html.pl

   1 #!/usr/bin/perl -X
   2
   3 # (C) 2016-2017 Tobias Girstmair
   4 # Extracts hypertext formatted news from ORF Teletext
   5 # uses a modified version of vtx2ascii to decode pages
   6 # from dvbtext's spool directory.
   7
   8 # Usage: ./plain.pl <VTX-file>
   9 # Output: HTML
  10
  11 use strict;
  12 use warnings;
  13 use 5.010;
  14 binmode STDOUT, ":encoding(utf8)";
  15
  16 # Seitenformat:
  17 #  100-109:
  18 #   Metadaten:      1
  19 #   Subressort:     2
  20 #   Ressort/Sparte: 3
  21 #   Leer            4
  22 #   Related:        5
  23 #   Leer            6
  24 #   Titel:          7
  25 #   Text:           8-24
  26 #
  27 #  112-899:
  28 #   Metadaten:      1
  29 #   Subressort:     2
  30 #   Ressort/Sparte: 3
  31 #   Leer            4
  32 #   Titel:          5
  33 #   Text:           6-24
  34 #
  35
  36 my %meta;
  37 my $title;
  38 my $text = "";
  39 my $page = shift;
  40 my $subp = 0; #TODO: allow subpages
  41 # run through vtx2ascii (has been modified to output correct ISO 8859-1 without national replacements)
  42 open (VTX, "./vtx2ascii -a $page |") || die ("Can'r run vtx2ascii");
  43 my $last = "";
  44 my $is_10x = 0;
  45 do {
  46         # transliterate from ETSI EN 300 706 G0 German to Latin-1 (will be converted to UTF-8 by perl):
  47         tr/[\\]{|}~/\N{U+C4}\N{U+D6}\N{U+DC}\N{U+E4}\N{U+F6}\N{U+FC}\N{U+DF}/;
  48         my $line = $_;
  49         $line =~ s/^\s+|\s+$//g;
  50         chomp ($line);
  51
  52         given ($.) {
  53                 when (1) { %meta = parse_metadata ($line) ; $is_10x = ($meta{'page'}<110) }
  54                 when (2) { $meta{'subres'} = $line }
  55                 when (3) { $meta{'res'} = $line }
  56                 when (4) {}
  57                 when (5 + (1*$is_10x)) { $title = $line }
  58                 when (4 + (1*$is_10x)) {}
  59                 when (4 + (3*$is_10x)) { $title .=$line }
  60                 default { $text .= $last . "_EOL_" . ($last eq ""?"":($line eq ""?"<br>":" ")) }
  61         }
  62         $last = $line unless $. == (5+(2*$is_10x));
  63 } while (<VTX>);
  64 $text .= $last;
  65
  66 #remove hyphenation at original line ending only when in between lowercase letters, replace with soft hyphen to still allow hyphenation when needed. ad _EOL_: linebreaks already stripped in loop above; wouldn't work either way due to single line regex.
  67 $text =~ s/([[:lower:]])-_EOL_ ([[:lower:]])/\1&shy;\2/g;
  68 $text =~ s/_EOL_//g;
  69
  70 # adblocker: just add more regexes
  71 $text =~ s/Kalendarium - t.glich neu \. 734//g;
  72
  73 print "<p>$meta{'page'}: <b>$title</b><br>$text</p>";
  74
  75 close (VTX);
  76
  77 sub parse_metadata {
  78         my @elems = split ' ', @_[0];
  79
  80         my %retval = (
  81                 'page' => shift @elems,
  82                 'channel' => shift @elems,
  83                 'date' => join (' ', @elems)
  84         );
  85
  86         return %retval;
  87 }