h2u

   1 #!/usr/bin/env perl
   2
   3 #=======================================================================
   4 # h2u
   5 # File ID: e93feb18-5d3a-11df-bda7-90e6ba3022ac
   6 #
   7 # Converts from numeric entities in HTML/SGML (&#x263A; and &#9786;) to
   8 # UTF-8.
   9 #
  10 # Character set: UTF-8
  11 # ©opyleft 2001– Øyvind A. Holm <sunny@sunbase.org>
  12 # License: GNU General Public License version 2 or later, see end of
  13 # file for legal stuff.
  14 #=======================================================================
  15
  16 use strict;
  17 use warnings;
  18 use Getopt::Long;
  19
  20 local $| = 1;
  21
  22 our %Opt = (
  23
  24     'help' => 0,
  25     'invalid' => 0,
  26     'latin1' => 0,
  27     'quiet' => 0,
  28     'verbose' => 0,
  29     'version' => 0,
  30
  31 );
  32
  33 our $progname = $0;
  34 $progname =~ s/^.*\/(.*?)$/$1/;
  35 our $VERSION = '0.1.0';
  36
  37 Getopt::Long::Configure('bundling');
  38 GetOptions(
  39
  40     'help|h' => \$Opt{'help'},
  41     'invalid|i' => \$Opt{'invalid'},
  42     'latin1|l' => \$Opt{'latin1'},
  43     'quiet|q+' => \$Opt{'quiet'},
  44     'verbose|v+' => \$Opt{'verbose'},
  45     'version' => \$Opt{'version'},
  46
  47 ) || die("$progname: Option error. Use -h for help.\n");
  48
  49 $Opt{'verbose'} -= $Opt{'quiet'};
  50 $Opt{'help'} && usage(0);
  51 if ($Opt{'version'}) {
  52     print_version();
  53     exit(0);
  54 }
  55
  56 exit(main());
  57
  58 sub main {
  59     # {{{
  60     my $Retval = 0;
  61
  62     while (<>) {
  63         $Opt{'latin1'} && s/([\x80-\xFF])/widechar(ord($1))/ge;
  64         s/&#(\d{1,10});/widechar($1)/ge;
  65         s/&#x([0-9a-f]{1,8});/widechar(hex($1))/gei;
  66         print;
  67     }
  68
  69     return $Retval;
  70     # }}}
  71 } # main()
  72
  73 sub widechar {
  74         # {{{
  75         my $Val = shift;
  76         if ($Val < 0x80) {
  77                 return sprintf("%c", $Val);
  78         } elsif ($Val < 0x800) {
  79                 return sprintf("%c%c", 0xC0 | ($Val >> 6),
  80                                        0x80 | ($Val & 0x3F));
  81         } elsif ($Val < 0x10000) {
  82                 unless ($Opt{'invalid'}) {
  83                         if  (($Val >= 0xD800 && $Val <= 0xDFFF) || ($Val eq 0xFFFE) || ($Val eq 0xFFFF)) {
  84                                 $Val = 0xFFFD;
  85                         }
  86                 }
  87                 return sprintf("%c%c%c", 0xE0 |  ($Val >> 12),
  88                                          0x80 | (($Val >>  6) & 0x3F),
  89                                          0x80 |  ($Val        & 0x3F));
  90         } elsif ($Val < 0x200000) {
  91                 return sprintf("%c%c%c%c", 0xF0 |  ($Val >> 18),
  92                                            0x80 | (($Val >> 12) & 0x3F),
  93                                            0x80 | (($Val >>  6) & 0x3F),
  94                                            0x80 |  ($Val        & 0x3F));
  95         } elsif ($Val < 0x4000000) {
  96                 return sprintf("%c%c%c%c%c", 0xF8 |  ($Val >> 24),
  97                                              0x80 | (($Val >> 18) & 0x3F),
  98                                              0x80 | (($Val >> 12) & 0x3F),
  99                                              0x80 | (($Val >>  6) & 0x3F),
 100                                              0x80 | ( $Val        & 0x3F));
 101         } elsif ($Val < 0x80000000) {
 102                 return sprintf("%c%c%c%c%c%c", 0xFC |  ($Val >> 30),
 103                                                0x80 | (($Val >> 24) & 0x3F),
 104                                                0x80 | (($Val >> 18) & 0x3F),
 105                                                0x80 | (($Val >> 12) & 0x3F),
 106                                                0x80 | (($Val >>  6) & 0x3F),
 107                                                0x80 | ( $Val        & 0x3F));
 108         } else {
 109                 return widechar(0xFFFD);
 110         }
 111         # }}}
 112 } # widechar()
 113
 114 sub print_version {
 115     # Print program version {{{
 116     print("$progname $VERSION\n");
 117     return;
 118     # }}}
 119 } # print_version()
 120
 121 sub usage {
 122     # Send the help message to stdout {{{
 123     my $Retval = shift;
 124
 125     if ($Opt{'verbose'}) {
 126         print("\n");
 127         print_version();
 128     }
 129     print(<<"END");
 130
 131 Converts from numeric entities in HTML/SGML (&#x263A; and &#9786;) to
 132 UTF-8.
 133
 134 Usage: $progname [options] [file [files [...]]]
 135
 136 Options:
 137
 138   -h, --help
 139     Show this help.
 140   -i, --invalid
 141     Allow invalid character range U+D800 through U+DFFF, U+FFFE and
 142     U+FFFF.
 143   -l, --latin1
 144     Also convert Latin-1 characters.
 145   -q, --quiet
 146     Be more quiet. Can be repeated to increase silence.
 147   -v, --verbose
 148     Increase level of verbosity. Can be repeated.
 149   --version
 150     Print version information.
 151
 152 END
 153     exit($Retval);
 154     # }}}
 155 } # usage()
 156
 157 sub msg {
 158     # Print a status message to stderr based on verbosity level {{{
 159     my ($verbose_level, $Txt) = @_;
 160
 161     if ($Opt{'verbose'} >= $verbose_level) {
 162         print(STDERR "$progname: $Txt\n");
 163     }
 164     return;
 165     # }}}
 166 } # msg()
 167
 168 __END__
 169
 170 # This program is free software; you can redistribute it and/or modify
 171 # it under the terms of the GNU General Public License as published by
 172 # the Free Software Foundation; either version 2 of the License, or (at
 173 # your option) any later version.
 174 #
 175 # This program is distributed in the hope that it will be useful, but
 176 # WITHOUT ANY WARRANTY; without even the implied warranty of
 177 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 178 # See the GNU General Public License for more details.
 179 #
 180 # You should have received a copy of the GNU General Public License
 181 # along with this program.
 182 # If not, see L<http://www.gnu.org/licenses/>.
 183
 184 # vim: set fenc=UTF-8 ft=perl fdm=marker ts=4 sw=4 sts=4 et fo+=w :