1 | #!/usr/bin/perl -w |
---|
2 | require 5.008; |
---|
3 | use bytes; |
---|
4 | use strict; |
---|
5 | |
---|
6 | use integer; |
---|
7 | |
---|
8 | # messages >= this value are written to a header file |
---|
9 | my $dontextract_threshold = 1000; |
---|
10 | |
---|
11 | # Magic identifier (12 bytes) |
---|
12 | my $magic = "Svx\nMsg\r\n\xfe\xff\0"; |
---|
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 $minor = 8; |
---|
19 | |
---|
20 | # File format (multi-byte numbers in network order (bigendian)): |
---|
21 | # 12 bytes: Magic identifier |
---|
22 | # 1 byte: File format major version number (0) |
---|
23 | # 1 byte: File format minor version number (8) |
---|
24 | # 2 bytes: Number of messages (N) |
---|
25 | # 4 bytes: Offset from XXXX to end of file |
---|
26 | # XXXX: |
---|
27 | # N*: |
---|
28 | # <message> NUL |
---|
29 | |
---|
30 | my %ent = (); |
---|
31 | |
---|
32 | open ENT, ($ENV{srcdir}?"$ENV{srcdir}/":"")."named-entities.txt" or die $!; |
---|
33 | while (<ENT>) { |
---|
34 | my ($e, $v) = /^(\w+),(\d+)/; |
---|
35 | $ent{$e} = $v; |
---|
36 | } |
---|
37 | close ENT; |
---|
38 | |
---|
39 | my %msgs = (); |
---|
40 | my %dontextract = (); |
---|
41 | |
---|
42 | my %raw = (); |
---|
43 | my $raw = ''; |
---|
44 | my $raw_comments = ''; |
---|
45 | my $curmsg = -1; |
---|
46 | |
---|
47 | while (<>) { |
---|
48 | if (/^\s*#/) { |
---|
49 | $raw_comments .= $_; |
---|
50 | next; # skip comments |
---|
51 | } |
---|
52 | |
---|
53 | # en: 0 0.81 the message |
---|
54 | # en-us: 0 0.81 " the message " |
---|
55 | my ($langs, $msgno, $dummy, $msg) = /^([-\w,]+):\s*(\d+)\s+("?)(.*)\3/; |
---|
56 | |
---|
57 | if ($msgno != $curmsg) { |
---|
58 | if ($msgno < $curmsg) { |
---|
59 | print STDERR "Warning: message number jumps back from $curmsg to $msgno\n"; |
---|
60 | } |
---|
61 | $raw{$curmsg} = $raw; |
---|
62 | $raw = ''; |
---|
63 | $curmsg = $msgno; |
---|
64 | } |
---|
65 | |
---|
66 | $raw .= $raw_comments . $_; |
---|
67 | $raw_comments = ''; |
---|
68 | |
---|
69 | unless (defined $langs) { |
---|
70 | chomp; |
---|
71 | print STDERR "Warning: Bad line: \"$_\"\n"; |
---|
72 | next; |
---|
73 | } |
---|
74 | |
---|
75 | $langs =~ tr/-/_/; |
---|
76 | |
---|
77 | if ($msg =~ /[\0-\x1f\x7f-\xff]/) { |
---|
78 | print STDERR "Warning: literal character in message $msgno\n"; |
---|
79 | } |
---|
80 | |
---|
81 | my $utf8 = string_to_utf8($msg); |
---|
82 | for (split /,/, $langs) { |
---|
83 | if ($msgno >= $dontextract_threshold) { |
---|
84 | if (${$dontextract{$_}}[$msgno - $dontextract_threshold]) { |
---|
85 | print STDERR "Warning: already had message $msgno for language $_\n"; |
---|
86 | } |
---|
87 | ${$dontextract{$_}}[$msgno - $dontextract_threshold] = $utf8; |
---|
88 | } else { |
---|
89 | if (${$msgs{$_}}[$msgno]) { |
---|
90 | print STDERR "Warning: already had message $msgno for language $_\n"; |
---|
91 | } |
---|
92 | ${$msgs{$_}}[$msgno] = $utf8; |
---|
93 | } |
---|
94 | } |
---|
95 | } |
---|
96 | $raw{$curmsg} = $raw; |
---|
97 | |
---|
98 | my $lang; |
---|
99 | my @langs = sort grep ! /_\*$/, keys %msgs; |
---|
100 | |
---|
101 | my $num_msgs = -1; |
---|
102 | foreach $lang (@langs) { |
---|
103 | my $aref = $msgs{$lang}; |
---|
104 | $num_msgs = scalar @$aref if scalar @$aref > $num_msgs; |
---|
105 | unlink "$lang.todo"; |
---|
106 | } |
---|
107 | |
---|
108 | foreach $lang (@langs) { |
---|
109 | my $fnm = $lang; |
---|
110 | $fnm =~ s/(_.*)$/\U$1/; |
---|
111 | open OUT, ">$fnm.msg" or die $!; |
---|
112 | |
---|
113 | my $aref = $msgs{$lang}; |
---|
114 | |
---|
115 | my $parentaref; |
---|
116 | my $mainlang = $lang; |
---|
117 | $parentaref = $msgs{$mainlang} if $mainlang =~ s/_.*$//; |
---|
118 | |
---|
119 | print OUT $magic or die $!; |
---|
120 | print OUT pack("CCn", $major, $minor, $num_msgs) or die $!; |
---|
121 | |
---|
122 | my $buff = ''; |
---|
123 | |
---|
124 | my $n; |
---|
125 | for $n (0 .. $num_msgs - 1) { |
---|
126 | my $warned = 0; |
---|
127 | my $msg = $$aref[$n]; |
---|
128 | if (!defined $msg) { |
---|
129 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
130 | if (!defined $msg) { |
---|
131 | $msg = ${$msgs{'en'}}[$n]; |
---|
132 | # don't report if we have a parent (as the omission will be |
---|
133 | # reported there) |
---|
134 | if (defined $msg && $msg ne '' && !defined $parentaref) { |
---|
135 | print STDERR "Warning: message $n not in language $lang\n"; |
---|
136 | open TODO, ">>$lang.todo" or die $!; |
---|
137 | print TODO $raw{$n}, "\n"; |
---|
138 | close TODO; |
---|
139 | $warned = 1; |
---|
140 | } else { |
---|
141 | $msg = '' if !defined $msg; |
---|
142 | } |
---|
143 | } |
---|
144 | } |
---|
145 | if (defined($raw{$n}) |
---|
146 | && $raw{$n} =~ /^\s*#\s*TRANSLATE\b[- \ta-z]*\b$lang\b/m) { |
---|
147 | if ($warned) { |
---|
148 | print STDERR "Warning: message $n missing and also marked for retranslation for language $lang\n"; |
---|
149 | } else { |
---|
150 | print STDERR "Warning: message $n needs retranslating for language $lang\n"; |
---|
151 | open TODO, ">>$lang.todo" or die $!; |
---|
152 | print TODO $raw{$n}, "\n"; |
---|
153 | close TODO; |
---|
154 | } |
---|
155 | } |
---|
156 | $buff .= $msg . "\0"; |
---|
157 | } |
---|
158 | |
---|
159 | print OUT pack('N',length($buff)), $buff or die $!; |
---|
160 | close OUT or die $!; |
---|
161 | } |
---|
162 | |
---|
163 | my $num_dontextract = -1; |
---|
164 | foreach $lang (@langs) { |
---|
165 | my $aref = $dontextract{$lang}; |
---|
166 | if (defined(@$aref)) { |
---|
167 | $num_dontextract = scalar @$aref if scalar @$aref > $num_dontextract; |
---|
168 | } |
---|
169 | } |
---|
170 | |
---|
171 | foreach $lang (@langs) { |
---|
172 | my $fnm = $lang; |
---|
173 | $fnm =~ s/(_.*)$/\U$1/; |
---|
174 | open OUT, ">$fnm.h" or die $!; |
---|
175 | print OUT "#define N_DONTEXTRACTMSGS ", $num_dontextract, "\n"; |
---|
176 | print OUT "static unsigned char dontextractmsgs[] ="; |
---|
177 | |
---|
178 | for my $n (0 .. $num_dontextract - 1) { |
---|
179 | print OUT "\n"; |
---|
180 | |
---|
181 | my $aref = $dontextract{$lang}; |
---|
182 | |
---|
183 | my $parentaref; |
---|
184 | my $mainlang = $lang; |
---|
185 | $parentaref = $dontextract{$mainlang} if $mainlang =~ s/_.*$//; |
---|
186 | |
---|
187 | my $msg = $$aref[$n]; |
---|
188 | if (!defined $msg) { |
---|
189 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
190 | if (!defined $msg) { |
---|
191 | $msg = ${$dontextract{'en'}}[$n]; |
---|
192 | # don't report if we have a parent (as the omission will be reported there) |
---|
193 | if (defined $msg && $msg ne '' && !defined $parentaref) { |
---|
194 | print STDERR "Warning: message ", $dontextract_threshold + $n, " not in language $lang\n"; |
---|
195 | open TODO, ">>$lang.todo" or die $!; |
---|
196 | print TODO $raw{$dontextract_threshold + $n}, "\n"; |
---|
197 | close TODO; |
---|
198 | } else { |
---|
199 | $msg = ''; |
---|
200 | } |
---|
201 | } |
---|
202 | } |
---|
203 | $msg =~ s/\\/\\\\/g; |
---|
204 | $msg =~ s/"/\\"/g; |
---|
205 | $msg =~ s/\t/\\t/g; |
---|
206 | $msg =~ s/\n/\\n/g; |
---|
207 | $msg =~ s/\r/\\r/g; |
---|
208 | if ($msg =~ /^ / || $msg =~ / $/) { |
---|
209 | $msg =~ s/\\"/\\\\\\"/g; |
---|
210 | $msg = '\\"'.$msg.'\\"'; |
---|
211 | } |
---|
212 | print OUT " /*", $dontextract_threshold + $n, "*/ \"$msg\\0\""; |
---|
213 | } |
---|
214 | print OUT ";\n"; |
---|
215 | close OUT or die $!; |
---|
216 | } |
---|
217 | |
---|
218 | #for (sort {$a<=>$b} keys %raw) {print "$_ = [$raw{$_}]\n";} |
---|
219 | |
---|
220 | sub string_to_utf8 { |
---|
221 | my $s = shift; |
---|
222 | $s =~ s/([\x80-\xff])/char_to_utf8(ord($1))/eg; |
---|
223 | $s =~ s/\&(#\d+|#x[a-f0-9]+|[a-z0-9]+);?/decode_entity($1)/eig; |
---|
224 | return $s; |
---|
225 | } |
---|
226 | |
---|
227 | sub decode_entity { |
---|
228 | my $ent = shift; |
---|
229 | return char_to_utf8($1) if $ent =~ /^#(\d+)$/; |
---|
230 | return char_to_utf8(hex($1)) if $ent =~ /^#x([a-f0-9]+)$/; |
---|
231 | return char_to_utf8($ent{$ent}) if exists $ent{$ent}; |
---|
232 | $ent = "\&$ent;"; |
---|
233 | print STDERR "Warning: entity \"$ent\" unknown\n"; |
---|
234 | return $ent; |
---|
235 | } |
---|
236 | |
---|
237 | sub char_to_utf8 { |
---|
238 | my $unicode = shift; |
---|
239 | # ASCII is easy, and the most common case |
---|
240 | return chr($unicode) if $unicode < 0x80; |
---|
241 | |
---|
242 | my $result = ''; |
---|
243 | my $n = 0x20; |
---|
244 | while (1) { |
---|
245 | $result = chr(0x80 | ($unicode & 0x3f)) . $result; |
---|
246 | $unicode >>= 6; |
---|
247 | last if $unicode < $n; |
---|
248 | $n >>= 1; |
---|
249 | } |
---|
250 | $result = chr((0x100 - $n*2) | $unicode) . $result; |
---|
251 | return $result; |
---|
252 | } |
---|
253 | |
---|