3 perluniintro - Perl Unicode introduction
7 This document gives a general idea of Unicode and how to use Unicode
12 Unicode is a character set standard with plans to cover all of the
13 writing systems of the world, plus many other symbols.
15 Unicode and ISO/IEC 10646 are coordinated standards that provide code
16 points for the characters in almost all modern character set standards,
17 covering more than 30 writing systems and hundreds of languages,
18 including all commercially important modern languages. All characters
19 in the largest Chinese, Japanese, and Korean dictionaries are also
20 encoded. The standards will eventually cover almost all characters in
21 more than 250 writing systems and thousands of languages.
23 A Unicode I<character> is an abstract entity. It is not bound to any
24 particular integer width, and especially not to the C language C<char>.
25 Unicode is language neutral and display neutral: it doesn't encode the
26 language of the text, and it doesn't define fonts or other graphical
27 layout details. Unicode operates on characters and on text built from
30 Unicode defines characters like C<LATIN CAPITAL LETTER A> or C<GREEK
31 SMALL LETTER ALPHA>, and then unique numbers for those, hexadecimal
32 0x0041 or 0x03B1 for those particular characters. Such unique
33 numbers are called I<code points>.
35 The Unicode standard prefers using hexadecimal notation for the code
36 points. (In case this notation, numbers like 0x0041, is unfamiliar to
37 you, take a peek at a later section, L</"Hexadecimal Notation">.)
38 The Unicode standard uses the notation C<U+0041 LATIN CAPITAL LETTER A>,
39 which gives the hexadecimal code point, and the normative name of
42 Unicode also defines various I<properties> for the characters, like
43 "uppercase" or "lowercase", "decimal digit", or "punctuation":
44 these properties are independent of the names of the characters.
45 Furthermore, various operations on the characters like uppercasing,
46 lowercasing, and collating (sorting), are defined.
48 A Unicode character consists either of a single code point, or a
49 I<base character> (like C<LATIN CAPITAL LETTER A>), followed by one or
50 more I<modifiers> (like C<COMBINING ACUTE ACCENT>). This sequence of
51 a base character and modifiers is called a I<combining character
54 Whether to call these combining character sequences, as a whole,
55 "characters" depends on your point of view. If you are a programmer, you
56 probably would tend towards seeing each element in the sequences as one
57 unit, one "character", but from the user viewpoint, the sequence as a
58 whole is probably considered one "character", since that's probably what
59 it looks like in the context of the user's language.
61 With this "as a whole" view of characters, the number of characters is
62 open-ended. But in the programmer's "one unit is one character" point of
63 view, the concept of "characters" is more deterministic, and so we take
64 that point of view in this document: one "character" is one Unicode
65 code point, be it a base character or a combining character.
67 For some of the combinations there are I<precomposed> characters,
68 for example C<LATIN CAPITAL LETTER A WITH ACUTE> is defined as
69 a single code point. These precomposed characters are, however,
70 often available only for some combinations, and mainly they are
71 meant to support round-trip conversions between Unicode and legacy
72 standards (like the ISO 8859), and in general case the composing
73 method is more extensible. To support conversion between the
74 different compositions of the characters, various I<normalization
75 forms> are also defined.
77 Because of backward compatibility with legacy encodings, the "a unique
78 number for every character" breaks down a bit: "at least one number
79 for every character" is closer to truth. (This happens when the same
80 character has been encoded in several legacy encodings.) The converse
81 is also not true: not every code point has an assigned character.
82 Firstly, there are unallocated code points within otherwise used
83 blocks. Secondly, there are special Unicode control characters that
84 do not represent true characters.
86 A common myth about Unicode is that it would be "16-bit", that is,
87 0x10000 (or 65536) characters from 0x0000 to 0xFFFF. B<This is untrue.>
88 Since Unicode 2.0 Unicode has been defined all the way up to 21 bits
89 (0x10FFFF), and since 3.1 characters have been defined beyond 0xFFFF.
90 The first 0x10000 characters are called the I<Plane 0>, or the I<Basic
91 Multilingual Plane> (BMP). With the Unicode 3.1, 17 planes in all are
92 defined (but nowhere near full of defined characters yet).
94 Another myth is that the 256-character blocks have something to do
95 with languages: a block per language. B<Also this is untrue.>
96 The division into the blocks exists but it is almost completely
97 accidental, an artifact of how the characters have been historically
98 allocated. Instead, there is a concept called I<scripts>, which may
99 be more useful: there is C<Latin> script, C<Greek> script, and so on.
100 Scripts usually span several parts of several blocks. For further
101 information see L<Unicode::UCD>.
103 The Unicode code points are just abstract numbers. To input and
104 output these abstract numbers, the numbers must be I<encoded> somehow.
105 Unicode defines several I<character encoding forms>, of which I<UTF-8>
106 is perhaps the most popular. UTF-8 is a variable length encoding that
107 encodes Unicode characters as 1 to 6 bytes (only 4 with the currently
108 defined characters). Other encodings are UTF-16 and UTF-32 and their
109 big and little endian variants (UTF-8 is byteorder independent).
110 The ISO/IEC 10646 defines the UCS-2 and UCS-4 encoding forms.
112 For more information about encodings, for example to learn what
113 I<surrogates> and I<byte order marks> (BOMs) are, see L<perlunicode>.
115 =head2 Perl's Unicode Support
117 Starting from Perl 5.6.0, Perl has had the capability of handling
118 Unicode natively. The first recommended release for serious Unicode
119 work is Perl 5.8.0, however. The maintenance release 5.6.1 fixed many
120 of the problems of the initial implementation of Unicode, but for
121 example regular expressions didn't really work with Unicode.
123 B<Starting from Perl 5.8.0, the use of C<use utf8> is no longer
124 necessary.> In earlier releases the C<utf8> pragma was used to declare
125 that operations in the current block or file would be Unicode-aware.
126 This model was found to be wrong, or at least clumsy: the Unicodeness
127 is now carried with the data, not attached to the operations. (There
128 is one remaining case where an explicit C<use utf8> is needed: if your
129 Perl script is in UTF-8, you can use UTF-8 in your variable and
130 subroutine names, and in your string and regular expression literals,
131 by saying C<use utf8>. This is not the default because that would
132 break existing scripts having legacy 8-bit data in them.)
134 =head2 Perl's Unicode Model
136 Perl supports both the old, pre-5.6, model of strings of eight-bit
137 native bytes, and strings of Unicode characters. The principle is
138 that Perl tries to keep its data as eight-bit bytes for as long as
139 possible, but as soon as Unicodeness cannot be avoided, the data is
140 transparently upgraded to Unicode.
142 The internal encoding of Unicode in Perl is UTF-8. The internal
143 encoding is normally hidden, however, and one need not and should not
144 worry about the internal encoding at all: it is all just characters.
146 Perl 5.8.0 will also support Unicode on EBCDIC platforms. There the
147 support is somewhat harder to implement since additional conversions
148 are needed at every step. Because of these difficulties the Unicode
149 support won't be quite as full as in other, mainly ASCII-based,
150 platforms (the Unicode support will be better than in the 5.6 series,
151 which didn't work much at all for EBCDIC platform). On EBCDIC
152 platforms the internal encoding form used is UTF-EBCDIC.
154 =head2 Creating Unicode
156 To create Unicode literals, use the C<\x{...}> notation in
157 doublequoted strings:
159 my $smiley = "\x{263a}";
161 Similarly for regular expression literals
163 $smiley =~ /\x{263a}/;
165 At run-time you can use C<chr()>:
167 my $hebrew_alef = chr(0x05d0);
169 (See L</"Further Resources"> for how to find all these numeric codes.)
171 Naturally, C<ord()> will do the reverse: turn a character to a code point.
173 Note that C<\x..> (no C<{}> and only two hexadecimal digits), C<\x{...}>
174 and C<chr(...)> for arguments less than 0x100 (decimal 256) will
175 generate an eight-bit character for backward compatibility with older
176 Perls. For arguments of 0x100 or more, Unicode will always be
177 produced. If you want UTF-8 always, use C<pack("U", ...)> instead of
178 C<\x..>, C<\x{...}>, or C<chr()>.
180 You can also use the C<charnames> pragma to invoke characters
181 by name in doublequoted strings:
183 use charnames ':full';
184 my $arabic_alef = "\N{ARABIC LETTER ALEF}";
186 And, as mentioned above, you can also C<pack()> numbers into Unicode
189 my $georgian_an = pack("U", 0x10a0);
191 Note that both C<\x{...}> and C<\N{...}> are compile-time string
192 constants: you cannot use variables in them. if you want similar
193 run-time functionality, use C<chr()> and C<charnames::vianame()>.
195 =head2 Handling Unicode
197 Handling Unicode is for the most part transparent: just use the
198 strings as usual. Functions like C<index()>, C<length()>, and
199 C<substr()> will work on the Unicode characters; regular expressions
200 will work on the Unicode characters (see L<perlunicode> and L<perlretut>).
202 Note that Perl does B<not> consider combining character sequences
203 to be characters, such for example
205 use charnames ':full';
206 print length("\N{LATIN CAPITAL LETTER A}\N{COMBINING ACUTE ACCENT}"), "\n";
208 will print 2, not 1. The only exception is that regular expressions
209 have C<\X> for matching a combining character sequence.
211 When life is not quite so transparent is working with legacy
212 encodings, and I/O, and certain special cases.
214 =head2 Legacy Encodings
216 When you combine legacy data and Unicode the legacy data needs
217 to be upgraded to Unicode. Normally ISO 8859-1 (or EBCDIC, if
218 applicable) is assumed. You can override this assumption by
219 using the C<encoding> pragma, for example
221 use encoding 'latin2'; # ISO 8859-2
223 in which case literals (string or regular expression) and chr/ord
224 in your whole script are assumed to produce Unicode characters from
225 ISO 8859-2 code points. Note that the matching for the encoding
226 names is forgiving: instead of C<latin2> you could have said
227 C<Latin 2>, or C<iso8859-2>, and so forth. With just
231 first the environment variable C<PERL_ENCODING> will be consulted,
232 and if that doesn't exist, ISO 8859-1 (Latin 1) will be assumed.
234 The C<Encode> module knows about many encodings and it has interfaces
235 for doing conversions between those encodings:
237 use Encode 'from_to';
238 from_to($data, "iso-8859-3", "utf-8"); # from legacy to utf-8
242 Normally writing out Unicode data
244 print FH chr(0x100), "\n";
246 will print out the raw UTF-8 bytes, but you will get a warning
247 out of that if you use C<-w> or C<use warnings>. To avoid the
248 warning open the stream explicitly in UTF-8:
250 open FH, ">:utf8", "file";
252 and on already open streams use C<binmode()>:
254 binmode(STDOUT, ":utf8");
256 Reading in correctly formed UTF-8 data will not magically turn
257 the data into Unicode in Perl's eyes.
259 You can use either the C<':utf8'> I/O discipline when opening files
261 open(my $fh,'<:utf8', 'anything');
262 my $line_of_utf8 = <$fh>;
264 The I/O disciplines can also be specified more flexibly with
265 the C<open> pragma; see L<open>:
267 use open ':utf8'; # input and output default discipline will be UTF-8
269 print X chr(0x100), "\n";
272 printf "%#x\n", ord(<Y>); # this should print 0x100
275 With the C<open> pragma you can use the C<:locale> discipline
277 $ENV{LANG} = 'ru_RU.KOI8-R';
278 # the :locale will probe the locale environment variables like LANG
279 use open OUT => ':locale'; # russki parusski
281 print O chr(0x430); # Unicode CYRILLIC SMALL LETTER A = KOI8-R 0xc1
284 printf "%#x\n", ord(<I>), "\n"; # this should print 0xc1
287 or you can also use the C<':encoding(...)'> discipline
289 open(my $epic,'<:encoding(iso-8859-7)','iliad.greek');
290 my $line_of_iliad = <$epic>;
292 Both of these methods install a transparent filter on the I/O stream that
293 will convert data from the specified encoding when it is read in from the
294 stream. In the first example the F<anything> file is assumed to be UTF-8
295 encoded Unicode, in the second example the F<iliad.greek> file is assumed
296 to be ISO-8858-7 encoded Greek, but the lines read in will be in both
299 The L<open> pragma affects all the C<open()> calls after the pragma by
300 setting default disciplines. If you want to affect only certain
301 streams, use explicit disciplines directly in the C<open()> call.
303 You can switch encodings on an already opened stream by using
304 C<binmode()>, see L<perlfunc/binmode>.
306 The C<:locale> does not currently work with C<open()> and
307 C<binmode()>, only with the C<open> pragma. The C<:utf8> and
308 C<:encoding(...)> do work with all of C<open()>, C<binmode()>,
309 and the C<open> pragma.
311 Similarly, you may use these I/O disciplines on input streams to
312 automatically convert data from the specified encoding when it is
313 written to the stream.
315 open(my $unicode, '<:utf8', 'japanese.uni');
316 open(my $nihongo, '>:encoding(iso2022-jp)', 'japanese.jp');
317 while (<$unicode>) { print $nihongo }
319 The naming of encodings, both by the C<open()> and by the C<open>
320 pragma, is similarly understanding as with the C<encoding> pragma:
321 C<koi8-r> and C<KOI8R> will both be understood.
323 Common encodings recognized by ISO, MIME, IANA, and various other
324 standardisation organisations are recognised, for a more detailed
327 C<read()> reads characters and returns the number of characters.
328 C<seek()> and C<tell()> operate on byte counts, as do C<sysread()>
331 Notice that because of the default behaviour "input is not UTF-8"
332 it is easy to mistakenly write code that keeps on expanding a file
333 by repeatedly encoding it in UTF-8:
337 local $/; # read in the whole file
340 open F, ">:utf8", "file";
344 If you run this code twice, the contents of the F<file> will be twice
345 UTF-8 encoded. A C<use open ':utf8'> would have avoided the bug, or
346 explicitly opening also the F<file> for input as UTF-8.
354 Bit Complement Operator ~ And vec()
356 The bit complement operator C<~> will produce surprising results if
357 used on strings containing Unicode characters. The results are
358 consistent with the internal UTF-8 encoding of the characters, but not
359 with much else. So don't do that. Similarly for vec(): you will be
360 operating on the UTF-8 bit patterns of the Unicode characters, not on
361 the bytes, which is very probably not what you want.
367 One way of peeking inside the internal encoding of Unicode characters
368 is to use C<unpack("C*", ...> to get the bytes, or C<unpack("H*", ...)>
369 to display the bytes:
371 # this will print c4 80 for the UTF-8 bytes 0xc4 0x80
372 print join(" ", unpack("H*", pack("U", 0x100))), "\n";
374 Yet another way would be to use the Devel::Peek module:
376 perl -MDevel::Peek -e 'Dump(chr(0x100))'
378 That will show the UTF8 flag in FLAGS and both the UTF-8 bytes
379 and Unicode characters in PV. See also later in this document
380 the discussion about the C<is_utf8> function of the C<Encode> module.
384 =head2 Advanced Topics
392 The question of string equivalence turns somewhat complicated
393 in Unicode: what do you mean by equal?
395 (Is C<LATIN CAPITAL LETTER A WITH ACUTE> equal to
396 C<LATIN CAPITAL LETTER A>?)
398 The short answer is that by default Perl compares equivalence
399 (C<eq>, C<ne>) based only on code points of the characters.
400 In the above case, no (because 0x00C1 != 0x0041). But sometimes any
401 CAPITAL LETTER As being considered equal, or even any As of any case,
404 The long answer is that you need to consider character normalization
405 and casing issues: see L<Unicode::Normalize>, and Unicode Technical
406 Reports #15 and #21, I<Unicode Normalization Forms> and I<Case
407 Mappings>, http://www.unicode.org/unicode/reports/tr15/
408 http://www.unicode.org/unicode/reports/tr21/
410 As of Perl 5.8.0, the's regular expression case-ignoring matching
411 implements only 1:1 semantics: one character matches one character.
412 In I<Case Mappings> both 1:N and N:1 matches are defined.
418 People like to see their strings nicely sorted, or as Unicode
419 parlance goes, collated. But again, what do you mean by collate?
421 (Does C<LATIN CAPITAL LETTER A WITH ACUTE> come before or after
422 C<LATIN CAPITAL LETTER A WITH GRAVE>?)
424 The short answer is that by default Perl compares strings (C<lt>,
425 C<le>, C<cmp>, C<ge>, C<gt>) based only on the code points of the
426 characters. In the above case, after, since 0x00C1 > 0x00C0.
428 The long answer is that "it depends", and a good answer cannot be
429 given without knowing (at the very least) the language context.
430 See L<Unicode::Collate>, and I<Unicode Collation Algorithm>
431 http://www.unicode.org/unicode/reports/tr10/
443 Character ranges in regular expression character classes (C</[a-z]/>)
444 and in the C<tr///> (also known as C<y///>) operator are not magically
445 Unicode-aware. What this means that C<[a-z]> will not magically start
446 to mean "all alphabetic letters" (not that it does mean that even for
447 8-bit characters, you should be using C</[[:alpha]]/> for that).
449 For specifying things like that in regular expressions you can use the
450 various Unicode properties, C<\pL> in this particular case. You can
451 use Unicode code points as the end points of character ranges, but
452 that means that particular code point range, nothing more. For
453 further information, see L<perlunicode>.
457 String-To-Number Conversions
459 Unicode does define several other decimal (and numeric) characters
460 than just the familiar 0 to 9, such as the Arabic and Indic digits.
461 Perl does not support string-to-number conversion for digits other
462 than the 0 to 9 (and a to f for hexadecimal).
466 =head2 Questions With Answers
470 =item Will My Old Scripts Break?
472 Very probably not. Unless you are generating Unicode characters
473 somehow, any old behaviour should be preserved. About the only
474 behaviour that has changed and which could start generating Unicode
475 is the old behaviour of C<chr()> where supplying an argument more
476 than 255 produced a character modulo 255 (for example, C<chr(300)>
477 was equal to C<chr(45)>).
479 =item How Do I Make My Scripts Work With Unicode?
481 Very little work should be needed since nothing changes until you
482 somehow generate Unicode data. The greatest trick will be getting
483 input as Unicode, and for that see the earlier I/O discussion.
485 =item How Do I Know Whether My String Is In Unicode?
487 You shouldn't care. No, you really shouldn't. If you have
488 to care (beyond the cases described above), it means that we
489 didn't get the transparency of Unicode quite right.
493 use Encode 'is_utf8';
494 print is_utf8($string) ? 1 : 0, "\n";
496 But note that this doesn't mean that any of the characters in the
497 string are necessary UTF-8 encoded, or that any of the characters have
498 code points greater than 0xFF (255) or even 0x80 (128), or that the
499 string has any characters at all. All the C<is_utf8()> does is to
500 return the value of the internal "utf8ness" flag attached to the
501 $string. If the flag is on, characters added to that string will be
502 automatically upgraded to UTF-8 (and even then only if they really
503 need to be upgraded, that is, if their code point is greater than 0xFF).
505 Sometimes you might really need to know the byte length of a string
506 instead of the character length. For that use the C<bytes> pragma
507 and its only defined function C<length()>:
509 my $unicode = chr(0x100);
510 print length($unicode), "\n"; # will print 1
512 print length($unicode), "\n"; # will print 2 (the 0xC4 0x80 of the UTF-8)
514 =item How Do I Detect Invalid UTF-8?
518 use Encode 'encode_utf8';
519 if (encode_utf8($string)) {
528 @chars = unpack("U0U*", "\xFF"); # will warn
530 The warning will be C<Malformed UTF-8 character (byte 0xff) in
531 unpack>. The "U0" means "expect strictly UTF-8 encoded Unicode".
532 Without that the C<unpack("U*", ...)> would accept also data like
535 =item How Do I Convert Data Into UTF-8? Or Vice Versa?
537 This probably isn't as useful (or simple) as you might think.
538 Also, normally you shouldn't need to.
540 In one sense what you are asking doesn't make much sense: UTF-8 is
541 (intended as an) Unicode encoding, so converting "data" into UTF-8
542 isn't meaningful unless you know in what character set and encoding
543 the binary data is in, and in this case you can use C<Encode>.
545 use Encode 'from_to';
546 from_to($data, "iso-8859-1", "utf-8"); # from latin-1 to utf-8
548 If you have ASCII (really 7-bit US-ASCII), you already have valid
549 UTF-8, the lowest 128 characters of UTF-8 encoded Unicode and US-ASCII
552 If you have Latin-1 (or want Latin-1), you can just use pack/unpack:
554 $latin1 = pack("C*", unpack("U*", $utf8));
555 $utf8 = pack("U*", unpack("C*", $latin1));
557 (The same works for EBCDIC.)
559 If you have a sequence of bytes you B<know> is valid UTF-8,
560 but Perl doesn't know it yet, you can make Perl a believer, too:
562 use Encode 'decode_utf8';
563 $utf8 = decode_utf8($bytes);
565 You can convert well-formed UTF-8 to a sequence of bytes, but if
566 you just want to convert random binary data into UTF-8, you can't.
567 Any random collection of bytes isn't well-formed UTF-8. You can
568 use C<unpack("C*", $string)> for the former, and you can create
569 well-formed Unicode/UTF-8 data by C<pack("U*", 0xff, ...)>.
571 =item How Do I Display Unicode? How Do I Input Unicode?
573 See http://www.hclrss.demon.co.uk/unicode/ and
574 http://www.cl.cam.ac.uk/~mgk25/unicode.html
576 =item How Does Unicode Work With Traditional Locales?
578 In Perl, not very well. Avoid using locales through the C<locale>
579 pragma. Use only one or the other.
583 =head2 Hexadecimal Notation
585 The Unicode standard prefers using hexadecimal notation because that
586 shows better the division of Unicode into blocks of 256 characters.
587 Hexadecimal is also simply shorter than decimal. You can use decimal
588 notation, too, but learning to use hexadecimal just makes life easier
589 with the Unicode standard.
591 The C<0x> prefix means a hexadecimal number, the digits are 0-9 I<and>
592 a-f (or A-F, case doesn't matter). Each hexadecimal digit represents
593 four bits, or half a byte. C<print 0x..., "\n"> will show a
594 hexadecimal number in decimal, and C<printf "%x\n", $decimal> will
595 show a decimal number in hexadecimal. If you have just the
596 "hexdigits" of a hexadecimal number, you can use the C<hex()>
599 print 0x0009, "\n"; # 9
600 print 0x000a, "\n"; # 10
601 print 0x000f, "\n"; # 15
602 print 0x0010, "\n"; # 16
603 print 0x0011, "\n"; # 17
604 print 0x0100, "\n"; # 256
606 print 0x0041, "\n"; # 65
608 printf "%x\n", 65; # 41
609 printf "%#x\n", 65; # 0x41
611 print hex("41"), "\n"; # 65
613 =head2 Further Resources
621 http://www.unicode.org/
627 http://www.unicode.org/unicode/faq/
633 http://www.unicode.org/glossary/
637 Unicode Useful Resources
639 http://www.unicode.org/unicode/onlinedat/resources.html
643 Unicode and Multilingual Support in HTML, Fonts, Web Browsers and Other Applications
645 http://www.hclrss.demon.co.uk/unicode/
649 UTF-8 and Unicode FAQ for Unix/Linux
651 http://www.cl.cam.ac.uk/~mgk25/unicode.html
655 Legacy Character Sets
657 http://www.czyborra.com/
658 http://www.eki.ee/letter/
662 The Unicode support files live within the Perl installation in the
665 $Config{installprivlib}/unicore
667 in Perl 5.8.0 or newer, and
669 $Config{installprivlib}/unicode
671 in the Perl 5.6 series. (The renaming to F<lib/unicore> was done to
672 avoid naming conflicts with lib/Unicode in case-insensitive filesystems.)
673 The main Unicode data file is F<Unicode.txt> (or F<Unicode.301> in
674 Perl 5.6.1.) You can find the C<$Config{installprivlib}> by
676 perl "-V:installprivlib"
678 Note that some of the files have been renamed from the Unicode
679 standard since the Perl installation tries to live by the "8.3"
680 filenaming restrictions. The renamings are shown in the
681 accompanying F<rename> file.
683 You can explore various information from the Unicode data files using
684 the C<Unicode::UCD> module.
690 L<perlunicode>, L<Encode>, L<encoding>, L<open>, L<utf8>, L<bytes>,
691 L<perlretut>, L<Unicode::Collate>, L<Unicode::Normalize>, L<Unicode::UCD>
693 =head1 ACKNOWLEDGEMENTS
695 Thanks to the kind readers of the perl5-porters@perl.org,
696 perl-unicode@perl.org, linux-utf8@nl.linux.org, and unicore@unicode.org
697 mailing lists for their valuable feedback.
699 =head1 AUTHOR, COPYRIGHT, AND LICENSE
701 Copyright 2001 Jarkko Hietaniemi <jhi@iki.fi>
703 This document may be distributed under the same terms as Perl itself.