This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
PATCH: Makefile in lib/unicore shouldn't clean .t file
[perl5.git] / pod / perlunicode.pod
CommitLineData
393fec97
GS
1=head1 NAME
2
3perlunicode - Unicode support in Perl
4
5=head1 DESCRIPTION
6
0a1f2d14 7=head2 Important Caveats
21bad921 8
376d9008 9Unicode support is an extensive requirement. While Perl does not
c349b1b9
JH
10implement the Unicode standard or the accompanying technical reports
11from cover to cover, Perl does support many Unicode features.
21bad921 12
2575c402 13People who want to learn to use Unicode in Perl, should probably read
e4911a48
RS
14L<the Perl Unicode tutorial, perlunitut|perlunitut>, before reading
15this reference document.
2575c402 16
13a2d996 17=over 4
21bad921 18
fae2c0fb 19=item Input and Output Layers
21bad921 20
376d9008 21Perl knows when a filehandle uses Perl's internal Unicode encodings
1bfb14c4
JH
22(UTF-8, or UTF-EBCDIC if in EBCDIC) if the filehandle is opened with
23the ":utf8" layer. Other encodings can be converted to Perl's
24encoding on input or from Perl's encoding on output by use of the
25":encoding(...)" layer. See L<open>.
c349b1b9 26
2575c402 27To indicate that Perl source itself is in UTF-8, use C<use utf8;>.
21bad921
GS
28
29=item Regular Expressions
30
c349b1b9 31The regular expression compiler produces polymorphic opcodes. That is,
376d9008 32the pattern adapts to the data and automatically switches to the Unicode
2575c402
JW
33character scheme when presented with data that is internally encoded in
34UTF-8 -- or instead uses a traditional byte scheme when presented with
35byte data.
21bad921 36
ad0029c4 37=item C<use utf8> still needed to enable UTF-8/UTF-EBCDIC in scripts
21bad921 38
376d9008
JB
39As a compatibility measure, the C<use utf8> pragma must be explicitly
40included to enable recognition of UTF-8 in the Perl scripts themselves
1bfb14c4
JH
41(in string or regular expression literals, or in identifier names) on
42ASCII-based machines or to recognize UTF-EBCDIC on EBCDIC-based
376d9008 43machines. B<These are the only times when an explicit C<use utf8>
8f8cf39c 44is needed.> See L<utf8>.
21bad921 45
7aa207d6
JH
46=item BOM-marked scripts and UTF-16 scripts autodetected
47
48If a Perl script begins marked with the Unicode BOM (UTF-16LE, UTF16-BE,
49or UTF-8), or if the script looks like non-BOM-marked UTF-16 of either
50endianness, Perl will correctly read in the script as Unicode.
51(BOMless UTF-8 cannot be effectively recognized or differentiated from
52ISO 8859-1 or other eight-bit encodings.)
53
990e18f7
AT
54=item C<use encoding> needed to upgrade non-Latin-1 byte strings
55
38a44b82 56By default, there is a fundamental asymmetry in Perl's Unicode model:
990e18f7
AT
57implicit upgrading from byte strings to Unicode strings assumes that
58they were encoded in I<ISO 8859-1 (Latin-1)>, but Unicode strings are
59downgraded with UTF-8 encoding. This happens because the first 256
51f494cc 60codepoints in Unicode happens to agree with Latin-1.
990e18f7 61
990e18f7
AT
62See L</"Byte and Character Semantics"> for more details.
63
21bad921
GS
64=back
65
376d9008 66=head2 Byte and Character Semantics
393fec97 67
376d9008 68Beginning with version 5.6, Perl uses logically-wide characters to
3e4dbfed 69represent strings internally.
393fec97 70
376d9008
JB
71In future, Perl-level operations will be expected to work with
72characters rather than bytes.
393fec97 73
376d9008 74However, as an interim compatibility measure, Perl aims to
75daf61c
JH
75provide a safe migration path from byte semantics to character
76semantics for programs. For operations where Perl can unambiguously
376d9008 77decide that the input data are characters, Perl switches to
75daf61c
JH
78character semantics. For operations where this determination cannot
79be made without additional information from the user, Perl decides in
376d9008 80favor of compatibility and chooses to use byte semantics.
8cbd9a7a 81
51f494cc 82Under byte semantics, when C<use locale> is in effect, Perl uses the
e1b711da
KW
83semantics associated with the current locale. Absent a C<use locale>, and
84absent a C<use feature 'unicode_strings'> pragma, Perl currently uses US-ASCII
85(or Basic Latin in Unicode terminology) byte semantics, meaning that characters
86whose ordinal numbers are in the range 128 - 255 are undefined except for their
87ordinal numbers. This means that none have case (upper and lower), nor are any
88a member of character classes, like C<[:alpha:]> or C<\w>. (But all do belong
89to the C<\W> class or the Perl regular expression extension C<[:^alpha:]>.)
2bbc8d55 90
8cbd9a7a 91This behavior preserves compatibility with earlier versions of Perl,
376d9008 92which allowed byte semantics in Perl operations only if
e1b711da 93none of the program's inputs were marked as being a source of Unicode
8cbd9a7a
GS
94character data. Such data may come from filehandles, from calls to
95external programs, from information provided by the system (such as %ENV),
21bad921 96or from literals and constants in the source text.
8cbd9a7a 97
376d9008
JB
98The C<bytes> pragma will always, regardless of platform, force byte
99semantics in a particular lexical scope. See L<bytes>.
8cbd9a7a 100
e1b711da
KW
101The C<use feature 'unicode_strings'> pragma is intended to always, regardless
102of platform, force Unicode semantics in a particular lexical scope. In
103release 5.12, it is partially implemented, applying only to case changes.
104See L</The "Unicode Bug"> below.
105
8cbd9a7a 106The C<utf8> pragma is primarily a compatibility device that enables
75daf61c 107recognition of UTF-(8|EBCDIC) in literals encountered by the parser.
376d9008
JB
108Note that this pragma is only required while Perl defaults to byte
109semantics; when character semantics become the default, this pragma
110may become a no-op. See L<utf8>.
111
112Unless explicitly stated, Perl operators use character semantics
113for Unicode data and byte semantics for non-Unicode data.
114The decision to use character semantics is made transparently. If
115input data comes from a Unicode source--for example, if a character
fae2c0fb 116encoding layer is added to a filehandle or a literal Unicode
376d9008
JB
117string constant appears in a program--character semantics apply.
118Otherwise, byte semantics are in effect. The C<bytes> pragma should
e1b711da
KW
119be used to force byte semantics on Unicode data, and the C<use feature
120'unicode_strings'> pragma to force Unicode semantics on byte data (though in
1215.12 it isn't fully implemented).
376d9008
JB
122
123If strings operating under byte semantics and strings with Unicode
51f494cc 124character data are concatenated, the new string will have
42bde815 125character semantics. This can cause surprises: See L</BUGS>, below
7dedd01f 126
feda178f 127Under character semantics, many operations that formerly operated on
376d9008 128bytes now operate on characters. A character in Perl is
feda178f 129logically just a number ranging from 0 to 2**31 or so. Larger
376d9008
JB
130characters may encode into longer sequences of bytes internally, but
131this internal detail is mostly hidden for Perl code.
132See L<perluniintro> for more.
393fec97 133
376d9008 134=head2 Effects of Character Semantics
393fec97
GS
135
136Character semantics have the following effects:
137
138=over 4
139
140=item *
141
376d9008 142Strings--including hash keys--and regular expression patterns may
574c8022 143contain characters that have an ordinal value larger than 255.
393fec97 144
2575c402
JW
145If you use a Unicode editor to edit your program, Unicode characters may
146occur directly within the literal strings in UTF-8 encoding, or UTF-16.
147(The former requires a BOM or C<use utf8>, the latter requires a BOM.)
3e4dbfed 148
2575c402
JW
149Unicode characters can also be added to a string by using the C<\x{...}>
150notation. The Unicode code for the desired character, in hexadecimal,
151should be placed in the braces. For instance, a smiley face is
2bbc8d55 152C<\x{263A}>. This encoding scheme works for all characters, but
2575c402
JW
153for characters under 0x100, note that Perl may use an 8 bit encoding
154internally, for optimization and/or backward compatibility.
3e4dbfed
JF
155
156Additionally, if you
574c8022 157
3e4dbfed 158 use charnames ':full';
574c8022 159
1bfb14c4
JH
160you can use the C<\N{...}> notation and put the official Unicode
161character name within the braces, such as C<\N{WHITE SMILING FACE}>.
376d9008 162
393fec97
GS
163=item *
164
574c8022
JH
165If an appropriate L<encoding> is specified, identifiers within the
166Perl script may contain Unicode alphanumeric characters, including
376d9008
JB
167ideographs. Perl does not currently attempt to canonicalize variable
168names.
393fec97 169
393fec97
GS
170=item *
171
1bfb14c4 172Regular expressions match characters instead of bytes. "." matches
2575c402 173a character instead of a byte.
393fec97 174
393fec97
GS
175=item *
176
177Character classes in regular expressions match characters instead of
376d9008 178bytes and match against the character properties specified in the
1bfb14c4 179Unicode properties database. C<\w> can be used to match a Japanese
75daf61c 180ideograph, for instance.
393fec97 181
393fec97
GS
182=item *
183
eb0cc9e3 184Named Unicode properties, scripts, and block ranges may be used like
376d9008 185character classes via the C<\p{}> "matches property" construct and
822502e5 186the C<\P{}> negation, "doesn't match property".
2575c402 187See L</"Unicode Character Properties"> for more details.
822502e5
ST
188
189You can define your own character properties and use them
190in the regular expression with the C<\p{}> or C<\P{}> construct.
822502e5
ST
191See L</"User-Defined Character Properties"> for more details.
192
193=item *
194
51f494cc
KW
195The special pattern C<\X> matches a logical character, an C<extended grapheme
196cluster> in Standardese. In Unicode what appears to the user to be a single
197character, for example an accented C<G>, may in fact be composed of a sequence
198of characters, in this case a C<G> followed by an accent character. C<\X>
199will match the entire sequence.
822502e5
ST
200
201=item *
202
203The C<tr///> operator translates characters instead of bytes. Note
204that the C<tr///CU> functionality has been removed. For similar
205functionality see pack('U0', ...) and pack('C0', ...).
206
207=item *
208
209Case translation operators use the Unicode case translation tables
210when character input is provided. Note that C<uc()>, or C<\U> in
211interpolated strings, translates to uppercase, while C<ucfirst>,
212or C<\u> in interpolated strings, translates to titlecase in languages
e1b711da
KW
213that make the distinction (which is equivalent to uppercase in languages
214without the distinction).
822502e5
ST
215
216=item *
217
218Most operators that deal with positions or lengths in a string will
219automatically switch to using character positions, including
220C<chop()>, C<chomp()>, C<substr()>, C<pos()>, C<index()>, C<rindex()>,
221C<sprintf()>, C<write()>, and C<length()>. An operator that
51f494cc
KW
222specifically does not switch is C<vec()>. Operators that really don't
223care include operators that treat strings as a bucket of bits such as
822502e5
ST
224C<sort()>, and operators dealing with filenames.
225
226=item *
227
51f494cc 228The C<pack()>/C<unpack()> letter C<C> does I<not> change, since it is often
822502e5
ST
229used for byte-oriented formats. Again, think C<char> in the C language.
230
231There is a new C<U> specifier that converts between Unicode characters
232and code points. There is also a C<W> specifier that is the equivalent of
233C<chr>/C<ord> and properly handles character values even if they are above 255.
234
235=item *
236
237The C<chr()> and C<ord()> functions work on characters, similar to
238C<pack("W")> and C<unpack("W")>, I<not> C<pack("C")> and
239C<unpack("C")>. C<pack("C")> and C<unpack("C")> are methods for
240emulating byte-oriented C<chr()> and C<ord()> on Unicode strings.
241While these methods reveal the internal encoding of Unicode strings,
242that is not something one normally needs to care about at all.
243
244=item *
245
246The bit string operators, C<& | ^ ~>, can operate on character data.
247However, for backward compatibility, such as when using bit string
248operations when characters are all less than 256 in ordinal value, one
249should not use C<~> (the bit complement) with characters of both
250values less than 256 and values greater than 256. Most importantly,
251DeMorgan's laws (C<~($x|$y) eq ~$x&~$y> and C<~($x&$y) eq ~$x|~$y>)
252will not hold. The reason for this mathematical I<faux pas> is that
253the complement cannot return B<both> the 8-bit (byte-wide) bit
254complement B<and> the full character-wide bit complement.
255
256=item *
257
e1b711da 258You can define your own mappings to be used in lc(),
822502e5 259lcfirst(), uc(), and ucfirst() (or their string-inlined versions).
822502e5
ST
260See L</"User-Defined Case Mappings"> for more details.
261
262=back
263
264=over 4
265
266=item *
267
268And finally, C<scalar reverse()> reverses by character rather than by byte.
269
270=back
271
272=head2 Unicode Character Properties
273
51f494cc
KW
274Most Unicode character properties are accessible by using regular expressions.
275They are used like character classes via the C<\p{}> "matches property"
276construct and the C<\P{}> negation, "doesn't match property".
277
278For instance, C<\p{Uppercase}> matches any character with the Unicode
279"Uppercase" property, while C<\p{L}> matches any character with a
280General_Category of "L" (letter) property. Brackets are not
281required for single letter properties, so C<\p{L}> is equivalent to C<\pL>.
282
e1b711da
KW
283More formally, C<\p{Uppercase}> matches any character whose Unicode Uppercase
284property value is True, and C<\P{Uppercase}> matches any character whose
285Uppercase property value is False, and they could have been written as
51f494cc
KW
286C<\p{Uppercase=True}> and C<\p{Uppercase=False}>, respectively
287
288This formality is needed when properties are not binary, that is if they can
289take on more values than just True and False. For example, the Bidi_Class (see
290L</"Bidirectional Character Types"> below), can take on a number of different
291values, such as Left, Right, Whitespace, and others. To match these, one needs
e1b711da 292to specify the property name (Bidi_Class), and the value being matched against
51f494cc
KW
293(Left, Right, etc.). This is done, as in the examples above, by having the two
294components separated by an equal sign (or interchangeably, a colon), like
295C<\p{Bidi_Class: Left}>.
296
297All Unicode-defined character properties may be written in these compound forms
298of C<\p{property=value}> or C<\p{property:value}>, but Perl provides some
299additional properties that are written only in the single form, as well as
300single-form short-cuts for all binary properties and certain others described
301below, in which you may omit the property name and the equals or colon
302separator.
303
304Most Unicode character properties have at least two synonyms (or aliases if you
305prefer), a short one that is easier to type, and a longer one which is more
306descriptive and hence it is easier to understand what it means. Thus the "L"
307and "Letter" above are equivalent and can be used interchangeably. Likewise,
308"Upper" is a synonym for "Uppercase", and we could have written
309C<\p{Uppercase}> equivalently as C<\p{Upper}>. Also, there are typically
310various synonyms for the values the property can be. For binary properties,
311"True" has 3 synonyms: "T", "Yes", and "Y"; and "False has correspondingly "F",
312"No", and "N". But be careful. A short form of a value for one property may
e1b711da 313not mean the same thing as the same short form for another. Thus, for the
51f494cc
KW
314General_Category property, "L" means "Letter", but for the Bidi_Class property,
315"L" means "Left". A complete list of properties and synonyms is in
316L<perluniprops>.
317
318Upper/lower case differences in the property names and values are irrelevant,
319thus C<\p{Upper}> means the same thing as C<\p{upper}> or even C<\p{UpPeR}>.
320Similarly, you can add or subtract underscores anywhere in the middle of a
321word, so that these are also equivalent to C<\p{U_p_p_e_r}>. And white space
322is irrelevant adjacent to non-word characters, such as the braces and the equals
323or colon separators so C<\p{ Upper }> and C<\p{ Upper_case : Y }> are
324equivalent to these as well. In fact, in most cases, white space and even
325hyphens can be added or deleted anywhere. So even C<\p{ Up-per case = Yes}> is
326equivalent. All this is called "loose-matching" by Unicode. The few places
327where stricter matching is employed is in the middle of numbers, and the Perl
328extension properties that begin or end with an underscore. Stricter matching
329cares about white space (except adjacent to the non-word characters) and
330hyphens, and non-interior underscores.
4193bef7 331
376d9008
JB
332You can also use negation in both C<\p{}> and C<\P{}> by introducing a caret
333(^) between the first brace and the property name: C<\p{^Tamil}> is
eb0cc9e3 334equal to C<\P{Tamil}>.
4193bef7 335
51f494cc 336=head3 B<General_Category>
14bb0a9a 337
51f494cc
KW
338Every Unicode character is assigned a general category, which is the "most
339usual categorization of a character" (from
340L<http://www.unicode.org/reports/tr44>).
822502e5 341
51f494cc
KW
342The compound way of writing these is like C<{\p{General_Category=Number}>
343(short, C<\p{gc:n}>). But Perl furnishes shortcuts in which everything up
344through the equal or colon separator is omitted. So you can instead just write
345C<\pN>.
822502e5 346
51f494cc 347Here are the short and long forms of the General Category properties:
393fec97 348
d73e5302
JH
349 Short Long
350
351 L Letter
51f494cc
KW
352 LC, L& Cased_Letter (that is: [\p{Ll}\p{Lu}\p{Lt}])
353 Lu Uppercase_Letter
354 Ll Lowercase_Letter
355 Lt Titlecase_Letter
356 Lm Modifier_Letter
357 Lo Other_Letter
d73e5302
JH
358
359 M Mark
51f494cc
KW
360 Mn Nonspacing_Mark
361 Mc Spacing_Mark
362 Me Enclosing_Mark
d73e5302
JH
363
364 N Number
51f494cc
KW
365 Nd Decimal_Number (also Digit)
366 Nl Letter_Number
367 No Other_Number
368
369 P Punctuation (also Punct)
370 Pc Connector_Punctuation
371 Pd Dash_Punctuation
372 Ps Open_Punctuation
373 Pe Close_Punctuation
374 Pi Initial_Punctuation
d73e5302 375 (may behave like Ps or Pe depending on usage)
51f494cc 376 Pf Final_Punctuation
d73e5302 377 (may behave like Ps or Pe depending on usage)
51f494cc 378 Po Other_Punctuation
d73e5302
JH
379
380 S Symbol
51f494cc
KW
381 Sm Math_Symbol
382 Sc Currency_Symbol
383 Sk Modifier_Symbol
384 So Other_Symbol
d73e5302
JH
385
386 Z Separator
51f494cc
KW
387 Zs Space_Separator
388 Zl Line_Separator
389 Zp Paragraph_Separator
d73e5302
JH
390
391 C Other
51f494cc 392 Cc Control (also Cntrl)
e150c829 393 Cf Format
eb0cc9e3 394 Cs Surrogate (not usable)
51f494cc 395 Co Private_Use
e150c829 396 Cn Unassigned
1ac13f9a 397
376d9008 398Single-letter properties match all characters in any of the
3e4dbfed 399two-letter sub-properties starting with the same letter.
12ac2576
JP
400C<LC> and C<L&> are special cases, which are aliases for the set of
401C<Ll>, C<Lu>, and C<Lt>.
32293815 402
eb0cc9e3 403Because Perl hides the need for the user to understand the internal
1bfb14c4
JH
404representation of Unicode characters, there is no need to implement
405the somewhat messy concept of surrogates. C<Cs> is therefore not
eb0cc9e3 406supported.
d73e5302 407
51f494cc 408=head3 B<Bidirectional Character Types>
822502e5 409
376d9008 410Because scripts differ in their directionality--Hebrew is
12ac2576 411written right to left, for example--Unicode supplies these properties in
51f494cc 412the Bidi_Class class:
32293815 413
eb0cc9e3 414 Property Meaning
92e830a9 415
12ac2576
JP
416 L Left-to-Right
417 LRE Left-to-Right Embedding
418 LRO Left-to-Right Override
419 R Right-to-Left
51f494cc 420 AL Arabic Letter
12ac2576
JP
421 RLE Right-to-Left Embedding
422 RLO Right-to-Left Override
423 PDF Pop Directional Format
424 EN European Number
51f494cc
KW
425 ES European Separator
426 ET European Terminator
12ac2576 427 AN Arabic Number
51f494cc 428 CS Common Separator
12ac2576
JP
429 NSM Non-Spacing Mark
430 BN Boundary Neutral
431 B Paragraph Separator
432 S Segment Separator
433 WS Whitespace
434 ON Other Neutrals
435
51f494cc
KW
436This property is always written in the compound form.
437For example, C<\p{Bidi_Class:R}> matches characters that are normally
eb0cc9e3
JH
438written right to left.
439
51f494cc
KW
440=head3 B<Scripts>
441
e1b711da
KW
442The world's languages are written in a number of scripts. This sentence
443(unless you're reading it in translation) is written in Latin, while Russian is
444written in Cyrllic, and Greek is written in, well, Greek; Japanese mainly in
445Hiragana or Katakana. There are many more.
51f494cc
KW
446
447The Unicode Script property gives what script a given character is in,
448and can be matched with the compound form like C<\p{Script=Hebrew}> (short:
449C<\p{sc=hebr}>). Perl furnishes shortcuts for all script names. You can omit
450everything up through the equals (or colon), and simply write C<\p{Latin}> or
451C<\P{Cyrillic}>.
452
453A complete list of scripts and their shortcuts is in L<perluniprops>.
454
455=head3 B<Extended property classes>
456
457There are many more property classes than the basic ones described here,
458including some Perl extensions.
459A complete list is in L<perluniprops>.
460The extensions are more fully described in L<perlrecharclass>
461
462=head3 B<Use of "Is" Prefix>
822502e5 463
1bfb14c4 464For backward compatibility (with Perl 5.6), all properties mentioned
51f494cc
KW
465so far may have C<Is> or C<Is_> prepended to their name, so C<\P{Is_Lu}>, for
466example, is equal to C<\P{Lu}>, and C<\p{IsScript:Arabic}> is equal to
467C<\p{Arabic}>.
eb0cc9e3 468
51f494cc 469=head3 B<Blocks>
2796c109 470
1bfb14c4
JH
471In addition to B<scripts>, Unicode also defines B<blocks> of
472characters. The difference between scripts and blocks is that the
473concept of scripts is closer to natural languages, while the concept
51f494cc
KW
474of blocks is more of an artificial grouping based on groups of Unicode
475characters with consecutive ordinal values. For example, the C<Basic Latin>
476block is all characters whose ordinals are between 0 and 127, inclusive, in
477other words, the ASCII characters. The C<Latin> script contains some letters
478from this block as well as several more, like C<Latin-1 Supplement>,
479C<Latin Extended-A>, I<etc.>, but it does not contain all the characters from
480those blocks. It does not, for example, contain digits, because digits are
481shared across many scripts. Digits and similar groups, like punctuation, are in
482the script called C<Common>. There is also a script called C<Inherited> for
483characters that modify other characters, and inherit the script value of the
484controlling character.
485
486For more about scripts versus blocks, see UAX#24 "Unicode Script Property":
487L<http://www.unicode.org/reports/tr24>
488
489The Script property is likely to be the one you want to use when processing
490natural language; the Block property may be useful in working with the nuts and
491bolts of Unicode.
492
493Block names are matched in the compound form, like C<\p{Block: Arrows}> or
494C<\p{Blk=Hebrew}>. Unlike most other properties only a few block names have a
495Unicode-defined short name. But Perl does provide a (slight) shortcut: You
496can say, for example C<\p{In_Arrows}> or C<\p{In_Hebrew}>. For backwards
497compatibility, the C<In> prefix may be omitted if there is no naming conflict
498with a script or any other property, and you can even use an C<Is> prefix
499instead in those cases. But it is not a good idea to do this, for a couple
500reasons:
501
502=over 4
503
504=item 1
505
506It is confusing. There are many naming conflicts, and you may forget some.
507For example, \p{Hebrew} means the I<script> Hebrew, and NOT the I<block>
508Hebrew. But would you remember that 6 months from now?
509
510=item 2
511
512It is unstable. A new version of Unicode may pre-empt the current meaning by
513creating a property with the same name. There was a time in very early Unicode
514releases when \p{Hebrew} would have matched the I<block> Hebrew; now it
515doesn't.
32293815 516
393fec97
GS
517=back
518
51f494cc
KW
519Some people just prefer to always use C<\p{Block: foo}> and C<\p{Script: bar}>
520instead of the shortcuts, for clarity, and because they can't remember the
521difference between 'In' and 'Is' anyway (or aren't confident that those who
522eventually will read their code will know).
523
524A complete list of blocks and their shortcuts is in L<perluniprops>.
525
376d9008 526=head2 User-Defined Character Properties
491fd90a 527
51f494cc
KW
528You can define your own binary character properties by defining subroutines
529whose names begin with "In" or "Is". The subroutines can be defined in any
530package. The user-defined properties can be used in the regular expression
531C<\p> and C<\P> constructs; if you are using a user-defined property from a
532package other than the one you are in, you must specify its package in the
533C<\p> or C<\P> construct.
bac0b425 534
51f494cc 535 # assuming property Is_Foreign defined in Lang::
bac0b425
JP
536 package main; # property package name required
537 if ($txt =~ /\p{Lang::IsForeign}+/) { ... }
538
539 package Lang; # property package name not required
540 if ($txt =~ /\p{IsForeign}+/) { ... }
541
542
543Note that the effect is compile-time and immutable once defined.
491fd90a 544
376d9008
JB
545The subroutines must return a specially-formatted string, with one
546or more newline-separated lines. Each line must be one of the following:
491fd90a
JH
547
548=over 4
549
550=item *
551
510254c9
A
552A single hexadecimal number denoting a Unicode code point to include.
553
554=item *
555
99a6b1f0 556Two hexadecimal numbers separated by horizontal whitespace (space or
376d9008 557tabular characters) denoting a range of Unicode code points to include.
491fd90a
JH
558
559=item *
560
376d9008 561Something to include, prefixed by "+": a built-in character
bac0b425
JP
562property (prefixed by "utf8::") or a user-defined character property,
563to represent all the characters in that property; two hexadecimal code
564points for a range; or a single hexadecimal code point.
491fd90a
JH
565
566=item *
567
376d9008 568Something to exclude, prefixed by "-": an existing character
bac0b425
JP
569property (prefixed by "utf8::") or a user-defined character property,
570to represent all the characters in that property; two hexadecimal code
571points for a range; or a single hexadecimal code point.
491fd90a
JH
572
573=item *
574
376d9008 575Something to negate, prefixed "!": an existing character
bac0b425
JP
576property (prefixed by "utf8::") or a user-defined character property,
577to represent all the characters in that property; two hexadecimal code
578points for a range; or a single hexadecimal code point.
579
580=item *
581
582Something to intersect with, prefixed by "&": an existing character
583property (prefixed by "utf8::") or a user-defined character property,
584for all the characters except the characters in the property; two
585hexadecimal code points for a range; or a single hexadecimal code point.
491fd90a
JH
586
587=back
588
589For example, to define a property that covers both the Japanese
590syllabaries (hiragana and katakana), you can define
591
592 sub InKana {
d5822f25
A
593 return <<END;
594 3040\t309F
595 30A0\t30FF
491fd90a
JH
596 END
597 }
598
d5822f25
A
599Imagine that the here-doc end marker is at the beginning of the line.
600Now you can use C<\p{InKana}> and C<\P{InKana}>.
491fd90a
JH
601
602You could also have used the existing block property names:
603
604 sub InKana {
605 return <<'END';
606 +utf8::InHiragana
607 +utf8::InKatakana
608 END
609 }
610
611Suppose you wanted to match only the allocated characters,
d5822f25 612not the raw block ranges: in other words, you want to remove
491fd90a
JH
613the non-characters:
614
615 sub InKana {
616 return <<'END';
617 +utf8::InHiragana
618 +utf8::InKatakana
619 -utf8::IsCn
620 END
621 }
622
623The negation is useful for defining (surprise!) negated classes.
624
625 sub InNotKana {
626 return <<'END';
627 !utf8::InHiragana
628 -utf8::InKatakana
629 +utf8::IsCn
630 END
631 }
632
bac0b425
JP
633Intersection is useful for getting the common characters matched by
634two (or more) classes.
635
636 sub InFooAndBar {
637 return <<'END';
638 +main::Foo
639 &main::Bar
640 END
641 }
642
643It's important to remember not to use "&" for the first set -- that
644would be intersecting with nothing (resulting in an empty set).
645
822502e5
ST
646=head2 User-Defined Case Mappings
647
3a2263fe
RGS
648You can also define your own mappings to be used in the lc(),
649lcfirst(), uc(), and ucfirst() (or their string-inlined versions).
822502e5 650The principle is similar to that of user-defined character
51f494cc 651properties: to define subroutines
3a2263fe
RGS
652with names like C<ToLower> (for lc() and lcfirst()), C<ToTitle> (for
653the first character in ucfirst()), and C<ToUpper> (for uc(), and the
654rest of the characters in ucfirst()).
655
51f494cc 656The string returned by the subroutines needs to be two hexadecimal numbers
e1b711da
KW
657separated by two tabulators: the two numbers being, respectively, the source
658code point and the destination code point. For example:
3a2263fe
RGS
659
660 sub ToUpper {
661 return <<END;
51f494cc 662 0061\t\t0041
3a2263fe
RGS
663 END
664 }
665
51f494cc
KW
666defines an uc() mapping that causes only the character "a"
667to be mapped to "A"; all other characters will remain unchanged.
3a2263fe 668
51f494cc
KW
669(For serious hackers only) The above means you have to furnish a complete
670mapping; you can't just override a couple of characters and leave the rest
671unchanged. You can find all the mappings in the directory
672C<$Config{privlib}>/F<unicore/To/>. The mapping data is returned as the
673here-document, and the C<utf8::ToSpecFoo> are special exception mappings
674derived from <$Config{privlib}>/F<unicore/SpecialCasing.txt>. The C<Digit> and
675C<Fold> mappings that one can see in the directory are not directly
676user-accessible, one can use either the C<Unicode::UCD> module, or just match
677case-insensitively (that's when the C<Fold> mapping is used).
3a2263fe 678
51f494cc
KW
679The mappings will only take effect on scalars that have been marked as having
680Unicode characters, for example by using C<utf8::upgrade()>.
681Old byte-style strings are not affected.
3a2263fe 682
51f494cc 683The mappings are in effect for the package they are defined in.
3a2263fe 684
376d9008 685=head2 Character Encodings for Input and Output
8cbd9a7a 686
7221edc9 687See L<Encode>.
8cbd9a7a 688
c29a771d 689=head2 Unicode Regular Expression Support Level
776f8809 690
376d9008
JB
691The following list of Unicode support for regular expressions describes
692all the features currently supported. The references to "Level N"
8158862b
ST
693and the section numbers refer to the Unicode Technical Standard #18,
694"Unicode Regular Expressions", version 11, in May 2005.
776f8809
JH
695
696=over 4
697
698=item *
699
700Level 1 - Basic Unicode Support
701
8158862b
ST
702 RL1.1 Hex Notation - done [1]
703 RL1.2 Properties - done [2][3]
704 RL1.2a Compatibility Properties - done [4]
705 RL1.3 Subtraction and Intersection - MISSING [5]
706 RL1.4 Simple Word Boundaries - done [6]
707 RL1.5 Simple Loose Matches - done [7]
708 RL1.6 Line Boundaries - MISSING [8]
709 RL1.7 Supplementary Code Points - done [9]
710
711 [1] \x{...}
712 [2] \p{...} \P{...}
e1b711da
KW
713 [3] supports not only minimal list, but all Unicode character
714 properties (see L</Unicode Character Properties>)
8158862b
ST
715 [4] \d \D \s \S \w \W \X [:prop:] [:^prop:]
716 [5] can use regular expression look-ahead [a] or
717 user-defined character properties [b] to emulate set operations
718 [6] \b \B
e1b711da
KW
719 [7] note that Perl does Full case-folding in matching (but with bugs),
720 not Simple: for example U+1F88 is equivalent to U+1F00 U+03B9,
2bbc8d55 721 not with 1F80. This difference matters mainly for certain Greek
376d9008
JB
722 capital letters with certain modifiers: the Full case-folding
723 decomposes the letter, while the Simple case-folding would map
e0f9d4a8 724 it to a single character.
8158862b
ST
725 [8] should do ^ and $ also on U+000B (\v in C), FF (\f), CR (\r),
726 CRLF (\r\n), NEL (U+0085), LS (U+2028), and PS (U+2029);
727 should also affect <>, $., and script line numbers;
728 should not split lines within CRLF [c] (i.e. there is no empty
729 line between \r and \n)
730 [9] UTF-8/UTF-EBDDIC used in perl allows not only U+10000 to U+10FFFF
731 but also beyond U+10FFFF [d]
7207e29d 732
237bad5b 733[a] You can mimic class subtraction using lookahead.
8158862b 734For example, what UTS#18 might write as
29bdacb8 735
dbe420b4
JH
736 [{Greek}-[{UNASSIGNED}]]
737
738in Perl can be written as:
739
1d81abf3
JH
740 (?!\p{Unassigned})\p{InGreekAndCoptic}
741 (?=\p{Assigned})\p{InGreekAndCoptic}
dbe420b4
JH
742
743But in this particular example, you probably really want
744
1bfb14c4 745 \p{GreekAndCoptic}
dbe420b4
JH
746
747which will match assigned characters known to be part of the Greek script.
29bdacb8 748
5ca1ac52 749Also see the Unicode::Regex::Set module, it does implement the full
8158862b
ST
750UTS#18 grouping, intersection, union, and removal (subtraction) syntax.
751
752[b] '+' for union, '-' for removal (set-difference), '&' for intersection
753(see L</"User-Defined Character Properties">)
754
755[c] Try the C<:crlf> layer (see L<PerlIO>).
5ca1ac52 756
c670e63a
KW
757[d] U+FFFF will currently generate a warning message if 'utf8' warnings are
758 enabled
237bad5b 759
776f8809
JH
760=item *
761
762Level 2 - Extended Unicode Support
763
8158862b 764 RL2.1 Canonical Equivalents - MISSING [10][11]
c670e63a 765 RL2.2 Default Grapheme Clusters - MISSING [12]
8158862b
ST
766 RL2.3 Default Word Boundaries - MISSING [14]
767 RL2.4 Default Loose Matches - MISSING [15]
768 RL2.5 Name Properties - MISSING [16]
769 RL2.6 Wildcard Properties - MISSING
770
771 [10] see UAX#15 "Unicode Normalization Forms"
772 [11] have Unicode::Normalize but not integrated to regexes
e1b711da 773 [12] have \X but we don't have a "Grapheme Cluster Mode"
8158862b
ST
774 [14] see UAX#29, Word Boundaries
775 [15] see UAX#21 "Case Mappings"
776 [16] have \N{...} but neither compute names of CJK Ideographs
777 and Hangul Syllables nor use a loose match [e]
778
779[e] C<\N{...}> allows namespaces (see L<charnames>).
776f8809
JH
780
781=item *
782
8158862b
ST
783Level 3 - Tailored Support
784
785 RL3.1 Tailored Punctuation - MISSING
786 RL3.2 Tailored Grapheme Clusters - MISSING [17][18]
787 RL3.3 Tailored Word Boundaries - MISSING
788 RL3.4 Tailored Loose Matches - MISSING
789 RL3.5 Tailored Ranges - MISSING
790 RL3.6 Context Matching - MISSING [19]
791 RL3.7 Incremental Matches - MISSING
792 ( RL3.8 Unicode Set Sharing )
793 RL3.9 Possible Match Sets - MISSING
794 RL3.10 Folded Matching - MISSING [20]
795 RL3.11 Submatchers - MISSING
796
797 [17] see UAX#10 "Unicode Collation Algorithms"
798 [18] have Unicode::Collate but not integrated to regexes
799 [19] have (?<=x) and (?=x), but look-aheads or look-behinds should see
800 outside of the target substring
801 [20] need insensitive matching for linguistic features other than case;
802 for example, hiragana to katakana, wide and narrow, simplified Han
803 to traditional Han (see UTR#30 "Character Foldings")
776f8809
JH
804
805=back
806
c349b1b9
JH
807=head2 Unicode Encodings
808
376d9008
JB
809Unicode characters are assigned to I<code points>, which are abstract
810numbers. To use these numbers, various encodings are needed.
c349b1b9
JH
811
812=over 4
813
c29a771d 814=item *
5cb3728c
RB
815
816UTF-8
c349b1b9 817
3e4dbfed 818UTF-8 is a variable-length (1 to 6 bytes, current character allocations
376d9008
JB
819require 4 bytes), byte-order independent encoding. For ASCII (and we
820really do mean 7-bit ASCII, not another 8-bit encoding), UTF-8 is
821transparent.
c349b1b9 822
8c007b5a 823The following table is from Unicode 3.2.
05632f9a 824
e1b711da 825 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
05632f9a 826
e1b711da
KW
827 U+0000..U+007F 00..7F
828 U+0080..U+07FF * C2..DF 80..BF
829 U+0800..U+0FFF E0 * A0..BF 80..BF
ec90690f
ST
830 U+1000..U+CFFF E1..EC 80..BF 80..BF
831 U+D000..U+D7FF ED 80..9F 80..BF
e1b711da 832 U+D800..U+DFFF +++++++ utf16 surrogates, not legal utf8 +++++++
ec90690f 833 U+E000..U+FFFF EE..EF 80..BF 80..BF
e1b711da
KW
834 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
835 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
836 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
837
838Note the gaps before several of the byte entries above marked by '*'. These are
839caused by legal UTF-8 avoiding non-shortest encodings: it is technically
840possible to UTF-8-encode a single code point in different ways, but that is
841explicitly forbidden, and the shortest possible encoding should always be used
842(and that is what Perl does).
37361303 843
376d9008 844Another way to look at it is via bits:
05632f9a
JH
845
846 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
847
848 0aaaaaaa 0aaaaaaa
849 00000bbbbbaaaaaa 110bbbbb 10aaaaaa
850 ccccbbbbbbaaaaaa 1110cccc 10bbbbbb 10aaaaaa
851 00000dddccccccbbbbbbaaaaaa 11110ddd 10cccccc 10bbbbbb 10aaaaaa
852
853As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 854leading bits of the start byte tell how many bytes there are in the
05632f9a
JH
855encoded character.
856
c29a771d 857=item *
5cb3728c
RB
858
859UTF-EBCDIC
dbe420b4 860
376d9008 861Like UTF-8 but EBCDIC-safe, in the way that UTF-8 is ASCII-safe.
dbe420b4 862
c29a771d 863=item *
5cb3728c 864
1e54db1a 865UTF-16, UTF-16BE, UTF-16LE, Surrogates, and BOMs (Byte Order Marks)
c349b1b9 866
1bfb14c4
JH
867The followings items are mostly for reference and general Unicode
868knowledge, Perl doesn't use these constructs internally.
dbe420b4 869
c349b1b9 870UTF-16 is a 2 or 4 byte encoding. The Unicode code points
1bfb14c4
JH
871C<U+0000..U+FFFF> are stored in a single 16-bit unit, and the code
872points C<U+10000..U+10FFFF> in two 16-bit units. The latter case is
c349b1b9
JH
873using I<surrogates>, the first 16-bit unit being the I<high
874surrogate>, and the second being the I<low surrogate>.
875
376d9008 876Surrogates are code points set aside to encode the C<U+10000..U+10FFFF>
c349b1b9 877range of Unicode code points in pairs of 16-bit units. The I<high
376d9008
JB
878surrogates> are the range C<U+D800..U+DBFF>, and the I<low surrogates>
879are the range C<U+DC00..U+DFFF>. The surrogate encoding is
c349b1b9
JH
880
881 $hi = ($uni - 0x10000) / 0x400 + 0xD800;
882 $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
883
884and the decoding is
885
1a3fa709 886 $uni = 0x10000 + ($hi - 0xD800) * 0x400 + ($lo - 0xDC00);
c349b1b9 887
feda178f 888If you try to generate surrogates (for example by using chr()), you
e1b711da 889will get a warning, if warnings are turned on, because those code
376d9008 890points are not valid for a Unicode character.
9466bab6 891
376d9008 892Because of the 16-bitness, UTF-16 is byte-order dependent. UTF-16
c349b1b9 893itself can be used for in-memory computations, but if storage or
376d9008
JB
894transfer is required either UTF-16BE (big-endian) or UTF-16LE
895(little-endian) encodings must be chosen.
c349b1b9
JH
896
897This introduces another problem: what if you just know that your data
376d9008
JB
898is UTF-16, but you don't know which endianness? Byte Order Marks, or
899BOMs, are a solution to this. A special character has been reserved
86bbd6d1 900in Unicode to function as a byte order marker: the character with the
376d9008 901code point C<U+FEFF> is the BOM.
042da322 902
c349b1b9 903The trick is that if you read a BOM, you will know the byte order,
376d9008
JB
904since if it was written on a big-endian platform, you will read the
905bytes C<0xFE 0xFF>, but if it was written on a little-endian platform,
906you will read the bytes C<0xFF 0xFE>. (And if the originating platform
907was writing in UTF-8, you will read the bytes C<0xEF 0xBB 0xBF>.)
042da322 908
86bbd6d1 909The way this trick works is that the character with the code point
376d9008
JB
910C<U+FFFE> is guaranteed not to be a valid Unicode character, so the
911sequence of bytes C<0xFF 0xFE> is unambiguously "BOM, represented in
1bfb14c4 912little-endian format" and cannot be C<U+FFFE>, represented in big-endian
e1b711da
KW
913format". (Actually, C<U+FFFE> is legal for use by your program, even for
914input/output, but better not use it if you need a BOM. But it is "illegal for
915interchange", so that an unsuspecting program won't get confused.)
c349b1b9 916
c29a771d 917=item *
5cb3728c 918
1e54db1a 919UTF-32, UTF-32BE, UTF-32LE
c349b1b9
JH
920
921The UTF-32 family is pretty much like the UTF-16 family, expect that
042da322 922the units are 32-bit, and therefore the surrogate scheme is not
376d9008
JB
923needed. The BOM signatures will be C<0x00 0x00 0xFE 0xFF> for BE and
924C<0xFF 0xFE 0x00 0x00> for LE.
c349b1b9 925
c29a771d 926=item *
5cb3728c
RB
927
928UCS-2, UCS-4
c349b1b9 929
86bbd6d1 930Encodings defined by the ISO 10646 standard. UCS-2 is a 16-bit
376d9008 931encoding. Unlike UTF-16, UCS-2 is not extensible beyond C<U+FFFF>,
339cfa0e
JH
932because it does not use surrogates. UCS-4 is a 32-bit encoding,
933functionally identical to UTF-32.
c349b1b9 934
c29a771d 935=item *
5cb3728c
RB
936
937UTF-7
c349b1b9 938
376d9008
JB
939A seven-bit safe (non-eight-bit) encoding, which is useful if the
940transport or storage is not eight-bit safe. Defined by RFC 2152.
c349b1b9 941
95a1a48b
JH
942=back
943
0d7c09bb
JH
944=head2 Security Implications of Unicode
945
e1b711da
KW
946Read L<Unicode Security Considerations|http://www.unicode.org/reports/tr36>.
947Also, note the following:
948
0d7c09bb
JH
949=over 4
950
951=item *
952
953Malformed UTF-8
bf0fa0b2
JH
954
955Unfortunately, the specification of UTF-8 leaves some room for
956interpretation of how many bytes of encoded output one should generate
376d9008
JB
957from one input Unicode character. Strictly speaking, the shortest
958possible sequence of UTF-8 bytes should be generated,
959because otherwise there is potential for an input buffer overflow at
feda178f 960the receiving end of a UTF-8 connection. Perl always generates the
e1b711da 961shortest length UTF-8, and with warnings on, Perl will warn about
376d9008
JB
962non-shortest length UTF-8 along with other malformations, such as the
963surrogates, which are not real Unicode code points.
bf0fa0b2 964
0d7c09bb
JH
965=item *
966
967Regular expressions behave slightly differently between byte data and
376d9008
JB
968character (Unicode) data. For example, the "word character" character
969class C<\w> will work differently depending on if data is eight-bit bytes
970or Unicode.
0d7c09bb 971
376d9008
JB
972In the first case, the set of C<\w> characters is either small--the
973default set of alphabetic characters, digits, and the "_"--or, if you
0d7c09bb
JH
974are using a locale (see L<perllocale>), the C<\w> might contain a few
975more letters according to your language and country.
976
376d9008 977In the second case, the C<\w> set of characters is much, much larger.
1bfb14c4
JH
978Most importantly, even in the set of the first 256 characters, it will
979probably match different characters: unlike most locales, which are
980specific to a language and country pair, Unicode classifies all the
981characters that are letters I<somewhere> as C<\w>. For example, your
982locale might not think that LATIN SMALL LETTER ETH is a letter (unless
983you happen to speak Icelandic), but Unicode does.
0d7c09bb 984
376d9008 985As discussed elsewhere, Perl has one foot (two hooves?) planted in
1bfb14c4
JH
986each of two worlds: the old world of bytes and the new world of
987characters, upgrading from bytes to characters when necessary.
376d9008
JB
988If your legacy code does not explicitly use Unicode, no automatic
989switch-over to characters should happen. Characters shouldn't get
1bfb14c4
JH
990downgraded to bytes, either. It is possible to accidentally mix bytes
991and characters, however (see L<perluniintro>), in which case C<\w> in
992regular expressions might start behaving differently. Review your
993code. Use warnings and the C<strict> pragma.
0d7c09bb
JH
994
995=back
996
c349b1b9
JH
997=head2 Unicode in Perl on EBCDIC
998
376d9008
JB
999The way Unicode is handled on EBCDIC platforms is still
1000experimental. On such platforms, references to UTF-8 encoding in this
1001document and elsewhere should be read as meaning the UTF-EBCDIC
1002specified in Unicode Technical Report 16, unless ASCII vs. EBCDIC issues
c349b1b9 1003are specifically discussed. There is no C<utfebcdic> pragma or
376d9008 1004":utfebcdic" layer; rather, "utf8" and ":utf8" are reused to mean
86bbd6d1
PN
1005the platform's "natural" 8-bit encoding of Unicode. See L<perlebcdic>
1006for more discussion of the issues.
c349b1b9 1007
b310b053
JH
1008=head2 Locales
1009
4616122b 1010Usually locale settings and Unicode do not affect each other, but
b310b053
JH
1011there are a couple of exceptions:
1012
1013=over 4
1014
1015=item *
1016
8aa8f774
JH
1017You can enable automatic UTF-8-ification of your standard file
1018handles, default C<open()> layer, and C<@ARGV> by using either
1019the C<-C> command line switch or the C<PERL_UNICODE> environment
1020variable, see L<perlrun> for the documentation of the C<-C> switch.
b310b053
JH
1021
1022=item *
1023
376d9008
JB
1024Perl tries really hard to work both with Unicode and the old
1025byte-oriented world. Most often this is nice, but sometimes Perl's
1026straddling of the proverbial fence causes problems.
b310b053
JH
1027
1028=back
1029
1aad1664
JH
1030=head2 When Unicode Does Not Happen
1031
1032While Perl does have extensive ways to input and output in Unicode,
1033and few other 'entry points' like the @ARGV which can be interpreted
1034as Unicode (UTF-8), there still are many places where Unicode (in some
1035encoding or another) could be given as arguments or received as
1036results, or both, but it is not.
1037
e1b711da
KW
1038The following are such interfaces. Also, see L</The "Unicode Bug">.
1039For all of these interfaces Perl
6cd4dd6c
JH
1040currently (as of 5.8.3) simply assumes byte strings both as arguments
1041and results, or UTF-8 strings if the C<encoding> pragma has been used.
1aad1664
JH
1042
1043One reason why Perl does not attempt to resolve the role of Unicode in
e1b711da 1044these cases is that the answers are highly dependent on the operating
1aad1664
JH
1045system and the file system(s). For example, whether filenames can be
1046in Unicode, and in exactly what kind of encoding, is not exactly a
1047portable concept. Similarly for the qx and system: how well will the
1048'command line interface' (and which of them?) handle Unicode?
1049
1050=over 4
1051
557a2462
RB
1052=item *
1053
51f494cc 1054chdir, chmod, chown, chroot, exec, link, lstat, mkdir,
1e8e8236 1055rename, rmdir, stat, symlink, truncate, unlink, utime, -X
557a2462
RB
1056
1057=item *
1058
1059%ENV
1060
1061=item *
1062
1063glob (aka the <*>)
1064
1065=item *
1aad1664 1066
557a2462 1067open, opendir, sysopen
1aad1664 1068
557a2462 1069=item *
1aad1664 1070
557a2462 1071qx (aka the backtick operator), system
1aad1664 1072
557a2462 1073=item *
1aad1664 1074
557a2462 1075readdir, readlink
1aad1664
JH
1076
1077=back
1078
e1b711da
KW
1079=head2 The "Unicode Bug"
1080
1081The term, the "Unicode bug" has been applied to an inconsistency with the
1082Unicode characters whose code points are in the Latin-1 Supplement block, that
1083is, between 128 and 255. Without a locale specified, unlike all other
1084characters or code points, these characters have very different semantics in
1085byte semantics versus character semantics.
1086
1087In character semantics they are interpreted as Unicode code points, which means
1088they have the same semantics as Latin-1 (ISO-8859-1).
1089
1090In byte semantics, they are considered to be unassigned characters, meaning
1091that the only semantics they have is their ordinal numbers, and that they are
1092not members of various character classes. None are considered to match C<\w>
1093for example, but all match C<\W>. (On EBCDIC platforms, the behavior may
1094be different from this, depending on the underlying C language library
1095functions.)
1096
1097The behavior is known to have effects on these areas:
1098
1099=over 4
1100
1101=item *
1102
1103Changing the case of a scalar, that is, using C<uc()>, C<ucfirst()>, C<lc()>,
1104and C<lcfirst()>, or C<\L>, C<\U>, C<\u> and C<\l> in regular expression
1105substitutions.
1106
1107=item *
1108
1109Using caseless (C</i>) regular expression matching
1110
1111=item *
1112
1113Matching a number of properties in regular expressions, such as C<\w>
1114
1115=item *
1116
1117User-defined case change mappings. You can create a C<ToUpper()> function, for
1118example, which overrides Perl's built-in case mappings. The scalar must be
1119encoded in utf8 for your function to actually be invoked.
1120
1121=back
1122
1123This behavior can lead to unexpected results in which a string's semantics
1124suddenly change if a code point above 255 is appended to or removed from it,
1125which changes the string's semantics from byte to character or vice versa. As
1126an example, consider the following program and its output:
1127
1128 $ perl -le'
1129 $s1 = "\xC2";
1130 $s2 = "\x{2660}";
1131 for ($s1, $s2, $s1.$s2) {
1132 print /\w/ || 0;
1133 }
1134 '
1135 0
1136 0
1137 1
1138
1139If there's no \w in s1 or in s2, why does their concatenation have one?
1140
1141This anomaly stems from Perl's attempt to not disturb older programs that
1142didn't use Unicode, and hence had no semantics for characters outside of the
1143ASCII range (except in a locale), along with Perl's desire to add Unicode
1144support seamlessly. The result wasn't seamless: these characters were
1145orphaned.
1146
1147Work is being done to correct this, but only some of it was complete in time
1148for the 5.12 release. What has been finished is the important part of the case
1149changing component. Due to concerns, and some evidence, that older code might
1150have come to rely on the existing behavior, the new behavior must be explicitly
1151enabled by the feature C<unicode_strings> in the L<feature> pragma, even though
1152no new syntax is involved.
1153
1154See L<perlfunc/lc> for details on how this pragma works in combination with
1155various others for casing. Even though the pragma only affects casing
1156operations in the 5.12 release, it is planned to have it affect all the
1157problematic behaviors in later releases: you can't have one without them all.
1158
1159In the meantime, a workaround is to always call utf8::upgrade($string), or to
1160use the standard modules L<Encode> or L<charnames>.
1161
1aad1664
JH
1162=head2 Forcing Unicode in Perl (Or Unforcing Unicode in Perl)
1163
e1b711da
KW
1164Sometimes (see L</"When Unicode Does Not Happen"> or L</The "Unicode Bug">)
1165there are situations where you simply need to force a byte
2bbc8d55
SP
1166string into UTF-8, or vice versa. The low-level calls
1167utf8::upgrade($bytestring) and utf8::downgrade($utf8string[, FAIL_OK]) are
1aad1664
JH
1168the answers.
1169
2bbc8d55
SP
1170Note that utf8::downgrade() can fail if the string contains characters
1171that don't fit into a byte.
1aad1664 1172
e1b711da
KW
1173Calling either function on a string that already is in the desired state is a
1174no-op.
1175
95a1a48b
JH
1176=head2 Using Unicode in XS
1177
3a2263fe
RGS
1178If you want to handle Perl Unicode in XS extensions, you may find the
1179following C APIs useful. See also L<perlguts/"Unicode Support"> for an
1180explanation about Unicode at the XS level, and L<perlapi> for the API
1181details.
95a1a48b
JH
1182
1183=over 4
1184
1185=item *
1186
1bfb14c4 1187C<DO_UTF8(sv)> returns true if the C<UTF8> flag is on and the bytes
2bbc8d55 1188pragma is not in effect. C<SvUTF8(sv)> returns true if the C<UTF8>
1bfb14c4
JH
1189flag is on; the bytes pragma is ignored. The C<UTF8> flag being on
1190does B<not> mean that there are any characters of code points greater
1191than 255 (or 127) in the scalar or that there are even any characters
1192in the scalar. What the C<UTF8> flag means is that the sequence of
1193octets in the representation of the scalar is the sequence of UTF-8
1194encoded code points of the characters of a string. The C<UTF8> flag
1195being off means that each octet in this representation encodes a
1196single character with code point 0..255 within the string. Perl's
1197Unicode model is not to use UTF-8 until it is absolutely necessary.
95a1a48b
JH
1198
1199=item *
1200
2bbc8d55 1201C<uvchr_to_utf8(buf, chr)> writes a Unicode character code point into
1bfb14c4 1202a buffer encoding the code point as UTF-8, and returns a pointer
2bbc8d55 1203pointing after the UTF-8 bytes. It works appropriately on EBCDIC machines.
95a1a48b
JH
1204
1205=item *
1206
2bbc8d55 1207C<utf8_to_uvchr(buf, lenp)> reads UTF-8 encoded bytes from a buffer and
376d9008 1208returns the Unicode character code point and, optionally, the length of
2bbc8d55 1209the UTF-8 byte sequence. It works appropriately on EBCDIC machines.
95a1a48b
JH
1210
1211=item *
1212
376d9008
JB
1213C<utf8_length(start, end)> returns the length of the UTF-8 encoded buffer
1214in characters. C<sv_len_utf8(sv)> returns the length of the UTF-8 encoded
95a1a48b
JH
1215scalar.
1216
1217=item *
1218
376d9008
JB
1219C<sv_utf8_upgrade(sv)> converts the string of the scalar to its UTF-8
1220encoded form. C<sv_utf8_downgrade(sv)> does the opposite, if
1221possible. C<sv_utf8_encode(sv)> is like sv_utf8_upgrade except that
1222it does not set the C<UTF8> flag. C<sv_utf8_decode()> does the
1223opposite of C<sv_utf8_encode()>. Note that none of these are to be
1224used as general-purpose encoding or decoding interfaces: C<use Encode>
1225for that. C<sv_utf8_upgrade()> is affected by the encoding pragma
1226but C<sv_utf8_downgrade()> is not (since the encoding pragma is
1227designed to be a one-way street).
95a1a48b
JH
1228
1229=item *
1230
376d9008 1231C<is_utf8_char(s)> returns true if the pointer points to a valid UTF-8
90f968e0 1232character.
95a1a48b
JH
1233
1234=item *
1235
376d9008 1236C<is_utf8_string(buf, len)> returns true if C<len> bytes of the buffer
95a1a48b
JH
1237are valid UTF-8.
1238
1239=item *
1240
376d9008
JB
1241C<UTF8SKIP(buf)> will return the number of bytes in the UTF-8 encoded
1242character in the buffer. C<UNISKIP(chr)> will return the number of bytes
1243required to UTF-8-encode the Unicode character code point. C<UTF8SKIP()>
90f968e0 1244is useful for example for iterating over the characters of a UTF-8
376d9008 1245encoded buffer; C<UNISKIP()> is useful, for example, in computing
90f968e0 1246the size required for a UTF-8 encoded buffer.
95a1a48b
JH
1247
1248=item *
1249
376d9008 1250C<utf8_distance(a, b)> will tell the distance in characters between the
95a1a48b
JH
1251two pointers pointing to the same UTF-8 encoded buffer.
1252
1253=item *
1254
2bbc8d55 1255C<utf8_hop(s, off)> will return a pointer to a UTF-8 encoded buffer
376d9008
JB
1256that is C<off> (positive or negative) Unicode characters displaced
1257from the UTF-8 buffer C<s>. Be careful not to overstep the buffer:
1258C<utf8_hop()> will merrily run off the end or the beginning of the
1259buffer if told to do so.
95a1a48b 1260
d2cc3551
JH
1261=item *
1262
376d9008
JB
1263C<pv_uni_display(dsv, spv, len, pvlim, flags)> and
1264C<sv_uni_display(dsv, ssv, pvlim, flags)> are useful for debugging the
1265output of Unicode strings and scalars. By default they are useful
1266only for debugging--they display B<all> characters as hexadecimal code
1bfb14c4
JH
1267points--but with the flags C<UNI_DISPLAY_ISPRINT>,
1268C<UNI_DISPLAY_BACKSLASH>, and C<UNI_DISPLAY_QQ> you can make the
1269output more readable.
d2cc3551
JH
1270
1271=item *
1272
2bbc8d55 1273C<ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2)> can be used to
376d9008
JB
1274compare two strings case-insensitively in Unicode. For case-sensitive
1275comparisons you can just use C<memEQ()> and C<memNE()> as usual.
d2cc3551 1276
c349b1b9
JH
1277=back
1278
95a1a48b
JH
1279For more information, see L<perlapi>, and F<utf8.c> and F<utf8.h>
1280in the Perl source code distribution.
1281
e1b711da
KW
1282=head2 Hacking Perl to work on earlier Unicode versions (for very serious hackers only)
1283
1284Perl by default comes with the latest supported Unicode version built in, but
1285you can change to use any earlier one.
1286
1287Download the files in the version of Unicode that you want from the Unicode web
1288site L<http://www.unicode.org>). These should replace the existing files in
1289C<\$Config{privlib}>/F<unicore>. (C<\%Config> is available from the Config
1290module.) Follow the instructions in F<README.perl> in that directory to change
1291some of their names, and then run F<make>.
1292
1293It is even possible to download them to a different directory, and then change
1294F<utf8_heavy.pl> in the directory C<\$Config{privlib}> to point to the new
1295directory, or maybe make a copy of that directory before making the change, and
1296using C<@INC> or the C<-I> run-time flag to switch between versions at will
1297(but because of caching, not in the middle of a process), but all this is
1298beyond the scope of these instructions.
1299
c29a771d
JH
1300=head1 BUGS
1301
376d9008 1302=head2 Interaction with Locales
7eabb34d 1303
376d9008
JB
1304Use of locales with Unicode data may lead to odd results. Currently,
1305Perl attempts to attach 8-bit locale info to characters in the range
13060..255, but this technique is demonstrably incorrect for locales that
1307use characters above that range when mapped into Unicode. Perl's
1308Unicode support will also tend to run slower. Use of locales with
1309Unicode is discouraged.
c29a771d 1310
e1b711da 1311=head2 Problems with characters in the C<Latin-1 Supplement> range
2bbc8d55 1312
e1b711da
KW
1313See L</The "Unicode Bug">
1314
1315=head2 Problems with case-insensitive regular expression matching
1316
1317There are problems with case-insensitive matches, including those involving
1318character classes (enclosed in [square brackets]), characters whose fold
1319is to multiple characters (such as the single character C<LATIN SMALL LIGATURE
1320FFL> matches case-insensitively with the 3-character string C<ffl>), and
1321characters in the C<Latin-1 Supplement>.
2bbc8d55 1322
376d9008 1323=head2 Interaction with Extensions
7eabb34d 1324
376d9008 1325When Perl exchanges data with an extension, the extension should be
2575c402 1326able to understand the UTF8 flag and act accordingly. If the
376d9008
JB
1327extension doesn't know about the flag, it's likely that the extension
1328will return incorrectly-flagged data.
7eabb34d
A
1329
1330So if you're working with Unicode data, consult the documentation of
1331every module you're using if there are any issues with Unicode data
1332exchange. If the documentation does not talk about Unicode at all,
a73d23f6 1333suspect the worst and probably look at the source to learn how the
376d9008 1334module is implemented. Modules written completely in Perl shouldn't
a73d23f6
RGS
1335cause problems. Modules that directly or indirectly access code written
1336in other programming languages are at risk.
7eabb34d 1337
376d9008 1338For affected functions, the simple strategy to avoid data corruption is
7eabb34d 1339to always make the encoding of the exchanged data explicit. Choose an
376d9008 1340encoding that you know the extension can handle. Convert arguments passed
7eabb34d
A
1341to the extensions to that encoding and convert results back from that
1342encoding. Write wrapper functions that do the conversions for you, so
1343you can later change the functions when the extension catches up.
1344
376d9008 1345To provide an example, let's say the popular Foo::Bar::escape_html
7eabb34d
A
1346function doesn't deal with Unicode data yet. The wrapper function
1347would convert the argument to raw UTF-8 and convert the result back to
376d9008 1348Perl's internal representation like so:
7eabb34d
A
1349
1350 sub my_escape_html ($) {
1351 my($what) = shift;
1352 return unless defined $what;
1353 Encode::decode_utf8(Foo::Bar::escape_html(Encode::encode_utf8($what)));
1354 }
1355
1356Sometimes, when the extension does not convert data but just stores
1357and retrieves them, you will be in a position to use the otherwise
1358dangerous Encode::_utf8_on() function. Let's say the popular
66b79f27 1359C<Foo::Bar> extension, written in C, provides a C<param> method that
7eabb34d
A
1360lets you store and retrieve data according to these prototypes:
1361
1362 $self->param($name, $value); # set a scalar
1363 $value = $self->param($name); # retrieve a scalar
1364
1365If it does not yet provide support for any encoding, one could write a
1366derived class with such a C<param> method:
1367
1368 sub param {
1369 my($self,$name,$value) = @_;
1370 utf8::upgrade($name); # make sure it is UTF-8 encoded
af55fc6a 1371 if (defined $value) {
7eabb34d
A
1372 utf8::upgrade($value); # make sure it is UTF-8 encoded
1373 return $self->SUPER::param($name,$value);
1374 } else {
1375 my $ret = $self->SUPER::param($name);
1376 Encode::_utf8_on($ret); # we know, it is UTF-8 encoded
1377 return $ret;
1378 }
1379 }
1380
a73d23f6
RGS
1381Some extensions provide filters on data entry/exit points, such as
1382DB_File::filter_store_key and family. Look out for such filters in
66b79f27 1383the documentation of your extensions, they can make the transition to
7eabb34d
A
1384Unicode data much easier.
1385
376d9008 1386=head2 Speed
7eabb34d 1387
c29a771d 1388Some functions are slower when working on UTF-8 encoded strings than
574c8022 1389on byte encoded strings. All functions that need to hop over
7c17141f
JH
1390characters such as length(), substr() or index(), or matching regular
1391expressions can work B<much> faster when the underlying data are
1392byte-encoded.
1393
1394In Perl 5.8.0 the slowness was often quite spectacular; in Perl 5.8.1
1395a caching scheme was introduced which will hopefully make the slowness
a104b433
JH
1396somewhat less spectacular, at least for some operations. In general,
1397operations with UTF-8 encoded strings are still slower. As an example,
1398the Unicode properties (character classes) like C<\p{Nd}> are known to
1399be quite a bit slower (5-20 times) than their simpler counterparts
1400like C<\d> (then again, there 268 Unicode characters matching C<Nd>
1401compared with the 10 ASCII characters matching C<d>).
666f95b9 1402
e1b711da
KW
1403=head2 Problems on EBCDIC platforms
1404
1405There are a number of known problems with Perl on EBCDIC platforms. If you
1406want to use Perl there, send email to perlbug@perl.org.
fe749c9a
KW
1407
1408In earlier versions, when byte and character data were concatenated,
1409the new string was sometimes created by
1410decoding the byte strings as I<ISO 8859-1 (Latin-1)>, even if the
1411old Unicode string used EBCDIC.
1412
1413If you find any of these, please report them as bugs.
1414
c8d992ba
A
1415=head2 Porting code from perl-5.6.X
1416
1417Perl 5.8 has a different Unicode model from 5.6. In 5.6 the programmer
1418was required to use the C<utf8> pragma to declare that a given scope
1419expected to deal with Unicode data and had to make sure that only
1420Unicode data were reaching that scope. If you have code that is
1421working with 5.6, you will need some of the following adjustments to
1422your code. The examples are written such that the code will continue
1423to work under 5.6, so you should be safe to try them out.
1424
1425=over 4
1426
1427=item *
1428
1429A filehandle that should read or write UTF-8
1430
1431 if ($] > 5.007) {
740d4bb2 1432 binmode $fh, ":encoding(utf8)";
c8d992ba
A
1433 }
1434
1435=item *
1436
1437A scalar that is going to be passed to some extension
1438
1439Be it Compress::Zlib, Apache::Request or any extension that has no
1440mention of Unicode in the manpage, you need to make sure that the
2575c402 1441UTF8 flag is stripped off. Note that at the time of this writing
c8d992ba
A
1442(October 2002) the mentioned modules are not UTF-8-aware. Please
1443check the documentation to verify if this is still true.
1444
1445 if ($] > 5.007) {
1446 require Encode;
1447 $val = Encode::encode_utf8($val); # make octets
1448 }
1449
1450=item *
1451
1452A scalar we got back from an extension
1453
1454If you believe the scalar comes back as UTF-8, you will most likely
2575c402 1455want the UTF8 flag restored:
c8d992ba
A
1456
1457 if ($] > 5.007) {
1458 require Encode;
1459 $val = Encode::decode_utf8($val);
1460 }
1461
1462=item *
1463
1464Same thing, if you are really sure it is UTF-8
1465
1466 if ($] > 5.007) {
1467 require Encode;
1468 Encode::_utf8_on($val);
1469 }
1470
1471=item *
1472
1473A wrapper for fetchrow_array and fetchrow_hashref
1474
1475When the database contains only UTF-8, a wrapper function or method is
1476a convenient way to replace all your fetchrow_array and
1477fetchrow_hashref calls. A wrapper function will also make it easier to
1478adapt to future enhancements in your database driver. Note that at the
1479time of this writing (October 2002), the DBI has no standardized way
1480to deal with UTF-8 data. Please check the documentation to verify if
1481that is still true.
1482
1483 sub fetchrow {
1484 my($self, $sth, $what) = @_; # $what is one of fetchrow_{array,hashref}
1485 if ($] < 5.007) {
1486 return $sth->$what;
1487 } else {
1488 require Encode;
1489 if (wantarray) {
1490 my @arr = $sth->$what;
1491 for (@arr) {
1492 defined && /[^\000-\177]/ && Encode::_utf8_on($_);
1493 }
1494 return @arr;
1495 } else {
1496 my $ret = $sth->$what;
1497 if (ref $ret) {
1498 for my $k (keys %$ret) {
1499 defined && /[^\000-\177]/ && Encode::_utf8_on($_) for $ret->{$k};
1500 }
1501 return $ret;
1502 } else {
1503 defined && /[^\000-\177]/ && Encode::_utf8_on($_) for $ret;
1504 return $ret;
1505 }
1506 }
1507 }
1508 }
1509
1510
1511=item *
1512
1513A large scalar that you know can only contain ASCII
1514
1515Scalars that contain only ASCII and are marked as UTF-8 are sometimes
1516a drag to your program. If you recognize such a situation, just remove
2575c402 1517the UTF8 flag:
c8d992ba
A
1518
1519 utf8::downgrade($val) if $] > 5.007;
1520
1521=back
1522
393fec97
GS
1523=head1 SEE ALSO
1524
51f494cc 1525L<perlunitut>, L<perluniintro>, L<perluniprops>, L<Encode>, L<open>, L<utf8>, L<bytes>,
a05d7ebb 1526L<perlretut>, L<perlvar/"${^UNICODE}">
51f494cc 1527L<http://www.unicode.org/reports/tr44>).
393fec97
GS
1528
1529=cut