#!/usr/bin/perl

# extract rhetoricaltext environment from XML
# THIS FILE NEEDS TO BE SAVED AS UNICODE!!! (UTF-8)
# args: extractdoc  xmlfile docid
 
 
$fn =  $ARGV[0];

warn "loading $fn\n";

{
	local(*INPUT, $/);
	open (INPUT, $fn) 	|| die "can't open $fn: $!";
	$f = <INPUT>;
	close INPUT;
}  

$docid = $ARGV[1];

if ($docid eq '')
{
print "printing all document ids: \n";
 
while ($f =~ /\<document\s+id\s*=\s*(\"?)([^\">]+)(\"?)\s*>/isg)
{
	print $2."\n";
}

} else
{

$f =~ /<document\s+id\s*=\s*(\"?)$docid(\"?)\s*>(.*?)<\/\s*document\s*>/is;

$doc = $3;

print '\begin{rhetoricaltext}\source{'."$docid}\n";
while($doc =~ /<segment\s+id\s*=\s*(\"?)([^\">]+)(\"?)[^>]*>(.*?)<\/\s*segment\s*>/isg )
{
	my $segid = $2;
	my $seg = $4;
	 
	$segid =~ s/$docid\.?//;
	 
	 $seg =~ s/\<\/?sign[^>]*>//isg;
	 
	$seg =~ s/(Ä|\&Auml;)/\\\"\{A\}/sg;
	$seg =~ s/(Ö|\&Ouml;)/\\\"\{O\}/sg;
	$seg =~ s/(Ü|\&Uuml;)/\\\"\{U\}/sg;
	$seg =~ s/(ä|\&auml;)/\\\"\{a\}/sg;
	$seg =~ s/(ö|\&ouml;)/\\\"\{o\}/sg;
	$seg =~ s/(ü|\&uuml;)/\\\"\{u\}/sg;
	$seg =~ s/(ß|\&szlig;)/\\\"\{ss\}/sg;
	if ($seg)
	{ print "\\unit[$segid]{$seg}\n";
	}
}
print '\end{rhetoricaltext}'."\n";

}