[adaa778] | 1 | #!/usr/bin/perl -w |
---|
[a269915] | 2 | require 5.008; |
---|
[96b0efa] | 3 | use bytes; |
---|
[adaa778] | 4 | use strict; |
---|
| 5 | |
---|
| 6 | use integer; |
---|
| 7 | |
---|
[d5bd3a7] | 8 | # messages >= this value are written to a header file |
---|
| 9 | my $dontextract_threshold = 1000; |
---|
| 10 | |
---|
[adaa778] | 11 | # Magic identifier (12 bytes) |
---|
[706b033] | 12 | my $magic = "Svx\nMsg\r\n\xfe\xff\0"; |
---|
[adaa778] | 13 | # Designed to be corrupted by ASCII ftp, top bit stripping (or |
---|
| 14 | # being used for parity). Contains a zero byte so more likely |
---|
| 15 | # to be flagged as data (e.g. by perl's "-B" test). |
---|
| 16 | |
---|
| 17 | my $major = 0; |
---|
| 18 | my $minor = 8; |
---|
| 19 | |
---|
| 20 | # File format (multi-byte numbers in network order (bigendian)): |
---|
| 21 | # 12 bytes: Magic identifier |
---|
| 22 | # 1 byte: File format major version number (0) |
---|
| 23 | # 1 byte: File format minor version number (8) |
---|
| 24 | # 2 bytes: Number of messages (N) |
---|
| 25 | # 4 bytes: Offset from XXXX to end of file |
---|
| 26 | # XXXX: |
---|
| 27 | # N*: |
---|
| 28 | # <message> NUL |
---|
| 29 | |
---|
| 30 | my %ent = (); |
---|
| 31 | |
---|
[abd7d47] | 32 | open ENT, ($ENV{srcdir}?"$ENV{srcdir}/":"")."named-entities.txt" or die $!; |
---|
[adaa778] | 33 | while (<ENT>) { |
---|
| 34 | my ($e, $v) = /^(\w+),(\d+)/; |
---|
| 35 | $ent{$e} = $v; |
---|
| 36 | } |
---|
| 37 | close ENT; |
---|
| 38 | |
---|
| 39 | my %msgs = (); |
---|
[d5bd3a7] | 40 | my %dontextract = (); |
---|
[adaa778] | 41 | |
---|
[2083f8b] | 42 | my %raw = (); |
---|
| 43 | my $raw = ''; |
---|
| 44 | my $raw_comments = ''; |
---|
| 45 | my $curmsg = -1; |
---|
| 46 | |
---|
[adaa778] | 47 | while (<>) { |
---|
[2083f8b] | 48 | if (/^\s*#/) { |
---|
| 49 | $raw_comments .= $_; |
---|
| 50 | next; # skip comments |
---|
| 51 | } |
---|
| 52 | |
---|
[adaa778] | 53 | # en: 0 0.81 the message |
---|
| 54 | # en-us: 0 0.81 " the message " |
---|
[eb695cd] | 55 | my ($langs, $msgno, $dummy, $msg) = /^([-\w,]+):\s*(\d+)\s+("?)(.*)\3/; |
---|
[adaa778] | 56 | |
---|
[2083f8b] | 57 | if ($msgno != $curmsg) { |
---|
[acf84e45] | 58 | if ($msgno < $curmsg) { |
---|
| 59 | print STDERR "Warning: message number jumps back from $curmsg to $msgno\n"; |
---|
| 60 | } |
---|
[2083f8b] | 61 | $raw{$curmsg} = $raw; |
---|
| 62 | $raw = ''; |
---|
| 63 | $curmsg = $msgno; |
---|
| 64 | } |
---|
| 65 | |
---|
| 66 | $raw .= $raw_comments . $_; |
---|
| 67 | $raw_comments = ''; |
---|
| 68 | |
---|
[eb695cd] | 69 | unless (defined $langs) { |
---|
[adaa778] | 70 | chomp; |
---|
| 71 | print STDERR "Warning: Bad line: \"$_\"\n"; |
---|
| 72 | next; |
---|
| 73 | } |
---|
| 74 | |
---|
[c0a9908] | 75 | $langs =~ tr/-/_/; |
---|
| 76 | |
---|
[adaa778] | 77 | if ($msg =~ /[\0-\x1f\x7f-\xff]/) { |
---|
| 78 | print STDERR "Warning: literal character in message $msgno\n"; |
---|
| 79 | } |
---|
| 80 | |
---|
[eb695cd] | 81 | my $utf8 = string_to_utf8($msg); |
---|
| 82 | for (split /,/, $langs) { |
---|
[d5bd3a7] | 83 | if ($msgno >= $dontextract_threshold) { |
---|
[acf84e45] | 84 | if (${$dontextract{$_}}[$msgno - $dontextract_threshold]) { |
---|
| 85 | print STDERR "Warning: already had message $msgno for language $_\n"; |
---|
| 86 | } |
---|
[d5bd3a7] | 87 | ${$dontextract{$_}}[$msgno - $dontextract_threshold] = $utf8; |
---|
| 88 | } else { |
---|
[acf84e45] | 89 | if (${$msgs{$_}}[$msgno]) { |
---|
| 90 | print STDERR "Warning: already had message $msgno for language $_\n"; |
---|
| 91 | } |
---|
[d5bd3a7] | 92 | ${$msgs{$_}}[$msgno] = $utf8; |
---|
| 93 | } |
---|
[eb695cd] | 94 | } |
---|
[adaa778] | 95 | } |
---|
[9e9d0a4] | 96 | $raw{$curmsg} = $raw; |
---|
[adaa778] | 97 | |
---|
| 98 | my $lang; |
---|
[c0a9908] | 99 | my @langs = sort grep ! /_\*$/, keys %msgs; |
---|
[adaa778] | 100 | |
---|
| 101 | my $num_msgs = -1; |
---|
| 102 | foreach $lang (@langs) { |
---|
| 103 | my $aref = $msgs{$lang}; |
---|
| 104 | $num_msgs = scalar @$aref if scalar @$aref > $num_msgs; |
---|
| 105 | } |
---|
| 106 | |
---|
| 107 | foreach $lang (@langs) { |
---|
[c0a9908] | 108 | my $fnm = $lang; |
---|
| 109 | $fnm =~ s/(_.*)$/\U$1/; |
---|
| 110 | open OUT, ">$fnm.msg" or die $!; |
---|
[adaa778] | 111 | |
---|
| 112 | my $aref = $msgs{$lang}; |
---|
| 113 | |
---|
| 114 | my $parentaref; |
---|
| 115 | my $mainlang = $lang; |
---|
[c0a9908] | 116 | $parentaref = $msgs{$mainlang} if $mainlang =~ s/_.*$//; |
---|
[adaa778] | 117 | |
---|
| 118 | print OUT $magic or die $!; |
---|
| 119 | print OUT pack("CCn", $major, $minor, $num_msgs) or die $!; |
---|
| 120 | |
---|
| 121 | my $buff = ''; |
---|
| 122 | |
---|
| 123 | my $n; |
---|
| 124 | for $n (0 .. $num_msgs - 1) { |
---|
[79c5f17] | 125 | my $warned = 0; |
---|
[adaa778] | 126 | my $msg = $$aref[$n]; |
---|
| 127 | if (!defined $msg) { |
---|
| 128 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
| 129 | if (!defined $msg) { |
---|
| 130 | $msg = ${$msgs{'en'}}[$n]; |
---|
[abd7d47] | 131 | # don't report if we have a parent (as the omission will be |
---|
| 132 | # reported there) |
---|
[2083f8b] | 133 | if (defined $msg && $msg ne '' && !defined $parentaref) { |
---|
| 134 | print STDERR "Warning: message $n not in language $lang\n"; |
---|
| 135 | open TODO, ">>$lang.todo" or die $!; |
---|
| 136 | print TODO $raw{$n}, "\n"; |
---|
| 137 | close TODO; |
---|
[79c5f17] | 138 | $warned = 1; |
---|
[adaa778] | 139 | } else { |
---|
[abd7d47] | 140 | $msg = '' if !defined $msg; |
---|
[adaa778] | 141 | } |
---|
| 142 | } |
---|
| 143 | } |
---|
[acab4b4] | 144 | if (defined($raw{$n}) |
---|
| 145 | && $raw{$n} =~ /^\s*#\s*TRANSLATE\b[- \ta-z]*\b$lang\b/m) { |
---|
[79c5f17] | 146 | if ($warned) { |
---|
[abd7d47] | 147 | print STDERR "Warning: message $n missing and also marked for retranslation for language $lang\n"; |
---|
[79c5f17] | 148 | } else { |
---|
| 149 | print STDERR "Warning: message $n needs retranslating for language $lang\n"; |
---|
| 150 | open TODO, ">>$lang.todo" or die $!; |
---|
| 151 | print TODO $raw{$n}, "\n"; |
---|
| 152 | close TODO; |
---|
| 153 | } |
---|
| 154 | } |
---|
[adaa778] | 155 | $buff .= $msg . "\0"; |
---|
| 156 | } |
---|
| 157 | |
---|
| 158 | print OUT pack('N',length($buff)), $buff or die $!; |
---|
| 159 | close OUT or die $!; |
---|
| 160 | } |
---|
| 161 | |
---|
[d5bd3a7] | 162 | my $num_dontextract = -1; |
---|
| 163 | foreach $lang (@langs) { |
---|
| 164 | my $aref = $dontextract{$lang}; |
---|
| 165 | if (defined(@$aref)) { |
---|
| 166 | $num_dontextract = scalar @$aref if scalar @$aref > $num_dontextract; |
---|
| 167 | } |
---|
| 168 | } |
---|
| 169 | |
---|
| 170 | foreach $lang (@langs) { |
---|
[dc639a8] | 171 | my $fnm = $lang; |
---|
| 172 | $fnm =~ s/(_.*)$/\U$1/; |
---|
| 173 | open OUT, ">$fnm.h" or die $!; |
---|
[d5bd3a7] | 174 | print OUT "#define N_DONTEXTRACTMSGS ", $num_dontextract, "\n"; |
---|
| 175 | print OUT "static unsigned char dontextractmsgs[] ="; |
---|
| 176 | |
---|
| 177 | for my $n (0 .. $num_dontextract - 1) { |
---|
| 178 | print OUT "\n"; |
---|
| 179 | |
---|
| 180 | my $aref = $dontextract{$lang}; |
---|
| 181 | |
---|
| 182 | my $parentaref; |
---|
| 183 | my $mainlang = $lang; |
---|
| 184 | $parentaref = $dontextract{$mainlang} if $mainlang =~ s/_.*$//; |
---|
| 185 | |
---|
| 186 | my $msg = $$aref[$n]; |
---|
| 187 | if (!defined $msg) { |
---|
| 188 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
| 189 | if (!defined $msg) { |
---|
| 190 | $msg = ${$dontextract{'en'}}[$n]; |
---|
[2083f8b] | 191 | # don't report if we have a parent (as the omission will be reported there) |
---|
| 192 | if (defined $msg && $msg ne '' && !defined $parentaref) { |
---|
| 193 | print STDERR "Warning: message ", $dontextract_threshold + $n, " not in language $lang\n"; |
---|
| 194 | open TODO, ">>$lang.todo" or die $!; |
---|
| 195 | print TODO $raw{$dontextract_threshold + $n}, "\n"; |
---|
| 196 | close TODO; |
---|
[d5bd3a7] | 197 | } else { |
---|
| 198 | $msg = ''; |
---|
| 199 | } |
---|
| 200 | } |
---|
| 201 | } |
---|
| 202 | $msg =~ s/\\/\\\\/g; |
---|
| 203 | $msg =~ s/"/\\"/g; |
---|
| 204 | $msg =~ s/\t/\\t/g; |
---|
| 205 | $msg =~ s/\n/\\n/g; |
---|
| 206 | $msg =~ s/\r/\\r/g; |
---|
| 207 | if ($msg =~ /^ / || $msg =~ / $/) { |
---|
| 208 | $msg =~ s/\\"/\\\\\\"/g; |
---|
| 209 | $msg = '\\"'.$msg.'\\"'; |
---|
| 210 | } |
---|
| 211 | print OUT " /*", $dontextract_threshold + $n, "*/ \"$msg\\0\""; |
---|
| 212 | } |
---|
| 213 | print OUT ";\n"; |
---|
| 214 | close OUT or die $!; |
---|
| 215 | } |
---|
| 216 | |
---|
[2083f8b] | 217 | #for (sort {$a<=>$b} keys %raw) {print "$_ = [$raw{$_}]\n";} |
---|
| 218 | |
---|
[adaa778] | 219 | sub string_to_utf8 { |
---|
| 220 | my $s = shift; |
---|
[7de787f] | 221 | $s =~ s/([\x80-\xff])/char_to_utf8(ord($1))/eg; |
---|
[adaa778] | 222 | $s =~ s/\&(#\d+|#x[a-f0-9]+|[a-z0-9]+);?/decode_entity($1)/eig; |
---|
| 223 | return $s; |
---|
| 224 | } |
---|
| 225 | |
---|
| 226 | sub decode_entity { |
---|
| 227 | my $ent = shift; |
---|
| 228 | return char_to_utf8($1) if $ent =~ /^#(\d+)$/; |
---|
| 229 | return char_to_utf8(hex($1)) if $ent =~ /^#x([a-f0-9]+)$/; |
---|
| 230 | return char_to_utf8($ent{$ent}) if exists $ent{$ent}; |
---|
| 231 | $ent = "\&$ent;"; |
---|
| 232 | print STDERR "Warning: entity \"$ent\" unknown\n"; |
---|
| 233 | return $ent; |
---|
| 234 | } |
---|
| 235 | |
---|
| 236 | sub char_to_utf8 { |
---|
| 237 | my $unicode = shift; |
---|
| 238 | # ASCII is easy, and the most common case |
---|
| 239 | return chr($unicode) if $unicode < 0x80; |
---|
| 240 | |
---|
| 241 | my $result = ''; |
---|
| 242 | my $n = 0x20; |
---|
| 243 | while (1) { |
---|
| 244 | $result = chr(0x80 | ($unicode & 0x3f)) . $result; |
---|
| 245 | $unicode >>= 6; |
---|
| 246 | last if $unicode < $n; |
---|
| 247 | $n >>= 1; |
---|
| 248 | } |
---|
| 249 | $result = chr((0x100 - $n*2) | $unicode) . $result; |
---|
| 250 | return $result; |
---|
| 251 | } |
---|
[d5bd3a7] | 252 | |
---|