The Fidelix Linux Distribution

Simple, Stable, and Secure

PERLUNICOOK(1) Perl Programmers Reference Guide PERLUNICOOK(1)

perlunicook - cookbookish examples of handling Unicode in Perl

This manpage contains short recipes demonstrating how to handle common Unicode operations in Perl, plus one complete program at the end. Any undeclared variables in individual recipes are assumed to have a previous appropriate value in them.

Unless otherwise notes, all examples below require this standard preamble to work correctly, with the "#!" adjusted to work on your system:

 #!/usr/bin/env perl

 use utf8;      # so literals and identifiers can be in UTF-8
 use v5.12;     # or later to get "unicode_strings" feature
 use strict;    # quote strings, declare variables
 use warnings;  # on by default
 use warnings  qw(FATAL utf8);    # fatalize encoding glitches
 use open      qw(:std :encoding(UTF-8)); # undeclared streams in UTF-8
 use charnames qw(:full :short);  # unneeded in v5.16

This does make even Unix programmers "binmode" your binary streams, or open them with ":raw", but that's the only way to get at them portably anyway.

WARNING: "use autodie" (pre 2.26) and "use open" do not get along with each other.

Always decompose on the way in, then recompose on the way out.

 use Unicode::Normalize;

 while (<>) {
     $_ = NFD($_);   # decompose + reorder canonically
 } continue {
     print NFC($_);  # recompose (where possible) + reorder canonically

As of v5.14, Perl distinguishes three subclasses of UTFX8 warnings.

 use v5.14;                  # subwarnings unavailable any earlier
 no warnings "nonchar";      # the 66 forbidden non-characters
 no warnings "surrogate";    # UTF-16/CESU-8 nonsense
 no warnings "non_unicode";  # for codepoints over 0x10_FFFF

Without the all-critical "use utf8" declaration, putting UTFX8 in your literals and identifiers wonXt work right. If you used the standard preamble just given above, this already happened. If you did, you can do things like this:

 use utf8;

 my $measure   = "Aangstroem";
 my @Xsoft     = qw( cp852 cp1251 cp1252 );
 my @X        = qw( koi8-f koi8-u koi8-r );

If you forget "use utf8", high bytes will be misunderstood as separate characters, and nothing will work right.

The "ord" and "chr" functions work transparently on all codepoints, not just on ASCII alone X nor in fact, not even just on Unicode alone.

 # ASCII characters

 # characters from the Basic Multilingual Plane

 # beyond the BMP
 ord("X")               # MATHEMATICAL ITALIC SMALL N

 # beyond Unicode! (up to MAXINT)

In an interpolated literal, whether a double-quoted string or a regex, you may specify a character by its number using the "\x{HHHHHH}" escape.

 String: "\x{3a3}"
 Regex:  /\x{3a3}/

 String: "\x{1d45b}"
 Regex:  /\x{1d45b}/

 # even non-BMP ranges in regex work fine

 use charnames ();
 my $name = charnames::viacode(0x03A3);

 use charnames ();
 my $number = charnames::vianame("GREEK CAPITAL LETTER SIGMA");

Use the "\N{charname}" notation to get the character by that name for use in interpolated literals (double-quoted strings and regexes). In v5.16, there is an implicit

 use charnames qw(:full :short);

But prior to v5.16, you must be explicit about which set of charnames you want. The ":full" names are the official Unicode character name, alias, or sequence, which all share a namespace.

 use charnames qw(:full :short latin greek);


Anything else is a Perl-specific convenience abbreviation. Specify one or more scripts by names if you want short names that are script-specific.

 "\N{Greek:Sigma}"                      # :short
 "\N{ae}"                               #  latin
 "\N{epsilon}"                          #  greek

The v5.16 release also supports a ":loose" import for loose matching of character names, which works just like loose matching of property names: that is, it disregards case, whitespace, and underscores:

 "\N{euro sign}"                        # :loose (from v5.16)

These look just like character names but return multiple codepoints. Notice the %vx vector-print functionality in "printf".

 use charnames qw(:full);
 printf "U+%v04X\n", $seq;

Use ":alias" to give your own lexically scoped nicknames to existing characters, or even to give unnamed private-use characters useful names.

 use charnames ":full", ":alias" => {
     "APPLE LOGO" => 0xF8FF, # private use character


Sinograms like XXXX come back with character names of "CJK UNIFIED IDEOGRAPH-6771" and "CJK UNIFIED IDEOGRAPH-4EAC", because their XnamesX vary. The CPAN "Unicode::Unihan" module has a large database for decoding these (and a whole lot more), provided you know how to understand its output.

 # cpan -i Unicode::Unihan
 use Unicode::Unihan;
 my $str = "XX";
 my $unhan = Unicode::Unihan->new;
 for my $lang (qw(Mandarin Cantonese Korean JapaneseOn JapaneseKun)) {
     printf "CJK $str in %-12s is ", $lang;
     say $unhan->$lang($str);


 CJK XX in Mandarin     is DONG1JING1
 CJK XX in Cantonese    is dung1ging1
 CJK XX in Korean       is TONGKYENG
 CJK XX in JapaneseOn   is TOUKYOU KEI KIN

If you have a specific romanization scheme in mind, use the specific module:

 # cpan -i Lingua::JA::Romanize::Japanese
 use Lingua::JA::Romanize::Japanese;
 my $k2r = Lingua::JA::Romanize::Japanese->new;
 my $str = "XX";
 say "Japanese for $str is ", $k2r->chars($str);


 Japanese for XX is toukyou

On rare occasion, such as a database read, you may be given encoded text you need to decode.

  use Encode qw(encode decode);

  my $chars = decode("shiftjis", $bytes, 1);
 # OR
  my $bytes = encode("MIME-Header-ISO_2022_JP", $chars, 1);

For streams all in the same encoding, don't use encode/decode; instead set the file encoding when you open the file or immediately after with "binmode" as described later below.

     $ perl -CA ...
     $ export PERL_UNICODE=A
    use Encode qw(decode);
    @ARGV = map { decode('UTF-8', $_, 1) } @ARGV;

    # cpan -i Encode::Locale
    use Encode qw(locale);
    use Encode::Locale;

    # use "locale" as an arg to encode/decode
    @ARGV = map { decode(locale => $_, 1) } @ARGV;

Use a command-line option, an environment variable, or else call "binmode" explicitly:

     $ perl -CS ...
     $ export PERL_UNICODE=S
     use open qw(:std :encoding(UTF-8));
     binmode(STDIN,  ":encoding(UTF-8)");
     binmode(STDOUT, ":utf8");
     binmode(STDERR, ":utf8");

    # cpan -i Encode::Locale
    use Encode;
    use Encode::Locale;

    # or as a stream for binmode or open
    binmode STDIN,  ":encoding(console_in)"  if -t STDIN;
    binmode STDOUT, ":encoding(console_out)" if -t STDOUT;
    binmode STDERR, ":encoding(console_out)" if -t STDERR;

Files opened without an encoding argument will be in UTF-8:

     $ perl -CD ...
     $ export PERL_UNICODE=D
     use open qw(:encoding(UTF-8));

     $ perl -CSDA ...
     $ export PERL_UNICODE=SDA
     use open qw(:std :encoding(UTF-8));
     use Encode qw(decode);
     @ARGV = map { decode('UTF-8', $_, 1) } @ARGV;

Specify stream encoding. This is the normal way to deal with encoded text, not by calling low-level functions.

 # input file
     open(my $in_file, "< :encoding(UTF-16)", "wintext");
     open(my $in_file, "<", "wintext");
     binmode($in_file, ":encoding(UTF-16)");
     my $line = <$in_file>;

 # output file
     open($out_file, "> :encoding(cp1252)", "wintext");
     open(my $out_file, ">", "wintext");
     binmode($out_file, ":encoding(cp1252)");
     print $out_file "some text\n";

More layers than just the encoding can be specified here. For example, the incantation ":raw :encoding(UTF-16LE) :crlf" includes implicit CRLF handling.

Unicode casing is very different from ASCII casing.

 uc("henry X")  # "HENRY X"
 uc("tschuess")   # "TSCHUeSS"  notice ss => SS

 # both are true:
 "tschuess"  =~ /TSCHUeSS/i   # notice ss => SS
 "XXXXXXX" =~ /XXXXXXX/i   # notice X,X,X sameness

Also available in the CPAN Unicode::CaseFold module, the new "fc" XfoldcaseX function from v5.16 grants access to the same Unicode casefolding as the "/i" pattern modifier has always used:

 use feature "fc"; # fc() function is from v5.16

 # sort case-insensitively
 my @sorted = sort { fc($a) cmp fc($b) } @list;

 # both are true:
 fc("tschuess")  eq fc("TSCHUeSS")
 fc("XXXXXXX") eq fc("XXXXXXX")

A Unicode linebreak matches the two-character CRLF grapheme or any of seven vertical whitespace characters. Good for dealing with textfiles coming from different operating systems.


 s/\R/\n/g;  # normalize all linebreaks to \n

Find the general category of a numeric codepoint.

 use Unicode::UCD qw(charinfo);
 my $cat = charinfo(0x3A3)->{category};  # "Lu"

Disable "\w", "\b", "\s", "\d", and the POSIX classes from working correctly on Unicode either in this scope, or in just one regex.

 use v5.14;
 use re "/a";

 # OR

 my($num) = $str =~ /(\d+)/a;

Or use specific un-Unicode properties, like "\p{ahex}" and "\p{POSIX_Digit"}. Properties still work normally no matter what charset modifiers ("/d /u /l /a /aa") should be effect.

These all match a single codepoint with the given property. Use "\P" in place of "\p" to match one codepoint lacking that property.

 \pL, \pN, \pS, \pP, \pM, \pZ, \pC
 \p{Sk}, \p{Ps}, \p{Lt}
 \p{alpha}, \p{upper}, \p{lower}
 \p{Latin}, \p{Greek}
 \p{script_extensions=Latin}, \p{scx=Greek}
 \p{East_Asian_Width=Wide}, \p{EA=W}
 \p{Line_Break=Hyphen}, \p{LB=HY}
 \p{Numeric_Value=4}, \p{NV=4}

Define at compile-time your own custom character properties for use in regexes.

 # using private-use characters
 sub In_Tengwar { "E000\tE07F\n" }

 if (/\p{In_Tengwar}/) { ... }

 # blending existing properties
 sub Is_GraecoRoman_Title {<<'END_OF_SET'}

 if (/\p{Is_GraecoRoman_Title}/ { ... }

Typically render into NFD on input and NFC on output. Using NFKC or NFKD functions improves recall on searches, assuming you've already done to the same text to be searched. Note that this is about much more than just pre- combined compatibility glyphs; it also reorders marks according to their canonical combining classes and weeds out singletons.

 use Unicode::Normalize;
 my $nfd  = NFD($orig);
 my $nfc  = NFC($orig);
 my $nfkd = NFKD($orig);
 my $nfkc = NFKC($orig);

Unless youXve used "/a" or "/aa", "\d" matches more than ASCII digits only, but PerlXs implicit string-to-number conversion does not current recognize these. HereXs how to convert such strings manually.

 use v5.14;  # needed for num() function
 use Unicode::UCD qw(num);
 my $str = "got X and XXXX and X and here";
 my @nums = ();
 while ($str =~ /(\d+|\N)/g) {  # not just ASCII!
    push @nums, num($1);
 say "@nums";   #     12      4567      0.875

 use charnames qw(:full);
 my $nv = num("\N{RUMI DIGIT ONE}\N{RUMI DIGIT TWO}");

Programmer-visible XcharactersX are codepoints matched by "/./s", but user-visible XcharactersX are graphemes matched by "/\X/".

 # Find vowel *plus* any combining diacritics,underlining,etc.
 my $nfd = NFD($orig);
 $nfd =~ / (?=[aeiou]) \X /xi

 # match and grab five first graphemes
 my($first_five) = $str =~ /^ ( \X{5} ) /x;

 # cpan -i Unicode::GCString
 use Unicode::GCString;
 my $gcs = Unicode::GCString->new($str);
 my $first_five = $gcs->substr(0, 5);

Reversing by codepoint messes up diacritics, mistakenly converting "creme brulee" into "eelXurb emXerc" instead of into "eelurb emerc"; so reverse by grapheme instead. Both these approaches work right no matter what normalization the string is in:

 $str = join("", reverse $str =~ /\X/g);

 # OR: cpan -i Unicode::GCString
 use Unicode::GCString;
 $str = reverse Unicode::GCString->new($str);

The string "brulee" has six graphemes but up to eight codepoints. This counts by grapheme, not by codepoint:

 my $str = "brulee";
 my $count = 0;
 while ($str =~ /\X/g) { $count++ }

  # OR: cpan -i Unicode::GCString
 use Unicode::GCString;
 my $gcs = Unicode::GCString->new($str);
 my $count = $gcs->length;

PerlXs "printf", "sprintf", and "format" think all codepoints take up 1 print column, but many take 0 or 2. Here to show that normalization makes no difference, we print out both forms:

 use Unicode::GCString;
 use Unicode::Normalize;

 my @words = qw/creme brulee/;
 @words = map { NFC($_), NFD($_) } @words;

 for my $str (@words) {
     my $gcs = Unicode::GCString->new($str);
     my $cols = $gcs->columns;
     my $pad = " " x (10 - $cols);
     say str, $pad, " |";

generates this to show that it pads correctly no matter the normalization:

 creme      |
 creXme      |
 brulee     |
 bruXleXe     |

Text sorted by numeric codepoint follows no reasonable alphabetic order; use the UCA for sorting text.

 use Unicode::Collate;
 my $col = Unicode::Collate->new();
 my @list = $col->sort(@old_list);

See the ucsort program from the Unicode::Tussle CPAN module for a convenient command-line interface to this module.

Specify a collation strength of level 1 to ignore case and diacritics, only looking at the basic character.

 use Unicode::Collate;
 my $col = Unicode::Collate->new(level => 1);
 my @list = $col->sort(@old_list);

Some locales have special sorting rules.

 # either use v5.12, OR: cpan -i Unicode::Collate::Locale
 use Unicode::Collate::Locale;
 my $col = Unicode::Collate::Locale->new(locale => "de__phonebook");
 my @list = $col->sort(@old_list);

The ucsort program mentioned above accepts a "--locale" parameter.

Instead of this:

 @srecs = sort {
     $b->{AGE}   <=>  $a->{AGE}
     $a->{NAME}  cmp  $b->{NAME}
 } @recs;

Use this:

 my $coll = Unicode::Collate->new();
 for my $rec (@recs) {
     $rec->{NAME_key} = $coll->getSortKey( $rec->{NAME} );
 @srecs = sort {
     $b->{AGE}       <=>  $a->{AGE}
     $a->{NAME_key}  cmp  $b->{NAME_key}
 } @recs;

Use a collator object to compare Unicode text by character instead of by codepoint.

 use Unicode::Collate;
 my $es = Unicode::Collate->new(
     level => 1,
     normalization => undef

  # now both are true:
 $es->eq("Garcia",  "GARCIA" );
 $es->eq("Marquez", "MARQUEZ");

Same, but in a specific locale.

 my $de = Unicode::Collate::Locale->new(
            locale => "de__phonebook",

 # now this is true:
 $de->eq("tschuess", "TSCHUESS");  # notice ue => UE, ss => SS

Break up text into lines according to Unicode rules.

 # cpan -i Unicode::LineBreak
 use Unicode::LineBreak;
 use charnames qw(:full);

 my $para = "This is a super\N{HYPHEN}long string. " x 20;
 my $fmt = Unicode::LineBreak->new;
 print $fmt->break($para), "\n";

Using a regular Perl string as a key or value for a DBM hash will trigger a wide character exception if any codepoints wonXt fit into a byte. HereXs how to manually manage the translation:

    use DB_File;
    use Encode qw(encode decode);
    tie %dbhash, "DB_File", "pathname";


    # assume $uni_key and $uni_value are abstract Unicode strings
    my $enc_key   = encode("UTF-8", $uni_key, 1);
    my $enc_value = encode("UTF-8", $uni_value, 1);
    $dbhash{$enc_key} = $enc_value;


    # assume $uni_key holds a normal Perl string (abstract Unicode)
    my $enc_key   = encode("UTF-8", $uni_key, 1);
    my $enc_value = $dbhash{$enc_key};
    my $uni_value = decode("UTF-8", $enc_value, 1);

HereXs how to implicitly manage the translation; all encoding and decoding is done automatically, just as with streams that have a particular encoding attached to them:

    use DB_File;
    use DBM_Filter;

    my $dbobj = tie %dbhash, "DB_File", "pathname";
    $dbobj->Filter_Value("utf8");  # this is the magic bit


    # assume $uni_key and $uni_value are abstract Unicode strings
    $dbhash{$uni_key} = $uni_value;


    # $uni_key holds a normal Perl string (abstract Unicode)
    my $uni_value = $dbhash{$uni_key};

HereXs a full program showing how to make use of locale-sensitive sorting, Unicode casing, and managing print widths when some of the characters take up zero or two columns, not just one column each time. When run, the following program produces this nicely aligned output:

    Creme Brulee....... X2.00
    Eclair............. X1.60
    Fideua............. X4.20
    Hamburger.......... X6.00
    Jamon Serrano...... X4.45
    Linguica........... X7.00
    Pate............... X4.15
    Pears.............. X2.00
    Peches............. X2.25
    Smorbrod........... X5.75
    Spaetzle............ X5.50
    Xorico............. X3.00
    XXXXX.............. X6.50
    XXX............. X4.00
    XXX............. X2.65
    XXXXX......... X8.00
    XXXXXXX..... X1.85
    XX............... X9.99
    XX............... X7.50

Here's that program; tested on v5.14.

 #!/usr/bin/env perl
 # umenu - demo sorting and printing of Unicode food
 # (obligatory and increasingly long preamble)
 use utf8;
 use v5.14;                       # for locale sorting
 use strict;
 use warnings;
 use warnings  qw(FATAL utf8);    # fatalize encoding faults
 use open      qw(:std :encoding(UTF-8)); # undeclared streams in UTF-8
 use charnames qw(:full :short);  # unneeded in v5.16

 # std modules
 use Unicode::Normalize;          # std perl distro as of v5.8
 use List::Util qw(max);          # std perl distro as of v5.10
 use Unicode::Collate::Locale;    # std perl distro as of v5.14

 # cpan modules
 use Unicode::GCString;           # from CPAN

 # forward defs
 sub pad($$$);
 sub colwidth(_);
 sub entitle(_);

 my %price = (
     "XXXXX"             => 6.50, # gyros
     "pears"             => 2.00, # like um, pears
     "linguica"          => 7.00, # spicy sausage, Portuguese
     "xorico"            => 3.00, # chorizo sausage, Catalan
     "hamburger"         => 6.00, # burgermeister meisterburger
     "eclair"            => 1.60, # dessert, French
     "smorbrod"          => 5.75, # sandwiches, Norwegian
     "spaetzle"           => 5.50, # Bayerisch noodles, little sparrows
     "XX"              => 7.50, # bao1 zi5, steamed pork buns, Mandarin
     "jamon serrano"     => 4.45, # country ham, Spanish
     "peches"            => 2.25, # peaches, French
     "XXXXXXX"    => 1.85, # cream-filled pastry like eclair
     "XXX"            => 4.00, # makgeolli, Korean rice wine
     "XX"              => 9.99, # sushi, Japanese
     "XXX"            => 2.65, # omochi, rice cakes, Japanese
     "creme brulee"      => 2.00, # crema catalana
     "fideua"            => 4.20, # more noodles, Valencian
                                  # (Catalan=fideuada)
     "pate"              => 4.15, # gooseliver paste, French
     "XXXXX"        => 8.00, # okonomiyaki, Japanese

 my $width = 5 + max map { colwidth } keys %price;

 # So the Asian stuff comes out in an order that someone
 # who reads those scripts won't freak out over; the
 # CJK stuff will be in JIS X 0208 order that way.
 my $coll  = Unicode::Collate::Locale->new(locale => "ja");

 for my $item ($coll->sort(keys %price)) {
     print pad(entitle($item), $width, ".");
     printf " X%.2f\n", $price{$item};

 sub pad($$$) {
     my($str, $width, $padchar) = @_;
     return $str . ($padchar x ($width - colwidth($str)));

 sub colwidth(_) {
     my($str) = @_;
     return Unicode::GCString->new($str)->columns;

 sub entitle(_) {
     my($str) = @_;
     $str =~ s{ (?=\pL)(\S)     (\S*) }
              { ucfirst($1) . lc($2)  }xge;
     return $str;

See these manpages, some of which are CPAN modules: perlunicode, perluniprops, perlre, perlrecharclass, perluniintro, perlunitut, perlunifaq, PerlIO, DB_File, DBM_Filter, DBM_Filter::utf8, Encode, Encode::Locale, Unicode::UCD, Unicode::Normalize, Unicode::GCString, Unicode::LineBreak, Unicode::Collate, Unicode::Collate::Locale, Unicode::Unihan, Unicode::CaseFold, Unicode::Tussle, Lingua::JA::Romanize::Japanese, Lingua::ZH::Romanize::Pinyin, Lingua::KO::Romanize::Hangul.

The Unicode::Tussle CPAN module includes many programs to help with working with Unicode, including these programs to fully or partly replace standard utilities: tcgrep instead of egrep, uniquote instead of cat -v or hexdump, uniwc instead of wc, unilook instead of look, unifmt instead of fmt, and ucsort instead of sort. For exploring Unicode character names and character properties, see its uniprops, unichars, and uninames programs. It also supplies these programs, all of which are general filters that do Unicode-y things: unititle and unicaps; uniwide and uninarrow; unisupers and unisubs; nfd, nfc, nfkd, and nfkc; and uc, lc, and tc.

Finally, see the published Unicode Standard (page numbers are from version 6.0.0), including these specific annexes and technical reports:

X3.13 Default Case Algorithms, page 113; X4.2 Case, pages 120X122; Case Mappings, page 166X172, especially Caseless Matching starting on page 170.
UAX #44: Unicode Character Database
UTS #18: Unicode Regular Expressions
UAX #15: Unicode Normalization Forms
UTS #10: Unicode Collation Algorithm
UAX #29: Unicode Text Segmentation
UAX #14: Unicode Line Breaking Algorithm
UAX #11: East Asian Width

Tom Christiansen <> wrote this, with occasional kibbitzing from Larry Wall and Jeffrey Friedl in the background.

Copyright X 2012 Tom Christiansen.

This program is free software; you may redistribute it and/or modify it under the same terms as Perl itself.

Most of these examples taken from the current edition of the XCamel BookX; that is, from the 4XX Edition of Programming Perl, Copyright X 2012 Tom Christiansen <et al.>, 2012-02-13 by OXReilly Media. The code itself is freely redistributable, and you are encouraged to transplant, fold, spindle, and mutilate any of the examples in this manpage however you please for inclusion into your own programs without any encumbrance whatsoever. Acknowledgement via code comment is polite but not required.

v1.0.0 X first public release, 2012-02-27
2019-10-21 perl v5.30.3