1 package Encode::Unicode;
6 our $VERSION = do { my @r = (q$Revision: 1.36 $ =~ /\d+/g); sprintf "%d."."%02d" x $#r, @r };
9 XSLoader::load(__PACKAGE__,$VERSION);
12 # Object Generator 8 transcoders all at once!
17 for my $name (qw(UTF-16 UTF-16BE UTF-16LE
18 UTF-32 UTF-32BE UTF-32LE
21 my ($size, $endian, $ucs2, $mask);
22 $name =~ /^(\w+)-(\d+)(\w*)$/o;
23 if ($ucs2 = ($1 eq 'UCS')){
28 $endian = ($3 eq 'BE') ? 'n' : ($3 eq 'LE') ? 'v' : '' ;
29 $size == 4 and $endian = uc($endian);
31 $Encode::Encoding{$name} =
41 use base qw(Encode::Encoding);
44 # three implementations of (en|de)code exist. The XS version is the
45 # fastest. *_modern uses an array and *_classic sticks with substr.
46 # *_classic is much slower but more memory conservative.
47 # *_xs is the default.
50 no warnings qw(redefine);
53 *decode = \&decode_xs;
54 *encode = \&encode_xs;
55 }elsif($type eq "modern"){
56 *decode = \&decode_modern;
57 *encode = \&encode_modern;
58 }elsif($type eq "classic"){
59 *decode = \&decode_classic;
60 *encode = \&encode_classic;
63 Carp::croak __PACKAGE__, "::set_transcoder(modern|classic|xs)";
70 # Aux. subs & constants
73 sub FBCHAR(){ 0xFFFd }
74 sub BOM_BE(){ 0xFeFF }
75 sub BOM16LE(){ 0xFFFe }
76 sub BOM32LE(){ 0xFFFe0000 }
80 (0 <= $_[0] && $_[0] < 0xD800)
81 || ( 0xDFFF < $_[0] && $_[0] <= 0xFFFF);
84 sub issurrogate($){ 0xD800 <= $_[0] && $_[0] <= 0xDFFF }
85 sub isHiSurrogate($){ 0xD800 <= $_[0] && $_[0] < 0xDC00 }
86 sub isLoSurrogate($){ 0xDC00 <= $_[0] && $_[0] <= 0xDFFF }
89 use integer; # we have divisions
91 my $hi = ($uni - 0x10000) / 0x400 + 0xD800;
92 my $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
98 return 0x10000 + ($hi - 0xD800)*0x400 + ($lo - 0xDC00);
101 sub Mask { {2 => 0xffff, 4 => 0xffffffff} }
104 # *_modern are much faster but guzzle more memory
107 sub decode_modern($$;$)
109 my ($obj, $str, $chk ) = @_;
110 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
112 # warn "$size, $endian, $ucs2";
113 $endian ||= BOMB($size, substr($str, 0, $size, ''))
114 or poisoned2death($obj, "Where's the BOM?");
115 my $mask = Mask->{$size};
117 my @ord = unpack("$endian*", $str);
118 undef $str; # to conserve memory
120 my $ord = shift @ord;
121 unless ($size == 4 or valid_ucs2($ord &= $mask)){
124 poisoned2death($obj, "no surrogates allowed", $ord);
125 shift @ord; # skip the next one as well
128 unless (isHiSurrogate($ord)){
129 poisoned2death($obj, "Malformed HI surrogate", $ord);
132 unless (isLoSurrogate($lo &= $mask)){
133 poisoned2death($obj, "Malformed LO surrogate", $ord, $lo);
135 $ord = desurrogate($ord, $lo);
140 utf8::upgrade($utf8);
144 sub encode_modern($$;$)
146 my ($obj, $utf8, $chk) = @_;
147 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
150 $endian = ($size == 4) ? 'N' : 'n';
153 my @ord = unpack("U*", $utf8);
154 undef $utf8; # to conserve memory
156 unless ($size == 4 or valid_ucs2($ord)) {
157 unless(issurrogate($ord)){
160 poisoned2death($obj, "code point too high", $ord);
165 push @str, ensurrogate($ord);
167 }else{ # not supposed to happen
174 return pack("$endian*", @str);
178 # *_classic are slower but more memory conservative
181 sub decode_classic($$;$)
183 my ($obj, $str, $chk ) = @_;
184 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
186 # warn "$size, $endian, $ucs2";
187 $endian ||= BOMB($size, substr($str, 0, $size, ''))
188 or poisoned2death($obj, "Where's the BOM?");
189 my $mask = Mask->{$size};
191 my @ord = unpack("$endian*", $str);
192 while (length($str)){
193 my $ord = unpack($endian, substr($str, 0, $size, ''));
194 unless ($size == 4 or valid_ucs2($ord &= $mask)){
197 poisoned2death($obj, "no surrogates allowed", $ord);
198 substr($str,0,$size,''); # skip the next one as well
201 unless (isHiSurrogate($ord)){
202 poisoned2death($obj, "Malformed HI surrogate", $ord);
204 my $lo = unpack($endian ,substr($str,0,$size,''));
205 unless (isLoSurrogate($lo &= $mask)){
206 poisoned2death($obj, "Malformed LO surrogate", $ord, $lo);
208 $ord = desurrogate($ord, $lo);
213 utf8::upgrade($utf8);
217 sub encode_classic($$;$)
219 my ($obj, $utf8, $chk) = @_;
220 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
221 # warn join ", ", $size, $ucs2, $endian, $mask;
224 $endian = ($size == 4) ? 'N' : 'n';
225 $str .= pack($endian, BOM_BE);
227 while (length($utf8)){
228 my $ord = ord(substr($utf8,0,1,''));
229 unless ($size == 4 or valid_ucs2($ord)) {
230 unless(issurrogate($ord)){
233 poisoned2death($obj, "code point too high", $ord);
234 $str .= pack($endian, FBCHAR);
236 $str .= pack($endian.2, ensurrogate($ord));
238 }else{ # not supposed to happen
239 $str .= pack($endian, FBCHAR);
242 $str .= pack($endian, $ord);
249 my ($size, $bom) = @_;
250 my $N = $size == 2 ? 'n' : 'N';
251 my $ord = unpack($N, $bom);
252 return ($ord eq BOM_BE) ? $N :
253 ($ord eq BOM16LE) ? 'v' : ($ord eq BOM32LE) ? 'V' : undef;
259 my $pair = join(", ", map {sprintf "\\x%x", $_} @_);
261 Carp::croak $obj->name, ":", $msg, "<$pair>.", caller;
269 Encode::Unicode -- Various Unicode Transformation Formats
275 use Encode qw/encode decode/;
276 $ucs2 = encode("UCS-2BE", $utf8);
277 $utf8 = decode("UCS-2BE", $ucs2);
281 This module implements all Character Encoding Schemes of Unicode that
282 are officially documented by Unicode Consortium (except, of course,
283 for UTF-8, which is a native format in perl).
287 =item L<http://www.unicode.org/glossary/> says:
289 I<Character Encoding Scheme> A character encoding form plus byte
290 serialization. There are seven character encoding schemes in Unicode:
291 UTF-8, UTF-16, UTF-16BE, UTF-16LE, UTF-32, UTF-32BE and UTF-32LE.
293 =item Quick Reference
295 Decodes from ord(N) Encodes chr(N) to...
296 octet/char BOM S.P d800-dfff ord > 0xffff \x{1abcd} ==
297 ---------------+-----------------+------------------------------
298 UCS-2BE 2 N N is bogus Not Available
299 UCS-2LE 2 N N bogus Not Available
300 UTF-16 2/4 Y Y is S.P S.P BE/LE
301 UTF-16BE 2/4 N Y S.P S.P 0xd82a,0xdfcd
302 UTF-16LE 2 N Y S.P S.P 0x2ad8,0xcddf
303 UTF-32 4 Y - is bogus As is BE/LE
304 UTF-32BE 4 N - bogus As is 0x0001abcd
305 UTF-32LE 4 N - bogus As is 0xcdab0100
306 UTF-8 1-4 - - bogus >= 4 octets \xf0\x9a\af\8d
307 ---------------+-----------------+------------------------------
311 =head1 Size, Endianness, and BOM
313 You can categorize these CES by 3 criteria: size of each character,
314 endianness, and Byte Order Mark.
318 UCS-2 is a fixed-length encoding with each character taking 16 bits.
319 It B<does not> support I<surrogate pairs>. When a surrogate pair
320 is encountered during decode(), its place is filled with \x{FFFD}
321 if I<CHECK> is 0, or the routine croaks if I<CHECK> is 1. When a
322 character whose ord value is larger than 0xFFFF is encountered,
323 its place is filled with \x{FFFD} if I<CHECK> is 0, or the routine
324 croaks if I<CHECK> is 1.
326 UTF-16 is almost the same as UCS-2 but it supports I<surrogate pairs>.
327 When it encounters a high surrogate (0xD800-0xDBFF), it fetches the
328 following low surrogate (0xDC00-0xDFFF) and C<desurrogate>s them to
329 form a character. Bogus surrogates result in death. When \x{10000}
330 or above is encountered during encode(), it C<ensurrogate>s them and
331 pushes the surrogate pair to the output stream.
333 UTF-32 is a fixed-length encoding with each character taking 32 bits.
334 Since it is 32-bit, there is no need for I<surrogate pairs>.
338 The first (and now failed) goal of Unicode was to map all character
339 repertoires into a fixed-length integer so that programmers are happy.
340 Since each character is either a I<short> or I<long> in C, you have to
341 pay attention to the endianness of each platform when you pass data
344 Anything marked as BE is Big Endian (or network byte order) and LE is
345 Little Endian (aka VAX byte order). For anything not marked either
346 BE or LE, a character called Byte Order Mark (BOM) indicating the
347 endianness is prepended to the string.
351 =item BOM as integer when fetched in network byte order
354 -------------------------
357 -------------------------
361 This modules handles the BOM as follows.
367 When BE or LE is explicitly stated as the name of encoding, BOM is
368 simply treated as a normal character (ZERO WIDTH NO-BREAK SPACE).
372 When BE or LE is omitted during decode(), it checks if BOM is at the
373 beginning of the string; if one is found, the endianness is set to
374 what the BOM says. If no BOM is found, the routine dies.
378 When BE or LE is omitted during encode(), it returns a BE-encoded
379 string with BOM prepended. So when you want to encode a whole text
380 file, make sure you encode() the whole text at once, not line by line
381 or each line, not file, will have a BOM prepended.
385 C<UCS-2> is an exception. Unlike others, this is an alias of UCS-2BE.
386 UCS-2 is already registered by IANA and others that way.
390 =head1 Surrogate Pairs
392 To say the least, surrogate pairs were the biggest mistake of the
393 Unicode Consortium. But according to the late Douglas Adams in I<The
394 Hitchhiker's Guide to the Galaxy> Trilogy, C<In the beginning the
395 Universe was created. This has made a lot of people very angry and
396 been widely regarded as a bad move>. Their mistake was not of this
397 magnitude so let's forgive them.
399 (I don't dare make any comparison with Unicode Consortium and the
400 Vogons here ;) Or, comparing Encode to Babel Fish is completely
401 appropriate -- if you can only stick this into your ear :)
403 Surrogate pairs were born when the Unicode Consortium finally
404 admitted that 16 bits were not big enough to hold all the world's
405 character repertoires. But they already made UCS-2 16-bit. What
408 Back then, the range 0xD800-0xDFFF was not allocated. Let's split
409 that range in half and use the first half to represent the C<upper
410 half of a character> and the second half to represent the C<lower
411 half of a character>. That way, you can represent 1024 * 1024 =
412 1048576 more characters. Now we can store character ranges up to
413 \x{10ffff} even with 16-bit encodings. This pair of half-character is
414 now called a I<surrogate pair> and UTF-16 is the name of the encoding
417 Here is a formula to ensurrogate a Unicode character \x{10000} and
420 $hi = ($uni - 0x10000) / 0x400 + 0xD800;
421 $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
425 $uni = 0x10000 + ($hi - 0xD800) * 0x400 + ($lo - 0xDC00);
427 Note this move has made \x{D800}-\x{DFFF} into a forbidden zone but
428 perl does not prohibit the use of characters within this range. To perl,
429 every one of \x{0000_0000} up to \x{ffff_ffff} (*) is I<a character>.
431 (*) or \x{ffff_ffff_ffff_ffff} if your perl is compiled with 64-bit
436 L<Encode>, L<http://www.unicode.org/glossary/>,
438 RFC 2781 L<http://rfc.net/rfc2781.html>,
440 L<http://www.unicode.org/unicode/faq/utf_bom.html>
442 Ch. 15, pp. 403 of C<Programming Perl (3rd Edition)>
443 by Larry Wall, Tom Christiansen, Jon Orwant;
444 O'Reilly & Associates; ISBN 0-596-00027-8