1=head1 NAME 2X<reference> X<pointer> X<data structure> X<structure> X<struct> 3 4perlref - Perl references and nested data structures 5 6=head1 NOTE 7 8This is complete documentation about all aspects of references. 9For a shorter, tutorial introduction to just the essential features, 10see L<perlreftut>. 11 12=head1 DESCRIPTION 13 14Before release 5 of Perl it was difficult to represent complex data 15structures, because all references had to be symbolic--and even then 16it was difficult to refer to a variable instead of a symbol table entry. 17Perl now not only makes it easier to use symbolic references to variables, 18but also lets you have "hard" references to any piece of data or code. 19Any scalar may hold a hard reference. Because arrays and hashes contain 20scalars, you can now easily build arrays of arrays, arrays of hashes, 21hashes of arrays, arrays of hashes of functions, and so on. 22 23Hard references are smart--they keep track of reference counts for you, 24automatically freeing the thing referred to when its reference count goes 25to zero. (Reference counts for values in self-referential or 26cyclic data structures may not go to zero without a little help; see 27L</"Circular References"> for a detailed explanation.) 28If that thing happens to be an object, the object is destructed. See 29L<perlobj> for more about objects. (In a sense, everything in Perl is an 30object, but we usually reserve the word for references to objects that 31have been officially "blessed" into a class package.) 32 33Symbolic references are names of variables or other objects, just as a 34symbolic link in a Unix filesystem contains merely the name of a file. 35The C<*glob> notation is something of a symbolic reference. (Symbolic 36references are sometimes called "soft references", but please don't call 37them that; references are confusing enough without useless synonyms.) 38X<reference, symbolic> X<reference, soft> 39X<symbolic reference> X<soft reference> 40 41In contrast, hard references are more like hard links in a Unix file 42system: They are used to access an underlying object without concern for 43what its (other) name is. When the word "reference" is used without an 44adjective, as in the following paragraph, it is usually talking about a 45hard reference. 46X<reference, hard> X<hard reference> 47 48References are easy to use in Perl. There is just one overriding 49principle: in general, Perl does no implicit referencing or dereferencing. 50When a scalar is holding a reference, it always behaves as a simple scalar. 51It doesn't magically start being an array or hash or subroutine; you have to 52tell it explicitly to do so, by dereferencing it. 53 54=head2 Making References 55X<reference, creation> X<referencing> 56 57References can be created in several ways. 58 59=head3 Backslash Operator 60X<\> X<backslash> 61 62By using the backslash operator on a variable, subroutine, or value. 63(This works much like the & (address-of) operator in C.) 64This typically creates I<another> reference to a variable, because 65there's already a reference to the variable in the symbol table. But 66the symbol table reference might go away, and you'll still have the 67reference that the backslash returned. Here are some examples: 68 69 $scalarref = \$foo; 70 $arrayref = \@ARGV; 71 $hashref = \%ENV; 72 $coderef = \&handler; 73 $globref = \*foo; 74 75It isn't possible to create a true reference to an IO handle (filehandle 76or dirhandle) using the backslash operator. The most you can get is a 77reference to a typeglob, which is actually a complete symbol table entry. 78But see the explanation of the C<*foo{THING}> syntax below. However, 79you can still use type globs and globrefs as though they were IO handles. 80 81=head3 Square Brackets 82X<array, anonymous> X<[> X<[]> X<square bracket> 83X<bracket, square> X<arrayref> X<array reference> X<reference, array> 84 85A reference to an anonymous array can be created using square 86brackets: 87 88 $arrayref = [1, 2, ['a', 'b', 'c']]; 89 90Here we've created a reference to an anonymous array of three elements 91whose final element is itself a reference to another anonymous array of three 92elements. (The multidimensional syntax described later can be used to 93access this. For example, after the above, C<< $arrayref->[2][1] >> would have 94the value "b".) 95 96Taking a reference to an enumerated list is not the same 97as using square brackets--instead it's the same as creating 98a list of references! 99 100 @list = (\$a, \@b, \%c); 101 @list = \($a, @b, %c); # same thing! 102 103As a special case, C<\(@foo)> returns a list of references to the contents 104of C<@foo>, not a reference to C<@foo> itself. Likewise for C<%foo>, 105except that the key references are to copies (since the keys are just 106strings rather than full-fledged scalars). 107 108=head3 Curly Brackets 109X<hash, anonymous> X<{> X<{}> X<curly bracket> 110X<bracket, curly> X<brace> X<hashref> X<hash reference> X<reference, hash> 111 112A reference to an anonymous hash can be created using curly 113brackets: 114 115 $hashref = { 116 'Adam' => 'Eve', 117 'Clyde' => 'Bonnie', 118 }; 119 120Anonymous hash and array composers like these can be intermixed freely to 121produce as complicated a structure as you want. The multidimensional 122syntax described below works for these too. The values above are 123literals, but variables and expressions would work just as well, because 124assignment operators in Perl (even within local() or my()) are executable 125statements, not compile-time declarations. 126 127Because curly brackets (braces) are used for several other things 128including BLOCKs, you may occasionally have to disambiguate braces at the 129beginning of a statement by putting a C<+> or a C<return> in front so 130that Perl realizes the opening brace isn't starting a BLOCK. The economy and 131mnemonic value of using curlies is deemed worth this occasional extra 132hassle. 133 134For example, if you wanted a function to make a new hash and return a 135reference to it, you have these options: 136 137 sub hashem { { @_ } } # silently wrong 138 sub hashem { +{ @_ } } # ok 139 sub hashem { return { @_ } } # ok 140 141On the other hand, if you want the other meaning, you can do this: 142 143 sub showem { { @_ } } # ambiguous (currently ok, 144 # but may change) 145 sub showem { {; @_ } } # ok 146 sub showem { { return @_ } } # ok 147 148The leading C<+{> and C<{;> always serve to disambiguate 149the expression to mean either the HASH reference, or the BLOCK. 150 151=head3 Anonymous Subroutines 152X<subroutine, anonymous> X<subroutine, reference> X<reference, subroutine> 153X<scope, lexical> X<closure> X<lexical> X<lexical scope> 154 155A reference to an anonymous subroutine can be created by using 156C<sub> without a subname: 157 158 $coderef = sub { print "Boink!\n" }; 159 160Note the semicolon. Except for the code 161inside not being immediately executed, a C<sub {}> is not so much a 162declaration as it is an operator, like C<do{}> or C<eval{}>. (However, no 163matter how many times you execute that particular line (unless you're in an 164C<eval("...")>), $coderef will still have a reference to the I<same> 165anonymous subroutine.) 166 167Anonymous subroutines act as closures with respect to my() variables, 168that is, variables lexically visible within the current scope. Closure 169is a notion out of the Lisp world that says if you define an anonymous 170function in a particular lexical context, it pretends to run in that 171context even when it's called outside the context. 172 173In human terms, it's a funny way of passing arguments to a subroutine when 174you define it as well as when you call it. It's useful for setting up 175little bits of code to run later, such as callbacks. You can even 176do object-oriented stuff with it, though Perl already provides a different 177mechanism to do that--see L<perlobj>. 178 179You might also think of closure as a way to write a subroutine 180template without using eval(). Here's a small example of how 181closures work: 182 183 sub newprint { 184 my $x = shift; 185 return sub { my $y = shift; print "$x, $y!\n"; }; 186 } 187 $h = newprint("Howdy"); 188 $g = newprint("Greetings"); 189 190 # Time passes... 191 192 &$h("world"); 193 &$g("earthlings"); 194 195This prints 196 197 Howdy, world! 198 Greetings, earthlings! 199 200Note particularly that $x continues to refer to the value passed 201into newprint() I<despite> "my $x" having gone out of scope by the 202time the anonymous subroutine runs. That's what a closure is all 203about. 204 205This applies only to lexical variables, by the way. Dynamic variables 206continue to work as they have always worked. Closure is not something 207that most Perl programmers need trouble themselves about to begin with. 208 209=head3 Constructors 210X<constructor> X<new> 211 212References are often returned by special subroutines called constructors. Perl 213objects are just references to a special type of object that happens to know 214which package it's associated with. Constructors are just special subroutines 215that know how to create that association. They do so by starting with an 216ordinary reference, and it remains an ordinary reference even while it's also 217being an object. Constructors are often named C<new()>. You I<can> call them 218indirectly: 219 220 $objref = new Doggie( Tail => 'short', Ears => 'long' ); 221 222But that can produce ambiguous syntax in certain cases, so it's often 223better to use the direct method invocation approach: 224 225 $objref = Doggie->new(Tail => 'short', Ears => 'long'); 226 227 use Term::Cap; 228 $terminal = Term::Cap->Tgetent( { OSPEED => 9600 }); 229 230 use Tk; 231 $main = MainWindow->new(); 232 $menubar = $main->Frame(-relief => "raised", 233 -borderwidth => 2) 234 235=head3 Autovivification 236X<autovivification> 237 238References of the appropriate type can spring into existence if you 239dereference them in a context that assumes they exist. Because we haven't 240talked about dereferencing yet, we can't show you any examples yet. 241 242=head3 Typeglob Slots 243X<*foo{THING}> X<*> 244 245A reference can be created by using a special syntax, lovingly known as 246the *foo{THING} syntax. *foo{THING} returns a reference to the THING 247slot in *foo (which is the symbol table entry which holds everything 248known as foo). 249 250 $scalarref = *foo{SCALAR}; 251 $arrayref = *ARGV{ARRAY}; 252 $hashref = *ENV{HASH}; 253 $coderef = *handler{CODE}; 254 $ioref = *STDIN{IO}; 255 $globref = *foo{GLOB}; 256 $formatref = *foo{FORMAT}; 257 $globname = *foo{NAME}; # "foo" 258 $pkgname = *foo{PACKAGE}; # "main" 259 260Most of these are self-explanatory, but C<*foo{IO}> 261deserves special attention. It returns 262the IO handle, used for file handles (L<perlfunc/open>), sockets 263(L<perlfunc/socket> and L<perlfunc/socketpair>), and directory 264handles (L<perlfunc/opendir>). For compatibility with previous 265versions of Perl, C<*foo{FILEHANDLE}> is a synonym for C<*foo{IO}>, though it 266is discouraged, to encourage a consistent use of one name: IO. On perls 267between v5.8 and v5.22, it will issue a deprecation warning, but this 268deprecation has since been rescinded. 269 270C<*foo{THING}> returns undef if that particular THING hasn't been used yet, 271except in the case of scalars. C<*foo{SCALAR}> returns a reference to an 272anonymous scalar if $foo hasn't been used yet. This might change in a 273future release. 274 275C<*foo{NAME}> and C<*foo{PACKAGE}> are the exception, in that they return 276strings, rather than references. These return the package and name of the 277typeglob itself, rather than one that has been assigned to it. So, after 278C<*foo=*Foo::bar>, C<*foo> will become "*Foo::bar" when used as a string, 279but C<*foo{PACKAGE}> and C<*foo{NAME}> will continue to produce "main" and 280"foo", respectively. 281 282C<*foo{IO}> is an alternative to the C<*HANDLE> mechanism given in 283L<perldata/"Typeglobs and Filehandles"> for passing filehandles 284into or out of subroutines, or storing into larger data structures. 285Its disadvantage is that it won't create a new filehandle for you. 286Its advantage is that you have less risk of clobbering more than 287you want to with a typeglob assignment. (It still conflates file 288and directory handles, though.) However, if you assign the incoming 289value to a scalar instead of a typeglob as we do in the examples 290below, there's no risk of that happening. 291 292 splutter(*STDOUT); # pass the whole glob 293 splutter(*STDOUT{IO}); # pass both file and dir handles 294 295 sub splutter { 296 my $fh = shift; 297 print $fh "her um well a hmmm\n"; 298 } 299 300 $rec = get_rec(*STDIN); # pass the whole glob 301 $rec = get_rec(*STDIN{IO}); # pass both file and dir handles 302 303 sub get_rec { 304 my $fh = shift; 305 return scalar <$fh>; 306 } 307 308=head2 Using References 309X<reference, use> X<dereferencing> X<dereference> 310 311That's it for creating references. By now you're probably dying to 312know how to use references to get back to your long-lost data. There 313are several basic methods. 314 315=head3 Simple Scalar 316 317Anywhere you'd put an identifier (or chain of identifiers) as part 318of a variable or subroutine name, you can replace the identifier with 319a simple scalar variable containing a reference of the correct type: 320 321 $bar = $$scalarref; 322 push(@$arrayref, $filename); 323 $$arrayref[0] = "January"; 324 $$hashref{"KEY"} = "VALUE"; 325 &$coderef(1,2,3); 326 print $globref "output\n"; 327 328It's important to understand that we are specifically I<not> dereferencing 329C<$arrayref[0]> or C<$hashref{"KEY"}> there. The dereference of the 330scalar variable happens I<before> it does any key lookups. Anything more 331complicated than a simple scalar variable must use methods 2 or 3 below. 332However, a "simple scalar" includes an identifier that itself uses method 3331 recursively. Therefore, the following prints "howdy". 334 335 $refrefref = \\\"howdy"; 336 print $$$$refrefref; 337 338=head3 Block 339 340Anywhere you'd put an identifier (or chain of identifiers) as part of a 341variable or subroutine name, you can replace the identifier with a 342BLOCK returning a reference of the correct type. In other words, the 343previous examples could be written like this: 344 345 $bar = ${$scalarref}; 346 push(@{$arrayref}, $filename); 347 ${$arrayref}[0] = "January"; 348 ${$hashref}{"KEY"} = "VALUE"; 349 &{$coderef}(1,2,3); 350 $globref->print("output\n"); # iff IO::Handle is loaded 351 352Admittedly, it's a little silly to use the curlies in this case, but 353the BLOCK can contain any arbitrary expression, in particular, 354subscripted expressions: 355 356 &{ $dispatch{$index} }(1,2,3); # call correct routine 357 358Because of being able to omit the curlies for the simple case of C<$$x>, 359people often make the mistake of viewing the dereferencing symbols as 360proper operators, and wonder about their precedence. If they were, 361though, you could use parentheses instead of braces. That's not the case. 362Consider the difference below; case 0 is a short-hand version of case 1, 363I<not> case 2: 364 365 $$hashref{"KEY"} = "VALUE"; # CASE 0 366 ${$hashref}{"KEY"} = "VALUE"; # CASE 1 367 ${$hashref{"KEY"}} = "VALUE"; # CASE 2 368 ${$hashref->{"KEY"}} = "VALUE"; # CASE 3 369 370Case 2 is also deceptive in that you're accessing a variable 371called %hashref, not dereferencing through $hashref to the hash 372it's presumably referencing. That would be case 3. 373 374=head3 Arrow Notation 375 376Subroutine calls and lookups of individual array elements arise often 377enough that it gets cumbersome to use method 2. As a form of 378syntactic sugar, the examples for method 2 may be written: 379 380 $arrayref->[0] = "January"; # Array element 381 $hashref->{"KEY"} = "VALUE"; # Hash element 382 $coderef->(1,2,3); # Subroutine call 383 384The left side of the arrow can be any expression returning a reference, 385including a previous dereference. Note that C<$array[$x]> is I<not> the 386same thing as C<< $array->[$x] >> here: 387 388 $array[$x]->{"foo"}->[0] = "January"; 389 390This is one of the cases we mentioned earlier in which references could 391spring into existence when in an lvalue context. Before this 392statement, C<$array[$x]> may have been undefined. If so, it's 393automatically defined with a hash reference so that we can look up 394C<{"foo"}> in it. Likewise C<< $array[$x]->{"foo"} >> will automatically get 395defined with an array reference so that we can look up C<[0]> in it. 396This process is called I<autovivification>. 397 398One more thing here. The arrow is optional I<between> brackets 399subscripts, so you can shrink the above down to 400 401 $array[$x]{"foo"}[0] = "January"; 402 403Which, in the degenerate case of using only ordinary arrays, gives you 404multidimensional arrays just like C's: 405 406 $score[$x][$y][$z] += 42; 407 408Well, okay, not entirely like C's arrays, actually. C doesn't know how 409to grow its arrays on demand. Perl does. 410 411=head3 Objects 412 413If a reference happens to be a reference to an object, then there are 414probably methods to access the things referred to, and you should probably 415stick to those methods unless you're in the class package that defines the 416object's methods. In other words, be nice, and don't violate the object's 417encapsulation without a very good reason. Perl does not enforce 418encapsulation. We are not totalitarians here. We do expect some basic 419civility though. 420 421=head3 Miscellaneous Usage 422 423Using a string or number as a reference produces a symbolic reference, 424as explained above. Using a reference as a number produces an 425integer representing its storage location in memory. The only 426useful thing to be done with this is to compare two references 427numerically to see whether they refer to the same location. 428X<reference, numeric context> 429 430 if ($ref1 == $ref2) { # cheap numeric compare of references 431 print "refs 1 and 2 refer to the same thing\n"; 432 } 433 434Using a reference as a string produces both its referent's type, 435including any package blessing as described in L<perlobj>, as well 436as the numeric address expressed in hex. The ref() operator returns 437just the type of thing the reference is pointing to, without the 438address. See L<perlfunc/ref> for details and examples of its use. 439X<reference, string context> 440 441The bless() operator may be used to associate the object a reference 442points to with a package functioning as an object class. See L<perlobj>. 443 444A typeglob may be dereferenced the same way a reference can, because 445the dereference syntax always indicates the type of reference desired. 446So C<${*foo}> and C<${\$foo}> both indicate the same scalar variable. 447 448Here's a trick for interpolating a subroutine call into a string: 449 450 print "My sub returned @{[mysub(1,2,3)]} that time.\n"; 451 452The way it works is that when the C<@{...}> is seen in the double-quoted 453string, it's evaluated as a block. The block creates a reference to an 454anonymous array containing the results of the call to C<mysub(1,2,3)>. So 455the whole block returns a reference to an array, which is then 456dereferenced by C<@{...}> and stuck into the double-quoted string. This 457chicanery is also useful for arbitrary expressions: 458 459 print "That yields @{[$n + 5]} widgets\n"; 460 461Similarly, an expression that returns a reference to a scalar can be 462dereferenced via C<${...}>. Thus, the above expression may be written 463as: 464 465 print "That yields ${\($n + 5)} widgets\n"; 466 467=head2 Circular References 468X<circular reference> X<reference, circular> 469 470It is possible to create a "circular reference" in Perl, which can lead 471to memory leaks. A circular reference occurs when two references 472contain a reference to each other, like this: 473 474 my $foo = {}; 475 my $bar = { foo => $foo }; 476 $foo->{bar} = $bar; 477 478You can also create a circular reference with a single variable: 479 480 my $foo; 481 $foo = \$foo; 482 483In this case, the reference count for the variables will never reach 0, 484and the references will never be garbage-collected. This can lead to 485memory leaks. 486 487Because objects in Perl are implemented as references, it's possible to 488have circular references with objects as well. Imagine a TreeNode class 489where each node references its parent and child nodes. Any node with a 490parent will be part of a circular reference. 491 492You can break circular references by creating a "weak reference". A 493weak reference does not increment the reference count for a variable, 494which means that the object can go out of scope and be destroyed. You 495can weaken a reference with the C<weaken> function exported by the 496L<Scalar::Util> module. 497 498Here's how we can make the first example safer: 499 500 use Scalar::Util 'weaken'; 501 502 my $foo = {}; 503 my $bar = { foo => $foo }; 504 $foo->{bar} = $bar; 505 506 weaken $foo->{bar}; 507 508The reference from C<$foo> to C<$bar> has been weakened. When the 509C<$bar> variable goes out of scope, it will be garbage-collected. The 510next time you look at the value of the C<< $foo->{bar} >> key, it will 511be C<undef>. 512 513This action at a distance can be confusing, so you should be careful 514with your use of weaken. You should weaken the reference in the 515variable that will go out of scope I<first>. That way, the longer-lived 516variable will contain the expected reference until it goes out of 517scope. 518 519=head2 Symbolic references 520X<reference, symbolic> X<reference, soft> 521X<symbolic reference> X<soft reference> 522 523We said that references spring into existence as necessary if they are 524undefined, but we didn't say what happens if a value used as a 525reference is already defined, but I<isn't> a hard reference. If you 526use it as a reference, it'll be treated as a symbolic 527reference. That is, the value of the scalar is taken to be the I<name> 528of a variable, rather than a direct link to a (possibly) anonymous 529value. 530 531People frequently expect it to work like this. So it does. 532 533 $name = "foo"; 534 $$name = 1; # Sets $foo 535 ${$name} = 2; # Sets $foo 536 ${$name x 2} = 3; # Sets $foofoo 537 $name->[0] = 4; # Sets $foo[0] 538 @$name = (); # Clears @foo 539 &$name(); # Calls &foo() 540 $pack = "THAT"; 541 ${"${pack}::$name"} = 5; # Sets $THAT::foo without eval 542 543This is powerful, and slightly dangerous, in that it's possible 544to intend (with the utmost sincerity) to use a hard reference, and 545accidentally use a symbolic reference instead. To protect against 546that, you can say 547 548 use strict 'refs'; 549 550and then only hard references will be allowed for the rest of the enclosing 551block. An inner block may countermand that with 552 553 no strict 'refs'; 554 555Only package variables (globals, even if localized) are visible to 556symbolic references. Lexical variables (declared with my()) aren't in 557a symbol table, and thus are invisible to this mechanism. For example: 558 559 local $value = 10; 560 $ref = "value"; 561 { 562 my $value = 20; 563 print $$ref; 564 } 565 566This will still print 10, not 20. Remember that local() affects package 567variables, which are all "global" to the package. 568 569=head2 Not-so-symbolic references 570 571Brackets around a symbolic reference can simply 572serve to isolate an identifier or variable name from the rest of an 573expression, just as they always have within a string. For example, 574 575 $push = "pop on "; 576 print "${push}over"; 577 578has always meant to print "pop on over", even though push is 579a reserved word. This is generalized to work the same 580without the enclosing double quotes, so that 581 582 print ${push} . "over"; 583 584and even 585 586 print ${ push } . "over"; 587 588will have the same effect. This 589construct is I<not> considered to be a symbolic reference when you're 590using strict refs: 591 592 use strict 'refs'; 593 ${ bareword }; # Okay, means $bareword. 594 ${ "bareword" }; # Error, symbolic reference. 595 596Similarly, because of all the subscripting that is done using single words, 597the same rule applies to any bareword that is used for subscripting a hash. 598So now, instead of writing 599 600 $hash{ "aaa" }{ "bbb" }{ "ccc" } 601 602you can write just 603 604 $hash{ aaa }{ bbb }{ ccc } 605 606and not worry about whether the subscripts are reserved words. In the 607rare event that you do wish to do something like 608 609 $hash{ shift } 610 611you can force interpretation as a reserved word by adding anything that 612makes it more than a bareword: 613 614 $hash{ shift() } 615 $hash{ +shift } 616 $hash{ shift @_ } 617 618The C<use warnings> pragma or the B<-w> switch will warn you if it 619interprets a reserved word as a string. 620But it will no longer warn you about using lowercase words, because the 621string is effectively quoted. 622 623=head2 Pseudo-hashes: Using an array as a hash 624X<pseudo-hash> X<pseudo hash> X<pseudohash> 625 626Pseudo-hashes have been removed from Perl. The 'fields' pragma 627remains available. 628 629=head2 Function Templates 630X<scope, lexical> X<closure> X<lexical> X<lexical scope> 631X<subroutine, nested> X<sub, nested> X<subroutine, local> X<sub, local> 632 633As explained above, an anonymous function with access to the lexical 634variables visible when that function was compiled, creates a closure. It 635retains access to those variables even though it doesn't get run until 636later, such as in a signal handler or a Tk callback. 637 638Using a closure as a function template allows us to generate many functions 639that act similarly. Suppose you wanted functions named after the colors 640that generated HTML font changes for the various colors: 641 642 print "Be ", red("careful"), "with that ", green("light"); 643 644The red() and green() functions would be similar. To create these, 645we'll assign a closure to a typeglob of the name of the function we're 646trying to build. 647 648 @colors = qw(red blue green yellow orange purple violet); 649 for my $name (@colors) { 650 no strict 'refs'; # allow symbol table manipulation 651 *$name = *{uc $name} = sub { "<FONT COLOR='$name'>@_</FONT>" }; 652 } 653 654Now all those different functions appear to exist independently. You can 655call red(), RED(), blue(), BLUE(), green(), etc. This technique saves on 656both compile time and memory use, and is less error-prone as well, since 657syntax checks happen at compile time. It's critical that any variables in 658the anonymous subroutine be lexicals in order to create a proper closure. 659That's the reasons for the C<my> on the loop iteration variable. 660 661This is one of the only places where giving a prototype to a closure makes 662much sense. If you wanted to impose scalar context on the arguments of 663these functions (probably not a wise idea for this particular example), 664you could have written it this way instead: 665 666 *$name = sub ($) { "<FONT COLOR='$name'>$_[0]</FONT>" }; 667 668However, since prototype checking happens at compile time, the assignment 669above happens too late to be of much use. You could address this by 670putting the whole loop of assignments within a BEGIN block, forcing it 671to occur during compilation. 672 673Access to lexicals that change over time--like those in the C<for> loop 674above, basically aliases to elements from the surrounding lexical scopes-- 675only works with anonymous subs, not with named subroutines. Generally 676said, named subroutines do not nest properly and should only be declared 677in the main package scope. 678 679This is because named subroutines are created at compile time so their 680lexical variables get assigned to the parent lexicals from the first 681execution of the parent block. If a parent scope is entered a second 682time, its lexicals are created again, while the nested subs still 683reference the old ones. 684 685Anonymous subroutines get to capture each time you execute the C<sub> 686operator, as they are created on the fly. If you are accustomed to using 687nested subroutines in other programming languages with their own private 688variables, you'll have to work at it a bit in Perl. The intuitive coding 689of this type of thing incurs mysterious warnings about "will not stay 690shared" due to the reasons explained above. 691For example, this won't work: 692 693 sub outer { 694 my $x = $_[0] + 35; 695 sub inner { return $x * 19 } # WRONG 696 return $x + inner(); 697 } 698 699A work-around is the following: 700 701 sub outer { 702 my $x = $_[0] + 35; 703 local *inner = sub { return $x * 19 }; 704 return $x + inner(); 705 } 706 707Now inner() can only be called from within outer(), because of the 708temporary assignments of the anonymous subroutine. But when it does, 709it has normal access to the lexical variable $x from the scope of 710outer() at the time outer is invoked. 711 712This has the interesting effect of creating a function local to another 713function, something not normally supported in Perl. 714 715=head2 Postfix Dereference Syntax 716 717Beginning in v5.20.0, a postfix syntax for using references is 718available. It behaves as described in L</Using References>, but instead 719of a prefixed sigil, a postfixed sigil-and-star is used. 720 721For example: 722 723 $r = \@a; 724 @b = $r->@*; # equivalent to @$r or @{ $r } 725 726 $r = [ 1, [ 2, 3 ], 4 ]; 727 $r->[1]->@*; # equivalent to @{ $r->[1] } 728 729In Perl 5.20 and 5.22, this syntax must be enabled with C<use feature 730'postderef'>. As of Perl 5.24, no feature declarations are required to make 731it available. 732 733Postfix dereference should work in all circumstances where block 734(circumfix) dereference worked, and should be entirely equivalent. This 735syntax allows dereferencing to be written and read entirely 736left-to-right. The following equivalencies are defined: 737 738 $sref->$*; # same as ${ $sref } 739 $aref->@*; # same as @{ $aref } 740 $aref->$#*; # same as $#{ $aref } 741 $href->%*; # same as %{ $href } 742 $cref->&*; # same as &{ $cref } 743 $gref->**; # same as *{ $gref } 744 745Note especially that C<< $cref->&* >> is I<not> equivalent to C<< 746$cref->() >>, and can serve different purposes. 747 748Glob elements can be extracted through the postfix dereferencing feature: 749 750 $gref->*{SCALAR}; # same as *{ $gref }{SCALAR} 751 752Postfix array and scalar dereferencing I<can> be used in interpolating 753strings (double quotes or the C<qq> operator), but only if the 754C<postderef_qq> feature is enabled. 755 756=head2 Postfix Reference Slicing 757 758Value slices of arrays and hashes may also be taken with postfix 759dereferencing notation, with the following equivalencies: 760 761 $aref->@[ ... ]; # same as @$aref[ ... ] 762 $href->@{ ... }; # same as @$href{ ... } 763 764Postfix key/value pair slicing, added in 5.20.0 and documented in 765L<the KeyE<sol>Value Hash Slices section of perldata|perldata/"Key/Value Hash 766Slices">, also behaves as expected: 767 768 $aref->%[ ... ]; # same as %$aref[ ... ] 769 $href->%{ ... }; # same as %$href{ ... } 770 771As with postfix array, postfix value slice dereferencing I<can> be used 772in interpolating strings (double quotes or the C<qq> operator), but only 773if the C<postderef_qq> L<feature> is enabled. 774 775=head2 Assigning to References 776 777Beginning in v5.22.0, the referencing operator can be assigned to. It 778performs an aliasing operation, so that the variable name referenced on the 779left-hand side becomes an alias for the thing referenced on the right-hand 780side: 781 782 \$a = \$b; # $a and $b now point to the same scalar 783 \&foo = \&bar; # foo() now means bar() 784 785This syntax must be enabled with C<use feature 'refaliasing'>. It is 786experimental, and will warn by default unless C<no warnings 787'experimental::refaliasing'> is in effect. 788 789These forms may be assigned to, and cause the right-hand side to be 790evaluated in scalar context: 791 792 \$scalar 793 \@array 794 \%hash 795 \&sub 796 \my $scalar 797 \my @array 798 \my %hash 799 \state $scalar # or @array, etc. 800 \our $scalar # etc. 801 \local $scalar # etc. 802 \local our $scalar # etc. 803 \$some_array[$index] 804 \$some_hash{$key} 805 \local $some_array[$index] 806 \local $some_hash{$key} 807 condition ? \$this : \$that[0] # etc. 808 809Slicing operations and parentheses cause 810the right-hand side to be evaluated in 811list context: 812 813 \@array[5..7] 814 (\@array[5..7]) 815 \(@array[5..7]) 816 \@hash{'foo','bar'} 817 (\@hash{'foo','bar'}) 818 \(@hash{'foo','bar'}) 819 (\$scalar) 820 \($scalar) 821 \(my $scalar) 822 \my($scalar) 823 (\@array) 824 (\%hash) 825 (\&sub) 826 \(&sub) 827 \($foo, @bar, %baz) 828 (\$foo, \@bar, \%baz) 829 830Each element on the right-hand side must be a reference to a datum of the 831right type. Parentheses immediately surrounding an array (and possibly 832also C<my>/C<state>/C<our>/C<local>) will make each element of the array an 833alias to the corresponding scalar referenced on the right-hand side: 834 835 \(@a) = \(@b); # @a and @b now have the same elements 836 \my(@a) = \(@b); # likewise 837 \(my @a) = \(@b); # likewise 838 push @a, 3; # but now @a has an extra element that @b lacks 839 \(@a) = (\$a, \$b, \$c); # @a now contains $a, $b, and $c 840 841Combining that form with C<local> and putting parentheses immediately 842around a hash are forbidden (because it is not clear what they should do): 843 844 \local(@array) = foo(); # WRONG 845 \(%hash) = bar(); # WRONG 846 847Assignment to references and non-references may be combined in lists and 848conditional ternary expressions, as long as the values on the right-hand 849side are the right type for each element on the left, though this may make 850for obfuscated code: 851 852 (my $tom, \my $dick, \my @harry) = (\1, \2, [1..3]); 853 # $tom is now \1 854 # $dick is now 2 (read-only) 855 # @harry is (1,2,3) 856 857 my $type = ref $thingy; 858 ($type ? $type eq 'ARRAY' ? \@foo : \$bar : $baz) = $thingy; 859 860The C<foreach> loop can also take a reference constructor for its loop 861variable, though the syntax is limited to one of the following, with an 862optional C<my>, C<state>, or C<our> after the backslash: 863 864 \$s 865 \@a 866 \%h 867 \&c 868 869No parentheses are permitted. This feature is particularly useful for 870arrays-of-arrays, or arrays-of-hashes: 871 872 foreach \my @a (@array_of_arrays) { 873 frobnicate($a[0], $a[-1]); 874 } 875 876 foreach \my %h (@array_of_hashes) { 877 $h{gelastic}++ if $h{type} eq 'funny'; 878 } 879 880B<CAVEAT:> Aliasing does not work correctly with closures. If you try to 881alias lexical variables from an inner subroutine or C<eval>, the aliasing 882will only be visible within that inner sub, and will not affect the outer 883subroutine where the variables are declared. This bizarre behavior is 884subject to change. 885 886=head2 Declaring a Reference to a Variable 887 888Beginning in v5.26.0, the referencing operator can come after C<my>, 889C<state>, C<our>, or C<local>. This syntax must be enabled with C<use 890feature 'declared_refs'>. It is experimental, and will warn by default 891unless C<no warnings 'experimental::refaliasing'> is in effect. 892 893This feature makes these: 894 895 my \$x; 896 our \$y; 897 898equivalent to: 899 900 \my $x; 901 \our $x; 902 903It is intended mainly for use in assignments to references (see 904L</Assigning to References>, above). It also allows the backslash to be 905used on just some items in a list of declared variables: 906 907 my ($foo, \@bar, \%baz); # equivalent to: my $foo, \my(@bar, %baz); 908 909=head1 WARNING: Don't use references as hash keys 910X<reference, string context> X<reference, use as hash key> 911 912You may not (usefully) use a reference as the key to a hash. It will be 913converted into a string: 914 915 $x{ \$a } = $a; 916 917If you try to dereference the key, it won't do a hard dereference, and 918you won't accomplish what you're attempting. You might want to do something 919more like 920 921 $r = \@a; 922 $x{ $r } = $r; 923 924And then at least you can use the values(), which will be 925real refs, instead of the keys(), which won't. 926 927The standard Tie::RefHash module provides a convenient workaround to this. 928 929=head1 SEE ALSO 930 931Besides the obvious documents, source code can be instructive. 932Some pathological examples of the use of references can be found 933in the F<t/op/ref.t> regression test in the Perl source directory. 934 935See also L<perldsc> and L<perllol> for how to use references to create 936complex data structures, and L<perlootut> and L<perlobj> 937for how to use them to create objects. 938