[adaa778] | 1 | #!/usr/bin/perl -w |
---|
| 2 | require 5.004; |
---|
| 3 | use strict; |
---|
| 4 | |
---|
| 5 | use integer; |
---|
| 6 | |
---|
| 7 | # Magic identifier (12 bytes) |
---|
[706b033] | 8 | my $magic = "Svx\nMsg\r\n\xfe\xff\0"; |
---|
[adaa778] | 9 | # Designed to be corrupted by ASCII ftp, top bit stripping (or |
---|
| 10 | # being used for parity). Contains a zero byte so more likely |
---|
| 11 | # to be flagged as data (e.g. by perl's "-B" test). |
---|
| 12 | |
---|
| 13 | my $major = 0; |
---|
| 14 | my $minor = 8; |
---|
| 15 | |
---|
| 16 | # File format (multi-byte numbers in network order (bigendian)): |
---|
| 17 | # 12 bytes: Magic identifier |
---|
| 18 | # 1 byte: File format major version number (0) |
---|
| 19 | # 1 byte: File format minor version number (8) |
---|
| 20 | # 2 bytes: Number of messages (N) |
---|
| 21 | # 4 bytes: Offset from XXXX to end of file |
---|
| 22 | # XXXX: |
---|
| 23 | # N*: |
---|
| 24 | # <message> NUL |
---|
| 25 | |
---|
| 26 | my %ent = (); |
---|
| 27 | |
---|
| 28 | open ENT, "named-entities.txt" or die $!; |
---|
| 29 | while (<ENT>) { |
---|
| 30 | my ($e, $v) = /^(\w+),(\d+)/; |
---|
| 31 | $ent{$e} = $v; |
---|
| 32 | } |
---|
| 33 | close ENT; |
---|
| 34 | |
---|
| 35 | my %msgs = (); |
---|
| 36 | |
---|
| 37 | while (<>) { |
---|
| 38 | next if /^\s*#/; # skip comments |
---|
| 39 | |
---|
| 40 | # en: 0 0.81 the message |
---|
| 41 | # en-us: 0 0.81 " the message " |
---|
[eb695cd] | 42 | my ($langs, $msgno, $dummy, $msg) = /^([-\w,]+):\s*(\d+)\s+("?)(.*)\3/; |
---|
[adaa778] | 43 | |
---|
[eb695cd] | 44 | unless (defined $langs) { |
---|
[adaa778] | 45 | chomp; |
---|
| 46 | print STDERR "Warning: Bad line: \"$_\"\n"; |
---|
| 47 | next; |
---|
| 48 | } |
---|
| 49 | |
---|
| 50 | if ($msg =~ /[\0-\x1f\x7f-\xff]/) { |
---|
| 51 | print STDERR "Warning: literal character in message $msgno\n"; |
---|
| 52 | } |
---|
| 53 | |
---|
[eb695cd] | 54 | my $utf8 = string_to_utf8($msg); |
---|
| 55 | for (split /,/, $langs) { |
---|
| 56 | ${$msgs{$_}}[$msgno] = $utf8; |
---|
| 57 | } |
---|
[adaa778] | 58 | } |
---|
| 59 | |
---|
| 60 | my $lang; |
---|
[eb695cd] | 61 | my @langs = sort grep ! /-\*$/, keys %msgs; |
---|
[adaa778] | 62 | |
---|
| 63 | my $num_msgs = -1; |
---|
| 64 | foreach $lang (@langs) { |
---|
| 65 | my $aref = $msgs{$lang}; |
---|
| 66 | $num_msgs = scalar @$aref if scalar @$aref > $num_msgs; |
---|
| 67 | } |
---|
| 68 | |
---|
| 69 | foreach $lang (@langs) { |
---|
| 70 | open OUT, ">$lang.msg" or die $!; |
---|
| 71 | |
---|
| 72 | my $aref = $msgs{$lang}; |
---|
| 73 | |
---|
| 74 | my $parentaref; |
---|
| 75 | my $mainlang = $lang; |
---|
| 76 | $parentaref = $msgs{$mainlang} if $mainlang =~ s/-.*$//; |
---|
| 77 | |
---|
| 78 | print OUT $magic or die $!; |
---|
| 79 | print OUT pack("CCn", $major, $minor, $num_msgs) or die $!; |
---|
| 80 | |
---|
| 81 | my $buff = ''; |
---|
| 82 | |
---|
| 83 | my $n; |
---|
| 84 | for $n (0 .. $num_msgs - 1) { |
---|
| 85 | my $msg = $$aref[$n]; |
---|
| 86 | if (!defined $msg) { |
---|
| 87 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
| 88 | if (!defined $msg) { |
---|
| 89 | $msg = ${$msgs{'en'}}[$n]; |
---|
| 90 | if (defined $msg && $msg ne '') { |
---|
| 91 | print STDERR "Warning: message $n not in language $lang\n"; |
---|
| 92 | } else { |
---|
| 93 | $msg = ''; |
---|
| 94 | } |
---|
| 95 | } |
---|
| 96 | } |
---|
| 97 | $buff .= $msg . "\0"; |
---|
| 98 | } |
---|
| 99 | |
---|
| 100 | print OUT pack('N',length($buff)), $buff or die $!; |
---|
| 101 | close OUT or die $!; |
---|
| 102 | } |
---|
| 103 | |
---|
| 104 | sub string_to_utf8 { |
---|
| 105 | my $s = shift; |
---|
| 106 | $s =~ s/[\x80-\xff]/char_to_utf8(ord$1)/eg; |
---|
| 107 | $s =~ s/\&(#\d+|#x[a-f0-9]+|[a-z0-9]+);?/decode_entity($1)/eig; |
---|
| 108 | return $s; |
---|
| 109 | } |
---|
| 110 | |
---|
| 111 | sub decode_entity { |
---|
| 112 | my $ent = shift; |
---|
| 113 | return char_to_utf8($1) if $ent =~ /^#(\d+)$/; |
---|
| 114 | return char_to_utf8(hex($1)) if $ent =~ /^#x([a-f0-9]+)$/; |
---|
| 115 | return char_to_utf8($ent{$ent}) if exists $ent{$ent}; |
---|
| 116 | $ent = "\&$ent;"; |
---|
| 117 | print STDERR "Warning: entity \"$ent\" unknown\n"; |
---|
| 118 | return $ent; |
---|
| 119 | } |
---|
| 120 | |
---|
| 121 | sub char_to_utf8 { |
---|
| 122 | my $unicode = shift; |
---|
| 123 | # ASCII is easy, and the most common case |
---|
| 124 | return chr($unicode) if $unicode < 0x80; |
---|
| 125 | |
---|
| 126 | my $result = ''; |
---|
| 127 | my $n = 0x20; |
---|
| 128 | while (1) { |
---|
| 129 | $result = chr(0x80 | ($unicode & 0x3f)) . $result; |
---|
| 130 | $unicode >>= 6; |
---|
| 131 | last if $unicode < $n; |
---|
| 132 | $n >>= 1; |
---|
| 133 | } |
---|
| 134 | $result = chr((0x100 - $n*2) | $unicode) . $result; |
---|
| 135 | return $result; |
---|
| 136 | } |
---|