1PERLUNICODE(1) Perl Programmers Reference Guide PERLUNICODE(1)
2
6 perlunicode - Unicode support in Perl
7
9 If you haven't already, before reading this document, you should become
10 familiar with both perlunitut and perluniintro.
11 Unicode aims to UNI-fy the en-CODE-ings of all the world's character
13 sets into a single Standard. For quite a few of the various coding
14 standards that existed when Unicode was first created, converting from
15 each to Unicode essentially meant adding a constant to each code point
16 in the original standard, and converting back meant just subtracting
17 that same constant. For ASCII and ISO-8859-1, the constant is 0. For
18 ISO-8859-5, (Cyrillic) the constant is 864; for Hebrew (ISO-8859-8),
19 it's 1488; Thai (ISO-8859-11), 3424; and so forth. This made it easy
20 to do the conversions, and facilitated the adoption of Unicode.
21 And it worked; nowadays, those legacy standards are rarely used. Most
23 everyone uses Unicode.
24 Unicode is a comprehensive standard. It specifies many things outside
26 the scope of Perl, such as how to display sequences of characters. For
27 a full discussion of all aspects of Unicode, see
28 <https://www.unicode.org>.
29 Important Caveats
31 Even though some of this section may not be understandable to you on
32 first reading, we think it's important enough to highlight some of the
33 gotchas before delving further, so here goes:
34 Unicode support is an extensive requirement. While Perl does not
36 implement the Unicode standard or the accompanying technical reports
37 from cover to cover, Perl does support many Unicode features.
38 Also, the use of Unicode may present security issues that aren't
40 obvious, see "Security Implications of Unicode" below.
41 Safest if you "use feature 'unicode_strings'"
43 In order to preserve backward compatibility, Perl does not turn on
44 full internal Unicode support unless the pragma
45 "use feature 'unicode_strings'" is specified. (This is
46 automatically selected if you "use 5.012" or higher.) Failure to
47 do this can trigger unexpected surprises. See "The "Unicode Bug""
48 below.
49 This pragma doesn't affect I/O. Nor does it change the internal
51 representation of strings, only their interpretation. There are
52 still several places where Unicode isn't fully supported, such as
53 in filenames.
54 Input and Output Layers
56 Use the ":encoding(...)" layer to read from and write to
57 filehandles using the specified encoding. (See open.)
58 You must convert your non-ASCII, non-UTF-8 Perl scripts to be UTF-8.
60 The encoding module has been deprecated since perl 5.18 and the
61 perl internals it requires have been removed with perl 5.26.
62 "use utf8" still needed to enable UTF-8 in scripts
64 If your Perl script is itself encoded in UTF-8, the "use utf8"
65 pragma must be explicitly included to enable recognition of that
66 (in string or regular expression literals, or in identifier names).
67 This is the only time when an explicit "use utf8" is needed. (See
68 utf8).
69 If a Perl script begins with the bytes that form the UTF-8 encoding
71 of the Unicode BYTE ORDER MARK ("BOM", see "Unicode Encodings"),
72 those bytes are completely ignored.
73 UTF-16 scripts autodetected
75 If a Perl script begins with the Unicode "BOM" (UTF-16LE,
76 UTF16-BE), or if the script looks like non-"BOM"-marked UTF-16 of
77 either endianness, Perl will correctly read in the script as the
78 appropriate Unicode encoding.
79 Byte and Character Semantics
81 Before Unicode, most encodings used 8 bits (a single byte) to encode
82 each character. Thus a character was a byte, and a byte was a
83 character, and there could be only 256 or fewer possible characters.
84 "Byte Semantics" in the title of this section refers to this behavior.
85 There was no need to distinguish between "Byte" and "Character".
86 Then along comes Unicode which has room for over a million characters
88 (and Perl allows for even more). This means that a character may
89 require more than a single byte to represent it, and so the two terms
90 are no longer equivalent. What matter are the characters as whole
91 entities, and not usually the bytes that comprise them. That's what
92 the term "Character Semantics" in the title of this section refers to.
93 Perl had to change internally to decouple "bytes" from "characters".
95 It is important that you too change your ideas, if you haven't already,
96 so that "byte" and "character" no longer mean the same thing in your
97 mind.
98 The basic building block of Perl strings has always been a "character".
100 The changes basically come down to that the implementation no longer
101 thinks that a character is always just a single byte.
102 There are various things to note:
104 • String handling functions, for the most part, continue to operate
106 in terms of characters. "length()", for example, returns the
107 number of characters in a string, just as before. But that number
108 no longer is necessarily the same as the number of bytes in the
109 string (there may be more bytes than characters). The other such
110 functions include "chop()", "chomp()", "substr()", "pos()",
111 "index()", "rindex()", "sort()", "sprintf()", and "write()".
112 The exceptions are:
114 • the bit-oriented "vec"
116 • the byte-oriented "pack"/"unpack" "C" format
120 However, the "W" specifier does operate on whole characters, as
122 does the "U" specifier.
123 • some operators that interact with the platform's operating
125 system
126 Operators dealing with filenames are examples.
128 • when the functions are called from within the scope of the
130 "use bytes" pragma
131 Likely, you should use this only for debugging anyway.
133 • Strings--including hash keys--and regular expression patterns may
135 contain characters that have ordinal values larger than 255.
136 If you use a Unicode editor to edit your program, Unicode
138 characters may occur directly within the literal strings in UTF-8
139 encoding, or UTF-16. (The former requires a "use utf8", the latter
140 may require a "BOM".)
141 "Creating Unicode" in perluniintro gives other ways to place non-
143 ASCII characters in your strings.
144 • The "chr()" and "ord()" functions work on whole characters.
146 • Regular expressions match whole characters. For example, "."
148 matches a whole character instead of only a single byte.
149 • The "tr///" operator translates whole characters. (Note that the
151 "tr///CU" functionality has been removed. For similar
152 functionality to that, see "pack('U0', ...)" and "pack('C0',
153 ...)").
154 • "scalar reverse()" reverses by character rather than by byte.
156 • The bit string operators, "& | ^ ~" and (starting in v5.22) "&. |.
158 ^. ~." can operate on bit strings encoded in UTF-8, but this can
159 give unexpected results if any of the strings contain code points
160 above 0xFF. Starting in v5.28, it is a fatal error to have such an
161 operand. Otherwise, the operation is performed on a non-UTF-8 copy
162 of the operand. If you're not sure about the encoding of a string,
163 downgrade it before using any of these operators; you can use
164 "utf8::utf8_downgrade()".
165 The bottom line is that Perl has always practiced "Character
167 Semantics", but with the advent of Unicode, that is now different than
168 "Byte Semantics".
169 ASCII Rules versus Unicode Rules
171 Before Unicode, when a character was a byte was a character, Perl knew
172 only about the 128 characters defined by ASCII, code points 0 through
173 127 (except for under "use locale"). That left the code points 128 to
174 255 as unassigned, and available for whatever use a program might want.
175 The only semantics they have is their ordinal numbers, and that they
176 are members of none of the non-negative character classes. None are
177 considered to match "\w" for example, but all match "\W".
178 Unicode, of course, assigns each of those code points a particular
180 meaning (along with ones above 255). To preserve backward
181 compatibility, Perl only uses the Unicode meanings when there is some
182 indication that Unicode is what is intended; otherwise the non-ASCII
183 code points remain treated as if they are unassigned.
184 Here are the ways that Perl knows that a string should be treated as
186 Unicode:
187 • Within the scope of "use utf8"
189 If the whole program is Unicode (signified by using 8-bit Unicode
191 Transformation Format), then all literal strings within it must be
192 Unicode.
193 • Within the scope of "use feature 'unicode_strings'"
195 This pragma was created so you can explicitly tell Perl that
197 operations executed within its scope are to use Unicode rules.
198 More operations are affected with newer perls. See "The "Unicode
199 Bug"".
200 • Within the scope of "use 5.012" or higher
202 This implicitly turns on "use feature 'unicode_strings'".
204 • Within the scope of "use locale 'not_characters'", or "use locale"
206 and the current locale is a UTF-8 locale.
207 The former is defined to imply Unicode handling; and the latter
209 indicates a Unicode locale, hence a Unicode interpretation of all
210 strings within it.
211 • When the string contains a Unicode-only code point
213 Perl has never accepted code points above 255 without them being
215 Unicode, so their use implies Unicode for the whole string.
216 • When the string contains a Unicode named code point "\N{...}"
218 The "\N{...}" construct explicitly refers to a Unicode code point,
220 even if it is one that is also in ASCII. Therefore the string
221 containing it must be Unicode.
222 • When the string has come from an external source marked as Unicode
224 The "-C" command line option can specify that certain inputs to the
226 program are Unicode, and the values of this can be read by your
227 Perl code, see "${^UNICODE}" in perlvar.
228 • When the string has been upgraded to UTF-8
230 The function "utf8::utf8_upgrade()" can be explicitly used to
232 permanently (unless a subsequent "utf8::utf8_downgrade()" is
233 called) cause a string to be treated as Unicode.
234 • There are additional methods for regular expression patterns
236 A pattern that is compiled with the "/u" or "/a" modifiers is
238 treated as Unicode (though there are some restrictions with "/a").
239 Under the "/d" and "/l" modifiers, there are several other
240 indications for Unicode; see "Character set modifiers" in perlre.
241 Note that all of the above are overridden within the scope of "use
243 bytes"; but you should be using this pragma only for debugging.
244 Note also that some interactions with the platform's operating system
246 never use Unicode rules.
247 When Unicode rules are in effect:
249 • Case translation operators use the Unicode case translation tables.
251 Note that "uc()", or "\U" in interpolated strings, translates to
253 uppercase, while "ucfirst", or "\u" in interpolated strings,
254 translates to titlecase in languages that make the distinction
255 (which is equivalent to uppercase in languages without the
256 distinction).
257 There is a CPAN module, "Unicode::Casing", which allows you to
259 define your own mappings to be used in "lc()", "lcfirst()", "uc()",
260 "ucfirst()", and "fc" (or their double-quoted string inlined
261 versions such as "\U"). (Prior to Perl 5.16, this functionality
262 was partially provided in the Perl core, but suffered from a number
263 of insurmountable drawbacks, so the CPAN module was written
264 instead.)
265 • Character classes in regular expressions match based on the
267 character properties specified in the Unicode properties database.
268 "\w" can be used to match a Japanese ideograph, for instance; and
270 "[[:digit:]]" a Bengali number.
271 • Named Unicode properties, scripts, and block ranges may be used
273 (like bracketed character classes) by using the "\p{}" "matches
274 property" construct and the "\P{}" negation, "doesn't match
275 property".
276 See "Unicode Character Properties" for more details.
278 You can define your own character properties and use them in the
280 regular expression with the "\p{}" or "\P{}" construct. See "User-
281 Defined Character Properties" for more details.
282 Extended Grapheme Clusters (Logical characters)
284 Consider a character, say "H". It could appear with various marks
285 around it, such as an acute accent, or a circumflex, or various hooks,
286 circles, arrows, etc., above, below, to one side or the other, etc.
287 There are many possibilities among the world's languages. The number
288 of combinations is astronomical, and if there were a character for each
289 combination, it would soon exhaust Unicode's more than a million
290 possible characters. So Unicode took a different approach: there is a
291 character for the base "H", and a character for each of the possible
292 marks, and these can be variously combined to get a final logical
293 character. So a logical character--what appears to be a single
294 character--can be a sequence of more than one individual characters.
295 The Unicode standard calls these "extended grapheme clusters" (which is
296 an improved version of the no-longer much used "grapheme cluster");
297 Perl furnishes the "\X" regular expression construct to match such
298 sequences in their entirety.
299 But Unicode's intent is to unify the existing character set standards
301 and practices, and several pre-existing standards have single
302 characters that mean the same thing as some of these combinations, like
303 ISO-8859-1, which has quite a few of them. For example, "LATIN CAPITAL
304 LETTER E WITH ACUTE" was already in this standard when Unicode came
305 along. Unicode therefore added it to its repertoire as that single
306 character. But this character is considered by Unicode to be
307 equivalent to the sequence consisting of the character "LATIN CAPITAL
308 LETTER E" followed by the character "COMBINING ACUTE ACCENT".
309 "LATIN CAPITAL LETTER E WITH ACUTE" is called a "pre-composed"
311 character, and its equivalence with the "E" and the "COMBINING ACCENT"
312 sequence is called canonical equivalence. All pre-composed characters
313 are said to have a decomposition (into the equivalent sequence), and
314 the decomposition type is also called canonical. A string may be
315 comprised as much as possible of precomposed characters, or it may be
316 comprised of entirely decomposed characters. Unicode calls these
317 respectively, "Normalization Form Composed" (NFC) and "Normalization
318 Form Decomposed". The "Unicode::Normalize" module contains functions
319 that convert between the two. A string may also have both composed
320 characters and decomposed characters; this module can be used to make
321 it all one or the other.
322 You may be presented with strings in any of these equivalent forms.
324 There is currently nothing in Perl 5 that ignores the differences. So
325 you'll have to specially handle it. The usual advice is to convert
326 your inputs to "NFD" before processing further.
327 For more detailed information, see <http://unicode.org/reports/tr15/>.
329 Unicode Character Properties
331 (The only time that Perl considers a sequence of individual code points
332 as a single logical character is in the "\X" construct, already
333 mentioned above. Therefore "character" in this discussion means a
334 single Unicode code point.)
335 Very nearly all Unicode character properties are accessible through
337 regular expressions by using the "\p{}" "matches property" construct
338 and the "\P{}" "doesn't match property" for its negation.
339 For instance, "\p{Uppercase}" matches any single character with the
341 Unicode "Uppercase" property, while "\p{L}" matches any character with
342 a "General_Category" of "L" (letter) property (see "General_Category"
343 below). Brackets are not required for single letter property names, so
344 "\p{L}" is equivalent to "\pL".
345 More formally, "\p{Uppercase}" matches any single character whose
347 Unicode "Uppercase" property value is "True", and "\P{Uppercase}"
348 matches any character whose "Uppercase" property value is "False", and
349 they could have been written as "\p{Uppercase=True}" and
350 "\p{Uppercase=False}", respectively.
351 This formality is needed when properties are not binary; that is, if
353 they can take on more values than just "True" and "False". For
354 example, the "Bidi_Class" property (see "Bidirectional Character Types"
355 below), can take on several different values, such as "Left", "Right",
356 "Whitespace", and others. To match these, one needs to specify both
357 the property name ("Bidi_Class"), AND the value being matched against
358 ("Left", "Right", etc.). This is done, as in the examples above, by
359 having the two components separated by an equal sign (or
360 interchangeably, a colon), like "\p{Bidi_Class: Left}".
361 All Unicode-defined character properties may be written in these
363 compound forms of "\p{property=value}" or "\p{property:value}", but
364 Perl provides some additional properties that are written only in the
365 single form, as well as single-form short-cuts for all binary
366 properties and certain others described below, in which you may omit
367 the property name and the equals or colon separator.
368 Most Unicode character properties have at least two synonyms (or
370 aliases if you prefer): a short one that is easier to type and a longer
371 one that is more descriptive and hence easier to understand. Thus the
372 "L" and "Letter" properties above are equivalent and can be used
373 interchangeably. Likewise, "Upper" is a synonym for "Uppercase", and
374 we could have written "\p{Uppercase}" equivalently as "\p{Upper}".
375 Also, there are typically various synonyms for the values the property
376 can be. For binary properties, "True" has 3 synonyms: "T", "Yes", and
377 "Y"; and "False" has correspondingly "F", "No", and "N". But be
378 careful. A short form of a value for one property may not mean the
379 same thing as the short form spelled the same for another. Thus, for
380 the "General_Category" property, "L" means "Letter", but for the
381 "Bidi_Class" property, "L" means "Left". A complete list of properties
382 and synonyms is in perluniprops.
383 Upper/lower case differences in property names and values are
385 irrelevant; thus "\p{Upper}" means the same thing as "\p{upper}" or
386 even "\p{UpPeR}". Similarly, you can add or subtract underscores
387 anywhere in the middle of a word, so that these are also equivalent to
388 "\p{U_p_p_e_r}". And white space is generally irrelevant adjacent to
389 non-word characters, such as the braces and the equals or colon
390 separators, so "\p{ Upper }" and "\p{ Upper_case : Y }" are
391 equivalent to these as well. In fact, white space and even hyphens can
392 usually be added or deleted anywhere. So even "\p{ Up-per case = Yes}"
393 is equivalent. All this is called "loose-matching" by Unicode. The
394 "name" property has some restrictions on this due to a few outlier
395 names. Full details are given in
396 <https://www.unicode.org/reports/tr44/tr44-24.html#UAX44-LM2>.
397 The few places where stricter matching is used is in the middle of
399 numbers, the "name" property, and in the Perl extension properties that
400 begin or end with an underscore. Stricter matching cares about white
401 space (except adjacent to non-word characters), hyphens, and non-
402 interior underscores.
403 You can also use negation in both "\p{}" and "\P{}" by introducing a
405 caret ("^") between the first brace and the property name: "\p{^Tamil}"
406 is equal to "\P{Tamil}".
407 Almost all properties are immune to case-insensitive matching. That
409 is, adding a "/i" regular expression modifier does not change what they
410 match. There are two sets that are affected. The first set is
411 "Uppercase_Letter", "Lowercase_Letter", and "Titlecase_Letter", all of
412 which match "Cased_Letter" under "/i" matching. And the second set is
413 "Uppercase", "Lowercase", and "Titlecase", all of which match "Cased"
414 under "/i" matching. This set also includes its subsets "PosixUpper"
415 and "PosixLower" both of which under "/i" match "PosixAlpha". (The
416 difference between these sets is that some things, such as Roman
417 numerals, come in both upper and lower case so they are "Cased", but
418 aren't considered letters, so they aren't "Cased_Letter"'s.)
419 See "Beyond Unicode code points" for special considerations when
421 matching Unicode properties against non-Unicode code points.
422 General_Category
424 Every Unicode character is assigned a general category, which is the
426 "most usual categorization of a character" (from
427 <https://www.unicode.org/reports/tr44>).
428 The compound way of writing these is like "\p{General_Category=Number}"
430 (short: "\p{gc:n}"). But Perl furnishes shortcuts in which everything
431 up through the equal or colon separator is omitted. So you can instead
432 just write "\pN".
433 Here are the short and long forms of the values the "General Category"
435 property can have:
436 Short Long
438 L Letter
440 LC, L& Cased_Letter (that is: [\p{Ll}\p{Lu}\p{Lt}])
441 Lu Uppercase_Letter
442 Ll Lowercase_Letter
443 Lt Titlecase_Letter
444 Lm Modifier_Letter
445 Lo Other_Letter
446 M Mark
448 Mn Nonspacing_Mark
449 Mc Spacing_Mark
450 Me Enclosing_Mark
451 N Number
453 Nd Decimal_Number (also Digit)
454 Nl Letter_Number
455 No Other_Number
456 P Punctuation (also Punct)
458 Pc Connector_Punctuation
459 Pd Dash_Punctuation
460 Ps Open_Punctuation
461 Pe Close_Punctuation
462 Pi Initial_Punctuation
463 (may behave like Ps or Pe depending on usage)
464 Pf Final_Punctuation
465 (may behave like Ps or Pe depending on usage)
466 Po Other_Punctuation
467 S Symbol
469 Sm Math_Symbol
470 Sc Currency_Symbol
471 Sk Modifier_Symbol
472 So Other_Symbol
473 Z Separator
475 Zs Space_Separator
476 Zl Line_Separator
477 Zp Paragraph_Separator
478 C Other
480 Cc Control (also Cntrl)
481 Cf Format
482 Cs Surrogate
483 Co Private_Use
484 Cn Unassigned
485 Single-letter properties match all characters in any of the two-letter
487 sub-properties starting with the same letter. "LC" and "L&" are
488 special: both are aliases for the set consisting of everything matched
489 by "Ll", "Lu", and "Lt".
490 Bidirectional Character Types
492 Because scripts differ in their directionality (Hebrew and Arabic are
494 written right to left, for example) Unicode supplies a "Bidi_Class"
495 property. Some of the values this property can have are:
496 Value Meaning
498 L Left-to-Right
500 LRE Left-to-Right Embedding
501 LRO Left-to-Right Override
502 R Right-to-Left
503 AL Arabic Letter
504 RLE Right-to-Left Embedding
505 RLO Right-to-Left Override
506 PDF Pop Directional Format
507 EN European Number
508 ES European Separator
509 ET European Terminator
510 AN Arabic Number
511 CS Common Separator
512 NSM Non-Spacing Mark
513 BN Boundary Neutral
514 B Paragraph Separator
515 S Segment Separator
516 WS Whitespace
517 ON Other Neutrals
518 This property is always written in the compound form. For example,
520 "\p{Bidi_Class:R}" matches characters that are normally written right
521 to left. Unlike the "General_Category" property, this property can
522 have more values added in a future Unicode release. Those listed above
523 comprised the complete set for many Unicode releases, but others were
524 added in Unicode 6.3; you can always find what the current ones are in
525 perluniprops. And <https://www.unicode.org/reports/tr9/> describes how
526 to use them.
527 Scripts
529 The world's languages are written in many different scripts. This
531 sentence (unless you're reading it in translation) is written in Latin,
532 while Russian is written in Cyrillic, and Greek is written in, well,
533 Greek; Japanese mainly in Hiragana or Katakana. There are many more.
534 The Unicode "Script" and "Script_Extensions" properties give what
536 script a given character is in. The "Script_Extensions" property is an
537 improved version of "Script", as demonstrated below. Either property
538 can be specified with the compound form like "\p{Script=Hebrew}"
539 (short: "\p{sc=hebr}"), or "\p{Script_Extensions=Javanese}" (short:
540 "\p{scx=java}"). In addition, Perl furnishes shortcuts for all
541 "Script_Extensions" property names. You can omit everything up through
542 the equals (or colon), and simply write "\p{Latin}" or "\P{Cyrillic}".
543 (This is not true for "Script", which is required to be written in the
544 compound form. Prior to Perl v5.26, the single form returned the plain
545 old "Script" version, but was changed because "Script_Extensions" gives
546 better results.)
547 The difference between these two properties involves characters that
549 are used in multiple scripts. For example the digits '0' through '9'
550 are used in many parts of the world. These are placed in a script
551 named "Common". Other characters are used in just a few scripts. For
552 example, the "KATAKANA-HIRAGANA DOUBLE HYPHEN" is used in both Japanese
553 scripts, Katakana and Hiragana, but nowhere else. The "Script"
554 property places all characters that are used in multiple scripts in the
555 "Common" script, while the "Script_Extensions" property places those
556 that are used in only a few scripts into each of those scripts; while
557 still using "Common" for those used in many scripts. Thus both these
558 match:
559 "0" =~ /\p{sc=Common}/ # Matches
561 "0" =~ /\p{scx=Common}/ # Matches
562 and only the first of these match:
564 "\N{KATAKANA-HIRAGANA DOUBLE HYPHEN}" =~ /\p{sc=Common} # Matches
566 "\N{KATAKANA-HIRAGANA DOUBLE HYPHEN}" =~ /\p{scx=Common} # No match
567 And only the last two of these match:
569 "\N{KATAKANA-HIRAGANA DOUBLE HYPHEN}" =~ /\p{sc=Hiragana} # No match
571 "\N{KATAKANA-HIRAGANA DOUBLE HYPHEN}" =~ /\p{sc=Katakana} # No match
572 "\N{KATAKANA-HIRAGANA DOUBLE HYPHEN}" =~ /\p{scx=Hiragana} # Matches
573 "\N{KATAKANA-HIRAGANA DOUBLE HYPHEN}" =~ /\p{scx=Katakana} # Matches
574 "Script_Extensions" is thus an improved "Script", in which there are
576 fewer characters in the "Common" script, and correspondingly more in
577 other scripts. It is new in Unicode version 6.0, and its data are
578 likely to change significantly in later releases, as things get sorted
579 out. New code should probably be using "Script_Extensions" and not
580 plain "Script". If you compile perl with a Unicode release that
581 doesn't have "Script_Extensions", the single form Perl extensions will
582 instead refer to the plain "Script" property. If you compile with a
583 version of Unicode that doesn't have the "Script" property, these
584 extensions will not be defined at all.
585 (Actually, besides "Common", the "Inherited" script, contains
587 characters that are used in multiple scripts. These are modifier
588 characters which inherit the script value of the controlling character.
589 Some of these are used in many scripts, and so go into "Inherited" in
590 both "Script" and "Script_Extensions". Others are used in just a few
591 scripts, so are in "Inherited" in "Script", but not in
592 "Script_Extensions".)
593 It is worth stressing that there are several different sets of digits
595 in Unicode that are equivalent to 0-9 and are matchable by "\d" in a
596 regular expression. If they are used in a single language only, they
597 are in that language's "Script" and "Script_Extensions". If they are
598 used in more than one script, they will be in "sc=Common", but only if
599 they are used in many scripts should they be in "scx=Common".
600 The explanation above has omitted some detail; refer to UAX#24 "Unicode
602 Script Property": <https://www.unicode.org/reports/tr24>.
603 A complete list of scripts and their shortcuts is in perluniprops.
605 Use of the "Is" Prefix
607 For backward compatibility (with ancient Perl 5.6), all properties
609 writable without using the compound form mentioned so far may have "Is"
610 or "Is_" prepended to their name, so "\P{Is_Lu}", for example, is equal
611 to "\P{Lu}", and "\p{IsScript:Arabic}" is equal to "\p{Arabic}".
612 Blocks
614 In addition to scripts, Unicode also defines blocks of characters. The
616 difference between scripts and blocks is that the concept of scripts is
617 closer to natural languages, while the concept of blocks is more of an
618 artificial grouping based on groups of Unicode characters with
619 consecutive ordinal values. For example, the "Basic Latin" block is all
620 the characters whose ordinals are between 0 and 127, inclusive; in
621 other words, the ASCII characters. The "Latin" script contains some
622 letters from this as well as several other blocks, like "Latin-1
623 Supplement", "Latin Extended-A", etc., but it does not contain all the
624 characters from those blocks. It does not, for example, contain the
625 digits 0-9, because those digits are shared across many scripts, and
626 hence are in the "Common" script.
627 For more about scripts versus blocks, see UAX#24 "Unicode Script
629 Property": <https://www.unicode.org/reports/tr24>
630 The "Script_Extensions" or "Script" properties are likely to be the
632 ones you want to use when processing natural language; the "Block"
633 property may occasionally be useful in working with the nuts and bolts
634 of Unicode.
635 Block names are matched in the compound form, like "\p{Block: Arrows}"
637 or "\p{Blk=Hebrew}". Unlike most other properties, only a few block
638 names have a Unicode-defined short name.
639 Perl also defines single form synonyms for the block property in cases
641 where these do not conflict with something else. But don't use any of
642 these, because they are unstable. Since these are Perl extensions,
643 they are subordinate to official Unicode property names; Unicode
644 doesn't know nor care about Perl's extensions. It may happen that a
645 name that currently means the Perl extension will later be changed
646 without warning to mean a different Unicode property in a future
647 version of the perl interpreter that uses a later Unicode release, and
648 your code would no longer work. The extensions are mentioned here for
649 completeness: Take the block name and prefix it with one of: "In" (for
650 example "\p{Blk=Arrows}" can currently be written as "\p{In_Arrows}");
651 or sometimes "Is" (like "\p{Is_Arrows}"); or sometimes no prefix at all
652 ("\p{Arrows}"). As of this writing (Unicode 9.0) there are no
653 conflicts with using the "In_" prefix, but there are plenty with the
654 other two forms. For example, "\p{Is_Hebrew}" and "\p{Hebrew}" mean
655 "\p{Script_Extensions=Hebrew}" which is NOT the same thing as
656 "\p{Blk=Hebrew}". Our advice used to be to use the "In_" prefix as a
657 single form way of specifying a block. But Unicode 8.0 added
658 properties whose names begin with "In", and it's now clear that it's
659 only luck that's so far prevented a conflict. Using "In" is only
660 marginally less typing than "Blk:", and the latter's meaning is clearer
661 anyway, and guaranteed to never conflict. So don't take chances. Use
662 "\p{Blk=foo}" for new code. And be sure that block is what you really
663 really want to do. In most cases scripts are what you want instead.
664 A complete list of blocks is in perluniprops.
666 Other Properties
668 There are many more properties than the very basic ones described here.
670 A complete list is in perluniprops.
671 Unicode defines all its properties in the compound form, so all single-
673 form properties are Perl extensions. Most of these are just synonyms
674 for the Unicode ones, but some are genuine extensions, including
675 several that are in the compound form. And quite a few of these are
676 actually recommended by Unicode (in
677 <https://www.unicode.org/reports/tr18>).
678 This section gives some details on all extensions that aren't just
680 synonyms for compound-form Unicode properties (for those properties,
681 you'll have to refer to the Unicode Standard
682 <https://www.unicode.org/reports/tr44>.
683 "\p{All}"
685 This matches every possible code point. It is equivalent to
686 "qr/./s". Unlike all the other non-user-defined "\p{}" property
687 matches, no warning is ever generated if this is property is
688 matched against a non-Unicode code point (see "Beyond Unicode code
689 points" below).
690 "\p{Alnum}"
692 This matches any "\p{Alphabetic}" or "\p{Decimal_Number}"
693 character.
694 "\p{Any}"
696 This matches any of the 1_114_112 Unicode code points. It is a
697 synonym for "\p{Unicode}".
698 "\p{ASCII}"
700 This matches any of the 128 characters in the US-ASCII character
701 set, which is a subset of Unicode.
702 "\p{Assigned}"
704 This matches any assigned code point; that is, any code point whose
705 general category is not "Unassigned" (or equivalently, not "Cn").
706 "\p{Blank}"
708 This is the same as "\h" and "\p{HorizSpace}": A character that
709 changes the spacing horizontally.
710 "\p{Decomposition_Type: Non_Canonical}" (Short: "\p{Dt=NonCanon}")
712 Matches a character that has a non-canonical decomposition.
713 The "Extended Grapheme Clusters (Logical characters)" section above
715 talked about canonical decompositions. However, many more
716 characters have a different type of decomposition, a "compatible"
717 or "non-canonical" decomposition. The sequences that form these
718 decompositions are not considered canonically equivalent to the
719 pre-composed character. An example is the "SUPERSCRIPT ONE". It
720 is somewhat like a regular digit 1, but not exactly; its
721 decomposition into the digit 1 is called a "compatible"
722 decomposition, specifically a "super" decomposition. There are
723 several such compatibility decompositions (see
724 <https://www.unicode.org/reports/tr44>), including one called
725 "compat", which means some miscellaneous type of decomposition that
726 doesn't fit into the other decomposition categories that Unicode
727 has chosen.
728 Note that most Unicode characters don't have a decomposition, so
730 their decomposition type is "None".
731 For your convenience, Perl has added the "Non_Canonical"
733 decomposition type to mean any of the several compatibility
734 decompositions.
735 "\p{Graph}"
737 Matches any character that is graphic. Theoretically, this means a
738 character that on a printer would cause ink to be used.
739 "\p{HorizSpace}"
741 This is the same as "\h" and "\p{Blank}": a character that changes
742 the spacing horizontally.
743 "\p{In=*}"
745 This is a synonym for "\p{Present_In=*}"
746 "\p{PerlSpace}"
748 This is the same as "\s", restricted to ASCII, namely "[ \f\n\r\t]"
749 and starting in Perl v5.18, a vertical tab.
750 Mnemonic: Perl's (original) space
752 "\p{PerlWord}"
754 This is the same as "\w", restricted to ASCII, namely
755 "[A-Za-z0-9_]"
756 Mnemonic: Perl's (original) word.
758 "\p{Posix...}"
760 There are several of these, which are equivalents, using the "\p{}"
761 notation, for Posix classes and are described in "POSIX Character
762 Classes" in perlrecharclass.
763 "\p{Present_In: *}" (Short: "\p{In=*}")
765 This property is used when you need to know in what Unicode
766 version(s) a character is.
767 The "*" above stands for some Unicode version number, such as 1.1
769 or 12.0; or the "*" can also be "Unassigned". This property will
770 match the code points whose final disposition has been settled as
771 of the Unicode release given by the version number; "\p{Present_In:
772 Unassigned}" will match those code points whose meaning has yet to
773 be assigned.
774 For example, "U+0041" "LATIN CAPITAL LETTER A" was present in the
776 very first Unicode release available, which is 1.1, so this
777 property is true for all valid "*" versions. On the other hand,
778 "U+1EFF" was not assigned until version 5.1 when it became "LATIN
779 SMALL LETTER Y WITH LOOP", so the only "*" that would match it are
780 5.1, 5.2, and later.
781 Unicode furnishes the "Age" property from which this is derived.
783 The problem with Age is that a strict interpretation of it (which
784 Perl takes) has it matching the precise release a code point's
785 meaning is introduced in. Thus "U+0041" would match only 1.1; and
786 "U+1EFF" only 5.1. This is not usually what you want.
787 Some non-Perl implementations of the Age property may change its
789 meaning to be the same as the Perl "Present_In" property; just be
790 aware of that.
791 Another confusion with both these properties is that the definition
793 is not that the code point has been assigned, but that the meaning
794 of the code point has been determined. This is because 66 code
795 points will always be unassigned, and so the "Age" for them is the
796 Unicode version in which the decision to make them so was made.
797 For example, "U+FDD0" is to be permanently unassigned to a
798 character, and the decision to do that was made in version 3.1, so
799 "\p{Age=3.1}" matches this character, as also does "\p{Present_In:
800 3.1}" and up.
801 "\p{Print}"
803 This matches any character that is graphical or blank, except
804 controls.
805 "\p{SpacePerl}"
807 This is the same as "\s", including beyond ASCII.
808 Mnemonic: Space, as modified by Perl. (It doesn't include the
810 vertical tab until v5.18, which both the Posix standard and Unicode
811 consider white space.)
812 "\p{Title}" and "\p{Titlecase}"
814 Under case-sensitive matching, these both match the same code
815 points as "\p{General Category=Titlecase_Letter}" ("\p{gc=lt}").
816 The difference is that under "/i" caseless matching, these match
817 the same as "\p{Cased}", whereas "\p{gc=lt}" matches
818 "\p{Cased_Letter").
819 "\p{Unicode}"
821 This matches any of the 1_114_112 Unicode code points. "\p{Any}".
822 "\p{VertSpace}"
824 This is the same as "\v": A character that changes the spacing
825 vertically.
826 "\p{Word}"
828 This is the same as "\w", including over 100_000 characters beyond
829 ASCII.
830 "\p{XPosix...}"
832 There are several of these, which are the standard Posix classes
833 extended to the full Unicode range. They are described in "POSIX
834 Character Classes" in perlrecharclass.
835 Comparison of "\N{...}" and "\p{name=...}"
837 Starting in Perl 5.32, you can specify a character by its name in
838 regular expression patterns using "\p{name=...}". This is in addition
839 to the longstanding method of using "\N{...}". The following
840 summarizes the differences between these two:
841 \N{...} \p{Name=...}
843 can interpolate only with eval yes [1]
844 custom names yes no [2]
845 name aliases yes yes [3]
846 named sequences yes yes [4]
847 name value parsing exact Unicode loose [5]
848 [1] The ability to interpolate means you can do something like
850 qr/\p{na=latin capital letter $which}/
852 and specify $which elsewhere.
854 [2] You can create your own names for characters, and override official
856 ones when using "\N{...}". See "CUSTOM ALIASES" in charnames.
857 [3] Some characters have multiple names (synonyms).
859 [4] Some particular sequences of characters are given a single name, in
861 addition to their individual ones.
862 [5] Exact name value matching means you have to specify case, hyphens,
864 underscores, and spaces precisely in the name you want. Loose
865 matching follows the Unicode rules
866 <https://www.unicode.org/reports/tr44/tr44-24.html#UAX44-LM2>,
867 where these are mostly irrelevant. Except for a few outlier
868 character names, these are the same rules as are already used for
869 any other "\p{...}" property.
870 Wildcards in Property Values
872 Starting in Perl 5.30, it is possible to do something like this:
873 qr!\p{numeric_value=/\A[0-5]\z/}!
875 or, by abbreviating and adding "/x",
877 qr! \p{nv= /(?x) \A [0-5] \z / }!
879 This matches all code points whose numeric value is one of 0, 1, 2, 3,
881 4, or 5. This particular example could instead have been written as
882 qr! \A [ \p{nv=0}\p{nv=1}\p{nv=2}\p{nv=3}\p{nv=4}\p{nv=5} ] \z !xx
884 in earlier perls, so in this case this feature just makes things easier
886 and shorter to write. If we hadn't included the "\A" and "\z", these
887 would have matched things like "1/2" because that contains a 1 (as well
888 as a 2). As written, it matches things like subscripts that have these
889 numeric values. If we only wanted the decimal digits with those
890 numeric values, we could say,
891 qr! (?[ \d & \p{nv=/[0-5]/ ]) }!x
893 The "\d" gets rid of needing to anchor the pattern, since it forces the
895 result to only match "[0-9]", and the "[0-5]" further restricts it.
896 The text in the above examples enclosed between the "/" characters can
898 be just about any regular expression. It is independent of the main
899 pattern, so doesn't share any capturing groups, etc. The delimiters
900 for it must be ASCII punctuation, but it may NOT be delimited by "{",
901 nor "}" nor contain a literal "}", as that delimits the end of the
902 enclosing "\p{}". Like any pattern, certain other delimiters are
903 terminated by their mirror images. These are "(", ""["", and "<". If
904 the delimiter is any of "-", "_", "+", or "\", or is the same delimiter
905 as is used for the enclosing pattern, it must be preceded by a
906 backslash escape, both fore and aft.
907 Beware of using "$" to indicate to match the end of the string. It can
909 too easily be interpreted as being a punctuation variable, like $/.
910 No modifiers may follow the final delimiter. Instead, use
912 "(?adlupimnsx-imnsx)" in perlre and/or "(?adluimnsx-imnsx:pattern)" in
913 perlre to specify modifiers. However, certain modifiers are illegal in
914 your wildcard subpattern. The only character set modifier specifiable
915 is "/aa"; any other character set, and "-m", and "p", and "s" are all
916 illegal. Specifying modifiers like "qr/.../gc" that aren't legal in
917 the "(?...)" notation normally raise a warning, but with wildcard
918 subpatterns, their use is an error. The "m" modifier is ineffective;
919 everything that matches will be a single line.
920 By default, your pattern is matched case-insensitively, as if "/i" had
922 been specified. You can change this by saying "(?-i)" in your pattern.
923 There are also certain operations that are illegal. You can't nest
925 "\p{...}" and "\P{...}" calls within a wildcard subpattern, and "\G"
926 doesn't make sense, so is also prohibited.
927 And the "*" quantifier (or its equivalent "(0,}") is illegal.
929 This feature is not available when the left-hand side is prefixed by
931 "Is_", nor for any form that is marked as "Discouraged" in
932 "Discouraged" in perluniprops.
933 This experimental feature has been added to begin to implement
935 <https://www.unicode.org/reports/tr18/#Wildcard_Properties>. Using it
936 will raise a (default-on) warning in the
937 "experimental::uniprop_wildcards" category. We reserve the right to
938 change its operation as we gain experience.
939 Your subpattern can be just about anything, but for it to have some
941 utility, it should match when called with either or both of a) the full
942 name of the property value with underscores (and/or spaces in the Block
943 property) and some things uppercase; or b) the property value in all
944 lowercase with spaces and underscores squeezed out. For example,
945 qr!\p{Blk=/Old I.*/}!
947 qr!\p{Blk=/oldi.*/}!
948 would match the same things.
950 Another example that shows that within "\p{...}", "/x" isn't needed to
952 have spaces:
953 qr!\p{scx= /Hebrew|Greek/ }!
955 To be safe, we should have anchored the above example, to prevent
957 matches for something like "Hebrew_Braille", but there aren't any
958 script names like that, so far. A warning is issued if none of the
959 legal values for a property are matched by your pattern. It's likely
960 that a future release will raise a warning if your pattern ends up
961 causing every possible code point to match.
962 Starting in 5.32, the Name, Name Aliases, and Named Sequences
964 properties are allowed to be matched. They are considered to be a
965 single combination property, just as has long been the case for "\N{}".
966 Loose matching doesn't work in exactly the same way for these as it
967 does for the values of other properties. The rules are given in
968 <https://www.unicode.org/reports/tr44/tr44-24.html#UAX44-LM2>. As a
969 result, Perl doesn't try loose matching for you, like it does in other
970 properties. All letters in names are uppercase, but you can add "(?i)"
971 to your subpattern to ignore case. If you're uncertain where a blank
972 is, you can use " ?" in your subpattern. No character name contains an
973 underscore, so don't bother trying to match one. The use of hyphens is
974 particularly problematic; refer to the above link. But note that, as
975 of Unicode 13.0, the only script in modern usage which has weirdnesses
976 with these is Tibetan; also the two Korean characters U+116C HANGUL
977 JUNGSEONG OE and U+1180 HANGUL JUNGSEONG O-E. Unicode makes no
978 promises to not add hyphen-problematic names in the future.
979 Using wildcards on these is resource intensive, given the hundreds of
981 thousands of legal names that must be checked against.
982 An example of using Name property wildcards is
984 qr!\p{name=/(SMILING|GRINNING) FACE/}!
986 Another is
988 qr/(?[ \p{name=\/CJK\/} - \p{ideographic} ])/
990 which is the 200-ish (as of Unicode 13.0) CJK characters that aren't
992 ideographs.
993 There are certain properties that wildcard subpatterns don't currently
995 work with. These are:
996 Bidi Mirroring Glyph
998 Bidi Paired Bracket
999 Case Folding
1000 Decomposition Mapping
1001 Equivalent Unified Ideograph
1002 Lowercase Mapping
1003 NFKC Case Fold
1004 Titlecase Mapping
1005 Uppercase Mapping
1006 Nor is the "@unicode_property@" form implemented.
1008 Here's a complete example of matching IPV4 internet protocol addresses
1010 in any (single) script
1011 no warnings 'experimental::regex_sets';
1013 no warnings 'experimental::uniprop_wildcards';
1014 # Can match a substring, so this intermediate regex needs to have
1016 # context or anchoring in its final use. Using nt=de yields decimal
1017 # digits. When specifying a subset of these, we must include \d to
1018 # prevent things like U+00B2 SUPERSCRIPT TWO from matching
1019 my $zero_through_255 =
1020 qr/ \b (*sr: # All from same sript
1021 (?[ \p{nv=0} & \d ])* # Optional leading zeros
1022 ( # Then one of:
1023 \d{1,2} # 0 - 99
1024 | (?[ \p{nv=1} & \d ]) \d{2} # 100 - 199
1025 | (?[ \p{nv=2} & \d ])
1026 ( (?[ \p{nv=:[0-4]:} & \d ]) \d # 200 - 249
1027 | (?[ \p{nv=5} & \d ])
1028 (?[ \p{nv=:[0-5]:} & \d ]) # 250 - 255
1029 )
1030 )
1031 )
1032 \b
1033 /x;
1034 my $ipv4 = qr/ \A (*sr: $zero_through_255
1036 (?: [.] $zero_through_255 ) {3}
1037 )
1038 \z
1039 /x;
1040 User-Defined Character Properties
1042 You can define your own binary character properties by defining
1043 subroutines whose names begin with "In" or "Is". (The experimental
1044 feature "(?[ ])" in perlre provides an alternative which allows more
1045 complex definitions.) The subroutines can be defined in any package.
1046 They override any Unicode properties expressed as the same names. The
1047 user-defined properties can be used in the regular expression "\p{}"
1048 and "\P{}" constructs; if you are using a user-defined property from a
1049 package other than the one you are in, you must specify its package in
1050 the "\p{}" or "\P{}" construct.
1051 # assuming property Is_Foreign defined in Lang::
1053 package main; # property package name required
1054 if ($txt =~ /\p{Lang::IsForeign}+/) { ... }
1055 package Lang; # property package name not required
1057 if ($txt =~ /\p{IsForeign}+/) { ... }
1058 Note that the effect is compile-time and immutable once defined.
1060 However, the subroutines are passed a single parameter, which is 0 if
1061 case-sensitive matching is in effect and non-zero if caseless matching
1062 is in effect. The subroutine may return different values depending on
1063 the value of the flag, and one set of values will immutably be in
1064 effect for all case-sensitive matches, and the other set for all case-
1065 insensitive matches.
1066 Note that if the regular expression is tainted, then Perl will die
1068 rather than calling the subroutine when the name of the subroutine is
1069 determined by the tainted data.
1070 The subroutines must return a specially-formatted string, with one or
1072 more newline-separated lines. Each line must be one of the following:
1073 • A single hexadecimal number denoting a code point to include.
1075 • Two hexadecimal numbers separated by horizontal whitespace (space
1077 or tabular characters) denoting a range of code points to include.
1078 The second number must not be smaller than the first.
1079 • Something to include, prefixed by "+": a built-in character
1081 property (prefixed by "utf8::") or a fully qualified (including
1082 package name) user-defined character property, to represent all the
1083 characters in that property; two hexadecimal code points for a
1084 range; or a single hexadecimal code point.
1085 • Something to exclude, prefixed by "-": an existing character
1087 property (prefixed by "utf8::") or a fully qualified (including
1088 package name) user-defined character property, to represent all the
1089 characters in that property; two hexadecimal code points for a
1090 range; or a single hexadecimal code point.
1091 • Something to negate, prefixed "!": an existing character property
1093 (prefixed by "utf8::") or a fully qualified (including package
1094 name) user-defined character property, to represent all the
1095 characters in that property; two hexadecimal code points for a
1096 range; or a single hexadecimal code point.
1097 • Something to intersect with, prefixed by "&": an existing character
1099 property (prefixed by "utf8::") or a fully qualified (including
1100 package name) user-defined character property, for all the
1101 characters except the characters in the property; two hexadecimal
1102 code points for a range; or a single hexadecimal code point.
1103 For example, to define a property that covers both the Japanese
1105 syllabaries (hiragana and katakana), you can define
1106 sub InKana {
1108 return <<END;
1109 3040\t309F
1110 30A0\t30FF
1111 END
1112 }
1113 Imagine that the here-doc end marker is at the beginning of the line.
1115 Now you can use "\p{InKana}" and "\P{InKana}".
1116 You could also have used the existing block property names:
1118 sub InKana {
1120 return <<'END';
1121 +utf8::InHiragana
1122 +utf8::InKatakana
1123 END
1124 }
1125 Suppose you wanted to match only the allocated characters, not the raw
1127 block ranges: in other words, you want to remove the unassigned
1128 characters:
1129 sub InKana {
1131 return <<'END';
1132 +utf8::InHiragana
1133 +utf8::InKatakana
1134 -utf8::IsCn
1135 END
1136 }
1137 The negation is useful for defining (surprise!) negated classes.
1139 sub InNotKana {
1141 return <<'END';
1142 !utf8::InHiragana
1143 -utf8::InKatakana
1144 +utf8::IsCn
1145 END
1146 }
1147 This will match all non-Unicode code points, since every one of them is
1149 not in Kana. You can use intersection to exclude these, if desired, as
1150 this modified example shows:
1151 sub InNotKana {
1153 return <<'END';
1154 !utf8::InHiragana
1155 -utf8::InKatakana
1156 +utf8::IsCn
1157 &utf8::Any
1158 END
1159 }
1160 &utf8::Any must be the last line in the definition.
1162 Intersection is used generally for getting the common characters
1164 matched by two (or more) classes. It's important to remember not to
1165 use "&" for the first set; that would be intersecting with nothing,
1166 resulting in an empty set. (Similarly using "-" for the first set does
1167 nothing).
1168 Unlike non-user-defined "\p{}" property matches, no warning is ever
1170 generated if these properties are matched against a non-Unicode code
1171 point (see "Beyond Unicode code points" below).
1172 User-Defined Case Mappings (for serious hackers only)
1174 This feature has been removed as of Perl 5.16. The CPAN module
1175 "Unicode::Casing" provides better functionality without the drawbacks
1176 that this feature had. If you are using a Perl earlier than 5.16, this
1177 feature was most fully documented in the 5.14 version of this pod:
1178 <http://perldoc.perl.org/5.14.0/perlunicode.html#User-Defined-Case-Mappings-%28for-serious-hackers-only%29>
1179 Character Encodings for Input and Output
1181 See Encode.
1182 Unicode Regular Expression Support Level
1184 The following list of Unicode supported features for regular
1185 expressions describes all features currently directly supported by core
1186 Perl. The references to "Level N" and the section numbers refer to
1187 UTS#18 "Unicode Regular Expressions"
1188 <https://www.unicode.org/reports/tr18>, version 18, October 2016.
1189 Level 1 - Basic Unicode Support
1191 RL1.1 Hex Notation - Done [1]
1193 RL1.2 Properties - Done [2]
1194 RL1.2a Compatibility Properties - Done [3]
1195 RL1.3 Subtraction and Intersection - Experimental [4]
1196 RL1.4 Simple Word Boundaries - Done [5]
1197 RL1.5 Simple Loose Matches - Done [6]
1198 RL1.6 Line Boundaries - Partial [7]
1199 RL1.7 Supplementary Code Points - Done [8]
1200 [1] "\N{U+...}" and "\x{...}"
1202 [2] "\p{...}" "\P{...}". This requirement is for a minimal list of
1203 properties. Perl supports these. See R2.7 for other properties.
1204 [3] Perl has "\d" "\D" "\s" "\S" "\w" "\W" "\X" "[:prop:]" "[:^prop:]",
1205 plus all the properties specified by
1206 <https://www.unicode.org/reports/tr18/#Compatibility_Properties>.
1207 These are described above in "Other Properties"
1208 [4] The experimental feature "(?[...])" starting in v5.18 accomplishes
1209 this.
1210 See "(?[ ])" in perlre. If you don't want to use an experimental
1212 feature, you can use one of the following:
1213 • Regular expression lookahead
1215 You can mimic class subtraction using lookahead. For example,
1217 what UTS#18 might write as
1218 [{Block=Greek}-[{UNASSIGNED}]]
1220 in Perl can be written as:
1222 (?!\p{Unassigned})\p{Block=Greek}
1224 (?=\p{Assigned})\p{Block=Greek}
1225 But in this particular example, you probably really want
1227 \p{Greek}
1229 which will match assigned characters known to be part of the
1231 Greek script.
1232 • CPAN module "Unicode::Regex::Set"
1234 It does implement the full UTS#18 grouping, intersection,
1236 union, and removal (subtraction) syntax.
1237 • "User-Defined Character Properties"
1239 "+" for union, "-" for removal (set-difference), "&" for
1241 intersection
1242 [5] "\b" "\B" meet most, but not all, the details of this requirement,
1244 but "\b{wb}" and "\B{wb}" do, as well as the stricter R2.3.
1245 [6] Note that Perl does Full case-folding in matching, not Simple:
1246 For example "U+1F88" is equivalent to "U+1F00 U+03B9", instead of
1248 just "U+1F80". This difference matters mainly for certain Greek
1249 capital letters with certain modifiers: the Full case-folding
1250 decomposes the letter, while the Simple case-folding would map it
1251 to a single character.
1252 [7] The reason this is considered to be only partially implemented is
1254 that Perl has "qr/\b{lb}/" and "Unicode::LineBreak" that are
1255 conformant with UAX#14 "Unicode Line Breaking Algorithm"
1256 <https://www.unicode.org/reports/tr14>. The regular expression
1257 construct provides default behavior, while the heavier-weight
1258 module provides customizable line breaking.
1259 But Perl treats "\n" as the start- and end-line delimiter, whereas
1261 Unicode specifies more characters that should be so-interpreted.
1262 These are:
1264 VT U+000B (\v in C)
1266 FF U+000C (\f)
1267 CR U+000D (\r)
1268 NEL U+0085
1269 LS U+2028
1270 PS U+2029
1271 "^" and "$" in regular expression patterns are supposed to match
1273 all these, but don't. These characters also don't, but should,
1274 affect "<>" $., and script line numbers.
1275 Also, lines should not be split within "CRLF" (i.e. there is no
1277 empty line between "\r" and "\n"). For "CRLF", try the ":crlf"
1278 layer (see PerlIO).
1279 [8] UTF-8/UTF-EBDDIC used in Perl allows not only "U+10000" to
1281 "U+10FFFF" but also beyond "U+10FFFF"
1282 Level 2 - Extended Unicode Support
1284 RL2.1 Canonical Equivalents - Retracted [9]
1286 by Unicode
1287 RL2.2 Extended Grapheme Clusters and - Partial [10]
1288 Character Classes with Strings
1289 RL2.3 Default Word Boundaries - Done [11]
1290 RL2.4 Default Case Conversion - Done
1291 RL2.5 Name Properties - Done
1292 RL2.6 Wildcards in Property Values - Partial [12]
1293 RL2.7 Full Properties - Partial [13]
1294 RL2.8 Optional Properties - Partial [14]
1295 [9] Unicode has rewritten this portion of UTS#18 to say that getting
1297 canonical equivalence (see UAX#15 "Unicode Normalization Forms"
1298 <https://www.unicode.org/reports/tr15>) is basically to be done at the
1299 programmer level. Use NFD to write both your regular expressions and
1300 text to match them against (you can use Unicode::Normalize).
1301 [10] Perl has "\X" and "\b{gcb}". Unicode has retracted their
1302 "Grapheme Cluster Mode", and recently added string properties, which
1303 Perl does not yet support.
1304 [11] see UAX#29 "Unicode Text Segmentation"
1305 <https://www.unicode.org/reports/tr29>,
1306 [12] see "Wildcards in Property Values" above.
1307 [13] Perl supports all the properties in the Unicode Character Database
1308 (UCD). It does not yet support the listed properties that come from
1309 other Unicode sources.
1310 [14] The only optional property that Perl supports is Named Sequence.
1311 None of these properties are in the UCD.
1312 Level 3 - Tailored Support
1314 This has been retracted by Unicode.
1316 Unicode Encodings
1318 Unicode characters are assigned to code points, which are abstract
1319 numbers. To use these numbers, various encodings are needed.
1320 • UTF-8
1322 UTF-8 is a variable-length (1 to 4 bytes), byte-order independent
1324 encoding. In most of Perl's documentation, including elsewhere in
1325 this document, the term "UTF-8" means also "UTF-EBCDIC". But in
1326 this section, "UTF-8" refers only to the encoding used on ASCII
1327 platforms. It is a superset of 7-bit US-ASCII, so anything encoded
1328 in ASCII has the identical representation when encoded in UTF-8.
1329 The following table is from Unicode 3.2.
1331 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
1333 U+0000..U+007F 00..7F
1335 U+0080..U+07FF * C2..DF 80..BF
1336 U+0800..U+0FFF E0 * A0..BF 80..BF
1337 U+1000..U+CFFF E1..EC 80..BF 80..BF
1338 U+D000..U+D7FF ED 80..9F 80..BF
1339 U+D800..U+DFFF +++++ utf16 surrogates, not legal utf8 +++++
1340 U+E000..U+FFFF EE..EF 80..BF 80..BF
1341 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
1342 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
1343 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
1344 Note the gaps marked by "*" before several of the byte entries
1346 above. These are caused by legal UTF-8 avoiding non-shortest
1347 encodings: it is technically possible to UTF-8-encode a single code
1348 point in different ways, but that is explicitly forbidden, and the
1349 shortest possible encoding should always be used (and that is what
1350 Perl does).
1351 Another way to look at it is via bits:
1353 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
1355 0aaaaaaa 0aaaaaaa
1357 00000bbbbbaaaaaa 110bbbbb 10aaaaaa
1358 ccccbbbbbbaaaaaa 1110cccc 10bbbbbb 10aaaaaa
1359 00000dddccccccbbbbbbaaaaaa 11110ddd 10cccccc 10bbbbbb 10aaaaaa
1360 As you can see, the continuation bytes all begin with "10", and the
1362 leading bits of the start byte tell how many bytes there are in the
1363 encoded character.
1364 The original UTF-8 specification allowed up to 6 bytes, to allow
1366 encoding of numbers up to "0x7FFF_FFFF". Perl continues to allow
1367 those, and has extended that up to 13 bytes to encode code points
1368 up to what can fit in a 64-bit word. However, Perl will warn if
1369 you output any of these as being non-portable; and under strict
1370 UTF-8 input protocols, they are forbidden. In addition, it is now
1371 illegal to use a code point larger than what a signed integer
1372 variable on your system can hold. On 32-bit ASCII systems, this
1373 means "0x7FFF_FFFF" is the legal maximum (much higher on 64-bit
1374 systems).
1375 • UTF-EBCDIC
1377 Like UTF-8, but EBCDIC-safe, in the way that UTF-8 is ASCII-safe.
1379 This means that all the basic characters (which includes all those
1380 that have ASCII equivalents (like "A", "0", "%", etc.) are the
1381 same in both EBCDIC and UTF-EBCDIC.)
1382 UTF-EBCDIC is used on EBCDIC platforms. It generally requires more
1384 bytes to represent a given code point than UTF-8 does; the largest
1385 Unicode code points take 5 bytes to represent (instead of 4 in
1386 UTF-8), and, extended for 64-bit words, it uses 14 bytes instead of
1387 13 bytes in UTF-8.
1388 • UTF-16, UTF-16BE, UTF-16LE, Surrogates, and "BOM"'s (Byte Order
1390 Marks)
1391 The followings items are mostly for reference and general Unicode
1393 knowledge, Perl doesn't use these constructs internally.
1394 Like UTF-8, UTF-16 is a variable-width encoding, but where UTF-8
1396 uses 8-bit code units, UTF-16 uses 16-bit code units. All code
1397 points occupy either 2 or 4 bytes in UTF-16: code points
1398 "U+0000..U+FFFF" are stored in a single 16-bit unit, and code
1399 points "U+10000..U+10FFFF" in two 16-bit units. The latter case is
1400 using surrogates, the first 16-bit unit being the high surrogate,
1401 and the second being the low surrogate.
1402 Surrogates are code points set aside to encode the
1404 "U+10000..U+10FFFF" range of Unicode code points in pairs of 16-bit
1405 units. The high surrogates are the range "U+D800..U+DBFF" and the
1406 low surrogates are the range "U+DC00..U+DFFF". The surrogate
1407 encoding is
1408 $hi = ($uni - 0x10000) / 0x400 + 0xD800;
1410 $lo = ($uni - 0x10000) % 0x400 + 0xDC00;
1411 and the decoding is
1413 $uni = 0x10000 + ($hi - 0xD800) * 0x400 + ($lo - 0xDC00);
1415 Because of the 16-bitness, UTF-16 is byte-order dependent. UTF-16
1417 itself can be used for in-memory computations, but if storage or
1418 transfer is required either UTF-16BE (big-endian) or UTF-16LE
1419 (little-endian) encodings must be chosen.
1420 This introduces another problem: what if you just know that your
1422 data is UTF-16, but you don't know which endianness? Byte Order
1423 Marks, or "BOM"'s, are a solution to this. A special character has
1424 been reserved in Unicode to function as a byte order marker: the
1425 character with the code point "U+FEFF" is the "BOM".
1426 The trick is that if you read a "BOM", you will know the byte
1428 order, since if it was written on a big-endian platform, you will
1429 read the bytes "0xFE 0xFF", but if it was written on a little-
1430 endian platform, you will read the bytes "0xFF 0xFE". (And if the
1431 originating platform was writing in ASCII platform UTF-8, you will
1432 read the bytes "0xEF 0xBB 0xBF".)
1433 The way this trick works is that the character with the code point
1435 "U+FFFE" is not supposed to be in input streams, so the sequence of
1436 bytes "0xFF 0xFE" is unambiguously ""BOM", represented in little-
1437 endian format" and cannot be "U+FFFE", represented in big-endian
1438 format".
1439 Surrogates have no meaning in Unicode outside their use in pairs to
1441 represent other code points. However, Perl allows them to be
1442 represented individually internally, for example by saying
1443 "chr(0xD801)", so that all code points, not just those valid for
1444 open interchange, are representable. Unicode does define semantics
1445 for them, such as their "General_Category" is "Cs". But because
1446 their use is somewhat dangerous, Perl will warn (using the warning
1447 category "surrogate", which is a sub-category of "utf8") if an
1448 attempt is made to do things like take the lower case of one, or
1449 match case-insensitively, or to output them. (But don't try this
1450 on Perls before 5.14.)
1451 • UTF-32, UTF-32BE, UTF-32LE
1453 The UTF-32 family is pretty much like the UTF-16 family, except
1455 that the units are 32-bit, and therefore the surrogate scheme is
1456 not needed. UTF-32 is a fixed-width encoding. The "BOM"
1457 signatures are "0x00 0x00 0xFE 0xFF" for BE and "0xFF 0xFE 0x00
1458 0x00" for LE.
1459 • UCS-2, UCS-4
1461 Legacy, fixed-width encodings defined by the ISO 10646 standard.
1463 UCS-2 is a 16-bit encoding. Unlike UTF-16, UCS-2 is not extensible
1464 beyond "U+FFFF", because it does not use surrogates. UCS-4 is a
1465 32-bit encoding, functionally identical to UTF-32 (the difference
1466 being that UCS-4 forbids neither surrogates nor code points larger
1467 than "0x10_FFFF").
1468 • UTF-7
1470 A seven-bit safe (non-eight-bit) encoding, which is useful if the
1472 transport or storage is not eight-bit safe. Defined by RFC 2152.
1473 Noncharacter code points
1475 66 code points are set aside in Unicode as "noncharacter code points".
1476 These all have the "Unassigned" ("Cn") "General_Category", and no
1477 character will ever be assigned to any of them. They are the 32 code
1478 points between "U+FDD0" and "U+FDEF" inclusive, and the 34 code points:
1479 U+FFFE U+FFFF
1481 U+1FFFE U+1FFFF
1482 U+2FFFE U+2FFFF
1483 ...
1484 U+EFFFE U+EFFFF
1485 U+FFFFE U+FFFFF
1486 U+10FFFE U+10FFFF
1487 Until Unicode 7.0, the noncharacters were "forbidden for use in open
1489 interchange of Unicode text data", so that code that processed those
1490 streams could use these code points as sentinels that could be mixed in
1491 with character data, and would always be distinguishable from that
1492 data. (Emphasis above and in the next paragraph are added in this
1493 document.)
1494 Unicode 7.0 changed the wording so that they are "not recommended for
1496 use in open interchange of Unicode text data". The 7.0 Standard goes
1497 on to say:
1498 "If a noncharacter is received in open interchange, an application
1500 is not required to interpret it in any way. It is good practice,
1501 however, to recognize it as a noncharacter and to take appropriate
1502 action, such as replacing it with "U+FFFD" replacement character,
1503 to indicate the problem in the text. It is not recommended to
1504 simply delete noncharacter code points from such text, because of
1505 the potential security issues caused by deleting uninterpreted
1506 characters. (See conformance clause C7 in Section 3.2, Conformance
1507 Requirements, and Unicode Technical Report #36, "Unicode Security
1508 Considerations"
1509 <https://www.unicode.org/reports/tr36/#Substituting_for_Ill_Formed_Subsequences>)."
1510 This change was made because it was found that various commercial tools
1512 like editors, or for things like source code control, had been written
1513 so that they would not handle program files that used these code
1514 points, effectively precluding their use almost entirely! And that was
1515 never the intent. They've always been meant to be usable within an
1516 application, or cooperating set of applications, at will.
1517 If you're writing code, such as an editor, that is supposed to be able
1519 to handle any Unicode text data, then you shouldn't be using these code
1520 points yourself, and instead allow them in the input. If you need
1521 sentinels, they should instead be something that isn't legal Unicode.
1522 For UTF-8 data, you can use the bytes 0xC1 and 0xC2 as sentinels, as
1523 they never appear in well-formed UTF-8. (There are equivalents for
1524 UTF-EBCDIC). You can also store your Unicode code points in integer
1525 variables and use negative values as sentinels.
1526 If you're not writing such a tool, then whether you accept
1528 noncharacters as input is up to you (though the Standard recommends
1529 that you not). If you do strict input stream checking with Perl, these
1530 code points continue to be forbidden. This is to maintain backward
1531 compatibility (otherwise potential security holes could open up, as an
1532 unsuspecting application that was written assuming the noncharacters
1533 would be filtered out before getting to it, could now, without warning,
1534 start getting them). To do strict checking, you can use the layer
1535 ":encoding('UTF-8')".
1536 Perl continues to warn (using the warning category "nonchar", which is
1538 a sub-category of "utf8") if an attempt is made to output
1539 noncharacters.
1540 Beyond Unicode code points
1542 The maximum Unicode code point is "U+10FFFF", and Unicode only defines
1543 operations on code points up through that. But Perl works on code
1544 points up to the maximum permissible signed number available on the
1545 platform. However, Perl will not accept these from input streams
1546 unless lax rules are being used, and will warn (using the warning
1547 category "non_unicode", which is a sub-category of "utf8") if any are
1548 output.
1549 Since Unicode rules are not defined on these code points, if a Unicode-
1551 defined operation is done on them, Perl uses what we believe are
1552 sensible rules, while generally warning, using the "non_unicode"
1553 category. For example, "uc("\x{11_0000}")" will generate such a
1554 warning, returning the input parameter as its result, since Perl
1555 defines the uppercase of every non-Unicode code point to be the code
1556 point itself. (All the case changing operations, not just uppercasing,
1557 work this way.)
1558 The situation with matching Unicode properties in regular expressions,
1560 the "\p{}" and "\P{}" constructs, against these code points is not as
1561 clear cut, and how these are handled has changed as we've gained
1562 experience.
1563 One possibility is to treat any match against these code points as
1565 undefined. But since Perl doesn't have the concept of a match being
1566 undefined, it converts this to failing or "FALSE". This is almost, but
1567 not quite, what Perl did from v5.14 (when use of these code points
1568 became generally reliable) through v5.18. The difference is that Perl
1569 treated all "\p{}" matches as failing, but all "\P{}" matches as
1570 succeeding.
1571 One problem with this is that it leads to unexpected, and confusing
1573 results in some cases:
1574 chr(0x110000) =~ \p{ASCII_Hex_Digit=True} # Failed on <= v5.18
1576 chr(0x110000) =~ \p{ASCII_Hex_Digit=False} # Failed! on <= v5.18
1577 That is, it treated both matches as undefined, and converted that to
1579 false (raising a warning on each). The first case is the expected
1580 result, but the second is likely counterintuitive: "How could both be
1581 false when they are complements?" Another problem was that the
1582 implementation optimized many Unicode property matches down to already
1583 existing simpler, faster operations, which don't raise the warning. We
1584 chose to not forgo those optimizations, which help the vast majority of
1585 matches, just to generate a warning for the unlikely event that an
1586 above-Unicode code point is being matched against.
1587 As a result of these problems, starting in v5.20, what Perl does is to
1589 treat non-Unicode code points as just typical unassigned Unicode
1590 characters, and matches accordingly. (Note: Unicode has atypical
1591 unassigned code points. For example, it has noncharacter code points,
1592 and ones that, when they do get assigned, are destined to be written
1593 Right-to-left, as Arabic and Hebrew are. Perl assumes that no non-
1594 Unicode code point has any atypical properties.)
1595 Perl, in most cases, will raise a warning when matching an above-
1597 Unicode code point against a Unicode property when the result is "TRUE"
1598 for "\p{}", and "FALSE" for "\P{}". For example:
1599 chr(0x110000) =~ \p{ASCII_Hex_Digit=True} # Fails, no warning
1601 chr(0x110000) =~ \p{ASCII_Hex_Digit=False} # Succeeds, with warning
1602 In both these examples, the character being matched is non-Unicode, so
1604 Unicode doesn't define how it should match. It clearly isn't an ASCII
1605 hex digit, so the first example clearly should fail, and so it does,
1606 with no warning. But it is arguable that the second example should
1607 have an undefined, hence "FALSE", result. So a warning is raised for
1608 it.
1609 Thus the warning is raised for many fewer cases than in earlier Perls,
1611 and only when what the result is could be arguable. It turns out that
1612 none of the optimizations made by Perl (or are ever likely to be made)
1613 cause the warning to be skipped, so it solves both problems of Perl's
1614 earlier approach. The most commonly used property that is affected by
1615 this change is "\p{Unassigned}" which is a short form for
1616 "\p{General_Category=Unassigned}". Starting in v5.20, all non-Unicode
1617 code points are considered "Unassigned". In earlier releases the
1618 matches failed because the result was considered undefined.
1619 The only place where the warning is not raised when it might ought to
1621 have been is if optimizations cause the whole pattern match to not even
1622 be attempted. For example, Perl may figure out that for a string to
1623 match a certain regular expression pattern, the string has to contain
1624 the substring "foobar". Before attempting the match, Perl may look for
1625 that substring, and if not found, immediately fail the match without
1626 actually trying it; so no warning gets generated even if the string
1627 contains an above-Unicode code point.
1628 This behavior is more "Do what I mean" than in earlier Perls for most
1630 applications. But it catches fewer issues for code that needs to be
1631 strictly Unicode compliant. Therefore there is an additional mode of
1632 operation available to accommodate such code. This mode is enabled if
1633 a regular expression pattern is compiled within the lexical scope where
1634 the "non_unicode" warning class has been made fatal, say by:
1635 use warnings FATAL => "non_unicode"
1637 (see warnings). In this mode of operation, Perl will raise the warning
1639 for all matches against a non-Unicode code point (not just the arguable
1640 ones), and it skips the optimizations that might cause the warning to
1641 not be output. (It currently still won't warn if the match isn't even
1642 attempted, like in the "foobar" example above.)
1643 In summary, Perl now normally treats non-Unicode code points as typical
1645 Unicode unassigned code points for regular expression matches, raising
1646 a warning only when it is arguable what the result should be. However,
1647 if this warning has been made fatal, it isn't skipped.
1648 There is one exception to all this. "\p{All}" looks like a Unicode
1650 property, but it is a Perl extension that is defined to be true for all
1651 possible code points, Unicode or not, so no warning is ever generated
1652 when matching this against a non-Unicode code point. (Prior to v5.20,
1653 it was an exact synonym for "\p{Any}", matching code points 0 through
1654 0x10FFFF.)
1655 Security Implications of Unicode
1657 First, read Unicode Security Considerations
1658 <https://www.unicode.org/reports/tr36>.
1659 Also, note the following:
1661 • Malformed UTF-8
1663 UTF-8 is very structured, so many combinations of bytes are
1665 invalid. In the past, Perl tried to soldier on and make some sense
1666 of invalid combinations, but this can lead to security holes, so
1667 now, if the Perl core needs to process an invalid combination, it
1668 will either raise a fatal error, or will replace those bytes by the
1669 sequence that forms the Unicode REPLACEMENT CHARACTER, for which
1670 purpose Unicode created it.
1671 Every code point can be represented by more than one possible
1673 syntactically valid UTF-8 sequence. Early on, both Unicode and
1674 Perl considered any of these to be valid, but now, all sequences
1675 longer than the shortest possible one are considered to be
1676 malformed.
1677 Unicode considers many code points to be illegal, or to be avoided.
1679 Perl generally accepts them, once they have passed through any
1680 input filters that may try to exclude them. These have been
1681 discussed above (see "Surrogates" under UTF-16 in "Unicode
1682 Encodings", "Noncharacter code points", and "Beyond Unicode code
1683 points").
1684 • Regular expression pattern matching may surprise you if you're not
1686 accustomed to Unicode. Starting in Perl 5.14, several pattern
1687 modifiers are available to control this, called the character set
1688 modifiers. Details are given in "Character set modifiers" in
1689 perlre.
1690 As discussed elsewhere, Perl has one foot (two hooves?) planted in each
1692 of two worlds: the old world of ASCII and single-byte locales, and the
1693 new world of Unicode, upgrading when necessary. If your legacy code
1694 does not explicitly use Unicode, no automatic switch-over to Unicode
1695 should happen.
1696 Unicode in Perl on EBCDIC
1698 Unicode is supported on EBCDIC platforms. See perlebcdic.
1699 Unless ASCII vs. EBCDIC issues are specifically being discussed,
1701 references to UTF-8 encoding in this document and elsewhere should be
1702 read as meaning UTF-EBCDIC on EBCDIC platforms. See "Unicode and UTF"
1703 in perlebcdic.
1704 Because UTF-EBCDIC is so similar to UTF-8, the differences are mostly
1706 hidden from you; "use utf8" (and NOT something like "use utfebcdic")
1707 declares the script is in the platform's "native" 8-bit encoding of
1708 Unicode. (Similarly for the ":utf8" layer.)
1709 Locales
1711 See "Unicode and UTF-8" in perllocale
1712 When Unicode Does Not Happen
1714 There are still many places where Unicode (in some encoding or another)
1715 could be given as arguments or received as results, or both in Perl,
1716 but it is not, in spite of Perl having extensive ways to input and
1717 output in Unicode, and a few other "entry points" like the @ARGV array
1718 (which can sometimes be interpreted as UTF-8).
1719 The following are such interfaces. Also, see "The "Unicode Bug"". For
1721 all of these interfaces Perl currently (as of v5.16.0) simply assumes
1722 byte strings both as arguments and results, or UTF-8 strings if the
1723 (deprecated) "encoding" pragma has been used.
1724 One reason that Perl does not attempt to resolve the role of Unicode in
1726 these situations is that the answers are highly dependent on the
1727 operating system and the file system(s). For example, whether
1728 filenames can be in Unicode and in exactly what kind of encoding, is
1729 not exactly a portable concept. Similarly for "qx" and "system": how
1730 well will the "command-line interface" (and which of them?) handle
1731 Unicode?
1732 • "chdir", "chmod", "chown", "chroot", "exec", "link", "lstat",
1734 "mkdir", "rename", "rmdir", "stat", "symlink", "truncate",
1735 "unlink", "utime", "-X"
1736 • %ENV
1738 • "glob" (aka the "<*>")
1740 • "open", "opendir", "sysopen"
1742 • "qx" (aka the backtick operator), "system"
1744 • "readdir", "readlink"
1746 The "Unicode Bug"
1748 The term, "Unicode bug" has been applied to an inconsistency with the
1749 code points in the "Latin-1 Supplement" block, that is, between 128 and
1750 255. Without a locale specified, unlike all other characters or code
1751 points, these characters can have very different semantics depending on
1752 the rules in effect. (Characters whose code points are above 255 force
1753 Unicode rules; whereas the rules for ASCII characters are the same
1754 under both ASCII and Unicode rules.)
1755 Under Unicode rules, these upper-Latin1 characters are interpreted as
1757 Unicode code points, which means they have the same semantics as
1758 Latin-1 (ISO-8859-1) and C1 controls.
1759 As explained in "ASCII Rules versus Unicode Rules", under ASCII rules,
1761 they are considered to be unassigned characters.
1762 This can lead to unexpected results. For example, a string's semantics
1764 can suddenly change if a code point above 255 is appended to it, which
1765 changes the rules from ASCII to Unicode. As an example, consider the
1766 following program and its output:
1767 $ perl -le'
1769 no feature "unicode_strings";
1770 $s1 = "\xC2";
1771 $s2 = "\x{2660}";
1772 for ($s1, $s2, $s1.$s2) {
1773 print /\w/ || 0;
1774 }
1775 '
1776 0
1777 0
1778 1
1779 If there's no "\w" in "s1" nor in "s2", why does their concatenation
1781 have one?
1782 This anomaly stems from Perl's attempt to not disturb older programs
1784 that didn't use Unicode, along with Perl's desire to add Unicode
1785 support seamlessly. But the result turned out to not be seamless. (By
1786 the way, you can choose to be warned when things like this happen. See
1787 "encoding::warnings".)
1788 "use feature 'unicode_strings'" was added, starting in Perl v5.12, to
1790 address this problem. It affects these things:
1791 • Changing the case of a scalar, that is, using "uc()", "ucfirst()",
1793 "lc()", and "lcfirst()", or "\L", "\U", "\u" and "\l" in double-
1794 quotish contexts, such as regular expression substitutions.
1795 Under "unicode_strings" starting in Perl 5.12.0, Unicode rules are
1797 generally used. See "lc" in perlfunc for details on how this works
1798 in combination with various other pragmas.
1799 • Using caseless ("/i") regular expression matching.
1801 Starting in Perl 5.14.0, regular expressions compiled within the
1803 scope of "unicode_strings" use Unicode rules even when executed or
1804 compiled into larger regular expressions outside the scope.
1805 • Matching any of several properties in regular expressions.
1807 These properties are "\b" (without braces), "\B" (without braces),
1809 "\s", "\S", "\w", "\W", and all the Posix character classes except
1810 "[[:ascii:]]".
1811 Starting in Perl 5.14.0, regular expressions compiled within the
1813 scope of "unicode_strings" use Unicode rules even when executed or
1814 compiled into larger regular expressions outside the scope.
1815 • In "quotemeta" or its inline equivalent "\Q".
1817 Starting in Perl 5.16.0, consistent quoting rules are used within
1819 the scope of "unicode_strings", as described in "quotemeta" in
1820 perlfunc. Prior to that, or outside its scope, no code points
1821 above 127 are quoted in UTF-8 encoded strings, but in byte encoded
1822 strings, code points between 128-255 are always quoted.
1823 • In the ".." or range operator.
1825 Starting in Perl 5.26.0, the range operator on strings treats their
1827 lengths consistently within the scope of "unicode_strings". Prior
1828 to that, or outside its scope, it could produce strings whose
1829 length in characters exceeded that of the right-hand side, where
1830 the right-hand side took up more bytes than the correct range
1831 endpoint.
1832 • In "split"'s special-case whitespace splitting.
1834 Starting in Perl 5.28.0, the "split" function with a pattern
1836 specified as a string containing a single space handles whitespace
1837 characters consistently within the scope of "unicode_strings".
1838 Prior to that, or outside its scope, characters that are whitespace
1839 according to Unicode rules but not according to ASCII rules were
1840 treated as field contents rather than field separators when they
1841 appear in byte-encoded strings.
1842 You can see from the above that the effect of "unicode_strings"
1844 increased over several Perl releases. (And Perl's support for Unicode
1845 continues to improve; it's best to use the latest available release in
1846 order to get the most complete and accurate results possible.) Note
1847 that "unicode_strings" is automatically chosen if you "use 5.012" or
1848 higher.
1849 For Perls earlier than those described above, or when a string is
1851 passed to a function outside the scope of "unicode_strings", see the
1852 next section.
1853 Forcing Unicode in Perl (Or Unforcing Unicode in Perl)
1855 Sometimes (see "When Unicode Does Not Happen" or "The "Unicode Bug"")
1856 there are situations where you simply need to force a byte string into
1857 UTF-8, or vice versa. The standard module Encode can be used for this,
1858 or the low-level calls "utf8::upgrade($bytestring)" and
1859 "utf8::downgrade($utf8string[, FAIL_OK])".
1860 Note that "utf8::downgrade()" can fail if the string contains
1862 characters that don't fit into a byte.
1863 Calling either function on a string that already is in the desired
1865 state is a no-op.
1866 "ASCII Rules versus Unicode Rules" gives all the ways that a string is
1868 made to use Unicode rules.
1869 Using Unicode in XS
1871 See "Unicode Support" in perlguts for an introduction to Unicode at the
1872 XS level, and "Unicode Support" in perlapi for the API details.
1873 Hacking Perl to work on earlier Unicode versions (for very serious hackers
1875 only)
1876 Perl by default comes with the latest supported Unicode version built-
1877 in, but the goal is to allow you to change to use any earlier one. In
1878 Perls v5.20 and v5.22, however, the earliest usable version is Unicode
1879 5.1. Perl v5.18 and v5.24 are able to handle all earlier versions.
1880 Download the files in the desired version of Unicode from the Unicode
1882 web site <https://www.unicode.org>). These should replace the existing
1883 files in lib/unicore in the Perl source tree. Follow the instructions
1884 in README.perl in that directory to change some of their names, and
1885 then build perl (see INSTALL).
1886 Porting code from perl-5.6.X
1888 Perls starting in 5.8 have a different Unicode model from 5.6. In 5.6
1889 the programmer was required to use the "utf8" pragma to declare that a
1890 given scope expected to deal with Unicode data and had to make sure
1891 that only Unicode data were reaching that scope. If you have code that
1892 is working with 5.6, you will need some of the following adjustments to
1893 your code. The examples are written such that the code will continue to
1894 work under 5.6, so you should be safe to try them out.
1895 • A filehandle that should read or write UTF-8
1897 if ($] > 5.008) {
1899 binmode $fh, ":encoding(UTF-8)";
1900 }
1901 • A scalar that is going to be passed to some extension
1903 Be it "Compress::Zlib", "Apache::Request" or any extension that has
1905 no mention of Unicode in the manpage, you need to make sure that the
1906 UTF8 flag is stripped off. Note that at the time of this writing
1907 (January 2012) the mentioned modules are not UTF-8-aware. Please
1908 check the documentation to verify if this is still true.
1909 if ($] > 5.008) {
1911 require Encode;
1912 $val = Encode::encode("UTF-8", $val); # make octets
1913 }
1914 • A scalar we got back from an extension
1916 If you believe the scalar comes back as UTF-8, you will most likely
1918 want the UTF8 flag restored:
1919 if ($] > 5.008) {
1921 require Encode;
1922 $val = Encode::decode("UTF-8", $val);
1923 }
1924 • Same thing, if you are really sure it is UTF-8
1926 if ($] > 5.008) {
1928 require Encode;
1929 Encode::_utf8_on($val);
1930 }
1931 • A wrapper for DBI "fetchrow_array" and "fetchrow_hashref"
1933 When the database contains only UTF-8, a wrapper function or method
1935 is a convenient way to replace all your "fetchrow_array" and
1936 "fetchrow_hashref" calls. A wrapper function will also make it
1937 easier to adapt to future enhancements in your database driver. Note
1938 that at the time of this writing (January 2012), the DBI has no
1939 standardized way to deal with UTF-8 data. Please check the DBI
1940 documentation to verify if that is still true.
1941 sub fetchrow {
1943 # $what is one of fetchrow_{array,hashref}
1944 my($self, $sth, $what) = @_;
1945 if ($] < 5.008) {
1946 return $sth->$what;
1947 } else {
1948 require Encode;
1949 if (wantarray) {
1950 my @arr = $sth->$what;
1951 for (@arr) {
1952 defined && /[^\000-\177]/ && Encode::_utf8_on($_);
1953 }
1954 return @arr;
1955 } else {
1956 my $ret = $sth->$what;
1957 if (ref $ret) {
1958 for my $k (keys %$ret) {
1959 defined
1960 && /[^\000-\177]/
1961 && Encode::_utf8_on($_) for $ret->{$k};
1962 }
1963 return $ret;
1964 } else {
1965 defined && /[^\000-\177]/ && Encode::_utf8_on($_) for $ret;
1966 return $ret;
1967 }
1968 }
1969 }
1970 }
1971 • A large scalar that you know can only contain ASCII
1973 Scalars that contain only ASCII and are marked as UTF-8 are
1975 sometimes a drag to your program. If you recognize such a situation,
1976 just remove the UTF8 flag:
1977 utf8::downgrade($val) if $] > 5.008;
1979
1981 See also "The "Unicode Bug"" above.
1982 Interaction with Extensions
1984 When Perl exchanges data with an extension, the extension should be
1985 able to understand the UTF8 flag and act accordingly. If the extension
1986 doesn't recognize that flag, it's likely that the extension will return
1987 incorrectly-flagged data.
1988 So if you're working with Unicode data, consult the documentation of
1990 every module you're using if there are any issues with Unicode data
1991 exchange. If the documentation does not talk about Unicode at all,
1992 suspect the worst and probably look at the source to learn how the
1993 module is implemented. Modules written completely in Perl shouldn't
1994 cause problems. Modules that directly or indirectly access code written
1995 in other programming languages are at risk.
1996 For affected functions, the simple strategy to avoid data corruption is
1998 to always make the encoding of the exchanged data explicit. Choose an
1999 encoding that you know the extension can handle. Convert arguments
2000 passed to the extensions to that encoding and convert results back from
2001 that encoding. Write wrapper functions that do the conversions for you,
2002 so you can later change the functions when the extension catches up.
2003 To provide an example, let's say the popular "Foo::Bar::escape_html"
2005 function doesn't deal with Unicode data yet. The wrapper function would
2006 convert the argument to raw UTF-8 and convert the result back to Perl's
2007 internal representation like so:
2008 sub my_escape_html ($) {
2010 my($what) = shift;
2011 return unless defined $what;
2012 Encode::decode("UTF-8", Foo::Bar::escape_html(
2013 Encode::encode("UTF-8", $what)));
2014 }
2015 Sometimes, when the extension does not convert data but just stores and
2017 retrieves it, you will be able to use the otherwise dangerous
2018 "Encode::_utf8_on()" function. Let's say the popular "Foo::Bar"
2019 extension, written in C, provides a "param" method that lets you store
2020 and retrieve data according to these prototypes:
2021 $self->param($name, $value); # set a scalar
2023 $value = $self->param($name); # retrieve a scalar
2024 If it does not yet provide support for any encoding, one could write a
2026 derived class with such a "param" method:
2027 sub param {
2029 my($self,$name,$value) = @_;
2030 utf8::upgrade($name); # make sure it is UTF-8 encoded
2031 if (defined $value) {
2032 utf8::upgrade($value); # make sure it is UTF-8 encoded
2033 return $self->SUPER::param($name,$value);
2034 } else {
2035 my $ret = $self->SUPER::param($name);
2036 Encode::_utf8_on($ret); # we know, it is UTF-8 encoded
2037 return $ret;
2038 }
2039 }
2040 Some extensions provide filters on data entry/exit points, such as
2042 "DB_File::filter_store_key" and family. Look out for such filters in
2043 the documentation of your extensions; they can make the transition to
2044 Unicode data much easier.
2045 Speed
2047 Some functions are slower when working on UTF-8 encoded strings than on
2048 byte encoded strings. All functions that need to hop over characters
2049 such as "length()", "substr()" or "index()", or matching regular
2050 expressions can work much faster when the underlying data are byte-
2051 encoded.
2052 In Perl 5.8.0 the slowness was often quite spectacular; in Perl 5.8.1 a
2054 caching scheme was introduced which improved the situation. In
2055 general, operations with UTF-8 encoded strings are still slower. As an
2056 example, the Unicode properties (character classes) like "\p{Nd}" are
2057 known to be quite a bit slower (5-20 times) than their simpler
2058 counterparts like "[0-9]" (then again, there are hundreds of Unicode
2059 characters matching "Nd" compared with the 10 ASCII characters matching
2060 "[0-9]").
2061
2063 perlunitut, perluniintro, perluniprops, Encode, open, utf8, bytes,
2064 perlretut, "${^UNICODE}" in perlvar,
2065 <https://www.unicode.org/reports/tr44>).
2066perl v5.32.1 2021-05-31 PERLUNICODE(1)