This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
[perl #47365] Docs for \$ prototypes
[perl5.git] / pod / perlsub.pod
CommitLineData
a0d0e21e 1=head1 NAME
d74e8afc 2X<subroutine> X<function>
a0d0e21e
LW
3
4perlsub - Perl subroutines
5
6=head1 SYNOPSIS
7
8To declare subroutines:
d74e8afc 9X<subroutine, declaration> X<sub>
a0d0e21e 10
09bef843
SB
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
cb1a09d0 15
09bef843
SB
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
a0d0e21e 20
748a9306 21To define an anonymous subroutine at runtime:
d74e8afc 22X<subroutine, anonymous>
748a9306 23
09bef843
SB
24 $subref = sub BLOCK; # no proto
25 $subref = sub (PROTO) BLOCK; # with proto
26 $subref = sub : ATTRS BLOCK; # with attributes
27 $subref = sub (PROTO) : ATTRS BLOCK; # with proto and attributes
748a9306 28
a0d0e21e 29To import subroutines:
d74e8afc 30X<import>
a0d0e21e 31
19799a22 32 use MODULE qw(NAME1 NAME2 NAME3);
a0d0e21e
LW
33
34To call subroutines:
d74e8afc 35X<subroutine, call> X<call>
a0d0e21e 36
5f05dabc 37 NAME(LIST); # & is optional with parentheses.
54310121 38 NAME LIST; # Parentheses optional if predeclared/imported.
19799a22 39 &NAME(LIST); # Circumvent prototypes.
5a964f20 40 &NAME; # Makes current @_ visible to called subroutine.
a0d0e21e
LW
41
42=head1 DESCRIPTION
43
19799a22
GS
44Like many languages, Perl provides for user-defined subroutines.
45These may be located anywhere in the main program, loaded in from
46other files via the C<do>, C<require>, or C<use> keywords, or
be3174d2 47generated on the fly using C<eval> or anonymous subroutines.
19799a22
GS
48You can even call a function indirectly using a variable containing
49its name or a CODE reference.
cb1a09d0
AD
50
51The Perl model for function call and return values is simple: all
52functions are passed as parameters one single flat list of scalars, and
53all functions likewise return to their caller one single flat list of
54scalars. Any arrays or hashes in these call and return lists will
55collapse, losing their identities--but you may always use
56pass-by-reference instead to avoid this. Both call and return lists may
57contain as many or as few scalar elements as you'd like. (Often a
58function without an explicit return statement is called a subroutine, but
19799a22 59there's really no difference from Perl's perspective.)
d74e8afc 60X<subroutine, parameter> X<parameter>
19799a22
GS
61
62Any arguments passed in show up in the array C<@_>. Therefore, if
63you called a function with two arguments, those would be stored in
64C<$_[0]> and C<$_[1]>. The array C<@_> is a local array, but its
65elements are aliases for the actual scalar parameters. In particular,
66if an element C<$_[0]> is updated, the corresponding argument is
67updated (or an error occurs if it is not updatable). If an argument
68is an array or hash element which did not exist when the function
69was called, that element is created only when (and if) it is modified
70or a reference to it is taken. (Some earlier versions of Perl
71created the element whether or not the element was assigned to.)
72Assigning to the whole array C<@_> removes that aliasing, and does
73not update any arguments.
d74e8afc 74X<subroutine, argument> X<argument> X<@_>
19799a22 75
dbb128be
XN
76A C<return> statement may be used to exit a subroutine, optionally
77specifying the returned value, which will be evaluated in the
78appropriate context (list, scalar, or void) depending on the context of
79the subroutine call. If you specify no return value, the subroutine
80returns an empty list in list context, the undefined value in scalar
81context, or nothing in void context. If you return one or more
82aggregates (arrays and hashes), these will be flattened together into
83one large indistinguishable list.
84
85If no C<return> is found and if the last statement is an expression, its
9a989771
RGS
86value is returned. If the last statement is a loop control structure
87like a C<foreach> or a C<while>, the returned value is unspecified. The
88empty sub returns the empty list.
d74e8afc 89X<subroutine, return value> X<return value> X<return>
19799a22
GS
90
91Perl does not have named formal parameters. In practice all you
92do is assign to a C<my()> list of these. Variables that aren't
93declared to be private are global variables. For gory details
94on creating private variables, see L<"Private Variables via my()">
95and L<"Temporary Values via local()">. To create protected
96environments for a set of functions in a separate package (and
97probably a separate file), see L<perlmod/"Packages">.
d74e8afc 98X<formal parameter> X<parameter, formal>
a0d0e21e
LW
99
100Example:
101
cb1a09d0
AD
102 sub max {
103 my $max = shift(@_);
a0d0e21e
LW
104 foreach $foo (@_) {
105 $max = $foo if $max < $foo;
106 }
cb1a09d0 107 return $max;
a0d0e21e 108 }
cb1a09d0 109 $bestday = max($mon,$tue,$wed,$thu,$fri);
a0d0e21e
LW
110
111Example:
112
113 # get a line, combining continuation lines
114 # that start with whitespace
115
116 sub get_line {
19799a22 117 $thisline = $lookahead; # global variables!
54310121 118 LINE: while (defined($lookahead = <STDIN>)) {
a0d0e21e
LW
119 if ($lookahead =~ /^[ \t]/) {
120 $thisline .= $lookahead;
121 }
122 else {
123 last LINE;
124 }
125 }
19799a22 126 return $thisline;
a0d0e21e
LW
127 }
128
129 $lookahead = <STDIN>; # get first line
19799a22 130 while (defined($line = get_line())) {
a0d0e21e
LW
131 ...
132 }
133
09bef843 134Assigning to a list of private variables to name your arguments:
a0d0e21e
LW
135
136 sub maybeset {
137 my($key, $value) = @_;
cb1a09d0 138 $Foo{$key} = $value unless $Foo{$key};
a0d0e21e
LW
139 }
140
19799a22
GS
141Because the assignment copies the values, this also has the effect
142of turning call-by-reference into call-by-value. Otherwise a
143function is free to do in-place modifications of C<@_> and change
144its caller's values.
d74e8afc 145X<call-by-reference> X<call-by-value>
cb1a09d0
AD
146
147 upcase_in($v1, $v2); # this changes $v1 and $v2
148 sub upcase_in {
54310121
PP
149 for (@_) { tr/a-z/A-Z/ }
150 }
cb1a09d0
AD
151
152You aren't allowed to modify constants in this way, of course. If an
153argument were actually literal and you tried to change it, you'd take a
154(presumably fatal) exception. For example, this won't work:
d74e8afc 155X<call-by-reference> X<call-by-value>
cb1a09d0
AD
156
157 upcase_in("frederick");
158
f86cebdf 159It would be much safer if the C<upcase_in()> function
cb1a09d0
AD
160were written to return a copy of its parameters instead
161of changing them in place:
162
19799a22 163 ($v3, $v4) = upcase($v1, $v2); # this doesn't change $v1 and $v2
cb1a09d0 164 sub upcase {
54310121 165 return unless defined wantarray; # void context, do nothing
cb1a09d0 166 my @parms = @_;
54310121 167 for (@parms) { tr/a-z/A-Z/ }
c07a80fd 168 return wantarray ? @parms : $parms[0];
54310121 169 }
cb1a09d0 170
19799a22 171Notice how this (unprototyped) function doesn't care whether it was
a2293a43 172passed real scalars or arrays. Perl sees all arguments as one big,
19799a22
GS
173long, flat parameter list in C<@_>. This is one area where
174Perl's simple argument-passing style shines. The C<upcase()>
175function would work perfectly well without changing the C<upcase()>
176definition even if we fed it things like this:
cb1a09d0
AD
177
178 @newlist = upcase(@list1, @list2);
179 @newlist = upcase( split /:/, $var );
180
181Do not, however, be tempted to do this:
182
183 (@a, @b) = upcase(@list1, @list2);
184
19799a22
GS
185Like the flattened incoming parameter list, the return list is also
186flattened on return. So all you have managed to do here is stored
17b63f68 187everything in C<@a> and made C<@b> empty. See
13a2d996 188L<Pass by Reference> for alternatives.
19799a22
GS
189
190A subroutine may be called using an explicit C<&> prefix. The
191C<&> is optional in modern Perl, as are parentheses if the
192subroutine has been predeclared. The C<&> is I<not> optional
193when just naming the subroutine, such as when it's used as
194an argument to defined() or undef(). Nor is it optional when you
195want to do an indirect subroutine call with a subroutine name or
196reference using the C<&$subref()> or C<&{$subref}()> constructs,
c47ff5f1 197although the C<< $subref->() >> notation solves that problem.
19799a22 198See L<perlref> for more about all that.
d74e8afc 199X<&>
19799a22
GS
200
201Subroutines may be called recursively. If a subroutine is called
202using the C<&> form, the argument list is optional, and if omitted,
203no C<@_> array is set up for the subroutine: the C<@_> array at the
204time of the call is visible to subroutine instead. This is an
205efficiency mechanism that new users may wish to avoid.
d74e8afc 206X<recursion>
a0d0e21e
LW
207
208 &foo(1,2,3); # pass three arguments
209 foo(1,2,3); # the same
210
211 foo(); # pass a null list
212 &foo(); # the same
a0d0e21e 213
cb1a09d0 214 &foo; # foo() get current args, like foo(@_) !!
54310121 215 foo; # like foo() IFF sub foo predeclared, else "foo"
cb1a09d0 216
19799a22
GS
217Not only does the C<&> form make the argument list optional, it also
218disables any prototype checking on arguments you do provide. This
c07a80fd 219is partly for historical reasons, and partly for having a convenient way
19799a22 220to cheat if you know what you're doing. See L<Prototypes> below.
d74e8afc 221X<&>
c07a80fd 222
ac90fb77
EM
223Subroutines whose names are in all upper case are reserved to the Perl
224core, as are modules whose names are in all lower case. A subroutine in
225all capitals is a loosely-held convention meaning it will be called
226indirectly by the run-time system itself, usually due to a triggered event.
227Subroutines that do special, pre-defined things include C<AUTOLOAD>, C<CLONE>,
228C<DESTROY> plus all functions mentioned in L<perltie> and L<PerlIO::via>.
229
3c10abe3
AG
230The C<BEGIN>, C<UNITCHECK>, C<CHECK>, C<INIT> and C<END> subroutines
231are not so much subroutines as named special code blocks, of which you
232can have more than one in a package, and which you can B<not> call
233explicitly. See L<perlmod/"BEGIN, UNITCHECK, CHECK, INIT and END">
5a964f20 234
b687b08b 235=head2 Private Variables via my()
d74e8afc
ITB
236X<my> X<variable, lexical> X<lexical> X<lexical variable> X<scope, lexical>
237X<lexical scope> X<attributes, my>
cb1a09d0
AD
238
239Synopsis:
240
241 my $foo; # declare $foo lexically local
242 my (@wid, %get); # declare list of variables local
243 my $foo = "flurp"; # declare $foo lexical, and init it
244 my @oof = @bar; # declare @oof lexical, and init it
09bef843
SB
245 my $x : Foo = $y; # similar, with an attribute applied
246
a0ae32d3
JH
247B<WARNING>: The use of attribute lists on C<my> declarations is still
248evolving. The current semantics and interface are subject to change.
249See L<attributes> and L<Attribute::Handlers>.
cb1a09d0 250
19799a22
GS
251The C<my> operator declares the listed variables to be lexically
252confined to the enclosing block, conditional (C<if/unless/elsif/else>),
253loop (C<for/foreach/while/until/continue>), subroutine, C<eval>,
254or C<do/require/use>'d file. If more than one value is listed, the
255list must be placed in parentheses. All listed elements must be
256legal lvalues. Only alphanumeric identifiers may be lexically
325192b1 257scoped--magical built-ins like C<$/> must currently be C<local>ized
19799a22
GS
258with C<local> instead.
259
260Unlike dynamic variables created by the C<local> operator, lexical
261variables declared with C<my> are totally hidden from the outside
262world, including any called subroutines. This is true if it's the
263same subroutine called from itself or elsewhere--every call gets
264its own copy.
d74e8afc 265X<local>
19799a22
GS
266
267This doesn't mean that a C<my> variable declared in a statically
268enclosing lexical scope would be invisible. Only dynamic scopes
269are cut off. For example, the C<bumpx()> function below has access
270to the lexical $x variable because both the C<my> and the C<sub>
271occurred at the same scope, presumably file scope.
5a964f20
TC
272
273 my $x = 10;
274 sub bumpx { $x++ }
275
19799a22
GS
276An C<eval()>, however, can see lexical variables of the scope it is
277being evaluated in, so long as the names aren't hidden by declarations within
278the C<eval()> itself. See L<perlref>.
d74e8afc 279X<eval, scope of>
cb1a09d0 280
19799a22 281The parameter list to my() may be assigned to if desired, which allows you
cb1a09d0
AD
282to initialize your variables. (If no initializer is given for a
283particular variable, it is created with the undefined value.) Commonly
19799a22 284this is used to name input parameters to a subroutine. Examples:
cb1a09d0
AD
285
286 $arg = "fred"; # "global" variable
287 $n = cube_root(27);
288 print "$arg thinks the root is $n\n";
289 fred thinks the root is 3
290
291 sub cube_root {
292 my $arg = shift; # name doesn't matter
293 $arg **= 1/3;
294 return $arg;
54310121 295 }
cb1a09d0 296
19799a22
GS
297The C<my> is simply a modifier on something you might assign to. So when
298you do assign to variables in its argument list, C<my> doesn't
6cc33c6d 299change whether those variables are viewed as a scalar or an array. So
cb1a09d0 300
5a964f20 301 my ($foo) = <STDIN>; # WRONG?
cb1a09d0
AD
302 my @FOO = <STDIN>;
303
5f05dabc 304both supply a list context to the right-hand side, while
cb1a09d0
AD
305
306 my $foo = <STDIN>;
307
5f05dabc 308supplies a scalar context. But the following declares only one variable:
748a9306 309
5a964f20 310 my $foo, $bar = 1; # WRONG
748a9306 311
cb1a09d0 312That has the same effect as
748a9306 313
cb1a09d0
AD
314 my $foo;
315 $bar = 1;
a0d0e21e 316
cb1a09d0
AD
317The declared variable is not introduced (is not visible) until after
318the current statement. Thus,
319
320 my $x = $x;
321
19799a22 322can be used to initialize a new $x with the value of the old $x, and
cb1a09d0
AD
323the expression
324
325 my $x = 123 and $x == 123
326
19799a22 327is false unless the old $x happened to have the value C<123>.
cb1a09d0 328
55497cff
PP
329Lexical scopes of control structures are not bounded precisely by the
330braces that delimit their controlled blocks; control expressions are
19799a22 331part of that scope, too. Thus in the loop
55497cff 332
19799a22 333 while (my $line = <>) {
55497cff
PP
334 $line = lc $line;
335 } continue {
336 print $line;
337 }
338
19799a22 339the scope of $line extends from its declaration throughout the rest of
55497cff
PP
340the loop construct (including the C<continue> clause), but not beyond
341it. Similarly, in the conditional
342
343 if ((my $answer = <STDIN>) =~ /^yes$/i) {
344 user_agrees();
345 } elsif ($answer =~ /^no$/i) {
346 user_disagrees();
347 } else {
348 chomp $answer;
349 die "'$answer' is neither 'yes' nor 'no'";
350 }
351
19799a22
GS
352the scope of $answer extends from its declaration through the rest
353of that conditional, including any C<elsif> and C<else> clauses,
457b36cb
MV
354but not beyond it. See L<perlsyn/"Simple statements"> for information
355on the scope of variables in statements with modifiers.
55497cff 356
5f05dabc 357The C<foreach> loop defaults to scoping its index variable dynamically
19799a22
GS
358in the manner of C<local>. However, if the index variable is
359prefixed with the keyword C<my>, or if there is already a lexical
360by that name in scope, then a new lexical is created instead. Thus
361in the loop
d74e8afc 362X<foreach> X<for>
55497cff
PP
363
364 for my $i (1, 2, 3) {
365 some_function();
366 }
367
19799a22
GS
368the scope of $i extends to the end of the loop, but not beyond it,
369rendering the value of $i inaccessible within C<some_function()>.
d74e8afc 370X<foreach> X<for>
55497cff 371
cb1a09d0 372Some users may wish to encourage the use of lexically scoped variables.
19799a22
GS
373As an aid to catching implicit uses to package variables,
374which are always global, if you say
cb1a09d0
AD
375
376 use strict 'vars';
377
19799a22
GS
378then any variable mentioned from there to the end of the enclosing
379block must either refer to a lexical variable, be predeclared via
77ca0c92 380C<our> or C<use vars>, or else must be fully qualified with the package name.
19799a22
GS
381A compilation error results otherwise. An inner block may countermand
382this with C<no strict 'vars'>.
383
384A C<my> has both a compile-time and a run-time effect. At compile
8593bda5 385time, the compiler takes notice of it. The principal usefulness
19799a22
GS
386of this is to quiet C<use strict 'vars'>, but it is also essential
387for generation of closures as detailed in L<perlref>. Actual
388initialization is delayed until run time, though, so it gets executed
389at the appropriate time, such as each time through a loop, for
390example.
391
392Variables declared with C<my> are not part of any package and are therefore
cb1a09d0
AD
393never fully qualified with the package name. In particular, you're not
394allowed to try to make a package variable (or other global) lexical:
395
396 my $pack::var; # ERROR! Illegal syntax
cb1a09d0
AD
397
398In fact, a dynamic variable (also known as package or global variables)
f86cebdf 399are still accessible using the fully qualified C<::> notation even while a
cb1a09d0
AD
400lexical of the same name is also visible:
401
402 package main;
403 local $x = 10;
404 my $x = 20;
405 print "$x and $::x\n";
406
f86cebdf 407That will print out C<20> and C<10>.
cb1a09d0 408
19799a22
GS
409You may declare C<my> variables at the outermost scope of a file
410to hide any such identifiers from the world outside that file. This
411is similar in spirit to C's static variables when they are used at
412the file level. To do this with a subroutine requires the use of
413a closure (an anonymous function that accesses enclosing lexicals).
414If you want to create a private subroutine that cannot be called
415from outside that block, it can declare a lexical variable containing
416an anonymous sub reference:
cb1a09d0
AD
417
418 my $secret_version = '1.001-beta';
419 my $secret_sub = sub { print $secret_version };
420 &$secret_sub();
421
422As long as the reference is never returned by any function within the
5f05dabc 423module, no outside module can see the subroutine, because its name is not in
cb1a09d0 424any package's symbol table. Remember that it's not I<REALLY> called
19799a22 425C<$some_pack::secret_version> or anything; it's just $secret_version,
cb1a09d0
AD
426unqualified and unqualifiable.
427
19799a22
GS
428This does not work with object methods, however; all object methods
429have to be in the symbol table of some package to be found. See
430L<perlref/"Function Templates"> for something of a work-around to
431this.
cb1a09d0 432
c2611fb3 433=head2 Persistent Private Variables
ba1f8e91
RGS
434X<state> X<state variable> X<static> X<variable, persistent> X<variable, static> X<closure>
435
436There are two ways to build persistent private variables in Perl 5.10.
437First, you can simply use the C<state> feature. Or, you can use closures,
438if you want to stay compatible with releases older than 5.10.
439
440=head3 Persistent variables via state()
441
442Beginning with perl 5.9.4, you can declare variables with the C<state>
443keyword in place of C<my>. For that to work, though, you must have
444enabled that feature beforehand, either by using the C<feature> pragma, or
445by using C<-E> on one-liners. (see L<feature>)
446
447For example, the following code maintains a private counter, incremented
448each time the gimme_another() function is called:
449
450 use feature 'state';
451 sub gimme_another { state $x; return ++$x }
452
453Also, since C<$x> is lexical, it can't be reached or modified by any Perl
454code outside.
455
f292fc7a
RS
456When combined with variable declaration, simple scalar assignment to C<state>
457variables (as in C<state $x = 42>) is executed only the first time. When such
458statements are evaluated subsequent times, the assignment is ignored. The
459behavior of this sort of assignment to non-scalar variables is undefined.
ba1f8e91
RGS
460
461=head3 Persistent variables with closures
5a964f20
TC
462
463Just because a lexical variable is lexically (also called statically)
f86cebdf 464scoped to its enclosing block, C<eval>, or C<do> FILE, this doesn't mean that
5a964f20
TC
465within a function it works like a C static. It normally works more
466like a C auto, but with implicit garbage collection.
467
468Unlike local variables in C or C++, Perl's lexical variables don't
469necessarily get recycled just because their scope has exited.
470If something more permanent is still aware of the lexical, it will
471stick around. So long as something else references a lexical, that
472lexical won't be freed--which is as it should be. You wouldn't want
473memory being free until you were done using it, or kept around once you
474were done. Automatic garbage collection takes care of this for you.
475
476This means that you can pass back or save away references to lexical
477variables, whereas to return a pointer to a C auto is a grave error.
478It also gives us a way to simulate C's function statics. Here's a
479mechanism for giving a function private variables with both lexical
480scoping and a static lifetime. If you do want to create something like
481C's static variables, just enclose the whole function in an extra block,
482and put the static variable outside the function but in the block.
cb1a09d0
AD
483
484 {
54310121 485 my $secret_val = 0;
cb1a09d0
AD
486 sub gimme_another {
487 return ++$secret_val;
54310121
PP
488 }
489 }
cb1a09d0
AD
490 # $secret_val now becomes unreachable by the outside
491 # world, but retains its value between calls to gimme_another
492
54310121 493If this function is being sourced in from a separate file
cb1a09d0 494via C<require> or C<use>, then this is probably just fine. If it's
19799a22 495all in the main program, you'll need to arrange for the C<my>
cb1a09d0 496to be executed early, either by putting the whole block above
f86cebdf 497your main program, or more likely, placing merely a C<BEGIN>
ac90fb77 498code block around it to make sure it gets executed before your program
cb1a09d0
AD
499starts to run:
500
ac90fb77 501 BEGIN {
54310121 502 my $secret_val = 0;
cb1a09d0
AD
503 sub gimme_another {
504 return ++$secret_val;
54310121
PP
505 }
506 }
cb1a09d0 507
3c10abe3
AG
508See L<perlmod/"BEGIN, UNITCHECK, CHECK, INIT and END"> about the
509special triggered code blocks, C<BEGIN>, C<UNITCHECK>, C<CHECK>,
510C<INIT> and C<END>.
cb1a09d0 511
19799a22
GS
512If declared at the outermost scope (the file scope), then lexicals
513work somewhat like C's file statics. They are available to all
514functions in that same file declared below them, but are inaccessible
515from outside that file. This strategy is sometimes used in modules
516to create private variables that the whole module can see.
5a964f20 517
cb1a09d0 518=head2 Temporary Values via local()
d74e8afc
ITB
519X<local> X<scope, dynamic> X<dynamic scope> X<variable, local>
520X<variable, temporary>
cb1a09d0 521
19799a22 522B<WARNING>: In general, you should be using C<my> instead of C<local>, because
6d28dffb 523it's faster and safer. Exceptions to this include the global punctuation
325192b1
RGS
524variables, global filehandles and formats, and direct manipulation of the
525Perl symbol table itself. C<local> is mostly used when the current value
526of a variable must be visible to called subroutines.
cb1a09d0
AD
527
528Synopsis:
529
325192b1
RGS
530 # localization of values
531
532 local $foo; # make $foo dynamically local
533 local (@wid, %get); # make list of variables local
534 local $foo = "flurp"; # make $foo dynamic, and init it
535 local @oof = @bar; # make @oof dynamic, and init it
536
537 local $hash{key} = "val"; # sets a local value for this hash entry
d361fafa 538 delete local $hash{key}; # delete this entry for the current block
325192b1
RGS
539 local ($cond ? $v1 : $v2); # several types of lvalues support
540 # localization
541
542 # localization of symbols
cb1a09d0
AD
543
544 local *FH; # localize $FH, @FH, %FH, &FH ...
545 local *merlyn = *randal; # now $merlyn is really $randal, plus
546 # @merlyn is really @randal, etc
547 local *merlyn = 'randal'; # SAME THING: promote 'randal' to *randal
54310121 548 local *merlyn = \$randal; # just alias $merlyn, not @merlyn etc
cb1a09d0 549
19799a22
GS
550A C<local> modifies its listed variables to be "local" to the
551enclosing block, C<eval>, or C<do FILE>--and to I<any subroutine
552called from within that block>. A C<local> just gives temporary
553values to global (meaning package) variables. It does I<not> create
554a local variable. This is known as dynamic scoping. Lexical scoping
555is done with C<my>, which works more like C's auto declarations.
cb1a09d0 556
325192b1
RGS
557Some types of lvalues can be localized as well : hash and array elements
558and slices, conditionals (provided that their result is always
559localizable), and symbolic references. As for simple variables, this
560creates new, dynamically scoped values.
561
562If more than one variable or expression is given to C<local>, they must be
563placed in parentheses. This operator works
cb1a09d0 564by saving the current values of those variables in its argument list on a
5f05dabc 565hidden stack and restoring them upon exiting the block, subroutine, or
cb1a09d0
AD
566eval. This means that called subroutines can also reference the local
567variable, but not the global one. The argument list may be assigned to if
568desired, which allows you to initialize your local variables. (If no
569initializer is given for a particular variable, it is created with an
325192b1 570undefined value.)
cb1a09d0 571
19799a22 572Because C<local> is a run-time operator, it gets executed each time
325192b1
RGS
573through a loop. Consequently, it's more efficient to localize your
574variables outside the loop.
575
576=head3 Grammatical note on local()
d74e8afc 577X<local, context>
cb1a09d0 578
f86cebdf
GS
579A C<local> is simply a modifier on an lvalue expression. When you assign to
580a C<local>ized variable, the C<local> doesn't change whether its list is viewed
cb1a09d0
AD
581as a scalar or an array. So
582
583 local($foo) = <STDIN>;
584 local @FOO = <STDIN>;
585
5f05dabc 586both supply a list context to the right-hand side, while
cb1a09d0
AD
587
588 local $foo = <STDIN>;
589
590supplies a scalar context.
591
325192b1 592=head3 Localization of special variables
d74e8afc 593X<local, special variable>
3e3baf6d 594
325192b1
RGS
595If you localize a special variable, you'll be giving a new value to it,
596but its magic won't go away. That means that all side-effects related
597to this magic still work with the localized value.
3e3baf6d 598
325192b1
RGS
599This feature allows code like this to work :
600
601 # Read the whole contents of FILE in $slurp
602 { local $/ = undef; $slurp = <FILE>; }
603
604Note, however, that this restricts localization of some values ; for
605example, the following statement dies, as of perl 5.9.0, with an error
606I<Modification of a read-only value attempted>, because the $1 variable is
607magical and read-only :
608
609 local $1 = 2;
610
611Similarly, but in a way more difficult to spot, the following snippet will
612die in perl 5.9.0 :
613
614 sub f { local $_ = "foo"; print }
615 for ($1) {
616 # now $_ is aliased to $1, thus is magic and readonly
617 f();
3e3baf6d 618 }
3e3baf6d 619
325192b1
RGS
620See next section for an alternative to this situation.
621
622B<WARNING>: Localization of tied arrays and hashes does not currently
623work as described.
fd5a896a
DM
624This will be fixed in a future release of Perl; in the meantime, avoid
625code that relies on any particular behaviour of localising tied arrays
626or hashes (localising individual elements is still okay).
325192b1 627See L<perl58delta/"Localising Tied Arrays and Hashes Is Broken"> for more
fd5a896a 628details.
d74e8afc 629X<local, tie>
fd5a896a 630
325192b1 631=head3 Localization of globs
d74e8afc 632X<local, glob> X<glob>
3e3baf6d 633
325192b1
RGS
634The construct
635
636 local *name;
637
638creates a whole new symbol table entry for the glob C<name> in the
639current package. That means that all variables in its glob slot ($name,
640@name, %name, &name, and the C<name> filehandle) are dynamically reset.
641
642This implies, among other things, that any magic eventually carried by
643those variables is locally lost. In other words, saying C<local */>
644will not have any effect on the internal value of the input record
645separator.
646
647Notably, if you want to work with a brand new value of the default scalar
648$_, and avoid the potential problem listed above about $_ previously
649carrying a magic value, you should use C<local *_> instead of C<local $_>.
a4fb8298
RGS
650As of perl 5.9.1, you can also use the lexical form of C<$_> (declaring it
651with C<my $_>), which avoids completely this problem.
325192b1
RGS
652
653=head3 Localization of elements of composite types
d74e8afc 654X<local, composite type element> X<local, array element> X<local, hash element>
3e3baf6d 655
6ee623d5 656It's also worth taking a moment to explain what happens when you
f86cebdf
GS
657C<local>ize a member of a composite type (i.e. an array or hash element).
658In this case, the element is C<local>ized I<by name>. This means that
6ee623d5
GS
659when the scope of the C<local()> ends, the saved value will be
660restored to the hash element whose key was named in the C<local()>, or
661the array element whose index was named in the C<local()>. If that
662element was deleted while the C<local()> was in effect (e.g. by a
663C<delete()> from a hash or a C<shift()> of an array), it will spring
664back into existence, possibly extending an array and filling in the
665skipped elements with C<undef>. For instance, if you say
666
667 %hash = ( 'This' => 'is', 'a' => 'test' );
668 @ary = ( 0..5 );
669 {
670 local($ary[5]) = 6;
671 local($hash{'a'}) = 'drill';
672 while (my $e = pop(@ary)) {
673 print "$e . . .\n";
674 last unless $e > 3;
675 }
676 if (@ary) {
677 $hash{'only a'} = 'test';
678 delete $hash{'a'};
679 }
680 }
681 print join(' ', map { "$_ $hash{$_}" } sort keys %hash),".\n";
682 print "The array has ",scalar(@ary)," elements: ",
683 join(', ', map { defined $_ ? $_ : 'undef' } @ary),"\n";
684
685Perl will print
686
687 6 . . .
688 4 . . .
689 3 . . .
690 This is a test only a test.
691 The array has 6 elements: 0, 1, 2, undef, undef, 5
692
19799a22 693The behavior of local() on non-existent members of composite
7185e5cc
GS
694types is subject to change in future.
695
d361fafa
VP
696=head3 Localized deletion of elements of composite types
697X<delete> X<local, composite type element> X<local, array element> X<local, hash element>
698
699You can use the C<delete local $array[$idx]> and C<delete local $hash{key}>
700constructs to delete a composite type entry for the current block and restore
701it when it ends. They return the array/hash value before the localization,
702which means that they are respectively equivalent to
703
704 do {
705 my $val = $array[$idx];
706 local $array[$idx];
707 delete $array[$idx];
708 $val
709 }
710
711and
712
713 do {
714 my $val = $hash{key};
715 local $hash{key};
716 delete $hash{key};
717 $val
718 }
719
720except that for those the C<local> is scoped to the C<do> block. Slices are
721also accepted.
722
723 my %hash = (
724 a => [ 7, 8, 9 ],
725 b => 1,
726 )
727
728 {
729 my $a = delete local $hash{a};
730 # $a is [ 7, 8, 9 ]
731 # %hash is (b => 1)
732
733 {
734 my @nums = delete local @$a[0, 2]
735 # @nums is (7, 9)
736 # $a is [ undef, 8 ]
737
738 $a[0] = 999; # will be erased when the scope ends
739 }
740 # $a is back to [ 7, 8, 9 ]
741
742 }
743 # %hash is back to its original state
744
cd06dffe 745=head2 Lvalue subroutines
d74e8afc 746X<lvalue> X<subroutine, lvalue>
cd06dffe 747
e6a32221
JC
748B<WARNING>: Lvalue subroutines are still experimental and the
749implementation may change in future versions of Perl.
cd06dffe
GS
750
751It is possible to return a modifiable value from a subroutine.
752To do this, you have to declare the subroutine to return an lvalue.
753
754 my $val;
755 sub canmod : lvalue {
e6a32221 756 # return $val; this doesn't work, don't say "return"
cd06dffe
GS
757 $val;
758 }
759 sub nomod {
760 $val;
761 }
762
763 canmod() = 5; # assigns to $val
764 nomod() = 5; # ERROR
765
766The scalar/list context for the subroutine and for the right-hand
767side of assignment is determined as if the subroutine call is replaced
768by a scalar. For example, consider:
769
770 data(2,3) = get_data(3,4);
771
772Both subroutines here are called in a scalar context, while in:
773
774 (data(2,3)) = get_data(3,4);
775
776and in:
777
778 (data(2),data(3)) = get_data(3,4);
779
780all the subroutines are called in a list context.
781
e6a32221
JC
782=over 4
783
784=item Lvalue subroutines are EXPERIMENTAL
785
786They appear to be convenient, but there are several reasons to be
787circumspect.
788
789You can't use the return keyword, you must pass out the value before
790falling out of subroutine scope. (see comment in example above). This
791is usually not a problem, but it disallows an explicit return out of a
792deeply nested loop, which is sometimes a nice way out.
793
794They violate encapsulation. A normal mutator can check the supplied
795argument before setting the attribute it is protecting, an lvalue
796subroutine never gets that chance. Consider;
797
798 my $some_array_ref = []; # protected by mutators ??
799
800 sub set_arr { # normal mutator
801 my $val = shift;
802 die("expected array, you supplied ", ref $val)
803 unless ref $val eq 'ARRAY';
804 $some_array_ref = $val;
805 }
806 sub set_arr_lv : lvalue { # lvalue mutator
807 $some_array_ref;
808 }
809
810 # set_arr_lv cannot stop this !
811 set_arr_lv() = { a => 1 };
818c4caa 812
e6a32221
JC
813=back
814
cb1a09d0 815=head2 Passing Symbol Table Entries (typeglobs)
d74e8afc 816X<typeglob> X<*>
cb1a09d0 817
19799a22
GS
818B<WARNING>: The mechanism described in this section was originally
819the only way to simulate pass-by-reference in older versions of
820Perl. While it still works fine in modern versions, the new reference
821mechanism is generally easier to work with. See below.
a0d0e21e
LW
822
823Sometimes you don't want to pass the value of an array to a subroutine
824but rather the name of it, so that the subroutine can modify the global
825copy of it rather than working with a local copy. In perl you can
cb1a09d0 826refer to all objects of a particular name by prefixing the name
5f05dabc 827with a star: C<*foo>. This is often known as a "typeglob", because the
a0d0e21e
LW
828star on the front can be thought of as a wildcard match for all the
829funny prefix characters on variables and subroutines and such.
830
55497cff 831When evaluated, the typeglob produces a scalar value that represents
5f05dabc 832all the objects of that name, including any filehandle, format, or
a0d0e21e 833subroutine. When assigned to, it causes the name mentioned to refer to
19799a22 834whatever C<*> value was assigned to it. Example:
a0d0e21e
LW
835
836 sub doubleary {
837 local(*someary) = @_;
838 foreach $elem (@someary) {
839 $elem *= 2;
840 }
841 }
842 doubleary(*foo);
843 doubleary(*bar);
844
19799a22 845Scalars are already passed by reference, so you can modify
a0d0e21e 846scalar arguments without using this mechanism by referring explicitly
1fef88e7 847to C<$_[0]> etc. You can modify all the elements of an array by passing
f86cebdf
GS
848all the elements as scalars, but you have to use the C<*> mechanism (or
849the equivalent reference mechanism) to C<push>, C<pop>, or change the size of
a0d0e21e
LW
850an array. It will certainly be faster to pass the typeglob (or reference).
851
852Even if you don't want to modify an array, this mechanism is useful for
5f05dabc 853passing multiple arrays in a single LIST, because normally the LIST
a0d0e21e 854mechanism will merge all the array values so that you can't extract out
55497cff 855the individual arrays. For more on typeglobs, see
2ae324a7 856L<perldata/"Typeglobs and Filehandles">.
cb1a09d0 857
5a964f20 858=head2 When to Still Use local()
d74e8afc 859X<local> X<variable, local>
5a964f20 860
19799a22
GS
861Despite the existence of C<my>, there are still three places where the
862C<local> operator still shines. In fact, in these three places, you
5a964f20
TC
863I<must> use C<local> instead of C<my>.
864
13a2d996 865=over 4
5a964f20 866
551e1d92
RB
867=item 1.
868
869You need to give a global variable a temporary value, especially $_.
5a964f20 870
f86cebdf
GS
871The global variables, like C<@ARGV> or the punctuation variables, must be
872C<local>ized with C<local()>. This block reads in F</etc/motd>, and splits
5a964f20 873it up into chunks separated by lines of equal signs, which are placed
f86cebdf 874in C<@Fields>.
5a964f20
TC
875
876 {
877 local @ARGV = ("/etc/motd");
878 local $/ = undef;
879 local $_ = <>;
880 @Fields = split /^\s*=+\s*$/;
881 }
882
19799a22 883It particular, it's important to C<local>ize $_ in any routine that assigns
5a964f20
TC
884to it. Look out for implicit assignments in C<while> conditionals.
885
551e1d92
RB
886=item 2.
887
888You need to create a local file or directory handle or a local function.
5a964f20 889
09bef843
SB
890A function that needs a filehandle of its own must use
891C<local()> on a complete typeglob. This can be used to create new symbol
5a964f20
TC
892table entries:
893
894 sub ioqueue {
895 local (*READER, *WRITER); # not my!
17b63f68 896 pipe (READER, WRITER) or die "pipe: $!";
5a964f20
TC
897 return (*READER, *WRITER);
898 }
899 ($head, $tail) = ioqueue();
900
901See the Symbol module for a way to create anonymous symbol table
902entries.
903
904Because assignment of a reference to a typeglob creates an alias, this
905can be used to create what is effectively a local function, or at least,
906a local alias.
907
908 {
f86cebdf
GS
909 local *grow = \&shrink; # only until this block exists
910 grow(); # really calls shrink()
911 move(); # if move() grow()s, it shrink()s too
5a964f20 912 }
f86cebdf 913 grow(); # get the real grow() again
5a964f20
TC
914
915See L<perlref/"Function Templates"> for more about manipulating
916functions by name in this way.
917
551e1d92
RB
918=item 3.
919
920You want to temporarily change just one element of an array or hash.
5a964f20 921
f86cebdf 922You can C<local>ize just one element of an aggregate. Usually this
5a964f20
TC
923is done on dynamics:
924
925 {
926 local $SIG{INT} = 'IGNORE';
927 funct(); # uninterruptible
928 }
929 # interruptibility automatically restored here
930
931But it also works on lexically declared aggregates. Prior to 5.005,
932this operation could on occasion misbehave.
933
934=back
935
cb1a09d0 936=head2 Pass by Reference
d74e8afc 937X<pass by reference> X<pass-by-reference> X<reference>
cb1a09d0 938
55497cff
PP
939If you want to pass more than one array or hash into a function--or
940return them from it--and have them maintain their integrity, then
941you're going to have to use an explicit pass-by-reference. Before you
942do that, you need to understand references as detailed in L<perlref>.
c07a80fd 943This section may not make much sense to you otherwise.
cb1a09d0 944
19799a22
GS
945Here are a few simple examples. First, let's pass in several arrays
946to a function and have it C<pop> all of then, returning a new list
947of all their former last elements:
cb1a09d0
AD
948
949 @tailings = popmany ( \@a, \@b, \@c, \@d );
950
951 sub popmany {
952 my $aref;
953 my @retlist = ();
954 foreach $aref ( @_ ) {
955 push @retlist, pop @$aref;
54310121 956 }
cb1a09d0 957 return @retlist;
54310121 958 }
cb1a09d0 959
54310121 960Here's how you might write a function that returns a
cb1a09d0
AD
961list of keys occurring in all the hashes passed to it:
962
54310121 963 @common = inter( \%foo, \%bar, \%joe );
cb1a09d0
AD
964 sub inter {
965 my ($k, $href, %seen); # locals
966 foreach $href (@_) {
967 while ( $k = each %$href ) {
968 $seen{$k}++;
54310121
PP
969 }
970 }
cb1a09d0 971 return grep { $seen{$_} == @_ } keys %seen;
54310121 972 }
cb1a09d0 973
5f05dabc 974So far, we're using just the normal list return mechanism.
54310121
PP
975What happens if you want to pass or return a hash? Well,
976if you're using only one of them, or you don't mind them
cb1a09d0 977concatenating, then the normal calling convention is ok, although
54310121 978a little expensive.
cb1a09d0
AD
979
980Where people get into trouble is here:
981
982 (@a, @b) = func(@c, @d);
983or
984 (%a, %b) = func(%c, %d);
985
19799a22
GS
986That syntax simply won't work. It sets just C<@a> or C<%a> and
987clears the C<@b> or C<%b>. Plus the function didn't get passed
988into two separate arrays or hashes: it got one long list in C<@_>,
989as always.
cb1a09d0
AD
990
991If you can arrange for everyone to deal with this through references, it's
992cleaner code, although not so nice to look at. Here's a function that
993takes two array references as arguments, returning the two array elements
994in order of how many elements they have in them:
995
996 ($aref, $bref) = func(\@c, \@d);
997 print "@$aref has more than @$bref\n";
998 sub func {
999 my ($cref, $dref) = @_;
1000 if (@$cref > @$dref) {
1001 return ($cref, $dref);
1002 } else {
c07a80fd 1003 return ($dref, $cref);
54310121
PP
1004 }
1005 }
cb1a09d0
AD
1006
1007It turns out that you can actually do this also:
1008
1009 (*a, *b) = func(\@c, \@d);
1010 print "@a has more than @b\n";
1011 sub func {
1012 local (*c, *d) = @_;
1013 if (@c > @d) {
1014 return (\@c, \@d);
1015 } else {
1016 return (\@d, \@c);
54310121
PP
1017 }
1018 }
cb1a09d0
AD
1019
1020Here we're using the typeglobs to do symbol table aliasing. It's
19799a22 1021a tad subtle, though, and also won't work if you're using C<my>
09bef843 1022variables, because only globals (even in disguise as C<local>s)
19799a22 1023are in the symbol table.
5f05dabc
PP
1024
1025If you're passing around filehandles, you could usually just use the bare
19799a22
GS
1026typeglob, like C<*STDOUT>, but typeglobs references work, too.
1027For example:
5f05dabc
PP
1028
1029 splutter(\*STDOUT);
1030 sub splutter {
1031 my $fh = shift;
1032 print $fh "her um well a hmmm\n";
1033 }
1034
1035 $rec = get_rec(\*STDIN);
1036 sub get_rec {
1037 my $fh = shift;
1038 return scalar <$fh>;
1039 }
1040
19799a22
GS
1041If you're planning on generating new filehandles, you could do this.
1042Notice to pass back just the bare *FH, not its reference.
5f05dabc
PP
1043
1044 sub openit {
19799a22 1045 my $path = shift;
5f05dabc 1046 local *FH;
e05a3a1e 1047 return open (FH, $path) ? *FH : undef;
54310121 1048 }
5f05dabc 1049
cb1a09d0 1050=head2 Prototypes
d74e8afc 1051X<prototype> X<subroutine, prototype>
cb1a09d0 1052
19799a22
GS
1053Perl supports a very limited kind of compile-time argument checking
1054using function prototyping. If you declare
cb1a09d0
AD
1055
1056 sub mypush (\@@)
1057
19799a22
GS
1058then C<mypush()> takes arguments exactly like C<push()> does. The
1059function declaration must be visible at compile time. The prototype
1060affects only interpretation of new-style calls to the function,
1061where new-style is defined as not using the C<&> character. In
1062other words, if you call it like a built-in function, then it behaves
1063like a built-in function. If you call it like an old-fashioned
1064subroutine, then it behaves like an old-fashioned subroutine. It
1065naturally falls out from this rule that prototypes have no influence
1066on subroutine references like C<\&foo> or on indirect subroutine
c47ff5f1 1067calls like C<&{$subref}> or C<< $subref->() >>.
c07a80fd
PP
1068
1069Method calls are not influenced by prototypes either, because the
19799a22
GS
1070function to be called is indeterminate at compile time, since
1071the exact code called depends on inheritance.
cb1a09d0 1072
19799a22
GS
1073Because the intent of this feature is primarily to let you define
1074subroutines that work like built-in functions, here are prototypes
1075for some other functions that parse almost exactly like the
1076corresponding built-in.
cb1a09d0
AD
1077
1078 Declared as Called as
1079
f86cebdf
GS
1080 sub mylink ($$) mylink $old, $new
1081 sub myvec ($$$) myvec $var, $offset, 1
1082 sub myindex ($$;$) myindex &getstring, "substr"
1083 sub mysyswrite ($$$;$) mysyswrite $buf, 0, length($buf) - $off, $off
1084 sub myreverse (@) myreverse $a, $b, $c
1085 sub myjoin ($@) myjoin ":", $a, $b, $c
1086 sub mypop (\@) mypop @array
803e1be1 1087 sub mysplice (\@$$@) mysplice @array, 0, 2, @pushme
f86cebdf
GS
1088 sub mykeys (\%) mykeys %{$hashref}
1089 sub myopen (*;$) myopen HANDLE, $name
1090 sub mypipe (**) mypipe READHANDLE, WRITEHANDLE
1091 sub mygrep (&@) mygrep { /foo/ } $a, $b, $c
d822fdf9 1092 sub myrand (;$) myrand 42
f86cebdf 1093 sub mytime () mytime
cb1a09d0 1094
c07a80fd 1095Any backslashed prototype character represents an actual argument
ae7a3cfa
FC
1096that must start with that character (optionally preceded by C<my>,
1097C<our> or C<local>), with the exception of C<$>, which will accept a
1098hash or array element even without a dollar sign, such as
1099C<my_function()->[0]>. The value passed as part of C<@_> will be a
1100reference to the actual argument given in the subroutine call,
1101obtained by applying C<\> to that argument.
c07a80fd 1102
5b794e05
JH
1103You can also backslash several argument types simultaneously by using
1104the C<\[]> notation:
1105
1106 sub myref (\[$@%&*])
1107
1108will allow calling myref() as
1109
1110 myref $var
1111 myref @array
1112 myref %hash
1113 myref &sub
1114 myref *glob
1115
1116and the first argument of myref() will be a reference to
1117a scalar, an array, a hash, a code, or a glob.
1118
c07a80fd 1119Unbackslashed prototype characters have special meanings. Any
19799a22 1120unbackslashed C<@> or C<%> eats all remaining arguments, and forces
f86cebdf
GS
1121list context. An argument represented by C<$> forces scalar context. An
1122C<&> requires an anonymous subroutine, which, if passed as the first
0df79f0c
GS
1123argument, does not require the C<sub> keyword or a subsequent comma.
1124
1125A C<*> allows the subroutine to accept a bareword, constant, scalar expression,
648ca4f7
GS
1126typeglob, or a reference to a typeglob in that slot. The value will be
1127available to the subroutine either as a simple scalar, or (in the latter
0df79f0c
GS
1128two cases) as a reference to the typeglob. If you wish to always convert
1129such arguments to a typeglob reference, use Symbol::qualify_to_ref() as
1130follows:
1131
1132 use Symbol 'qualify_to_ref';
1133
1134 sub foo (*) {
1135 my $fh = qualify_to_ref(shift, caller);
1136 ...
1137 }
c07a80fd 1138
859a4967 1139A semicolon (C<;>) separates mandatory arguments from optional arguments.
19799a22 1140It is redundant before C<@> or C<%>, which gobble up everything else.
cb1a09d0 1141
7adf2bcd
RGS
1142As the last character of a prototype, or just before a semicolon, you can
1143use C<_> in place of C<$>: if this argument is not provided, C<$_> will be
1144used instead.
859a4967 1145
19799a22
GS
1146Note how the last three examples in the table above are treated
1147specially by the parser. C<mygrep()> is parsed as a true list
1148operator, C<myrand()> is parsed as a true unary operator with unary
1149precedence the same as C<rand()>, and C<mytime()> is truly without
1150arguments, just like C<time()>. That is, if you say
cb1a09d0
AD
1151
1152 mytime +2;
1153
f86cebdf 1154you'll get C<mytime() + 2>, not C<mytime(2)>, which is how it would be parsed
19799a22 1155without a prototype.
cb1a09d0 1156
19799a22
GS
1157The interesting thing about C<&> is that you can generate new syntax with it,
1158provided it's in the initial position:
d74e8afc 1159X<&>
cb1a09d0 1160
6d28dffb 1161 sub try (&@) {
cb1a09d0
AD
1162 my($try,$catch) = @_;
1163 eval { &$try };
1164 if ($@) {
1165 local $_ = $@;
1166 &$catch;
1167 }
1168 }
55497cff 1169 sub catch (&) { $_[0] }
cb1a09d0
AD
1170
1171 try {
1172 die "phooey";
1173 } catch {
1174 /phooey/ and print "unphooey\n";
1175 };
1176
f86cebdf 1177That prints C<"unphooey">. (Yes, there are still unresolved
19799a22 1178issues having to do with visibility of C<@_>. I'm ignoring that
f86cebdf 1179question for the moment. (But note that if we make C<@_> lexically
cb1a09d0 1180scoped, those anonymous subroutines can act like closures... (Gee,
5f05dabc 1181is this sounding a little Lispish? (Never mind.))))
cb1a09d0 1182
19799a22 1183And here's a reimplementation of the Perl C<grep> operator:
d74e8afc 1184X<grep>
cb1a09d0
AD
1185
1186 sub mygrep (&@) {
1187 my $code = shift;
1188 my @result;
1189 foreach $_ (@_) {
6e47f808 1190 push(@result, $_) if &$code;
cb1a09d0
AD
1191 }
1192 @result;
1193 }
a0d0e21e 1194
cb1a09d0
AD
1195Some folks would prefer full alphanumeric prototypes. Alphanumerics have
1196been intentionally left out of prototypes for the express purpose of
1197someday in the future adding named, formal parameters. The current
1198mechanism's main goal is to let module writers provide better diagnostics
1199for module users. Larry feels the notation quite understandable to Perl
1200programmers, and that it will not intrude greatly upon the meat of the
1201module, nor make it harder to read. The line noise is visually
1202encapsulated into a small pill that's easy to swallow.
1203
420cdfc1
ST
1204If you try to use an alphanumeric sequence in a prototype you will
1205generate an optional warning - "Illegal character in prototype...".
1206Unfortunately earlier versions of Perl allowed the prototype to be
1207used as long as its prefix was a valid prototype. The warning may be
1208upgraded to a fatal error in a future version of Perl once the
1209majority of offending code is fixed.
1210
cb1a09d0
AD
1211It's probably best to prototype new functions, not retrofit prototyping
1212into older ones. That's because you must be especially careful about
1213silent impositions of differing list versus scalar contexts. For example,
1214if you decide that a function should take just one parameter, like this:
1215
1216 sub func ($) {
1217 my $n = shift;
1218 print "you gave me $n\n";
54310121 1219 }
cb1a09d0
AD
1220
1221and someone has been calling it with an array or expression
1222returning a list:
1223
1224 func(@foo);
1225 func( split /:/ );
1226
19799a22 1227Then you've just supplied an automatic C<scalar> in front of their
f86cebdf 1228argument, which can be more than a bit surprising. The old C<@foo>
cb1a09d0 1229which used to hold one thing doesn't get passed in. Instead,
19799a22
GS
1230C<func()> now gets passed in a C<1>; that is, the number of elements
1231in C<@foo>. And the C<split> gets called in scalar context so it
1232starts scribbling on your C<@_> parameter list. Ouch!
cb1a09d0 1233
5f05dabc 1234This is all very powerful, of course, and should be used only in moderation
54310121 1235to make the world a better place.
44a8e56a
PP
1236
1237=head2 Constant Functions
d74e8afc 1238X<constant>
44a8e56a
PP
1239
1240Functions with a prototype of C<()> are potential candidates for
19799a22
GS
1241inlining. If the result after optimization and constant folding
1242is either a constant or a lexically-scoped scalar which has no other
54310121 1243references, then it will be used in place of function calls made
19799a22
GS
1244without C<&>. Calls made using C<&> are never inlined. (See
1245F<constant.pm> for an easy way to declare most constants.)
44a8e56a 1246
5a964f20 1247The following functions would all be inlined:
44a8e56a 1248
699e6cd4
TP
1249 sub pi () { 3.14159 } # Not exact, but close.
1250 sub PI () { 4 * atan2 1, 1 } # As good as it gets,
1251 # and it's inlined, too!
44a8e56a
PP
1252 sub ST_DEV () { 0 }
1253 sub ST_INO () { 1 }
1254
1255 sub FLAG_FOO () { 1 << 8 }
1256 sub FLAG_BAR () { 1 << 9 }
1257 sub FLAG_MASK () { FLAG_FOO | FLAG_BAR }
54310121
PP
1258
1259 sub OPT_BAZ () { not (0x1B58 & FLAG_MASK) }
88267271 1260
1261 sub N () { int(OPT_BAZ) / 3 }
1262
1263 sub FOO_SET () { 1 if FLAG_MASK & FLAG_FOO }
1264
1265Be aware that these will not be inlined; as they contain inner scopes,
1266the constant folding doesn't reduce them to a single constant:
1267
1268 sub foo_set () { if (FLAG_MASK & FLAG_FOO) { 1 } }
1269
1270 sub baz_val () {
44a8e56a
PP
1271 if (OPT_BAZ) {
1272 return 23;
1273 }
1274 else {
1275 return 42;
1276 }
1277 }
cb1a09d0 1278
5a964f20 1279If you redefine a subroutine that was eligible for inlining, you'll get
4cee8e80
CS
1280a mandatory warning. (You can use this warning to tell whether or not a
1281particular subroutine is considered constant.) The warning is
1282considered severe enough not to be optional because previously compiled
1283invocations of the function will still be using the old value of the
19799a22 1284function. If you need to be able to redefine the subroutine, you need to
4cee8e80 1285ensure that it isn't inlined, either by dropping the C<()> prototype
19799a22 1286(which changes calling semantics, so beware) or by thwarting the
4cee8e80
CS
1287inlining mechanism in some other way, such as
1288
4cee8e80 1289 sub not_inlined () {
54310121 1290 23 if $];
4cee8e80
CS
1291 }
1292
19799a22 1293=head2 Overriding Built-in Functions
d74e8afc 1294X<built-in> X<override> X<CORE> X<CORE::GLOBAL>
a0d0e21e 1295
19799a22 1296Many built-in functions may be overridden, though this should be tried
5f05dabc 1297only occasionally and for good reason. Typically this might be
19799a22 1298done by a package attempting to emulate missing built-in functionality
a0d0e21e
LW
1299on a non-Unix system.
1300
163e3a99
JP
1301Overriding may be done only by importing the name from a module at
1302compile time--ordinary predeclaration isn't good enough. However, the
19799a22
GS
1303C<use subs> pragma lets you, in effect, predeclare subs
1304via the import syntax, and these names may then override built-in ones:
a0d0e21e
LW
1305
1306 use subs 'chdir', 'chroot', 'chmod', 'chown';
1307 chdir $somewhere;
1308 sub chdir { ... }
1309
19799a22
GS
1310To unambiguously refer to the built-in form, precede the
1311built-in name with the special package qualifier C<CORE::>. For example,
1312saying C<CORE::open()> always refers to the built-in C<open()>, even
fb73857a 1313if the current package has imported some other subroutine called
19799a22 1314C<&open()> from elsewhere. Even though it looks like a regular
09bef843 1315function call, it isn't: you can't take a reference to it, such as
19799a22 1316the incorrect C<\&CORE::open> might appear to produce.
fb73857a 1317
19799a22
GS
1318Library modules should not in general export built-in names like C<open>
1319or C<chdir> as part of their default C<@EXPORT> list, because these may
a0d0e21e 1320sneak into someone else's namespace and change the semantics unexpectedly.
19799a22 1321Instead, if the module adds that name to C<@EXPORT_OK>, then it's
a0d0e21e
LW
1322possible for a user to import the name explicitly, but not implicitly.
1323That is, they could say
1324
1325 use Module 'open';
1326
19799a22 1327and it would import the C<open> override. But if they said
a0d0e21e
LW
1328
1329 use Module;
1330
19799a22 1331they would get the default imports without overrides.
a0d0e21e 1332
19799a22 1333The foregoing mechanism for overriding built-in is restricted, quite
95d94a4f 1334deliberately, to the package that requests the import. There is a second
19799a22 1335method that is sometimes applicable when you wish to override a built-in
95d94a4f
GS
1336everywhere, without regard to namespace boundaries. This is achieved by
1337importing a sub into the special namespace C<CORE::GLOBAL::>. Here is an
1338example that quite brazenly replaces the C<glob> operator with something
1339that understands regular expressions.
1340
1341 package REGlob;
1342 require Exporter;
1343 @ISA = 'Exporter';
1344 @EXPORT_OK = 'glob';
1345
1346 sub import {
1347 my $pkg = shift;
1348 return unless @_;
1349 my $sym = shift;
1350 my $where = ($sym =~ s/^GLOBAL_// ? 'CORE::GLOBAL' : caller(0));
1351 $pkg->export($where, $sym, @_);
1352 }
1353
1354 sub glob {
1355 my $pat = shift;
1356 my @got;
7b815c67
RGS
1357 if (opendir my $d, '.') {
1358 @got = grep /$pat/, readdir $d;
1359 closedir $d;
19799a22
GS
1360 }
1361 return @got;
95d94a4f
GS
1362 }
1363 1;
1364
1365And here's how it could be (ab)used:
1366
1367 #use REGlob 'GLOBAL_glob'; # override glob() in ALL namespaces
1368 package Foo;
1369 use REGlob 'glob'; # override glob() in Foo:: only
1370 print for <^[a-z_]+\.pm\$>; # show all pragmatic modules
1371
19799a22 1372The initial comment shows a contrived, even dangerous example.
95d94a4f 1373By overriding C<glob> globally, you would be forcing the new (and
19799a22 1374subversive) behavior for the C<glob> operator for I<every> namespace,
95d94a4f
GS
1375without the complete cognizance or cooperation of the modules that own
1376those namespaces. Naturally, this should be done with extreme caution--if
1377it must be done at all.
1378
1379The C<REGlob> example above does not implement all the support needed to
19799a22 1380cleanly override perl's C<glob> operator. The built-in C<glob> has
95d94a4f 1381different behaviors depending on whether it appears in a scalar or list
19799a22 1382context, but our C<REGlob> doesn't. Indeed, many perl built-in have such
95d94a4f
GS
1383context sensitive behaviors, and these must be adequately supported by
1384a properly written override. For a fully functional example of overriding
1385C<glob>, study the implementation of C<File::DosGlob> in the standard
1386library.
1387
77bc9082
RGS
1388When you override a built-in, your replacement should be consistent (if
1389possible) with the built-in native syntax. You can achieve this by using
1390a suitable prototype. To get the prototype of an overridable built-in,
1391use the C<prototype> function with an argument of C<"CORE::builtin_name">
1392(see L<perlfunc/prototype>).
1393
1394Note however that some built-ins can't have their syntax expressed by a
1395prototype (such as C<system> or C<chomp>). If you override them you won't
1396be able to fully mimic their original syntax.
1397
fe854a6f 1398The built-ins C<do>, C<require> and C<glob> can also be overridden, but due
77bc9082
RGS
1399to special magic, their original syntax is preserved, and you don't have
1400to define a prototype for their replacements. (You can't override the
1401C<do BLOCK> syntax, though).
1402
1403C<require> has special additional dark magic: if you invoke your
1404C<require> replacement as C<require Foo::Bar>, it will actually receive
1405the argument C<"Foo/Bar.pm"> in @_. See L<perlfunc/require>.
1406
1407And, as you'll have noticed from the previous example, if you override
593b9c14 1408C<glob>, the C<< <*> >> glob operator is overridden as well.
77bc9082 1409
9b3023bc 1410In a similar fashion, overriding the C<readline> function also overrides
e3f73d4e
RGS
1411the equivalent I/O operator C<< <FILEHANDLE> >>. Also, overriding
1412C<readpipe> also overrides the operators C<``> and C<qx//>.
9b3023bc 1413
fe854a6f 1414Finally, some built-ins (e.g. C<exists> or C<grep>) can't be overridden.
77bc9082 1415
a0d0e21e 1416=head2 Autoloading
d74e8afc 1417X<autoloading> X<AUTOLOAD>
a0d0e21e 1418
19799a22
GS
1419If you call a subroutine that is undefined, you would ordinarily
1420get an immediate, fatal error complaining that the subroutine doesn't
1421exist. (Likewise for subroutines being used as methods, when the
1422method doesn't exist in any base class of the class's package.)
1423However, if an C<AUTOLOAD> subroutine is defined in the package or
1424packages used to locate the original subroutine, then that
1425C<AUTOLOAD> subroutine is called with the arguments that would have
1426been passed to the original subroutine. The fully qualified name
1427of the original subroutine magically appears in the global $AUTOLOAD
1428variable of the same package as the C<AUTOLOAD> routine. The name
1429is not passed as an ordinary argument because, er, well, just
593b9c14 1430because, that's why. (As an exception, a method call to a nonexistent
80ee23cd
RH
1431C<import> or C<unimport> method is just skipped instead. Also, if
1432the AUTOLOAD subroutine is an XSUB, C<$AUTOLOAD> is not populated;
1433instead, you should call L<< C<SvPVX>E<sol>C<SvCUR>|perlapi >> on the
1434C<CV> for C<AUTOLOAD> to retrieve the method name.)
1435
19799a22
GS
1436
1437Many C<AUTOLOAD> routines load in a definition for the requested
1438subroutine using eval(), then execute that subroutine using a special
1439form of goto() that erases the stack frame of the C<AUTOLOAD> routine
1440without a trace. (See the source to the standard module documented
1441in L<AutoLoader>, for example.) But an C<AUTOLOAD> routine can
1442also just emulate the routine and never define it. For example,
1443let's pretend that a function that wasn't defined should just invoke
1444C<system> with those arguments. All you'd do is:
cb1a09d0
AD
1445
1446 sub AUTOLOAD {
1447 my $program = $AUTOLOAD;
1448 $program =~ s/.*:://;
1449 system($program, @_);
54310121 1450 }
cb1a09d0 1451 date();
6d28dffb 1452 who('am', 'i');
cb1a09d0
AD
1453 ls('-l');
1454
19799a22
GS
1455In fact, if you predeclare functions you want to call that way, you don't
1456even need parentheses:
cb1a09d0
AD
1457
1458 use subs qw(date who ls);
1459 date;
1460 who "am", "i";
593b9c14 1461 ls '-l';
cb1a09d0
AD
1462
1463A more complete example of this is the standard Shell module, which
19799a22 1464can treat undefined subroutine calls as calls to external programs.
a0d0e21e 1465
19799a22
GS
1466Mechanisms are available to help modules writers split their modules
1467into autoloadable files. See the standard AutoLoader module
6d28dffb
PP
1468described in L<AutoLoader> and in L<AutoSplit>, the standard
1469SelfLoader modules in L<SelfLoader>, and the document on adding C
19799a22 1470functions to Perl code in L<perlxs>.
cb1a09d0 1471
09bef843 1472=head2 Subroutine Attributes
d74e8afc 1473X<attribute> X<subroutine, attribute> X<attrs>
09bef843
SB
1474
1475A subroutine declaration or definition may have a list of attributes
1476associated with it. If such an attribute list is present, it is
0120eecf 1477broken up at space or colon boundaries and treated as though a
09bef843
SB
1478C<use attributes> had been seen. See L<attributes> for details
1479about what attributes are currently supported.
1480Unlike the limitation with the obsolescent C<use attrs>, the
1481C<sub : ATTRLIST> syntax works to associate the attributes with
1482a pre-declaration, and not just with a subroutine definition.
1483
1484The attributes must be valid as simple identifier names (without any
1485punctuation other than the '_' character). They may have a parameter
1486list appended, which is only checked for whether its parentheses ('(',')')
1487nest properly.
1488
1489Examples of valid syntax (even though the attributes are unknown):
1490
4358a253
SS
1491 sub fnord (&\%) : switch(10,foo(7,3)) : expensive;
1492 sub plugh () : Ugly('\(") :Bad;
09bef843
SB
1493 sub xyzzy : _5x5 { ... }
1494
1495Examples of invalid syntax:
1496
4358a253
SS
1497 sub fnord : switch(10,foo(); # ()-string not balanced
1498 sub snoid : Ugly('('); # ()-string not balanced
1499 sub xyzzy : 5x5; # "5x5" not a valid identifier
1500 sub plugh : Y2::north; # "Y2::north" not a simple identifier
1501 sub snurt : foo + bar; # "+" not a colon or space
09bef843
SB
1502
1503The attribute list is passed as a list of constant strings to the code
1504which associates them with the subroutine. In particular, the second example
1505of valid syntax above currently looks like this in terms of how it's
1506parsed and invoked:
1507
1508 use attributes __PACKAGE__, \&plugh, q[Ugly('\(")], 'Bad';
1509
1510For further details on attribute lists and their manipulation,
a0ae32d3 1511see L<attributes> and L<Attribute::Handlers>.
09bef843 1512
cb1a09d0 1513=head1 SEE ALSO
a0d0e21e 1514
19799a22
GS
1515See L<perlref/"Function Templates"> for more about references and closures.
1516See L<perlxs> if you'd like to learn about calling C subroutines from Perl.
a2293a43 1517See L<perlembed> if you'd like to learn about calling Perl subroutines from C.
19799a22
GS
1518See L<perlmod> to learn about bundling up your functions in separate files.
1519See L<perlmodlib> to learn what library modules come standard on your system.
1520See L<perltoot> to learn how to make object method calls.