xref: /openbsd/gnu/usr.bin/perl/pod/perlvar.pod (revision 09467b48)
1=head1 NAME
2
3perlvar - Perl predefined variables
4
5=head1 DESCRIPTION
6
7=head2 The Syntax of Variable Names
8
9Variable names in Perl can have several formats.  Usually, they
10must begin with a letter or underscore, in which case they can be
11arbitrarily long (up to an internal limit of 251 characters) and
12may contain letters, digits, underscores, or the special sequence
13C<::> or C<'>.  In this case, the part before the last C<::> or
14C<'> is taken to be a I<package qualifier>; see L<perlmod>.
15A Unicode letter that is not ASCII is not considered to be a letter
16unless S<C<"use utf8">> is in effect, and somewhat more complicated
17rules apply; see L<perldata/Identifier parsing> for details.
18
19Perl variable names may also be a sequence of digits, a single
20punctuation character, or the two-character sequence: C<^> (caret or
21CIRCUMFLEX ACCENT) followed by any one of the characters C<[][A-Z^_?\]>.
22These names are all reserved for
23special uses by Perl; for example, the all-digits names are used
24to hold data captured by backreferences after a regular expression
25match.
26
27Since Perl v5.6.0, Perl variable names may also be alphanumeric strings
28preceded by a caret.  These must all be written in the form C<${^Foo}>;
29the braces are not optional.  C<${^Foo}> denotes the scalar variable
30whose name is considered to be a control-C<F> followed by two C<o>'s.
31These variables are
32reserved for future special uses by Perl, except for the ones that
33begin with C<^_> (caret-underscore).  No
34name that begins with C<^_> will acquire a special
35meaning in any future version of Perl; such names may therefore be
36used safely in programs.  C<$^_> itself, however, I<is> reserved.
37
38Perl identifiers that begin with digits or
39punctuation characters are exempt from the effects of the C<package>
40declaration and are always forced to be in package C<main>; they are
41also exempt from C<strict 'vars'> errors.  A few other names are also
42exempt in these ways:
43
44    ENV      STDIN
45    INC      STDOUT
46    ARGV     STDERR
47    ARGVOUT
48    SIG
49
50In particular, the special C<${^_XYZ}> variables are always taken
51to be in package C<main>, regardless of any C<package> declarations
52presently in scope.
53
54=head1 SPECIAL VARIABLES
55
56The following names have special meaning to Perl.  Most punctuation
57names have reasonable mnemonics, or analogs in the shells.
58Nevertheless, if you wish to use long variable names, you need only say:
59
60    use English;
61
62at the top of your program.  This aliases all the short names to the long
63names in the current package.  Some even have medium names, generally
64borrowed from B<awk>.  For more info, please see L<English>.
65
66Before you continue, note the sort order for variables.  In general, we
67first list the variables in case-insensitive, almost-lexigraphical
68order (ignoring the C<{> or C<^> preceding words, as in C<${^UNICODE}>
69or C<$^T>), although C<$_> and C<@_> move up to the top of the pile.
70For variables with the same identifier, we list it in order of scalar,
71array, hash, and bareword.
72
73=head2 General Variables
74
75=over 8
76
77=item $ARG
78
79=item $_
80X<$_> X<$ARG>
81
82The default input and pattern-searching space.  The following pairs are
83equivalent:
84
85    while (<>) {...}    # equivalent only in while!
86    while (defined($_ = <>)) {...}
87
88    /^Subject:/
89    $_ =~ /^Subject:/
90
91    tr/a-z/A-Z/
92    $_ =~ tr/a-z/A-Z/
93
94    chomp
95    chomp($_)
96
97Here are the places where Perl will assume C<$_> even if you don't use it:
98
99=over 3
100
101=item *
102
103The following functions use C<$_> as a default argument:
104
105abs, alarm, chomp, chop, chr, chroot,
106cos, defined, eval, evalbytes, exp, fc, glob, hex, int, lc,
107lcfirst, length, log, lstat, mkdir, oct, ord, pos, print, printf,
108quotemeta, readlink, readpipe, ref, require, reverse (in scalar context only),
109rmdir, say, sin, split (for its second
110argument), sqrt, stat, study, uc, ucfirst,
111unlink, unpack.
112
113=item *
114
115All file tests (C<-f>, C<-d>) except for C<-t>, which defaults to STDIN.
116See L<perlfunc/-X>
117
118=item *
119
120The pattern matching operations C<m//>, C<s///> and C<tr///> (aka C<y///>)
121when used without an C<=~> operator.
122
123=item *
124
125The default iterator variable in a C<foreach> loop if no other
126variable is supplied.
127
128=item *
129
130The implicit iterator variable in the C<grep()> and C<map()> functions.
131
132=item *
133
134The implicit variable of C<given()>.
135
136=item *
137
138The default place to put the next value or input record
139when a C<< <FH> >>, C<readline>, C<readdir> or C<each>
140operation's result is tested by itself as the sole criterion of a C<while>
141test.  Outside a C<while> test, this will not happen.
142
143=back
144
145C<$_> is a global variable.
146
147However, between perl v5.10.0 and v5.24.0, it could be used lexically by
148writing C<my $_>.  Making C<$_> refer to the global C<$_> in the same scope
149was then possible with C<our $_>.  This experimental feature was removed and is
150now a fatal error, but you may encounter it in older code.
151
152Mnemonic: underline is understood in certain operations.
153
154=item @ARG
155
156=item @_
157X<@_> X<@ARG>
158
159Within a subroutine the array C<@_> contains the parameters passed to
160that subroutine.  Inside a subroutine, C<@_> is the default array for
161the array operators C<pop> and C<shift>.
162
163See L<perlsub>.
164
165=item $LIST_SEPARATOR
166
167=item $"
168X<$"> X<$LIST_SEPARATOR>
169
170When an array or an array slice is interpolated into a double-quoted
171string or a similar context such as C</.../>, its elements are
172separated by this value.  Default is a space.  For example, this:
173
174    print "The array is: @array\n";
175
176is equivalent to this:
177
178    print "The array is: " . join($", @array) . "\n";
179
180Mnemonic: works in double-quoted context.
181
182=item $PROCESS_ID
183
184=item $PID
185
186=item $$
187X<$$> X<$PID> X<$PROCESS_ID>
188
189The process number of the Perl running this script.  Though you I<can> set
190this variable, doing so is generally discouraged, although it can be
191invaluable for some testing purposes.  It will be reset automatically
192across C<fork()> calls.
193
194Note for Linux and Debian GNU/kFreeBSD users: Before Perl v5.16.0 perl
195would emulate POSIX semantics on Linux systems using LinuxThreads, a
196partial implementation of POSIX Threads that has since been superseded
197by the Native POSIX Thread Library (NPTL).
198
199LinuxThreads is now obsolete on Linux, and caching C<getpid()>
200like this made embedding perl unnecessarily complex (since you'd have
201to manually update the value of $$), so now C<$$> and C<getppid()>
202will always return the same values as the underlying C library.
203
204Debian GNU/kFreeBSD systems also used LinuxThreads up until and
205including the 6.0 release, but after that moved to FreeBSD thread
206semantics, which are POSIX-like.
207
208To see if your system is affected by this discrepancy check if
209C<getconf GNU_LIBPTHREAD_VERSION | grep -q NPTL> returns a false
210value.  NTPL threads preserve the POSIX semantics.
211
212Mnemonic: same as shells.
213
214=item $PROGRAM_NAME
215
216=item $0
217X<$0> X<$PROGRAM_NAME>
218
219Contains the name of the program being executed.
220
221On some (but not all) operating systems assigning to C<$0> modifies
222the argument area that the C<ps> program sees.  On some platforms you
223may have to use special C<ps> options or a different C<ps> to see the
224changes.  Modifying the C<$0> is more useful as a way of indicating the
225current program state than it is for hiding the program you're
226running.
227
228Note that there are platform-specific limitations on the maximum
229length of C<$0>.  In the most extreme case it may be limited to the
230space occupied by the original C<$0>.
231
232In some platforms there may be arbitrary amount of padding, for
233example space characters, after the modified name as shown by C<ps>.
234In some platforms this padding may extend all the way to the original
235length of the argument area, no matter what you do (this is the case
236for example with Linux 2.2).
237
238Note for BSD users: setting C<$0> does not completely remove "perl"
239from the ps(1) output.  For example, setting C<$0> to C<"foobar"> may
240result in C<"perl: foobar (perl)"> (whether both the C<"perl: "> prefix
241and the " (perl)" suffix are shown depends on your exact BSD variant
242and version).  This is an operating system feature, Perl cannot help it.
243
244In multithreaded scripts Perl coordinates the threads so that any
245thread may modify its copy of the C<$0> and the change becomes visible
246to ps(1) (assuming the operating system plays along).  Note that
247the view of C<$0> the other threads have will not change since they
248have their own copies of it.
249
250If the program has been given to perl via the switches C<-e> or C<-E>,
251C<$0> will contain the string C<"-e">.
252
253On Linux as of perl v5.14.0 the legacy process name will be set with
254C<prctl(2)>, in addition to altering the POSIX name via C<argv[0]> as
255perl has done since version 4.000.  Now system utilities that read the
256legacy process name such as ps, top and killall will recognize the
257name you set when assigning to C<$0>.  The string you supply will be
258cut off at 16 bytes, this is a limitation imposed by Linux.
259
260Mnemonic: same as B<sh> and B<ksh>.
261
262=item $REAL_GROUP_ID
263
264=item $GID
265
266=item $(
267X<$(> X<$GID> X<$REAL_GROUP_ID>
268
269The real gid of this process.  If you are on a machine that supports
270membership in multiple groups simultaneously, gives a space separated
271list of groups you are in.  The first number is the one returned by
272C<getgid()>, and the subsequent ones by C<getgroups()>, one of which may be
273the same as the first number.
274
275However, a value assigned to C<$(> must be a single number used to
276set the real gid.  So the value given by C<$(> should I<not> be assigned
277back to C<$(> without being forced numeric, such as by adding zero.  Note
278that this is different to the effective gid (C<$)>) which does take a
279list.
280
281You can change both the real gid and the effective gid at the same
282time by using C<POSIX::setgid()>.  Changes
283to C<$(> require a check to C<$!>
284to detect any possible errors after an attempted change.
285
286Mnemonic: parentheses are used to I<group> things.  The real gid is the
287group you I<left>, if you're running setgid.
288
289=item $EFFECTIVE_GROUP_ID
290
291=item $EGID
292
293=item $)
294X<$)> X<$EGID> X<$EFFECTIVE_GROUP_ID>
295
296The effective gid of this process.  If you are on a machine that
297supports membership in multiple groups simultaneously, gives a space
298separated list of groups you are in.  The first number is the one
299returned by C<getegid()>, and the subsequent ones by C<getgroups()>,
300one of which may be the same as the first number.
301
302Similarly, a value assigned to C<$)> must also be a space-separated
303list of numbers.  The first number sets the effective gid, and
304the rest (if any) are passed to C<setgroups()>.  To get the effect of an
305empty list for C<setgroups()>, just repeat the new effective gid; that is,
306to force an effective gid of 5 and an effectively empty C<setgroups()>
307list, say C< $) = "5 5" >.
308
309You can change both the effective gid and the real gid at the same
310time by using C<POSIX::setgid()> (use only a single numeric argument).
311Changes to C<$)> require a check to C<$!> to detect any possible errors
312after an attempted change.
313
314C<< $< >>, C<< $> >>, C<$(> and C<$)> can be set only on
315machines that support the corresponding I<set[re][ug]id()> routine.  C<$(>
316and C<$)> can be swapped only on machines supporting C<setregid()>.
317
318Mnemonic: parentheses are used to I<group> things.  The effective gid
319is the group that's I<right> for you, if you're running setgid.
320
321=item $REAL_USER_ID
322
323=item $UID
324
325=item $<
326X<< $< >> X<$UID> X<$REAL_USER_ID>
327
328The real uid of this process.  You can change both the real uid and the
329effective uid at the same time by using C<POSIX::setuid()>.  Since
330changes to C<< $< >> require a system call, check C<$!> after a change
331attempt to detect any possible errors.
332
333Mnemonic: it's the uid you came I<from>, if you're running setuid.
334
335=item $EFFECTIVE_USER_ID
336
337=item $EUID
338
339=item $>
340X<< $> >> X<$EUID> X<$EFFECTIVE_USER_ID>
341
342The effective uid of this process.  For example:
343
344    $< = $>;            # set real to effective uid
345    ($<,$>) = ($>,$<);  # swap real and effective uids
346
347You can change both the effective uid and the real uid at the same
348time by using C<POSIX::setuid()>.  Changes to C<< $> >> require a check
349to C<$!> to detect any possible errors after an attempted change.
350
351C<< $< >> and C<< $> >> can be swapped only on machines
352supporting C<setreuid()>.
353
354Mnemonic: it's the uid you went I<to>, if you're running setuid.
355
356=item $SUBSCRIPT_SEPARATOR
357
358=item $SUBSEP
359
360=item $;
361X<$;> X<$SUBSEP> X<SUBSCRIPT_SEPARATOR>
362
363The subscript separator for multidimensional array emulation.  If you
364refer to a hash element as
365
366    $foo{$x,$y,$z}
367
368it really means
369
370    $foo{join($;, $x, $y, $z)}
371
372But don't put
373
374    @foo{$x,$y,$z}	# a slice--note the @
375
376which means
377
378    ($foo{$x},$foo{$y},$foo{$z})
379
380Default is "\034", the same as SUBSEP in B<awk>.  If your keys contain
381binary data there might not be any safe value for C<$;>.
382
383Consider using "real" multidimensional arrays as described
384in L<perllol>.
385
386Mnemonic: comma (the syntactic subscript separator) is a semi-semicolon.
387
388=item $a
389
390=item $b
391X<$a> X<$b>
392
393Special package variables when using C<sort()>, see L<perlfunc/sort>.
394Because of this specialness C<$a> and C<$b> don't need to be declared
395(using C<use vars>, or C<our()>) even when using the C<strict 'vars'>
396pragma.  Don't lexicalize them with C<my $a> or C<my $b> if you want to
397be able to use them in the C<sort()> comparison block or function.
398
399=item %ENV
400X<%ENV>
401
402The hash C<%ENV> contains your current environment.  Setting a
403value in C<ENV> changes the environment for any child processes
404you subsequently C<fork()> off.
405
406As of v5.18.0, both keys and values stored in C<%ENV> are stringified.
407
408    my $foo = 1;
409    $ENV{'bar'} = \$foo;
410    if( ref $ENV{'bar'} ) {
411        say "Pre 5.18.0 Behaviour";
412    } else {
413        say "Post 5.18.0 Behaviour";
414    }
415
416Previously, only child processes received stringified values:
417
418    my $foo = 1;
419    $ENV{'bar'} = \$foo;
420
421    # Always printed 'non ref'
422    system($^X, '-e',
423           q/print ( ref $ENV{'bar'}  ? 'ref' : 'non ref' ) /);
424
425This happens because you can't really share arbitrary data structures with
426foreign processes.
427
428=item $OLD_PERL_VERSION
429
430=item $]
431X<$]> X<$OLD_PERL_VERSION>
432
433The revision, version, and subversion of the Perl interpreter, represented
434as a decimal of the form 5.XXXYYY, where XXX is the version / 1e3 and YYY
435is the subversion / 1e6.  For example, Perl v5.10.1 would be "5.010001".
436
437This variable can be used to determine whether the Perl interpreter
438executing a script is in the right range of versions:
439
440    warn "No PerlIO!\n" if $] lt '5.008';
441
442When comparing C<$]>, string comparison operators are B<highly
443recommended>.  The inherent limitations of binary floating point
444representation can sometimes lead to incorrect comparisons for some
445numbers on some architectures.
446
447See also the documentation of C<use VERSION> and C<require VERSION>
448for a convenient way to fail if the running Perl interpreter is too old.
449
450See L</$^V> for a representation of the Perl version as a L<version>
451object, which allows more flexible string comparisons.
452
453The main advantage of C<$]> over C<$^V> is that it works the same on any
454version of Perl.  The disadvantages are that it can't easily be compared
455to versions in other formats (e.g. literal v-strings, "v1.2.3" or
456version objects) and numeric comparisons can occasionally fail; it's good
457for string literal version checks and bad for comparing to a variable
458that hasn't been sanity-checked.
459
460The C<$OLD_PERL_VERSION> form was added in Perl v5.20.0 for historical
461reasons but its use is discouraged. (If your reason to use C<$]> is to
462run code on old perls then referring to it as C<$OLD_PERL_VERSION> would
463be self-defeating.)
464
465Mnemonic: Is this version of perl in the right bracket?
466
467=item $SYSTEM_FD_MAX
468
469=item $^F
470X<$^F> X<$SYSTEM_FD_MAX>
471
472The maximum system file descriptor, ordinarily 2.  System file
473descriptors are passed to C<exec()>ed processes, while higher file
474descriptors are not.  Also, during an
475C<open()>, system file descriptors are
476preserved even if the C<open()> fails (ordinary file descriptors are
477closed before the C<open()> is attempted).  The close-on-exec
478status of a file descriptor will be decided according to the value of
479C<$^F> when the corresponding file, pipe, or socket was opened, not the
480time of the C<exec()>.
481
482=item @F
483X<@F>
484
485The array C<@F> contains the fields of each line read in when autosplit
486mode is turned on.  See L<perlrun> for the B<-a> switch.  This array
487is package-specific, and must be declared or given a full package name
488if not in package main when running under C<strict 'vars'>.
489
490=item @INC
491X<@INC>
492
493The array C<@INC> contains the list of places that the C<do EXPR>,
494C<require>, or C<use> constructs look for their library files.  It
495initially consists of the arguments to any B<-I> command-line
496switches, followed by the default Perl library, probably
497F</usr/local/lib/perl>.
498Prior to Perl 5.26, C<.> -which represents the current directory, was included
499in C<@INC>; it has been removed. This change in behavior is documented
500in L<C<PERL_USE_UNSAFE_INC>|perlrun/PERL_USE_UNSAFE_INC> and it is
501not recommended that C<.> be re-added to C<@INC>.
502If you need to modify C<@INC> at runtime, you should use the C<use lib> pragma
503to get the machine-dependent library properly loaded as well:
504
505    use lib '/mypath/libdir/';
506    use SomeMod;
507
508You can also insert hooks into the file inclusion system by putting Perl
509code directly into C<@INC>.  Those hooks may be subroutine references,
510array references or blessed objects.  See L<perlfunc/require> for details.
511
512=item %INC
513X<%INC>
514
515The hash C<%INC> contains entries for each filename included via the
516C<do>, C<require>, or C<use> operators.  The key is the filename
517you specified (with module names converted to pathnames), and the
518value is the location of the file found.  The C<require>
519operator uses this hash to determine whether a particular file has
520already been included.
521
522If the file was loaded via a hook (e.g. a subroutine reference, see
523L<perlfunc/require> for a description of these hooks), this hook is
524by default inserted into C<%INC> in place of a filename.  Note, however,
525that the hook may have set the C<%INC> entry by itself to provide some more
526specific info.
527
528=item $INPLACE_EDIT
529
530=item $^I
531X<$^I> X<$INPLACE_EDIT>
532
533The current value of the inplace-edit extension.  Use C<undef> to disable
534inplace editing.
535
536Mnemonic: value of B<-i> switch.
537
538=item @ISA
539X<@ISA>
540
541Each package contains a special array called C<@ISA> which contains a list
542of that class's parent classes, if any. This array is simply a list of
543scalars, each of which is a string that corresponds to a package name. The
544array is examined when Perl does method resolution, which is covered in
545L<perlobj>.
546
547To load packages while adding them to C<@ISA>, see the L<parent> pragma. The
548discouraged L<base> pragma does this as well, but should not be used except
549when compatibility with the discouraged L<fields> pragma is required.
550
551=item $^M
552X<$^M>
553
554By default, running out of memory is an untrappable, fatal error.
555However, if suitably built, Perl can use the contents of C<$^M>
556as an emergency memory pool after C<die()>ing.  Suppose that your Perl
557were compiled with C<-DPERL_EMERGENCY_SBRK> and used Perl's malloc.
558Then
559
560    $^M = 'a' x (1 << 16);
561
562would allocate a 64K buffer for use in an emergency.  See the
563F<INSTALL> file in the Perl distribution for information on how to
564add custom C compilation flags when compiling perl.  To discourage casual
565use of this advanced feature, there is no L<English|English> long name for
566this variable.
567
568This variable was added in Perl 5.004.
569
570=item $OSNAME
571
572=item $^O
573X<$^O> X<$OSNAME>
574
575The name of the operating system under which this copy of Perl was
576built, as determined during the configuration process.  For examples
577see L<perlport/PLATFORMS>.
578
579The value is identical to C<$Config{'osname'}>.  See also L<Config>
580and the B<-V> command-line switch documented in L<perlrun>.
581
582In Windows platforms, C<$^O> is not very helpful: since it is always
583C<MSWin32>, it doesn't tell the difference between
58495/98/ME/NT/2000/XP/CE/.NET.  Use C<Win32::GetOSName()> or
585Win32::GetOSVersion() (see L<Win32> and L<perlport>) to distinguish
586between the variants.
587
588This variable was added in Perl 5.003.
589
590=item %SIG
591X<%SIG>
592
593The hash C<%SIG> contains signal handlers for signals.  For example:
594
595    sub handler {   # 1st argument is signal name
596	my($sig) = @_;
597	print "Caught a SIG$sig--shutting down\n";
598	close(LOG);
599	exit(0);
600	}
601
602    $SIG{'INT'}  = \&handler;
603    $SIG{'QUIT'} = \&handler;
604    ...
605    $SIG{'INT'}  = 'DEFAULT';   # restore default action
606    $SIG{'QUIT'} = 'IGNORE';    # ignore SIGQUIT
607
608Using a value of C<'IGNORE'> usually has the effect of ignoring the
609signal, except for the C<CHLD> signal.  See L<perlipc> for more about
610this special case.
611
612Here are some other examples:
613
614    $SIG{"PIPE"} = "Plumber";   # assumes main::Plumber (not
615				# recommended)
616    $SIG{"PIPE"} = \&Plumber;   # just fine; assume current
617				# Plumber
618    $SIG{"PIPE"} = *Plumber;    # somewhat esoteric
619    $SIG{"PIPE"} = Plumber();   # oops, what did Plumber()
620				# return??
621
622Be sure not to use a bareword as the name of a signal handler,
623lest you inadvertently call it.
624
625If your system has the C<sigaction()> function then signal handlers
626are installed using it.  This means you get reliable signal handling.
627
628The default delivery policy of signals changed in Perl v5.8.0 from
629immediate (also known as "unsafe") to deferred, also known as "safe
630signals".  See L<perlipc> for more information.
631
632Certain internal hooks can be also set using the C<%SIG> hash.  The
633routine indicated by C<$SIG{__WARN__}> is called when a warning
634message is about to be printed.  The warning message is passed as the
635first argument.  The presence of a C<__WARN__> hook causes the
636ordinary printing of warnings to C<STDERR> to be suppressed.  You can
637use this to save warnings in a variable, or turn warnings into fatal
638errors, like this:
639
640    local $SIG{__WARN__} = sub { die $_[0] };
641    eval $proggie;
642
643As the C<'IGNORE'> hook is not supported by C<__WARN__>, you can
644disable warnings using the empty subroutine:
645
646    local $SIG{__WARN__} = sub {};
647
648The routine indicated by C<$SIG{__DIE__}> is called when a fatal
649exception is about to be thrown.  The error message is passed as the
650first argument.  When a C<__DIE__> hook routine returns, the exception
651processing continues as it would have in the absence of the hook,
652unless the hook routine itself exits via a C<goto &sub>, a loop exit,
653or a C<die()>.  The C<__DIE__> handler is explicitly disabled during
654the call, so that you can die from a C<__DIE__> handler.  Similarly
655for C<__WARN__>.
656
657The C<$SIG{__DIE__}> hook is called even inside an C<eval()>. It was
658never intended to happen this way, but an implementation glitch made
659this possible. This used to be deprecated, as it allowed strange action
660at a distance like rewriting a pending exception in C<$@>. Plans to
661rectify this have been scrapped, as users found that rewriting a
662pending exception is actually a useful feature, and not a bug.
663
664C<__DIE__>/C<__WARN__> handlers are very special in one respect: they
665may be called to report (probable) errors found by the parser.  In such
666a case the parser may be in inconsistent state, so any attempt to
667evaluate Perl code from such a handler will probably result in a
668segfault.  This means that warnings or errors that result from parsing
669Perl should be used with extreme caution, like this:
670
671    require Carp if defined $^S;
672    Carp::confess("Something wrong") if defined &Carp::confess;
673    die "Something wrong, but could not load Carp to give "
674      . "backtrace...\n\t"
675      . "To see backtrace try starting Perl with -MCarp switch";
676
677Here the first line will load C<Carp> I<unless> it is the parser who
678called the handler.  The second line will print backtrace and die if
679C<Carp> was available.  The third line will be executed only if C<Carp> was
680not available.
681
682Having to even think about the C<$^S> variable in your exception
683handlers is simply wrong.  C<$SIG{__DIE__}> as currently implemented
684invites grievous and difficult to track down errors.  Avoid it
685and use an C<END{}> or CORE::GLOBAL::die override instead.
686
687See L<perlfunc/die>, L<perlfunc/warn>, L<perlfunc/eval>, and
688L<warnings> for additional information.
689
690=item $BASETIME
691
692=item $^T
693X<$^T> X<$BASETIME>
694
695The time at which the program began running, in seconds since the
696epoch (beginning of 1970).  The values returned by the B<-M>, B<-A>,
697and B<-C> filetests are based on this value.
698
699=item $PERL_VERSION
700
701=item $^V
702X<$^V> X<$PERL_VERSION>
703
704The revision, version, and subversion of the Perl interpreter,
705represented as a L<version> object.
706
707This variable first appeared in perl v5.6.0; earlier versions of perl
708will see an undefined value.  Before perl v5.10.0 C<$^V> was represented
709as a v-string rather than a L<version> object.
710
711C<$^V> can be used to determine whether the Perl interpreter executing
712a script is in the right range of versions.  For example:
713
714    warn "Hashes not randomized!\n" if !$^V or $^V lt v5.8.1
715
716While version objects overload stringification, to portably convert
717C<$^V> into its string representation, use C<sprintf()>'s C<"%vd">
718conversion, which works for both v-strings or version objects:
719
720    printf "version is v%vd\n", $^V;  # Perl's version
721
722See the documentation of C<use VERSION> and C<require VERSION>
723for a convenient way to fail if the running Perl interpreter is too old.
724
725See also C<L</$]>> for a decimal representation of the Perl version.
726
727The main advantage of C<$^V> over C<$]> is that, for Perl v5.10.0 or
728later, it overloads operators, allowing easy comparison against other
729version representations (e.g. decimal, literal v-string, "v1.2.3", or
730objects).  The disadvantage is that prior to v5.10.0, it was only a
731literal v-string, which can't be easily printed or compared, whereas
732the behavior of C<$]> is unchanged on all versions of Perl.
733
734Mnemonic: use ^V for a version object.
735
736=item ${^WIN32_SLOPPY_STAT}
737X<${^WIN32_SLOPPY_STAT}> X<sitecustomize> X<sitecustomize.pl>
738
739If this variable is set to a true value, then C<stat()> on Windows will
740not try to open the file.  This means that the link count cannot be
741determined and file attributes may be out of date if additional
742hardlinks to the file exist.  On the other hand, not opening the file
743is considerably faster, especially for files on network drives.
744
745This variable could be set in the F<sitecustomize.pl> file to
746configure the local Perl installation to use "sloppy" C<stat()> by
747default.  See the documentation for B<-f> in
748L<perlrun|perlrun/"Command Switches"> for more information about site
749customization.
750
751This variable was added in Perl v5.10.0.
752
753=item $EXECUTABLE_NAME
754
755=item $^X
756X<$^X> X<$EXECUTABLE_NAME>
757
758The name used to execute the current copy of Perl, from C's
759C<argv[0]> or (where supported) F</proc/self/exe>.
760
761Depending on the host operating system, the value of C<$^X> may be
762a relative or absolute pathname of the perl program file, or may
763be the string used to invoke perl but not the pathname of the
764perl program file.  Also, most operating systems permit invoking
765programs that are not in the PATH environment variable, so there
766is no guarantee that the value of C<$^X> is in PATH.  For VMS, the
767value may or may not include a version number.
768
769You usually can use the value of C<$^X> to re-invoke an independent
770copy of the same perl that is currently running, e.g.,
771
772    @first_run = `$^X -le "print int rand 100 for 1..100"`;
773
774But recall that not all operating systems support forking or
775capturing of the output of commands, so this complex statement
776may not be portable.
777
778It is not safe to use the value of C<$^X> as a path name of a file,
779as some operating systems that have a mandatory suffix on
780executable files do not require use of the suffix when invoking
781a command.  To convert the value of C<$^X> to a path name, use the
782following statements:
783
784    # Build up a set of file names (not command names).
785    use Config;
786    my $this_perl = $^X;
787    if ($^O ne 'VMS') {
788	$this_perl .= $Config{_exe}
789	  unless $this_perl =~ m/$Config{_exe}$/i;
790	}
791
792Because many operating systems permit anyone with read access to
793the Perl program file to make a copy of it, patch the copy, and
794then execute the copy, the security-conscious Perl programmer
795should take care to invoke the installed copy of perl, not the
796copy referenced by C<$^X>.  The following statements accomplish
797this goal, and produce a pathname that can be invoked as a
798command or referenced as a file.
799
800    use Config;
801    my $secure_perl_path = $Config{perlpath};
802    if ($^O ne 'VMS') {
803	$secure_perl_path .= $Config{_exe}
804	    unless $secure_perl_path =~ m/$Config{_exe}$/i;
805	}
806
807=back
808
809=head2 Variables related to regular expressions
810
811Most of the special variables related to regular expressions are side
812effects.  Perl sets these variables when it has a successful match, so
813you should check the match result before using them.  For instance:
814
815    if( /P(A)TT(ER)N/ ) {
816	print "I found $1 and $2\n";
817	}
818
819These variables are read-only and dynamically-scoped, unless we note
820otherwise.
821
822The dynamic nature of the regular expression variables means that
823their value is limited to the block that they are in, as demonstrated
824by this bit of code:
825
826    my $outer = 'Wallace and Grommit';
827    my $inner = 'Mutt and Jeff';
828
829    my $pattern = qr/(\S+) and (\S+)/;
830
831    sub show_n { print "\$1 is $1; \$2 is $2\n" }
832
833    {
834    OUTER:
835	show_n() if $outer =~ m/$pattern/;
836
837	INNER: {
838	    show_n() if $inner =~ m/$pattern/;
839	    }
840
841	show_n();
842    }
843
844The output shows that while in the C<OUTER> block, the values of C<$1>
845and C<$2> are from the match against C<$outer>.  Inside the C<INNER>
846block, the values of C<$1> and C<$2> are from the match against
847C<$inner>, but only until the end of the block (i.e. the dynamic
848scope).  After the C<INNER> block completes, the values of C<$1> and
849C<$2> return to the values for the match against C<$outer> even though
850we have not made another match:
851
852    $1 is Wallace; $2 is Grommit
853    $1 is Mutt; $2 is Jeff
854    $1 is Wallace; $2 is Grommit
855
856=head3 Performance issues
857
858Traditionally in Perl, any use of any of the three variables  C<$`>, C<$&>
859or C<$'> (or their C<use English> equivalents) anywhere in the code, caused
860all subsequent successful pattern matches to make a copy of the matched
861string, in case the code might subsequently access one of those variables.
862This imposed a considerable performance penalty across the whole program,
863so generally the use of these variables has been discouraged.
864
865In Perl 5.6.0 the C<@-> and C<@+> dynamic arrays were introduced that
866supply the indices of successful matches. So you could for example do
867this:
868
869    $str =~ /pattern/;
870
871    print $`, $&, $'; # bad: performance hit
872
873    print             # good: no performance hit
874	substr($str, 0,     $-[0]),
875	substr($str, $-[0], $+[0]-$-[0]),
876	substr($str, $+[0]);
877
878In Perl 5.10.0 the C</p> match operator flag and the C<${^PREMATCH}>,
879C<${^MATCH}>, and C<${^POSTMATCH}> variables were introduced, that allowed
880you to suffer the penalties only on patterns marked with C</p>.
881
882In Perl 5.18.0 onwards, perl started noting the presence of each of the
883three variables separately, and only copied that part of the string
884required; so in
885
886    $`; $&; "abcdefgh" =~ /d/
887
888perl would only copy the "abcd" part of the string. That could make a big
889difference in something like
890
891    $str = 'x' x 1_000_000;
892    $&; # whoops
893    $str =~ /x/g # one char copied a million times, not a million chars
894
895In Perl 5.20.0 a new copy-on-write system was enabled by default, which
896finally fixes all performance issues with these three variables, and makes
897them safe to use anywhere.
898
899The C<Devel::NYTProf> and C<Devel::FindAmpersand> modules can help you
900find uses of these problematic match variables in your code.
901
902=over 8
903
904=item $<I<digits>> ($1, $2, ...)
905X<$1> X<$2> X<$3> X<$I<digits>>
906
907Contains the subpattern from the corresponding set of capturing
908parentheses from the last successful pattern match, not counting patterns
909matched in nested blocks that have been exited already.
910
911Note there is a distinction between a capture buffer which matches
912the empty string a capture buffer which is optional. Eg, C<(x?)> and
913C<(x)?> The latter may be undef, the former not.
914
915These variables are read-only and dynamically-scoped.
916
917Mnemonic: like \digits.
918
919=item @{^CAPTURE}
920X<@{^CAPTURE}> X<@^CAPTURE>
921
922An array which exposes the contents of the capture buffers, if any, of
923the last successful pattern match, not counting patterns matched
924in nested blocks that have been exited already.
925
926Note that the 0 index of @{^CAPTURE} is equivalent to $1, the 1 index
927is equivalent to $2, etc.
928
929    if ("foal"=~/(.)(.)(.)(.)/) {
930        print join "-", @{^CAPTURE};
931    }
932
933should output "f-o-a-l".
934
935See also L<<< /$<I<digits>> ($1, $2, ...) >>>, L</%{^CAPTURE}> and
936L</%{^CAPTURE_ALL}>.
937
938Note that unlike most other regex magic variables there is no single
939letter equivalent to C<@{^CAPTURE}>.
940
941This variable was added in 5.25.7
942
943=item $MATCH
944
945=item $&
946X<$&> X<$MATCH>
947
948The string matched by the last successful pattern match (not counting
949any matches hidden within a BLOCK or C<eval()> enclosed by the current
950BLOCK).
951
952See L</Performance issues> above for the serious performance implications
953of using this variable (even once) in your code.
954
955This variable is read-only and dynamically-scoped.
956
957Mnemonic: like C<&> in some editors.
958
959=item ${^MATCH}
960X<${^MATCH}>
961
962This is similar to C<$&> (C<$MATCH>) except that it does not incur the
963performance penalty associated with that variable.
964
965See L</Performance issues> above.
966
967In Perl v5.18 and earlier, it is only guaranteed
968to return a defined value when the pattern was compiled or executed with
969the C</p> modifier.  In Perl v5.20, the C</p> modifier does nothing, so
970C<${^MATCH}> does the same thing as C<$MATCH>.
971
972This variable was added in Perl v5.10.0.
973
974This variable is read-only and dynamically-scoped.
975
976=item $PREMATCH
977
978=item $`
979X<$`> X<$PREMATCH> X<${^PREMATCH}>
980
981The string preceding whatever was matched by the last successful
982pattern match, not counting any matches hidden within a BLOCK or C<eval>
983enclosed by the current BLOCK.
984
985See L</Performance issues> above for the serious performance implications
986of using this variable (even once) in your code.
987
988This variable is read-only and dynamically-scoped.
989
990Mnemonic: C<`> often precedes a quoted string.
991
992=item ${^PREMATCH}
993X<$`> X<${^PREMATCH}>
994
995This is similar to C<$`> ($PREMATCH) except that it does not incur the
996performance penalty associated with that variable.
997
998See L</Performance issues> above.
999
1000In Perl v5.18 and earlier, it is only guaranteed
1001to return a defined value when the pattern was compiled or executed with
1002the C</p> modifier.  In Perl v5.20, the C</p> modifier does nothing, so
1003C<${^PREMATCH}> does the same thing as C<$PREMATCH>.
1004
1005This variable was added in Perl v5.10.0.
1006
1007This variable is read-only and dynamically-scoped.
1008
1009=item $POSTMATCH
1010
1011=item $'
1012X<$'> X<$POSTMATCH> X<${^POSTMATCH}> X<@->
1013
1014The string following whatever was matched by the last successful
1015pattern match (not counting any matches hidden within a BLOCK or C<eval()>
1016enclosed by the current BLOCK).  Example:
1017
1018    local $_ = 'abcdefghi';
1019    /def/;
1020    print "$`:$&:$'\n";  	# prints abc:def:ghi
1021
1022See L</Performance issues> above for the serious performance implications
1023of using this variable (even once) in your code.
1024
1025This variable is read-only and dynamically-scoped.
1026
1027Mnemonic: C<'> often follows a quoted string.
1028
1029=item ${^POSTMATCH}
1030X<${^POSTMATCH}> X<$'> X<$POSTMATCH>
1031
1032This is similar to C<$'> (C<$POSTMATCH>) except that it does not incur the
1033performance penalty associated with that variable.
1034
1035See L</Performance issues> above.
1036
1037In Perl v5.18 and earlier, it is only guaranteed
1038to return a defined value when the pattern was compiled or executed with
1039the C</p> modifier.  In Perl v5.20, the C</p> modifier does nothing, so
1040C<${^POSTMATCH}> does the same thing as C<$POSTMATCH>.
1041
1042This variable was added in Perl v5.10.0.
1043
1044This variable is read-only and dynamically-scoped.
1045
1046=item $LAST_PAREN_MATCH
1047
1048=item $+
1049X<$+> X<$LAST_PAREN_MATCH>
1050
1051The text matched by the highest used capture group of the last
1052successful search pattern.  It is logically equivalent to the highest
1053numbered capture variable (C<$1>, C<$2>, ...) which has a defined value.
1054
1055This is useful if you don't know which one of a set of alternative patterns
1056matched.  For example:
1057
1058    /Version: (.*)|Revision: (.*)/ && ($rev = $+);
1059
1060This variable is read-only and dynamically-scoped.
1061
1062Mnemonic: be positive and forward looking.
1063
1064=item $LAST_SUBMATCH_RESULT
1065
1066=item $^N
1067X<$^N> X<$LAST_SUBMATCH_RESULT>
1068
1069The text matched by the used group most-recently closed (i.e. the group
1070with the rightmost closing parenthesis) of the last successful search
1071pattern. This is subtly different from C<$+>. For example in
1072
1073    "ab" =~ /^((.)(.))$/
1074
1075we have
1076
1077    $1,$^N   have the value "ab"
1078    $2       has  the value "a"
1079    $3,$+    have the value "b"
1080
1081This is primarily used inside C<(?{...})> blocks for examining text
1082recently matched.  For example, to effectively capture text to a variable
1083(in addition to C<$1>, C<$2>, etc.), replace C<(...)> with
1084
1085    (?:(...)(?{ $var = $^N }))
1086
1087By setting and then using C<$var> in this way relieves you from having to
1088worry about exactly which numbered set of parentheses they are.
1089
1090This variable was added in Perl v5.8.0.
1091
1092Mnemonic: the (possibly) Nested parenthesis that most recently closed.
1093
1094=item @LAST_MATCH_END
1095
1096=item @+
1097X<@+> X<@LAST_MATCH_END>
1098
1099This array holds the offsets of the ends of the last successful
1100submatches in the currently active dynamic scope.  C<$+[0]> is
1101the offset into the string of the end of the entire match.  This
1102is the same value as what the C<pos> function returns when called
1103on the variable that was matched against.  The I<n>th element
1104of this array holds the offset of the I<n>th submatch, so
1105C<$+[1]> is the offset past where C<$1> ends, C<$+[2]> the offset
1106past where C<$2> ends, and so on.  You can use C<$#+> to determine
1107how many subgroups were in the last successful match.  See the
1108examples given for the C<@-> variable.
1109
1110This variable was added in Perl v5.6.0.
1111
1112=item %{^CAPTURE}
1113
1114=item %LAST_PAREN_MATCH
1115
1116=item %+
1117X<%+> X<%LAST_PAREN_MATCH> X<%{^CAPTURE}>
1118
1119Similar to C<@+>, the C<%+> hash allows access to the named capture
1120buffers, should they exist, in the last successful match in the
1121currently active dynamic scope.
1122
1123For example, C<$+{foo}> is equivalent to C<$1> after the following match:
1124
1125    'foo' =~ /(?<foo>foo)/;
1126
1127The keys of the C<%+> hash list only the names of buffers that have
1128captured (and that are thus associated to defined values).
1129
1130If multiple distinct capture groups have the same name, then
1131C<$+{NAME}> will refer to the leftmost defined group in the match.
1132
1133The underlying behaviour of C<%+> is provided by the
1134L<Tie::Hash::NamedCapture> module.
1135
1136B<Note:> C<%-> and C<%+> are tied views into a common internal hash
1137associated with the last successful regular expression.  Therefore mixing
1138iterative access to them via C<each> may have unpredictable results.
1139Likewise, if the last successful match changes, then the results may be
1140surprising.
1141
1142This variable was added in Perl v5.10.0. The C<%{^CAPTURE}> alias was
1143added in 5.25.7.
1144
1145This variable is read-only and dynamically-scoped.
1146
1147=item @LAST_MATCH_START
1148
1149=item @-
1150X<@-> X<@LAST_MATCH_START>
1151
1152C<$-[0]> is the offset of the start of the last successful match.
1153C<$-[I<n>]> is the offset of the start of the substring matched by
1154I<n>-th subpattern, or undef if the subpattern did not match.
1155
1156Thus, after a match against C<$_>, C<$&> coincides with C<substr $_, $-[0],
1157$+[0] - $-[0]>.  Similarly, $I<n> coincides with C<substr $_, $-[n],
1158$+[n] - $-[n]> if C<$-[n]> is defined, and $+ coincides with
1159C<substr $_, $-[$#-], $+[$#-] - $-[$#-]>.  One can use C<$#-> to find the
1160last matched subgroup in the last successful match.  Contrast with
1161C<$#+>, the number of subgroups in the regular expression.  Compare
1162with C<@+>.
1163
1164This array holds the offsets of the beginnings of the last
1165successful submatches in the currently active dynamic scope.
1166C<$-[0]> is the offset into the string of the beginning of the
1167entire match.  The I<n>th element of this array holds the offset
1168of the I<n>th submatch, so C<$-[1]> is the offset where C<$1>
1169begins, C<$-[2]> the offset where C<$2> begins, and so on.
1170
1171After a match against some variable C<$var>:
1172
1173=over 5
1174
1175=item C<$`> is the same as C<substr($var, 0, $-[0])>
1176
1177=item C<$&> is the same as C<substr($var, $-[0], $+[0] - $-[0])>
1178
1179=item C<$'> is the same as C<substr($var, $+[0])>
1180
1181=item C<$1> is the same as C<substr($var, $-[1], $+[1] - $-[1])>
1182
1183=item C<$2> is the same as C<substr($var, $-[2], $+[2] - $-[2])>
1184
1185=item C<$3> is the same as C<substr($var, $-[3], $+[3] - $-[3])>
1186
1187=back
1188
1189This variable was added in Perl v5.6.0.
1190
1191=item %{^CAPTURE_ALL}
1192X<%{^CAPTURE_ALL}>
1193
1194=item %-
1195X<%->
1196
1197Similar to C<%+>, this variable allows access to the named capture groups
1198in the last successful match in the currently active dynamic scope.  To
1199each capture group name found in the regular expression, it associates a
1200reference to an array containing the list of values captured by all
1201buffers with that name (should there be several of them), in the order
1202where they appear.
1203
1204Here's an example:
1205
1206    if ('1234' =~ /(?<A>1)(?<B>2)(?<A>3)(?<B>4)/) {
1207        foreach my $bufname (sort keys %-) {
1208            my $ary = $-{$bufname};
1209            foreach my $idx (0..$#$ary) {
1210                print "\$-{$bufname}[$idx] : ",
1211                      (defined($ary->[$idx])
1212                          ? "'$ary->[$idx]'"
1213                          : "undef"),
1214                      "\n";
1215            }
1216        }
1217    }
1218
1219would print out:
1220
1221    $-{A}[0] : '1'
1222    $-{A}[1] : '3'
1223    $-{B}[0] : '2'
1224    $-{B}[1] : '4'
1225
1226The keys of the C<%-> hash correspond to all buffer names found in
1227the regular expression.
1228
1229The behaviour of C<%-> is implemented via the
1230L<Tie::Hash::NamedCapture> module.
1231
1232B<Note:> C<%-> and C<%+> are tied views into a common internal hash
1233associated with the last successful regular expression.  Therefore mixing
1234iterative access to them via C<each> may have unpredictable results.
1235Likewise, if the last successful match changes, then the results may be
1236surprising.
1237
1238This variable was added in Perl v5.10.0. The C<%{^CAPTURE_ALL}> alias was
1239added in 5.25.7.
1240
1241This variable is read-only and dynamically-scoped.
1242
1243=item $LAST_REGEXP_CODE_RESULT
1244
1245=item $^R
1246X<$^R> X<$LAST_REGEXP_CODE_RESULT>
1247
1248The result of evaluation of the last successful C<(?{ code })>
1249regular expression assertion (see L<perlre>).  May be written to.
1250
1251This variable was added in Perl 5.005.
1252
1253=item ${^RE_COMPILE_RECURSION_LIMIT}
1254X<${^RE_COMPILE_RECURSION_LIMIT}>
1255
1256The current value giving the maximum number of open but unclosed
1257parenthetical groups there may be at any point during a regular
1258expression compilation.  The default is currently 1000 nested groups.
1259You may adjust it depending on your needs and the amount of memory
1260available.
1261
1262This variable was added in Perl v5.30.0.
1263
1264=item ${^RE_DEBUG_FLAGS}
1265X<${^RE_DEBUG_FLAGS}>
1266
1267The current value of the regex debugging flags.  Set to 0 for no debug output
1268even when the C<re 'debug'> module is loaded.  See L<re> for details.
1269
1270This variable was added in Perl v5.10.0.
1271
1272=item ${^RE_TRIE_MAXBUF}
1273X<${^RE_TRIE_MAXBUF}>
1274
1275Controls how certain regex optimisations are applied and how much memory they
1276utilize.  This value by default is 65536 which corresponds to a 512kB
1277temporary cache.  Set this to a higher value to trade
1278memory for speed when matching large alternations.  Set
1279it to a lower value if you want the optimisations to
1280be as conservative of memory as possible but still occur, and set it to a
1281negative value to prevent the optimisation and conserve the most memory.
1282Under normal situations this variable should be of no interest to you.
1283
1284This variable was added in Perl v5.10.0.
1285
1286=back
1287
1288=head2 Variables related to filehandles
1289
1290Variables that depend on the currently selected filehandle may be set
1291by calling an appropriate object method on the C<IO::Handle> object,
1292although this is less efficient than using the regular built-in
1293variables.  (Summary lines below for this contain the word HANDLE.)
1294First you must say
1295
1296    use IO::Handle;
1297
1298after which you may use either
1299
1300    method HANDLE EXPR
1301
1302or more safely,
1303
1304    HANDLE->method(EXPR)
1305
1306Each method returns the old value of the C<IO::Handle> attribute.  The
1307methods each take an optional EXPR, which, if supplied, specifies the
1308new value for the C<IO::Handle> attribute in question.  If not
1309supplied, most methods do nothing to the current value--except for
1310C<autoflush()>, which will assume a 1 for you, just to be different.
1311
1312Because loading in the C<IO::Handle> class is an expensive operation,
1313you should learn how to use the regular built-in variables.
1314
1315A few of these variables are considered "read-only".  This means that
1316if you try to assign to this variable, either directly or indirectly
1317through a reference, you'll raise a run-time exception.
1318
1319You should be very careful when modifying the default values of most
1320special variables described in this document.  In most cases you want
1321to localize these variables before changing them, since if you don't,
1322the change may affect other modules which rely on the default values
1323of the special variables that you have changed.  This is one of the
1324correct ways to read the whole file at once:
1325
1326    open my $fh, "<", "foo" or die $!;
1327    local $/; # enable localized slurp mode
1328    my $content = <$fh>;
1329    close $fh;
1330
1331But the following code is quite bad:
1332
1333    open my $fh, "<", "foo" or die $!;
1334    undef $/; # enable slurp mode
1335    my $content = <$fh>;
1336    close $fh;
1337
1338since some other module, may want to read data from some file in the
1339default "line mode", so if the code we have just presented has been
1340executed, the global value of C<$/> is now changed for any other code
1341running inside the same Perl interpreter.
1342
1343Usually when a variable is localized you want to make sure that this
1344change affects the shortest scope possible.  So unless you are already
1345inside some short C<{}> block, you should create one yourself.  For
1346example:
1347
1348    my $content = '';
1349    open my $fh, "<", "foo" or die $!;
1350    {
1351	local $/;
1352	$content = <$fh>;
1353    }
1354    close $fh;
1355
1356Here is an example of how your own code can go broken:
1357
1358    for ( 1..3 ){
1359	$\ = "\r\n";
1360	nasty_break();
1361	print "$_";
1362    }
1363
1364    sub nasty_break {
1365	$\ = "\f";
1366	# do something with $_
1367    }
1368
1369You probably expect this code to print the equivalent of
1370
1371    "1\r\n2\r\n3\r\n"
1372
1373but instead you get:
1374
1375    "1\f2\f3\f"
1376
1377Why? Because C<nasty_break()> modifies C<$\> without localizing it
1378first.  The value you set in  C<nasty_break()> is still there when you
1379return.  The fix is to add C<local()> so the value doesn't leak out of
1380C<nasty_break()>:
1381
1382    local $\ = "\f";
1383
1384It's easy to notice the problem in such a short example, but in more
1385complicated code you are looking for trouble if you don't localize
1386changes to the special variables.
1387
1388=over 8
1389
1390=item $ARGV
1391X<$ARGV>
1392
1393Contains the name of the current file when reading from C<< <> >>.
1394
1395=item @ARGV
1396X<@ARGV>
1397
1398The array C<@ARGV> contains the command-line arguments intended for
1399the script.  C<$#ARGV> is generally the number of arguments minus
1400one, because C<$ARGV[0]> is the first argument, I<not> the program's
1401command name itself.  See L</$0> for the command name.
1402
1403=item ARGV
1404X<ARGV>
1405
1406The special filehandle that iterates over command-line filenames in
1407C<@ARGV>.  Usually written as the null filehandle in the angle operator
1408C<< <> >>.  Note that currently C<ARGV> only has its magical effect
1409within the C<< <> >> operator; elsewhere it is just a plain filehandle
1410corresponding to the last file opened by C<< <> >>.  In particular,
1411passing C<\*ARGV> as a parameter to a function that expects a filehandle
1412may not cause your function to automatically read the contents of all the
1413files in C<@ARGV>.
1414
1415=item ARGVOUT
1416X<ARGVOUT>
1417
1418The special filehandle that points to the currently open output file
1419when doing edit-in-place processing with B<-i>.  Useful when you have
1420to do a lot of inserting and don't want to keep modifying C<$_>.  See
1421L<perlrun> for the B<-i> switch.
1422
1423=item IO::Handle->output_field_separator( EXPR )
1424
1425=item $OUTPUT_FIELD_SEPARATOR
1426
1427=item $OFS
1428
1429=item $,
1430X<$,> X<$OFS> X<$OUTPUT_FIELD_SEPARATOR>
1431
1432The output field separator for the print operator.  If defined, this
1433value is printed between each of print's arguments.  Default is C<undef>.
1434
1435You cannot call C<output_field_separator()> on a handle, only as a
1436static method.  See L<IO::Handle|IO::Handle>.
1437
1438Mnemonic: what is printed when there is a "," in your print statement.
1439
1440=item HANDLE->input_line_number( EXPR )
1441
1442=item $INPUT_LINE_NUMBER
1443
1444=item $NR
1445
1446=item $.
1447X<$.> X<$NR> X<$INPUT_LINE_NUMBER> X<line number>
1448
1449Current line number for the last filehandle accessed.
1450
1451Each filehandle in Perl counts the number of lines that have been read
1452from it.  (Depending on the value of C<$/>, Perl's idea of what
1453constitutes a line may not match yours.)  When a line is read from a
1454filehandle (via C<readline()> or C<< <> >>), or when C<tell()> or
1455C<seek()> is called on it, C<$.> becomes an alias to the line counter
1456for that filehandle.
1457
1458You can adjust the counter by assigning to C<$.>, but this will not
1459actually move the seek pointer.  I<Localizing C<$.> will not localize
1460the filehandle's line count>.  Instead, it will localize perl's notion
1461of which filehandle C<$.> is currently aliased to.
1462
1463C<$.> is reset when the filehandle is closed, but B<not> when an open
1464filehandle is reopened without an intervening C<close()>.  For more
1465details, see L<perlop/"IE<sol>O Operators">.  Because C<< <> >> never does
1466an explicit close, line numbers increase across C<ARGV> files (but see
1467examples in L<perlfunc/eof>).
1468
1469You can also use C<< HANDLE->input_line_number(EXPR) >> to access the
1470line counter for a given filehandle without having to worry about
1471which handle you last accessed.
1472
1473Mnemonic: many programs use "." to mean the current line number.
1474
1475=item IO::Handle->input_record_separator( EXPR )
1476
1477=item $INPUT_RECORD_SEPARATOR
1478
1479=item $RS
1480
1481=item $/
1482X<$/> X<$RS> X<$INPUT_RECORD_SEPARATOR>
1483
1484The input record separator, newline by default.  This influences Perl's
1485idea of what a "line" is.  Works like B<awk>'s RS variable, including
1486treating empty lines as a terminator if set to the null string (an
1487empty line cannot contain any spaces or tabs).  You may set it to a
1488multi-character string to match a multi-character terminator, or to
1489C<undef> to read through the end of file.  Setting it to C<"\n\n">
1490means something slightly different than setting to C<"">, if the file
1491contains consecutive empty lines.  Setting to C<""> will treat two or
1492more consecutive empty lines as a single empty line.  Setting to
1493C<"\n\n"> will blindly assume that the next input character belongs to
1494the next paragraph, even if it's a newline.
1495
1496    local $/;           # enable "slurp" mode
1497    local $_ = <FH>;    # whole file now here
1498    s/\n[ \t]+/ /g;
1499
1500Remember: the value of C<$/> is a string, not a regex.  B<awk> has to
1501be better for something. :-)
1502
1503Setting C<$/> to an empty string -- the so-called I<paragraph mode> -- merits
1504special attention.  When C<$/> is set to C<""> and the entire file is read in
1505with that setting, any sequence of consecutive newlines C<"\n\n"> at the
1506beginning of the file is discarded.  With the exception of the final record in
1507the file, each sequence of characters ending in two or more newlines is
1508treated as one record and is read in to end in exactly two newlines.  If the
1509last record in the file ends in zero or one consecutive newlines, that record
1510is read in with that number of newlines.  If the last record ends in two or
1511more consecutive newlines, it is read in with two newlines like all preceding
1512records.
1513
1514Suppose we wrote the following string to a file:
1515
1516    my $string = "\n\n\n";
1517    $string .= "alpha beta\ngamma delta\n\n\n";
1518    $string .= "epsilon zeta eta\n\n";
1519    $string .= "theta\n";
1520
1521    my $file = 'simple_file.txt';
1522    open my $OUT, '>', $file or die;
1523    print $OUT $string;
1524    close $OUT or die;
1525
1526Now we read that file in paragraph mode:
1527
1528    local $/ = ""; # paragraph mode
1529    open my $IN, '<', $file or die;
1530    my @records = <$IN>;
1531    close $IN or die;
1532
1533C<@records> will consist of these 3 strings:
1534
1535    (
1536      "alpha beta\ngamma delta\n\n",
1537      "epsilon zeta eta\n\n",
1538      "theta\n",
1539    )
1540
1541Setting C<$/> to a reference to an integer, scalar containing an
1542integer, or scalar that's convertible to an integer will attempt to
1543read records instead of lines, with the maximum record size being the
1544referenced integer number of characters.  So this:
1545
1546    local $/ = \32768; # or \"32768", or \$var_containing_32768
1547    open my $fh, "<", $myfile or die $!;
1548    local $_ = <$fh>;
1549
1550will read a record of no more than 32768 characters from $fh.  If you're
1551not reading from a record-oriented file (or your OS doesn't have
1552record-oriented files), then you'll likely get a full chunk of data
1553with every read.  If a record is larger than the record size you've
1554set, you'll get the record back in pieces.  Trying to set the record
1555size to zero or less is deprecated and will cause $/ to have the value
1556of "undef", which will cause reading in the (rest of the) whole file.
1557
1558As of 5.19.9 setting C<$/> to any other form of reference will throw a
1559fatal exception. This is in preparation for supporting new ways to set
1560C<$/> in the future.
1561
1562On VMS only, record reads bypass PerlIO layers and any associated
1563buffering, so you must not mix record and non-record reads on the
1564same filehandle.  Record mode mixes with line mode only when the
1565same buffering layer is in use for both modes.
1566
1567You cannot call C<input_record_separator()> on a handle, only as a
1568static method.  See L<IO::Handle|IO::Handle>.
1569
1570See also L<perlport/"Newlines">.  Also see L</$.>.
1571
1572Mnemonic: / delimits line boundaries when quoting poetry.
1573
1574=item IO::Handle->output_record_separator( EXPR )
1575
1576=item $OUTPUT_RECORD_SEPARATOR
1577
1578=item $ORS
1579
1580=item $\
1581X<$\> X<$ORS> X<$OUTPUT_RECORD_SEPARATOR>
1582
1583The output record separator for the print operator.  If defined, this
1584value is printed after the last of print's arguments.  Default is C<undef>.
1585
1586You cannot call C<output_record_separator()> on a handle, only as a
1587static method.  See L<IO::Handle|IO::Handle>.
1588
1589Mnemonic: you set C<$\> instead of adding "\n" at the end of the print.
1590Also, it's just like C<$/>, but it's what you get "back" from Perl.
1591
1592=item HANDLE->autoflush( EXPR )
1593
1594=item $OUTPUT_AUTOFLUSH
1595
1596=item $|
1597X<$|> X<autoflush> X<flush> X<$OUTPUT_AUTOFLUSH>
1598
1599If set to nonzero, forces a flush right away and after every write or
1600print on the currently selected output channel.  Default is 0
1601(regardless of whether the channel is really buffered by the system or
1602not; C<$|> tells you only whether you've asked Perl explicitly to
1603flush after each write).  STDOUT will typically be line buffered if
1604output is to the terminal and block buffered otherwise.  Setting this
1605variable is useful primarily when you are outputting to a pipe or
1606socket, such as when you are running a Perl program under B<rsh> and
1607want to see the output as it's happening.  This has no effect on input
1608buffering.  See L<perlfunc/getc> for that.  See L<perlfunc/select> on
1609how to select the output channel.  See also L<IO::Handle>.
1610
1611Mnemonic: when you want your pipes to be piping hot.
1612
1613=item ${^LAST_FH}
1614X<${^LAST_FH}>
1615
1616This read-only variable contains a reference to the last-read filehandle.
1617This is set by C<< <HANDLE> >>, C<readline>, C<tell>, C<eof> and C<seek>.
1618This is the same handle that C<$.> and C<tell> and C<eof> without arguments
1619use.  It is also the handle used when Perl appends ", <STDIN> line 1" to
1620an error or warning message.
1621
1622This variable was added in Perl v5.18.0.
1623
1624=back
1625
1626=head3 Variables related to formats
1627
1628The special variables for formats are a subset of those for
1629filehandles.  See L<perlform> for more information about Perl's
1630formats.
1631
1632=over 8
1633
1634=item $ACCUMULATOR
1635
1636=item $^A
1637X<$^A> X<$ACCUMULATOR>
1638
1639The current value of the C<write()> accumulator for C<format()> lines.
1640A format contains C<formline()> calls that put their result into
1641C<$^A>.  After calling its format, C<write()> prints out the contents
1642of C<$^A> and empties.  So you never really see the contents of C<$^A>
1643unless you call C<formline()> yourself and then look at it.  See
1644L<perlform> and L<perlfunc/"formline PICTURE,LIST">.
1645
1646=item IO::Handle->format_formfeed(EXPR)
1647
1648=item $FORMAT_FORMFEED
1649
1650=item $^L
1651X<$^L> X<$FORMAT_FORMFEED>
1652
1653What formats output as a form feed.  The default is C<\f>.
1654
1655You cannot call C<format_formfeed()> on a handle, only as a static
1656method.  See L<IO::Handle|IO::Handle>.
1657
1658=item HANDLE->format_page_number(EXPR)
1659
1660=item $FORMAT_PAGE_NUMBER
1661
1662=item $%
1663X<$%> X<$FORMAT_PAGE_NUMBER>
1664
1665The current page number of the currently selected output channel.
1666
1667Mnemonic: C<%> is page number in B<nroff>.
1668
1669=item HANDLE->format_lines_left(EXPR)
1670
1671=item $FORMAT_LINES_LEFT
1672
1673=item $-
1674X<$-> X<$FORMAT_LINES_LEFT>
1675
1676The number of lines left on the page of the currently selected output
1677channel.
1678
1679Mnemonic: lines_on_page - lines_printed.
1680
1681=item IO::Handle->format_line_break_characters EXPR
1682
1683=item $FORMAT_LINE_BREAK_CHARACTERS
1684
1685=item $:
1686X<$:> X<FORMAT_LINE_BREAK_CHARACTERS>
1687
1688The current set of characters after which a string may be broken to
1689fill continuation fields (starting with C<^>) in a format.  The default is
1690S<" \n-">, to break on a space, newline, or a hyphen.
1691
1692You cannot call C<format_line_break_characters()> on a handle, only as
1693a static method.  See L<IO::Handle|IO::Handle>.
1694
1695Mnemonic: a "colon" in poetry is a part of a line.
1696
1697=item HANDLE->format_lines_per_page(EXPR)
1698
1699=item $FORMAT_LINES_PER_PAGE
1700
1701=item $=
1702X<$=> X<$FORMAT_LINES_PER_PAGE>
1703
1704The current page length (printable lines) of the currently selected
1705output channel.  The default is 60.
1706
1707Mnemonic: = has horizontal lines.
1708
1709=item HANDLE->format_top_name(EXPR)
1710
1711=item $FORMAT_TOP_NAME
1712
1713=item $^
1714X<$^> X<$FORMAT_TOP_NAME>
1715
1716The name of the current top-of-page format for the currently selected
1717output channel.  The default is the name of the filehandle with C<_TOP>
1718appended.  For example, the default format top name for the C<STDOUT>
1719filehandle is C<STDOUT_TOP>.
1720
1721Mnemonic: points to top of page.
1722
1723=item HANDLE->format_name(EXPR)
1724
1725=item $FORMAT_NAME
1726
1727=item $~
1728X<$~> X<$FORMAT_NAME>
1729
1730The name of the current report format for the currently selected
1731output channel.  The default format name is the same as the filehandle
1732name.  For example, the default format name for the C<STDOUT>
1733filehandle is just C<STDOUT>.
1734
1735Mnemonic: brother to C<$^>.
1736
1737=back
1738
1739=head2 Error Variables
1740X<error> X<exception>
1741
1742The variables C<$@>, C<$!>, C<$^E>, and C<$?> contain information
1743about different types of error conditions that may appear during
1744execution of a Perl program.  The variables are shown ordered by
1745the "distance" between the subsystem which reported the error and
1746the Perl process.  They correspond to errors detected by the Perl
1747interpreter, C library, operating system, or an external program,
1748respectively.
1749
1750To illustrate the differences between these variables, consider the
1751following Perl expression, which uses a single-quoted string.  After
1752execution of this statement, perl may have set all four special error
1753variables:
1754
1755    eval q{
1756	open my $pipe, "/cdrom/install |" or die $!;
1757	my @res = <$pipe>;
1758	close $pipe or die "bad pipe: $?, $!";
1759    };
1760
1761When perl executes the C<eval()> expression, it translates the
1762C<open()>, C<< <PIPE> >>, and C<close> calls in the C run-time library
1763and thence to the operating system kernel.  perl sets C<$!> to
1764the C library's C<errno> if one of these calls fails.
1765
1766C<$@> is set if the string to be C<eval>-ed did not compile (this may
1767happen if C<open> or C<close> were imported with bad prototypes), or
1768if Perl code executed during evaluation C<die()>d.  In these cases the
1769value of C<$@> is the compile error, or the argument to C<die> (which
1770will interpolate C<$!> and C<$?>).  (See also L<Fatal>, though.)
1771
1772Under a few operating systems, C<$^E> may contain a more verbose error
1773indicator, such as in this case, "CDROM tray not closed."  Systems that
1774do not support extended error messages leave C<$^E> the same as C<$!>.
1775
1776Finally, C<$?> may be set to a non-0 value if the external program
1777F</cdrom/install> fails.  The upper eight bits reflect specific error
1778conditions encountered by the program (the program's C<exit()> value).
1779The lower eight bits reflect mode of failure, like signal death and
1780core dump information.  See L<wait(2)> for details.  In contrast to
1781C<$!> and C<$^E>, which are set only if an error condition is detected,
1782the variable C<$?> is set on each C<wait> or pipe C<close>,
1783overwriting the old value.  This is more like C<$@>, which on every
1784C<eval()> is always set on failure and cleared on success.
1785
1786For more details, see the individual descriptions at C<$@>, C<$!>,
1787C<$^E>, and C<$?>.
1788
1789=over 8
1790
1791=item ${^CHILD_ERROR_NATIVE}
1792X<$^CHILD_ERROR_NATIVE>
1793
1794The native status returned by the last pipe close, backtick (C<``>)
1795command, successful call to C<wait()> or C<waitpid()>, or from the
1796C<system()> operator.  On POSIX-like systems this value can be decoded
1797with the WIFEXITED, WEXITSTATUS, WIFSIGNALED, WTERMSIG, WIFSTOPPED,
1798WSTOPSIG and WIFCONTINUED functions provided by the L<POSIX> module.
1799
1800Under VMS this reflects the actual VMS exit status; i.e. it is the
1801same as C<$?> when the pragma C<use vmsish 'status'> is in effect.
1802
1803This variable was added in Perl v5.10.0.
1804
1805=item $EXTENDED_OS_ERROR
1806
1807=item $^E
1808X<$^E> X<$EXTENDED_OS_ERROR>
1809
1810Error information specific to the current operating system.  At the
1811moment, this differs from C<L</$!>> under only VMS, OS/2, and Win32 (and
1812for MacPerl).  On all other platforms, C<$^E> is always just the same
1813as C<$!>.
1814
1815Under VMS, C<$^E> provides the VMS status value from the last system
1816error.  This is more specific information about the last system error
1817than that provided by C<$!>.  This is particularly important when C<$!>
1818is set to B<EVMSERR>.
1819
1820Under OS/2, C<$^E> is set to the error code of the last call to OS/2
1821API either via CRT, or directly from perl.
1822
1823Under Win32, C<$^E> always returns the last error information reported
1824by the Win32 call C<GetLastError()> which describes the last error
1825from within the Win32 API.  Most Win32-specific code will report errors
1826via C<$^E>.  ANSI C and Unix-like calls set C<errno> and so most
1827portable Perl code will report errors via C<$!>.
1828
1829Caveats mentioned in the description of C<L</$!>> generally apply to
1830C<$^E>, also.
1831
1832This variable was added in Perl 5.003.
1833
1834Mnemonic: Extra error explanation.
1835
1836=item $EXCEPTIONS_BEING_CAUGHT
1837
1838=item $^S
1839X<$^S> X<$EXCEPTIONS_BEING_CAUGHT>
1840
1841Current state of the interpreter.
1842
1843	$^S         State
1844	---------   -------------------------------------
1845	undef       Parsing module, eval, or main program
1846	true (1)    Executing an eval
1847	false (0)   Otherwise
1848
1849The first state may happen in C<$SIG{__DIE__}> and C<$SIG{__WARN__}>
1850handlers.
1851
1852The English name $EXCEPTIONS_BEING_CAUGHT is slightly misleading, because
1853the C<undef> value does not indicate whether exceptions are being caught,
1854since compilation of the main program does not catch exceptions.
1855
1856This variable was added in Perl 5.004.
1857
1858=item $WARNING
1859
1860=item $^W
1861X<$^W> X<$WARNING>
1862
1863The current value of the warning switch, initially true if B<-w> was
1864used, false otherwise, but directly modifiable.
1865
1866See also L<warnings>.
1867
1868Mnemonic: related to the B<-w> switch.
1869
1870=item ${^WARNING_BITS}
1871X<${^WARNING_BITS}>
1872
1873The current set of warning checks enabled by the C<use warnings> pragma.
1874It has the same scoping as the C<$^H> and C<%^H> variables.  The exact
1875values are considered internal to the L<warnings> pragma and may change
1876between versions of Perl.
1877
1878This variable was added in Perl v5.6.0.
1879
1880=item $OS_ERROR
1881
1882=item $ERRNO
1883
1884=item $!
1885X<$!> X<$ERRNO> X<$OS_ERROR>
1886
1887When referenced, C<$!> retrieves the current value
1888of the C C<errno> integer variable.
1889If C<$!> is assigned a numerical value, that value is stored in C<errno>.
1890When referenced as a string, C<$!> yields the system error string
1891corresponding to C<errno>.
1892
1893Many system or library calls set C<errno> if they fail,
1894to indicate the cause of failure.  They usually do B<not>
1895set C<errno> to zero if they succeed.  This means C<errno>,
1896hence C<$!>, is meaningful only I<immediately> after a B<failure>:
1897
1898    if (open my $fh, "<", $filename) {
1899		# Here $! is meaningless.
1900		...
1901    }
1902    else {
1903		# ONLY here is $! meaningful.
1904		...
1905		# Already here $! might be meaningless.
1906    }
1907    # Since here we might have either success or failure,
1908    # $! is meaningless.
1909
1910Here, I<meaningless> means that C<$!> may be unrelated to the outcome
1911of the C<open()> operator.  Assignment to C<$!> is similarly ephemeral.
1912It can be used immediately before invoking the C<die()> operator,
1913to set the exit value, or to inspect the system error string
1914corresponding to error I<n>, or to restore C<$!> to a meaningful state.
1915
1916Mnemonic: What just went bang?
1917
1918=item %OS_ERROR
1919
1920=item %ERRNO
1921
1922=item %!
1923X<%!> X<%OS_ERROR> X<%ERRNO>
1924
1925Each element of C<%!> has a true value only if C<$!> is set to that
1926value.  For example, C<$!{ENOENT}> is true if and only if the current
1927value of C<$!> is C<ENOENT>; that is, if the most recent error was "No
1928such file or directory" (or its moral equivalent: not all operating
1929systems give that exact error, and certainly not all languages).  The
1930specific true value is not guaranteed, but in the past has generally
1931been the numeric value of C<$!>.  To check if a particular key is
1932meaningful on your system, use C<exists $!{the_key}>; for a list of legal
1933keys, use C<keys %!>.  See L<Errno> for more information, and also see
1934L</$!>.
1935
1936This variable was added in Perl 5.005.
1937
1938=item $CHILD_ERROR
1939
1940=item $?
1941X<$?> X<$CHILD_ERROR>
1942
1943The status returned by the last pipe close, backtick (C<``>) command,
1944successful call to C<wait()> or C<waitpid()>, or from the C<system()>
1945operator.  This is just the 16-bit status word returned by the
1946traditional Unix C<wait()> system call (or else is made up to look
1947like it).  Thus, the exit value of the subprocess is really (C<<< $? >>
19488 >>>), and C<$? & 127> gives which signal, if any, the process died
1949from, and C<$? & 128> reports whether there was a core dump.
1950
1951Additionally, if the C<h_errno> variable is supported in C, its value
1952is returned via C<$?> if any C<gethost*()> function fails.
1953
1954If you have installed a signal handler for C<SIGCHLD>, the
1955value of C<$?> will usually be wrong outside that handler.
1956
1957Inside an C<END> subroutine C<$?> contains the value that is going to be
1958given to C<exit()>.  You can modify C<$?> in an C<END> subroutine to
1959change the exit status of your program.  For example:
1960
1961    END {
1962	$? = 1 if $? == 255;  # die would make it 255
1963    }
1964
1965Under VMS, the pragma C<use vmsish 'status'> makes C<$?> reflect the
1966actual VMS exit status, instead of the default emulation of POSIX
1967status; see L<perlvms/$?> for details.
1968
1969Mnemonic: similar to B<sh> and B<ksh>.
1970
1971=item $EVAL_ERROR
1972
1973=item $@
1974X<$@> X<$EVAL_ERROR>
1975
1976The Perl error from the last C<eval> operator, i.e. the last exception that
1977was caught.  For C<eval BLOCK>, this is either a runtime error message or the
1978string or reference C<die> was called with.  The C<eval STRING> form also
1979catches syntax errors and other compile time exceptions.
1980
1981If no error occurs, C<eval> sets C<$@> to the empty string.
1982
1983Warning messages are not collected in this variable.  You can, however,
1984set up a routine to process warnings by setting C<$SIG{__WARN__}> as
1985described in L</%SIG>.
1986
1987Mnemonic: Where was the error "at"?
1988
1989=back
1990
1991=head2 Variables related to the interpreter state
1992
1993These variables provide information about the current interpreter state.
1994
1995=over 8
1996
1997=item $COMPILING
1998
1999=item $^C
2000X<$^C> X<$COMPILING>
2001
2002The current value of the flag associated with the B<-c> switch.
2003Mainly of use with B<-MO=...> to allow code to alter its behavior
2004when being compiled, such as for example to C<AUTOLOAD> at compile
2005time rather than normal, deferred loading.  Setting
2006C<$^C = 1> is similar to calling C<B::minus_c>.
2007
2008This variable was added in Perl v5.6.0.
2009
2010=item $DEBUGGING
2011
2012=item $^D
2013X<$^D> X<$DEBUGGING>
2014
2015The current value of the debugging flags.  May be read or set.  Like its
2016L<command-line equivalent|perlrun/B<-D>I<letters>>, you can use numeric
2017or symbolic values, e.g. C<$^D = 10> or C<$^D = "st">.  See
2018L<perlrun/B<-D>I<number>>.  The contents of this variable also affects the
2019debugger operation.  See L<perldebguts/Debugger Internals>.
2020
2021Mnemonic: value of B<-D> switch.
2022
2023=item ${^ENCODING}
2024X<${^ENCODING}>
2025
2026This variable is no longer supported.
2027
2028It used to hold the I<object reference> to the C<Encode> object that was
2029used to convert the source code to Unicode.
2030
2031Its purpose was to allow your non-ASCII Perl
2032scripts not to have to be written in UTF-8; this was
2033useful before editors that worked on UTF-8 encoded text were common, but
2034that was long ago.  It caused problems, such as affecting the operation
2035of other modules that weren't expecting it, causing general mayhem.
2036
2037If you need something like this functionality, it is recommended that use
2038you a simple source filter, such as L<Filter::Encoding>.
2039
2040If you are coming here because code of yours is being adversely affected
2041by someone's use of this variable, you can usually work around it by
2042doing this:
2043
2044 local ${^ENCODING};
2045
2046near the beginning of the functions that are getting broken.  This
2047undefines the variable during the scope of execution of the including
2048function.
2049
2050This variable was added in Perl 5.8.2 and removed in 5.26.0.
2051Setting it to anything other than C<undef> was made fatal in Perl 5.28.0.
2052
2053=item ${^GLOBAL_PHASE}
2054X<${^GLOBAL_PHASE}>
2055
2056The current phase of the perl interpreter.
2057
2058Possible values are:
2059
2060=over 8
2061
2062=item CONSTRUCT
2063
2064The C<PerlInterpreter*> is being constructed via C<perl_construct>.  This
2065value is mostly there for completeness and for use via the
2066underlying C variable C<PL_phase>.  It's not really possible for Perl
2067code to be executed unless construction of the interpreter is
2068finished.
2069
2070=item START
2071
2072This is the global compile-time.  That includes, basically, every
2073C<BEGIN> block executed directly or indirectly from during the
2074compile-time of the top-level program.
2075
2076This phase is not called "BEGIN" to avoid confusion with
2077C<BEGIN>-blocks, as those are executed during compile-time of any
2078compilation unit, not just the top-level program.  A new, localised
2079compile-time entered at run-time, for example by constructs as
2080C<eval "use SomeModule"> are not global interpreter phases, and
2081therefore aren't reflected by C<${^GLOBAL_PHASE}>.
2082
2083=item CHECK
2084
2085Execution of any C<CHECK> blocks.
2086
2087=item INIT
2088
2089Similar to "CHECK", but for C<INIT>-blocks, not C<CHECK> blocks.
2090
2091=item RUN
2092
2093The main run-time, i.e. the execution of C<PL_main_root>.
2094
2095=item END
2096
2097Execution of any C<END> blocks.
2098
2099=item DESTRUCT
2100
2101Global destruction.
2102
2103=back
2104
2105Also note that there's no value for UNITCHECK-blocks.  That's because
2106those are run for each compilation unit individually, and therefore is
2107not a global interpreter phase.
2108
2109Not every program has to go through each of the possible phases, but
2110transition from one phase to another can only happen in the order
2111described in the above list.
2112
2113An example of all of the phases Perl code can see:
2114
2115    BEGIN { print "compile-time: ${^GLOBAL_PHASE}\n" }
2116
2117    INIT  { print "init-time: ${^GLOBAL_PHASE}\n" }
2118
2119    CHECK { print "check-time: ${^GLOBAL_PHASE}\n" }
2120
2121    {
2122        package Print::Phase;
2123
2124        sub new {
2125            my ($class, $time) = @_;
2126            return bless \$time, $class;
2127        }
2128
2129        sub DESTROY {
2130            my $self = shift;
2131            print "$$self: ${^GLOBAL_PHASE}\n";
2132        }
2133    }
2134
2135    print "run-time: ${^GLOBAL_PHASE}\n";
2136
2137    my $runtime = Print::Phase->new(
2138        "lexical variables are garbage collected before END"
2139    );
2140
2141    END   { print "end-time: ${^GLOBAL_PHASE}\n" }
2142
2143    our $destruct = Print::Phase->new(
2144        "package variables are garbage collected after END"
2145    );
2146
2147This will print out
2148
2149    compile-time: START
2150    check-time: CHECK
2151    init-time: INIT
2152    run-time: RUN
2153    lexical variables are garbage collected before END: RUN
2154    end-time: END
2155    package variables are garbage collected after END: DESTRUCT
2156
2157This variable was added in Perl 5.14.0.
2158
2159=item $^H
2160X<$^H>
2161
2162WARNING: This variable is strictly for
2163internal use only.  Its availability,
2164behavior, and contents are subject to change without notice.
2165
2166This variable contains compile-time hints for the Perl interpreter.  At the
2167end of compilation of a BLOCK the value of this variable is restored to the
2168value when the interpreter started to compile the BLOCK.
2169
2170When perl begins to parse any block construct that provides a lexical scope
2171(e.g., eval body, required file, subroutine body, loop body, or conditional
2172block), the existing value of C<$^H> is saved, but its value is left unchanged.
2173When the compilation of the block is completed, it regains the saved value.
2174Between the points where its value is saved and restored, code that
2175executes within BEGIN blocks is free to change the value of C<$^H>.
2176
2177This behavior provides the semantic of lexical scoping, and is used in,
2178for instance, the C<use strict> pragma.
2179
2180The contents should be an integer; different bits of it are used for
2181different pragmatic flags.  Here's an example:
2182
2183    sub add_100 { $^H |= 0x100 }
2184
2185    sub foo {
2186	BEGIN { add_100() }
2187	bar->baz($boon);
2188    }
2189
2190Consider what happens during execution of the BEGIN block.  At this point
2191the BEGIN block has already been compiled, but the body of C<foo()> is still
2192being compiled.  The new value of C<$^H>
2193will therefore be visible only while
2194the body of C<foo()> is being compiled.
2195
2196Substitution of C<BEGIN { add_100() }> block with:
2197
2198    BEGIN { require strict; strict->import('vars') }
2199
2200demonstrates how C<use strict 'vars'> is implemented.  Here's a conditional
2201version of the same lexical pragma:
2202
2203    BEGIN {
2204	require strict; strict->import('vars') if $condition
2205    }
2206
2207This variable was added in Perl 5.003.
2208
2209=item %^H
2210X<%^H>
2211
2212The C<%^H> hash provides the same scoping semantic as C<$^H>.  This makes
2213it useful for implementation of lexically scoped pragmas.  See
2214L<perlpragma>.   All the entries are stringified when accessed at
2215runtime, so only simple values can be accommodated.  This means no
2216pointers to objects, for example.
2217
2218When putting items into C<%^H>, in order to avoid conflicting with other
2219users of the hash there is a convention regarding which keys to use.
2220A module should use only keys that begin with the module's name (the
2221name of its main package) and a "/" character.  For example, a module
2222C<Foo::Bar> should use keys such as C<Foo::Bar/baz>.
2223
2224This variable was added in Perl v5.6.0.
2225
2226=item ${^OPEN}
2227X<${^OPEN}>
2228
2229An internal variable used by PerlIO.  A string in two parts, separated
2230by a C<\0> byte, the first part describes the input layers, the second
2231part describes the output layers.
2232
2233This variable was added in Perl v5.8.0.
2234
2235=item $PERLDB
2236
2237=item $^P
2238X<$^P> X<$PERLDB>
2239
2240The internal variable for debugging support.  The meanings of the
2241various bits are subject to change, but currently indicate:
2242
2243=over 6
2244
2245=item 0x01
2246
2247Debug subroutine enter/exit.
2248
2249=item 0x02
2250
2251Line-by-line debugging.  Causes C<DB::DB()> subroutine to be called for
2252each statement executed.  Also causes saving source code lines (like
22530x400).
2254
2255=item 0x04
2256
2257Switch off optimizations.
2258
2259=item 0x08
2260
2261Preserve more data for future interactive inspections.
2262
2263=item 0x10
2264
2265Keep info about source lines on which a subroutine is defined.
2266
2267=item 0x20
2268
2269Start with single-step on.
2270
2271=item 0x40
2272
2273Use subroutine address instead of name when reporting.
2274
2275=item 0x80
2276
2277Report C<goto &subroutine> as well.
2278
2279=item 0x100
2280
2281Provide informative "file" names for evals based on the place they were compiled.
2282
2283=item 0x200
2284
2285Provide informative names to anonymous subroutines based on the place they
2286were compiled.
2287
2288=item 0x400
2289
2290Save source code lines into C<@{"_<$filename"}>.
2291
2292=item 0x800
2293
2294When saving source, include evals that generate no subroutines.
2295
2296=item 0x1000
2297
2298When saving source, include source that did not compile.
2299
2300=back
2301
2302Some bits may be relevant at compile-time only, some at
2303run-time only.  This is a new mechanism and the details may change.
2304See also L<perldebguts>.
2305
2306=item ${^TAINT}
2307X<${^TAINT}>
2308
2309Reflects if taint mode is on or off.  1 for on (the program was run with
2310B<-T>), 0 for off, -1 when only taint warnings are enabled (i.e. with
2311B<-t> or B<-TU>).
2312
2313This variable is read-only.
2314
2315This variable was added in Perl v5.8.0.
2316
2317=item ${^SAFE_LOCALES}
2318X<${^SAFE_LOCALES}>
2319
2320Reflects if safe locale operations are available to this perl (when the
2321value is 1) or not (the value is 0).  This variable is always 1 if the
2322perl has been compiled without threads.  It is also 1 if this perl is
2323using thread-safe locale operations.  Note that an individual thread may
2324choose to use the global locale (generally unsafe) by calling
2325L<perlapi/switch_to_global_locale>.  This variable currently is still
2326set to 1 in such threads.
2327
2328This variable is read-only.
2329
2330This variable was added in Perl v5.28.0.
2331
2332=item ${^UNICODE}
2333X<${^UNICODE}>
2334
2335Reflects certain Unicode settings of Perl.  See L<perlrun>
2336documentation for the C<-C> switch for more information about
2337the possible values.
2338
2339This variable is set during Perl startup and is thereafter read-only.
2340
2341This variable was added in Perl v5.8.2.
2342
2343=item ${^UTF8CACHE}
2344X<${^UTF8CACHE}>
2345
2346This variable controls the state of the internal UTF-8 offset caching code.
23471 for on (the default), 0 for off, -1 to debug the caching code by checking
2348all its results against linear scans, and panicking on any discrepancy.
2349
2350This variable was added in Perl v5.8.9.  It is subject to change or
2351removal without notice, but is currently used to avoid recalculating the
2352boundaries of multi-byte UTF-8-encoded characters.
2353
2354=item ${^UTF8LOCALE}
2355X<${^UTF8LOCALE}>
2356
2357This variable indicates whether a UTF-8 locale was detected by perl at
2358startup.  This information is used by perl when it's in
2359adjust-utf8ness-to-locale mode (as when run with the C<-CL> command-line
2360switch); see L<perlrun> for more info on this.
2361
2362This variable was added in Perl v5.8.8.
2363
2364=back
2365
2366=head2 Deprecated and removed variables
2367
2368Deprecating a variable announces the intent of the perl maintainers to
2369eventually remove the variable from the language.  It may still be
2370available despite its status.  Using a deprecated variable triggers
2371a warning.
2372
2373Once a variable is removed, its use triggers an error telling you
2374the variable is unsupported.
2375
2376See L<perldiag> for details about error messages.
2377
2378=over 8
2379
2380=item $#
2381X<$#>
2382
2383C<$#> was a variable that could be used to format printed numbers.
2384After a deprecation cycle, its magic was removed in Perl v5.10.0 and
2385using it now triggers a warning: C<$# is no longer supported>.
2386
2387This is not the sigil you use in front of an array name to get the
2388last index, like C<$#array>.  That's still how you get the last index
2389of an array in Perl.  The two have nothing to do with each other.
2390
2391Deprecated in Perl 5.
2392
2393Removed in Perl v5.10.0.
2394
2395=item $*
2396X<$*>
2397
2398C<$*> was a variable that you could use to enable multiline matching.
2399After a deprecation cycle, its magic was removed in Perl v5.10.0.
2400Using it now triggers a warning: C<$* is no longer supported>.
2401You should use the C</s> and C</m> regexp modifiers instead.
2402
2403Deprecated in Perl 5.
2404
2405Removed in Perl v5.10.0.
2406
2407=item $[
2408X<$[>
2409
2410This variable stores the index of the first element in an array, and
2411of the first character in a substring.  The default is 0, but you could
2412theoretically set it to 1 to make Perl behave more like B<awk> (or Fortran)
2413when subscripting and when evaluating the index() and substr() functions.
2414
2415As of release 5 of Perl, assignment to C<$[> is treated as a compiler
2416directive, and cannot influence the behavior of any other file.
2417(That's why you can only assign compile-time constants to it.)
2418Its use is highly discouraged.
2419
2420Prior to Perl v5.10.0, assignment to C<$[> could be seen from outer lexical
2421scopes in the same file, unlike other compile-time directives (such as
2422L<strict>).  Using local() on it would bind its value strictly to a lexical
2423block.  Now it is always lexically scoped.
2424
2425As of Perl v5.16.0, it is implemented by the L<arybase> module.
2426
2427As of Perl v5.30.0, or under C<use v5.16>, or C<no feature "array_base">,
2428C<$[> no longer has any effect, and always contains 0.
2429Assigning 0 to it is permitted, but any other value will produce an error.
2430
2431Mnemonic: [ begins subscripts.
2432
2433Deprecated in Perl v5.12.0.
2434
2435=back
2436
2437=cut
2438