This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add thread-safe locale handling
[perl5.git] / pod / perllocale.pod
CommitLineData
66cbab2c
KW
1=encoding utf8
2
5f05dabc
PP
3=head1 NAME
4
b0c42ed9 5perllocale - Perl locale handling (internationalization and localization)
5f05dabc
PP
6
7=head1 DESCRIPTION
8
66cbab2c
KW
9In the beginning there was ASCII, the "American Standard Code for
10Information Interchange", which works quite well for Americans with
11their English alphabet and dollar-denominated currency. But it doesn't
12work so well even for other English speakers, who may use different
13currencies, such as the pound sterling (as the symbol for that currency
14is not in ASCII); and it's hopelessly inadequate for many of the
15thousands of the world's other languages.
16
17To address these deficiencies, the concept of locales was invented
18(formally the ISO C, XPG4, POSIX 1.c "locale system"). And applications
19were and are being written that use the locale mechanism. The process of
20making such an application take account of its users' preferences in
21these kinds of matters is called B<internationalization> (often
22abbreviated as B<i18n>); telling such an application about a particular
23set of preferences is known as B<localization> (B<l10n>).
24
0c880285
KW
25Perl has been extended to support certain types of locales available in
26the locale system. This is controlled per application by using one
27pragma, one function call, and several environment variables.
28
29Perl supports single-byte locales that are supersets of ASCII, such as
30the ISO 8859 ones, and one multi-byte-type locale, UTF-8 ones, described
31in the next paragraph. Perl doesn't support any other multi-byte
32locales, such as the ones for East Asian languages.
66cbab2c
KW
33
34Unfortunately, there are quite a few deficiencies with the design (and
31f05a37
KW
35often, the implementations) of locales. Unicode was invented (see
36L<perlunitut> for an introduction to that) in part to address these
37design deficiencies, and nowadays, there is a series of "UTF-8
38locales", based on Unicode. These are locales whose character set is
39Unicode, encoded in UTF-8. Starting in v5.20, Perl fully supports
9accf6df 40UTF-8 locales, except for sorting and string comparisons like C<lt> and
a4a439fb
KW
41C<ge>. Starting in v5.26, Perl can handle these reasonably as well,
42depending on the platform's implementation. However, for earlier
0c880285 43releases or for better control, use L<Unicode::Collate>. Perl continues to
a4a439fb
KW
44support the old non UTF-8 locales as well. There are currently no UTF-8
45locales for EBCDIC platforms.
31f05a37 46
66cbab2c
KW
47(Unicode is also creating C<CLDR>, the "Common Locale Data Repository",
48L<http://cldr.unicode.org/> which includes more types of information than
49are available in the POSIX locale system. At the time of this writing,
50there was no CPAN module that provides access to this XML-encoded data.
e784ce68
KW
51However, it is possible to compute the POSIX locale data from them, and
52earlier CLDR versions had these already extracted for you as UTF-8 locales
53L<http://unicode.org/Public/cldr/2.0.1/>.)
66cbab2c
KW
54
55=head1 WHAT IS A LOCALE
56
57A locale is a set of data that describes various aspects of how various
58communities in the world categorize their world. These categories are
59broken down into the following types (some of which include a brief
60note here):
61
62=over
63
cb88b78e 64=item Category C<LC_NUMERIC>: Numeric formatting
66cbab2c
KW
65
66This indicates how numbers should be formatted for human readability,
67for example the character used as the decimal point.
68
cb88b78e 69=item Category C<LC_MONETARY>: Formatting of monetary amounts
66cbab2c
KW
70
71=for comment
ebc3223b 72The nbsp below makes this look better (though not great)
66cbab2c
KW
73
74E<160>
75
cb88b78e 76=item Category C<LC_TIME>: Date/Time formatting
66cbab2c
KW
77
78=for comment
ebc3223b 79The nbsp below makes this look better (though not great)
66cbab2c
KW
80
81E<160>
82
cb88b78e 83=item Category C<LC_MESSAGES>: Error and other messages
66cbab2c 84
2619d284 85This is used by Perl itself only for accessing operating system error
03c702c5 86messages via L<$!|perlvar/$ERRNO> and L<$^E|perlvar/$EXTENDED_OS_ERROR>.
66cbab2c 87
cb88b78e 88=item Category C<LC_COLLATE>: Collation
66cbab2c 89
76073c88 90This indicates the ordering of letters for comparison and sorting.
66cbab2c
KW
91In Latin alphabets, for example, "b", generally follows "a".
92
cb88b78e 93=item Category C<LC_CTYPE>: Character Types
66cbab2c
KW
94
95This indicates, for example if a character is an uppercase letter.
96
2619d284
KW
97=item Other categories
98
99Some platforms have other categories, dealing with such things as
100measurement units and paper sizes. None of these are used directly by
101Perl, but outside operations that Perl interacts with may use
d6ded950 102these. See L</Not within the scope of "use locale"> below.
2619d284 103
66cbab2c
KW
104=back
105
2619d284
KW
106More details on the categories used by Perl are given below in L</LOCALE
107CATEGORIES>.
66cbab2c
KW
108
109Together, these categories go a long way towards being able to customize
110a single program to run in many different locations. But there are
111deficiencies, so keep reading.
5f05dabc
PP
112
113=head1 PREPARING TO USE LOCALES
114
c5c88224
KW
115Perl itself (outside the L<POSIX> module) will not use locales unless
116specifically requested to (but
2619d284
KW
117again note that Perl may interact with code that does use them). Even
118if there is such a request, B<all> of the following must be true
b960a36e 119for it to work properly:
5f05dabc
PP
120
121=over 4
122
123=item *
124
125B<Your operating system must support the locale system>. If it does,
39332f68 126you should find that the C<setlocale()> function is a documented part of
5f05dabc
PP
127its C library.
128
129=item *
130
5a964f20 131B<Definitions for locales that you use must be installed>. You, or
14280422
DD
132your system administrator, must make sure that this is the case. The
133available locales, the location in which they are kept, and the manner
5a964f20
TC
134in which they are installed all vary from system to system. Some systems
135provide only a few, hard-wired locales and do not allow more to be
136added. Others allow you to add "canned" locales provided by the system
137supplier. Still others allow you or the system administrator to define
14280422 138and add arbitrary locales. (You may have to ask your supplier to
5a964f20 139provide canned locales that are not delivered with your operating
14280422 140system.) Read your system documentation for further illumination.
5f05dabc
PP
141
142=item *
143
144B<Perl must believe that the locale system is supported>. If it does,
145C<perl -V:d_setlocale> will say that the value for C<d_setlocale> is
146C<define>.
147
148=back
149
150If you want a Perl application to process and present your data
151according to a particular locale, the application code should include
5a0de581 152the S<C<use locale>> pragma (see L</The "use locale" pragma>) where
5f05dabc
PP
153appropriate, and B<at least one> of the following must be true:
154
155=over 4
156
c052850d 157=item 1
5f05dabc 158
66cbab2c 159B<The locale-determining environment variables (see L</"ENVIRONMENT">)
5a964f20 160must be correctly set up> at the time the application is started, either
ef3087ec 161by yourself or by whomever set up your system account; or
5f05dabc 162
c052850d 163=item 2
5f05dabc 164
14280422 165B<The application must set its own locale> using the method described in
5a0de581 166L</The setlocale function>.
5f05dabc
PP
167
168=back
169
170=head1 USING LOCALES
171
d6ded950 172=head2 The C<"use locale"> pragma
5f05dabc 173
e9bc6d6b
KW
174Starting in Perl 5.28, this pragma may be used in
175L<multi-threaded|threads> applications on systems that have thread-safe
176locale ability. Some caveats apply, see L</Multi-threaded> below. On
177systems without this capability, or in earlier Perls, do NOT use this
178pragma in scripts that have multiple L<threads|threads> active. The
179locale in these cases is not local to a single thread. Another thread
180may change the locale at any time, which could cause at a minimum that a
181given thread is operating in a locale it isn't expecting to be in. On
182some platforms, segfaults can also occur. The locale change need not be
183explicit; some operations cause perl to change the locale itself. You
184are vulnerable simply by having done a S<C<"use locale">>.
fc82b82e 185
c5c88224
KW
186By default, Perl itself (outside the L<POSIX> module)
187ignores the current locale. The S<C<use locale>>
66cbab2c 188pragma tells Perl to use the current locale for some operations.
d6ded950
KW
189Starting in v5.16, there are optional parameters to this pragma,
190described below, which restrict which operations are affected by it.
c052850d
KW
191
192The current locale is set at execution time by
193L<setlocale()|/The setlocale function> described below. If that function
194hasn't yet been called in the course of the program's execution, the
66cbab2c 195current locale is that which was determined by the L</"ENVIRONMENT"> in
ebc3223b 196effect at the start of the program.
dfcc8045 197If there is no valid environment, the current locale is whatever the
65ebb059
KW
198system default has been set to. On POSIX systems, it is likely, but
199not necessarily, the "C" locale. On Windows, the default is set via the
200computer's S<C<Control Panel-E<gt>Regional and Language Options>> (or its
201current equivalent).
c052850d
KW
202
203The operations that are affected by locale are:
5f05dabc
PP
204
205=over 4
206
d6ded950 207=item B<Not within the scope of C<"use locale">>
b960a36e 208
d6ded950
KW
209Only certain operations originating outside Perl should be affected, as
210follows:
1d2ab946
KW
211
212=over 4
213
214=item *
2619d284 215
663d437a 216The current locale is used when going outside of Perl with
2619d284
KW
217operations like L<system()|perlfunc/system LIST> or
218L<qxE<sol>E<sol>|perlop/qxE<sol>STRINGE<sol>>, if those operations are
219locale-sensitive.
220
1d2ab946
KW
221=item *
222
2619d284
KW
223Also Perl gives access to various C library functions through the
224L<POSIX> module. Some of those functions are always affected by the
225current locale. For example, C<POSIX::strftime()> uses C<LC_TIME>;
226C<POSIX::strtod()> uses C<LC_NUMERIC>; C<POSIX::strcoll()> and
9accf6df 227C<POSIX::strxfrm()> use C<LC_COLLATE>. All such functions
2619d284 228will behave according to the current underlying locale, even if that
1d2ab946 229locale isn't exposed to Perl space.
2619d284 230
d9e2553c
KW
231This applies as well to L<I18N::Langinfo>.
232
1d2ab946
KW
233=item *
234
52686f2a
KW
235XS modules for all categories but C<LC_NUMERIC> get the underlying
236locale, and hence any C library functions they call will use that
8a384d3a 237underlying locale. For more discussion, see L<perlxs/CAVEATS>.
1d2ab946
KW
238
239=back
240
c5c88224
KW
241Note that all C programs (including the perl interpreter, which is
242written in C) always have an underlying locale. That locale is the "C"
243locale unless changed by a call to L<setlocale()|/The setlocale
244function>. When Perl starts up, it changes the underlying locale to the
245one which is indicated by the L</ENVIRONMENT>. When using the L<POSIX>
246module or writing XS code, it is important to keep in mind that the
247underlying locale may be something other than "C", even if the program
248hasn't explicitly changed it.
249
ebc3223b
KW
250=for comment
251The nbsp below makes this look better (though not great)
252
253E<160>
254
255=item B<Lingering effects of C<S<use locale>>>
1d2ab946
KW
256
257Certain Perl operations that are set-up within the scope of a
d6ded950 258C<use locale> retain that effect even outside the scope.
b960a36e
KW
259These include:
260
261=over 4
262
263=item *
264
265The output format of a L<write()|perlfunc/write> is determined by an
266earlier format declaration (L<perlfunc/format>), so whether or not the
267output is affected by locale is determined by if the C<format()> is
d6ded950 268within the scope of a C<use locale>, not whether the C<write()>
b960a36e
KW
269is.
270
271=item *
272
273Regular expression patterns can be compiled using
33be4c61 274L<qrE<sol>E<sol>|perlop/qrE<sol>STRINGE<sol>msixpodualn> with actual
b960a36e
KW
275matching deferred to later. Again, it is whether or not the compilation
276was done within the scope of C<use locale> that determines the match
277behavior, not if the matches are done within such a scope or not.
278
279=back
280
ebc3223b
KW
281=for comment
282The nbsp below makes this look better (though not great)
283
d6ded950 284
ebc3223b
KW
285E<160>
286
d6ded950 287=item B<Under C<"use locale";>>
66cbab2c
KW
288
289=over 4
290
291=item *
292
d6ded950 293All the above operations
b960a36e
KW
294
295=item *
296
297B<Format declarations> (L<perlfunc/format>) and hence any subsequent
298C<write()>s use C<LC_NUMERIC>.
66cbab2c
KW
299
300=item *
301
b960a36e
KW
302B<stringification and output> use C<LC_NUMERIC>.
303These include the results of
304C<print()>,
305C<printf()>,
306C<say()>,
307and
308C<sprintf()>.
66cbab2c 309
b960a36e
KW
310=item *
311
312B<The comparison operators> (C<lt>, C<le>, C<cmp>, C<ge>, and C<gt>) use
39332f68 313C<LC_COLLATE>. C<sort()> is also affected if used without an
5a964f20 314explicit comparison function, because it uses C<cmp> by default.
14280422 315
5a964f20 316B<Note:> C<eq> and C<ne> are unaffected by locale: they always
de108802 317perform a char-by-char comparison of their scalar operands. What's
14280422
DD
318more, if C<cmp> finds that its operands are equal according to the
319collation sequence specified by the current locale, it goes on to
de108802
RGS
320perform a char-by-char comparison, and only returns I<0> (equal) if the
321operands are char-for-char identical. If you really want to know whether
5a964f20 322two strings--which C<eq> and C<cmp> may consider different--are equal
14280422 323as far as collation in the locale is concerned, see the discussion in
cb88b78e 324L<Category C<LC_COLLATE>: Collation>.
5f05dabc
PP
325
326=item *
327
39332f68
KW
328B<Regular expressions and case-modification functions> (C<uc()>, C<lc()>,
329C<ucfirst()>, and C<lcfirst()>) use C<LC_CTYPE>
5f05dabc 330
2c6ee1a7
KW
331=item *
332
dbf3c4d7
KW
333B<The variables L<$!|perlvar/$ERRNO>> (and its synonyms C<$ERRNO> and
334C<$OS_ERROR>) B<and L<$^E|perlvar/$EXTENDED_OS_ERROR>> (and its synonym
2c6ee1a7
KW
335C<$EXTENDED_OS_ERROR>) when used as strings use C<LC_MESSAGES>.
336
5f05dabc
PP
337=back
338
66cbab2c 339=back
5f05dabc 340
5a964f20 341The default behavior is restored with the S<C<no locale>> pragma, or
ef3087ec 342upon reaching the end of the block enclosing C<use locale>.
3cd61afa 343Note that C<use locale> calls may be
66cbab2c
KW
344nested, and that what is in effect within an inner scope will revert to
345the outer scope's rules at the end of the inner scope.
5f05dabc 346
5a964f20 347The string result of any operation that uses locale
14280422 348information is tainted, as it is possible for a locale to be
5a0de581 349untrustworthy. See L</"SECURITY">.
5f05dabc 350
d6ded950
KW
351Starting in Perl v5.16 in a very limited way, and more generally in
352v5.22, you can restrict which category or categories are enabled by this
353particular instance of the pragma by adding parameters to it. For
354example,
355
356 use locale qw(:ctype :numeric);
357
358enables locale awareness within its scope of only those operations
359(listed above) that are affected by C<LC_CTYPE> and C<LC_NUMERIC>.
360
361The possible categories are: C<:collate>, C<:ctype>, C<:messages>,
362C<:monetary>, C<:numeric>, C<:time>, and the pseudo category
363C<:characters> (described below).
364
365Thus you can say
366
367 use locale ':messages';
368
369and only L<$!|perlvar/$ERRNO> and L<$^E|perlvar/$EXTENDED_OS_ERROR>
370will be locale aware. Everything else is unaffected.
371
372Since Perl doesn't currently do anything with the C<LC_MONETARY>
373category, specifying C<:monetary> does effectively nothing. Some
22803c6a 374systems have other categories, such as C<LC_PAPER>, but Perl
e9bc6d6b 375also doesn't do anything with them, and there is no way to specify
d6ded950
KW
376them in this pragma's arguments.
377
378You can also easily say to use all categories but one, by either, for
379example,
380
381 use locale ':!ctype';
382 use locale ':not_ctype';
383
384both of which mean to enable locale awarness of all categories but
385C<LC_CTYPE>. Only one category argument may be specified in a
386S<C<use locale>> if it is of the negated form.
387
388Prior to v5.22 only one form of the pragma with arguments is available:
389
390 use locale ':not_characters';
391
392(and you have to say C<not_>; you can't use the bang C<!> form). This
393pseudo category is a shorthand for specifying both C<:collate> and
394C<:ctype>. Hence, in the negated form, it is nearly the same thing as
395saying
396
397 use locale qw(:messages :monetary :numeric :time);
398
399We use the term "nearly", because C<:not_characters> also turns on
400S<C<use feature 'unicode_strings'>> within its scope. This form is
401less useful in v5.20 and later, and is described fully in
402L</Unicode and UTF-8>, but briefly, it tells Perl to not use the
403character portions of the locale definition, that is the C<LC_CTYPE> and
404C<LC_COLLATE> categories. Instead it will use the native character set
405(extended by Unicode). When using this parameter, you are responsible
406for getting the external character set translated into the
407native/Unicode one (which it already will be if it is one of the
408increasingly popular UTF-8 locales). There are convenient ways of doing
409this, as described in L</Unicode and UTF-8>.
410
5f05dabc
PP
411=head2 The setlocale function
412
e9bc6d6b
KW
413WARNING! Prior to Perl 5.28 or on a system that does not support
414thread-safe locale operations, do NOT use this function in a
415L<thread|threads>. The locale will change in all other threads at the
416same time, and should your thread get paused by the operating system,
417and another started, that thread will not have the locale it is
418expecting. On some platforms, there can be a race leading to segfaults
419if two threads call this function nearly simultaneously.
fc82b82e 420
14280422 421You can switch locales as often as you wish at run time with the
39332f68 422C<POSIX::setlocale()> function:
5f05dabc 423
5f05dabc
PP
424 # Import locale-handling tool set from POSIX module.
425 # This example uses: setlocale -- the function call
426 # LC_CTYPE -- explained below
2619d284
KW
427 # (Showing the testing for success/failure of operations is
428 # omitted in these examples to avoid distracting from the main
ebc3223b 429 # point)
6ea81ccf 430
5f05dabc 431 use POSIX qw(locale_h);
dfcc8045
KW
432 use locale;
433 my $old_locale;
5f05dabc 434
14280422 435 # query and save the old locale
5f05dabc
PP
436 $old_locale = setlocale(LC_CTYPE);
437
438 setlocale(LC_CTYPE, "fr_CA.ISO8859-1");
439 # LC_CTYPE now in locale "French, Canada, codeset ISO 8859-1"
440
441 setlocale(LC_CTYPE, "");
65ebb059
KW
442 # LC_CTYPE now reset to the default defined by the
443 # LC_ALL/LC_CTYPE/LANG environment variables, or to the system
444 # default. See below for documentation.
5f05dabc
PP
445
446 # restore the old locale
447 setlocale(LC_CTYPE, $old_locale);
448
39332f68 449The first argument of C<setlocale()> gives the B<category>, the second the
14280422
DD
450B<locale>. The category tells in what aspect of data processing you
451want to apply locale-specific rules. Category names are discussed in
66cbab2c 452L</LOCALE CATEGORIES> and L</"ENVIRONMENT">. The locale is the name of a
14280422
DD
453collection of customization information corresponding to a particular
454combination of language, country or territory, and codeset. Read on for
455hints on the naming of locales: not all systems name locales as in the
456example.
457
39332f68 458If no second argument is provided and the category is something other
cb88b78e 459than C<LC_ALL>, the function returns a string naming the current locale
502a173a 460for the category. You can use this value as the second argument in a
f170b852
KW
461subsequent call to C<setlocale()>, B<but> on some platforms the string
462is opaque, not something that most people would be able to decipher as
463to what locale it means.
502a173a 464
cb88b78e 465If no second argument is provided and the category is C<LC_ALL>, the
502a173a 466result is implementation-dependent. It may be a string of
c052850d 467concatenated locale names (separator also implementation-dependent)
39332f68 468or a single locale name. Please consult your L<setlocale(3)> man page for
502a173a
JH
469details.
470
471If a second argument is given and it corresponds to a valid locale,
472the locale for the category is set to that value, and the function
473returns the now-current locale value. You can then use this in yet
39332f68 474another call to C<setlocale()>. (In some implementations, the return
502a173a
JH
475value may sometimes differ from the value you gave as the second
476argument--think of it as an alias for the value you gave.)
5f05dabc
PP
477
478As the example shows, if the second argument is an empty string, the
479category's locale is returned to the default specified by the
480corresponding environment variables. Generally, this results in a
5a964f20 481return to the default that was in force when Perl started up: changes
54310121 482to the environment made by the application after startup may or may not
5a964f20 483be noticed, depending on your system's C library.
5f05dabc 484
d6ded950
KW
485Note that when a form of C<use locale> that doesn't include all
486categories is specified, Perl ignores the excluded categories.
66cbab2c 487
f170b852 488If C<set_locale()> fails for some reason (for example, an attempt to set
dfcc8045
KW
489to a locale unknown to the system), the locale for the category is not
490changed, and the function returns C<undef>.
491
e9bc6d6b
KW
492Starting in Perl 5.28, on multi-threaded perls compiled on systems that
493implement POSIX 2008 thread-safe locale operations, this function
494doesn't actually call the system C<setlocale>. Instead those
495thread-safe operations are used to emulate the C<setlocale> function,
496but in a thread-safe manner.
2619d284 497
39332f68 498For further information about the categories, consult L<setlocale(3)>.
3e6e419a 499
e9bc6d6b
KW
500=head2 Multi-threaded operation
501
502Beginning in Perl 5.28, multi-threaded locale operation is supported on
503systems that implement either the POSIX 2008 or Windows-specific
504thread-safe locale operations. Many modern systems, such as various
505Unix variants and Darwin do have this.
506
507You can tell if using locales is safe on your system by looking at the
508read-only boolean variable C<${^SAFE_LOCALES}>. The value is 1 if the
509perl is not threaded, or if it is using thread-safe locale operations.
510
511Thread-safe operations are supported in Windows starting in Visual Studio
5122005, and in systems compatible with POSIX 2008. Some platforms claim
513to support POSIX 2008, but have buggy implementations, so that the hints
514files for compiling to run on them turn off attempting to use
515thread-safety. C<${^SAFE_LOCALES}> will be 0 on them.
516
517Be aware that writing a multi-threaded application will not be portable
518to a platform which lacks the native thread-safe locale support. On
519systems that do have it, you automatically get this behavior for
520threaded perls, without having to do anything. If for some reason, you
521don't want to use this capability (perhaps the POSIX 2008 support is
522buggy on your system), you can manually compile Perl to use the old
523non-thread-safe implementation by passing the argument
524C<-Accflags='-DNO_THREAD_SAFE_LOCALE'> to F<Configure>.
525Except on Windows, this will continue to use certain of the POSIX 2008
526functions in some situations. If these are buggy, you can pass the
527following to F<Configure> instead or additionally:
528C<-Accflags='-DNO_POSIX_2008_LOCALE'>. This will also keep the code
529from using thread-safe locales.
530C<${^SAFE_LOCALES}> will be 0 on systems that turn off the thread-safe
531operations.
532
533The initial program is started up using the locale specified from the
534environment, as currently, described in L</ENVIRONMENT>. All newly
535created threads start with C<LC_ALL> set to C<"C">>. Each thread may
536use C<POSIX::setlocale()> to query or switch its locale at any time,
537without affecting any other thread. All locale-dependent operations
538automatically use their thread's locale.
539
540This should be completely transparent to any applications written
541entirely in Perl (minus a few rarely encountered caveats given in the
542L</Multi-threaded> section). Information for XS module writers is given
543in L<perlxs/Locale-aware XS code>.
544
3e6e419a
JH
545=head2 Finding locales
546
39332f68 547For locales available in your system, consult also L<setlocale(3)> to
5a964f20
TC
548see whether it leads to the list of available locales (search for the
549I<SEE ALSO> section). If that fails, try the following command lines:
5f05dabc
PP
550
551 locale -a
552
553 nlsinfo
554
555 ls /usr/lib/nls/loc
556
557 ls /usr/lib/locale
558
559 ls /usr/lib/nls
560
b478f28d
JH
561 ls /usr/share/locale
562
5f05dabc
PP
563and see whether they list something resembling these
564
2bdf8add 565 en_US.ISO8859-1 de_DE.ISO8859-1 ru_RU.ISO8859-5
502a173a 566 en_US.iso88591 de_DE.iso88591 ru_RU.iso88595
2bdf8add 567 en_US de_DE ru_RU
14280422 568 en de ru
2bdf8add
JH
569 english german russian
570 english.iso88591 german.iso88591 russian.iso88595
502a173a 571 english.roman8 russian.koi8r
5f05dabc 572
39332f68 573Sadly, even though the calling interface for C<setlocale()> has been
528d65ad 574standardized, names of locales and the directories where the
5a964f20 575configuration resides have not been. The basic form of the name is
528d65ad
JH
576I<language_territory>B<.>I<codeset>, but the latter parts after
577I<language> are not always present. The I<language> and I<country>
578are usually from the standards B<ISO 3166> and B<ISO 639>, the
579two-letter abbreviations for the countries and the languages of the
580world, respectively. The I<codeset> part often mentions some B<ISO
5818859> character set, the Latin codesets. For example, C<ISO 8859-1>
582is the so-called "Western European codeset" that can be used to encode
583most Western European languages adequately. Again, there are several
584ways to write even the name of that one standard. Lamentably.
5f05dabc 585
14280422
DD
586Two special locales are worth particular mention: "C" and "POSIX".
587Currently these are effectively the same locale: the difference is
5a964f20
TC
588mainly that the first one is defined by the C standard, the second by
589the POSIX standard. They define the B<default locale> in which
14280422 590every program starts in the absence of locale information in its
5a964f20 591environment. (The I<default> default locale, if you will.) Its language
39332f68
KW
592is (American) English and its character codeset ASCII or, rarely, a
593superset thereof (such as the "DEC Multinational Character Set
594(DEC-MCS)"). B<Warning>. The C locale delivered by some vendors
595may not actually exactly match what the C standard calls for. So
596beware.
5f05dabc 597
14280422
DD
598B<NOTE>: Not all systems have the "POSIX" locale (not all systems are
599POSIX-conformant), so use "C" when you need explicitly to specify this
600default locale.
5f05dabc 601
3e6e419a
JH
602=head2 LOCALE PROBLEMS
603
5a964f20 604You may encounter the following warning message at Perl startup:
3e6e419a
JH
605
606 perl: warning: Setting locale failed.
607 perl: warning: Please check that your locale settings:
608 LC_ALL = "En_US",
609 LANG = (unset)
610 are supported and installed on your system.
611 perl: warning: Falling back to the standard locale ("C").
612
cb88b78e 613This means that your locale settings had C<LC_ALL> set to "En_US" and
5a964f20
TC
614LANG exists but has no value. Perl tried to believe you but could not.
615Instead, Perl gave up and fell back to the "C" locale, the default locale
65ebb059
KW
616that is supposed to work no matter what. (On Windows, it first tries
617falling back to the system default locale.) This usually means your
618locale settings were wrong, they mention locales your system has never
619heard of, or the locale installation in your system has problems (for
620example, some system files are broken or missing). There are quick and
621temporary fixes to these problems, as well as more thorough and lasting
622fixes.
3e6e419a 623
83fb1bf0
KW
624=head2 Testing for broken locales
625
626If you are building Perl from source, the Perl test suite file
627F<lib/locale.t> can be used to test the locales on your system.
628Setting the environment variable C<PERL_DEBUG_FULL_TEST> to 1
629will cause it to output detailed results. For example, on Linux, you
630could say
631
1d2ab946 632 PERL_DEBUG_FULL_TEST=1 ./perl -T -Ilib lib/locale.t > locale.log 2>&1
83fb1bf0
KW
633
634Besides many other tests, it will test every locale it finds on your
635system to see if they conform to the POSIX standard. If any have
636errors, it will include a summary near the end of the output of which
637locales passed all its tests, and which failed, and why.
638
3e6e419a
JH
639=head2 Temporarily fixing locale problems
640
5a964f20 641The two quickest fixes are either to render Perl silent about any
3e6e419a
JH
642locale inconsistencies or to run Perl under the default locale "C".
643
644Perl's moaning about locale problems can be silenced by setting the
22ff3130 645environment variable C<PERL_BADLANG> to "0" or "".
900bd440
JH
646This method really just sweeps the problem under the carpet: you tell
647Perl to shut up even when Perl sees that something is wrong. Do not
648be surprised if later something locale-dependent misbehaves.
3e6e419a
JH
649
650Perl can be run under the "C" locale by setting the environment
cb88b78e
KW
651variable C<LC_ALL> to "C". This method is perhaps a bit more civilized
652than the C<PERL_BADLANG> approach, but setting C<LC_ALL> (or
5a964f20
TC
653other locale variables) may affect other programs as well, not just
654Perl. In particular, external programs run from within Perl will see
3e6e419a 655these changes. If you make the new settings permanent (read on), all
5a0de581
LM
656programs you run see the changes. See L</"ENVIRONMENT"> for
657the full list of relevant environment variables and L</"USING LOCALES">
e05ffc7d 658for their effects in Perl. Effects in other programs are
cb88b78e 659easily deducible. For example, the variable C<LC_COLLATE> may well affect
b432a672 660your B<sort> program (or whatever the program that arranges "records"
3e6e419a
JH
661alphabetically in your system is called).
662
5a964f20
TC
663You can test out changing these variables temporarily, and if the
664new settings seem to help, put those settings into your shell startup
663d437a 665files. Consult your local documentation for the exact details. For
5a964f20 666Bourne-like shells (B<sh>, B<ksh>, B<bash>, B<zsh>):
3e6e419a
JH
667
668 LC_ALL=en_US.ISO8859-1
669 export LC_ALL
670
5a964f20
TC
671This assumes that we saw the locale "en_US.ISO8859-1" using the commands
672discussed above. We decided to try that instead of the above faulty
673locale "En_US"--and in Cshish shells (B<csh>, B<tcsh>)
3e6e419a
JH
674
675 setenv LC_ALL en_US.ISO8859-1
c47ff5f1 676
663d437a 677or if you have the "env" application you can do (in any shell)
c406981e
JH
678
679 env LC_ALL=en_US.ISO8859-1 perl ...
680
5a964f20 681If you do not know what shell you have, consult your local
3e6e419a
JH
682helpdesk or the equivalent.
683
684=head2 Permanently fixing locale problems
685
5a964f20
TC
686The slower but superior fixes are when you may be able to yourself
687fix the misconfiguration of your own environment variables. The
3e6e419a
JH
688mis(sing)configuration of the whole system's locales usually requires
689the help of your friendly system administrator.
690
5a0de581 691First, see earlier in this document about L</Finding locales>. That tells
5a964f20
TC
692how to find which locales are really supported--and more importantly,
693installed--on your system. In our example error message, environment
694variables affecting the locale are listed in the order of decreasing
695importance (and unset variables do not matter). Therefore, having
696LC_ALL set to "En_US" must have been the bad choice, as shown by the
697error message. First try fixing locale settings listed first.
3e6e419a 698
5a964f20
TC
699Second, if using the listed commands you see something B<exactly>
700(prefix matches do not count and case usually counts) like "En_US"
701without the quotes, then you should be okay because you are using a
702locale name that should be installed and available in your system.
5a0de581 703In this case, see L</Permanently fixing your system's locale configuration>.
3e6e419a 704
4a4eefd0 705=head2 Permanently fixing your system's locale configuration
3e6e419a 706
5a964f20 707This is when you see something like:
3e6e419a
JH
708
709 perl: warning: Please check that your locale settings:
710 LC_ALL = "En_US",
711 LANG = (unset)
712 are supported and installed on your system.
713
714but then cannot see that "En_US" listed by the above-mentioned
5a964f20
TC
715commands. You may see things like "en_US.ISO8859-1", but that isn't
716the same. In this case, try running under a locale
717that you can list and which somehow matches what you tried. The
3e6e419a 718rules for matching locale names are a bit vague because
e05ffc7d 719standardization is weak in this area. See again the
5a0de581 720L</Finding locales> about general rules.
3e6e419a 721
b687b08b 722=head2 Fixing system locale configuration
3e6e419a 723
5a964f20
TC
724Contact a system administrator (preferably your own) and report the exact
725error message you get, and ask them to read this same documentation you
726are now reading. They should be able to check whether there is something
5a0de581 727wrong with the locale configuration of the system. The L</Finding locales>
5a964f20
TC
728section is unfortunately a bit vague about the exact commands and places
729because these things are not that standardized.
3e6e419a 730
5f05dabc
PP
731=head2 The localeconv function
732
39332f68 733The C<POSIX::localeconv()> function allows you to get particulars of the
14280422 734locale-dependent numeric formatting information specified by the current
a835cd47
KW
735underlying C<LC_NUMERIC> and C<LC_MONETARY> locales (regardless of
736whether called from within the scope of C<S<use locale>> or not). (If
737you just want the name of
39332f68 738the current locale for a particular category, use C<POSIX::setlocale()>
5a0de581 739with a single parameter--see L</The setlocale function>.)
5f05dabc
PP
740
741 use POSIX qw(locale_h);
5f05dabc
PP
742
743 # Get a reference to a hash of locale-dependent info
744 $locale_values = localeconv();
745
746 # Output sorted list of the values
747 for (sort keys %$locale_values) {
14280422 748 printf "%-20s = %s\n", $_, $locale_values->{$_}
5f05dabc
PP
749 }
750
39332f68 751C<localeconv()> takes no arguments, and returns B<a reference to> a hash.
5a964f20 752The keys of this hash are variable names for formatting, such as
502a173a 753C<decimal_point> and C<thousands_sep>. The values are the
cea6626f 754corresponding, er, values. See L<POSIX/localeconv> for a longer
502a173a
JH
755example listing the categories an implementation might be expected to
756provide; some provide more and others fewer. You don't need an
39332f68 757explicit C<use locale>, because C<localeconv()> always observes the
502a173a 758current locale.
5f05dabc 759
5a964f20
TC
760Here's a simple-minded example program that rewrites its command-line
761parameters as integers correctly formatted in the current locale:
5f05dabc 762
ef3087ec
KW
763 use POSIX qw(locale_h);
764
765 # Get some of locale's numeric formatting parameters
766 my ($thousands_sep, $grouping) =
767 @{localeconv()}{'thousands_sep', 'grouping'};
768
769 # Apply defaults if values are missing
770 $thousands_sep = ',' unless $thousands_sep;
771
772 # grouping and mon_grouping are packed lists
773 # of small integers (characters) telling the
774 # grouping (thousand_seps and mon_thousand_seps
775 # being the group dividers) of numbers and
776 # monetary quantities. The integers' meanings:
777 # 255 means no more grouping, 0 means repeat
778 # the previous grouping, 1-254 means use that
779 # as the current grouping. Grouping goes from
780 # right to left (low to high digits). In the
781 # below we cheat slightly by never using anything
782 # else than the first grouping (whatever that is).
783 if ($grouping) {
784 @grouping = unpack("C*", $grouping);
785 } else {
786 @grouping = (3);
787 }
788
789 # Format command line params for current locale
790 for (@ARGV) {
791 $_ = int; # Chop non-integer part
792 1 while
793 s/(\d)(\d{$grouping[0]}($|$thousands_sep))/$1$thousands_sep$2/;
794 print "$_";
795 }
796 print "\n";
5f05dabc 797
03ceeedf
KW
798Note that if the platform doesn't have C<LC_NUMERIC> and/or
799C<LC_MONETARY> available or enabled, the corresponding elements of the
800hash will be missing.
801
74c76037 802=head2 I18N::Langinfo
4bbcc6e8
JH
803
804Another interface for querying locale-dependent information is the
39332f68 805C<I18N::Langinfo::langinfo()> function, available at least in Unix-like
4bbcc6e8
JH
806systems and VMS.
807
39332f68
KW
808The following example will import the C<langinfo()> function itself and
809three constants to be used as arguments to C<langinfo()>: a constant for
74c76037
JH
810the abbreviated first day of the week (the numbering starts from
811Sunday = 1) and two more constants for the affirmative and negative
812answers for a yes/no question in the current locale.
4bbcc6e8 813
74c76037 814 use I18N::Langinfo qw(langinfo ABDAY_1 YESSTR NOSTR);
4bbcc6e8 815
ef3087ec
KW
816 my ($abday_1, $yesstr, $nostr)
817 = map { langinfo } qw(ABDAY_1 YESSTR NOSTR);
4bbcc6e8 818
74c76037 819 print "$abday_1? [$yesstr/$nostr] ";
4bbcc6e8 820
74c76037
JH
821In other words, in the "C" (or English) locale the above will probably
822print something like:
823
e05ffc7d 824 Sun? [yes/no]
4bbcc6e8
JH
825
826See L<I18N::Langinfo> for more information.
827
5f05dabc
PP
828=head1 LOCALE CATEGORIES
829
5a964f20
TC
830The following subsections describe basic locale categories. Beyond these,
831some combination categories allow manipulation of more than one
5a0de581 832basic category at a time. See L</"ENVIRONMENT"> for a discussion of these.
5f05dabc 833
ff52fcf1 834=head2 Category C<LC_COLLATE>: Collation: Text Comparisons and Sorting
5f05dabc 835
d6ded950
KW
836In the scope of a S<C<use locale>> form that includes collation, Perl
837looks to the C<LC_COLLATE>
5a964f20 838environment variable to determine the application's notions on collation
b4ffc3db
TC
839(ordering) of characters. For example, "b" follows "a" in Latin
840alphabets, but where do "E<aacute>" and "E<aring>" belong? And while
f87fa335 841"color" follows "chocolate" in English, what about in traditional Spanish?
5f05dabc 842
60f0fa02 843The following collations all make sense and you may meet any of them
dbf3c4d7 844if you C<"use locale">.
60f0fa02
JH
845
846 A B C D E a b c d e
35316ca3 847 A a B b C c D d E e
60f0fa02
JH
848 a A b B c C d D e E
849 a b c d e A B C D E
850
f1cbbd6e 851Here is a code snippet to tell what "word"
5a964f20 852characters are in the current locale, in that locale's order:
5f05dabc
PP
853
854 use locale;
35316ca3 855 print +(sort grep /\w/, map { chr } 0..255), "\n";
5f05dabc 856
14280422
DD
857Compare this with the characters that you see and their order if you
858state explicitly that the locale should be ignored:
5f05dabc
PP
859
860 no locale;
35316ca3 861 print +(sort grep /\w/, map { chr } 0..255), "\n";
5f05dabc
PP
862
863This machine-native collation (which is what you get unless S<C<use
864locale>> has appeared earlier in the same block) must be used for
865sorting raw binary data, whereas the locale-dependent collation of the
b0c42ed9 866first example is useful for natural text.
5f05dabc 867
5a0de581 868As noted in L</USING LOCALES>, C<cmp> compares according to the current
14280422 869collation locale when C<use locale> is in effect, but falls back to a
de108802 870char-by-char comparison for strings that the locale says are equal. You
39332f68 871can use C<POSIX::strcoll()> if you don't want this fall-back:
14280422
DD
872
873 use POSIX qw(strcoll);
874 $equal_in_locale =
875 !strcoll("space and case ignored", "SpaceAndCaseIgnored");
876
39332f68 877C<$equal_in_locale> will be true if the collation locale specifies a
5a964f20 878dictionary-like ordering that ignores space characters completely and
9e3a2af8 879which folds case.
14280422 880
a4a439fb
KW
881Perl uses the platform's C library collation functions C<strcoll()> and
882C<strxfrm()>. That means you get whatever they give. On some
883platforms, these functions work well on UTF-8 locales, giving
884a reasonable default collation for the code points that are important in
885that locale. (And if they aren't working well, the problem may only be
886that the locale definition is deficient, so can be fixed by using a
887better definition file. Unicode's definitions (see L</Freely available
888locale definitions>) provide reasonable UTF-8 locale collation
889definitions.) Starting in Perl v5.26, Perl's use of these functions has
890been made more seamless. This may be sufficient for your needs. For
891more control, and to make sure strings containing any code point (not
892just the ones important in the locale) collate properly, the
893L<Unicode::Collate> module is suggested.
894
895In non-UTF-8 locales (hence single byte), code points above 0xFF are
896technically invalid. But if present, again starting in v5.26, they will
897collate to the same position as the highest valid code point does. This
898generally gives good results, but the collation order may be skewed if
899the valid code point gets special treatment when it forms particular
900sequences with other characters as defined by the locale.
901When two strings collate identically, the code point order is used as a
902tie breaker.
903
904If Perl detects that there are problems with the locale collation order,
905it reverts to using non-locale collation rules for that locale.
31f05a37 906
5a964f20 907If you have a single string that you want to check for "equality in
14280422 908locale" against several others, you might think you could gain a little
39332f68 909efficiency by using C<POSIX::strxfrm()> in conjunction with C<eq>:
14280422
DD
910
911 use POSIX qw(strxfrm);
912 $xfrm_string = strxfrm("Mixed-case string");
913 print "locale collation ignores spaces\n"
914 if $xfrm_string eq strxfrm("Mixed-casestring");
915 print "locale collation ignores hyphens\n"
916 if $xfrm_string eq strxfrm("Mixedcase string");
917 print "locale collation ignores case\n"
918 if $xfrm_string eq strxfrm("mixed-case string");
919
39332f68 920C<strxfrm()> takes a string and maps it into a transformed string for use
de108802 921in char-by-char comparisons against other transformed strings during
14280422 922collation. "Under the hood", locale-affected Perl comparison operators
39332f68
KW
923call C<strxfrm()> for both operands, then do a char-by-char
924comparison of the transformed strings. By calling C<strxfrm()> explicitly
14280422 925and using a non locale-affected comparison, the example attempts to save
5a964f20 926a couple of transformations. But in fact, it doesn't save anything: Perl
2ae324a7 927magic (see L<perlguts/Magic Variables>) creates the transformed version of a
5a964f20 928string the first time it's needed in a comparison, then keeps this version around
14280422 929in case it's needed again. An example rewritten the easy way with
e38874e2 930C<cmp> runs just about as fast. It also copes with null characters
39332f68 931embedded in strings; if you call C<strxfrm()> directly, it treats the first
0c880285 932null it finds as a terminator. Don't expect the transformed strings
5a964f20 933it produces to be portable across systems--or even from one revision
39332f68 934of your operating system to the next. In short, don't call C<strxfrm()>
e38874e2 935directly: let Perl do it for you.
14280422 936
5a964f20 937Note: C<use locale> isn't shown in some of these examples because it isn't
dfcc8045
KW
938needed: C<strcoll()> and C<strxfrm()> are POSIX functions
939which use the standard system-supplied C<libc> functions that
940always obey the current C<LC_COLLATE> locale.
5f05dabc 941
cb88b78e 942=head2 Category C<LC_CTYPE>: Character Types
5f05dabc 943
d6ded950
KW
944In the scope of a S<C<use locale>> form that includes C<LC_CTYPE>, Perl
945obeys the C<LC_CTYPE> locale
14280422 946setting. This controls the application's notion of which characters are
ebc3223b
KW
947alphabetic, numeric, punctuation, I<etc>. This affects Perl's C<\w>
948regular expression metanotation,
f1cbbd6e 949which stands for alphanumeric characters--that is, alphabetic,
ebc3223b
KW
950numeric, and the platform's native underscore.
951(Consult L<perlre> for more information about
14280422 952regular expressions.) Thanks to C<LC_CTYPE>, depending on your locale
b4ffc3db
TC
953setting, characters like "E<aelig>", "E<eth>", "E<szlig>", and
954"E<oslash>" may be understood as C<\w> characters.
ebc3223b
KW
955It also affects things like C<\s>, C<\D>, and the POSIX character
956classes, like C<[[:graph:]]>. (See L<perlrecharclass> for more
957information on all these.)
5f05dabc 958
2c268ad5 959The C<LC_CTYPE> locale also provides the map used in transliterating
68dc0745 960characters between lower and uppercase. This affects the case-mapping
663d437a
KW
961functions--C<fc()>, C<lc()>, C<lcfirst()>, C<uc()>, and C<ucfirst()>;
962case-mapping
b9cc4f69 963interpolation with C<\F>, C<\l>, C<\L>, C<\u>, or C<\U> in double-quoted
26a9b33a 964strings and C<s///> substitutions; and case-insensitive regular expression
e38874e2
DD
965pattern matching using the C<i> modifier.
966
31f05a37
KW
967Starting in v5.20, Perl supports UTF-8 locales for C<LC_CTYPE>, but
968otherwise Perl only supports single-byte locales, such as the ISO 8859
969series. This means that wide character locales, for example for Asian
b5eea289
KW
970languages, are not well-supported. Use of these locales may cause core
971dumps. If the platform has the capability for Perl to detect such a
972locale, starting in Perl v5.22, L<Perl will warn, default
973enabled|warnings/Category Hierarchy>, using the C<locale> warning
974category, whenever such a locale is switched into. The UTF-8 locale
975support is actually a
31f05a37 976superset of POSIX locales, because it is really full Unicode behavior
13af4fd9
KW
977as if no C<LC_CTYPE> locale were in effect at all (except for tainting;
978see L</SECURITY>). POSIX locales, even UTF-8 ones,
31f05a37
KW
979are lacking certain concepts in Unicode, such as the idea that changing
980the case of a character could expand to be more than one character.
981Perl in a UTF-8 locale, will give you that expansion. Prior to v5.20,
982Perl treated a UTF-8 locale on some platforms like an ISO 8859-1 one,
983with some restrictions, and on other platforms more like the "C" locale.
984For releases v5.16 and v5.18, C<S<use locale 'not_characters>> could be
985used as a workaround for this (see L</Unicode and UTF-8>).
986
5d63e270 987Note that there are quite a few things that are unaffected by the
663d437a
KW
988current locale. Any literal character is the native character for the
989given platform. Hence 'A' means the character at code point 65 on ASCII
990platforms, and 193 on EBCDIC. That may or may not be an 'A' in the
991current locale, if that locale even has an 'A'.
992Similarly, all the escape sequences for particular characters,
5d63e270
KW
993C<\n> for example, always mean the platform's native one. This means,
994for example, that C<\N> in regular expressions (every character
1d2ab946 995but new-line) works on the platform character set.
5d63e270 996
8c6180a9
KW
997Starting in v5.22, Perl will by default warn when switching into a
998locale that redefines any ASCII printable character (plus C<\t> and
84035de0
KW
999C<\n>) into a different class than expected. This is likely to
1000happen on modern locales only on EBCDIC platforms, where, for example,
1001a CCSID 0037 locale on a CCSID 1047 machine moves C<"[">, but it can
1002happen on ASCII platforms with the ISO 646 and other
8c6180a9
KW
10037-bit locales that are essentially obsolete. Things may still work,
1004depending on what features of Perl are used by the program. For
1005example, in the example from above where C<"|"> becomes a C<\w>, and
1006there are no regular expressions where this matters, the program may
1007still work properly. The warning lists all the characters that
1008it can determine could be adversely affected.
1009
14280422
DD
1010B<Note:> A broken or malicious C<LC_CTYPE> locale definition may result
1011in clearly ineligible characters being considered to be alphanumeric by
e199995e 1012your application. For strict matching of (mundane) ASCII letters and
5a964f20 1013digits--for example, in command strings--locale-aware applications
5a0de581 1014should use C<\w> with the C</a> regular expression modifier. See L</"SECURITY">.
5f05dabc 1015
fee33030 1016=head2 Category C<LC_NUMERIC>: Numeric Formatting
5f05dabc 1017
d6ded950
KW
1018After a proper C<POSIX::setlocale()> call, and within the scope of
1019of a C<use locale> form that includes numerics, Perl obeys the
1020C<LC_NUMERIC> locale information, which controls an application's idea
1021of how numbers should be formatted for human readability.
b960a36e 1022In most implementations the only effect is to
b4ffc3db 1023change the character used for the decimal point--perhaps from "." to ",".
b960a36e 1024The functions aren't aware of such niceties as thousands separation and
5a0de581 1025so on. (See L</The localeconv function> if you care about these things.)
5a964f20 1026
b960a36e
KW
1027 use POSIX qw(strtod setlocale LC_NUMERIC);
1028 use locale;
5f05dabc 1029
b960a36e 1030 setlocale LC_NUMERIC, "";
14280422 1031
b960a36e 1032 $n = 5/2; # Assign numeric 2.5 to $n
5f05dabc 1033
b960a36e 1034 $a = " $n"; # Locale-dependent conversion to string
5f05dabc 1035
b960a36e 1036 print "half five is $n\n"; # Locale-dependent output
5f05dabc 1037
b960a36e 1038 printf "half five is %g\n", $n; # Locale-dependent output
5f05dabc 1039
b960a36e
KW
1040 print "DECIMAL POINT IS COMMA\n"
1041 if $n == (strtod("2,5"))[0]; # Locale-dependent conversion
5f05dabc 1042
4bbcc6e8
JH
1043See also L<I18N::Langinfo> and C<RADIXCHAR>.
1044
cb88b78e 1045=head2 Category C<LC_MONETARY>: Formatting of monetary amounts
5f05dabc 1046
e199995e 1047The C standard defines the C<LC_MONETARY> category, but not a function
5a964f20 1048that is affected by its contents. (Those with experience of standards
b0c42ed9 1049committees will recognize that the working group decided to punt on the
fa9b773e
KW
1050issue.) Consequently, Perl essentially takes no notice of it. If you
1051really want to use C<LC_MONETARY>, you can query its contents--see
5a0de581 1052L</The localeconv function>--and use the information that it returns in your
e05ffc7d
KW
1053application's own formatting of currency amounts. However, you may well
1054find that the information, voluminous and complex though it may be, still
1055does not quite meet your requirements: currency formatting is a hard nut
13a2d996 1056to crack.
5f05dabc 1057
4bbcc6e8
JH
1058See also L<I18N::Langinfo> and C<CRNCYSTR>.
1059
ff52fcf1 1060=head2 Category C<LC_TIME>: Respresentation of time
5f05dabc 1061
39332f68 1062Output produced by C<POSIX::strftime()>, which builds a formatted
5f05dabc
PP
1063human-readable date/time string, is affected by the current C<LC_TIME>
1064locale. Thus, in a French locale, the output produced by the C<%B>
1065format element (full month name) for the first month of the year would
5a964f20 1066be "janvier". Here's how to get a list of long month names in the
5f05dabc
PP
1067current locale:
1068
1069 use POSIX qw(strftime);
14280422
DD
1070 for (0..11) {
1071 $long_month_name[$_] =
1072 strftime("%B", 0, 0, 0, 1, $_, 96);
5f05dabc
PP
1073 }
1074
2619d284
KW
1075Note: C<use locale> isn't needed in this example: C<strftime()> is a POSIX
1076function which uses the standard system-supplied C<libc> function that
1077always obeys the current C<LC_TIME> locale.
5f05dabc 1078
4bbcc6e8 1079See also L<I18N::Langinfo> and C<ABDAY_1>..C<ABDAY_7>, C<DAY_1>..C<DAY_7>,
2a2bf5f4 1080C<ABMON_1>..C<ABMON_12>, and C<ABMON_1>..C<ABMON_12>.
4bbcc6e8 1081
5f05dabc
PP
1082=head2 Other categories
1083
2619d284
KW
1084The remaining locale categories are not currently used by Perl itself.
1085But again note that things Perl interacts with may use these, including
1086extensions outside the standard Perl distribution, and by the
98a6f11e 1087operating system and its utilities. Note especially that the string
1088value of C<$!> and the error messages given by external utilities may
1089be changed by C<LC_MESSAGES>. If you want to have portable error
265f5c4a 1090codes, use C<%!>. See L<Errno>.
14280422
DD
1091
1092=head1 SECURITY
1093
5a964f20 1094Although the main discussion of Perl security issues can be found in
14280422
DD
1095L<perlsec>, a discussion of Perl's locale handling would be incomplete
1096if it did not draw your attention to locale-dependent security issues.
5a964f20
TC
1097Locales--particularly on systems that allow unprivileged users to
1098build their own locales--are untrustworthy. A malicious (or just plain
14280422
DD
1099broken) locale can make a locale-aware application give unexpected
1100results. Here are a few possibilities:
1101
1102=over 4
1103
1104=item *
1105
1106Regular expression checks for safe file names or mail addresses using
5a964f20 1107C<\w> may be spoofed by an C<LC_CTYPE> locale that claims that
dbf3c4d7 1108characters such as C<"E<gt>"> and C<"|"> are alphanumeric.
14280422
DD
1109
1110=item *
1111
e38874e2 1112String interpolation with case-mapping, as in, say, C<$dest =
cb88b78e 1113"C:\U$name.$ext">, may produce dangerous results if a bogus C<LC_CTYPE>
e38874e2
DD
1114case-mapping table is in effect.
1115
1116=item *
1117
14280422
DD
1118A sneaky C<LC_COLLATE> locale could result in the names of students with
1119"D" grades appearing ahead of those with "A"s.
1120
1121=item *
1122
5a964f20 1123An application that takes the trouble to use information in
14280422 1124C<LC_MONETARY> may format debits as if they were credits and vice versa
5a964f20 1125if that locale has been subverted. Or it might make payments in US
14280422
DD
1126dollars instead of Hong Kong dollars.
1127
1128=item *
1129
39332f68 1130The date and day names in dates formatted by C<strftime()> could be
14280422 1131manipulated to advantage by a malicious user able to subvert the
5a964f20 1132C<LC_DATE> locale. ("Look--it says I wasn't in the building on
14280422
DD
1133Sunday.")
1134
1135=back
1136
1137Such dangers are not peculiar to the locale system: any aspect of an
5a964f20 1138application's environment which may be modified maliciously presents
14280422 1139similar challenges. Similarly, they are not specific to Perl: any
5a964f20 1140programming language that allows you to write programs that take
14280422
DD
1141account of their environment exposes you to these issues.
1142
5a964f20
TC
1143Perl cannot protect you from all possibilities shown in the
1144examples--there is no substitute for your own vigilance--but, when
14280422 1145C<use locale> is in effect, Perl uses the tainting mechanism (see
5a964f20 1146L<perlsec>) to mark string results that become locale-dependent, and
14280422 1147which may be untrustworthy in consequence. Here is a summary of the
5a964f20 1148tainting behavior of operators and functions that may be affected by
14280422
DD
1149the locale:
1150
1151=over 4
1152
551e1d92
RB
1153=item *
1154
1155B<Comparison operators> (C<lt>, C<le>, C<ge>, C<gt> and C<cmp>):
14280422
DD
1156
1157Scalar true/false (or less/equal/greater) result is never tainted.
1158
551e1d92
RB
1159=item *
1160
1d2ab946 1161B<Case-mapping interpolation> (with C<\l>, C<\L>, C<\u>, C<\U>, or C<\F>)
e38874e2 1162
d6ded950
KW
1163The result string containing interpolated material is tainted if
1164a C<use locale> form that includes C<LC_CTYPE> is in effect.
e38874e2 1165
551e1d92
RB
1166=item *
1167
1168B<Matching operator> (C<m//>):
14280422
DD
1169
1170Scalar true/false result never tainted.
1171
1d2ab946 1172All subpatterns, either delivered as a list-context result or as C<$1>
d6ded950
KW
1173I<etc>., are tainted if a C<use locale> form that includes
1174C<LC_CTYPE> is in effect, and the subpattern
63baef57
KW
1175regular expression contains a locale-dependent construct. These
1176constructs include C<\w> (to match an alphanumeric character), C<\W>
1177(non-alphanumeric character), C<\b> and C<\B> (word-boundary and
1178non-boundardy, which depend on what C<\w> and C<\W> match), C<\s>
1179(whitespace character), C<\S> (non whitespace character), C<\d> and
1180C<\D> (digits and non-digits), and the POSIX character classes, such as
1181C<[:alpha:]> (see L<perlrecharclass/POSIX Character Classes>).
1182
1183Tainting is also likely if the pattern is to be matched
1184case-insensitively (via C</i>). The exception is if all the code points
1185to be matched this way are above 255 and do not have folds under Unicode
1186rules to below 256. Tainting is not done for these because Perl
1187only uses Unicode rules for such code points, and those rules are the
1188same no matter what the current locale.
1189
1d2ab946
KW
1190The matched-pattern variables, C<$&>, C<$`> (pre-match), C<$'>
1191(post-match), and C<$+> (last match) also are tainted.
14280422 1192
551e1d92
RB
1193=item *
1194
1195B<Substitution operator> (C<s///>):
14280422 1196
e38874e2 1197Has the same behavior as the match operator. Also, the left
d6ded950
KW
1198operand of C<=~> becomes tainted when a C<use locale>
1199form that includes C<LC_CTYPE> is in effect, if modified as
66cbab2c 1200a result of a substitution based on a regular
1d2ab946
KW
1201expression match involving any of the things mentioned in the previous
1202item, or of case-mapping, such as C<\l>, C<\L>,C<\u>, C<\U>, or C<\F>.
14280422 1203
551e1d92
RB
1204=item *
1205
39332f68 1206B<Output formatting functions> (C<printf()> and C<write()>):
14280422 1207
3cf03d68
JH
1208Results are never tainted because otherwise even output from print,
1209for example C<print(1/7)>, should be tainted if C<use locale> is in
1210effect.
14280422 1211
551e1d92
RB
1212=item *
1213
39332f68 1214B<Case-mapping functions> (C<lc()>, C<lcfirst()>, C<uc()>, C<ucfirst()>):
14280422 1215
d6ded950
KW
1216Results are tainted if a C<use locale> form that includes C<LC_CTYPE> is
1217in effect.
14280422 1218
551e1d92
RB
1219=item *
1220
39332f68
KW
1221B<POSIX locale-dependent functions> (C<localeconv()>, C<strcoll()>,
1222C<strftime()>, C<strxfrm()>):
14280422
DD
1223
1224Results are never tainted.
1225
14280422
DD
1226=back
1227
1228Three examples illustrate locale-dependent tainting.
1229The first program, which ignores its locale, won't run: a value taken
54310121 1230directly from the command line may not be used to name an output file
14280422
DD
1231when taint checks are enabled.
1232
1233 #/usr/local/bin/perl -T
1234 # Run with taint checking
1235
54310121 1236 # Command line sanity check omitted...
14280422
DD
1237 $tainted_output_file = shift;
1238
1239 open(F, ">$tainted_output_file")
3183d96c 1240 or warn "Open of $tainted_output_file failed: $!\n";
14280422
DD
1241
1242The program can be made to run by "laundering" the tainted value through
5a964f20
TC
1243a regular expression: the second example--which still ignores locale
1244information--runs, creating the file named on its command line
14280422
DD
1245if it can.
1246
1247 #/usr/local/bin/perl -T
1248
1249 $tainted_output_file = shift;
1250 $tainted_output_file =~ m%[\w/]+%;
1251 $untainted_output_file = $&;
1252
1253 open(F, ">$untainted_output_file")
1254 or warn "Open of $untainted_output_file failed: $!\n";
1255
5a964f20 1256Compare this with a similar but locale-aware program:
14280422
DD
1257
1258 #/usr/local/bin/perl -T
1259
1260 $tainted_output_file = shift;
1261 use locale;
1262 $tainted_output_file =~ m%[\w/]+%;
1263 $localized_output_file = $&;
1264
1265 open(F, ">$localized_output_file")
1266 or warn "Open of $localized_output_file failed: $!\n";
1267
1d2ab946 1268This third program fails to run because C<$&> is tainted: it is the result
5a964f20 1269of a match involving C<\w> while C<use locale> is in effect.
5f05dabc
PP
1270
1271=head1 ENVIRONMENT
1272
1273=over 12
1274
ee1ec05f
KW
1275=item PERL_SKIP_LOCALE_INIT
1276
c5c88224
KW
1277This environment variable, available starting in Perl v5.20, if set
1278(to any value), tells Perl to not use the rest of the
ee1ec05f
KW
1279environment variables to initialize with. Instead, Perl uses whatever
1280the current locale settings are. This is particularly useful in
1281embedded environments, see
1282L<perlembed/Using embedded Perl with POSIX locales>.
1283
5f05dabc
PP
1284=item PERL_BADLANG
1285
14280422 1286A string that can suppress Perl's warning about failed locale settings
54310121 1287at startup. Failure can occur if the locale support in the operating
5a964f20 1288system is lacking (broken) in some way--or if you mistyped the name of
900bd440 1289a locale when you set up your environment. If this environment
22ff3130
HS
1290variable is absent, or has a value other than "0" or "", Perl will
1291complain about locale setting failures.
5f05dabc 1292
cb88b78e 1293B<NOTE>: C<PERL_BADLANG> only gives you a way to hide the warning message.
14280422
DD
1294The message tells about some problem in your system's locale support,
1295and you should investigate what the problem is.
5f05dabc
PP
1296
1297=back
1298
1299The following environment variables are not specific to Perl: They are
39332f68 1300part of the standardized (ISO C, XPG4, POSIX 1.c) C<setlocale()> method
b385bb4d
KW
1301for controlling an application's opinion on data. Windows is non-POSIX,
1302but Perl arranges for the following to work as described anyway.
65ebb059
KW
1303If the locale given by an environment variable is not valid, Perl tries
1304the next lower one in priority. If none are valid, on Windows, the
1305system default locale is then tried. If all else fails, the C<"C">
1306locale is used. If even that doesn't work, something is badly broken,
c5e9a8e7 1307but Perl tries to forge ahead with whatever the locale settings might
65ebb059 1308be.
5f05dabc
PP
1309
1310=over 12
1311
cb88b78e 1312=item C<LC_ALL>
5f05dabc 1313
5a964f20 1314C<LC_ALL> is the "override-all" locale environment variable. If
5f05dabc
PP
1315set, it overrides all the rest of the locale environment variables.
1316
cb88b78e 1317=item C<LANGUAGE>
528d65ad
JH
1318
1319B<NOTE>: C<LANGUAGE> is a GNU extension, it affects you only if you
1320are using the GNU libc. This is the case if you are using e.g. Linux.
e1020413 1321If you are using "commercial" Unixes you are most probably I<not>
22b6f60d
JH
1322using GNU libc and you can ignore C<LANGUAGE>.
1323
1324However, in the case you are using C<LANGUAGE>: it affects the
1325language of informational, warning, and error messages output by
1326commands (in other words, it's like C<LC_MESSAGES>) but it has higher
96090e4f 1327priority than C<LC_ALL>. Moreover, it's not a single value but
22b6f60d
JH
1328instead a "path" (":"-separated list) of I<languages> (not locales).
1329See the GNU C<gettext> library documentation for more information.
528d65ad 1330
3ee1a09c 1331=item C<LC_CTYPE>
5f05dabc
PP
1332
1333In the absence of C<LC_ALL>, C<LC_CTYPE> chooses the character type
1334locale. In the absence of both C<LC_ALL> and C<LC_CTYPE>, C<LANG>
1335chooses the character type locale.
1336
cb88b78e 1337=item C<LC_COLLATE>
5f05dabc 1338
14280422
DD
1339In the absence of C<LC_ALL>, C<LC_COLLATE> chooses the collation
1340(sorting) locale. In the absence of both C<LC_ALL> and C<LC_COLLATE>,
1341C<LANG> chooses the collation locale.
5f05dabc 1342
cb88b78e 1343=item C<LC_MONETARY>
5f05dabc 1344
14280422
DD
1345In the absence of C<LC_ALL>, C<LC_MONETARY> chooses the monetary
1346formatting locale. In the absence of both C<LC_ALL> and C<LC_MONETARY>,
1347C<LANG> chooses the monetary formatting locale.
5f05dabc 1348
cb88b78e 1349=item C<LC_NUMERIC>
5f05dabc
PP
1350
1351In the absence of C<LC_ALL>, C<LC_NUMERIC> chooses the numeric format
1352locale. In the absence of both C<LC_ALL> and C<LC_NUMERIC>, C<LANG>
1353chooses the numeric format.
1354
cb88b78e 1355=item C<LC_TIME>
5f05dabc 1356
14280422
DD
1357In the absence of C<LC_ALL>, C<LC_TIME> chooses the date and time
1358formatting locale. In the absence of both C<LC_ALL> and C<LC_TIME>,
1359C<LANG> chooses the date and time formatting locale.
5f05dabc 1360
cb88b78e 1361=item C<LANG>
5f05dabc 1362
14280422
DD
1363C<LANG> is the "catch-all" locale environment variable. If it is set, it
1364is used as the last resort after the overall C<LC_ALL> and the
3ee1a09c 1365category-specific C<LC_I<foo>>.
5f05dabc
PP
1366
1367=back
1368
7e4353e9
RGS
1369=head2 Examples
1370
cb88b78e 1371The C<LC_NUMERIC> controls the numeric output:
7e4353e9 1372
ef3087ec
KW
1373 use locale;
1374 use POSIX qw(locale_h); # Imports setlocale() and the LC_ constants.
1375 setlocale(LC_NUMERIC, "fr_FR") or die "Pardon";
1376 printf "%g\n", 1.23; # If the "fr_FR" succeeded, probably shows 1,23.
7e4353e9 1377
39332f68 1378and also how strings are parsed by C<POSIX::strtod()> as numbers:
7e4353e9 1379
ef3087ec
KW
1380 use locale;
1381 use POSIX qw(locale_h strtod);
1382 setlocale(LC_NUMERIC, "de_DE") or die "Entschuldigung";
1383 my $x = strtod("2,34") + 5;
1384 print $x, "\n"; # Probably shows 7,34.
7e4353e9 1385
5f05dabc
PP
1386=head1 NOTES
1387
b960a36e
KW
1388=head2 String C<eval> and C<LC_NUMERIC>
1389
1390A string L<eval|perlfunc/eval EXPR> parses its expression as standard
1391Perl. It is therefore expecting the decimal point to be a dot. If
1392C<LC_NUMERIC> is set to have this be a comma instead, the parsing will
1393be confused, perhaps silently.
1394
1395 use locale;
1396 use POSIX qw(locale_h);
1397 setlocale(LC_NUMERIC, "fr_FR") or die "Pardon";
1398 my $a = 1.2;
1399 print eval "$a + 1.5";
1400 print "\n";
1401
1402prints C<13,5>. This is because in that locale, the comma is the
1403decimal point character. The C<eval> thus expands to:
1404
1405 eval "1,2 + 1.5"
1406
1407and the result is not what you likely expected. No warnings are
1408generated. If you do string C<eval>'s within the scope of
1409S<C<use locale>>, you should instead change the C<eval> line to do
1410something like:
1411
1412 print eval "no locale; $a + 1.5";
1413
1414This prints C<2.7>.
1415
d6ded950
KW
1416You could also exclude C<LC_NUMERIC>, if you don't need it, by
1417
1418 use locale ':!numeric';
1419
5f05dabc
PP
1420=head2 Backward compatibility
1421
b0c42ed9 1422Versions of Perl prior to 5.004 B<mostly> ignored locale information,
5a964f20
TC
1423generally behaving as if something similar to the C<"C"> locale were
1424always in force, even if the program environment suggested otherwise
5a0de581 1425(see L</The setlocale function>). By default, Perl still behaves this
5a964f20
TC
1426way for backward compatibility. If you want a Perl application to pay
1427attention to locale information, you B<must> use the S<C<use locale>>
5a0de581 1428pragma (see L</The "use locale" pragma>) or, in the unlikely event
062ca197 1429that you want to do so for just pattern matching, the
70709c68
KW
1430C</l> regular expression modifier (see L<perlre/Character set
1431modifiers>) to instruct it to do so.
b0c42ed9
JH
1432
1433Versions of Perl from 5.002 to 5.003 did use the C<LC_CTYPE>
5a964f20
TC
1434information if available; that is, C<\w> did understand what
1435were the letters according to the locale environment variables.
b0c42ed9
JH
1436The problem was that the user had no control over the feature:
1437if the C library supported locales, Perl used them.
1438
1439=head2 I18N:Collate obsolete
1440
5a964f20 1441In versions of Perl prior to 5.004, per-locale collation was possible
b0c42ed9
JH
1442using the C<I18N::Collate> library module. This module is now mildly
1443obsolete and should be avoided in new applications. The C<LC_COLLATE>
1444functionality is now integrated into the Perl core language: One can
1445use locale-specific scalar data completely normally with C<use locale>,
1446so there is no longer any need to juggle with the scalar references of
1447C<I18N::Collate>.
5f05dabc 1448
14280422 1449=head2 Sort speed and memory use impacts
5f05dabc
PP
1450
1451Comparing and sorting by locale is usually slower than the default
14280422
DD
1452sorting; slow-downs of two to four times have been observed. It will
1453also consume more memory: once a Perl scalar variable has participated
1454in any string comparison or sorting operation obeying the locale
1455collation rules, it will take 3-15 times more memory than before. (The
1456exact multiplier depends on the string's contents, the operating system
1457and the locale.) These downsides are dictated more by the operating
1458system's implementation of the locale system than by Perl.
5f05dabc 1459
5f05dabc
PP
1460=head2 Freely available locale definitions
1461
66cbab2c
KW
1462The Unicode CLDR project extracts the POSIX portion of many of its
1463locales, available at
1464
e784ce68
KW
1465 http://unicode.org/Public/cldr/2.0.1/
1466
1467(Newer versions of CLDR require you to compute the POSIX data yourself.
1468See L<http://unicode.org/Public/cldr/latest/>.)
66cbab2c 1469
08d7a6b2
LB
1470There is a large collection of locale definitions at:
1471
1472 http://std.dkuug.dk/i18n/WG15-collection/locales/
1473
1474You should be aware that it is
14280422 1475unsupported, and is not claimed to be fit for any purpose. If your
5a964f20 1476system allows installation of arbitrary locales, you may find the
14280422
DD
1477definitions useful as they are, or as a basis for the development of
1478your own locales.
5f05dabc 1479
14280422 1480=head2 I18n and l10n
5f05dabc 1481
b0c42ed9
JH
1482"Internationalization" is often abbreviated as B<i18n> because its first
1483and last letters are separated by eighteen others. (You may guess why
1484the internalin ... internaliti ... i18n tends to get abbreviated.) In
1485the same way, "localization" is often abbreviated to B<l10n>.
14280422
DD
1486
1487=head2 An imperfect standard
1488
1489Internationalization, as defined in the C and POSIX standards, can be
e9bc6d6b
KW
1490criticized as incomplete and ungainly. They also have a tendency, like
1491standards groups, to divide the world into nations, when we all know
1492that the world can equally well be divided into bankers, bikers, gamers,
1493and so on.
5f05dabc 1494
b310b053
JH
1495=head1 Unicode and UTF-8
1496
7ee2ae1e 1497The support of Unicode is new starting from Perl version v5.6, and more fully
31f05a37
KW
1498implemented in versions v5.8 and later. See L<perluniintro>.
1499
a4a439fb
KW
1500Starting in Perl v5.20, UTF-8 locales are supported in Perl, except
1501C<LC_COLLATE> is only partially supported; collation support is improved
1502in Perl v5.26 to a level that may be sufficient for your needs
1503(see L</Category C<LC_COLLATE>: Collation: Text Comparisons and Sorting>).
1504
1505If you have Perl v5.16 or v5.18 and can't upgrade, you can use
66cbab2c
KW
1506
1507 use locale ':not_characters';
1508
1509When this form of the pragma is used, only the non-character portions of
1510locales are used by Perl, for example C<LC_NUMERIC>. Perl assumes that
1511you have translated all the characters it is to operate on into Unicode
1512(actually the platform's native character set (ASCII or EBCDIC) plus
1513Unicode). For data in files, this can conveniently be done by also
1514specifying
1515
1516 use open ':locale';
1517
1518This pragma arranges for all inputs from files to be translated into
1519Unicode from the current locale as specified in the environment (see
1520L</ENVIRONMENT>), and all outputs to files to be translated back
1521into the locale. (See L<open>). On a per-filehandle basis, you can
1522instead use the L<PerlIO::locale> module, or the L<Encode::Locale>
1523module, both available from CPAN. The latter module also has methods to
1524ease the handling of C<ARGV> and environment variables, and can be used
31f05a37 1525on individual strings. If you know that all your locales will be
66cbab2c
KW
1526UTF-8, as many are these days, you can use the L<B<-C>|perlrun/-C>
1527command line switch.
1528
1529This form of the pragma allows essentially seamless handling of locales
31f05a37 1530with Unicode. The collation order will be by Unicode code point order.
a4a439fb 1531L<Unicode::Collate> can be used to get Unicode rules collation.
66cbab2c 1532
31f05a37
KW
1533All the modules and switches just described can be used in v5.20 with
1534just plain C<use locale>, and, should the input locales not be UTF-8,
1535you'll get the less than ideal behavior, described below, that you get
1536with pre-v5.16 Perls, or when you use the locale pragma without the
1537C<:not_characters> parameter in v5.16 and v5.18. If you are using
1538exclusively UTF-8 locales in v5.20 and higher, the rest of this section
1539does not apply to you.
1540
1541There are two cases, multi-byte and single-byte locales. First
1542multi-byte:
1543
1544The only multi-byte (or wide character) locale that Perl is ever likely
1545to support is UTF-8. This is due to the difficulty of implementation,
1546the fact that high quality UTF-8 locales are now published for every
e784ce68
KW
1547area of the world (L<http://unicode.org/Public/cldr/2.0.1/> for
1548ones that are already set-up, but from an earlier version;
1549L<http://unicode.org/Public/cldr/latest/> for the most up-to-date, but
1550you have to extract the POSIX information yourself), and that
31f05a37
KW
1551failing all that you can use the L<Encode> module to translate to/from
1552your locale. So, you'll have to do one of those things if you're using
1553one of these locales, such as Big5 or Shift JIS. For UTF-8 locales, in
1554Perls (pre v5.20) that don't have full UTF-8 locale support, they may
1555work reasonably well (depending on your C library implementation)
1556simply because both
dc4bfc4b
KW
1557they and Perl store characters that take up multiple bytes the same way.
1558However, some, if not most, C library implementations may not process
1559the characters in the upper half of the Latin-1 range (128 - 255)
cb88b78e 1560properly under C<LC_CTYPE>. To see if a character is a particular type
dc4bfc4b
KW
1561under a locale, Perl uses the functions like C<isalnum()>. Your C
1562library may not work for UTF-8 locales with those functions, instead
dbf3c4d7
KW
1563only working under the newer wide library functions like C<iswalnum()>,
1564which Perl does not use.
1565These multi-byte locales are treated like single-byte locales, and will
1566have the restrictions described below. Starting in Perl v5.22 a warning
1567message is raised when Perl detects a multi-byte locale that it doesn't
1568fully support.
e199995e 1569
31f05a37 1570For single-byte locales,
e199995e 1571Perl generally takes the tack to use locale rules on code points that can fit
66cbab2c
KW
1572in a single byte, and Unicode rules for those that can't (though this
1573isn't uniformly applied, see the note at the end of this section). This
1574prevents many problems in locales that aren't UTF-8. Suppose the locale
1575is ISO8859-7, Greek. The character at 0xD7 there is a capital Chi. But
1576in the ISO8859-1 locale, Latin1, it is a multiplication sign. The POSIX
1577regular expression character class C<[[:alpha:]]> will magically match
15780xD7 in the Greek locale but not in the Latin one.
e199995e 1579
1d2ab946 1580However, there are places where this breaks down. Certain Perl constructs are
b4ffc3db
TC
1581for Unicode only, such as C<\p{Alpha}>. They assume that 0xD7 always has its
1582Unicode meaning (or the equivalent on EBCDIC platforms). Since Latin1 is a
1583subset of Unicode and 0xD7 is the multiplication sign in both Latin1 and
1584Unicode, C<\p{Alpha}> will never match it, regardless of locale. A similar
0c880285 1585issue occurs with C<\N{...}>. Prior to v5.20, it is therefore a bad
31f05a37 1586idea to use C<\p{}> or
66cbab2c 1587C<\N{}> under plain C<use locale>--I<unless> you can guarantee that the
dbf3c4d7 1588locale will be ISO8859-1. Use POSIX character classes instead.
66cbab2c
KW
1589
1590Another problem with this approach is that operations that cross the
1591single byte/multiple byte boundary are not well-defined, and so are
4a70680a 1592disallowed. (This boundary is between the codepoints at 255/256.)
66cbab2c
KW
1593For example, lower casing LATIN CAPITAL LETTER Y WITH DIAERESIS (U+0178)
1594should return LATIN SMALL LETTER Y WITH DIAERESIS (U+00FF). But in the
1595Greek locale, for example, there is no character at 0xFF, and Perl
1596has no way of knowing what the character at 0xFF is really supposed to
1597represent. Thus it disallows the operation. In this mode, the
1598lowercase of U+0178 is itself.
1599
1600The same problems ensue if you enable automatic UTF-8-ification of your
e199995e 1601standard file handles, default C<open()> layer, and C<@ARGV> on non-ISO8859-1,
b4ffc3db
TC
1602non-UTF-8 locales (by using either the B<-C> command line switch or the
1603C<PERL_UNICODE> environment variable; see L<perlrun>).
1604Things are read in as UTF-8, which would normally imply a Unicode
1605interpretation, but the presence of a locale causes them to be interpreted
1606in that locale instead. For example, a 0xD7 code point in the Unicode
1607input, which should mean the multiplication sign, won't be interpreted by
66cbab2c 1608Perl that way under the Greek locale. This is not a problem
b4ffc3db 1609I<provided> you make certain that all locales will always and only be either
66cbab2c 1610an ISO8859-1, or, if you don't have a deficient C library, a UTF-8 locale.
b4ffc3db 1611
1d2ab946
KW
1612Still another problem is that this approach can lead to two code
1613points meaning the same character. Thus in a Greek locale, both U+03A7
1614and U+00D7 are GREEK CAPITAL LETTER CHI.
1615
613abc6d
KW
1616Because of all these problems, starting in v5.22, Perl will raise a
1617warning if a multi-byte (hence Unicode) code point is used when a
1618single-byte locale is in effect. (Although it doesn't check for this if
1619doing so would unreasonably slow execution down.)
1620
b4ffc3db
TC
1621Vendor locales are notoriously buggy, and it is difficult for Perl to test
1622its locale-handling code because this interacts with code that Perl has no
1623control over; therefore the locale-handling code in Perl may be buggy as
66cbab2c
KW
1624well. (However, the Unicode-supplied locales should be better, and
1625there is a feed back mechanism to correct any problems. See
1626L</Freely available locale definitions>.)
1627
7ee2ae1e 1628If you have Perl v5.16, the problems mentioned above go away if you use
66cbab2c 1629the C<:not_characters> parameter to the locale pragma (except for vendor
7ee2ae1e 1630bugs in the non-character portions). If you don't have v5.16, and you
66cbab2c
KW
1631I<do> have locales that work, using them may be worthwhile for certain
1632specific purposes, as long as you keep in mind the gotchas already
1633mentioned. For example, if the collation for your locales works, it
1634runs faster under locales than under L<Unicode::Collate>; and you gain
1635access to such things as the local currency symbol and the names of the
7ee2ae1e 1636months and days of the week. (But to hammer home the point, in v5.16,
66cbab2c
KW
1637you get this access without the downsides of locales by using the
1638C<:not_characters> form of the pragma.)
1639
1640Note: The policy of using locale rules for code points that can fit in a
1641byte, and Unicode rules for those that can't is not uniformly applied.
7ee2ae1e 1642Pre-v5.12, it was somewhat haphazard; in v5.12 it was applied fairly
66cbab2c 1643consistently to regular expression matching except for bracketed
7ee2ae1e 1644character classes; in v5.14 it was extended to all regex matches; and in
663d437a 1645v5.16 to the casing operations such as C<\L> and C<uc()>. For
dbf3c4d7 1646collation, in all releases so far, the system's C<strxfrm()> function is
a4a439fb
KW
1647called, and whatever it does is what you get. Starting in v5.26, various
1648bugs are fixed with the way perl uses this function.
b310b053 1649
5f05dabc
PP
1650=head1 BUGS
1651
4e615abd
KW
1652=head2 Collation of strings containing embedded C<NUL> characters
1653
6696cfa7
KW
1654C<NUL> characters will sort the same as the lowest collating control
1655character does, or to C<"\001"> in the unlikely event that there are no
1656control characters at all in the locale. In cases where the strings
1657don't contain this non-C<NUL> control, the results will be correct, and
1658in many locales, this control, whatever it might be, will rarely be
1659encountered. But there are cases where a C<NUL> should sort before this
1660control, but doesn't. If two strings do collate identically, the one
0c880285
KW
1661containing the C<NUL> will sort to earlier. Prior to 5.26, there were
1662more bugs.
4e615abd 1663
e9bc6d6b
KW
1664=head2 Multi-threaded
1665
1666XS code or C-language libraries called from it that use the system
1667L<C<setlocale(3)>> function (except on Windows) likely will not work
1668from a multi-threaded application without changes. See
1669L<perlxs/Locale-aware XS code>.
1670
1671An XS module that is locale-dependent could have been written under the
1672assumption that it will never be called in a multi-threaded environment,
1673and so uses other non-locale constructs that aren't multi-thread-safe.
1674See L<perlxs/Thread-aware system interfaces>.
1675
1676POSIX does not define a way to get the name of the current per-thread
1677locale. Some systems, such as Darwin and NetBSD do implement a
1678function, L<querylocale(3)> to do this. On non-Windows systems without
1679it, such as Linux, there are some additional caveats:
1680
1681=over
1682
1683=item *
1684
1685An embedded perl needs to be started up while the global locale is in
1686effect. See L<perlembed/Using embedded Perl with POSIX locales>.
1687
1688=item *
1689
1690It becomes more important for perl to know about all the possible
1691locale categories on the platform, even if they aren't apparently used
1692in your program. Perl knows all of the Linux ones. If your platform
1693has others, you can send email to L<mailto:perlbug@perl.org> for
1694inclusion of it in the next release. In the meantime, it is possible to
1695edit the Perl source to teach it about the category, and then recompile.
1696Search for instances of, say, C<LC_PAPER> in the source, and use that as
1697a template to add the omitted one.
1698
1699=item *
1700
1701It is possible, though hard to do, to call C<POSIX::setlocale> with a
1702locale that it doesn't recognize as syntactically legal, but actually is
1703legal on that system. This should happen only with embedded perls, or
1704if you hand-craft a locale name yourself.
1705
1706=back
1707
5f05dabc
PP
1708=head2 Broken systems
1709
5a964f20 1710In certain systems, the operating system's locale support
2bdf8add 1711is broken and cannot be fixed or used by Perl. Such deficiencies can
b4ffc3db 1712and will result in mysterious hangs and/or Perl core dumps when
2bdf8add 1713C<use locale> is in effect. When confronted with such a system,
7f2de2d2 1714please report in excruciating detail to <F<perlbug@perl.org>>, and
b4ffc3db 1715also contact your vendor: bug fixes may exist for these problems
2bdf8add 1716in your operating system. Sometimes such bug fixes are called an
83fb1bf0
KW
1717operating system upgrade. If you have the source for Perl, include in
1718the perlbug email the output of the test described above in L</Testing
1719for broken locales>.
5f05dabc
PP
1720
1721=head1 SEE ALSO
1722
b310b053 1723L<I18N::Langinfo>, L<perluniintro>, L<perlunicode>, L<open>,
106ab961 1724L<POSIX/localeconv>,
4bbcc6e8
JH
1725L<POSIX/setlocale>, L<POSIX/strcoll>, L<POSIX/strftime>,
1726L<POSIX/strtod>, L<POSIX/strxfrm>.
5f05dabc 1727
ccd65d51
KW
1728For special considerations when Perl is embedded in a C program,
1729see L<perlembed/Using embedded Perl with POSIX locales>.
1730
5f05dabc
PP
1731=head1 HISTORY
1732
b0c42ed9 1733Jarkko Hietaniemi's original F<perli18n.pod> heavily hacked by Dominic
5a964f20 1734Dunlop, assisted by the perl5-porters. Prose worked over a bit by
106ab961 1735Tom Christiansen, and now maintained by Perl 5 porters.