2 X<subroutine> X<function>
4 perlsub - Perl subroutines
8 To declare subroutines:
9 X<subroutine, declaration> X<sub>
11 sub NAME; # A "forward" declaration.
12 sub NAME(PROTO); # ditto, but with prototypes
13 sub NAME : ATTRS; # with attributes
14 sub NAME(PROTO) : ATTRS; # with attributes and prototypes
16 sub NAME BLOCK # A declaration and a definition.
17 sub NAME(PROTO) BLOCK # ditto, but with prototypes
18 sub NAME : ATTRS BLOCK # with attributes
19 sub NAME(PROTO) : ATTRS BLOCK # with prototypes and attributes
21 use feature 'signatures';
22 sub NAME(SIG) BLOCK # with signature
23 sub NAME :ATTRS (SIG) BLOCK # with signature, attributes
24 sub NAME :prototype(PROTO) (SIG) BLOCK # with signature, prototype
26 To define an anonymous subroutine at runtime:
27 X<subroutine, anonymous>
29 $subref = sub BLOCK; # no proto
30 $subref = sub (PROTO) BLOCK; # with proto
31 $subref = sub : ATTRS BLOCK; # with attributes
32 $subref = sub (PROTO) : ATTRS BLOCK; # with proto and attributes
34 use feature 'signatures';
35 $subref = sub (SIG) BLOCK; # with signature
36 $subref = sub : ATTRS(SIG) BLOCK; # with signature, attributes
38 To import subroutines:
41 use MODULE qw(NAME1 NAME2 NAME3);
44 X<subroutine, call> X<call>
46 NAME(LIST); # & is optional with parentheses.
47 NAME LIST; # Parentheses optional if predeclared/imported.
48 &NAME(LIST); # Circumvent prototypes.
49 &NAME; # Makes current @_ visible to called subroutine.
53 Like many languages, Perl provides for user-defined subroutines.
54 These may be located anywhere in the main program, loaded in from
55 other files via the C<do>, C<require>, or C<use> keywords, or
56 generated on the fly using C<eval> or anonymous subroutines.
57 You can even call a function indirectly using a variable containing
58 its name or a CODE reference.
60 The Perl model for function call and return values is simple: all
61 functions are passed as parameters one single flat list of scalars, and
62 all functions likewise return to their caller one single flat list of
63 scalars. Any arrays or hashes in these call and return lists will
64 collapse, losing their identities--but you may always use
65 pass-by-reference instead to avoid this. Both call and return lists may
66 contain as many or as few scalar elements as you'd like. (Often a
67 function without an explicit return statement is called a subroutine, but
68 there's really no difference from Perl's perspective.)
69 X<subroutine, parameter> X<parameter>
71 Any arguments passed in show up in the array C<@_>.
72 (They may also show up in lexical variables introduced by a signature;
73 see L</Signatures> below.) Therefore, if
74 you called a function with two arguments, those would be stored in
75 C<$_[0]> and C<$_[1]>. The array C<@_> is a local array, but its
76 elements are aliases for the actual scalar parameters. In particular,
77 if an element C<$_[0]> is updated, the corresponding argument is
78 updated (or an error occurs if it is not updatable). If an argument
79 is an array or hash element which did not exist when the function
80 was called, that element is created only when (and if) it is modified
81 or a reference to it is taken. (Some earlier versions of Perl
82 created the element whether or not the element was assigned to.)
83 Assigning to the whole array C<@_> removes that aliasing, and does
84 not update any arguments.
85 X<subroutine, argument> X<argument> X<@_>
87 A C<return> statement may be used to exit a subroutine, optionally
88 specifying the returned value, which will be evaluated in the
89 appropriate context (list, scalar, or void) depending on the context of
90 the subroutine call. If you specify no return value, the subroutine
91 returns an empty list in list context, the undefined value in scalar
92 context, or nothing in void context. If you return one or more
93 aggregates (arrays and hashes), these will be flattened together into
94 one large indistinguishable list.
96 If no C<return> is found and if the last statement is an expression, its
97 value is returned. If the last statement is a loop control structure
98 like a C<foreach> or a C<while>, the returned value is unspecified. The
99 empty sub returns the empty list.
100 X<subroutine, return value> X<return value> X<return>
102 Aside from an experimental facility (see L</Signatures> below),
103 Perl does not have named formal parameters. In practice all you
104 do is assign to a C<my()> list of these. Variables that aren't
105 declared to be private are global variables. For gory details
106 on creating private variables, see L</"Private Variables via my()">
107 and L</"Temporary Values via local()">. To create protected
108 environments for a set of functions in a separate package (and
109 probably a separate file), see L<perlmod/"Packages">.
110 X<formal parameter> X<parameter, formal>
117 $max = $foo if $max < $foo;
121 $bestday = max($mon,$tue,$wed,$thu,$fri);
125 # get a line, combining continuation lines
126 # that start with whitespace
129 $thisline = $lookahead; # global variables!
130 LINE: while (defined($lookahead = <STDIN>)) {
131 if ($lookahead =~ /^[ \t]/) {
132 $thisline .= $lookahead;
141 $lookahead = <STDIN>; # get first line
142 while (defined($line = get_line())) {
146 Assigning to a list of private variables to name your arguments:
149 my($key, $value) = @_;
150 $Foo{$key} = $value unless $Foo{$key};
153 Because the assignment copies the values, this also has the effect
154 of turning call-by-reference into call-by-value. Otherwise a
155 function is free to do in-place modifications of C<@_> and change
157 X<call-by-reference> X<call-by-value>
159 upcase_in($v1, $v2); # this changes $v1 and $v2
161 for (@_) { tr/a-z/A-Z/ }
164 You aren't allowed to modify constants in this way, of course. If an
165 argument were actually literal and you tried to change it, you'd take a
166 (presumably fatal) exception. For example, this won't work:
167 X<call-by-reference> X<call-by-value>
169 upcase_in("frederick");
171 It would be much safer if the C<upcase_in()> function
172 were written to return a copy of its parameters instead
173 of changing them in place:
175 ($v3, $v4) = upcase($v1, $v2); # this doesn't change $v1 and $v2
177 return unless defined wantarray; # void context, do nothing
179 for (@parms) { tr/a-z/A-Z/ }
180 return wantarray ? @parms : $parms[0];
183 Notice how this (unprototyped) function doesn't care whether it was
184 passed real scalars or arrays. Perl sees all arguments as one big,
185 long, flat parameter list in C<@_>. This is one area where
186 Perl's simple argument-passing style shines. The C<upcase()>
187 function would work perfectly well without changing the C<upcase()>
188 definition even if we fed it things like this:
190 @newlist = upcase(@list1, @list2);
191 @newlist = upcase( split /:/, $var );
193 Do not, however, be tempted to do this:
195 (@a, @b) = upcase(@list1, @list2);
197 Like the flattened incoming parameter list, the return list is also
198 flattened on return. So all you have managed to do here is stored
199 everything in C<@a> and made C<@b> empty. See
200 L</Pass by Reference> for alternatives.
202 A subroutine may be called using an explicit C<&> prefix. The
203 C<&> is optional in modern Perl, as are parentheses if the
204 subroutine has been predeclared. The C<&> is I<not> optional
205 when just naming the subroutine, such as when it's used as
206 an argument to defined() or undef(). Nor is it optional when you
207 want to do an indirect subroutine call with a subroutine name or
208 reference using the C<&$subref()> or C<&{$subref}()> constructs,
209 although the C<< $subref->() >> notation solves that problem.
210 See L<perlref> for more about all that.
213 Subroutines may be called recursively. If a subroutine is called
214 using the C<&> form, the argument list is optional, and if omitted,
215 no C<@_> array is set up for the subroutine: the C<@_> array at the
216 time of the call is visible to subroutine instead. This is an
217 efficiency mechanism that new users may wish to avoid.
220 &foo(1,2,3); # pass three arguments
221 foo(1,2,3); # the same
223 foo(); # pass a null list
226 &foo; # foo() get current args, like foo(@_) !!
227 foo; # like foo() IFF sub foo predeclared, else "foo"
229 Not only does the C<&> form make the argument list optional, it also
230 disables any prototype checking on arguments you do provide. This
231 is partly for historical reasons, and partly for having a convenient way
232 to cheat if you know what you're doing. See L</Prototypes> below.
235 Since Perl 5.16.0, the C<__SUB__> token is available under C<use feature
236 'current_sub'> and C<use 5.16.0>. It will evaluate to a reference to the
237 currently-running sub, which allows for recursive calls without knowing
238 your subroutine's name.
241 my $factorial = sub {
244 return($x * __SUB__->( $x - 1 ) );
247 The behavior of C<__SUB__> within a regex code block (such as C</(?{...})/>)
248 is subject to change.
250 Subroutines whose names are in all upper case are reserved to the Perl
251 core, as are modules whose names are in all lower case. A subroutine in
252 all capitals is a loosely-held convention meaning it will be called
253 indirectly by the run-time system itself, usually due to a triggered event.
254 Subroutines whose name start with a left parenthesis are also reserved the
255 same way. The following is a list of some subroutines that currently do
256 special, pre-defined things.
260 =item documented later in this document
264 =item documented in L<perlmod>
266 C<CLONE>, C<CLONE_SKIP>
268 =item documented in L<perlobj>
272 =item documented in L<perltie>
274 C<BINMODE>, C<CLEAR>, C<CLOSE>, C<DELETE>, C<DESTROY>, C<EOF>, C<EXISTS>,
275 C<EXTEND>, C<FETCH>, C<FETCHSIZE>, C<FILENO>, C<FIRSTKEY>, C<GETC>,
276 C<NEXTKEY>, C<OPEN>, C<POP>, C<PRINT>, C<PRINTF>, C<PUSH>, C<READ>,
277 C<READLINE>, C<SCALAR>, C<SEEK>, C<SHIFT>, C<SPLICE>, C<STORE>,
278 C<STORESIZE>, C<TELL>, C<TIEARRAY>, C<TIEHANDLE>, C<TIEHASH>,
279 C<TIESCALAR>, C<UNSHIFT>, C<UNTIE>, C<WRITE>
281 =item documented in L<PerlIO::via>
283 C<BINMODE>, C<CLEARERR>, C<CLOSE>, C<EOF>, C<ERROR>, C<FDOPEN>, C<FILENO>,
284 C<FILL>, C<FLUSH>, C<OPEN>, C<POPPED>, C<PUSHED>, C<READ>, C<SEEK>,
285 C<SETLINEBUF>, C<SYSOPEN>, C<TELL>, C<UNREAD>, C<UTF8>, C<WRITE>
287 =item documented in L<perlfunc>
289 L<< C<import> | perlfunc/use >>, L<< C<unimport> | perlfunc/use >>,
290 L<< C<INC> | perlfunc/require >>
292 =item documented in L<UNIVERSAL>
296 =item documented in L<perldebguts>
298 C<DB::DB>, C<DB::sub>, C<DB::lsub>, C<DB::goto>, C<DB::postponed>
300 =item undocumented, used internally by the L<overload> feature
302 any starting with C<(>
306 The C<BEGIN>, C<UNITCHECK>, C<CHECK>, C<INIT> and C<END> subroutines
307 are not so much subroutines as named special code blocks, of which you
308 can have more than one in a package, and which you can B<not> call
309 explicitly. See L<perlmod/"BEGIN, UNITCHECK, CHECK, INIT and END">
313 B<WARNING>: Subroutine signatures are experimental. The feature may be
314 modified or removed in future versions of Perl.
316 Perl has an experimental facility to allow a subroutine's formal
317 parameters to be introduced by special syntax, separate from the
318 procedural code of the subroutine body. The formal parameter list
319 is known as a I<signature>. The facility must be enabled first by a
320 pragmatic declaration, C<use feature 'signatures'>, and it will produce
321 a warning unless the "experimental::signatures" warnings category is
324 The signature is part of a subroutine's body. Normally the body of a
325 subroutine is simply a braced block of code, but when using a signature,
326 the signature is a parenthesised list that goes immediately before the
327 block, after any name or attributes.
331 sub foo :lvalue ($a, $b = 1, @c) { .... }
333 The signature declares lexical variables that are
334 in scope for the block. When the subroutine is called, the signature
335 takes control first. It populates the signature variables from the
336 list of arguments that were passed. If the argument list doesn't meet
337 the requirements of the signature, then it will throw an exception.
338 When the signature processing is complete, control passes to the block.
340 Positional parameters are handled by simply naming scalar variables in
341 the signature. For example,
343 sub foo ($left, $right) {
344 return $left + $right;
347 takes two positional parameters, which must be filled at runtime by
348 two arguments. By default the parameters are mandatory, and it is
349 not permitted to pass more arguments than expected. So the above is
353 die "Too many arguments for subroutine" unless @_ <= 2;
354 die "Too few arguments for subroutine" unless @_ >= 2;
357 return $left + $right;
360 An argument can be ignored by omitting the main part of the name from
361 a parameter declaration, leaving just a bare C<$> sigil. For example,
363 sub foo ($first, $, $third) {
364 return "first=$first, third=$third";
367 Although the ignored argument doesn't go into a variable, it is still
368 mandatory for the caller to pass it.
370 A positional parameter is made optional by giving a default value,
371 separated from the parameter name by C<=>:
373 sub foo ($left, $right = 0) {
374 return $left + $right;
377 The above subroutine may be called with either one or two arguments.
378 The default value expression is evaluated when the subroutine is called,
379 so it may provide different default values for different calls. It is
380 only evaluated if the argument was actually omitted from the call.
384 sub foo ($thing, $id = $auto_id++) {
385 print "$thing has ID $id";
388 automatically assigns distinct sequential IDs to things for which no
389 ID was supplied by the caller. A default value expression may also
390 refer to parameters earlier in the signature, making the default for
391 one parameter vary according to the earlier parameters. For example,
393 sub foo ($first_name, $surname, $nickname = $first_name) {
394 print "$first_name $surname is known as \"$nickname\"";
397 An optional parameter can be nameless just like a mandatory parameter.
400 sub foo ($thing, $ = 1) {
404 The parameter's default value will still be evaluated if the corresponding
405 argument isn't supplied, even though the value won't be stored anywhere.
406 This is in case evaluating it has important side effects. However, it
407 will be evaluated in void context, so if it doesn't have side effects
408 and is not trivial it will generate a warning if the "void" warning
409 category is enabled. If a nameless optional parameter's default value
410 is not important, it may be omitted just as the parameter's name was:
412 sub foo ($thing, $=) {
416 Optional positional parameters must come after all mandatory positional
417 parameters. (If there are no mandatory positional parameters then an
418 optional positional parameters can be the first thing in the signature.)
419 If there are multiple optional positional parameters and not enough
420 arguments are supplied to fill them all, they will be filled from left
423 After positional parameters, additional arguments may be captured in a
424 slurpy parameter. The simplest form of this is just an array variable:
426 sub foo ($filter, @inputs) {
427 print $filter->($_) foreach @inputs;
430 With a slurpy parameter in the signature, there is no upper limit on how
431 many arguments may be passed. A slurpy array parameter may be nameless
432 just like a positional parameter, in which case its only effect is to
433 turn off the argument limit that would otherwise apply:
435 sub foo ($thing, @) {
439 A slurpy parameter may instead be a hash, in which case the arguments
440 available to it are interpreted as alternating keys and values.
441 There must be as many keys as values: if there is an odd argument then
442 an exception will be thrown. Keys will be stringified, and if there are
443 duplicates then the later instance takes precedence over the earlier,
444 as with standard hash construction.
446 sub foo ($filter, %inputs) {
447 print $filter->($_, $inputs{$_}) foreach sort keys %inputs;
450 A slurpy hash parameter may be nameless just like other kinds of
451 parameter. It still insists that the number of arguments available to
452 it be even, even though they're not being put into a variable.
454 sub foo ($thing, %) {
458 A slurpy parameter, either array or hash, must be the last thing in the
459 signature. It may follow mandatory and optional positional parameters;
460 it may also be the only thing in the signature. Slurpy parameters cannot
461 have default values: if no arguments are supplied for them then you get
462 an empty array or empty hash.
464 A signature may be entirely empty, in which case all it does is check
465 that the caller passed no arguments:
471 When using a signature, the arguments are still available in the special
472 array variable C<@_>, in addition to the lexical variables of the
473 signature. There is a difference between the two ways of accessing the
474 arguments: C<@_> I<aliases> the arguments, but the signature variables
475 get I<copies> of the arguments. So writing to a signature variable
476 only changes that variable, and has no effect on the caller's variables,
477 but writing to an element of C<@_> modifies whatever the caller used to
478 supply that argument.
480 There is a potential syntactic ambiguity between signatures and prototypes
481 (see L</Prototypes>), because both start with an opening parenthesis and
482 both can appear in some of the same places, such as just after the name
483 in a subroutine declaration. For historical reasons, when signatures
484 are not enabled, any opening parenthesis in such a context will trigger
485 very forgiving prototype parsing. Most signatures will be interpreted
486 as prototypes in those circumstances, but won't be valid prototypes.
487 (A valid prototype cannot contain any alphabetic character.) This will
488 lead to somewhat confusing error messages.
490 To avoid ambiguity, when signatures are enabled the special syntax
491 for prototypes is disabled. There is no attempt to guess whether a
492 parenthesised group was intended to be a prototype or a signature.
493 To give a subroutine a prototype under these circumstances, use a
494 L<prototype attribute|attributes/Built-in Attributes>. For example,
496 sub foo :prototype($) { $_[0] }
498 It is entirely possible for a subroutine to have both a prototype and
499 a signature. They do different jobs: the prototype affects compilation
500 of calls to the subroutine, and the signature puts argument values into
501 lexical variables at runtime. You can therefore write
503 sub foo :prototype($$) ($left, $right) {
504 return $left + $right;
507 The prototype attribute, and any other attributes, must come before
508 the signature. The signature always immediately precedes the block of
509 the subroutine's body.
511 =head2 Private Variables via my()
512 X<my> X<variable, lexical> X<lexical> X<lexical variable> X<scope, lexical>
513 X<lexical scope> X<attributes, my>
517 my $foo; # declare $foo lexically local
518 my (@wid, %get); # declare list of variables local
519 my $foo = "flurp"; # declare $foo lexical, and init it
520 my @oof = @bar; # declare @oof lexical, and init it
521 my $x : Foo = $y; # similar, with an attribute applied
523 B<WARNING>: The use of attribute lists on C<my> declarations is still
524 evolving. The current semantics and interface are subject to change.
525 See L<attributes> and L<Attribute::Handlers>.
527 The C<my> operator declares the listed variables to be lexically
528 confined to the enclosing block, conditional
529 (C<if>/C<unless>/C<elsif>/C<else>), loop
530 (C<for>/C<foreach>/C<while>/C<until>/C<continue>), subroutine, C<eval>,
531 or C<do>/C<require>/C<use>'d file. If more than one value is listed, the
532 list must be placed in parentheses. All listed elements must be
533 legal lvalues. Only alphanumeric identifiers may be lexically
534 scoped--magical built-ins like C<$/> must currently be C<local>ized
535 with C<local> instead.
537 Unlike dynamic variables created by the C<local> operator, lexical
538 variables declared with C<my> are totally hidden from the outside
539 world, including any called subroutines. This is true if it's the
540 same subroutine called from itself or elsewhere--every call gets
544 This doesn't mean that a C<my> variable declared in a statically
545 enclosing lexical scope would be invisible. Only dynamic scopes
546 are cut off. For example, the C<bumpx()> function below has access
547 to the lexical $x variable because both the C<my> and the C<sub>
548 occurred at the same scope, presumably file scope.
553 An C<eval()>, however, can see lexical variables of the scope it is
554 being evaluated in, so long as the names aren't hidden by declarations within
555 the C<eval()> itself. See L<perlref>.
558 The parameter list to my() may be assigned to if desired, which allows you
559 to initialize your variables. (If no initializer is given for a
560 particular variable, it is created with the undefined value.) Commonly
561 this is used to name input parameters to a subroutine. Examples:
563 $arg = "fred"; # "global" variable
565 print "$arg thinks the root is $n\n";
566 fred thinks the root is 3
569 my $arg = shift; # name doesn't matter
574 The C<my> is simply a modifier on something you might assign to. So when
575 you do assign to variables in its argument list, C<my> doesn't
576 change whether those variables are viewed as a scalar or an array. So
578 my ($foo) = <STDIN>; # WRONG?
581 both supply a list context to the right-hand side, while
585 supplies a scalar context. But the following declares only one variable:
587 my $foo, $bar = 1; # WRONG
589 That has the same effect as
594 The declared variable is not introduced (is not visible) until after
595 the current statement. Thus,
599 can be used to initialize a new $x with the value of the old $x, and
602 my $x = 123 and $x == 123
604 is false unless the old $x happened to have the value C<123>.
606 Lexical scopes of control structures are not bounded precisely by the
607 braces that delimit their controlled blocks; control expressions are
608 part of that scope, too. Thus in the loop
610 while (my $line = <>) {
616 the scope of $line extends from its declaration throughout the rest of
617 the loop construct (including the C<continue> clause), but not beyond
618 it. Similarly, in the conditional
620 if ((my $answer = <STDIN>) =~ /^yes$/i) {
622 } elsif ($answer =~ /^no$/i) {
626 die "'$answer' is neither 'yes' nor 'no'";
629 the scope of $answer extends from its declaration through the rest
630 of that conditional, including any C<elsif> and C<else> clauses,
631 but not beyond it. See L<perlsyn/"Simple Statements"> for information
632 on the scope of variables in statements with modifiers.
634 The C<foreach> loop defaults to scoping its index variable dynamically
635 in the manner of C<local>. However, if the index variable is
636 prefixed with the keyword C<my>, or if there is already a lexical
637 by that name in scope, then a new lexical is created instead. Thus
641 for my $i (1, 2, 3) {
645 the scope of $i extends to the end of the loop, but not beyond it,
646 rendering the value of $i inaccessible within C<some_function()>.
649 Some users may wish to encourage the use of lexically scoped variables.
650 As an aid to catching implicit uses to package variables,
651 which are always global, if you say
655 then any variable mentioned from there to the end of the enclosing
656 block must either refer to a lexical variable, be predeclared via
657 C<our> or C<use vars>, or else must be fully qualified with the package name.
658 A compilation error results otherwise. An inner block may countermand
659 this with C<no strict 'vars'>.
661 A C<my> has both a compile-time and a run-time effect. At compile
662 time, the compiler takes notice of it. The principal usefulness
663 of this is to quiet C<use strict 'vars'>, but it is also essential
664 for generation of closures as detailed in L<perlref>. Actual
665 initialization is delayed until run time, though, so it gets executed
666 at the appropriate time, such as each time through a loop, for
669 Variables declared with C<my> are not part of any package and are therefore
670 never fully qualified with the package name. In particular, you're not
671 allowed to try to make a package variable (or other global) lexical:
673 my $pack::var; # ERROR! Illegal syntax
675 In fact, a dynamic variable (also known as package or global variables)
676 are still accessible using the fully qualified C<::> notation even while a
677 lexical of the same name is also visible:
682 print "$x and $::x\n";
684 That will print out C<20> and C<10>.
686 You may declare C<my> variables at the outermost scope of a file
687 to hide any such identifiers from the world outside that file. This
688 is similar in spirit to C's static variables when they are used at
689 the file level. To do this with a subroutine requires the use of
690 a closure (an anonymous function that accesses enclosing lexicals).
691 If you want to create a private subroutine that cannot be called
692 from outside that block, it can declare a lexical variable containing
693 an anonymous sub reference:
695 my $secret_version = '1.001-beta';
696 my $secret_sub = sub { print $secret_version };
699 As long as the reference is never returned by any function within the
700 module, no outside module can see the subroutine, because its name is not in
701 any package's symbol table. Remember that it's not I<REALLY> called
702 C<$some_pack::secret_version> or anything; it's just $secret_version,
703 unqualified and unqualifiable.
705 This does not work with object methods, however; all object methods
706 have to be in the symbol table of some package to be found. See
707 L<perlref/"Function Templates"> for something of a work-around to
710 =head2 Persistent Private Variables
711 X<state> X<state variable> X<static> X<variable, persistent> X<variable, static> X<closure>
713 There are two ways to build persistent private variables in Perl 5.10.
714 First, you can simply use the C<state> feature. Or, you can use closures,
715 if you want to stay compatible with releases older than 5.10.
717 =head3 Persistent variables via state()
719 Beginning with Perl 5.10.0, you can declare variables with the C<state>
720 keyword in place of C<my>. For that to work, though, you must have
721 enabled that feature beforehand, either by using the C<feature> pragma, or
722 by using C<-E> on one-liners (see L<feature>). Beginning with Perl 5.16,
723 the C<CORE::state> form does not require the
726 The C<state> keyword creates a lexical variable (following the same scoping
727 rules as C<my>) that persists from one subroutine call to the next. If a
728 state variable resides inside an anonymous subroutine, then each copy of
729 the subroutine has its own copy of the state variable. However, the value
730 of the state variable will still persist between calls to the same copy of
731 the anonymous subroutine. (Don't forget that C<sub { ... }> creates a new
732 subroutine each time it is executed.)
734 For example, the following code maintains a private counter, incremented
735 each time the gimme_another() function is called:
738 sub gimme_another { state $x; return ++$x }
740 And this example uses anonymous subroutines to create separate counters:
744 return sub { state $x; return ++$x }
747 Also, since C<$x> is lexical, it can't be reached or modified by any Perl
750 When combined with variable declaration, simple assignment to C<state>
751 variables (as in C<state $x = 42>) is executed only the first time. When such
752 statements are evaluated subsequent times, the assignment is ignored. The
753 behavior of assignment to C<state> declarations where the left hand side
754 of the assignment involves any parentheses is currently undefined.
756 =head3 Persistent variables with closures
758 Just because a lexical variable is lexically (also called statically)
759 scoped to its enclosing block, C<eval>, or C<do> FILE, this doesn't mean that
760 within a function it works like a C static. It normally works more
761 like a C auto, but with implicit garbage collection.
763 Unlike local variables in C or C++, Perl's lexical variables don't
764 necessarily get recycled just because their scope has exited.
765 If something more permanent is still aware of the lexical, it will
766 stick around. So long as something else references a lexical, that
767 lexical won't be freed--which is as it should be. You wouldn't want
768 memory being free until you were done using it, or kept around once you
769 were done. Automatic garbage collection takes care of this for you.
771 This means that you can pass back or save away references to lexical
772 variables, whereas to return a pointer to a C auto is a grave error.
773 It also gives us a way to simulate C's function statics. Here's a
774 mechanism for giving a function private variables with both lexical
775 scoping and a static lifetime. If you do want to create something like
776 C's static variables, just enclose the whole function in an extra block,
777 and put the static variable outside the function but in the block.
782 return ++$secret_val;
785 # $secret_val now becomes unreachable by the outside
786 # world, but retains its value between calls to gimme_another
788 If this function is being sourced in from a separate file
789 via C<require> or C<use>, then this is probably just fine. If it's
790 all in the main program, you'll need to arrange for the C<my>
791 to be executed early, either by putting the whole block above
792 your main program, or more likely, placing merely a C<BEGIN>
793 code block around it to make sure it gets executed before your program
799 return ++$secret_val;
803 See L<perlmod/"BEGIN, UNITCHECK, CHECK, INIT and END"> about the
804 special triggered code blocks, C<BEGIN>, C<UNITCHECK>, C<CHECK>,
807 If declared at the outermost scope (the file scope), then lexicals
808 work somewhat like C's file statics. They are available to all
809 functions in that same file declared below them, but are inaccessible
810 from outside that file. This strategy is sometimes used in modules
811 to create private variables that the whole module can see.
813 =head2 Temporary Values via local()
814 X<local> X<scope, dynamic> X<dynamic scope> X<variable, local>
815 X<variable, temporary>
817 B<WARNING>: In general, you should be using C<my> instead of C<local>, because
818 it's faster and safer. Exceptions to this include the global punctuation
819 variables, global filehandles and formats, and direct manipulation of the
820 Perl symbol table itself. C<local> is mostly used when the current value
821 of a variable must be visible to called subroutines.
825 # localization of values
827 local $foo; # make $foo dynamically local
828 local (@wid, %get); # make list of variables local
829 local $foo = "flurp"; # make $foo dynamic, and init it
830 local @oof = @bar; # make @oof dynamic, and init it
832 local $hash{key} = "val"; # sets a local value for this hash entry
833 delete local $hash{key}; # delete this entry for the current block
834 local ($cond ? $v1 : $v2); # several types of lvalues support
837 # localization of symbols
839 local *FH; # localize $FH, @FH, %FH, &FH ...
840 local *merlyn = *randal; # now $merlyn is really $randal, plus
841 # @merlyn is really @randal, etc
842 local *merlyn = 'randal'; # SAME THING: promote 'randal' to *randal
843 local *merlyn = \$randal; # just alias $merlyn, not @merlyn etc
845 A C<local> modifies its listed variables to be "local" to the
846 enclosing block, C<eval>, or C<do FILE>--and to I<any subroutine
847 called from within that block>. A C<local> just gives temporary
848 values to global (meaning package) variables. It does I<not> create
849 a local variable. This is known as dynamic scoping. Lexical scoping
850 is done with C<my>, which works more like C's auto declarations.
852 Some types of lvalues can be localized as well: hash and array elements
853 and slices, conditionals (provided that their result is always
854 localizable), and symbolic references. As for simple variables, this
855 creates new, dynamically scoped values.
857 If more than one variable or expression is given to C<local>, they must be
858 placed in parentheses. This operator works
859 by saving the current values of those variables in its argument list on a
860 hidden stack and restoring them upon exiting the block, subroutine, or
861 eval. This means that called subroutines can also reference the local
862 variable, but not the global one. The argument list may be assigned to if
863 desired, which allows you to initialize your local variables. (If no
864 initializer is given for a particular variable, it is created with an
867 Because C<local> is a run-time operator, it gets executed each time
868 through a loop. Consequently, it's more efficient to localize your
869 variables outside the loop.
871 =head3 Grammatical note on local()
874 A C<local> is simply a modifier on an lvalue expression. When you assign to
875 a C<local>ized variable, the C<local> doesn't change whether its list is viewed
876 as a scalar or an array. So
878 local($foo) = <STDIN>;
879 local @FOO = <STDIN>;
881 both supply a list context to the right-hand side, while
883 local $foo = <STDIN>;
885 supplies a scalar context.
887 =head3 Localization of special variables
888 X<local, special variable>
890 If you localize a special variable, you'll be giving a new value to it,
891 but its magic won't go away. That means that all side-effects related
892 to this magic still work with the localized value.
894 This feature allows code like this to work :
896 # Read the whole contents of FILE in $slurp
897 { local $/ = undef; $slurp = <FILE>; }
899 Note, however, that this restricts localization of some values ; for
900 example, the following statement dies, as of perl 5.10.0, with an error
901 I<Modification of a read-only value attempted>, because the $1 variable is
902 magical and read-only :
906 One exception is the default scalar variable: starting with perl 5.14
907 C<local($_)> will always strip all magic from $_, to make it possible
908 to safely reuse $_ in a subroutine.
910 B<WARNING>: Localization of tied arrays and hashes does not currently
912 This will be fixed in a future release of Perl; in the meantime, avoid
913 code that relies on any particular behavior of localising tied arrays
914 or hashes (localising individual elements is still okay).
915 See L<perl58delta/"Localising Tied Arrays and Hashes Is Broken"> for more
919 =head3 Localization of globs
920 X<local, glob> X<glob>
926 creates a whole new symbol table entry for the glob C<name> in the
927 current package. That means that all variables in its glob slot ($name,
928 @name, %name, &name, and the C<name> filehandle) are dynamically reset.
930 This implies, among other things, that any magic eventually carried by
931 those variables is locally lost. In other words, saying C<local */>
932 will not have any effect on the internal value of the input record
935 =head3 Localization of elements of composite types
936 X<local, composite type element> X<local, array element> X<local, hash element>
938 It's also worth taking a moment to explain what happens when you
939 C<local>ize a member of a composite type (i.e. an array or hash element).
940 In this case, the element is C<local>ized I<by name>. This means that
941 when the scope of the C<local()> ends, the saved value will be
942 restored to the hash element whose key was named in the C<local()>, or
943 the array element whose index was named in the C<local()>. If that
944 element was deleted while the C<local()> was in effect (e.g. by a
945 C<delete()> from a hash or a C<shift()> of an array), it will spring
946 back into existence, possibly extending an array and filling in the
947 skipped elements with C<undef>. For instance, if you say
949 %hash = ( 'This' => 'is', 'a' => 'test' );
953 local($hash{'a'}) = 'drill';
954 while (my $e = pop(@ary)) {
959 $hash{'only a'} = 'test';
963 print join(' ', map { "$_ $hash{$_}" } sort keys %hash),".\n";
964 print "The array has ",scalar(@ary)," elements: ",
965 join(', ', map { defined $_ ? $_ : 'undef' } @ary),"\n";
972 This is a test only a test.
973 The array has 6 elements: 0, 1, 2, undef, undef, 5
975 The behavior of local() on non-existent members of composite
976 types is subject to change in future.
978 =head3 Localized deletion of elements of composite types
979 X<delete> X<local, composite type element> X<local, array element> X<local, hash element>
981 You can use the C<delete local $array[$idx]> and C<delete local $hash{key}>
982 constructs to delete a composite type entry for the current block and restore
983 it when it ends. They return the array/hash value before the localization,
984 which means that they are respectively equivalent to
987 my $val = $array[$idx];
996 my $val = $hash{key};
1002 except that for those the C<local> is
1003 scoped to the C<do> block. Slices are
1012 my $a = delete local $hash{a};
1017 my @nums = delete local @$a[0, 2]
1019 # $a is [ undef, 8 ]
1021 $a[0] = 999; # will be erased when the scope ends
1023 # $a is back to [ 7, 8, 9 ]
1026 # %hash is back to its original state
1028 =head2 Lvalue subroutines
1029 X<lvalue> X<subroutine, lvalue>
1031 It is possible to return a modifiable value from a subroutine.
1032 To do this, you have to declare the subroutine to return an lvalue.
1035 sub canmod : lvalue {
1036 $val; # or: return $val;
1042 canmod() = 5; # assigns to $val
1043 nomod() = 5; # ERROR
1045 The scalar/list context for the subroutine and for the right-hand
1046 side of assignment is determined as if the subroutine call is replaced
1047 by a scalar. For example, consider:
1049 data(2,3) = get_data(3,4);
1051 Both subroutines here are called in a scalar context, while in:
1053 (data(2,3)) = get_data(3,4);
1057 (data(2),data(3)) = get_data(3,4);
1059 all the subroutines are called in a list context.
1061 Lvalue subroutines are convenient, but you have to keep in mind that,
1062 when used with objects, they may violate encapsulation. A normal
1063 mutator can check the supplied argument before setting the attribute
1064 it is protecting, an lvalue subroutine cannot. If you require any
1065 special processing when storing and retrieving the values, consider
1066 using the CPAN module Sentinel or something similar.
1068 =head2 Lexical Subroutines
1069 X<my sub> X<state sub> X<our sub> X<subroutine, lexical>
1071 Beginning with Perl 5.18, you can declare a private subroutine with C<my>
1072 or C<state>. As with state variables, the C<state> keyword is only
1073 available under C<use feature 'state'> or C<use 5.010> or higher.
1075 Prior to Perl 5.26, lexical subroutines were deemed experimental and were
1076 available only under the C<use feature 'lexical_subs'> pragma. They also
1077 produced a warning unless the "experimental::lexical_subs" warnings
1078 category was disabled.
1080 These subroutines are only visible within the block in which they are
1081 declared, and only after that declaration:
1083 # Include these two lines if your code is intended to run under Perl
1084 # versions earlier than 5.26.
1085 no warnings "experimental::lexical_subs";
1086 use feature 'lexical_subs';
1088 foo(); # calls the package/global subroutine
1090 foo(); # also calls the package subroutine
1092 foo(); # calls "state" sub
1093 my $ref = \&foo; # take a reference to "state" sub
1096 bar(); # calls "my" sub
1098 You can't (directly) write a recursive lexical subroutine:
1105 This example fails because C<baz()> refers to the package/global subroutine
1106 C<baz>, not the lexical subroutine currently being defined.
1108 The solution is to use L<C<__SUB__>|perlfunc/__SUB__>:
1111 __SUB__->(); # calls itself
1114 It is possible to predeclare a lexical subroutine. The C<sub foo {...}>
1115 subroutine definition syntax respects any previous C<my sub;> or C<state sub;>
1116 declaration. Using this to define recursive subroutines is a bad idea,
1119 my sub baz; # predeclaration
1120 sub baz { # define the "my" sub
1121 baz(); # WRONG: calls itself, but leaks memory
1124 Just like C<< my $f; $f = sub { $f->() } >>, this example leaks memory. The
1125 name C<baz> is a reference to the subroutine, and the subroutine uses the name
1126 C<baz>; they keep each other alive (see L<perlref/Circular References>).
1128 =head3 C<state sub> vs C<my sub>
1130 What is the difference between "state" subs and "my" subs? Each time that
1131 execution enters a block when "my" subs are declared, a new copy of each
1132 sub is created. "State" subroutines persist from one execution of the
1133 containing block to the next.
1135 So, in general, "state" subroutines are faster. But "my" subs are
1136 necessary if you want to create closures:
1141 ... do something with $x ...
1146 In this example, a new C<$x> is created when C<whatever> is called, and
1147 also a new C<inner>, which can see the new C<$x>. A "state" sub will only
1148 see the C<$x> from the first call to C<whatever>.
1150 =head3 C<our> subroutines
1152 Like C<our $variable>, C<our sub> creates a lexical alias to the package
1153 subroutine of the same name.
1155 The two main uses for this are to switch back to using the package sub
1156 inside an inner scope:
1163 # need to use the outer foo here
1169 and to make a subroutine visible to other packages in the same scope:
1171 package MySneakyModule;
1173 our sub do_something { ... }
1175 sub do_something_with_caller {
1177 () = caller 1; # sets @DB::args
1178 do_something(@args); # uses MySneakyModule::do_something
1181 =head2 Passing Symbol Table Entries (typeglobs)
1184 B<WARNING>: The mechanism described in this section was originally
1185 the only way to simulate pass-by-reference in older versions of
1186 Perl. While it still works fine in modern versions, the new reference
1187 mechanism is generally easier to work with. See below.
1189 Sometimes you don't want to pass the value of an array to a subroutine
1190 but rather the name of it, so that the subroutine can modify the global
1191 copy of it rather than working with a local copy. In perl you can
1192 refer to all objects of a particular name by prefixing the name
1193 with a star: C<*foo>. This is often known as a "typeglob", because the
1194 star on the front can be thought of as a wildcard match for all the
1195 funny prefix characters on variables and subroutines and such.
1197 When evaluated, the typeglob produces a scalar value that represents
1198 all the objects of that name, including any filehandle, format, or
1199 subroutine. When assigned to, it causes the name mentioned to refer to
1200 whatever C<*> value was assigned to it. Example:
1203 local(*someary) = @_;
1204 foreach $elem (@someary) {
1211 Scalars are already passed by reference, so you can modify
1212 scalar arguments without using this mechanism by referring explicitly
1213 to C<$_[0]> etc. You can modify all the elements of an array by passing
1214 all the elements as scalars, but you have to use the C<*> mechanism (or
1215 the equivalent reference mechanism) to C<push>, C<pop>, or change the size of
1216 an array. It will certainly be faster to pass the typeglob (or reference).
1218 Even if you don't want to modify an array, this mechanism is useful for
1219 passing multiple arrays in a single LIST, because normally the LIST
1220 mechanism will merge all the array values so that you can't extract out
1221 the individual arrays. For more on typeglobs, see
1222 L<perldata/"Typeglobs and Filehandles">.
1224 =head2 When to Still Use local()
1225 X<local> X<variable, local>
1227 Despite the existence of C<my>, there are still three places where the
1228 C<local> operator still shines. In fact, in these three places, you
1229 I<must> use C<local> instead of C<my>.
1235 You need to give a global variable a temporary value, especially $_.
1237 The global variables, like C<@ARGV> or the punctuation variables, must be
1238 C<local>ized with C<local()>. This block reads in F</etc/motd>, and splits
1239 it up into chunks separated by lines of equal signs, which are placed
1243 local @ARGV = ("/etc/motd");
1246 @Fields = split /^\s*=+\s*$/;
1249 It particular, it's important to C<local>ize $_ in any routine that assigns
1250 to it. Look out for implicit assignments in C<while> conditionals.
1254 You need to create a local file or directory handle or a local function.
1256 A function that needs a filehandle of its own must use
1257 C<local()> on a complete typeglob. This can be used to create new symbol
1261 local (*READER, *WRITER); # not my!
1262 pipe (READER, WRITER) or die "pipe: $!";
1263 return (*READER, *WRITER);
1265 ($head, $tail) = ioqueue();
1267 See the Symbol module for a way to create anonymous symbol table
1270 Because assignment of a reference to a typeglob creates an alias, this
1271 can be used to create what is effectively a local function, or at least,
1275 local *grow = \&shrink; # only until this block exits
1276 grow(); # really calls shrink()
1277 move(); # if move() grow()s, it shrink()s too
1279 grow(); # get the real grow() again
1281 See L<perlref/"Function Templates"> for more about manipulating
1282 functions by name in this way.
1286 You want to temporarily change just one element of an array or hash.
1288 You can C<local>ize just one element of an aggregate. Usually this
1289 is done on dynamics:
1292 local $SIG{INT} = 'IGNORE';
1293 funct(); # uninterruptible
1295 # interruptibility automatically restored here
1297 But it also works on lexically declared aggregates.
1301 =head2 Pass by Reference
1302 X<pass by reference> X<pass-by-reference> X<reference>
1304 If you want to pass more than one array or hash into a function--or
1305 return them from it--and have them maintain their integrity, then
1306 you're going to have to use an explicit pass-by-reference. Before you
1307 do that, you need to understand references as detailed in L<perlref>.
1308 This section may not make much sense to you otherwise.
1310 Here are a few simple examples. First, let's pass in several arrays
1311 to a function and have it C<pop> all of then, returning a new list
1312 of all their former last elements:
1314 @tailings = popmany ( \@a, \@b, \@c, \@d );
1319 foreach $aref ( @_ ) {
1320 push @retlist, pop @$aref;
1325 Here's how you might write a function that returns a
1326 list of keys occurring in all the hashes passed to it:
1328 @common = inter( \%foo, \%bar, \%joe );
1330 my ($k, $href, %seen); # locals
1331 foreach $href (@_) {
1332 while ( $k = each %$href ) {
1336 return grep { $seen{$_} == @_ } keys %seen;
1339 So far, we're using just the normal list return mechanism.
1340 What happens if you want to pass or return a hash? Well,
1341 if you're using only one of them, or you don't mind them
1342 concatenating, then the normal calling convention is ok, although
1345 Where people get into trouble is here:
1347 (@a, @b) = func(@c, @d);
1349 (%a, %b) = func(%c, %d);
1351 That syntax simply won't work. It sets just C<@a> or C<%a> and
1352 clears the C<@b> or C<%b>. Plus the function didn't get passed
1353 into two separate arrays or hashes: it got one long list in C<@_>,
1356 If you can arrange for everyone to deal with this through references, it's
1357 cleaner code, although not so nice to look at. Here's a function that
1358 takes two array references as arguments, returning the two array elements
1359 in order of how many elements they have in them:
1361 ($aref, $bref) = func(\@c, \@d);
1362 print "@$aref has more than @$bref\n";
1364 my ($cref, $dref) = @_;
1365 if (@$cref > @$dref) {
1366 return ($cref, $dref);
1368 return ($dref, $cref);
1372 It turns out that you can actually do this also:
1374 (*a, *b) = func(\@c, \@d);
1375 print "@a has more than @b\n";
1377 local (*c, *d) = @_;
1385 Here we're using the typeglobs to do symbol table aliasing. It's
1386 a tad subtle, though, and also won't work if you're using C<my>
1387 variables, because only globals (even in disguise as C<local>s)
1388 are in the symbol table.
1390 If you're passing around filehandles, you could usually just use the bare
1391 typeglob, like C<*STDOUT>, but typeglobs references work, too.
1397 print $fh "her um well a hmmm\n";
1400 $rec = get_rec(\*STDIN);
1403 return scalar <$fh>;
1406 If you're planning on generating new filehandles, you could do this.
1407 Notice to pass back just the bare *FH, not its reference.
1412 return open (FH, $path) ? *FH : undef;
1416 X<prototype> X<subroutine, prototype>
1418 Perl supports a very limited kind of compile-time argument checking
1419 using function prototyping. This can be declared in either the PROTO
1420 section or with a L<prototype attribute|attributes/Built-in Attributes>.
1421 If you declare either of
1424 sub mypush :prototype(\@@)
1426 then C<mypush()> takes arguments exactly like C<push()> does.
1428 If subroutine signatures are enabled (see L</Signatures>), then
1429 the shorter PROTO syntax is unavailable, because it would clash with
1430 signatures. In that case, a prototype can only be declared in the form
1434 function declaration must be visible at compile time. The prototype
1435 affects only interpretation of new-style calls to the function,
1436 where new-style is defined as not using the C<&> character. In
1437 other words, if you call it like a built-in function, then it behaves
1438 like a built-in function. If you call it like an old-fashioned
1439 subroutine, then it behaves like an old-fashioned subroutine. It
1440 naturally falls out from this rule that prototypes have no influence
1441 on subroutine references like C<\&foo> or on indirect subroutine
1442 calls like C<&{$subref}> or C<< $subref->() >>.
1444 Method calls are not influenced by prototypes either, because the
1445 function to be called is indeterminate at compile time, since
1446 the exact code called depends on inheritance.
1448 Because the intent of this feature is primarily to let you define
1449 subroutines that work like built-in functions, here are prototypes
1450 for some other functions that parse almost exactly like the
1451 corresponding built-in.
1453 Declared as Called as
1455 sub mylink ($$) mylink $old, $new
1456 sub myvec ($$$) myvec $var, $offset, 1
1457 sub myindex ($$;$) myindex &getstring, "substr"
1458 sub mysyswrite ($$$;$) mysyswrite $buf, 0, length($buf) - $off, $off
1459 sub myreverse (@) myreverse $a, $b, $c
1460 sub myjoin ($@) myjoin ":", $a, $b, $c
1461 sub mypop (\@) mypop @array
1462 sub mysplice (\@$$@) mysplice @array, 0, 2, @pushme
1463 sub mykeys (\[%@]) mykeys %{$hashref}
1464 sub myopen (*;$) myopen HANDLE, $name
1465 sub mypipe (**) mypipe READHANDLE, WRITEHANDLE
1466 sub mygrep (&@) mygrep { /foo/ } $a, $b, $c
1467 sub myrand (;$) myrand 42
1468 sub mytime () mytime
1470 Any backslashed prototype character represents an actual argument
1471 that must start with that character (optionally preceded by C<my>,
1472 C<our> or C<local>), with the exception of C<$>, which will
1473 accept any scalar lvalue expression, such as C<$foo = 7> or
1474 C<< my_function()->[0] >>. The value passed as part of C<@_> will be a
1475 reference to the actual argument given in the subroutine call,
1476 obtained by applying C<\> to that argument.
1478 You can use the C<\[]> backslash group notation to specify more than one
1479 allowed argument type. For example:
1481 sub myref (\[$@%&*])
1483 will allow calling myref() as
1491 and the first argument of myref() will be a reference to
1492 a scalar, an array, a hash, a code, or a glob.
1494 Unbackslashed prototype characters have special meanings. Any
1495 unbackslashed C<@> or C<%> eats all remaining arguments, and forces
1496 list context. An argument represented by C<$> forces scalar context. An
1497 C<&> requires an anonymous subroutine, which, if passed as the first
1498 argument, does not require the C<sub> keyword or a subsequent comma.
1500 A C<*> allows the subroutine to accept a bareword, constant, scalar expression,
1501 typeglob, or a reference to a typeglob in that slot. The value will be
1502 available to the subroutine either as a simple scalar, or (in the latter
1503 two cases) as a reference to the typeglob. If you wish to always convert
1504 such arguments to a typeglob reference, use Symbol::qualify_to_ref() as
1507 use Symbol 'qualify_to_ref';
1510 my $fh = qualify_to_ref(shift, caller);
1514 The C<+> prototype is a special alternative to C<$> that will act like
1515 C<\[@%]> when given a literal array or hash variable, but will otherwise
1516 force scalar context on the argument. This is useful for functions which
1517 should accept either a literal array or an array reference as the argument:
1521 die "Not an array or arrayref" unless ref $aref eq 'ARRAY';
1525 When using the C<+> prototype, your function must check that the argument
1526 is of an acceptable type.
1528 A semicolon (C<;>) separates mandatory arguments from optional arguments.
1529 It is redundant before C<@> or C<%>, which gobble up everything else.
1531 As the last character of a prototype, or just before a semicolon, a C<@>
1532 or a C<%>, you can use C<_> in place of C<$>: if this argument is not
1533 provided, C<$_> will be used instead.
1535 Note how the last three examples in the table above are treated
1536 specially by the parser. C<mygrep()> is parsed as a true list
1537 operator, C<myrand()> is parsed as a true unary operator with unary
1538 precedence the same as C<rand()>, and C<mytime()> is truly without
1539 arguments, just like C<time()>. That is, if you say
1543 you'll get C<mytime() + 2>, not C<mytime(2)>, which is how it would be parsed
1544 without a prototype. If you want to force a unary function to have the
1545 same precedence as a list operator, add C<;> to the end of the prototype:
1547 sub mygetprotobynumber($;);
1548 mygetprotobynumber $a > $b; # parsed as mygetprotobynumber($a > $b)
1550 The interesting thing about C<&> is that you can generate new syntax with it,
1551 provided it's in the initial position:
1555 my($try,$catch) = @_;
1562 sub catch (&) { $_[0] }
1567 /phooey/ and print "unphooey\n";
1570 That prints C<"unphooey">. (Yes, there are still unresolved
1571 issues having to do with visibility of C<@_>. I'm ignoring that
1572 question for the moment. (But note that if we make C<@_> lexically
1573 scoped, those anonymous subroutines can act like closures... (Gee,
1574 is this sounding a little Lispish? (Never mind.))))
1576 And here's a reimplementation of the Perl C<grep> operator:
1583 push(@result, $_) if &$code;
1588 Some folks would prefer full alphanumeric prototypes. Alphanumerics have
1589 been intentionally left out of prototypes for the express purpose of
1590 someday in the future adding named, formal parameters. The current
1591 mechanism's main goal is to let module writers provide better diagnostics
1592 for module users. Larry feels the notation quite understandable to Perl
1593 programmers, and that it will not intrude greatly upon the meat of the
1594 module, nor make it harder to read. The line noise is visually
1595 encapsulated into a small pill that's easy to swallow.
1597 If you try to use an alphanumeric sequence in a prototype you will
1598 generate an optional warning - "Illegal character in prototype...".
1599 Unfortunately earlier versions of Perl allowed the prototype to be
1600 used as long as its prefix was a valid prototype. The warning may be
1601 upgraded to a fatal error in a future version of Perl once the
1602 majority of offending code is fixed.
1604 It's probably best to prototype new functions, not retrofit prototyping
1605 into older ones. That's because you must be especially careful about
1606 silent impositions of differing list versus scalar contexts. For example,
1607 if you decide that a function should take just one parameter, like this:
1611 print "you gave me $n\n";
1614 and someone has been calling it with an array or expression
1618 func( $text =~ /\w+/g );
1620 Then you've just supplied an automatic C<scalar> in front of their
1621 argument, which can be more than a bit surprising. The old C<@foo>
1622 which used to hold one thing doesn't get passed in. Instead,
1623 C<func()> now gets passed in a C<1>; that is, the number of elements
1624 in C<@foo>. And the C<m//g> gets called in scalar context so instead of a
1625 list of words it returns a boolean result and advances C<pos($text)>. Ouch!
1627 If a sub has both a PROTO and a BLOCK, the prototype is not applied
1628 until after the BLOCK is completely defined. This means that a recursive
1629 function with a prototype has to be predeclared for the prototype to take
1637 This is all very powerful, of course, and should be used only in moderation
1638 to make the world a better place.
1640 =head2 Constant Functions
1643 Functions with a prototype of C<()> are potential candidates for
1644 inlining. If the result after optimization and constant folding
1645 is either a constant or a lexically-scoped scalar which has no other
1646 references, then it will be used in place of function calls made
1647 without C<&>. Calls made using C<&> are never inlined. (See
1648 F<constant.pm> for an easy way to declare most constants.)
1650 The following functions would all be inlined:
1652 sub pi () { 3.14159 } # Not exact, but close.
1653 sub PI () { 4 * atan2 1, 1 } # As good as it gets,
1654 # and it's inlined, too!
1658 sub FLAG_FOO () { 1 << 8 }
1659 sub FLAG_BAR () { 1 << 9 }
1660 sub FLAG_MASK () { FLAG_FOO | FLAG_BAR }
1662 sub OPT_BAZ () { not (0x1B58 & FLAG_MASK) }
1664 sub N () { int(OPT_BAZ) / 3 }
1666 sub FOO_SET () { 1 if FLAG_MASK & FLAG_FOO }
1667 sub FOO_SET2 () { if (FLAG_MASK & FLAG_FOO) { 1 } }
1669 (Be aware that the last example was not always inlined in Perl 5.20 and
1670 earlier, which did not behave consistently with subroutines containing
1671 inner scopes.) You can countermand inlining by using an explicit
1682 sub bonk_val () { return 12345 }
1684 As alluded to earlier you can also declare inlined subs dynamically at
1685 BEGIN time if their body consists of a lexically-scoped scalar which
1686 has no other references. Only the first example here will be inlined:
1691 *INLINED = sub () { $var };
1698 *NOT_INLINED = sub () { $var };
1701 A not so obvious caveat with this (see [RT #79908]) is that the
1702 variable will be immediately inlined, and will stop behaving like a
1703 normal lexical variable, e.g. this will print C<79907>, not C<79908>:
1707 *RT_79908 = sub () { $x };
1710 print RT_79908(); # prints 79907
1712 As of Perl 5.22, this buggy behavior, while preserved for backward
1713 compatibility, is detected and emits a deprecation warning. If you want
1714 the subroutine to be inlined (with no warning), make sure the variable is
1715 not used in a context where it could be modified aside from where it is
1721 *INLINED = sub () { $x };
1723 # Warns. Future Perl versions will stop inlining it.
1727 *ALSO_INLINED = sub () { $x };
1730 Perl 5.22 also introduces the experimental "const" attribute as an
1731 alternative. (Disable the "experimental::const_attr" warnings if you want
1732 to use it.) When applied to an anonymous subroutine, it forces the sub to
1733 be called when the C<sub> expression is evaluated. The return value is
1734 captured and turned into a constant subroutine:
1737 *INLINED = sub : const { $x };
1740 The return value of C<INLINED> in this example will always be 54321,
1741 regardless of later modifications to $x. You can also put any arbitrary
1742 code inside the sub, at it will be executed immediately and its return
1743 value captured the same way.
1745 If you really want a subroutine with a C<()> prototype that returns a
1746 lexical variable you can easily force it to not be inlined by adding
1747 an explicit C<return>:
1751 *RT_79908 = sub () { return $x };
1754 print RT_79908(); # prints 79908
1756 The easiest way to tell if a subroutine was inlined is by using
1757 L<B::Deparse>. Consider this example of two subroutines returning
1758 C<1>, one with a C<()> prototype causing it to be inlined, and one
1759 without (with deparse output truncated for clarity):
1761 $ perl -MO=Deparse -le 'sub ONE { 1 } if (ONE) { print ONE if ONE }'
1766 print ONE() if ONE ;
1768 $ perl -MO=Deparse -le 'sub ONE () { 1 } if (ONE) { print ONE if ONE }'
1774 If you redefine a subroutine that was eligible for inlining, you'll
1775 get a warning by default. You can use this warning to tell whether or
1776 not a particular subroutine is considered inlinable, since it's
1777 different than the warning for overriding non-inlined subroutines:
1779 $ perl -e 'sub one () {1} sub one () {2}'
1780 Constant subroutine one redefined at -e line 1.
1781 $ perl -we 'sub one {1} sub one {2}'
1782 Subroutine one redefined at -e line 1.
1784 The warning is considered severe enough not to be affected by the
1785 B<-w> switch (or its absence) because previously compiled invocations
1786 of the function will still be using the old value of the function. If
1787 you need to be able to redefine the subroutine, you need to ensure
1788 that it isn't inlined, either by dropping the C<()> prototype (which
1789 changes calling semantics, so beware) or by thwarting the inlining
1790 mechanism in some other way, e.g. by adding an explicit C<return>, as
1793 sub not_inlined () { return 23 }
1795 =head2 Overriding Built-in Functions
1796 X<built-in> X<override> X<CORE> X<CORE::GLOBAL>
1798 Many built-in functions may be overridden, though this should be tried
1799 only occasionally and for good reason. Typically this might be
1800 done by a package attempting to emulate missing built-in functionality
1801 on a non-Unix system.
1803 Overriding may be done only by importing the name from a module at
1804 compile time--ordinary predeclaration isn't good enough. However, the
1805 C<use subs> pragma lets you, in effect, predeclare subs
1806 via the import syntax, and these names may then override built-in ones:
1808 use subs 'chdir', 'chroot', 'chmod', 'chown';
1812 To unambiguously refer to the built-in form, precede the
1813 built-in name with the special package qualifier C<CORE::>. For example,
1814 saying C<CORE::open()> always refers to the built-in C<open()>, even
1815 if the current package has imported some other subroutine called
1816 C<&open()> from elsewhere. Even though it looks like a regular
1817 function call, it isn't: the CORE:: prefix in that case is part of Perl's
1818 syntax, and works for any keyword, regardless of what is in the CORE
1819 package. Taking a reference to it, that is, C<\&CORE::open>, only works
1820 for some keywords. See L<CORE>.
1822 Library modules should not in general export built-in names like C<open>
1823 or C<chdir> as part of their default C<@EXPORT> list, because these may
1824 sneak into someone else's namespace and change the semantics unexpectedly.
1825 Instead, if the module adds that name to C<@EXPORT_OK>, then it's
1826 possible for a user to import the name explicitly, but not implicitly.
1827 That is, they could say
1831 and it would import the C<open> override. But if they said
1835 they would get the default imports without overrides.
1837 The foregoing mechanism for overriding built-in is restricted, quite
1838 deliberately, to the package that requests the import. There is a second
1839 method that is sometimes applicable when you wish to override a built-in
1840 everywhere, without regard to namespace boundaries. This is achieved by
1841 importing a sub into the special namespace C<CORE::GLOBAL::>. Here is an
1842 example that quite brazenly replaces the C<glob> operator with something
1843 that understands regular expressions.
1848 @EXPORT_OK = 'glob';
1854 my $where = ($sym =~ s/^GLOBAL_// ? 'CORE::GLOBAL' : caller(0));
1855 $pkg->export($where, $sym, @_);
1861 if (opendir my $d, '.') {
1862 @got = grep /$pat/, readdir $d;
1869 And here's how it could be (ab)used:
1871 #use REGlob 'GLOBAL_glob'; # override glob() in ALL namespaces
1873 use REGlob 'glob'; # override glob() in Foo:: only
1874 print for <^[a-z_]+\.pm\$>; # show all pragmatic modules
1876 The initial comment shows a contrived, even dangerous example.
1877 By overriding C<glob> globally, you would be forcing the new (and
1878 subversive) behavior for the C<glob> operator for I<every> namespace,
1879 without the complete cognizance or cooperation of the modules that own
1880 those namespaces. Naturally, this should be done with extreme caution--if
1881 it must be done at all.
1883 The C<REGlob> example above does not implement all the support needed to
1884 cleanly override perl's C<glob> operator. The built-in C<glob> has
1885 different behaviors depending on whether it appears in a scalar or list
1886 context, but our C<REGlob> doesn't. Indeed, many perl built-in have such
1887 context sensitive behaviors, and these must be adequately supported by
1888 a properly written override. For a fully functional example of overriding
1889 C<glob>, study the implementation of C<File::DosGlob> in the standard
1892 When you override a built-in, your replacement should be consistent (if
1893 possible) with the built-in native syntax. You can achieve this by using
1894 a suitable prototype. To get the prototype of an overridable built-in,
1895 use the C<prototype> function with an argument of C<"CORE::builtin_name">
1896 (see L<perlfunc/prototype>).
1898 Note however that some built-ins can't have their syntax expressed by a
1899 prototype (such as C<system> or C<chomp>). If you override them you won't
1900 be able to fully mimic their original syntax.
1902 The built-ins C<do>, C<require> and C<glob> can also be overridden, but due
1903 to special magic, their original syntax is preserved, and you don't have
1904 to define a prototype for their replacements. (You can't override the
1905 C<do BLOCK> syntax, though).
1907 C<require> has special additional dark magic: if you invoke your
1908 C<require> replacement as C<require Foo::Bar>, it will actually receive
1909 the argument C<"Foo/Bar.pm"> in @_. See L<perlfunc/require>.
1911 And, as you'll have noticed from the previous example, if you override
1912 C<glob>, the C<< <*> >> glob operator is overridden as well.
1914 In a similar fashion, overriding the C<readline> function also overrides
1915 the equivalent I/O operator C<< <FILEHANDLE> >>. Also, overriding
1916 C<readpipe> also overrides the operators C<``> and C<qx//>.
1918 Finally, some built-ins (e.g. C<exists> or C<grep>) can't be overridden.
1921 X<autoloading> X<AUTOLOAD>
1923 If you call a subroutine that is undefined, you would ordinarily
1924 get an immediate, fatal error complaining that the subroutine doesn't
1925 exist. (Likewise for subroutines being used as methods, when the
1926 method doesn't exist in any base class of the class's package.)
1927 However, if an C<AUTOLOAD> subroutine is defined in the package or
1928 packages used to locate the original subroutine, then that
1929 C<AUTOLOAD> subroutine is called with the arguments that would have
1930 been passed to the original subroutine. The fully qualified name
1931 of the original subroutine magically appears in the global $AUTOLOAD
1932 variable of the same package as the C<AUTOLOAD> routine. The name
1933 is not passed as an ordinary argument because, er, well, just
1934 because, that's why. (As an exception, a method call to a nonexistent
1935 C<import> or C<unimport> method is just skipped instead. Also, if
1936 the AUTOLOAD subroutine is an XSUB, there are other ways to retrieve the
1937 subroutine name. See L<perlguts/Autoloading with XSUBs> for details.)
1940 Many C<AUTOLOAD> routines load in a definition for the requested
1941 subroutine using eval(), then execute that subroutine using a special
1942 form of goto() that erases the stack frame of the C<AUTOLOAD> routine
1943 without a trace. (See the source to the standard module documented
1944 in L<AutoLoader>, for example.) But an C<AUTOLOAD> routine can
1945 also just emulate the routine and never define it. For example,
1946 let's pretend that a function that wasn't defined should just invoke
1947 C<system> with those arguments. All you'd do is:
1950 our $AUTOLOAD; # keep 'use strict' happy
1951 my $program = $AUTOLOAD;
1952 $program =~ s/.*:://;
1953 system($program, @_);
1959 In fact, if you predeclare functions you want to call that way, you don't
1960 even need parentheses:
1962 use subs qw(date who ls);
1967 A more complete example of this is the Shell module on CPAN, which
1968 can treat undefined subroutine calls as calls to external programs.
1970 Mechanisms are available to help modules writers split their modules
1971 into autoloadable files. See the standard AutoLoader module
1972 described in L<AutoLoader> and in L<AutoSplit>, the standard
1973 SelfLoader modules in L<SelfLoader>, and the document on adding C
1974 functions to Perl code in L<perlxs>.
1976 =head2 Subroutine Attributes
1977 X<attribute> X<subroutine, attribute> X<attrs>
1979 A subroutine declaration or definition may have a list of attributes
1980 associated with it. If such an attribute list is present, it is
1981 broken up at space or colon boundaries and treated as though a
1982 C<use attributes> had been seen. See L<attributes> for details
1983 about what attributes are currently supported.
1984 Unlike the limitation with the obsolescent C<use attrs>, the
1985 C<sub : ATTRLIST> syntax works to associate the attributes with
1986 a pre-declaration, and not just with a subroutine definition.
1988 The attributes must be valid as simple identifier names (without any
1989 punctuation other than the '_' character). They may have a parameter
1990 list appended, which is only checked for whether its parentheses ('(',')')
1993 Examples of valid syntax (even though the attributes are unknown):
1995 sub fnord (&\%) : switch(10,foo(7,3)) : expensive;
1996 sub plugh () : Ugly('\(") :Bad;
1997 sub xyzzy : _5x5 { ... }
1999 Examples of invalid syntax:
2001 sub fnord : switch(10,foo(); # ()-string not balanced
2002 sub snoid : Ugly('('); # ()-string not balanced
2003 sub xyzzy : 5x5; # "5x5" not a valid identifier
2004 sub plugh : Y2::north; # "Y2::north" not a simple identifier
2005 sub snurt : foo + bar; # "+" not a colon or space
2007 The attribute list is passed as a list of constant strings to the code
2008 which associates them with the subroutine. In particular, the second example
2009 of valid syntax above currently looks like this in terms of how it's
2012 use attributes __PACKAGE__, \&plugh, q[Ugly('\(")], 'Bad';
2014 For further details on attribute lists and their manipulation,
2015 see L<attributes> and L<Attribute::Handlers>.
2019 See L<perlref/"Function Templates"> for more about references and closures.
2020 See L<perlxs> if you'd like to learn about calling C subroutines from Perl.
2021 See L<perlembed> if you'd like to learn about calling Perl subroutines from C.
2022 See L<perlmod> to learn about bundling up your functions in separate files.
2023 See L<perlmodlib> to learn what library modules come standard on your system.
2024 See L<perlootut> to learn how to make object method calls.