$i]) {
         $tokens[$i] .= $t;
      } else { $tokens[$i] = $t; }
     if ($t eq ">") { $i++; }
   }
   if (defined $tokens[$i] and $tokens[$i] =~ /^</) { 
      die "open tag: cannot happen\n";
   }
   # store text (not xml tags) in output file
   $outFile = $file;
   $outFile =~ s/xml$/txt/i;
   if (not open(OUTFILE,"|../bin/htmlize -r > $outFile")) {
      print STDERR "cannot open $outFile for writing\n";
      next LOOP1;
   }
   $headlineSeen = 0;
   LOOP2: foreach $t (@tokens) {
      if ($t =~ /<headline/i) { $headlineSeen = 1; }
      if ($headlineSeen and 
          $t !~ /^\s*$/ and $t !~ /^<.*>$/) { print OUTFILE "$t\n\n"; }
      if ($t =~ /<\/text>/i) { last LOOP2; } # stop when </text> was seen
   }
   close(OUTFILE);
}}   
exit(0);