Unicode and ISO/IEC 10646 are coordinated standards that provide code points for characters in almost all modern character set standards, covering more than 30 writing systems and hundreds of languages, including all commercially-important modern languages. All characters in the largest Chinese, Japanese, and Korean dictionaries are also encoded. The standards will eventually cover almost all characters in more than 250 writing systems and thousands of languages. Unicode 1.0 was released in October 1991, and 4.0 in April 2003.

A Unicode character is an abstract entity. It is not bound to any particular integer width, especially not to the C language char. Unicode is language-neutral and display-neutral: it does not encode the language of the text and it does not define fonts or other graphical layout details. Unicode operates on characters and on text built from those characters.

Unicode defines characters like LATIN CAPITAL LETTER A or GREEK SMALL LETTER ALPHA and unique numbers for the characters, in this case 0x0041 and 0x03B1, respectively. These unique numbers are called code points.

The Unicode standard prefers using hexadecimal notation for the code points. If numbers like 0x0041 are unfamiliar to you, take a peek at a later section, Hexadecimal Notation. The Unicode standard uses the notation U+0041 LATIN CAPITAL LETTER A, to give the hexadecimal code point and the normative name of the character.

Unicode also defines various properties for the characters, like "uppercase" or "lowercase", "decimal digit", or "punctuation"; these properties are independent of the names of the characters. Furthermore, various operations on the characters like uppercasing, lowercasing, and collating (sorting) are defined.

A Unicode character consists either of a single code point, or a base character (like LATIN CAPITAL LETTER A), followed by one or more modifiers (like COMBINING ACUTE ACCENT). This sequence of base character and modifiers is called a combining character sequence.

Whether to call these combining character sequences "characters" depends on your point of view. If you are a programmer, you probably would tend towards seeing each element in the sequences as one unit, or "character". The whole sequence could be seen as one "character", however, from the user's point of view, since that's probably what it looks like in the context of the user's language.

With this "whole sequence" view of characters, the total number of characters is open-ended. But in the programmer's "one unit is one character" point of view, the concept of "characters" is more deterministic. In this document, we take that second point of view: one "character" is one Unicode code point, be it a base character or a combining character.

For some combinations, there are precomposed characters. LATIN CAPITAL LETTER A WITH ACUTE, for example, is defined as a single code point. These precomposed characters are, however, only available for some combinations, and are mainly meant to support round-trip conversions between Unicode and legacy standards (like the ISO 8859). In the general case, the composing method is more extensible. To support conversion between different compositions of the characters, various normalization forms to standardize representations are also defined.

Because of backward compatibility with legacy encodings, the "a unique number for every character" idea breaks down a bit: instead, there is "at least one number for every character". The same character could be represented differently in several legacy encodings. The converse is also not true: some code points do not have an assigned character. Firstly, there are unallocated code points within otherwise used blocks. Secondly, there are special Unicode control characters that do not represent true characters.

A common myth about Unicode is that it would be "16-bit", that is, Unicode is only represented as 0x10000 (or 65536) characters from 0x0000 to 0xFFFF. This is untrue. Since Unicode 2.0 (July 1996), Unicode has been defined all the way up to 21 bits (0x10FFFF), and since Unicode 3.1 (March 2001), characters have been defined beyond 0xFFFF. The first 0x10000 characters are called the Plane 0, or the Basic Multilingual Plane (BMP). With Unicode 3.1, 17 (yes, seventeen) planes in all were defined--but they are nowhere near full of defined characters, yet.

Another myth is that the 256-character blocks have something to do with languages--that each block would define the characters used by a language or a set of languages. This is also untrue. The division into blocks exists, but it is almost completely accidental--an artifact of how the characters have been and still are allocated. Instead, there is a concept called scripts, which is more useful: there is Latin script, Greek script, and so on. Scripts usually span varied parts of several blocks. For further information see the Unicode::UCD manpage.

The Unicode code points are just abstract numbers. To input and output these abstract numbers, the numbers must be encoded or serialised somehow. Unicode defines several character encoding forms, of which UTF-8 is perhaps the most popular. UTF-8 is a variable length encoding that encodes Unicode characters as 1 to 6 bytes (only 4 with the currently defined characters). Other encodings include UTF-16 and UTF-32 and their big- and little-endian variants (UTF-8 is byte-order independent) The ISO/IEC 10646 defines the UCS-2 and UCS-4 encoding forms.

For more information about encodings--for instance, to learn what surrogates and byte order marks (BOMs) are--see the perlunicode manpage.

Perl's Unicode Support

Starting from Perl 5.6.0, Perl has had the capacity to handle Unicode natively. Perl 5.8.0, however, is the first recommended release for serious Unicode work. The maintenance release 5.6.1 fixed many of the problems of the initial Unicode implementation, but for example regular expressions still do not work with Unicode in 5.6.1.

Starting from Perl 5.8.0, the use of use utf8 is no longer necessary. In earlier releases the utf8 pragma was used to declare that operations in the current block or file would be Unicode-aware. This model was found to be wrong, or at least clumsy: the "Unicodeness" is now carried with the data, instead of being attached to the operations. Only one case remains where an explicit use utf8 is needed: if your Perl script itself is encoded in UTF-8, you can use UTF-8 in your identifier names, and in string and regular expression literals, by saying use utf8. This is not the default because scripts with legacy 8-bit data in them would break. See the utf8 manpage.

Perl's Unicode Model

Perl supports both pre-5.6 strings of eight-bit native bytes, and strings of Unicode characters. The principle is that Perl tries to keep its data as eight-bit bytes for as long as possible, but as soon as Unicodeness cannot be avoided, the data is transparently upgraded to Unicode.

Internally, Perl currently uses either whatever the native eight-bit character set of the platform (for example Latin-1) is, defaulting to UTF-8, to encode Unicode strings. Specifically, if all code points in the string are 0xFF or less, Perl uses the native eight-bit character set. Otherwise, it uses UTF-8.

A user of Perl does not normally need to know nor care how Perl happens to encode its internal strings, but it becomes relevant when outputting Unicode strings to a stream without a PerlIO layer -- one with the "default" encoding. In such a case, the raw bytes used internally (the native character set or UTF-8, as appropriate for each string) will be used, and a "Wide character" warning will be issued if those strings contain a character beyond 0x00FF.

For example,

      perl -e 'print "\x{DF}\n", "\x{0100}\x{DF}\n"'

produces a fairly useless mixture of native bytes and UTF-8, as well as a warning:

     Wide character in print at ...

To output UTF-8, use the :encoding or :utf8 output layer. Prepending

      binmode(STDOUT, ":utf8");

to this sample program ensures that the output is completely UTF-8, and removes the program's warning.

You can enable automatic UTF-8-ification of your standard file handles, default open() layer, and @ARGV by using either the -C command line switch or the PERL_UNICODE environment variable, see the perlrun manpage for the documentation of the -C switch.

Note that this means that Perl expects other software to work, too: if Perl has been led to believe that STDIN should be UTF-8, but then STDIN coming in from another command is not UTF-8, Perl will complain about the malformed UTF-8.

All features that combine Unicode and I/O also require using the new PerlIO feature. Almost all Perl 5.8 platforms do use PerlIO, though: you can see whether yours is by running "perl -V" and looking for useperlio=define.

Unicode and EBCDIC

Perl 5.8.0 also supports Unicode on EBCDIC platforms. There, Unicode support is somewhat more complex to implement since additional conversions are needed at every step. Some problems remain, see the perlebcdic manpage for details.

In any case, the Unicode support on EBCDIC platforms is better than in the 5.6 series, which didn't work much at all for EBCDIC platform. On EBCDIC platforms, the internal Unicode encoding form is UTF-EBCDIC instead of UTF-8. The difference is that as UTF-8 is "ASCII-safe" in that ASCII characters encode to UTF-8 as-is, while UTF-EBCDIC is "EBCDIC-safe".

Creating Unicode

To create Unicode characters in literals for code points above 0xFF, use the \x{...} notation in double-quoted strings:

    my $smiley = "\x{263a}";

Similarly, it can be used in regular expression literals

    $smiley =~ /\x{263a}/;

At run-time you can use chr():

    my $hebrew_alef = chr(0x05d0);

See Further Resources for how to find all these numeric codes.

Naturally, ord() will do the reverse: it turns a character into a code point.

Note that \x.. (no {} and only two hexadecimal digits), \x{...}, and chr(...) for arguments less than 0x100 (decimal 256) generate an eight-bit character for backward compatibility with older Perls. For arguments of 0x100 or more, Unicode characters are always produced. If you want to force the production of Unicode characters regardless of the numeric value, use pack("U", ...) instead of \x.., \x{...}, or chr().

You can also use the charnames pragma to invoke characters by name in double-quoted strings:

    use charnames ':full';
    my $arabic_alef = "\N{ARABIC LETTER ALEF}";

And, as mentioned above, you can also pack() numbers into Unicode characters:

   my $georgian_an  = pack("U", 0x10a0);

Note that both \x{...} and \N{...} are compile-time string constants: you cannot use variables in them. if you want similar run-time functionality, use chr() and charnames::vianame().

If you want to force the result to Unicode characters, use the special "U0" prefix. It consumes no arguments but causes the following bytes to be interpreted as the UTF-8 encoding of Unicode characters:

   my $chars = pack("U0W*", 0x80, 0x42);

Likewise, you can stop such UTF-8 interpretation by using the special "C0" prefix.

Handling Unicode

Handling Unicode is for the most part transparent: just use the strings as usual. Functions like index(), length(), and substr() will work on the Unicode characters; regular expressions will work on the Unicode characters (see the perlunicode manpage and the perlretut manpage).

Note that Perl considers combining character sequences to be separate characters, so for example

    use charnames ':full';
    print length("\N{LATIN CAPITAL LETTER A}\N{COMBINING ACUTE ACCENT}"), "\n";

will print 2, not 1. The only exception is that regular expressions have \X for matching a combining character sequence.

Life is not quite so transparent, however, when working with legacy encodings, I/O, and certain special cases:

Legacy Encodings

When you combine legacy data and Unicode the legacy data needs to be upgraded to Unicode. Normally ISO 8859-1 (or EBCDIC, if applicable) is assumed.

The Encode module knows about many encodings and has interfaces for doing conversions between those encodings:

    use Encode 'decode';
    $data = decode("iso-8859-3", $data); # convert from legacy to utf-8

Unicode I/O

Normally, writing out Unicode data

    print FH $some_string_with_unicode, "\n";

produces raw bytes that Perl happens to use to internally encode the Unicode string. Perl's internal encoding depends on the system as well as what characters happen to be in the string at the time. If any of the characters are at code points 0x100 or above, you will get a warning. To ensure that the output is explicitly rendered in the encoding you desire--and to avoid the warning--open the stream with the desired encoding. Some examples:

    open FH, ">:utf8", "file";

    open FH, ">:encoding(ucs2)",      "file";
    open FH, ">:encoding(UTF-8)",     "file";
    open FH, ">:encoding(shift_jis)", "file";

and on already open streams, use binmode():

    binmode(STDOUT, ":utf8");

    binmode(STDOUT, ":encoding(ucs2)");
    binmode(STDOUT, ":encoding(UTF-8)");
    binmode(STDOUT, ":encoding(shift_jis)");

The matching of encoding names is loose: case does not matter, and many encodings have several aliases. Note that the :utf8 layer must always be specified exactly like that; it is not subject to the loose matching of encoding names. Also note that :utf8 is unsafe for input, because it accepts the data without validating that it is indeed valid UTF8.

See the PerlIO manpage for the :utf8 layer, the PerlIO::encoding manpage and the Encode::PerlIO manpage for the :encoding() layer, and the Encode::Supported manpage for many encodings supported by the Encode module.

Reading in a file that you know happens to be encoded in one of the Unicode or legacy encodings does not magically turn the data into Unicode in Perl's eyes. To do that, specify the appropriate layer when opening files

    open(my $fh,'<:encoding(utf8)', 'anything');
    my $line_of_unicode = <$fh>;

    open(my $fh,'<:encoding(Big5)', 'anything');
    my $line_of_unicode = <$fh>;

The I/O layers can also be specified more flexibly with the open pragma. See the open manpage, or look at the following example.

    use open ':encoding(utf8)'; # input/output default encoding will be UTF-8
    open X, ">file";
    print X chr(0x100), "\n";
    close X;
    open Y, "<file";
    printf "%#x\n", ord(<Y>); # this should print 0x100
    close Y;

With the open pragma you can use the :locale layer

    BEGIN { $ENV{LC_ALL} = $ENV{LANG} = 'ru_RU.KOI8-R' }
    # the :locale will probe the locale environment variables like LC_ALL
    use open OUT => ':locale'; # russki parusski
    open(O, ">koi8");
    print O chr(0x430); # Unicode CYRILLIC SMALL LETTER A = KOI8-R 0xc1
    close O;
    open(I, "<koi8");
    printf "%#x\n", ord(<I>), "\n"; # this should print 0xc1
    close I;

These methods install a transparent filter on the I/O stream that converts data from the specified encoding when it is read in from the stream. The result is always Unicode.

The the open manpage pragma affects all the open() calls after the pragma by setting default layers. If you want to affect only certain streams, use explicit layers directly in the open() call.

You can switch encodings on an already opened stream by using binmode(); see binmode in the perlfunc manpage.

The :locale does not currently (as of Perl 5.8.0) work with