1 | #!/usr/bin/perl -w |
---|
2 | require 5.004; |
---|
3 | use strict; |
---|
4 | |
---|
5 | use integer; |
---|
6 | |
---|
7 | # messages >= this value are written to a header file |
---|
8 | my $dontextract_threshold = 1000; |
---|
9 | |
---|
10 | # Magic identifier (12 bytes) |
---|
11 | my $magic = "Svx\nMsg\r\n\xfe\xff\0"; |
---|
12 | # Designed to be corrupted by ASCII ftp, top bit stripping (or |
---|
13 | # being used for parity). Contains a zero byte so more likely |
---|
14 | # to be flagged as data (e.g. by perl's "-B" test). |
---|
15 | |
---|
16 | my $major = 0; |
---|
17 | my $minor = 8; |
---|
18 | |
---|
19 | # File format (multi-byte numbers in network order (bigendian)): |
---|
20 | # 12 bytes: Magic identifier |
---|
21 | # 1 byte: File format major version number (0) |
---|
22 | # 1 byte: File format minor version number (8) |
---|
23 | # 2 bytes: Number of messages (N) |
---|
24 | # 4 bytes: Offset from XXXX to end of file |
---|
25 | # XXXX: |
---|
26 | # N*: |
---|
27 | # <message> NUL |
---|
28 | |
---|
29 | my %ent = (); |
---|
30 | |
---|
31 | open ENT, "named-entities.txt" or die $!; |
---|
32 | while (<ENT>) { |
---|
33 | my ($e, $v) = /^(\w+),(\d+)/; |
---|
34 | $ent{$e} = $v; |
---|
35 | } |
---|
36 | close ENT; |
---|
37 | |
---|
38 | my %msgs = (); |
---|
39 | my %dontextract = (); |
---|
40 | |
---|
41 | while (<>) { |
---|
42 | next if /^\s*#/; # skip comments |
---|
43 | |
---|
44 | # en: 0 0.81 the message |
---|
45 | # en-us: 0 0.81 " the message " |
---|
46 | my ($langs, $msgno, $dummy, $msg) = /^([-\w,]+):\s*(\d+)\s+("?)(.*)\3/; |
---|
47 | |
---|
48 | unless (defined $langs) { |
---|
49 | chomp; |
---|
50 | print STDERR "Warning: Bad line: \"$_\"\n"; |
---|
51 | next; |
---|
52 | } |
---|
53 | |
---|
54 | $langs =~ tr/-/_/; |
---|
55 | |
---|
56 | if ($msg =~ /[\0-\x1f\x7f-\xff]/) { |
---|
57 | print STDERR "Warning: literal character in message $msgno\n"; |
---|
58 | } |
---|
59 | |
---|
60 | my $utf8 = string_to_utf8($msg); |
---|
61 | for (split /,/, $langs) { |
---|
62 | if ($msgno >= $dontextract_threshold) { |
---|
63 | ${$dontextract{$_}}[$msgno - $dontextract_threshold] = $utf8; |
---|
64 | } else { |
---|
65 | ${$msgs{$_}}[$msgno] = $utf8; |
---|
66 | } |
---|
67 | } |
---|
68 | } |
---|
69 | |
---|
70 | my $lang; |
---|
71 | my @langs = sort grep ! /_\*$/, keys %msgs; |
---|
72 | |
---|
73 | my $num_msgs = -1; |
---|
74 | foreach $lang (@langs) { |
---|
75 | my $aref = $msgs{$lang}; |
---|
76 | $num_msgs = scalar @$aref if scalar @$aref > $num_msgs; |
---|
77 | } |
---|
78 | |
---|
79 | foreach $lang (@langs) { |
---|
80 | my $fnm = $lang; |
---|
81 | $fnm =~ s/(_.*)$/\U$1/; |
---|
82 | open OUT, ">$fnm.msg" or die $!; |
---|
83 | |
---|
84 | my $aref = $msgs{$lang}; |
---|
85 | |
---|
86 | my $parentaref; |
---|
87 | my $mainlang = $lang; |
---|
88 | $parentaref = $msgs{$mainlang} if $mainlang =~ s/_.*$//; |
---|
89 | |
---|
90 | print OUT $magic or die $!; |
---|
91 | print OUT pack("CCn", $major, $minor, $num_msgs) or die $!; |
---|
92 | |
---|
93 | my $buff = ''; |
---|
94 | |
---|
95 | my $n; |
---|
96 | for $n (0 .. $num_msgs - 1) { |
---|
97 | my $msg = $$aref[$n]; |
---|
98 | if (!defined $msg) { |
---|
99 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
100 | if (!defined $msg) { |
---|
101 | $msg = ${$msgs{'en'}}[$n]; |
---|
102 | if (defined $msg && $msg ne '') { |
---|
103 | # don't report if we have a parent (as the omission will be reported there) |
---|
104 | print STDERR "Warning: message $n not in language $lang\n" unless defined $parentaref; |
---|
105 | } else { |
---|
106 | $msg = ''; |
---|
107 | } |
---|
108 | } |
---|
109 | } |
---|
110 | $buff .= $msg . "\0"; |
---|
111 | } |
---|
112 | |
---|
113 | print OUT pack('N',length($buff)), $buff or die $!; |
---|
114 | close OUT or die $!; |
---|
115 | } |
---|
116 | |
---|
117 | my $num_dontextract = -1; |
---|
118 | foreach $lang (@langs) { |
---|
119 | my $aref = $dontextract{$lang}; |
---|
120 | if (defined(@$aref)) { |
---|
121 | $num_dontextract = scalar @$aref if scalar @$aref > $num_dontextract; |
---|
122 | } |
---|
123 | } |
---|
124 | |
---|
125 | foreach $lang (@langs) { |
---|
126 | my $fnm = $lang; |
---|
127 | $fnm =~ s/(_.*)$/\U$1/; |
---|
128 | open OUT, ">$fnm.h" or die $!; |
---|
129 | print OUT "#define N_DONTEXTRACTMSGS ", $num_dontextract, "\n"; |
---|
130 | print OUT "static unsigned char dontextractmsgs[] ="; |
---|
131 | |
---|
132 | for my $n (0 .. $num_dontextract - 1) { |
---|
133 | print OUT "\n"; |
---|
134 | |
---|
135 | my $aref = $dontextract{$lang}; |
---|
136 | |
---|
137 | my $parentaref; |
---|
138 | my $mainlang = $lang; |
---|
139 | $parentaref = $dontextract{$mainlang} if $mainlang =~ s/_.*$//; |
---|
140 | |
---|
141 | my $msg = $$aref[$n]; |
---|
142 | if (!defined $msg) { |
---|
143 | $msg = $$parentaref[$n] if defined $parentaref; |
---|
144 | if (!defined $msg) { |
---|
145 | $msg = ${$dontextract{'en'}}[$n]; |
---|
146 | if (defined $msg && $msg ne '') { |
---|
147 | # don't report if we have a parent (as the omission will be reported there) |
---|
148 | print STDERR "Warning: message ", $dontextract_threshold + $n, " not in language $lang\n" unless defined $parentaref; |
---|
149 | } else { |
---|
150 | $msg = ''; |
---|
151 | } |
---|
152 | } |
---|
153 | } |
---|
154 | $msg =~ s/\\/\\\\/g; |
---|
155 | $msg =~ s/"/\\"/g; |
---|
156 | $msg =~ s/\t/\\t/g; |
---|
157 | $msg =~ s/\n/\\n/g; |
---|
158 | $msg =~ s/\r/\\r/g; |
---|
159 | if ($msg =~ /^ / || $msg =~ / $/) { |
---|
160 | $msg =~ s/\\"/\\\\\\"/g; |
---|
161 | $msg = '\\"'.$msg.'\\"'; |
---|
162 | } |
---|
163 | print OUT " /*", $dontextract_threshold + $n, "*/ \"$msg\\0\""; |
---|
164 | } |
---|
165 | print OUT ";\n"; |
---|
166 | close OUT or die $!; |
---|
167 | } |
---|
168 | |
---|
169 | sub string_to_utf8 { |
---|
170 | my $s = shift; |
---|
171 | $s =~ s/([\x80-\xff])/char_to_utf8(ord($1))/eg; |
---|
172 | $s =~ s/\&(#\d+|#x[a-f0-9]+|[a-z0-9]+);?/decode_entity($1)/eig; |
---|
173 | return $s; |
---|
174 | } |
---|
175 | |
---|
176 | sub decode_entity { |
---|
177 | my $ent = shift; |
---|
178 | return char_to_utf8($1) if $ent =~ /^#(\d+)$/; |
---|
179 | return char_to_utf8(hex($1)) if $ent =~ /^#x([a-f0-9]+)$/; |
---|
180 | return char_to_utf8($ent{$ent}) if exists $ent{$ent}; |
---|
181 | $ent = "\&$ent;"; |
---|
182 | print STDERR "Warning: entity \"$ent\" unknown\n"; |
---|
183 | return $ent; |
---|
184 | } |
---|
185 | |
---|
186 | sub char_to_utf8 { |
---|
187 | my $unicode = shift; |
---|
188 | # ASCII is easy, and the most common case |
---|
189 | return chr($unicode) if $unicode < 0x80; |
---|
190 | |
---|
191 | my $result = ''; |
---|
192 | my $n = 0x20; |
---|
193 | while (1) { |
---|
194 | $result = chr(0x80 | ($unicode & 0x3f)) . $result; |
---|
195 | $unicode >>= 6; |
---|
196 | last if $unicode < $n; |
---|
197 | $n >>= 1; |
---|
198 | } |
---|
199 | $result = chr((0x100 - $n*2) | $unicode) . $result; |
---|
200 | return $result; |
---|
201 | } |
---|
202 | |
---|