3 perlref - Perl references and nested data structures
7 This is complete documentation about all aspects of references.
8 For a shorter, tutorial introduction to just the essential features,
13 Before release 5 of Perl it was difficult to represent complex data
14 structures, because all references had to be symbolic--and even then
15 it was difficult to refer to a variable instead of a symbol table entry.
16 Perl now not only makes it easier to use symbolic references to variables,
17 but also lets you have "hard" references to any piece of data or code.
18 Any scalar may hold a hard reference. Because arrays and hashes contain
19 scalars, you can now easily build arrays of arrays, arrays of hashes,
20 hashes of arrays, arrays of hashes of functions, and so on.
22 Hard references are smart--they keep track of reference counts for you,
23 automatically freeing the thing referred to when its reference count goes
24 to zero. (Reference counts for values in self-referential or
25 cyclic data structures may not go to zero without a little help; see
26 L<perlobj/"Two-Phased Garbage Collection"> for a detailed explanation.)
27 If that thing happens to be an object, the object is destructed. See
28 L<perlobj> for more about objects. (In a sense, everything in Perl is an
29 object, but we usually reserve the word for references to objects that
30 have been officially "blessed" into a class package.)
32 Symbolic references are names of variables or other objects, just as a
33 symbolic link in a Unix filesystem contains merely the name of a file.
34 The C<*glob> notation is something of a symbolic reference. (Symbolic
35 references are sometimes called "soft references", but please don't call
36 them that; references are confusing enough without useless synonyms.)
38 In contrast, hard references are more like hard links in a Unix file
39 system: They are used to access an underlying object without concern for
40 what its (other) name is. When the word "reference" is used without an
41 adjective, as in the following paragraph, it is usually talking about a
44 References are easy to use in Perl. There is just one overriding
45 principle: Perl does no implicit referencing or dereferencing. When a
46 scalar is holding a reference, it always behaves as a simple scalar. It
47 doesn't magically start being an array or hash or subroutine; you have to
48 tell it explicitly to do so, by dereferencing it.
50 =head2 Making References
52 References can be created in several ways.
58 By using the backslash operator on a variable, subroutine, or value.
59 (This works much like the & (address-of) operator in C.)
60 This typically creates I<another> reference to a variable, because
61 there's already a reference to the variable in the symbol table. But
62 the symbol table reference might go away, and you'll still have the
63 reference that the backslash returned. Here are some examples:
71 It isn't possible to create a true reference to an IO handle (filehandle
72 or dirhandle) using the backslash operator. The most you can get is a
73 reference to a typeglob, which is actually a complete symbol table entry.
74 But see the explanation of the C<*foo{THING}> syntax below. However,
75 you can still use type globs and globrefs as though they were IO handles.
79 A reference to an anonymous array can be created using square
82 $arrayref = [1, 2, ['a', 'b', 'c']];
84 Here we've created a reference to an anonymous array of three elements
85 whose final element is itself a reference to another anonymous array of three
86 elements. (The multidimensional syntax described later can be used to
87 access this. For example, after the above, C<< $arrayref->[2][1] >> would have
90 Taking a reference to an enumerated list is not the same
91 as using square brackets--instead it's the same as creating
94 @list = (\$a, \@b, \%c);
95 @list = \($a, @b, %c); # same thing!
97 As a special case, C<\(@foo)> returns a list of references to the contents
98 of C<@foo>, not a reference to C<@foo> itself. Likewise for C<%foo>,
99 except that the key references are to copies (since the keys are just
100 strings rather than full-fledged scalars).
104 A reference to an anonymous hash can be created using curly
112 Anonymous hash and array composers like these can be intermixed freely to
113 produce as complicated a structure as you want. The multidimensional
114 syntax described below works for these too. The values above are
115 literals, but variables and expressions would work just as well, because
116 assignment operators in Perl (even within local() or my()) are executable
117 statements, not compile-time declarations.
119 Because curly brackets (braces) are used for several other things
120 including BLOCKs, you may occasionally have to disambiguate braces at the
121 beginning of a statement by putting a C<+> or a C<return> in front so
122 that Perl realizes the opening brace isn't starting a BLOCK. The economy and
123 mnemonic value of using curlies is deemed worth this occasional extra
126 For example, if you wanted a function to make a new hash and return a
127 reference to it, you have these options:
129 sub hashem { { @_ } } # silently wrong
130 sub hashem { +{ @_ } } # ok
131 sub hashem { return { @_ } } # ok
133 On the other hand, if you want the other meaning, you can do this:
135 sub showem { { @_ } } # ambiguous (currently ok, but may change)
136 sub showem { {; @_ } } # ok
137 sub showem { { return @_ } } # ok
139 The leading C<+{> and C<{;> always serve to disambiguate
140 the expression to mean either the HASH reference, or the BLOCK.
144 A reference to an anonymous subroutine can be created by using
145 C<sub> without a subname:
147 $coderef = sub { print "Boink!\n" };
149 Note the semicolon. Except for the code
150 inside not being immediately executed, a C<sub {}> is not so much a
151 declaration as it is an operator, like C<do{}> or C<eval{}>. (However, no
152 matter how many times you execute that particular line (unless you're in an
153 C<eval("...")>), $coderef will still have a reference to the I<same>
154 anonymous subroutine.)
156 Anonymous subroutines act as closures with respect to my() variables,
157 that is, variables lexically visible within the current scope. Closure
158 is a notion out of the Lisp world that says if you define an anonymous
159 function in a particular lexical context, it pretends to run in that
160 context even when it's called outside the context.
162 In human terms, it's a funny way of passing arguments to a subroutine when
163 you define it as well as when you call it. It's useful for setting up
164 little bits of code to run later, such as callbacks. You can even
165 do object-oriented stuff with it, though Perl already provides a different
166 mechanism to do that--see L<perlobj>.
168 You might also think of closure as a way to write a subroutine
169 template without using eval(). Here's a small example of how
174 return sub { my $y = shift; print "$x, $y!\n"; };
176 $h = newprint("Howdy");
177 $g = newprint("Greetings");
187 Greetings, earthlings!
189 Note particularly that $x continues to refer to the value passed
190 into newprint() I<despite> "my $x" having gone out of scope by the
191 time the anonymous subroutine runs. That's what a closure is all
194 This applies only to lexical variables, by the way. Dynamic variables
195 continue to work as they have always worked. Closure is not something
196 that most Perl programmers need trouble themselves about to begin with.
200 References are often returned by special subroutines called constructors.
201 Perl objects are just references to a special type of object that happens to know
202 which package it's associated with. Constructors are just special
203 subroutines that know how to create that association. They do so by
204 starting with an ordinary reference, and it remains an ordinary reference
205 even while it's also being an object. Constructors are often
206 named new() and called indirectly:
208 $objref = new Doggie (Tail => 'short', Ears => 'long');
210 But don't have to be:
212 $objref = Doggie->new(Tail => 'short', Ears => 'long');
215 $terminal = Term::Cap->Tgetent( { OSPEED => 9600 });
218 $main = MainWindow->new();
219 $menubar = $main->Frame(-relief => "raised",
224 References of the appropriate type can spring into existence if you
225 dereference them in a context that assumes they exist. Because we haven't
226 talked about dereferencing yet, we can't show you any examples yet.
230 A reference can be created by using a special syntax, lovingly known as
231 the *foo{THING} syntax. *foo{THING} returns a reference to the THING
232 slot in *foo (which is the symbol table entry which holds everything
235 $scalarref = *foo{SCALAR};
236 $arrayref = *ARGV{ARRAY};
237 $hashref = *ENV{HASH};
238 $coderef = *handler{CODE};
240 $globref = *foo{GLOB};
242 All of these are self-explanatory except for C<*foo{IO}>. It returns
243 the IO handle, used for file handles (L<perlfunc/open>), sockets
244 (L<perlfunc/socket> and L<perlfunc/socketpair>), and directory
245 handles (L<perlfunc/opendir>). For compatibility with previous
246 versions of Perl, C<*foo{FILEHANDLE}> is a synonym for C<*foo{IO}>, though it
247 is deprecated as of 5.8.0. If deprecation warnings are in effect, it will warn
250 C<*foo{THING}> returns undef if that particular THING hasn't been used yet,
251 except in the case of scalars. C<*foo{SCALAR}> returns a reference to an
252 anonymous scalar if $foo hasn't been used yet. This might change in a
255 C<*foo{IO}> is an alternative to the C<*HANDLE> mechanism given in
256 L<perldata/"Typeglobs and Filehandles"> for passing filehandles
257 into or out of subroutines, or storing into larger data structures.
258 Its disadvantage is that it won't create a new filehandle for you.
259 Its advantage is that you have less risk of clobbering more than
260 you want to with a typeglob assignment. (It still conflates file
261 and directory handles, though.) However, if you assign the incoming
262 value to a scalar instead of a typeglob as we do in the examples
263 below, there's no risk of that happening.
265 splutter(*STDOUT); # pass the whole glob
266 splutter(*STDOUT{IO}); # pass both file and dir handles
270 print $fh "her um well a hmmm\n";
273 $rec = get_rec(*STDIN); # pass the whole glob
274 $rec = get_rec(*STDIN{IO}); # pass both file and dir handles
283 =head2 Using References
285 That's it for creating references. By now you're probably dying to
286 know how to use references to get back to your long-lost data. There
287 are several basic methods.
293 Anywhere you'd put an identifier (or chain of identifiers) as part
294 of a variable or subroutine name, you can replace the identifier with
295 a simple scalar variable containing a reference of the correct type:
298 push(@$arrayref, $filename);
299 $$arrayref[0] = "January";
300 $$hashref{"KEY"} = "VALUE";
302 print $globref "output\n";
304 It's important to understand that we are specifically I<not> dereferencing
305 C<$arrayref[0]> or C<$hashref{"KEY"}> there. The dereference of the
306 scalar variable happens I<before> it does any key lookups. Anything more
307 complicated than a simple scalar variable must use methods 2 or 3 below.
308 However, a "simple scalar" includes an identifier that itself uses method
309 1 recursively. Therefore, the following prints "howdy".
311 $refrefref = \\\"howdy";
316 Anywhere you'd put an identifier (or chain of identifiers) as part of a
317 variable or subroutine name, you can replace the identifier with a
318 BLOCK returning a reference of the correct type. In other words, the
319 previous examples could be written like this:
321 $bar = ${$scalarref};
322 push(@{$arrayref}, $filename);
323 ${$arrayref}[0] = "January";
324 ${$hashref}{"KEY"} = "VALUE";
326 $globref->print("output\n"); # iff IO::Handle is loaded
328 Admittedly, it's a little silly to use the curlies in this case, but
329 the BLOCK can contain any arbitrary expression, in particular,
330 subscripted expressions:
332 &{ $dispatch{$index} }(1,2,3); # call correct routine
334 Because of being able to omit the curlies for the simple case of C<$$x>,
335 people often make the mistake of viewing the dereferencing symbols as
336 proper operators, and wonder about their precedence. If they were,
337 though, you could use parentheses instead of braces. That's not the case.
338 Consider the difference below; case 0 is a short-hand version of case 1,
341 $$hashref{"KEY"} = "VALUE"; # CASE 0
342 ${$hashref}{"KEY"} = "VALUE"; # CASE 1
343 ${$hashref{"KEY"}} = "VALUE"; # CASE 2
344 ${$hashref->{"KEY"}} = "VALUE"; # CASE 3
346 Case 2 is also deceptive in that you're accessing a variable
347 called %hashref, not dereferencing through $hashref to the hash
348 it's presumably referencing. That would be case 3.
352 Subroutine calls and lookups of individual array elements arise often
353 enough that it gets cumbersome to use method 2. As a form of
354 syntactic sugar, the examples for method 2 may be written:
356 $arrayref->[0] = "January"; # Array element
357 $hashref->{"KEY"} = "VALUE"; # Hash element
358 $coderef->(1,2,3); # Subroutine call
360 The left side of the arrow can be any expression returning a reference,
361 including a previous dereference. Note that C<$array[$x]> is I<not> the
362 same thing as C<< $array->[$x] >> here:
364 $array[$x]->{"foo"}->[0] = "January";
366 This is one of the cases we mentioned earlier in which references could
367 spring into existence when in an lvalue context. Before this
368 statement, C<$array[$x]> may have been undefined. If so, it's
369 automatically defined with a hash reference so that we can look up
370 C<{"foo"}> in it. Likewise C<< $array[$x]->{"foo"} >> will automatically get
371 defined with an array reference so that we can look up C<[0]> in it.
372 This process is called I<autovivification>.
374 One more thing here. The arrow is optional I<between> brackets
375 subscripts, so you can shrink the above down to
377 $array[$x]{"foo"}[0] = "January";
379 Which, in the degenerate case of using only ordinary arrays, gives you
380 multidimensional arrays just like C's:
382 $score[$x][$y][$z] += 42;
384 Well, okay, not entirely like C's arrays, actually. C doesn't know how
385 to grow its arrays on demand. Perl does.
389 If a reference happens to be a reference to an object, then there are
390 probably methods to access the things referred to, and you should probably
391 stick to those methods unless you're in the class package that defines the
392 object's methods. In other words, be nice, and don't violate the object's
393 encapsulation without a very good reason. Perl does not enforce
394 encapsulation. We are not totalitarians here. We do expect some basic
399 Using a string or number as a reference produces a symbolic reference,
400 as explained above. Using a reference as a number produces an
401 integer representing its storage location in memory. The only
402 useful thing to be done with this is to compare two references
403 numerically to see whether they refer to the same location.
405 if ($ref1 == $ref2) { # cheap numeric compare of references
406 print "refs 1 and 2 refer to the same thing\n";
409 Using a reference as a string produces both its referent's type,
410 including any package blessing as described in L<perlobj>, as well
411 as the numeric address expressed in hex. The ref() operator returns
412 just the type of thing the reference is pointing to, without the
413 address. See L<perlfunc/ref> for details and examples of its use.
415 The bless() operator may be used to associate the object a reference
416 points to with a package functioning as an object class. See L<perlobj>.
418 A typeglob may be dereferenced the same way a reference can, because
419 the dereference syntax always indicates the type of reference desired.
420 So C<${*foo}> and C<${\$foo}> both indicate the same scalar variable.
422 Here's a trick for interpolating a subroutine call into a string:
424 print "My sub returned @{[mysub(1,2,3)]} that time.\n";
426 The way it works is that when the C<@{...}> is seen in the double-quoted
427 string, it's evaluated as a block. The block creates a reference to an
428 anonymous array containing the results of the call to C<mysub(1,2,3)>. So
429 the whole block returns a reference to an array, which is then
430 dereferenced by C<@{...}> and stuck into the double-quoted string. This
431 chicanery is also useful for arbitrary expressions:
433 print "That yields @{[$n + 5]} widgets\n";
435 =head2 Symbolic references
437 We said that references spring into existence as necessary if they are
438 undefined, but we didn't say what happens if a value used as a
439 reference is already defined, but I<isn't> a hard reference. If you
440 use it as a reference, it'll be treated as a symbolic
441 reference. That is, the value of the scalar is taken to be the I<name>
442 of a variable, rather than a direct link to a (possibly) anonymous
445 People frequently expect it to work like this. So it does.
448 $$name = 1; # Sets $foo
449 ${$name} = 2; # Sets $foo
450 ${$name x 2} = 3; # Sets $foofoo
451 $name->[0] = 4; # Sets $foo[0]
452 @$name = (); # Clears @foo
453 &$name(); # Calls &foo() (as in Perl 4)
455 ${"${pack}::$name"} = 5; # Sets $THAT::foo without eval
457 This is powerful, and slightly dangerous, in that it's possible
458 to intend (with the utmost sincerity) to use a hard reference, and
459 accidentally use a symbolic reference instead. To protect against
464 and then only hard references will be allowed for the rest of the enclosing
465 block. An inner block may countermand that with
469 Only package variables (globals, even if localized) are visible to
470 symbolic references. Lexical variables (declared with my()) aren't in
471 a symbol table, and thus are invisible to this mechanism. For example:
480 This will still print 10, not 20. Remember that local() affects package
481 variables, which are all "global" to the package.
483 =head2 Not-so-symbolic references
485 A new feature contributing to readability in perl version 5.001 is that the
486 brackets around a symbolic reference behave more like quotes, just as they
487 always have within a string. That is,
492 has always meant to print "pop on over", even though push is
493 a reserved word. This has been generalized to work the same outside
496 print ${push} . "over";
500 print ${ push } . "over";
502 will have the same effect. (This would have been a syntax error in
503 Perl 5.000, though Perl 4 allowed it in the spaceless form.) This
504 construct is I<not> considered to be a symbolic reference when you're
508 ${ bareword }; # Okay, means $bareword.
509 ${ "bareword" }; # Error, symbolic reference.
511 Similarly, because of all the subscripting that is done using single
512 words, we've applied the same rule to any bareword that is used for
513 subscripting a hash. So now, instead of writing
515 $array{ "aaa" }{ "bbb" }{ "ccc" }
519 $array{ aaa }{ bbb }{ ccc }
521 and not worry about whether the subscripts are reserved words. In the
522 rare event that you do wish to do something like
526 you can force interpretation as a reserved word by adding anything that
527 makes it more than a bareword:
533 The C<use warnings> pragma or the B<-w> switch will warn you if it
534 interprets a reserved word as a string.
535 But it will no longer warn you about using lowercase words, because the
536 string is effectively quoted.
538 =head2 Pseudo-hashes: Using an array as a hash
540 B<WARNING>: This section describes an experimental feature. Details may
541 change without notice in future versions.
543 B<NOTE>: The current user-visible implementation of pseudo-hashes
544 (the weird use of the first array element) is deprecated starting from
545 Perl 5.8.0 and will be removed in Perl 5.10.0, and the feature will be
546 implemented differently. Not only is the current interface rather ugly,
547 but the current implementation slows down normal array and hash use quite
548 noticeably. The 'fields' pragma interface will remain available.
550 Beginning with release 5.005 of Perl, you may use an array reference
551 in some contexts that would normally require a hash reference. This
552 allows you to access array elements using symbolic names, as if they
553 were fields in a structure.
555 For this to work, the array must contain extra information. The first
556 element of the array has to be a hash reference that maps field names
557 to array indices. Here is an example:
559 $struct = [{foo => 1, bar => 2}, "FOO", "BAR"];
561 $struct->{foo}; # same as $struct->[1], i.e. "FOO"
562 $struct->{bar}; # same as $struct->[2], i.e. "BAR"
564 keys %$struct; # will return ("foo", "bar") in some order
565 values %$struct; # will return ("FOO", "BAR") in same some order
567 while (my($k,$v) = each %$struct) {
571 Perl will raise an exception if you try to access nonexistent fields.
572 To avoid inconsistencies, always use the fields::phash() function
573 provided by the C<fields> pragma.
576 $pseudohash = fields::phash(foo => "FOO", bar => "BAR");
578 For better performance, Perl can also do the translation from field
579 names to array indices at compile time for typed object references.
582 There are two ways to check for the existence of a key in a
583 pseudo-hash. The first is to use exists(). This checks to see if the
584 given field has ever been set. It acts this way to match the behavior
585 of a regular hash. For instance:
588 $phash = fields::phash([qw(foo bar pants)], ['FOO']);
589 $phash->{pants} = undef;
591 print exists $phash->{foo}; # true, 'foo' was set in the declaration
592 print exists $phash->{bar}; # false, 'bar' has not been used.
593 print exists $phash->{pants}; # true, your 'pants' have been touched
595 The second is to use exists() on the hash reference sitting in the
596 first array element. This checks to see if the given key is a valid
597 field in the pseudo-hash.
599 print exists $phash->[0]{bar}; # true, 'bar' is a valid field
600 print exists $phash->[0]{shoes};# false, 'shoes' can't be used
602 delete() on a pseudo-hash element only deletes the value corresponding
603 to the key, not the key itself. To delete the key, you'll have to
604 explicitly delete it from the first hash element.
606 print delete $phash->{foo}; # prints $phash->[1], "FOO"
607 print exists $phash->{foo}; # false
608 print exists $phash->[0]{foo}; # true, key still exists
609 print delete $phash->[0]{foo}; # now key is gone
610 print $phash->{foo}; # runtime exception
612 =head2 Function Templates
614 As explained above, a closure is an anonymous function with access to the
615 lexical variables visible when that function was compiled. It retains
616 access to those variables even though it doesn't get run until later,
617 such as in a signal handler or a Tk callback.
619 Using a closure as a function template allows us to generate many functions
620 that act similarly. Suppose you wanted functions named after the colors
621 that generated HTML font changes for the various colors:
623 print "Be ", red("careful"), "with that ", green("light");
625 The red() and green() functions would be similar. To create these,
626 we'll assign a closure to a typeglob of the name of the function we're
629 @colors = qw(red blue green yellow orange purple violet);
630 for my $name (@colors) {
631 no strict 'refs'; # allow symbol table manipulation
632 *$name = *{uc $name} = sub { "<FONT COLOR='$name'>@_</FONT>" };
635 Now all those different functions appear to exist independently. You can
636 call red(), RED(), blue(), BLUE(), green(), etc. This technique saves on
637 both compile time and memory use, and is less error-prone as well, since
638 syntax checks happen at compile time. It's critical that any variables in
639 the anonymous subroutine be lexicals in order to create a proper closure.
640 That's the reasons for the C<my> on the loop iteration variable.
642 This is one of the only places where giving a prototype to a closure makes
643 much sense. If you wanted to impose scalar context on the arguments of
644 these functions (probably not a wise idea for this particular example),
645 you could have written it this way instead:
647 *$name = sub ($) { "<FONT COLOR='$name'>$_[0]</FONT>" };
649 However, since prototype checking happens at compile time, the assignment
650 above happens too late to be of much use. You could address this by
651 putting the whole loop of assignments within a BEGIN block, forcing it
652 to occur during compilation.
654 Access to lexicals that change over type--like those in the C<for> loop
655 above--only works with closures, not general subroutines. In the general
656 case, then, named subroutines do not nest properly, although anonymous
657 ones do. If you are accustomed to using nested subroutines in other
658 programming languages with their own private variables, you'll have to
659 work at it a bit in Perl. The intuitive coding of this type of thing
660 incurs mysterious warnings about ``will not stay shared''. For example,
665 sub inner { return $x * 19 } # WRONG
669 A work-around is the following:
673 local *inner = sub { return $x * 19 };
677 Now inner() can only be called from within outer(), because of the
678 temporary assignments of the closure (anonymous subroutine). But when
679 it does, it has normal access to the lexical variable $x from the scope
682 This has the interesting effect of creating a function local to another
683 function, something not normally supported in Perl.
687 You may not (usefully) use a reference as the key to a hash. It will be
688 converted into a string:
692 If you try to dereference the key, it won't do a hard dereference, and
693 you won't accomplish what you're attempting. You might want to do something
699 And then at least you can use the values(), which will be
700 real refs, instead of the keys(), which won't.
702 The standard Tie::RefHash module provides a convenient workaround to this.
706 Besides the obvious documents, source code can be instructive.
707 Some pathological examples of the use of references can be found
708 in the F<t/op/ref.t> regression test in the Perl source directory.
710 See also L<perldsc> and L<perllol> for how to use references to create
711 complex data structures, and L<perltoot>, L<perlobj>, and L<perlbot>
712 for how to use them to create objects.