1 package Encode::Unicode;
6 our $VERSION = do { my @r = (q$Revision: 1.35 $ =~ /\d+/g); sprintf "%d."."%02d" x $#r, @r };
9 XSLoader::load(__PACKAGE__,$VERSION);
12 # Object Generator 8 transcoders all at once!
16 for my $name (qw(UTF-16 UTF-16BE UTF-16LE
17 UTF-32 UTF-32BE UTF-32LE
20 my ($size, $endian, $ucs2, $mask);
21 $name =~ /^(\w+)-(\d+)(\w*)$/o;
22 if ($ucs2 = ($1 eq 'UCS')){
27 $endian = ($3 eq 'BE') ? 'n' : ($3 eq 'LE') ? 'v' : '' ;
28 $size == 4 and $endian = uc($endian);
30 $Encode::Encoding{$name} =
40 sub name { shift->{'Name'} }
44 # Return the original if endian known
45 return $self if ($self->{endian});
47 return bless {%$self},ref($self);
50 sub needs_lines { 0 };
53 eval{ require PerlIO::encoding };
63 # three implementations of (en|de)code exist. The XS version is the
64 # fastest. *_modern uses an array and *_classic sticks with substr.
65 # *_classic is much slower but more memory conservative.
66 # *_xs is the default.
69 no warnings qw(redefine);
72 *decode = \&decode_xs;
73 *encode = \&encode_xs;
74 }elsif($type eq "modern"){
75 *decode = \&decode_modern;
76 *encode = \&encode_modern;
77 }elsif($type eq "classic"){
78 *decode = \&decode_classic;
79 *encode = \&encode_classic;
82 Carp::croak __PACKAGE__, "::set_transcoder(modern|classic|xs)";
89 # Aux. subs & constants
92 sub FBCHAR(){ 0xFFFd }
93 sub BOM_BE(){ 0xFeFF }
94 sub BOM16LE(){ 0xFFFe }
95 sub BOM32LE(){ 0xFFFe0000 }
99 (0 <= $_[0] && $_[0] < 0xD800)
100 || ( 0xDFFF < $_[0] && $_[0] <= 0xFFFF);
103 sub issurrogate($){ 0xD800 <= $_[0] && $_[0] <= 0xDFFF }
104 sub isHiSurrogate($){ 0xD800 <= $_[0] && $_[0] < 0xDC00 }
105 sub isLoSurrogate($){ 0xDC00 <= $_[0] && $_[0] <= 0xDFFF }
108 use integer; # we have divisions
110 my $hi = ($uni - 0x10000) / 0x400 + 0xD800;
111 my $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
117 return 0x10000 + ($hi - 0xD800)*0x400 + ($lo - 0xDC00);
120 sub Mask { {2 => 0xffff, 4 => 0xffffffff} }
123 # *_modern are much faster but guzzle more memory
126 sub decode_modern($$;$)
128 my ($obj, $str, $chk ) = @_;
129 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
131 # warn "$size, $endian, $ucs2";
132 $endian ||= BOMB($size, substr($str, 0, $size, ''))
133 or poisoned2death($obj, "Where's the BOM?");
134 my $mask = Mask->{$size};
136 my @ord = unpack("$endian*", $str);
137 undef $str; # to conserve memory
139 my $ord = shift @ord;
140 unless ($size == 4 or valid_ucs2($ord &= $mask)){
143 poisoned2death($obj, "no surrogates allowed", $ord);
144 shift @ord; # skip the next one as well
147 unless (isHiSurrogate($ord)){
148 poisoned2death($obj, "Malformed HI surrogate", $ord);
151 unless (isLoSurrogate($lo &= $mask)){
152 poisoned2death($obj, "Malformed LO surrogate", $ord, $lo);
154 $ord = desurrogate($ord, $lo);
159 utf8::upgrade($utf8);
163 sub encode_modern($$;$)
165 my ($obj, $utf8, $chk) = @_;
166 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
169 $endian = ($size == 4) ? 'N' : 'n';
172 my @ord = unpack("U*", $utf8);
173 undef $utf8; # to conserve memory
175 unless ($size == 4 or valid_ucs2($ord)) {
176 unless(issurrogate($ord)){
179 poisoned2death($obj, "code point too high", $ord);
184 push @str, ensurrogate($ord);
186 }else{ # not supposed to happen
193 return pack("$endian*", @str);
197 # *_classic are slower but more memory conservative
200 sub decode_classic($$;$)
202 my ($obj, $str, $chk ) = @_;
203 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
205 # warn "$size, $endian, $ucs2";
206 $endian ||= BOMB($size, substr($str, 0, $size, ''))
207 or poisoned2death($obj, "Where's the BOM?");
208 my $mask = Mask->{$size};
210 my @ord = unpack("$endian*", $str);
211 while (length($str)){
212 my $ord = unpack($endian, substr($str, 0, $size, ''));
213 unless ($size == 4 or valid_ucs2($ord &= $mask)){
216 poisoned2death($obj, "no surrogates allowed", $ord);
217 substr($str,0,$size,''); # skip the next one as well
220 unless (isHiSurrogate($ord)){
221 poisoned2death($obj, "Malformed HI surrogate", $ord);
223 my $lo = unpack($endian ,substr($str,0,$size,''));
224 unless (isLoSurrogate($lo &= $mask)){
225 poisoned2death($obj, "Malformed LO surrogate", $ord, $lo);
227 $ord = desurrogate($ord, $lo);
232 utf8::upgrade($utf8);
236 sub encode_classic($$;$)
238 my ($obj, $utf8, $chk) = @_;
239 my ($size, $endian, $ucs2) = @$obj{qw(size endian ucs2)};
240 # warn join ", ", $size, $ucs2, $endian, $mask;
243 $endian = ($size == 4) ? 'N' : 'n';
244 $str .= pack($endian, BOM_BE);
246 while (length($utf8)){
247 my $ord = ord(substr($utf8,0,1,''));
248 unless ($size == 4 or valid_ucs2($ord)) {
249 unless(issurrogate($ord)){
252 poisoned2death($obj, "code point too high", $ord);
253 $str .= pack($endian, FBCHAR);
255 $str .= pack($endian.2, ensurrogate($ord));
257 }else{ # not supposed to happen
258 $str .= pack($endian, FBCHAR);
261 $str .= pack($endian, $ord);
268 my ($size, $bom) = @_;
269 my $N = $size == 2 ? 'n' : 'N';
270 my $ord = unpack($N, $bom);
271 return ($ord eq BOM_BE) ? $N :
272 ($ord eq BOM16LE) ? 'v' : ($ord eq BOM32LE) ? 'V' : undef;
278 my $pair = join(", ", map {sprintf "\\x%x", $_} @_);
280 Carp::croak $obj->name, ":", $msg, "<$pair>.", caller;
288 Encode::Unicode -- Various Unicode Transformation Formats
294 use Encode qw/encode decode/;
295 $ucs2 = encode("UCS-2BE", $utf8);
296 $utf8 = decode("UCS-2BE", $ucs2);
300 This module implements all Character Encoding Schemes of Unicode that
301 are officially documented by Unicode Consortium (except, of course,
302 for UTF-8, which is a native format in perl).
306 =item L<http://www.unicode.org/glossary/> says:
308 I<Character Encoding Scheme> A character encoding form plus byte
309 serialization. There are seven character encoding schemes in Unicode:
310 UTF-8, UTF-16, UTF-16BE, UTF-16LE, UTF-32, UTF-32BE and UTF-32LE.
312 =item Quick Reference
314 Decodes from ord(N) Encodes chr(N) to...
315 octet/char BOM S.P d800-dfff ord > 0xffff \x{1abcd} ==
316 ---------------+-----------------+------------------------------
317 UCS-2BE 2 N N is bogus Not Available
318 UCS-2LE 2 N N bogus Not Available
319 UTF-16 2/4 Y Y is S.P S.P BE/LE
320 UTF-16BE 2/4 N Y S.P S.P 0xd82a,0xdfcd
321 UTF-16LE 2 N Y S.P S.P 0x2ad8,0xcddf
322 UTF-32 4 Y - is bogus As is BE/LE
323 UTF-32BE 4 N - bogus As is 0x0001abcd
324 UTF-32LE 4 N - bogus As is 0xcdab0100
325 UTF-8 1-4 - - bogus >= 4 octets \xf0\x9a\af\8d
326 ---------------+-----------------+------------------------------
330 =head1 Size, Endianness, and BOM
332 You can categorize these CES by 3 criteria: size of each character,
333 endianness, and Byte Order Mark.
337 UCS-2 is a fixed-length encoding with each character taking 16 bits.
338 It B<does not> support I<surrogate pairs>. When a surrogate pair
339 is encountered during decode(), its place is filled with \x{FFFD}
340 if I<CHECK> is 0, or the routine croaks if I<CHECK> is 1. When a
341 character whose ord value is larger than 0xFFFF is encountered,
342 its place is filled with \x{FFFD} if I<CHECK> is 0, or the routine
343 croaks if I<CHECK> is 1.
345 UTF-16 is almost the same as UCS-2 but it supports I<surrogate pairs>.
346 When it encounters a high surrogate (0xD800-0xDBFF), it fetches the
347 following low surrogate (0xDC00-0xDFFF) and C<desurrogate>s them to
348 form a character. Bogus surrogates result in death. When \x{10000}
349 or above is encountered during encode(), it C<ensurrogate>s them and
350 pushes the surrogate pair to the output stream.
352 UTF-32 is a fixed-length encoding with each character taking 32 bits.
353 Since it is 32-bit, there is no need for I<surrogate pairs>.
357 The first (and now failed) goal of Unicode was to map all character
358 repertoires into a fixed-length integer so that programmers are happy.
359 Since each character is either a I<short> or I<long> in C, you have to
360 pay attention to the endianness of each platform when you pass data
363 Anything marked as BE is Big Endian (or network byte order) and LE is
364 Little Endian (aka VAX byte order). For anything not marked either
365 BE or LE, a character called Byte Order Mark (BOM) indicating the
366 endianness is prepended to the string.
370 =item BOM as integer when fetched in network byte order
373 -------------------------
376 -------------------------
380 This modules handles the BOM as follows.
386 When BE or LE is explicitly stated as the name of encoding, BOM is
387 simply treated as a normal character (ZERO WIDTH NO-BREAK SPACE).
391 When BE or LE is omitted during decode(), it checks if BOM is at the
392 beginning of the string; if one is found, the endianness is set to
393 what the BOM says. If no BOM is found, the routine dies.
397 When BE or LE is omitted during encode(), it returns a BE-encoded
398 string with BOM prepended. So when you want to encode a whole text
399 file, make sure you encode() the whole text at once, not line by line
400 or each line, not file, will have a BOM prepended.
404 C<UCS-2> is an exception. Unlike others, this is an alias of UCS-2BE.
405 UCS-2 is already registered by IANA and others that way.
409 =head1 Surrogate Pairs
411 To say the least, surrogate pairs were the biggest mistake of the
412 Unicode Consortium. But according to the late Douglas Adams in I<The
413 Hitchhiker's Guide to the Galaxy> Trilogy, C<In the beginning the
414 Universe was created. This has made a lot of people very angry and
415 been widely regarded as a bad move>. Their mistake was not of this
416 magnitude so let's forgive them.
418 (I don't dare make any comparison with Unicode Consortium and the
419 Vogons here ;) Or, comparing Encode to Babel Fish is completely
420 appropriate -- if you can only stick this into your ear :)
422 Surrogate pairs were born when the Unicode Consortium finally
423 admitted that 16 bits were not big enough to hold all the world's
424 character repertoires. But they already made UCS-2 16-bit. What
427 Back then, the range 0xD800-0xDFFF was not allocated. Let's split
428 that range in half and use the first half to represent the C<upper
429 half of a character> and the second half to represent the C<lower
430 half of a character>. That way, you can represent 1024 * 1024 =
431 1048576 more characters. Now we can store character ranges up to
432 \x{10ffff} even with 16-bit encodings. This pair of half-character is
433 now called a I<surrogate pair> and UTF-16 is the name of the encoding
436 Here is a formula to ensurrogate a Unicode character \x{10000} and
439 $hi = ($uni - 0x10000) / 0x400 + 0xD800;
440 $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
444 $uni = 0x10000 + ($hi - 0xD800) * 0x400 + ($lo - 0xDC00);
446 Note this move has made \x{D800}-\x{DFFF} into a forbidden zone but
447 perl does not prohibit the use of characters within this range. To perl,
448 every one of \x{0000_0000} up to \x{ffff_ffff} (*) is I<a character>.
450 (*) or \x{ffff_ffff_ffff_ffff} if your perl is compiled with 64-bit
455 L<Encode>, L<http://www.unicode.org/glossary/>,
457 RFC 2781 L<http://rfc.net/rfc2781.html>,
459 L<http://www.unicode.org/unicode/faq/utf_bom.html>
461 Ch. 15, pp. 403 of C<Programming Perl (3rd Edition)>
462 by Larry Wall, Tom Christiansen, Jon Orwant;
463 O'Reilly & Associates; ISBN 0-596-00027-8