mingw/html/pod/perlunicode.html

   1 <?xml version="1.0" ?>
   2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
   3 <html xmlns="http://www.w3.org/1999/xhtml">
   4 <head>
   5 <title>perlunicode - Unicode support in Perl</title>
   6 <meta http-equiv="content-type" content="text/html; charset=utf-8" />
   7 <link rev="made" href="mailto:" />
   8 </head>
   9
  10 <body style="background-color: white">
  11 <table border="0" width="100%" cellspacing="0" cellpadding="3">
  12 <tr><td class="block" style="background-color: #cccccc" valign="middle">
  13 <big><strong><span class="block">&nbsp;perlunicode - Unicode support in Perl</span></strong></big>
  14 </td></tr>
  15 </table>
  16
  17 <p><a name="__index__"></a></p>
  18 <!-- INDEX BEGIN -->
  19
  20 <ul>
  21
  22         <li><a href="#name">NAME</a></li>
  23         <li><a href="#description">DESCRIPTION</a></li>
  24         <ul>
  25
  26                 <li><a href="#important_caveats">Important Caveats</a></li>
  27                 <li><a href="#byte_and_character_semantics">Byte and Character Semantics</a></li>
  28                 <li><a href="#effects_of_character_semantics">Effects of Character Semantics</a></li>
  29                 <li><a href="#scripts">Scripts</a></li>
  30                 <li><a href="#blocks">Blocks</a></li>
  31                 <li><a href="#userdefined_character_properties">User-Defined Character Properties</a></li>
  32                 <li><a href="#character_encodings_for_input_and_output">Character Encodings for Input and Output</a></li>
  33                 <li><a href="#unicode_regular_expression_support_level">Unicode Regular Expression Support Level</a></li>
  34                 <li><a href="#unicode_encodings">Unicode Encodings</a></li>
  35                 <li><a href="#security_implications_of_unicode">Security Implications of Unicode</a></li>
  36                 <li><a href="#unicode_in_perl_on_ebcdic">Unicode in Perl on EBCDIC</a></li>
  37                 <li><a href="#locales">Locales</a></li>
  38                 <li><a href="#when_unicode_does_not_happen">When Unicode Does Not Happen</a></li>
  39                 <li><a href="#forcing_unicode_in_perl__or_unforcing_unicode_in_perl_">Forcing Unicode in Perl (Or Unforcing Unicode in Perl)</a></li>
  40                 <li><a href="#using_unicode_in_xs">Using Unicode in XS</a></li>
  41         </ul>
  42
  43         <li><a href="#bugs">BUGS</a></li>
  44         <ul>
  45
  46                 <li><a href="#interaction_with_locales">Interaction with Locales</a></li>
  47                 <li><a href="#interaction_with_extensions">Interaction with Extensions</a></li>
  48                 <li><a href="#speed">Speed</a></li>
  49                 <li><a href="#porting_code_from_perl5_6_x">Porting code from perl-5.6.X</a></li>
  50         </ul>
  51
  52         <li><a href="#see_also">SEE ALSO</a></li>
  53 </ul>
  54 <!-- INDEX END -->
  55
  56 <hr />
  57 <p>
  58 </p>
  59 <h1><a name="name">NAME</a></h1>
  60 <p>perlunicode - Unicode support in Perl</p>
  61 <p>
  62 </p>
  63 <hr />
  64 <h1><a name="description">DESCRIPTION</a></h1>
  65 <p>
  66 </p>
  67 <h2><a name="important_caveats">Important Caveats</a></h2>
  68 <p>Unicode support is an extensive requirement. While Perl does not
  69 implement the Unicode standard or the accompanying technical reports
  70 from cover to cover, Perl does support many Unicode features.</p>
  71 <dl>
  72 <dt><strong><a name="item_input_and_output_layers">Input and Output Layers</a></strong>
  73
  74 <dd>
  75 <p>Perl knows when a filehandle uses Perl's internal Unicode encodings
  76 (UTF-8, or UTF-EBCDIC if in EBCDIC) if the filehandle is opened with
  77 the ``:utf8'' layer.  Other encodings can be converted to Perl's
  78 encoding on input or from Perl's encoding on output by use of the
  79 ``:encoding(...)''  layer.  See <a href="file://C|\msysgit\mingw\html/lib/open.html">the open manpage</a>.</p>
  80 </dd>
  81 <dd>
  82 <p>To indicate that Perl source itself is using a particular encoding,
  83 see <a href="file://C|\msysgit\mingw\html/lib/encoding.html">the encoding manpage</a>.</p>
  84 </dd>
  85 </li>
  86 <dt><strong><a name="item_regular_expressions">Regular Expressions</a></strong>
  87
  88 <dd>
  89 <p>The regular expression compiler produces polymorphic opcodes.  That is,
  90 the pattern adapts to the data and automatically switches to the Unicode
  91 character scheme when presented with Unicode data--or instead uses
  92 a traditional byte scheme when presented with byte data.</p>
  93 </dd>
  94 </li>
  95 <dt><strong><a name="item_use_utf8_still_needed_to_enable_utf_2d8_2futf_2deb"><code>use utf8</code> still needed to enable UTF-8/UTF-EBCDIC in scripts</a></strong>
  96
  97 <dd>
  98 <p>As a compatibility measure, the <code>use utf8</code> pragma must be explicitly
  99 included to enable recognition of UTF-8 in the Perl scripts themselves
 100 (in string or regular expression literals, or in identifier names) on
 101 ASCII-based machines or to recognize UTF-EBCDIC on EBCDIC-based
 102 machines.  <strong>These are the only times when an explicit <code>use utf8</code>
 103 is needed.</strong>  See <a href="file://C|\msysgit\mingw\html/lib/utf8.html">the utf8 manpage</a>.</p>
 104 </dd>
 105 <dd>
 106 <p>You can also use the <code>encoding</code> pragma to change the default encoding
 107 of the data in your script; see <a href="file://C|\msysgit\mingw\html/lib/encoding.html">the encoding manpage</a>.</p>
 108 </dd>
 109 </li>
 110 <dt><strong><a name="item_bom_2dmarked_scripts_and_utf_2d16_scripts_autodete">BOM-marked scripts and UTF-16 scripts autodetected</a></strong>
 111
 112 <dd>
 113 <p>If a Perl script begins marked with the Unicode BOM (UTF-16LE, UTF16-BE,
 114 or UTF-8), or if the script looks like non-BOM-marked UTF-16 of either
 115 endianness, Perl will correctly read in the script as Unicode.
 116 (BOMless UTF-8 cannot be effectively recognized or differentiated from
 117 ISO 8859-1 or other eight-bit encodings.)</p>
 118 </dd>
 119 </li>
 120 <dt><strong><a name="item_use_encoding_needed_to_upgrade_non_2dlatin_2d1_byt"><code>use encoding</code> needed to upgrade non-Latin-1 byte strings</a></strong>
 121
 122 <dd>
 123 <p>By default, there is a fundamental asymmetry in Perl's unicode model:
 124 implicit upgrading from byte strings to Unicode strings assumes that
 125 they were encoded in <em>ISO 8859-1 (Latin-1)</em>, but Unicode strings are
 126 downgraded with UTF-8 encoding.  This happens because the first 256
 127 codepoints in Unicode happens to agree with Latin-1.</p>
 128 </dd>
 129 <dd>
 130 <p>If you wish to interpret byte strings as UTF-8 instead, use the
 131 <code>encoding</code> pragma:</p>
 132 </dd>
 133 <dd>
 134 <pre>
 135     use encoding 'utf8';</pre>
 136 </dd>
 137 <dd>
 138 <p>See <a href="#byte_and_character_semantics">Byte and Character Semantics</a> for more details.</p>
 139 </dd>
 140 </li>
 141 </dl>
 142 <p>
 143 </p>
 144 <h2><a name="byte_and_character_semantics">Byte and Character Semantics</a></h2>
 145 <p>Beginning with version 5.6, Perl uses logically-wide characters to
 146 represent strings internally.</p>
 147 <p>In future, Perl-level operations will be expected to work with
 148 characters rather than bytes.</p>
 149 <p>However, as an interim compatibility measure, Perl aims to
 150 provide a safe migration path from byte semantics to character
 151 semantics for programs.  For operations where Perl can unambiguously
 152 decide that the input data are characters, Perl switches to
 153 character semantics.  For operations where this determination cannot
 154 be made without additional information from the user, Perl decides in
 155 favor of compatibility and chooses to use byte semantics.</p>
 156 <p>This behavior preserves compatibility with earlier versions of Perl,
 157 which allowed byte semantics in Perl operations only if
 158 none of the program's inputs were marked as being as source of Unicode
 159 character data.  Such data may come from filehandles, from calls to
 160 external programs, from information provided by the system (such as %ENV),
 161 or from literals and constants in the source text.</p>
 162 <p>The <code>bytes</code> pragma will always, regardless of platform, force byte
 163 semantics in a particular lexical scope.  See <a href="file://C|\msysgit\mingw\html/lib/bytes.html">the bytes manpage</a>.</p>
 164 <p>The <code>utf8</code> pragma is primarily a compatibility device that enables
 165 recognition of UTF-(8|EBCDIC) in literals encountered by the parser.
 166 Note that this pragma is only required while Perl defaults to byte
 167 semantics; when character semantics become the default, this pragma
 168 may become a no-op.  See <a href="file://C|\msysgit\mingw\html/lib/utf8.html">the utf8 manpage</a>.</p>
 169 <p>Unless explicitly stated, Perl operators use character semantics
 170 for Unicode data and byte semantics for non-Unicode data.
 171 The decision to use character semantics is made transparently.  If
 172 input data comes from a Unicode source--for example, if a character
 173 encoding layer is added to a filehandle or a literal Unicode
 174 string constant appears in a program--character semantics apply.
 175 Otherwise, byte semantics are in effect.  The <code>bytes</code> pragma should
 176 be used to force byte semantics on Unicode data.</p>
 177 <p>If strings operating under byte semantics and strings with Unicode
 178 character data are concatenated, the new string will be created by
 179 decoding the byte strings as <em>ISO 8859-1 (Latin-1)</em>, even if the
 180 old Unicode string used EBCDIC.  This translation is done without
 181 regard to the system's native 8-bit encoding.  To change this for
 182 systems with non-Latin-1 and non-EBCDIC native encodings, use the
 183 <code>encoding</code> pragma.  See <a href="file://C|\msysgit\mingw\html/lib/encoding.html">the encoding manpage</a>.</p>
 184 <p>Under character semantics, many operations that formerly operated on
 185 bytes now operate on characters. A character in Perl is
 186 logically just a number ranging from 0 to 2**31 or so. Larger
 187 characters may encode into longer sequences of bytes internally, but
 188 this internal detail is mostly hidden for Perl code.
 189 See <a href="file://C|\msysgit\mingw\html/pod/perluniintro.html">the perluniintro manpage</a> for more.</p>
 190 <p>
 191 </p>
 192 <h2><a name="effects_of_character_semantics">Effects of Character Semantics</a></h2>
 193 <p>Character semantics have the following effects:</p>
 194 <ul>
 195 <li>
 196 <p>Strings--including hash keys--and regular expression patterns may
 197 contain characters that have an ordinal value larger than 255.</p>
 198 <p>If you use a Unicode editor to edit your program, Unicode characters
 199 may occur directly within the literal strings in one of the various
 200 Unicode encodings (UTF-8, UTF-EBCDIC, UCS-2, etc.), but will be recognized
 201 as such and converted to Perl's internal representation only if the
 202 appropriate <a href="file://C|\msysgit\mingw\html/lib/encoding.html">the encoding manpage</a> is specified.</p>
 203 <p>Unicode characters can also be added to a string by using the
 204 <code>\x{...}</code> notation.  The Unicode code for the desired character, in
 205 hexadecimal, should be placed in the braces. For instance, a smiley
 206 face is <code>\x{263A}</code>.  This encoding scheme only works for characters
 207 with a code of 0x100 or above.</p>
 208 <p>Additionally, if you</p>
 209 <pre>
 210    use charnames ':full';</pre>
 211 <p>you can use the <code>\N{...}</code> notation and put the official Unicode
 212 character name within the braces, such as <code>\N{WHITE SMILING FACE}</code>.</p>
 213 </li>
 214 <li>
 215 <p>If an appropriate <a href="file://C|\msysgit\mingw\html/lib/encoding.html">the encoding manpage</a> is specified, identifiers within the
 216 Perl script may contain Unicode alphanumeric characters, including
 217 ideographs.  Perl does not currently attempt to canonicalize variable
 218 names.</p>
 219 </li>
 220 <li>
 221 <p>Regular expressions match characters instead of bytes.  ``.'' matches
 222 a character instead of a byte.  The <code>\C</code> pattern is provided to force
 223 a match a single byte--a <code>char</code> in C, hence <code>\C</code>.</p>
 224 </li>
 225 <li>
 226 <p>Character classes in regular expressions match characters instead of
 227 bytes and match against the character properties specified in the
 228 Unicode properties database.  <code>\w</code> can be used to match a Japanese
 229 ideograph, for instance.</p>
 230 <p>(However, and as a limitation of the current implementation, using
 231 <code>\w</code> or <code>\W</code> <em>inside</em> a <code>[...]</code> character class will still match
 232 with byte semantics.)</p>
 233 </li>
 234 <li>
 235 <p>Named Unicode properties, scripts, and block ranges may be used like
 236 character classes via the <code>\p{}</code> ``matches property'' construct and
 237 the  <code>\P{}</code> negation, ``doesn't match property''.</p>
 238 <p>For instance, <code>\p{Lu}</code> matches any character with the Unicode ``Lu''
 239 (Letter, uppercase) property, while <code>\p{M}</code> matches any character
 240 with an ``M'' (mark--accents and such) property.  Brackets are not
 241 required for single letter properties, so <code>\p{M}</code> is equivalent to
 242 <code>\pM</code>. Many predefined properties are available, such as
 243 <code>\p{Mirrored}</code> and <code>\p{Tibetan}</code>.</p>
 244 <p>The official Unicode script and block names have spaces and dashes as
 245 separators, but for convenience you can use dashes, spaces, or
 246 underbars, and case is unimportant. It is recommended, however, that
 247 for consistency you use the following naming: the official Unicode
 248 script, property, or block name (see below for the additional rules
 249 that apply to block names) with whitespace and dashes removed, and the
 250 words ``uppercase-first-lowercase-rest''. <code>Latin-1 Supplement</code> thus
 251 becomes <code>Latin1Supplement</code>.</p>
 252 <p>You can also use negation in both <code>\p{}</code> and <code>\P{}</code> by introducing a caret
 253 (^) between the first brace and the property name: <code>\p{^Tamil}</code> is
 254 equal to <code>\P{Tamil}</code>.</p>
 255 <p><strong>NOTE: the properties, scripts, and blocks listed here are as of
 256 Unicode 3.2.0, March 2002, or Perl 5.8.0, July 2002.  Unicode 4.0.0
 257 came out in April 2003, and Perl 5.8.1 in September 2003.</strong></p>
 258 <p>Here are the basic Unicode General Category properties, followed by their
 259 long form.  You can use either; <code>\p{Lu}</code> and <code>\p{UppercaseLetter}</code>,
 260 for instance, are identical.</p>
 261 <pre>
 262     Short       Long</pre>
 263 <pre>
 264     L           Letter
 265     LC          CasedLetter
 266     Lu          UppercaseLetter
 267     Ll          LowercaseLetter
 268     Lt          TitlecaseLetter
 269     Lm          ModifierLetter
 270     Lo          OtherLetter</pre>
 271 <pre>
 272     M           Mark
 273     Mn          NonspacingMark
 274     Mc          SpacingMark
 275     Me          EnclosingMark</pre>
 276 <pre>
 277     N           Number
 278     Nd          DecimalNumber
 279     Nl          LetterNumber
 280     No          OtherNumber</pre>
 281 <pre>
 282     P           Punctuation
 283     Pc          ConnectorPunctuation
 284     Pd          DashPunctuation
 285     Ps          OpenPunctuation
 286     Pe          ClosePunctuation
 287     Pi          InitialPunctuation
 288                 (may behave like Ps or Pe depending on usage)
 289     Pf          FinalPunctuation
 290                 (may behave like Ps or Pe depending on usage)
 291     Po          OtherPunctuation</pre>
 292 <pre>
 293     S           Symbol
 294     Sm          MathSymbol
 295     Sc          CurrencySymbol
 296     Sk          ModifierSymbol
 297     So          OtherSymbol</pre>
 298 <pre>
 299     Z           Separator
 300     Zs          SpaceSeparator
 301     Zl          LineSeparator
 302     Zp          ParagraphSeparator</pre>
 303 <pre>
 304     C           Other
 305     Cc          Control
 306     Cf          Format
 307     Cs          Surrogate   (not usable)
 308     Co          PrivateUse
 309     Cn          Unassigned</pre>
 310 <p>Single-letter properties match all characters in any of the
 311 two-letter sub-properties starting with the same letter.
 312 <code>LC</code> and <code>L&amp;</code> are special cases, which are aliases for the set of
 313 <code>Ll</code>, <code>Lu</code>, and <code>Lt</code>.</p>
 314 <p>Because Perl hides the need for the user to understand the internal
 315 representation of Unicode characters, there is no need to implement
 316 the somewhat messy concept of surrogates. <code>Cs</code> is therefore not
 317 supported.</p>
 318 <p>Because scripts differ in their directionality--Hebrew is
 319 written right to left, for example--Unicode supplies these properties in
 320 the BidiClass class:</p>
 321 <pre>
 322     Property    Meaning</pre>
 323 <pre>
 324     L           Left-to-Right
 325     LRE         Left-to-Right Embedding
 326     LRO         Left-to-Right Override
 327     R           Right-to-Left
 328     AL          Right-to-Left Arabic
 329     RLE         Right-to-Left Embedding
 330     RLO         Right-to-Left Override
 331     PDF         Pop Directional Format
 332     EN          European Number
 333     ES          European Number Separator
 334     ET          European Number Terminator
 335     AN          Arabic Number
 336     CS          Common Number Separator
 337     NSM         Non-Spacing Mark
 338     BN          Boundary Neutral
 339     B           Paragraph Separator
 340     S           Segment Separator
 341     WS          Whitespace
 342     ON          Other Neutrals</pre>
 343 <p>For example, <code>\p{BidiClass:R}</code> matches characters that are normally
 344 written right to left.</p>
 345 </li>
 346 </ul>
 347 <p>
 348 </p>
 349 <h2><a name="scripts">Scripts</a></h2>
 350 <p>The script names which can be used by <code>\p{...}</code> and <code>\P{...}</code>,
 351 such as in <code>\p{Latin}</code> or <code>\p{Cyrillic}</code>, are as follows:</p>
 352 <pre>
 353     Arabic
 354     Armenian
 355     Bengali
 356     Bopomofo
 357     Buhid
 358     CanadianAboriginal
 359     Cherokee
 360     Cyrillic
 361     Deseret
 362     Devanagari
 363     Ethiopic
 364     Georgian
 365     Gothic
 366     Greek
 367     Gujarati
 368     Gurmukhi
 369     Han
 370     Hangul
 371     Hanunoo
 372     Hebrew
 373     Hiragana
 374     Inherited
 375     Kannada
 376     Katakana
 377     Khmer
 378     Lao
 379     Latin
 380     Malayalam
 381     Mongolian
 382     Myanmar
 383     Ogham
 384     OldItalic
 385     Oriya
 386     Runic
 387     Sinhala
 388     Syriac
 389     Tagalog
 390     Tagbanwa
 391     Tamil
 392     Telugu
 393     Thaana
 394     Thai
 395     Tibetan
 396     Yi</pre>
 397 <p>Extended property classes can supplement the basic
 398 properties, defined by the <em>PropList</em> Unicode database:</p>
 399 <pre>
 400     ASCIIHexDigit
 401     BidiControl
 402     Dash
 403     Deprecated
 404     Diacritic
 405     Extender
 406     GraphemeLink
 407     HexDigit
 408     Hyphen
 409     Ideographic
 410     IDSBinaryOperator
 411     IDSTrinaryOperator
 412     JoinControl
 413     LogicalOrderException
 414     NoncharacterCodePoint
 415     OtherAlphabetic
 416     OtherDefaultIgnorableCodePoint
 417     OtherGraphemeExtend
 418     OtherLowercase
 419     OtherMath
 420     OtherUppercase
 421     QuotationMark
 422     Radical
 423     SoftDotted
 424     TerminalPunctuation
 425     UnifiedIdeograph
 426     WhiteSpace</pre>
 427 <p>and there are further derived properties:</p>
 428 <pre>
 429     Alphabetic      Lu + Ll + Lt + Lm + Lo + OtherAlphabetic
 430     Lowercase       Ll + OtherLowercase
 431     Uppercase       Lu + OtherUppercase
 432     Math            Sm + OtherMath</pre>
 433 <pre>
 434     ID_Start        Lu + Ll + Lt + Lm + Lo + Nl
 435     ID_Continue     ID_Start + Mn + Mc + Nd + Pc</pre>
 436 <pre>
 437     Any             Any character
 438     Assigned        Any non-Cn character (i.e. synonym for \P{Cn})
 439     Unassigned      Synonym for \p{Cn}
 440     Common          Any character (or unassigned code point)
 441                     not explicitly assigned to a script</pre>
 442 <p>For backward compatibility (with Perl 5.6), all properties mentioned
 443 so far may have <code>Is</code> prepended to their name, so <code>\P{IsLu}</code>, for
 444 example, is equal to <code>\P{Lu}</code>.</p>
 445 <p>
 446 </p>
 447 <h2><a name="blocks">Blocks</a></h2>
 448 <p>In addition to <strong>scripts</strong>, Unicode also defines <strong>blocks</strong> of
 449 characters.  The difference between scripts and blocks is that the
 450 concept of scripts is closer to natural languages, while the concept
 451 of blocks is more of an artificial grouping based on groups of 256
 452 Unicode characters. For example, the <code>Latin</code> script contains letters
 453 from many blocks but does not contain all the characters from those
 454 blocks. It does not, for example, contain digits, because digits are
 455 shared across many scripts. Digits and similar groups, like
 456 punctuation, are in a category called <code>Common</code>.</p>
 457 <p>For more about scripts, see the UTR #24:</p>
 458 <pre>
 459    <a href="http://www.unicode.org/unicode/reports/tr24/">http://www.unicode.org/unicode/reports/tr24/</a></pre>
 460 <p>For more about blocks, see:</p>
 461 <pre>
 462    <a href="http://www.unicode.org/Public/UNIDATA/Blocks.txt">http://www.unicode.org/Public/UNIDATA/Blocks.txt</a></pre>
 463 <p>Block names are given with the <code>In</code> prefix. For example, the
 464 Katakana block is referenced via <code>\p{InKatakana}</code>.  The <code>In</code>
 465 prefix may be omitted if there is no naming conflict with a script
 466 or any other property, but it is recommended that <code>In</code> always be used
 467 for block tests to avoid confusion.</p>
 468 <p>These block names are supported:</p>
 469 <pre>
 470     InAlphabeticPresentationForms
 471     InArabic
 472     InArabicPresentationFormsA
 473     InArabicPresentationFormsB
 474     InArmenian
 475     InArrows
 476     InBasicLatin
 477     InBengali
 478     InBlockElements
 479     InBopomofo
 480     InBopomofoExtended
 481     InBoxDrawing
 482     InBraillePatterns
 483     InBuhid
 484     InByzantineMusicalSymbols
 485     InCJKCompatibility
 486     InCJKCompatibilityForms
 487     InCJKCompatibilityIdeographs
 488     InCJKCompatibilityIdeographsSupplement
 489     InCJKRadicalsSupplement
 490     InCJKSymbolsAndPunctuation
 491     InCJKUnifiedIdeographs
 492     InCJKUnifiedIdeographsExtensionA
 493     InCJKUnifiedIdeographsExtensionB
 494     InCherokee
 495     InCombiningDiacriticalMarks
 496     InCombiningDiacriticalMarksforSymbols
 497     InCombiningHalfMarks
 498     InControlPictures
 499     InCurrencySymbols
 500     InCyrillic
 501     InCyrillicSupplementary
 502     InDeseret
 503     InDevanagari
 504     InDingbats
 505     InEnclosedAlphanumerics
 506     InEnclosedCJKLettersAndMonths
 507     InEthiopic
 508     InGeneralPunctuation
 509     InGeometricShapes
 510     InGeorgian
 511     InGothic
 512     InGreekExtended
 513     InGreekAndCoptic
 514     InGujarati
 515     InGurmukhi
 516     InHalfwidthAndFullwidthForms
 517     InHangulCompatibilityJamo
 518     InHangulJamo
 519     InHangulSyllables
 520     InHanunoo
 521     InHebrew
 522     InHighPrivateUseSurrogates
 523     InHighSurrogates
 524     InHiragana
 525     InIPAExtensions
 526     InIdeographicDescriptionCharacters
 527     InKanbun
 528     InKangxiRadicals
 529     InKannada
 530     InKatakana
 531     InKatakanaPhoneticExtensions
 532     InKhmer
 533     InLao
 534     InLatin1Supplement
 535     InLatinExtendedA
 536     InLatinExtendedAdditional
 537     InLatinExtendedB
 538     InLetterlikeSymbols
 539     InLowSurrogates
 540     InMalayalam
 541     InMathematicalAlphanumericSymbols
 542     InMathematicalOperators
 543     InMiscellaneousMathematicalSymbolsA
 544     InMiscellaneousMathematicalSymbolsB
 545     InMiscellaneousSymbols
 546     InMiscellaneousTechnical
 547     InMongolian
 548     InMusicalSymbols
 549     InMyanmar
 550     InNumberForms
 551     InOgham
 552     InOldItalic
 553     InOpticalCharacterRecognition
 554     InOriya
 555     InPrivateUseArea
 556     InRunic
 557     InSinhala
 558     InSmallFormVariants
 559     InSpacingModifierLetters
 560     InSpecials
 561     InSuperscriptsAndSubscripts
 562     InSupplementalArrowsA
 563     InSupplementalArrowsB
 564     InSupplementalMathematicalOperators
 565     InSupplementaryPrivateUseAreaA
 566     InSupplementaryPrivateUseAreaB
 567     InSyriac
 568     InTagalog
 569     InTagbanwa
 570     InTags
 571     InTamil
 572     InTelugu
 573     InThaana
 574     InThai
 575     InTibetan
 576     InUnifiedCanadianAboriginalSyllabics
 577     InVariationSelectors
 578     InYiRadicals
 579     InYiSyllables</pre>
 580 <ul>
 581 <li>
 582 <p>The special pattern <code>\X</code> matches any extended Unicode
 583 sequence--``a combining character sequence'' in Standardese--where the
 584 first character is a base character and subsequent characters are mark
 585 characters that apply to the base character.  <code>\X</code> is equivalent to
 586 <code>(?:\PM\pM*)</code>.</p>
 587 </li>
 588 <li>
 589 <p>The <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_tr_"><code>tr///</code></a> operator translates characters instead of bytes.  Note
 590 that the <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_tr_"><code>tr///CU</code></a> functionality has been removed.  For similar
 591 functionality see pack('U0', ...) and pack('C0', ...).</p>
 592 </li>
 593 <li>
 594 <p>Case translation operators use the Unicode case translation tables
 595 when character input is provided.  Note that <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_uc"><code>uc()</code></a>, or <code>\U</code> in
 596 interpolated strings, translates to uppercase, while <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ucfirst"><code>ucfirst</code></a>,
 597 or <code>\u</code> in interpolated strings, translates to titlecase in languages
 598 that make the distinction.</p>
 599 </li>
 600 <li>
 601 <p>Most operators that deal with positions or lengths in a string will
 602 automatically switch to using character positions, including
 603 <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_chop"><code>chop()</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_chomp"><code>chomp()</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item_substr"><code>substr()</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_pos"><code>pos()</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_index"><code>index()</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_rindex"><code>rindex()</code></a>,
 604 <code>sprintf()</code>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_write"><code>write()</code></a>, and <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_length"><code>length()</code></a>.  Operators that
 605 specifically do not switch include <code>vec()</code>, <code>pack()</code>, and
 606 <code>unpack()</code>.  Operators that really don't care include
 607 operators that treats strings as a bucket of bits such as <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_sort"><code>sort()</code></a>,
 608 and operators dealing with filenames.</p>
 609 </li>
 610 <li>
 611 <p>The <code>pack()</code>/<code>unpack()</code> letters <code>c</code> and <code>C</code> do <em>not</em> change,
 612 since they are often used for byte-oriented formats.  Again, think
 613 <code>char</code> in the C language.</p>
 614 <p>There is a new <code>U</code> specifier that converts between Unicode characters
 615 and code points.</p>
 616 </li>
 617 <li>
 618 <p>The <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_chr"><code>chr()</code></a> and <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ord"><code>ord()</code></a> functions work on characters, similar to
 619 <code>pack(&quot;U&quot;)</code> and <code>unpack(&quot;U&quot;)</code>, <em>not</em> <code>pack(&quot;C&quot;)</code> and
 620 <code>unpack(&quot;C&quot;)</code>.  <code>pack(&quot;C&quot;)</code> and <code>unpack(&quot;C&quot;)</code> are methods for
 621 emulating byte-oriented <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_chr"><code>chr()</code></a> and <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ord"><code>ord()</code></a> on Unicode strings.
 622 While these methods reveal the internal encoding of Unicode strings,
 623 that is not something one normally needs to care about at all.</p>
 624 </li>
 625 <li>
 626 <p>The bit string operators, <code>&amp; | ^ ~</code>, can operate on character data.
 627 However, for backward compatibility, such as when using bit string
 628 operations when characters are all less than 256 in ordinal value, one
 629 should not use <code>~</code> (the bit complement) with characters of both
 630 values less than 256 and values greater than 256.  Most importantly,
 631 DeMorgan's laws (<code>~($x|$y) eq ~$x&amp;~$y</code> and <code>~($x&amp;$y) eq ~$x|~$y</code>)
 632 will not hold.  The reason for this mathematical <em>faux pas</em> is that
 633 the complement cannot return <strong>both</strong> the 8-bit (byte-wide) bit
 634 complement <strong>and</strong> the full character-wide bit complement.</p>
 635 </li>
 636 <li>
 637 <p>lc(), uc(), lcfirst(), and <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ucfirst"><code>ucfirst()</code></a> work for the following cases:</p>
 638 <ul>
 639 <li>
 640 <p>the case mapping is from a single Unicode character to another
 641 single Unicode character, or</p>
 642 </li>
 643 <li>
 644 <p>the case mapping is from a single Unicode character to more
 645 than one Unicode character.</p>
 646 </li>
 647 </ul>
 648 <p>Things to do with locales (Lithuanian, Turkish, Azeri) do <strong>not</strong> work
 649 since Perl does not understand the concept of Unicode locales.</p>
 650 <p>See the Unicode Technical Report #21, Case Mappings, for more details.</p>
 651 </ul>
 652 <ul>
 653 <li>
 654 <p>And finally, <code>scalar reverse()</code> reverses by character rather than by byte.</p>
 655 </li>
 656 </ul>
 657 <p>
 658 </p>
 659 <h2><a name="userdefined_character_properties">User-Defined Character Properties</a></h2>
 660 <p>You can define your own character properties by defining subroutines
 661 whose names begin with ``In'' or ``Is''.  The subroutines can be defined in
 662 any package.  The user-defined properties can be used in the regular
 663 expression <code>\p</code> and <code>\P</code> constructs; if you are using a user-defined
 664 property from a package other than the one you are in, you must specify
 665 its package in the <code>\p</code> or <code>\P</code> construct.</p>
 666 <pre>
 667     # assuming property IsForeign defined in Lang::
 668     package main;  # property package name required
 669     if ($txt =~ /\p{Lang::IsForeign}+/) { ... }</pre>
 670 <pre>
 671     package Lang;  # property package name not required
 672     if ($txt =~ /\p{IsForeign}+/) { ... }</pre>
 673 <p>Note that the effect is compile-time and immutable once defined.</p>
 674 <p>The subroutines must return a specially-formatted string, with one
 675 or more newline-separated lines.  Each line must be one of the following:</p>
 676 <ul>
 677 <li>
 678 <p>Two hexadecimal numbers separated by horizontal whitespace (space or
 679 tabular characters) denoting a range of Unicode code points to include.</p>
 680 </li>
 681 <li>
 682 <p>Something to include, prefixed by ``+'': a built-in character
 683 property (prefixed by ``utf8::'') or a user-defined character property,
 684 to represent all the characters in that property; two hexadecimal code
 685 points for a range; or a single hexadecimal code point.</p>
 686 </li>
 687 <li>
 688 <p>Something to exclude, prefixed by ``-'': an existing character
 689 property (prefixed by ``utf8::'') or a user-defined character property,
 690 to represent all the characters in that property; two hexadecimal code
 691 points for a range; or a single hexadecimal code point.</p>
 692 </li>
 693 <li>
 694 <p>Something to negate, prefixed ``!'': an existing character
 695 property (prefixed by ``utf8::'') or a user-defined character property,
 696 to represent all the characters in that property; two hexadecimal code
 697 points for a range; or a single hexadecimal code point.</p>
 698 </li>
 699 <li>
 700 <p>Something to intersect with, prefixed by ``&amp;'': an existing character
 701 property (prefixed by ``utf8::'') or a user-defined character property,
 702 for all the characters except the characters in the property; two
 703 hexadecimal code points for a range; or a single hexadecimal code point.</p>
 704 </li>
 705 </ul>
 706 <p>For example, to define a property that covers both the Japanese
 707 syllabaries (hiragana and katakana), you can define</p>
 708 <pre>
 709     sub InKana {
 710         return &lt;&lt;END;
 711     3040\t309F
 712     30A0\t30FF
 713     END
 714     }</pre>
 715 <p>Imagine that the here-doc end marker is at the beginning of the line.
 716 Now you can use <code>\p{InKana}</code> and <code>\P{InKana}</code>.</p>
 717 <p>You could also have used the existing block property names:</p>
 718 <pre>
 719     sub InKana {
 720         return &lt;&lt;'END';
 721     +utf8::InHiragana
 722     +utf8::InKatakana
 723     END
 724     }</pre>
 725 <p>Suppose you wanted to match only the allocated characters,
 726 not the raw block ranges: in other words, you want to remove
 727 the non-characters:</p>
 728 <pre>
 729     sub InKana {
 730         return &lt;&lt;'END';
 731     +utf8::InHiragana
 732     +utf8::InKatakana
 733     -utf8::IsCn
 734     END
 735     }</pre>
 736 <p>The negation is useful for defining (surprise!) negated classes.</p>
 737 <pre>
 738     sub InNotKana {
 739         return &lt;&lt;'END';
 740     !utf8::InHiragana
 741     -utf8::InKatakana
 742     +utf8::IsCn
 743     END
 744     }</pre>
 745 <p>Intersection is useful for getting the common characters matched by
 746 two (or more) classes.</p>
 747 <pre>
 748     sub InFooAndBar {
 749         return &lt;&lt;'END';
 750     +main::Foo
 751     &amp;main::Bar
 752     END
 753     }</pre>
 754 <p>It's important to remember not to use ``&amp;'' for the first set -- that
 755 would be intersecting with nothing (resulting in an empty set).</p>
 756 <p>You can also define your own mappings to be used in the lc(),
 757 lcfirst(), uc(), and <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ucfirst"><code>ucfirst()</code></a> (or their string-inlined versions).
 758 The principle is the same: define subroutines in the <code>main</code> package
 759 with names like <code>ToLower</code> (for <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_lc"><code>lc()</code></a> and lcfirst()), <code>ToTitle</code> (for
 760 the first character in ucfirst()), and <code>ToUpper</code> (for uc(), and the
 761 rest of the characters in ucfirst()).</p>
 762 <p>The string returned by the subroutines needs now to be three
 763 hexadecimal numbers separated by tabulators: start of the source
 764 range, end of the source range, and start of the destination range.
 765 For example:</p>
 766 <pre>
 767     sub ToUpper {
 768         return &lt;&lt;END;
 769     0061\t0063\t0041
 770     END
 771     }</pre>
 772 <p>defines an <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_uc"><code>uc()</code></a> mapping that causes only the characters ``a'', ``b'', and
 773 ``c'' to be mapped to ``A'', ``B'', ``C'', all other characters will remain
 774 unchanged.</p>
 775 <p>If there is no source range to speak of, that is, the mapping is from
 776 a single character to another single character, leave the end of the
 777 source range empty, but the two tabulator characters are still needed.
 778 For example:</p>
 779 <pre>
 780     sub ToLower {
 781         return &lt;&lt;END;
 782     0041\t\t0061
 783     END
 784     }</pre>
 785 <p>defines a <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_lc"><code>lc()</code></a> mapping that causes only ``A'' to be mapped to ``a'', all
 786 other characters will remain unchanged.</p>
 787 <p>(For serious hackers only)  If you want to introspect the default
 788 mappings, you can find the data in the directory
 789 <code>$Config{privlib}</code>/<em>unicore/To/</em>.  The mapping data is returned as
 790 the here-document, and the <code>utf8::ToSpecFoo</code> are special exception
 791 mappings derived from &lt;$Config{privlib}&gt;/<em>unicore/SpecialCasing.txt</em>.
 792 The <code>Digit</code> and <code>Fold</code> mappings that one can see in the directory
 793 are not directly user-accessible, one can use either the
 794 <code>Unicode::UCD</code> module, or just match case-insensitively (that's when
 795 the <code>Fold</code> mapping is used).</p>
 796 <p>A final note on the user-defined property tests and mappings: they
 797 will be used only if the scalar has been marked as having Unicode
 798 characters.  Old byte-style strings will not be affected.</p>
 799 <p>
 800 </p>
 801 <h2><a name="character_encodings_for_input_and_output">Character Encodings for Input and Output</a></h2>
 802 <p>See <a href="file://C|\msysgit\mingw\html/lib/Encode.html">the Encode manpage</a>.</p>
 803 <p>
 804 </p>
 805 <h2><a name="unicode_regular_expression_support_level">Unicode Regular Expression Support Level</a></h2>
 806 <p>The following list of Unicode support for regular expressions describes
 807 all the features currently supported.  The references to ``Level N''
 808 and the section numbers refer to the Unicode Technical Report 18,
 809 ``Unicode Regular Expression Guidelines'', version 6 (Unicode 3.2.0,
 810 Perl 5.8.0).</p>
 811 <ul>
 812 <li>
 813 <p>Level 1 - Basic Unicode Support</p>
 814 <pre>
 815         2.1 Hex Notation                        - done          [1]
 816             Named Notation                      - done          [2]
 817         2.2 Categories                          - done          [3][4]
 818         2.3 Subtraction                         - MISSING       [5][6]
 819         2.4 Simple Word Boundaries              - done          [7]
 820         2.5 Simple Loose Matches                - done          [8]
 821         2.6 End of Line                         - MISSING       [9][10]</pre>
 822 <pre>
 823         [ 1] \x{...}
 824         [ 2] \N{...}
 825         [ 3] . \p{...} \P{...}
 826         [ 4] support for scripts (see UTR#24 Script Names), blocks,
 827              binary properties, enumerated non-binary properties, and
 828              numeric properties (as listed in UTR#18 Other Properties)
 829         [ 5] have negation
 830         [ 6] can use regular expression look-ahead [a]
 831              or user-defined character properties [b] to emulate subtraction
 832         [ 7] include Letters in word characters
 833         [ 8] note that Perl does Full case-folding in matching, not Simple:
 834              for example U+1F88 is equivalent with U+1F00 U+03B9,
 835              not with 1F80.  This difference matters for certain Greek
 836              capital letters with certain modifiers: the Full case-folding
 837              decomposes the letter, while the Simple case-folding would map
 838              it to a single character.
 839         [ 9] see UTR #13 Unicode Newline Guidelines
 840         [10] should do ^ and $ also on \x{85}, \x{2028} and \x{2029}
 841              (should also affect &lt;&gt;, $., and script line numbers)
 842              (the \x{85}, \x{2028} and \x{2029} do match \s)</pre>
 843 <p>[a] You can mimic class subtraction using lookahead.
 844 For example, what UTR #18 might write as</p>
 845 <pre>
 846     [{Greek}-[{UNASSIGNED}]]</pre>
 847 <p>in Perl can be written as:</p>
 848 <pre>
 849     (?!\p{Unassigned})\p{InGreekAndCoptic}
 850     (?=\p{Assigned})\p{InGreekAndCoptic}</pre>
 851 <p>But in this particular example, you probably really want</p>
 852 <pre>
 853     \p{GreekAndCoptic}</pre>
 854 <p>which will match assigned characters known to be part of the Greek script.</p>
 855 <p>Also see the Unicode::Regex::Set module, it does implement the full
 856 UTR #18 grouping, intersection, union, and removal (subtraction) syntax.</p>
 857 <p>[b] See <a href="#userdefined_character_properties">User-Defined Character Properties</a>.</p>
 858 </li>
 859 <li>
 860 <p>Level 2 - Extended Unicode Support</p>
 861 <pre>
 862         3.1 Surrogates                          - MISSING       [11]
 863         3.2 Canonical Equivalents               - MISSING       [12][13]
 864         3.3 Locale-Independent Graphemes        - MISSING       [14]
 865         3.4 Locale-Independent Words            - MISSING       [15]
 866         3.5 Locale-Independent Loose Matches    - MISSING       [16]</pre>
 867 <pre>
 868         [11] Surrogates are solely a UTF-16 concept and Perl's internal
 869              representation is UTF-8.  The Encode module does UTF-16, though.
 870         [12] see UTR#15 Unicode Normalization
 871         [13] have Unicode::Normalize but not integrated to regexes
 872         [14] have \X but at this level . should equal that
 873         [15] need three classes, not just \w and \W
 874         [16] see UTR#21 Case Mappings</pre>
 875 </li>
 876 <li>
 877 <p>Level 3 - Locale-Sensitive Support</p>
 878 <pre>
 879         4.1 Locale-Dependent Categories         - MISSING
 880         4.2 Locale-Dependent Graphemes          - MISSING       [16][17]
 881         4.3 Locale-Dependent Words              - MISSING
 882         4.4 Locale-Dependent Loose Matches      - MISSING
 883         4.5 Locale-Dependent Ranges             - MISSING</pre>
 884 <pre>
 885         [16] see UTR#10 Unicode Collation Algorithms
 886         [17] have Unicode::Collate but not integrated to regexes</pre>
 887 </li>
 888 </ul>
 889 <p>
 890 </p>
 891 <h2><a name="unicode_encodings">Unicode Encodings</a></h2>
 892 <p>Unicode characters are assigned to <em>code points</em>, which are abstract
 893 numbers.  To use these numbers, various encodings are needed.</p>
 894 <ul>
 895 <li>
 896 <p>UTF-8</p>
 897 <p>UTF-8 is a variable-length (1 to 6 bytes, current character allocations
 898 require 4 bytes), byte-order independent encoding. For ASCII (and we
 899 really do mean 7-bit ASCII, not another 8-bit encoding), UTF-8 is
 900 transparent.</p>
 901 <p>The following table is from Unicode 3.2.</p>
 902 <pre>
 903  Code Points            1st Byte  2nd Byte  3rd Byte  4th Byte</pre>
 904 <pre>
 905    U+0000..U+007F       00..7F
 906    U+0080..U+07FF       C2..DF    80..BF
 907    U+0800..U+0FFF       E0        A0..BF    80..BF
 908    U+1000..U+CFFF       E1..EC    80..BF    80..BF
 909    U+D000..U+D7FF       ED        80..9F    80..BF
 910    U+D800..U+DFFF       ******* ill-formed *******
 911    U+E000..U+FFFF       EE..EF    80..BF    80..BF
 912   U+10000..U+3FFFF      F0        90..BF    80..BF    80..BF
 913   U+40000..U+FFFFF      F1..F3    80..BF    80..BF    80..BF
 914  U+100000..U+10FFFF     F4        80..8F    80..BF    80..BF</pre>
 915 <p>Note the <code>A0..BF</code> in <code>U+0800..U+0FFF</code>, the <code>80..9F</code> in
 916 <code>U+D000...U+D7FF</code>, the <code>90..B</code>F in <code>U+10000..U+3FFFF</code>, and the
 917 <code>80...8F</code> in <code>U+100000..U+10FFFF</code>.  The ``gaps'' are caused by legal
 918 UTF-8 avoiding non-shortest encodings: it is technically possible to
 919 UTF-8-encode a single code point in different ways, but that is
 920 explicitly forbidden, and the shortest possible encoding should always
 921 be used.  So that's what Perl does.</p>
 922 <p>Another way to look at it is via bits:</p>
 923 <pre>
 924  Code Points                    1st Byte   2nd Byte  3rd Byte  4th Byte</pre>
 925 <pre>
 926                     0aaaaaaa     0aaaaaaa
 927             00000bbbbbaaaaaa     110bbbbb  10aaaaaa
 928             ccccbbbbbbaaaaaa     1110cccc  10bbbbbb  10aaaaaa
 929   00000dddccccccbbbbbbaaaaaa     11110ddd  10cccccc  10bbbbbb  10aaaaaa</pre>
 930 <p>As you can see, the continuation bytes all begin with <code>10</code>, and the
 931 leading bits of the start byte tell how many bytes the are in the
 932 encoded character.</p>
 933 </li>
 934 <li>
 935 <p>UTF-EBCDIC</p>
 936 <p>Like UTF-8 but EBCDIC-safe, in the way that UTF-8 is ASCII-safe.</p>
 937 </li>
 938 <li>
 939 <p>UTF-16, UTF-16BE, UTF-16LE, Surrogates, and BOMs (Byte Order Marks)</p>
 940 <p>The followings items are mostly for reference and general Unicode
 941 knowledge, Perl doesn't use these constructs internally.</p>
 942 <p>UTF-16 is a 2 or 4 byte encoding.  The Unicode code points
 943 <code>U+0000..U+FFFF</code> are stored in a single 16-bit unit, and the code
 944 points <code>U+10000..U+10FFFF</code> in two 16-bit units.  The latter case is
 945 using <em>surrogates</em>, the first 16-bit unit being the <em>high
 946 surrogate</em>, and the second being the <em>low surrogate</em>.</p>
 947 <p>Surrogates are code points set aside to encode the <code>U+10000..U+10FFFF</code>
 948 range of Unicode code points in pairs of 16-bit units.  The <em>high
 949 surrogates</em> are the range <code>U+D800..U+DBFF</code>, and the <em>low surrogates</em>
 950 are the range <code>U+DC00..U+DFFF</code>.  The surrogate encoding is</p>
 951 <pre>
 952         $hi = ($uni - 0x10000) / 0x400 + 0xD800;
 953         $lo = ($uni - 0x10000) % 0x400 + 0xDC00;</pre>
 954 <p>and the decoding is</p>
 955 <pre>
 956         $uni = 0x10000 + ($hi - 0xD800) * 0x400 + ($lo - 0xDC00);</pre>
 957 <p>If you try to generate surrogates (for example by using chr()), you
 958 will get a warning if warnings are turned on, because those code
 959 points are not valid for a Unicode character.</p>
 960 <p>Because of the 16-bitness, UTF-16 is byte-order dependent.  UTF-16
 961 itself can be used for in-memory computations, but if storage or
 962 transfer is required either UTF-16BE (big-endian) or UTF-16LE
 963 (little-endian) encodings must be chosen.</p>
 964 <p>This introduces another problem: what if you just know that your data
 965 is UTF-16, but you don't know which endianness?  Byte Order Marks, or
 966 BOMs, are a solution to this.  A special character has been reserved
 967 in Unicode to function as a byte order marker: the character with the
 968 code point <code>U+FEFF</code> is the BOM.</p>
 969 <p>The trick is that if you read a BOM, you will know the byte order,
 970 since if it was written on a big-endian platform, you will read the
 971 bytes <code>0xFE 0xFF</code>, but if it was written on a little-endian platform,
 972 you will read the bytes <code>0xFF 0xFE</code>.  (And if the originating platform
 973 was writing in UTF-8, you will read the bytes <code>0xEF 0xBB 0xBF</code>.)</p>
 974 <p>The way this trick works is that the character with the code point
 975 <code>U+FFFE</code> is guaranteed not to be a valid Unicode character, so the
 976 sequence of bytes <code>0xFF 0xFE</code> is unambiguously ``BOM, represented in
 977 little-endian format'' and cannot be <code>U+FFFE</code>, represented in big-endian
 978 format``.</p>
 979 </li>
 980 <li>
 981 <p>UTF-32, UTF-32BE, UTF-32LE</p>
 982 <p>The UTF-32 family is pretty much like the UTF-16 family, expect that
 983 the units are 32-bit, and therefore the surrogate scheme is not
 984 needed.  The BOM signatures will be <code>0x00 0x00 0xFE 0xFF</code> for BE and
 985 <code>0xFF 0xFE 0x00 0x00</code> for LE.</p>
 986 </li>
 987 <li>
 988 <p>UCS-2, UCS-4</p>
 989 <p>Encodings defined by the ISO 10646 standard.  UCS-2 is a 16-bit
 990 encoding.  Unlike UTF-16, UCS-2 is not extensible beyond <code>U+FFFF</code>,
 991 because it does not use surrogates.  UCS-4 is a 32-bit encoding,
 992 functionally identical to UTF-32.</p>
 993 </li>
 994 <li>
 995 <p>UTF-7</p>
 996 <p>A seven-bit safe (non-eight-bit) encoding, which is useful if the
 997 transport or storage is not eight-bit safe.  Defined by RFC 2152.</p>
 998 </li>
 999 </ul>
1000 <p>
1001 </p>
1002 <h2><a name="security_implications_of_unicode">Security Implications of Unicode</a></h2>
1003 <ul>
1004 <li>
1005 <p>Malformed UTF-8</p>
1006 <p>Unfortunately, the specification of UTF-8 leaves some room for
1007 interpretation of how many bytes of encoded output one should generate
1008 from one input Unicode character.  Strictly speaking, the shortest
1009 possible sequence of UTF-8 bytes should be generated,
1010 because otherwise there is potential for an input buffer overflow at
1011 the receiving end of a UTF-8 connection.  Perl always generates the
1012 shortest length UTF-8, and with warnings on Perl will warn about
1013 non-shortest length UTF-8 along with other malformations, such as the
1014 surrogates, which are not real Unicode code points.</p>
1015 </li>
1016 <li>
1017 <p>Regular expressions behave slightly differently between byte data and
1018 character (Unicode) data.  For example, the ``word character'' character
1019 class <code>\w</code> will work differently depending on if data is eight-bit bytes
1020 or Unicode.</p>
1021 <p>In the first case, the set of <code>\w</code> characters is either small--the
1022 default set of alphabetic characters, digits, and the ``_''--or, if you
1023 are using a locale (see <a href="file://C|\msysgit\mingw\html/pod/perllocale.html">the perllocale manpage</a>), the <code>\w</code> might contain a few
1024 more letters according to your language and country.</p>
1025 <p>In the second case, the <code>\w</code> set of characters is much, much larger.
1026 Most importantly, even in the set of the first 256 characters, it will
1027 probably match different characters: unlike most locales, which are
1028 specific to a language and country pair, Unicode classifies all the
1029 characters that are letters <em>somewhere</em> as <code>\w</code>.  For example, your
1030 locale might not think that LATIN SMALL LETTER ETH is a letter (unless
1031 you happen to speak Icelandic), but Unicode does.</p>
1032 <p>As discussed elsewhere, Perl has one foot (two hooves?) planted in
1033 each of two worlds: the old world of bytes and the new world of
1034 characters, upgrading from bytes to characters when necessary.
1035 If your legacy code does not explicitly use Unicode, no automatic
1036 switch-over to characters should happen.  Characters shouldn't get
1037 downgraded to bytes, either.  It is possible to accidentally mix bytes
1038 and characters, however (see <a href="file://C|\msysgit\mingw\html/pod/perluniintro.html">the perluniintro manpage</a>), in which case <code>\w</code> in
1039 regular expressions might start behaving differently.  Review your
1040 code.  Use warnings and the <code>strict</code> pragma.</p>
1041 </li>
1042 </ul>
1043 <p>
1044 </p>
1045 <h2><a name="unicode_in_perl_on_ebcdic">Unicode in Perl on EBCDIC</a></h2>
1046 <p>The way Unicode is handled on EBCDIC platforms is still
1047 experimental.  On such platforms, references to UTF-8 encoding in this
1048 document and elsewhere should be read as meaning the UTF-EBCDIC
1049 specified in Unicode Technical Report 16, unless ASCII vs. EBCDIC issues
1050 are specifically discussed. There is no <code>utfebcdic</code> pragma or
1051 ``:utfebcdic'' layer; rather, ``utf8'' and ``:utf8'' are reused to mean
1052 the platform's ``natural'' 8-bit encoding of Unicode. See <a href="file://C|\msysgit\mingw\html/pod/perlebcdic.html">the perlebcdic manpage</a>
1053 for more discussion of the issues.</p>
1054 <p>
1055 </p>
1056 <h2><a name="locales">Locales</a></h2>
1057 <p>Usually locale settings and Unicode do not affect each other, but
1058 there are a couple of exceptions:</p>
1059 <ul>
1060 <li>
1061 <p>You can enable automatic UTF-8-ification of your standard file
1062 handles, default <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_open"><code>open()</code></a> layer, and <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item__argv"><code>@ARGV</code></a> by using either
1063 the <code>-C</code> command line switch or the <code>PERL_UNICODE</code> environment
1064 variable, see <a href="file://C|\msysgit\mingw\html/pod/perlrun.html">the perlrun manpage</a> for the documentation of the <code>-C</code> switch.</p>
1065 </li>
1066 <li>
1067 <p>Perl tries really hard to work both with Unicode and the old
1068 byte-oriented world. Most often this is nice, but sometimes Perl's
1069 straddling of the proverbial fence causes problems.</p>
1070 </li>
1071 </ul>
1072 <p>
1073 </p>
1074 <h2><a name="when_unicode_does_not_happen">When Unicode Does Not Happen</a></h2>
1075 <p>While Perl does have extensive ways to input and output in Unicode,
1076 and few other 'entry points' like the @ARGV which can be interpreted
1077 as Unicode (UTF-8), there still are many places where Unicode (in some
1078 encoding or another) could be given as arguments or received as
1079 results, or both, but it is not.</p>
1080 <p>The following are such interfaces.  For all of these interfaces Perl
1081 currently (as of 5.8.3) simply assumes byte strings both as arguments
1082 and results, or UTF-8 strings if the <code>encoding</code> pragma has been used.</p>
1083 <p>One reason why Perl does not attempt to resolve the role of Unicode in
1084 this cases is that the answers are highly dependent on the operating
1085 system and the file system(s).  For example, whether filenames can be
1086 in Unicode, and in exactly what kind of encoding, is not exactly a
1087 portable concept.  Similarly for the qx and system: how well will the
1088 'command line interface' (and which of them?) handle Unicode?</p>
1089 <ul>
1090 <li>
1091 <p>chdir, chmod, chown, chroot, exec, link, lstat, mkdir,
1092 rename, rmdir, stat, symlink, truncate, unlink, utime, -X</p>
1093 </li>
1094 <li>
1095 <p>%ENV</p>
1096 </li>
1097 <li>
1098 <p>glob (aka the &lt;*&gt;)</p>
1099 </li>
1100 <li>
1101 <p>open, opendir, sysopen</p>
1102 </li>
1103 <li>
1104 <p>qx (aka the backtick operator), system</p>
1105 </li>
1106 <li>
1107 <p>readdir, readlink</p>
1108 </li>
1109 </ul>
1110 <p>
1111 </p>
1112 <h2><a name="forcing_unicode_in_perl__or_unforcing_unicode_in_perl_">Forcing Unicode in Perl (Or Unforcing Unicode in Perl)</a></h2>
1113 <p>Sometimes (see <a href="#when_unicode_does_not_happen">When Unicode Does Not Happen</a>) there are
1114 situations where you simply need to force Perl to believe that a byte
1115 string is UTF-8, or vice versa.  The low-level calls
1116 utf8::upgrade($bytestring) and utf8::downgrade($utf8string) are
1117 the answers.</p>
1118 <p>Do not use them without careful thought, though: Perl may easily get
1119 very confused, angry, or even crash, if you suddenly change the 'nature'
1120 of scalar like that.  Especially careful you have to be if you use the
1121 utf8::upgrade(): any random byte string is not valid UTF-8.</p>
1122 <p>
1123 </p>
1124 <h2><a name="using_unicode_in_xs">Using Unicode in XS</a></h2>
1125 <p>If you want to handle Perl Unicode in XS extensions, you may find the
1126 following C APIs useful.  See also <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#unicode_support">Unicode Support in the perlguts manpage</a> for an
1127 explanation about Unicode at the XS level, and <a href="file://C|\msysgit\mingw\html/pod/perlapi.html">the perlapi manpage</a> for the API
1128 details.</p>
1129 <ul>
1130 <li>
1131 <p><code>DO_UTF8(sv)</code> returns true if the <code>UTF8</code> flag is on and the bytes
1132 pragma is not in effect.  <code>SvUTF8(sv)</code> returns true is the <code>UTF8</code>
1133 flag is on; the bytes pragma is ignored.  The <code>UTF8</code> flag being on
1134 does <strong>not</strong> mean that there are any characters of code points greater
1135 than 255 (or 127) in the scalar or that there are even any characters
1136 in the scalar.  What the <code>UTF8</code> flag means is that the sequence of
1137 octets in the representation of the scalar is the sequence of UTF-8
1138 encoded code points of the characters of a string.  The <code>UTF8</code> flag
1139 being off means that each octet in this representation encodes a
1140 single character with code point 0..255 within the string.  Perl's
1141 Unicode model is not to use UTF-8 until it is absolutely necessary.</p>
1142 </li>
1143 <li>
1144 <p><code>uvuni_to_utf8(buf, chr)</code> writes a Unicode character code point into
1145 a buffer encoding the code point as UTF-8, and returns a pointer
1146 pointing after the UTF-8 bytes.</p>
1147 </li>
1148 <li>
1149 <p><code>utf8_to_uvuni(buf, lenp)</code> reads UTF-8 encoded bytes from a buffer and
1150 returns the Unicode character code point and, optionally, the length of
1151 the UTF-8 byte sequence.</p>
1152 </li>
1153 <li>
1154 <p><code>utf8_length(start, end)</code> returns the length of the UTF-8 encoded buffer
1155 in characters.  <code>sv_len_utf8(sv)</code> returns the length of the UTF-8 encoded
1156 scalar.</p>
1157 </li>
1158 <li>
1159 <p><code>sv_utf8_upgrade(sv)</code> converts the string of the scalar to its UTF-8
1160 encoded form.  <code>sv_utf8_downgrade(sv)</code> does the opposite, if
1161 possible.  <code>sv_utf8_encode(sv)</code> is like sv_utf8_upgrade except that
1162 it does not set the <code>UTF8</code> flag.  <code>sv_utf8_decode()</code> does the
1163 opposite of <code>sv_utf8_encode()</code>.  Note that none of these are to be
1164 used as general-purpose encoding or decoding interfaces: <code>use Encode</code>
1165 for that.  <code>sv_utf8_upgrade()</code> is affected by the encoding pragma
1166 but <code>sv_utf8_downgrade()</code> is not (since the encoding pragma is
1167 designed to be a one-way street).</p>
1168 </li>
1169 <li>
1170 <p><code>is_utf8_char(s)</code> returns true if the pointer points to a valid UTF-8
1171 character.</p>
1172 </li>
1173 <li>
1174 <p><code>is_utf8_string(buf, len)</code> returns true if <code>len</code> bytes of the buffer
1175 are valid UTF-8.</p>
1176 </li>
1177 <li>
1178 <p><code>UTF8SKIP(buf)</code> will return the number of bytes in the UTF-8 encoded
1179 character in the buffer.  <code>UNISKIP(chr)</code> will return the number of bytes
1180 required to UTF-8-encode the Unicode character code point.  <code>UTF8SKIP()</code>
1181 is useful for example for iterating over the characters of a UTF-8
1182 encoded buffer; <code>UNISKIP()</code> is useful, for example, in computing
1183 the size required for a UTF-8 encoded buffer.</p>
1184 </li>
1185 <li>
1186 <p><code>utf8_distance(a, b)</code> will tell the distance in characters between the
1187 two pointers pointing to the same UTF-8 encoded buffer.</p>
1188 </li>
1189 <li>
1190 <p><code>utf8_hop(s, off)</code> will return a pointer to an UTF-8 encoded buffer
1191 that is <code>off</code> (positive or negative) Unicode characters displaced
1192 from the UTF-8 buffer <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_s"><code>s</code></a>.  Be careful not to overstep the buffer:
1193 <code>utf8_hop()</code> will merrily run off the end or the beginning of the
1194 buffer if told to do so.</p>
1195 </li>
1196 <li>
1197 <p><code>pv_uni_display(dsv, spv, len, pvlim, flags)</code> and
1198 <code>sv_uni_display(dsv, ssv, pvlim, flags)</code> are useful for debugging the
1199 output of Unicode strings and scalars.  By default they are useful
1200 only for debugging--they display <strong>all</strong> characters as hexadecimal code
1201 points--but with the flags <code>UNI_DISPLAY_ISPRINT</code>,
1202 <code>UNI_DISPLAY_BACKSLASH</code>, and <code>UNI_DISPLAY_QQ</code> you can make the
1203 output more readable.</p>
1204 </li>
1205 <li>
1206 <p><code>ibcmp_utf8(s1, pe1, u1, l1, u1, s2, pe2, l2, u2)</code> can be used to
1207 compare two strings case-insensitively in Unicode.  For case-sensitive
1208 comparisons you can just use <code>memEQ()</code> and <code>memNE()</code> as usual.</p>
1209 </li>
1210 </ul>
1211 <p>For more information, see <a href="file://C|\msysgit\mingw\html/pod/perlapi.html">the perlapi manpage</a>, and <em>utf8.c</em> and <em>utf8.h</em>
1212 in the Perl source code distribution.</p>
1213 <p>
1214 </p>
1215 <hr />
1216 <h1><a name="bugs">BUGS</a></h1>
1217 <p>
1218 </p>
1219 <h2><a name="interaction_with_locales">Interaction with Locales</a></h2>
1220 <p>Use of locales with Unicode data may lead to odd results.  Currently,
1221 Perl attempts to attach 8-bit locale info to characters in the range
1222 0..255, but this technique is demonstrably incorrect for locales that
1223 use characters above that range when mapped into Unicode.  Perl's
1224 Unicode support will also tend to run slower.  Use of locales with
1225 Unicode is discouraged.</p>
1226 <p>
1227 </p>
1228 <h2><a name="interaction_with_extensions">Interaction with Extensions</a></h2>
1229 <p>When Perl exchanges data with an extension, the extension should be
1230 able to understand the UTF-8 flag and act accordingly. If the
1231 extension doesn't know about the flag, it's likely that the extension
1232 will return incorrectly-flagged data.</p>
1233 <p>So if you're working with Unicode data, consult the documentation of
1234 every module you're using if there are any issues with Unicode data
1235 exchange. If the documentation does not talk about Unicode at all,
1236 suspect the worst and probably look at the source to learn how the
1237 module is implemented. Modules written completely in Perl shouldn't
1238 cause problems. Modules that directly or indirectly access code written
1239 in other programming languages are at risk.</p>
1240 <p>For affected functions, the simple strategy to avoid data corruption is
1241 to always make the encoding of the exchanged data explicit. Choose an
1242 encoding that you know the extension can handle. Convert arguments passed
1243 to the extensions to that encoding and convert results back from that
1244 encoding. Write wrapper functions that do the conversions for you, so
1245 you can later change the functions when the extension catches up.</p>
1246 <p>To provide an example, let's say the popular Foo::Bar::escape_html
1247 function doesn't deal with Unicode data yet. The wrapper function
1248 would convert the argument to raw UTF-8 and convert the result back to
1249 Perl's internal representation like so:</p>
1250 <pre>
1251     sub my_escape_html ($) {
1252       my($what) = shift;
1253       return unless defined $what;
1254       Encode::decode_utf8(Foo::Bar::escape_html(Encode::encode_utf8($what)));
1255     }</pre>
1256 <p>Sometimes, when the extension does not convert data but just stores
1257 and retrieves them, you will be in a position to use the otherwise
1258 dangerous Encode::_utf8_on() function. Let's say the popular
1259 <code>Foo::Bar</code> extension, written in C, provides a <code>param</code> method that
1260 lets you store and retrieve data according to these prototypes:</p>
1261 <pre>
1262     $self-&gt;param($name, $value);            # set a scalar
1263     $value = $self-&gt;param($name);           # retrieve a scalar</pre>
1264 <p>If it does not yet provide support for any encoding, one could write a
1265 derived class with such a <code>param</code> method:</p>
1266 <pre>
1267     sub param {
1268       my($self,$name,$value) = @_;
1269       utf8::upgrade($name);     # make sure it is UTF-8 encoded
1270       if (defined $value)
1271         utf8::upgrade($value);  # make sure it is UTF-8 encoded
1272         return $self-&gt;SUPER::param($name,$value);
1273       } else {
1274         my $ret = $self-&gt;SUPER::param($name);
1275         Encode::_utf8_on($ret); # we know, it is UTF-8 encoded
1276         return $ret;
1277       }
1278     }</pre>
1279 <p>Some extensions provide filters on data entry/exit points, such as
1280 DB_File::filter_store_key and family. Look out for such filters in
1281 the documentation of your extensions, they can make the transition to
1282 Unicode data much easier.</p>
1283 <p>
1284 </p>
1285 <h2><a name="speed">Speed</a></h2>
1286 <p>Some functions are slower when working on UTF-8 encoded strings than
1287 on byte encoded strings.  All functions that need to hop over
1288 characters such as length(), <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item_substr"><code>substr()</code></a> or index(), or matching regular
1289 expressions can work <strong>much</strong> faster when the underlying data are
1290 byte-encoded.</p>
1291 <p>In Perl 5.8.0 the slowness was often quite spectacular; in Perl 5.8.1
1292 a caching scheme was introduced which will hopefully make the slowness
1293 somewhat less spectacular, at least for some operations.  In general,
1294 operations with UTF-8 encoded strings are still slower. As an example,
1295 the Unicode properties (character classes) like <code>\p{Nd}</code> are known to
1296 be quite a bit slower (5-20 times) than their simpler counterparts
1297 like <code>\d</code> (then again, there 268 Unicode characters matching <code>Nd</code>
1298 compared with the 10 ASCII characters matching <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_d"><code>d</code></a>).</p>
1299 <p>
1300 </p>
1301 <h2><a name="porting_code_from_perl5_6_x">Porting code from perl-5.6.X</a></h2>
1302 <p>Perl 5.8 has a different Unicode model from 5.6. In 5.6 the programmer
1303 was required to use the <code>utf8</code> pragma to declare that a given scope
1304 expected to deal with Unicode data and had to make sure that only
1305 Unicode data were reaching that scope. If you have code that is
1306 working with 5.6, you will need some of the following adjustments to
1307 your code. The examples are written such that the code will continue
1308 to work under 5.6, so you should be safe to try them out.</p>
1309 <ul>
1310 <li>
1311 <p>A filehandle that should read or write UTF-8</p>
1312 <pre>
1313   if ($] &gt; 5.007) {
1314     binmode $fh, &quot;:utf8&quot;;
1315   }</pre>
1316 </li>
1317 <li>
1318 <p>A scalar that is going to be passed to some extension</p>
1319 <p>Be it Compress::Zlib, Apache::Request or any extension that has no
1320 mention of Unicode in the manpage, you need to make sure that the
1321 UTF-8 flag is stripped off. Note that at the time of this writing
1322 (October 2002) the mentioned modules are not UTF-8-aware. Please
1323 check the documentation to verify if this is still true.</p>
1324 <pre>
1325   if ($] &gt; 5.007) {
1326     require Encode;
1327     $val = Encode::encode_utf8($val); # make octets
1328   }</pre>
1329 </li>
1330 <li>
1331 <p>A scalar we got back from an extension</p>
1332 <p>If you believe the scalar comes back as UTF-8, you will most likely
1333 want the UTF-8 flag restored:</p>
1334 <pre>
1335   if ($] &gt; 5.007) {
1336     require Encode;
1337     $val = Encode::decode_utf8($val);
1338   }</pre>
1339 </li>
1340 <li>
1341 <p>Same thing, if you are really sure it is UTF-8</p>
1342 <pre>
1343   if ($] &gt; 5.007) {
1344     require Encode;
1345     Encode::_utf8_on($val);
1346   }</pre>
1347 </li>
1348 <li>
1349 <p>A wrapper for fetchrow_array and fetchrow_hashref</p>
1350 <p>When the database contains only UTF-8, a wrapper function or method is
1351 a convenient way to replace all your fetchrow_array and
1352 fetchrow_hashref calls. A wrapper function will also make it easier to
1353 adapt to future enhancements in your database driver. Note that at the
1354 time of this writing (October 2002), the DBI has no standardized way
1355 to deal with UTF-8 data. Please check the documentation to verify if
1356 that is still true.</p>
1357 <pre>
1358   sub fetchrow {
1359     my($self, $sth, $what) = @_; # $what is one of fetchrow_{array,hashref}
1360     if ($] &lt; 5.007) {
1361       return $sth-&gt;$what;
1362     } else {
1363       require Encode;
1364       if (wantarray) {
1365         my @arr = $sth-&gt;$what;
1366         for (@arr) {
1367           defined &amp;&amp; /[^\000-\177]/ &amp;&amp; Encode::_utf8_on($_);
1368         }
1369         return @arr;
1370       } else {
1371         my $ret = $sth-&gt;$what;
1372         if (ref $ret) {
1373           for my $k (keys %$ret) {
1374             defined &amp;&amp; /[^\000-\177]/ &amp;&amp; Encode::_utf8_on($_) for $ret-&gt;{$k};
1375           }
1376           return $ret;
1377         } else {
1378           defined &amp;&amp; /[^\000-\177]/ &amp;&amp; Encode::_utf8_on($_) for $ret;
1379           return $ret;
1380         }
1381       }
1382     }
1383   }</pre>
1384 </li>
1385 <li>
1386 <p>A large scalar that you know can only contain ASCII</p>
1387 <p>Scalars that contain only ASCII and are marked as UTF-8 are sometimes
1388 a drag to your program. If you recognize such a situation, just remove
1389 the UTF-8 flag:</p>
1390 <pre>
1391   utf8::downgrade($val) if $] &gt; 5.007;</pre>
1392 </li>
1393 </ul>
1394 <p>
1395 </p>
1396 <hr />
1397 <h1><a name="see_also">SEE ALSO</a></h1>
1398 <p><a href="file://C|\msysgit\mingw\html/pod/perluniintro.html">the perluniintro manpage</a>, <a href="file://C|\msysgit\mingw\html/lib/encoding.html">the encoding manpage</a>, <a href="file://C|\msysgit\mingw\html/lib/Encode.html">the Encode manpage</a>, <a href="file://C|\msysgit\mingw\html/lib/open.html">the open manpage</a>, <a href="file://C|\msysgit\mingw\html/lib/utf8.html">the utf8 manpage</a>, <a href="file://C|\msysgit\mingw\html/lib/bytes.html">the bytes manpage</a>,
1399 <a href="file://C|\msysgit\mingw\html/pod/perlretut.html">the perlretut manpage</a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item____unicode_">${^UNICODE} in the perlvar manpage</a></p>
1400 <table border="0" width="100%" cellspacing="0" cellpadding="3">
1401 <tr><td class="block" style="background-color: #cccccc" valign="middle">
1402 <big><strong><span class="block">&nbsp;perlunicode - Unicode support in Perl</span></strong></big>
1403 </td></tr>
1404 </table>
1405
1406 </body>
1407
1408 </html>