3 perlunicode - Unicode support in Perl
7 =head2 Important Caveats
9 Unicode support is an extensive requirement. While Perl does not
10 implement the Unicode standard or the accompanying technical reports
11 from cover to cover, Perl does support many Unicode features.
15 =item Input and Output Layers
17 Perl knows when a filehandle uses Perl's internal Unicode encodings
18 (UTF-8, or UTF-EBCDIC if in EBCDIC) if the filehandle is opened with
19 the ":utf8" layer. Other encodings can be converted to Perl's
20 encoding on input or from Perl's encoding on output by use of the
21 ":encoding(...)" layer. See L<open>.
23 To indicate that Perl source itself is using a particular encoding,
26 =item Regular Expressions
28 The regular expression compiler produces polymorphic opcodes. That is,
29 the pattern adapts to the data and automatically switches to the Unicode
30 character scheme when presented with Unicode data--or instead uses
31 a traditional byte scheme when presented with byte data.
33 =item C<use utf8> still needed to enable UTF-8/UTF-EBCDIC in scripts
35 As a compatibility measure, the C<use utf8> pragma must be explicitly
36 included to enable recognition of UTF-8 in the Perl scripts themselves
37 (in string or regular expression literals, or in identifier names) on
38 ASCII-based machines or to recognize UTF-EBCDIC on EBCDIC-based
39 machines. B<These are the only times when an explicit C<use utf8>
40 is needed.> See L<utf8>.
42 You can also use the C<encoding> pragma to change the default encoding
43 of the data in your script; see L<encoding>.
47 =head2 Byte and Character Semantics
49 Beginning with version 5.6, Perl uses logically-wide characters to
50 represent strings internally.
52 In future, Perl-level operations will be expected to work with
53 characters rather than bytes.
55 However, as an interim compatibility measure, Perl aims to
56 provide a safe migration path from byte semantics to character
57 semantics for programs. For operations where Perl can unambiguously
58 decide that the input data are characters, Perl switches to
59 character semantics. For operations where this determination cannot
60 be made without additional information from the user, Perl decides in
61 favor of compatibility and chooses to use byte semantics.
63 This behavior preserves compatibility with earlier versions of Perl,
64 which allowed byte semantics in Perl operations only if
65 none of the program's inputs were marked as being as source of Unicode
66 character data. Such data may come from filehandles, from calls to
67 external programs, from information provided by the system (such as %ENV),
68 or from literals and constants in the source text.
70 On Windows platforms, if the C<-C> command line switch is used or the
71 ${^WIDE_SYSTEM_CALLS} global flag is set to C<1>, all system calls
72 will use the corresponding wide-character APIs. This feature is
73 available only on Windows to conform to the API standard already
74 established for that platform--and there are very few non-Windows
75 platforms that have Unicode-aware APIs.
77 The C<bytes> pragma will always, regardless of platform, force byte
78 semantics in a particular lexical scope. See L<bytes>.
80 The C<utf8> pragma is primarily a compatibility device that enables
81 recognition of UTF-(8|EBCDIC) in literals encountered by the parser.
82 Note that this pragma is only required while Perl defaults to byte
83 semantics; when character semantics become the default, this pragma
84 may become a no-op. See L<utf8>.
86 Unless explicitly stated, Perl operators use character semantics
87 for Unicode data and byte semantics for non-Unicode data.
88 The decision to use character semantics is made transparently. If
89 input data comes from a Unicode source--for example, if a character
90 encoding layer is added to a filehandle or a literal Unicode
91 string constant appears in a program--character semantics apply.
92 Otherwise, byte semantics are in effect. The C<bytes> pragma should
93 be used to force byte semantics on Unicode data.
95 If strings operating under byte semantics and strings with Unicode
96 character data are concatenated, the new string will be upgraded to
97 I<ISO 8859-1 (Latin-1)>, even if the old Unicode string used EBCDIC.
98 This translation is done without regard to the system's native 8-bit
99 encoding, so to change this for systems with non-Latin-1 and
100 non-EBCDIC native encodings use the C<encoding> pragma. See
103 Under character semantics, many operations that formerly operated on
104 bytes now operate on characters. A character in Perl is
105 logically just a number ranging from 0 to 2**31 or so. Larger
106 characters may encode into longer sequences of bytes internally, but
107 this internal detail is mostly hidden for Perl code.
108 See L<perluniintro> for more.
110 =head2 Effects of Character Semantics
112 Character semantics have the following effects:
118 Strings--including hash keys--and regular expression patterns may
119 contain characters that have an ordinal value larger than 255.
121 If you use a Unicode editor to edit your program, Unicode characters
122 may occur directly within the literal strings in one of the various
123 Unicode encodings (UTF-8, UTF-EBCDIC, UCS-2, etc.), but will be recognized
124 as such and converted to Perl's internal representation only if the
125 appropriate L<encoding> is specified.
127 Unicode characters can also be added to a string by using the
128 C<\x{...}> notation. The Unicode code for the desired character, in
129 hexadecimal, should be placed in the braces. For instance, a smiley
130 face is C<\x{263A}>. This encoding scheme only works for characters
131 with a code of 0x100 or above.
135 use charnames ':full';
137 you can use the C<\N{...}> notation and put the official Unicode
138 character name within the braces, such as C<\N{WHITE SMILING FACE}>.
143 If an appropriate L<encoding> is specified, identifiers within the
144 Perl script may contain Unicode alphanumeric characters, including
145 ideographs. Perl does not currently attempt to canonicalize variable
150 Regular expressions match characters instead of bytes. "." matches
151 a character instead of a byte. The C<\C> pattern is provided to force
152 a match a single byte--a C<char> in C, hence C<\C>.
156 Character classes in regular expressions match characters instead of
157 bytes and match against the character properties specified in the
158 Unicode properties database. C<\w> can be used to match a Japanese
159 ideograph, for instance.
163 Named Unicode properties, scripts, and block ranges may be used like
164 character classes via the C<\p{}> "matches property" construct and
165 the C<\P{}> negation, "doesn't match property".
167 For instance, C<\p{Lu}> matches any character with the Unicode "Lu"
168 (Letter, uppercase) property, while C<\p{M}> matches any character
169 with an "M" (mark--accents and such) property. Brackets are not
170 required for single letter properties, so C<\p{M}> is equivalent to
171 C<\pM>. Many predefined properties are available, such as
172 C<\p{Mirrored}> and C<\p{Tibetan}>.
174 The official Unicode script and block names have spaces and dashes as
175 separators, but for convenience you can use dashes, spaces, or
176 underbars, and case is unimportant. It is recommended, however, that
177 for consistency you use the following naming: the official Unicode
178 script, property, or block name (see below for the additional rules
179 that apply to block names) with whitespace and dashes removed, and the
180 words "uppercase-first-lowercase-rest". C<Latin-1 Supplement> thus
181 becomes C<Latin1Supplement>.
183 You can also use negation in both C<\p{}> and C<\P{}> by introducing a caret
184 (^) between the first brace and the property name: C<\p{^Tamil}> is
185 equal to C<\P{Tamil}>.
187 Here are the basic Unicode General Category properties, followed by their
188 long form. You can use either; C<\p{Lu}> and C<\p{LowercaseLetter}>,
189 for instance, are identical.
211 Pc ConnectorPunctuation
215 Pi InitialPunctuation
216 (may behave like Ps or Pe depending on usage)
218 (may behave like Ps or Pe depending on usage)
230 Zp ParagraphSeparator
235 Cs Surrogate (not usable)
239 Single-letter properties match all characters in any of the
240 two-letter sub-properties starting with the same letter.
241 C<L&> is a special case, which is an alias for C<Ll>, C<Lu>, and C<Lt>.
243 Because Perl hides the need for the user to understand the internal
244 representation of Unicode characters, there is no need to implement
245 the somewhat messy concept of surrogates. C<Cs> is therefore not
248 Because scripts differ in their directionality--Hebrew is
249 written right to left, for example--Unicode supplies these properties:
254 BidiLRE Left-to-Right Embedding
255 BidiLRO Left-to-Right Override
257 BidiAL Right-to-Left Arabic
258 BidiRLE Right-to-Left Embedding
259 BidiRLO Right-to-Left Override
260 BidiPDF Pop Directional Format
261 BidiEN European Number
262 BidiES European Number Separator
263 BidiET European Number Terminator
265 BidiCS Common Number Separator
266 BidiNSM Non-Spacing Mark
267 BidiBN Boundary Neutral
268 BidiB Paragraph Separator
269 BidiS Segment Separator
271 BidiON Other Neutrals
273 For example, C<\p{BidiR}> matches characters that are normally
274 written right to left.
280 The script names which can be used by C<\p{...}> and C<\P{...}>,
281 such as in C<\p{Latin}> or C<\p{Cyrillic}>, are as follows:
328 Extended property classes can supplement the basic
329 properties, defined by the F<PropList> Unicode database:
344 LogicalOrderException
345 NoncharacterCodePoint
347 OtherDefaultIgnorableCodePoint
359 and there are further derived properties:
361 Alphabetic Lu + Ll + Lt + Lm + Lo + OtherAlphabetic
362 Lowercase Ll + OtherLowercase
363 Uppercase Lu + OtherUppercase
366 ID_Start Lu + Ll + Lt + Lm + Lo + Nl
367 ID_Continue ID_Start + Mn + Mc + Nd + Pc
370 Assigned Any non-Cn character (i.e. synonym for \P{Cn})
371 Unassigned Synonym for \p{Cn}
372 Common Any character (or unassigned code point)
373 not explicitly assigned to a script
375 For backward compatibility (with Perl 5.6), all properties mentioned
376 so far may have C<Is> prepended to their name, so C<\P{IsLu}>, for
377 example, is equal to C<\P{Lu}>.
381 In addition to B<scripts>, Unicode also defines B<blocks> of
382 characters. The difference between scripts and blocks is that the
383 concept of scripts is closer to natural languages, while the concept
384 of blocks is more of an artificial grouping based on groups of 256
385 Unicode characters. For example, the C<Latin> script contains letters
386 from many blocks but does not contain all the characters from those
387 blocks. It does not, for example, contain digits, because digits are
388 shared across many scripts. Digits and similar groups, like
389 punctuation, are in a category called C<Common>.
391 For more about scripts, see the UTR #24:
393 http://www.unicode.org/unicode/reports/tr24/
395 For more about blocks, see:
397 http://www.unicode.org/Public/UNIDATA/Blocks.txt
399 Block names are given with the C<In> prefix. For example, the
400 Katakana block is referenced via C<\p{InKatakana}>. The C<In>
401 prefix may be omitted if there is no naming conflict with a script
402 or any other property, but it is recommended that C<In> always be used
403 for block tests to avoid confusion.
405 These block names are supported:
407 InAlphabeticPresentationForms
409 InArabicPresentationFormsA
410 InArabicPresentationFormsB
421 InByzantineMusicalSymbols
423 InCJKCompatibilityForms
424 InCJKCompatibilityIdeographs
425 InCJKCompatibilityIdeographsSupplement
426 InCJKRadicalsSupplement
427 InCJKSymbolsAndPunctuation
428 InCJKUnifiedIdeographs
429 InCJKUnifiedIdeographsExtensionA
430 InCJKUnifiedIdeographsExtensionB
432 InCombiningDiacriticalMarks
433 InCombiningDiacriticalMarksforSymbols
438 InCyrillicSupplementary
442 InEnclosedAlphanumerics
443 InEnclosedCJKLettersAndMonths
453 InHalfwidthAndFullwidthForms
454 InHangulCompatibilityJamo
459 InHighPrivateUseSurrogates
463 InIdeographicDescriptionCharacters
468 InKatakanaPhoneticExtensions
473 InLatinExtendedAdditional
478 InMathematicalAlphanumericSymbols
479 InMathematicalOperators
480 InMiscellaneousMathematicalSymbolsA
481 InMiscellaneousMathematicalSymbolsB
482 InMiscellaneousSymbols
483 InMiscellaneousTechnical
490 InOpticalCharacterRecognition
496 InSpacingModifierLetters
498 InSuperscriptsAndSubscripts
499 InSupplementalArrowsA
500 InSupplementalArrowsB
501 InSupplementalMathematicalOperators
502 InSupplementaryPrivateUseAreaA
503 InSupplementaryPrivateUseAreaB
513 InUnifiedCanadianAboriginalSyllabics
522 The special pattern C<\X> matches any extended Unicode
523 sequence--"a combining character sequence" in Standardese--where the
524 first character is a base character and subsequent characters are mark
525 characters that apply to the base character. C<\X> is equivalent to
530 The C<tr///> operator translates characters instead of bytes. Note
531 that the C<tr///CU> functionality has been removed. For similar
532 functionality see pack('U0', ...) and pack('C0', ...).
536 Case translation operators use the Unicode case translation tables
537 when character input is provided. Note that C<uc()>, or C<\U> in
538 interpolated strings, translates to uppercase, while C<ucfirst>,
539 or C<\u> in interpolated strings, translates to titlecase in languages
540 that make the distinction.
544 Most operators that deal with positions or lengths in a string will
545 automatically switch to using character positions, including
546 C<chop()>, C<substr()>, C<pos()>, C<index()>, C<rindex()>,
547 C<sprintf()>, C<write()>, and C<length()>. Operators that
548 specifically do not switch include C<vec()>, C<pack()>, and
549 C<unpack()>. Operators that really don't care include C<chomp()>,
550 operators that treats strings as a bucket of bits such as C<sort()>,
551 and operators dealing with filenames.
555 The C<pack()>/C<unpack()> letters C<c> and C<C> do I<not> change,
556 since they are often used for byte-oriented formats. Again, think
557 C<char> in the C language.
559 There is a new C<U> specifier that converts between Unicode characters
564 The C<chr()> and C<ord()> functions work on characters, similar to
565 C<pack("U")> and C<unpack("U")>, I<not> C<pack("C")> and
566 C<unpack("C")>. C<pack("C")> and C<unpack("C")> are methods for
567 emulating byte-oriented C<chr()> and C<ord()> on Unicode strings.
568 While these methods reveal the internal encoding of Unicode strings,
569 that is not something one normally needs to care about at all.
573 The bit string operators, C<& | ^ ~>, can operate on character data.
574 However, for backward compatibility, such as when using bit string
575 operations when characters are all less than 256 in ordinal value, one
576 should not use C<~> (the bit complement) with characters of both
577 values less than 256 and values greater than 256. Most importantly,
578 DeMorgan's laws (C<~($x|$y) eq ~$x&~$y> and C<~($x&$y) eq ~$x|~$y>)
579 will not hold. The reason for this mathematical I<faux pas> is that
580 the complement cannot return B<both> the 8-bit (byte-wide) bit
581 complement B<and> the full character-wide bit complement.
585 lc(), uc(), lcfirst(), and ucfirst() work for the following cases:
591 the case mapping is from a single Unicode character to another
592 single Unicode character, or
596 the case mapping is from a single Unicode character to more
597 than one Unicode character.
601 Things to do with locales (Lithuanian, Turkish, Azeri) do B<not> work
602 since Perl does not understand the concept of Unicode locales.
604 See the Unicode Technical Report #21, Case Mappings, for more details.
612 And finally, C<scalar reverse()> reverses by character rather than by byte.
616 =head2 User-Defined Character Properties
618 You can define your own character properties by defining subroutines
619 whose names begin with "In" or "Is". The subroutines must be defined
620 in the C<main> package. The user-defined properties can be used in the
621 regular expression C<\p> and C<\P> constructs. Note that the effect
622 is compile-time and immutable once defined.
624 The subroutines must return a specially-formatted string, with one
625 or more newline-separated lines. Each line must be one of the following:
631 Two hexadecimal numbers separated by horizontal whitespace (space or
632 tabular characters) denoting a range of Unicode code points to include.
636 Something to include, prefixed by "+": a built-in character
637 property (prefixed by "utf8::"), to represent all the characters in that
638 property; two hexadecimal code points for a range; or a single
639 hexadecimal code point.
643 Something to exclude, prefixed by "-": an existing character
644 property (prefixed by "utf8::"), for all the characters in that
645 property; two hexadecimal code points for a range; or a single
646 hexadecimal code point.
650 Something to negate, prefixed "!": an existing character
651 property (prefixed by "utf8::") for all the characters except the
652 characters in the property; two hexadecimal code points for a range;
653 or a single hexadecimal code point.
657 For example, to define a property that covers both the Japanese
658 syllabaries (hiragana and katakana), you can define
667 Imagine that the here-doc end marker is at the beginning of the line.
668 Now you can use C<\p{InKana}> and C<\P{InKana}>.
670 You could also have used the existing block property names:
679 Suppose you wanted to match only the allocated characters,
680 not the raw block ranges: in other words, you want to remove
691 The negation is useful for defining (surprise!) negated classes.
701 You can also define your own mappings to be used in the lc(),
702 lcfirst(), uc(), and ucfirst() (or their string-inlined versions).
703 The principle is the same: define subroutines in the C<main> package
704 with names like C<ToLower> (for lc() and lcfirst()), C<ToTitle> (for
705 the first character in ucfirst()), and C<ToUpper> (for uc(), and the
706 rest of the characters in ucfirst()).
708 The string returned by the subroutines needs now to be three
709 hexadecimal numbers separated by tabulators: start of the source
710 range, end of the source range, and start of the destination range.
719 defines an uc() mapping that causes only the characters "a", "b", and
720 "c" to be mapped to "A", "B", "C", all other characters will remain
723 If there is no source range to speak of, that is, the mapping is from
724 a single character to another single character, leave the end of the
725 source range empty, but the two tabulator characters are still needed.
734 defines a lc() mapping that causes only "A" to be mapped to "a", all
735 other characters will remain unchanged.
737 (For serious hackers only) If you want to introspect the default
738 mappings, you can find the data in the directory
739 C<$Config{privlib}>/F<unicore/To/>. The mapping data is returned as
740 the here-document, and the C<utf8::ToSpecFoo> are special exception
741 mappings derived from <$Config{privlib}>/F<unicore/SpecialCasing.txt>.
742 The C<Digit> and C<Fold> mappings that one can see in the directory
743 are not directly user-accessible, one can use either the
744 C<Unicode::UCD> module, or just match case-insensitively (that's when
745 the C<Fold> mapping is used).
747 A final note on the user-defined property tests and mappings: they
748 will be used only if the scalar has been marked as having Unicode
749 characters. Old byte-style strings will not be affected.
751 =head2 Character Encodings for Input and Output
755 =head2 Unicode Regular Expression Support Level
757 The following list of Unicode support for regular expressions describes
758 all the features currently supported. The references to "Level N"
759 and the section numbers refer to the Unicode Technical Report 18,
760 "Unicode Regular Expression Guidelines".
766 Level 1 - Basic Unicode Support
768 2.1 Hex Notation - done [1]
769 Named Notation - done [2]
770 2.2 Categories - done [3][4]
771 2.3 Subtraction - MISSING [5][6]
772 2.4 Simple Word Boundaries - done [7]
773 2.5 Simple Loose Matches - done [8]
774 2.6 End of Line - MISSING [9][10]
778 [ 3] . \p{...} \P{...}
779 [ 4] now scripts (see UTR#24 Script Names) in addition to blocks
781 [ 6] can use regular expression look-ahead [a]
782 or user-defined character properties [b] to emulate subtraction
783 [ 7] include Letters in word characters
784 [ 8] note that Perl does Full case-folding in matching, not Simple:
785 for example U+1F88 is equivalent with U+1F00 U+03B9,
786 not with 1F80. This difference matters for certain Greek
787 capital letters with certain modifiers: the Full case-folding
788 decomposes the letter, while the Simple case-folding would map
789 it to a single character.
790 [ 9] see UTR#13 Unicode Newline Guidelines
791 [10] should do ^ and $ also on \x{85}, \x{2028} and \x{2029}
792 (should also affect <>, $., and script line numbers)
793 (the \x{85}, \x{2028} and \x{2029} do match \s)
795 [a] You can mimic class subtraction using lookahead.
796 For example, what TR18 might write as
798 [{Greek}-[{UNASSIGNED}]]
800 in Perl can be written as:
802 (?!\p{Unassigned})\p{InGreekAndCoptic}
803 (?=\p{Assigned})\p{InGreekAndCoptic}
805 But in this particular example, you probably really want
809 which will match assigned characters known to be part of the Greek script.
811 [b] See L</"User-Defined Character Properties">.
815 Level 2 - Extended Unicode Support
817 3.1 Surrogates - MISSING [11]
818 3.2 Canonical Equivalents - MISSING [12][13]
819 3.3 Locale-Independent Graphemes - MISSING [14]
820 3.4 Locale-Independent Words - MISSING [15]
821 3.5 Locale-Independent Loose Matches - MISSING [16]
823 [11] Surrogates are solely a UTF-16 concept and Perl's internal
824 representation is UTF-8. The Encode module does UTF-16, though.
825 [12] see UTR#15 Unicode Normalization
826 [13] have Unicode::Normalize but not integrated to regexes
827 [14] have \X but at this level . should equal that
828 [15] need three classes, not just \w and \W
829 [16] see UTR#21 Case Mappings
833 Level 3 - Locale-Sensitive Support
835 4.1 Locale-Dependent Categories - MISSING
836 4.2 Locale-Dependent Graphemes - MISSING [16][17]
837 4.3 Locale-Dependent Words - MISSING
838 4.4 Locale-Dependent Loose Matches - MISSING
839 4.5 Locale-Dependent Ranges - MISSING
841 [16] see UTR#10 Unicode Collation Algorithms
842 [17] have Unicode::Collate but not integrated to regexes
846 =head2 Unicode Encodings
848 Unicode characters are assigned to I<code points>, which are abstract
849 numbers. To use these numbers, various encodings are needed.
857 UTF-8 is a variable-length (1 to 6 bytes, current character allocations
858 require 4 bytes), byte-order independent encoding. For ASCII (and we
859 really do mean 7-bit ASCII, not another 8-bit encoding), UTF-8 is
862 The following table is from Unicode 3.2.
864 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
866 U+0000..U+007F 00..7F
867 U+0080..U+07FF C2..DF 80..BF
868 U+0800..U+0FFF E0 A0..BF 80..BF
869 U+1000..U+CFFF E1..EC 80..BF 80..BF
870 U+D000..U+D7FF ED 80..9F 80..BF
871 U+D800..U+DFFF ******* ill-formed *******
872 U+E000..U+FFFF EE..EF 80..BF 80..BF
873 U+10000..U+3FFFF F0 90..BF 80..BF 80..BF
874 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
875 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
877 Note the C<A0..BF> in C<U+0800..U+0FFF>, the C<80..9F> in
878 C<U+D000...U+D7FF>, the C<90..B>F in C<U+10000..U+3FFFF>, and the
879 C<80...8F> in C<U+100000..U+10FFFF>. The "gaps" are caused by legal
880 UTF-8 avoiding non-shortest encodings: it is technically possible to
881 UTF-8-encode a single code point in different ways, but that is
882 explicitly forbidden, and the shortest possible encoding should always
883 be used. So that's what Perl does.
885 Another way to look at it is via bits:
887 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
890 00000bbbbbaaaaaa 110bbbbb 10aaaaaa
891 ccccbbbbbbaaaaaa 1110cccc 10bbbbbb 10aaaaaa
892 00000dddccccccbbbbbbaaaaaa 11110ddd 10cccccc 10bbbbbb 10aaaaaa
894 As you can see, the continuation bytes all begin with C<10>, and the
895 leading bits of the start byte tell how many bytes the are in the
902 Like UTF-8 but EBCDIC-safe, in the way that UTF-8 is ASCII-safe.
906 UTF-16, UTF-16BE, UTF16-LE, Surrogates, and BOMs (Byte Order Marks)
908 The followings items are mostly for reference and general Unicode
909 knowledge, Perl doesn't use these constructs internally.
911 UTF-16 is a 2 or 4 byte encoding. The Unicode code points
912 C<U+0000..U+FFFF> are stored in a single 16-bit unit, and the code
913 points C<U+10000..U+10FFFF> in two 16-bit units. The latter case is
914 using I<surrogates>, the first 16-bit unit being the I<high
915 surrogate>, and the second being the I<low surrogate>.
917 Surrogates are code points set aside to encode the C<U+10000..U+10FFFF>
918 range of Unicode code points in pairs of 16-bit units. The I<high
919 surrogates> are the range C<U+D800..U+DBFF>, and the I<low surrogates>
920 are the range C<U+DC00..U+DFFF>. The surrogate encoding is
922 $hi = ($uni - 0x10000) / 0x400 + 0xD800;
923 $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
927 $uni = 0x10000 + ($hi - 0xD800) * 0x400 + ($lo - 0xDC00);
929 If you try to generate surrogates (for example by using chr()), you
930 will get a warning if warnings are turned on, because those code
931 points are not valid for a Unicode character.
933 Because of the 16-bitness, UTF-16 is byte-order dependent. UTF-16
934 itself can be used for in-memory computations, but if storage or
935 transfer is required either UTF-16BE (big-endian) or UTF-16LE
936 (little-endian) encodings must be chosen.
938 This introduces another problem: what if you just know that your data
939 is UTF-16, but you don't know which endianness? Byte Order Marks, or
940 BOMs, are a solution to this. A special character has been reserved
941 in Unicode to function as a byte order marker: the character with the
942 code point C<U+FEFF> is the BOM.
944 The trick is that if you read a BOM, you will know the byte order,
945 since if it was written on a big-endian platform, you will read the
946 bytes C<0xFE 0xFF>, but if it was written on a little-endian platform,
947 you will read the bytes C<0xFF 0xFE>. (And if the originating platform
948 was writing in UTF-8, you will read the bytes C<0xEF 0xBB 0xBF>.)
950 The way this trick works is that the character with the code point
951 C<U+FFFE> is guaranteed not to be a valid Unicode character, so the
952 sequence of bytes C<0xFF 0xFE> is unambiguously "BOM, represented in
953 little-endian format" and cannot be C<U+FFFE>, represented in big-endian
958 UTF-32, UTF-32BE, UTF32-LE
960 The UTF-32 family is pretty much like the UTF-16 family, expect that
961 the units are 32-bit, and therefore the surrogate scheme is not
962 needed. The BOM signatures will be C<0x00 0x00 0xFE 0xFF> for BE and
963 C<0xFF 0xFE 0x00 0x00> for LE.
969 Encodings defined by the ISO 10646 standard. UCS-2 is a 16-bit
970 encoding. Unlike UTF-16, UCS-2 is not extensible beyond C<U+FFFF>,
971 because it does not use surrogates. UCS-4 is a 32-bit encoding,
972 functionally identical to UTF-32.
978 A seven-bit safe (non-eight-bit) encoding, which is useful if the
979 transport or storage is not eight-bit safe. Defined by RFC 2152.
983 =head2 Security Implications of Unicode
991 Unfortunately, the specification of UTF-8 leaves some room for
992 interpretation of how many bytes of encoded output one should generate
993 from one input Unicode character. Strictly speaking, the shortest
994 possible sequence of UTF-8 bytes should be generated,
995 because otherwise there is potential for an input buffer overflow at
996 the receiving end of a UTF-8 connection. Perl always generates the
997 shortest length UTF-8, and with warnings on Perl will warn about
998 non-shortest length UTF-8 along with other malformations, such as the
999 surrogates, which are not real Unicode code points.
1003 Regular expressions behave slightly differently between byte data and
1004 character (Unicode) data. For example, the "word character" character
1005 class C<\w> will work differently depending on if data is eight-bit bytes
1008 In the first case, the set of C<\w> characters is either small--the
1009 default set of alphabetic characters, digits, and the "_"--or, if you
1010 are using a locale (see L<perllocale>), the C<\w> might contain a few
1011 more letters according to your language and country.
1013 In the second case, the C<\w> set of characters is much, much larger.
1014 Most importantly, even in the set of the first 256 characters, it will
1015 probably match different characters: unlike most locales, which are
1016 specific to a language and country pair, Unicode classifies all the
1017 characters that are letters I<somewhere> as C<\w>. For example, your
1018 locale might not think that LATIN SMALL LETTER ETH is a letter (unless
1019 you happen to speak Icelandic), but Unicode does.
1021 As discussed elsewhere, Perl has one foot (two hooves?) planted in
1022 each of two worlds: the old world of bytes and the new world of
1023 characters, upgrading from bytes to characters when necessary.
1024 If your legacy code does not explicitly use Unicode, no automatic
1025 switch-over to characters should happen. Characters shouldn't get
1026 downgraded to bytes, either. It is possible to accidentally mix bytes
1027 and characters, however (see L<perluniintro>), in which case C<\w> in
1028 regular expressions might start behaving differently. Review your
1029 code. Use warnings and the C<strict> pragma.
1033 =head2 Unicode in Perl on EBCDIC
1035 The way Unicode is handled on EBCDIC platforms is still
1036 experimental. On such platforms, references to UTF-8 encoding in this
1037 document and elsewhere should be read as meaning the UTF-EBCDIC
1038 specified in Unicode Technical Report 16, unless ASCII vs. EBCDIC issues
1039 are specifically discussed. There is no C<utfebcdic> pragma or
1040 ":utfebcdic" layer; rather, "utf8" and ":utf8" are reused to mean
1041 the platform's "natural" 8-bit encoding of Unicode. See L<perlebcdic>
1042 for more discussion of the issues.
1046 Usually locale settings and Unicode do not affect each other, but
1047 there are a couple of exceptions:
1053 If your locale environment variables (LANGUAGE, LC_ALL, LC_CTYPE, LANG)
1054 contain the strings 'UTF-8' or 'UTF8' (case-insensitive matching),
1055 the default encodings of your STDIN, STDOUT, and STDERR, and of
1056 B<any subsequent file open>, are considered to be UTF-8.
1060 Perl tries really hard to work both with Unicode and the old
1061 byte-oriented world. Most often this is nice, but sometimes Perl's
1062 straddling of the proverbial fence causes problems.
1066 =head2 Using Unicode in XS
1068 If you want to handle Perl Unicode in XS extensions, you may find the
1069 following C APIs useful. See also L<perlguts/"Unicode Support"> for an
1070 explanation about Unicode at the XS level, and L<perlapi> for the API
1077 C<DO_UTF8(sv)> returns true if the C<UTF8> flag is on and the bytes
1078 pragma is not in effect. C<SvUTF8(sv)> returns true is the C<UTF8>
1079 flag is on; the bytes pragma is ignored. The C<UTF8> flag being on
1080 does B<not> mean that there are any characters of code points greater
1081 than 255 (or 127) in the scalar or that there are even any characters
1082 in the scalar. What the C<UTF8> flag means is that the sequence of
1083 octets in the representation of the scalar is the sequence of UTF-8
1084 encoded code points of the characters of a string. The C<UTF8> flag
1085 being off means that each octet in this representation encodes a
1086 single character with code point 0..255 within the string. Perl's
1087 Unicode model is not to use UTF-8 until it is absolutely necessary.
1091 C<uvuni_to_utf8(buf, chr>) writes a Unicode character code point into
1092 a buffer encoding the code point as UTF-8, and returns a pointer
1093 pointing after the UTF-8 bytes.
1097 C<utf8_to_uvuni(buf, lenp)> reads UTF-8 encoded bytes from a buffer and
1098 returns the Unicode character code point and, optionally, the length of
1099 the UTF-8 byte sequence.
1103 C<utf8_length(start, end)> returns the length of the UTF-8 encoded buffer
1104 in characters. C<sv_len_utf8(sv)> returns the length of the UTF-8 encoded
1109 C<sv_utf8_upgrade(sv)> converts the string of the scalar to its UTF-8
1110 encoded form. C<sv_utf8_downgrade(sv)> does the opposite, if
1111 possible. C<sv_utf8_encode(sv)> is like sv_utf8_upgrade except that
1112 it does not set the C<UTF8> flag. C<sv_utf8_decode()> does the
1113 opposite of C<sv_utf8_encode()>. Note that none of these are to be
1114 used as general-purpose encoding or decoding interfaces: C<use Encode>
1115 for that. C<sv_utf8_upgrade()> is affected by the encoding pragma
1116 but C<sv_utf8_downgrade()> is not (since the encoding pragma is
1117 designed to be a one-way street).
1121 C<is_utf8_char(s)> returns true if the pointer points to a valid UTF-8
1126 C<is_utf8_string(buf, len)> returns true if C<len> bytes of the buffer
1131 C<UTF8SKIP(buf)> will return the number of bytes in the UTF-8 encoded
1132 character in the buffer. C<UNISKIP(chr)> will return the number of bytes
1133 required to UTF-8-encode the Unicode character code point. C<UTF8SKIP()>
1134 is useful for example for iterating over the characters of a UTF-8
1135 encoded buffer; C<UNISKIP()> is useful, for example, in computing
1136 the size required for a UTF-8 encoded buffer.
1140 C<utf8_distance(a, b)> will tell the distance in characters between the
1141 two pointers pointing to the same UTF-8 encoded buffer.
1145 C<utf8_hop(s, off)> will return a pointer to an UTF-8 encoded buffer
1146 that is C<off> (positive or negative) Unicode characters displaced
1147 from the UTF-8 buffer C<s>. Be careful not to overstep the buffer:
1148 C<utf8_hop()> will merrily run off the end or the beginning of the
1149 buffer if told to do so.
1153 C<pv_uni_display(dsv, spv, len, pvlim, flags)> and
1154 C<sv_uni_display(dsv, ssv, pvlim, flags)> are useful for debugging the
1155 output of Unicode strings and scalars. By default they are useful
1156 only for debugging--they display B<all> characters as hexadecimal code
1157 points--but with the flags C<UNI_DISPLAY_ISPRINT>,
1158 C<UNI_DISPLAY_BACKSLASH>, and C<UNI_DISPLAY_QQ> you can make the
1159 output more readable.
1163 C<ibcmp_utf8(s1, pe1, u1, l1, u1, s2, pe2, l2, u2)> can be used to
1164 compare two strings case-insensitively in Unicode. For case-sensitive
1165 comparisons you can just use C<memEQ()> and C<memNE()> as usual.
1169 For more information, see L<perlapi>, and F<utf8.c> and F<utf8.h>
1170 in the Perl source code distribution.
1174 =head2 Interaction with Locales
1176 Use of locales with Unicode data may lead to odd results. Currently,
1177 Perl attempts to attach 8-bit locale info to characters in the range
1178 0..255, but this technique is demonstrably incorrect for locales that
1179 use characters above that range when mapped into Unicode. Perl's
1180 Unicode support will also tend to run slower. Use of locales with
1181 Unicode is discouraged.
1183 =head2 Interaction with Extensions
1185 When Perl exchanges data with an extension, the extension should be
1186 able to understand the UTF-8 flag and act accordingly. If the
1187 extension doesn't know about the flag, it's likely that the extension
1188 will return incorrectly-flagged data.
1190 So if you're working with Unicode data, consult the documentation of
1191 every module you're using if there are any issues with Unicode data
1192 exchange. If the documentation does not talk about Unicode at all,
1193 suspect the worst and probably look at the source to learn how the
1194 module is implemented. Modules written completely in Perl shouldn't
1195 cause problems. Modules that directly or indirectly access code written
1196 in other programming languages are at risk.
1198 For affected functions, the simple strategy to avoid data corruption is
1199 to always make the encoding of the exchanged data explicit. Choose an
1200 encoding that you know the extension can handle. Convert arguments passed
1201 to the extensions to that encoding and convert results back from that
1202 encoding. Write wrapper functions that do the conversions for you, so
1203 you can later change the functions when the extension catches up.
1205 To provide an example, let's say the popular Foo::Bar::escape_html
1206 function doesn't deal with Unicode data yet. The wrapper function
1207 would convert the argument to raw UTF-8 and convert the result back to
1208 Perl's internal representation like so:
1210 sub my_escape_html ($) {
1212 return unless defined $what;
1213 Encode::decode_utf8(Foo::Bar::escape_html(Encode::encode_utf8($what)));
1216 Sometimes, when the extension does not convert data but just stores
1217 and retrieves them, you will be in a position to use the otherwise
1218 dangerous Encode::_utf8_on() function. Let's say the popular
1219 C<Foo::Bar> extension, written in C, provides a C<param> method that
1220 lets you store and retrieve data according to these prototypes:
1222 $self->param($name, $value); # set a scalar
1223 $value = $self->param($name); # retrieve a scalar
1225 If it does not yet provide support for any encoding, one could write a
1226 derived class with such a C<param> method:
1229 my($self,$name,$value) = @_;
1230 utf8::upgrade($name); # make sure it is UTF-8 encoded
1232 utf8::upgrade($value); # make sure it is UTF-8 encoded
1233 return $self->SUPER::param($name,$value);
1235 my $ret = $self->SUPER::param($name);
1236 Encode::_utf8_on($ret); # we know, it is UTF-8 encoded
1241 Some extensions provide filters on data entry/exit points, such as
1242 DB_File::filter_store_key and family. Look out for such filters in
1243 the documentation of your extensions, they can make the transition to
1244 Unicode data much easier.
1248 Some functions are slower when working on UTF-8 encoded strings than
1249 on byte encoded strings. All functions that need to hop over
1250 characters such as length(), substr() or index() can work B<much>
1251 faster when the underlying data are byte-encoded. Witness the
1252 following benchmark:
1258 our $u = our $b = "x" x $l;
1259 substr($u,0,1) = "\x{100}";
1261 LENGTH_B => q{ length($b) },
1262 LENGTH_U => q{ length($u) },
1263 SUBSTR_B => q{ substr($b, $l/4, $l/2) },
1264 SUBSTR_U => q{ substr($u, $l/4, $l/2) },
1267 Benchmark: running LENGTH_B, LENGTH_U, SUBSTR_B, SUBSTR_U for at least 2 CPU seconds...
1268 LENGTH_B: 2 wallclock secs ( 2.36 usr + 0.00 sys = 2.36 CPU) @ 5649983.05/s (n=13333960)
1269 LENGTH_U: 2 wallclock secs ( 2.11 usr + 0.00 sys = 2.11 CPU) @ 12155.45/s (n=25648)
1270 SUBSTR_B: 3 wallclock secs ( 2.16 usr + 0.00 sys = 2.16 CPU) @ 374480.09/s (n=808877)
1271 SUBSTR_U: 2 wallclock secs ( 2.11 usr + 0.00 sys = 2.11 CPU) @ 6791.00/s (n=14329)
1273 The numbers show an incredible slowness on long UTF-8 strings. You
1274 should carefully avoid using these functions in tight loops. If you
1275 want to iterate over characters, the superior coding technique would
1276 split the characters into an array instead of using substr, as the following
1283 our $u = our $b = "x" x $l;
1284 substr($u,0,1) = "\x{100}";
1286 SPLIT_B => q{ for my $c (split //, $b){} },
1287 SPLIT_U => q{ for my $c (split //, $u){} },
1288 SUBSTR_B => q{ for my $i (0..length($b)-1){my $c = substr($b,$i,1);} },
1289 SUBSTR_U => q{ for my $i (0..length($u)-1){my $c = substr($u,$i,1);} },
1292 Benchmark: running SPLIT_B, SPLIT_U, SUBSTR_B, SUBSTR_U for at least 5 CPU seconds...
1293 SPLIT_B: 6 wallclock secs ( 5.29 usr + 0.00 sys = 5.29 CPU) @ 56.14/s (n=297)
1294 SPLIT_U: 5 wallclock secs ( 5.17 usr + 0.01 sys = 5.18 CPU) @ 55.21/s (n=286)
1295 SUBSTR_B: 5 wallclock secs ( 5.34 usr + 0.00 sys = 5.34 CPU) @ 123.22/s (n=658)
1296 SUBSTR_U: 7 wallclock secs ( 6.20 usr + 0.00 sys = 6.20 CPU) @ 0.81/s (n=5)
1298 Even though the algorithm based on C<substr()> is faster than
1299 C<split()> for byte-encoded data, it pales in comparison to the speed
1300 of C<split()> when used with UTF-8 data.
1302 =head2 Porting code from perl-5.6.X
1304 Perl 5.8 has a different Unicode model from 5.6. In 5.6 the programmer
1305 was required to use the C<utf8> pragma to declare that a given scope
1306 expected to deal with Unicode data and had to make sure that only
1307 Unicode data were reaching that scope. If you have code that is
1308 working with 5.6, you will need some of the following adjustments to
1309 your code. The examples are written such that the code will continue
1310 to work under 5.6, so you should be safe to try them out.
1316 A filehandle that should read or write UTF-8
1319 binmode $fh, ":utf8";
1324 A scalar that is going to be passed to some extension
1326 Be it Compress::Zlib, Apache::Request or any extension that has no
1327 mention of Unicode in the manpage, you need to make sure that the
1328 UTF-8 flag is stripped off. Note that at the time of this writing
1329 (October 2002) the mentioned modules are not UTF-8-aware. Please
1330 check the documentation to verify if this is still true.
1334 $val = Encode::encode_utf8($val); # make octets
1339 A scalar we got back from an extension
1341 If you believe the scalar comes back as UTF-8, you will most likely
1342 want the UTF-8 flag restored:
1346 $val = Encode::decode_utf8($val);
1351 Same thing, if you are really sure it is UTF-8
1355 Encode::_utf8_on($val);
1360 A wrapper for fetchrow_array and fetchrow_hashref
1362 When the database contains only UTF-8, a wrapper function or method is
1363 a convenient way to replace all your fetchrow_array and
1364 fetchrow_hashref calls. A wrapper function will also make it easier to
1365 adapt to future enhancements in your database driver. Note that at the
1366 time of this writing (October 2002), the DBI has no standardized way
1367 to deal with UTF-8 data. Please check the documentation to verify if
1371 my($self, $sth, $what) = @_; # $what is one of fetchrow_{array,hashref}
1377 my @arr = $sth->$what;
1379 defined && /[^\000-\177]/ && Encode::_utf8_on($_);
1383 my $ret = $sth->$what;
1385 for my $k (keys %$ret) {
1386 defined && /[^\000-\177]/ && Encode::_utf8_on($_) for $ret->{$k};
1390 defined && /[^\000-\177]/ && Encode::_utf8_on($_) for $ret;
1400 A large scalar that you know can only contain ASCII
1402 Scalars that contain only ASCII and are marked as UTF-8 are sometimes
1403 a drag to your program. If you recognize such a situation, just remove
1406 utf8::downgrade($val) if $] > 5.007;
1412 L<perluniintro>, L<encoding>, L<Encode>, L<open>, L<utf8>, L<bytes>,
1413 L<perlretut>, L<perlvar/"${^WIDE_SYSTEM_CALLS}">