6 with_assign => "+ - * / % ** << >> x .",
7 assign => "+= -= *= /= %= **= <<= >>= x= .=",
8 num_comparison => "< <= > >= == !=",
9 '3way_comparison' => "<=> cmp",
10 str_comparison => "lt le gt ge eq ne",
11 binary => '& &= | |= ^ ^= &. &.= |. |.= ^. ^.=',
12 unary => "neg ! ~ ~.",
14 func => "atan2 cos sin exp abs log sqrt int",
15 conversion => 'bool "" 0+ qr',
18 dereferencing => '${} @{} %{} &{} *{}',
20 special => 'nomethod fallback =',
24 for $category (keys %ops) {
25 $ops_seen{$_}++ for (split /\s+/, $ops{$category});
34 *{$package . "::(("} = \&nil; # Make it findable via fetchmethod.
36 if ($_ eq 'fallback') {
37 for my $sym (*{$package . "::()"}) {
38 *$sym = \&nil; # Make it findable via fetchmethod.
42 warnings::warnif("overload arg '$_' is invalid")
46 $ {$package . "::(" . $_} = $sub;
49 #print STDERR "Setting '$ {'package'}::\cO$_' to \\&'$sub'.\n";
50 *{$package . "::(" . $_} = \&{ $sub };
56 $package = (caller())[0];
57 # *{$package . "::OVERLOAD"} = \&OVERLOAD;
59 $package->overload::OVERLOAD(@_);
63 $package = (caller())[0];
65 *{$package . "::(("} = \&nil;
67 warnings::warnif("overload arg '$_' is invalid")
69 delete $ {$package . "::"}{$_ eq 'fallback' ? '()' : "(" .$_};
75 $package = ref $package if ref $package;
76 mycan ($package, '()') || mycan ($package, '((');
81 return undef unless $globref;
82 my $sub = \&{*$globref};
84 return $sub if $sub != \&nil;
85 return shift->can($ {*$globref});
88 sub OverloadedStringify {
90 $package = ref $package if ref $package;
92 ov_method mycan($package, '(""'), $package
93 or ov_method mycan($package, '(0+'), $package
94 or ov_method mycan($package, '(bool'), $package
95 or ov_method mycan($package, '(nomethod'), $package;
103 require Scalar::Util;
104 $package = Scalar::Util::blessed($package);
105 return undef if !defined $package;
107 #my $meth = $package->can('(' . shift);
108 ov_method mycan($package, '(' . shift), $package;
109 #return $meth if $meth ne \&nil;
110 #return $ {*{$meth}};
120 sub mycan { # Real can would leave stubs.
121 my ($package, $meth) = @_;
127 my $mro = mro::get_linear_isa($package);
128 foreach my $p (@$mro) {
129 my $fqmeth = $p . q{::} . $meth;
130 return \*{$fqmeth} if defined &{$fqmeth};
137 'integer' => 0x1000, # HINT_NEW_INTEGER
138 'float' => 0x2000, # HINT_NEW_FLOAT
139 'binary' => 0x4000, # HINT_NEW_BINARY
140 'q' => 0x8000, # HINT_NEW_STRING
141 'qr' => 0x10000, # HINT_NEW_RE
144 use warnings::register;
146 # Arguments: what, sub
149 warnings::warnif ("Odd number of arguments for overload::constant");
152 elsif (!exists $constants {$_ [0]}) {
153 warnings::warnif ("'$_[0]' is not an overloadable type");
155 elsif (!ref $_ [1] || "$_[1]" !~ /(^|=)CODE\(0x[0-9a-f]+\)$/) {
156 # Can't use C<ref $_[1] eq "CODE"> above as code references can be
157 # blessed, and C<ref> would return the package the ref is blessed into.
158 if (warnings::enabled) {
159 $_ [1] = "undef" unless defined $_ [1];
160 warnings::warn ("'$_[1]' is not a code reference");
165 $^H |= $constants{$_[0]};
171 sub remove_constant {
172 # Arguments: what, sub
175 $^H &= ~ $constants{$_[0]};
186 overload - Package for overloading Perl operations
199 $a = SomeThing->new( 57 );
202 if (overload::Overloaded $b) {...}
204 $strval = overload::StrVal $b;
208 This pragma allows overloading of Perl's operators for a class.
209 To overload built-in functions, see L<perlsub/Overriding Built-in Functions> instead.
215 Arguments of the C<use overload> directive are (key, value) pairs.
216 For the full set of legal keys, see L<Overloadable Operations> below.
218 Operator implementations (the values) can be subroutines,
219 references to subroutines, or anonymous subroutines
220 - in other words, anything legal inside a C<&{ ... }> call.
221 Values specified as strings are interpreted as method names.
228 '""' => sub { ...; };
230 declares that subtraction is to be implemented by method C<minus()>
231 in the class C<Number> (or one of its base classes),
232 and that the function C<Number::muas()> is to be used for the
233 assignment form of multiplication, C<*=>.
234 It also defines an anonymous subroutine to implement stringification:
235 this is called whenever an object blessed into the package C<Number>
236 is used in a string context (this subroutine might, for example,
237 return the number as a Roman numeral).
239 =head3 Calling Conventions and Magic Autogeneration
241 The following sample implementation of C<minus()> (which assumes
242 that C<Number> objects are simply blessed references to scalars)
243 illustrates the calling conventions:
247 my ($self, $other, $swap) = @_;
248 my $result = $$self - $other; # *
249 $result = -$result if $swap;
250 ref $result ? $result : bless \$result;
252 # * may recurse once - see table below
254 Three arguments are passed to all subroutines specified in the
255 C<use overload> directive (with exceptions - see below, particularly
258 The first of these is the operand providing the overloaded
259 operator implementation -
260 in this case, the object whose C<minus()> method is being called.
262 The second argument is the other operand, or C<undef> in the
263 case of a unary operator.
265 The third argument is set to TRUE if (and only if) the two
266 operands have been swapped. Perl may do this to ensure that the
267 first argument (C<$self>) is an object implementing the overloaded
268 operation, in line with general object calling conventions.
269 For example, if C<$x> and C<$y> are C<Number>s:
271 operation | generates a call to
272 ============|======================
273 $x - $y | minus($x, $y, '')
274 $x - 7 | minus($x, 7, '')
275 7 - $x | minus($x, 7, 1)
277 Perl may also use C<minus()> to implement other operators which
278 have not been specified in the C<use overload> directive,
279 according to the rules for L<Magic Autogeneration> described later.
280 For example, the C<use overload> above declared no subroutine
281 for any of the operators C<-->, C<neg> (the overload key for
282 unary minus), or C<-=>. Thus
284 operation | generates a call to
285 ============|======================
286 -$x | minus($x, 0, 1)
287 $x-- | minus($x, 1, undef)
288 $x -= 3 | minus($x, 3, undef)
291 where autogeneration results in the method for a standard
292 operator which does not change either of its operands, such
293 as C<->, being used to implement an operator which changes
294 the operand ("mutators": here, C<--> and C<-=>),
295 Perl passes undef as the third argument.
296 This still evaluates as FALSE, consistent with the fact that
297 the operands have not been swapped, but gives the subroutine
298 a chance to alter its behaviour in these cases.
300 In all the above examples, C<minus()> is required
301 only to return the result of the subtraction:
302 Perl takes care of the assignment to $x.
303 In fact, such methods should I<not> modify their operands,
304 even if C<undef> is passed as the third argument
305 (see L<Overloadable Operations>).
307 The same is not true of implementations of C<++> and C<-->:
308 these are expected to modify their operand.
309 An appropriate implementation of C<--> might look like
311 use overload '--' => "decr",
313 sub decr { --${$_[0]}; }
315 If the experimental "bitwise" feature is enabled (see L<feature>), a fifth
316 TRUE argument is passed to subroutines handling C<&>, C<|>, C<^> and C<~>.
317 This indicates that the caller is expecting numeric behaviour. The fourth
318 argument will be C<undef>, as that position (C<$_[3]>) is reserved for use
321 =head3 Mathemagic, Mutators, and Copy Constructors
323 The term 'mathemagic' describes the overloaded implementation
324 of mathematical operators.
325 Mathemagical operations raise an issue.
331 If C<$a> and C<$b> are scalars then after these statements
335 An object, however, is a reference to blessed data, so if
336 C<$a> and C<$b> are objects then the assignment C<$a = $b>
337 copies only the reference, leaving C<$a> and C<$b> referring
338 to the same object data.
339 One might therefore expect the operation C<--$a> to decrement
340 C<$b> as well as C<$a>.
341 However, this would not be consistent with how we expect the
342 mathematical operators to work.
344 Perl resolves this dilemma by transparently calling a copy
345 constructor before calling a method defined to implement
346 a mutator (C<-->, C<+=>, and so on.).
347 In the above example, when Perl reaches the decrement
348 statement, it makes a copy of the object data in C<$a> and
349 assigns to C<$a> a reference to the copied data.
350 Only then does it call C<decr()>, which alters the copied
351 data, leaving C<$b> unchanged.
352 Thus the object metaphor is preserved as far as possible,
353 while mathemagical operations still work according to the
356 Note: the preceding paragraph describes what happens when
357 Perl autogenerates the copy constructor for an object based
359 For other cases, see L<Copy Constructor>.
361 =head2 Overloadable Operations
363 The complete list of keys that can be specified in the C<use overload>
364 directive are given, separated by spaces, in the values of the
365 hash C<%overload::ops>:
367 with_assign => '+ - * / % ** << >> x .',
368 assign => '+= -= *= /= %= **= <<= >>= x= .=',
369 num_comparison => '< <= > >= == !=',
370 '3way_comparison'=> '<=> cmp',
371 str_comparison => 'lt le gt ge eq ne',
372 binary => '& &= | |= ^ ^= &. &.= |. |.= ^. ^.=',
373 unary => 'neg ! ~ ~.',
375 func => 'atan2 cos sin exp abs log sqrt int',
376 conversion => 'bool "" 0+ qr',
379 dereferencing => '${} @{} %{} &{} *{}',
381 special => 'nomethod fallback ='
383 Most of the overloadable operators map one-to-one to these keys.
384 Exceptions, including additional overloadable operations not
385 apparent from this hash, are included in the notes which follow.
387 A warning is issued if an attempt is made to register an operator not found
394 The operator C<not> is not a valid key for C<use overload>.
395 However, if the operator C<!> is overloaded then the same
396 implementation will be used for C<not>
397 (since the two operators differ only in precedence).
401 The key C<neg> is used for unary minus to disambiguate it from
406 Assuming they are to behave analogously to Perl's C<++> and C<-->,
407 overloaded implementations of these operators are required to
408 mutate their operands.
410 No distinction is made between prefix and postfix forms of the
411 increment and decrement operators: these differ only in the
412 point at which Perl calls the associated subroutine when
413 evaluating an expression.
415 =item * I<Assignments>
417 += -= *= /= %= **= <<= >>= x= .=
420 Simple assignment is not overloadable (the C<'='> key is used
421 for the L<Copy Constructor>).
422 Perl does have a way to make assignments to an object do whatever
423 you want, but this involves using tie(), not overload -
424 see L<perlfunc/tie> and the L</COOKBOOK> examples below.
426 The subroutine for the assignment variant of an operator is
427 required only to return the result of the operation.
428 It is permitted to change the value of its operand
429 (this is safe because Perl calls the copy constructor first),
430 but this is optional since Perl assigns the returned value to
431 the left-hand operand anyway.
433 An object that overloads an assignment operator does so only in
434 respect of assignments to that object.
435 In other words, Perl never calls the corresponding methods with
436 the third argument (the "swap" argument) set to TRUE.
437 For example, the operation
441 cannot lead to C<$b>'s implementation of C<*=> being called,
442 even if C<$a> is a scalar.
443 (It can, however, generate a call to C<$b>'s method for C<*>).
445 =item * I<Non-mutators with a mutator variant>
447 + - * / % ** << >> x .
450 As described L<above|"Calling Conventions and Magic Autogeneration">,
451 Perl may call methods for operators like C<+> and C<&> in the course
452 of implementing missing operations like C<++>, C<+=>, and C<&=>.
453 While these methods may detect this usage by testing the definedness
454 of the third argument, they should in all cases avoid changing their
456 This is because Perl does not call the copy constructor before
457 invoking these methods.
461 Traditionally, the Perl function C<int> rounds to 0
462 (see L<perlfunc/int>), and so for floating-point-like types one
463 should follow the same semantic.
465 =item * I<String, numeric, boolean, and regexp conversions>
469 These conversions are invoked according to context as necessary.
470 For example, the subroutine for C<'""'> (stringify) may be used
471 where the overloaded object is passed as an argument to C<print>,
472 and that for C<'bool'> where it is tested in the condition of a flow
473 control statement (like C<while>) or the ternary C<?:> operation.
475 Of course, in contexts like, for example, C<$obj + 1>, Perl will
476 invoke C<$obj>'s implementation of C<+> rather than (in this
477 example) converting C<$obj> to a number using the numify method
478 C<'0+'> (an exception to this is when no method has been provided
479 for C<'+'> and L</fallback> is set to TRUE).
481 The subroutines for C<'""'>, C<'0+'>, and C<'bool'> can return
482 any arbitrary Perl value.
483 If the corresponding operation for this value is overloaded too,
484 the operation will be called again with this value.
486 As a special case if the overload returns the object itself then it will
487 be used directly. An overloaded conversion returning the object is
488 probably a bug, because you're likely to get something that looks like
489 C<YourPackage=HASH(0x8172b34)>.
493 The subroutine for C<'qr'> is used wherever the object is
494 interpolated into or used as a regexp, including when it
495 appears on the RHS of a C<=~> or C<!~> operator.
497 C<qr> must return a compiled regexp, or a ref to a compiled regexp
498 (such as C<qr//> returns), and any further overloading on the return
499 value will be ignored.
503 If C<E<lt>E<gt>> is overloaded then the same implementation is used
504 for both the I<read-filehandle> syntax C<E<lt>$varE<gt>> and
505 I<globbing> syntax C<E<lt>${var}E<gt>>.
507 =item * I<File tests>
509 The key C<'-X'> is used to specify a subroutine to handle all the
510 filetest operators (C<-f>, C<-x>, and so on: see L<perlfunc/-X> for
512 it is not possible to overload any filetest operator individually.
513 To distinguish them, the letter following the '-' is passed as the
514 second argument (that is, in the slot that for binary operators
515 is used to pass the second operand).
517 Calling an overloaded filetest operator does not affect the stat value
518 associated with the special filehandle C<_>. It still refers to the
519 result of the last C<stat>, C<lstat> or unoverloaded filetest.
521 This overload was introduced in Perl 5.12.
525 The key C<"~~"> allows you to override the smart matching logic used by
526 the C<~~> operator and the switch construct (C<given>/C<when>). See
527 L<perlsyn/Switch Statements> and L<feature>.
529 Unusually, the overloaded implementation of the smart match operator
530 does not get full control of the smart match behaviour.
531 In particular, in the following code:
534 use overload '~~' => 'match';
536 my $obj = Foo->new();
539 the smart match does I<not> invoke the method call like this:
541 $obj->match([1,2,3],0);
543 rather, the smart match distributive rule takes precedence, so $obj is
544 smart matched against each array element in turn until a match is found,
545 so you may see between one and three of these calls instead:
551 Consult the match table in L<perlop/"Smartmatch Operator"> for
552 details of when overloading is invoked.
554 =item * I<Dereferencing>
558 If these operators are not explicitly overloaded then they
559 work in the normal way, yielding the underlying scalar,
560 array, or whatever stores the object data (or the appropriate
561 error message if the dereference operator doesn't match it).
562 Defining a catch-all C<'nomethod'> (see L<below|/nomethod>)
563 makes no difference to this as the catch-all function will
564 not be called to implement a missing dereference operator.
566 If a dereference operator is overloaded then it must return a
567 I<reference> of the appropriate type (for example, the
568 subroutine for key C<'${}'> should return a reference to a
569 scalar, not a scalar), or another object which overloads the
570 operator: that is, the subroutine only determines what is
571 dereferenced and the actual dereferencing is left to Perl.
572 As a special case, if the subroutine returns the object itself
573 then it will not be called again - avoiding infinite recursion.
579 See L<Special Keys for C<use overload>>.
583 =head2 Magic Autogeneration
585 If a method for an operation is not found then Perl tries to
586 autogenerate a substitute implementation from the operations
587 that have been defined.
589 Note: the behaviour described in this section can be disabled
590 by setting C<fallback> to FALSE (see L</fallback>).
592 In the following tables, numbers indicate priority.
593 For example, the table below states that,
594 if no implementation for C<'!'> has been defined then Perl will
595 implement it using C<'bool'> (that is, by inverting the value
596 returned by the method for C<'bool'>);
597 if boolean conversion is also unimplemented then Perl will
598 use C<'0+'> or, failing that, C<'""'>.
600 operator | can be autogenerated from
603 =========|==========================
617 Note: The iterator (C<'E<lt>E<gt>'>) and file test (C<'-X'>)
618 operators work as normal: if the operand is not a blessed glob or
619 IO reference then it is converted to a string (using the method
620 for C<'""'>, C<'0+'>, or C<'bool'>) to be interpreted as a glob
623 operator | can be autogenerated from
626 =========|==========================
630 abs | a1 a2 b1 b2 [*]
638 * one from [a1, a2] and one from [b1, b2]
640 Just as numeric comparisons can be autogenerated from the method
641 for C<< '<=>' >>, string comparisons can be autogenerated from
644 operators | can be autogenerated from
645 ====================|===========================
646 lt gt le ge eq ne | cmp
648 Similarly, autogeneration for keys C<'+='> and C<'++'> is analogous
649 to C<'-='> and C<'--'> above:
651 operator | can be autogenerated from
654 =========|==========================
658 And other assignment variations are analogous to
659 C<'+='> and C<'-='> (and similar to C<'.='> and C<'x='> above):
661 operator || *= /= %= **= <<= >>= &= ^= |= &.= ^.= |.=
662 -------------------||-------------------------------------------
663 autogenerated from || * / % ** << >> & ^ | &. ^. |.
665 Note also that the copy constructor (key C<'='>) may be
666 autogenerated, but only for objects based on scalars.
667 See L<Copy Constructor>.
669 =head3 Minimal Set of Overloaded Operations
671 Since some operations can be automatically generated from others, there is
672 a minimal set of operations that need to be overloaded in order to have
673 the complete set of overloaded operations at one's disposal.
674 Of course, the autogenerated operations may not do exactly what the user
675 expects. The minimal set is:
680 atan2 cos sin exp log sqrt int
684 Of the conversions, only one of string, boolean or numeric is
685 needed because each can be generated from either of the other two.
687 =head2 Special Keys for C<use overload>
691 The C<'nomethod'> key is used to specify a catch-all function to
692 be called for any operator that is not individually overloaded.
693 The specified function will be passed four parameters.
694 The first three arguments coincide with those that would have been
695 passed to the corresponding method if it had been defined.
696 The fourth argument is the C<use overload> key for that missing
697 method. If the experimental "bitwise" feature is enabled (see L<feature>),
698 a fifth TRUE argument is passed to subroutines handling C<&>, C<|>, C<^> and C<~> to indicate that the caller is expecting numeric behaviour.
700 For example, if C<$a> is an object blessed into a package declaring
702 use overload 'nomethod' => 'catch_all', # ...
708 could (unless a method is specifically declared for the key
709 C<'+'>) result in a call
711 catch_all($a, 3, 1, '+')
713 See L<How Perl Chooses an Operator Implementation>.
717 The value assigned to the key C<'fallback'> tells Perl how hard
718 it should try to find an alternative way to implement a missing
723 =item * defined, but FALSE
725 use overload "fallback" => 0, # ... ;
727 This disables L<Magic Autogeneration>.
731 In the default case where no value is explicitly assigned to
732 C<fallback>, magic autogeneration is enabled.
736 The same as for C<undef>, but if a missing operator cannot be
737 autogenerated then, instead of issuing an error message, Perl
738 is allowed to revert to what it would have done for that
739 operator if there had been no C<use overload> directive.
741 Note: in most cases, particularly the L<Copy Constructor>,
742 this is unlikely to be appropriate behaviour.
746 See L<How Perl Chooses an Operator Implementation>.
748 =head3 Copy Constructor
750 As mentioned L<above|"Mathemagic, Mutators, and Copy Constructors">,
751 this operation is called when a mutator is applied to a reference
752 that shares its object with some other reference.
753 For example, if C<$b> is mathemagical, and C<'++'> is overloaded
754 with C<'incr'>, and C<'='> is overloaded with C<'clone'>, then the
758 # ... (other code which does not modify $a or $b) ...
761 would be executed in a manner equivalent to
765 $b = $b->clone(undef, "");
774 The subroutine for C<'='> does not overload the Perl assignment
775 operator: it is used only to allow mutators to work as described
776 here. (See L</Assignments> above.)
780 As for other operations, the subroutine implementing '=' is passed
781 three arguments, though the last two are always C<undef> and C<''>.
785 The copy constructor is called only before a call to a function
786 declared to implement a mutator, for example, if C<++$b;> in the
787 code above is effected via a method declared for key C<'++'>
788 (or 'nomethod', passed C<'++'> as the fourth argument) or, by
789 autogeneration, C<'+='>.
790 It is not called if the increment operation is effected by a call
791 to the method for C<'+'> since, in the equivalent code,
796 the data referred to by C<$a> is unchanged by the assignment to
797 C<$b> of a reference to new object data.
801 The copy constructor is not called if Perl determines that it is
802 unnecessary because there is no other reference to the data being
807 If C<'fallback'> is undefined or TRUE then a copy constructor
808 can be autogenerated, but only for objects based on scalars.
809 In other cases it needs to be defined explicitly.
810 Where an object's data is stored as, for example, an array of
811 scalars, the following might be appropriate:
813 use overload '=' => sub { bless [ @{$_[0]} ] }, # ...
817 If C<'fallback'> is TRUE and no copy constructor is defined then,
818 for objects not based on scalars, Perl may silently fall back on
819 simple assignment - that is, assignment of the object reference.
820 In effect, this disables the copy constructor mechanism since
821 no new copy of the object data is created.
822 This is almost certainly not what you want.
823 (It is, however, consistent: for example, Perl's fallback for the
824 C<++> operator is to increment the reference itself.)
828 =head2 How Perl Chooses an Operator Implementation
830 Which is checked first, C<nomethod> or C<fallback>?
831 If the two operands of an operator are of different types and
832 both overload the operator, which implementation is used?
833 The following are the precedence rules:
839 If the first operand has declared a subroutine to overload the
840 operator then use that implementation.
844 Otherwise, if fallback is TRUE or undefined for the
845 first operand then see if the
846 L<rules for autogeneration|"Magic Autogeneration">
847 allows another of its operators to be used instead.
851 Unless the operator is an assignment (C<+=>, C<-=>, etc.),
852 repeat step (1) in respect of the second operand.
856 Repeat Step (2) in respect of the second operand.
860 If the first operand has a "nomethod" method then use that.
864 If the second operand has a "nomethod" method then use that.
868 If C<fallback> is TRUE for both operands
869 then perform the usual operation for the operator,
870 treating the operands as numbers, strings, or booleans
871 as appropriate for the operator (see note).
875 Nothing worked - die.
879 Where there is only one operand (or only one operand with
880 overloading) the checks in respect of the other operand above are
883 There are exceptions to the above rules for dereference operations
884 (which, if Step 1 fails, always fall back to the normal, built-in
885 implementations - see Dereferencing), and for C<~~> (which has its
886 own set of rules - see C<Matching> under L</Overloadable Operations>
889 Note on Step 7: some operators have a different semantic depending
890 on the type of their operands.
891 As there is no way to instruct Perl to treat the operands as, e.g.,
892 numbers instead of strings, the result here may not be what you
894 See L<BUGS AND PITFALLS>.
896 =head2 Losing Overloading
898 The restriction for the comparison operation is that even if, for example,
899 C<cmp> should return a blessed reference, the autogenerated C<lt>
900 function will produce only a standard logical value based on the
901 numerical value of the result of C<cmp>. In particular, a working
902 numeric conversion is needed in this case (possibly expressed in terms of
905 Similarly, C<.=> and C<x=> operators lose their mathemagical properties
906 if the string conversion substitution is applied.
908 When you chop() a mathemagical object it is promoted to a string and its
909 mathemagical properties are lost. The same can happen with other
912 =head2 Inheritance and Overloading
914 Overloading respects inheritance via the @ISA hierarchy.
915 Inheritance interacts with overloading in two ways.
919 =item Method names in the C<use overload> directive
923 use overload key => value;
925 is a string, it is interpreted as a method name - which may
926 (in the usual way) be inherited from another class.
928 =item Overloading of an operation is inherited by derived classes
930 Any class derived from an overloaded class is also overloaded
931 and inherits its operator implementations.
932 If the same operator is overloaded in more than one ancestor
933 then the implementation is determined by the usual inheritance
936 For example, if C<A> inherits from C<B> and C<C> (in that order),
937 C<B> overloads C<+> with C<\&D::plus_sub>, and C<C> overloads
938 C<+> by C<"plus_meth">, then the subroutine C<D::plus_sub> will
939 be called to implement operation C<+> for an object in package C<A>.
943 Note that in Perl version prior to 5.18 inheritance of the C<fallback> key
944 was not governed by the above rules. The value of C<fallback> in the first
945 overloaded ancestor was used. This was fixed in 5.18 to follow the usual
946 rules of inheritance.
948 =head2 Run-time Overloading
950 Since all C<use> directives are executed at compile-time, the only way to
951 change overloading during run-time is to
953 eval 'use overload "+" => \&addmethod';
957 eval 'no overload "+", "--", "<="';
959 though the use of these constructs during run-time is questionable.
961 =head2 Public Functions
963 Package C<overload.pm> provides the following public functions:
967 =item overload::StrVal(arg)
969 Gives the string value of C<arg> as in the
970 absence of stringify overloading. If you
971 are using this to get the address of a reference (useful for checking if two
972 references point to the same thing) then you may be better off using
973 C<Scalar::Util::refaddr()>, which is faster.
975 =item overload::Overloaded(arg)
977 Returns true if C<arg> is subject to overloading of some operations.
979 =item overload::Method(obj,op)
981 Returns C<undef> or a reference to the method that implements C<op>.
985 =head2 Overloading Constants
987 For some applications, the Perl parser mangles constants too much.
988 It is possible to hook into this process via C<overload::constant()>
989 and C<overload::remove_constant()> functions.
991 These functions take a hash as an argument. The recognized keys of this hash
998 to overload integer constants,
1002 to overload floating point constants,
1006 to overload octal and hexadecimal constants,
1010 to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted
1011 strings and here-documents,
1015 to overload constant pieces of regular expressions.
1019 The corresponding values are references to functions which take three arguments:
1020 the first one is the I<initial> string form of the constant, the second one
1021 is how Perl interprets this constant, the third one is how the constant is used.
1022 Note that the initial string form does not
1023 contain string delimiters, and has backslashes in backslash-delimiter
1024 combinations stripped (thus the value of delimiter is not relevant for
1025 processing of this string). The return value of this function is how this
1026 constant is going to be interpreted by Perl. The third argument is undefined
1027 unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote
1028 context (comes from strings, regular expressions, and single-quote HERE
1029 documents), it is C<tr> for arguments of C<tr>/C<y> operators,
1030 it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise.
1032 Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>,
1033 it is expected that overloaded constant strings are equipped with reasonable
1034 overloaded catenation operator, otherwise absurd results will result.
1035 Similarly, negative numbers are considered as negations of positive constants.
1037 Note that it is probably meaningless to call the functions overload::constant()
1038 and overload::remove_constant() from anywhere but import() and unimport() methods.
1039 From these methods they may be called as
1044 die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant';
1045 overload::constant integer => sub {Math::BigInt->new(shift)};
1048 =head1 IMPLEMENTATION
1050 What follows is subject to change RSN.
1052 The table of methods for all operations is cached in magic for the
1053 symbol table hash for the package. The cache is invalidated during
1054 processing of C<use overload>, C<no overload>, new function
1055 definitions, and changes in @ISA.
1057 (Every SVish thing has a magic queue, and magic is an entry in that
1058 queue. This is how a single variable may participate in multiple
1059 forms of magic simultaneously. For instance, environment variables
1060 regularly have two forms at once: their %ENV magic and their taint
1061 magic. However, the magic which implements overloading is applied to
1062 the stashes, which are rarely used directly, thus should not slow down
1065 If a package uses overload, it carries a special flag. This flag is also
1066 set when new functions are defined or @ISA is modified. There will be a
1067 slight speed penalty on the very first operation thereafter that supports
1068 overloading, while the overload tables are updated. If there is no
1069 overloading present, the flag is turned off. Thus the only speed penalty
1070 thereafter is the checking of this flag.
1072 It is expected that arguments to methods that are not explicitly supposed
1073 to be changed are constant (but this is not enforced).
1077 Please add examples to what follows!
1079 =head2 Two-face Scalars
1081 Put this in F<two_face.pm> in your Perl library directory:
1083 package two_face; # Scalars with separate string and
1085 sub new { my $p = shift; bless [@_], $p }
1086 use overload '""' => \&str, '0+' => \&num, fallback => 1;
1087 sub num {shift->[1]}
1088 sub str {shift->[0]}
1093 my $seven = two_face->new("vii", 7);
1094 printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1;
1095 print "seven contains 'i'\n" if $seven =~ /i/;
1097 (The second line creates a scalar which has both a string value, and a
1098 numeric value.) This prints:
1100 seven=vii, seven=7, eight=8
1103 =head2 Two-face References
1105 Suppose you want to create an object which is accessible as both an
1106 array reference and a hash reference.
1109 use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} };
1117 tie %h, ref $self, $self;
1121 sub TIEHASH { my $p = shift; bless \ shift, $p }
1124 $fields{$_} = $i++ foreach qw{zero one two three};
1126 my $self = ${shift()};
1127 my $key = $fields{shift()};
1128 defined $key or die "Out of band access";
1129 $$self->[$key] = shift;
1132 my $self = ${shift()};
1133 my $key = $fields{shift()};
1134 defined $key or die "Out of band access";
1138 Now one can access an object using both the array and hash syntax:
1140 my $bar = two_refs->new(3,4,5,6);
1142 $bar->{two} == 11 or die 'bad hash fetch';
1144 Note several important features of this example. First of all, the
1145 I<actual> type of $bar is a scalar reference, and we do not overload
1146 the scalar dereference. Thus we can get the I<actual> non-overloaded
1147 contents of $bar by just using C<$$bar> (what we do in functions which
1148 overload dereference). Similarly, the object returned by the
1149 TIEHASH() method is a scalar reference.
1151 Second, we create a new tied hash each time the hash syntax is used.
1152 This allows us not to worry about a possibility of a reference loop,
1153 which would lead to a memory leak.
1155 Both these problems can be cured. Say, if we want to overload hash
1156 dereference on a reference to an object which is I<implemented> as a
1157 hash itself, the only problem one has to circumvent is how to access
1158 this I<actual> hash (as opposed to the I<virtual> hash exhibited by the
1159 overloaded dereference operator). Here is one possible fetching routine:
1162 my ($self, $key) = (shift, shift);
1163 my $class = ref $self;
1164 bless $self, 'overload::dummy'; # Disable overloading of %{}
1165 my $out = $self->{$key};
1166 bless $self, $class; # Restore overloading
1170 To remove creation of the tied hash on each access, one may an extra
1171 level of indirection which allows a non-circular structure of references:
1174 use overload '%{}' => sub { ${shift()}->[1] },
1175 '@{}' => sub { ${shift()}->[0] };
1181 bless \ [$a, \%h], $p;
1186 tie %h, ref $self, $self;
1190 sub TIEHASH { my $p = shift; bless \ shift, $p }
1193 $fields{$_} = $i++ foreach qw{zero one two three};
1196 my $key = $fields{shift()};
1197 defined $key or die "Out of band access";
1202 my $key = $fields{shift()};
1203 defined $key or die "Out of band access";
1207 Now if $baz is overloaded like this, then C<$baz> is a reference to a
1208 reference to the intermediate array, which keeps a reference to an
1209 actual array, and the access hash. The tie()ing object for the access
1210 hash is a reference to a reference to the actual array, so
1216 There are no loops of references.
1220 Both "objects" which are blessed into the class C<two_refs1> are
1221 references to a reference to an array, thus references to a I<scalar>.
1222 Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no
1223 overloaded operations.
1227 =head2 Symbolic Calculator
1229 Put this in F<symbolic.pm> in your Perl library directory:
1231 package symbolic; # Primitive symbolic calculator
1232 use overload nomethod => \&wrap;
1234 sub new { shift; bless ['n', @_] }
1236 my ($obj, $other, $inv, $meth) = @_;
1237 ($obj, $other) = ($other, $obj) if $inv;
1238 bless [$meth, $obj, $other];
1241 This module is very unusual as overloaded modules go: it does not
1242 provide any usual overloaded operators, instead it provides an
1243 implementation for L</C<nomethod>>. In this example the C<nomethod>
1244 subroutine returns an object which encapsulates operations done over
1245 the objects: C<< symbolic->new(3) >> contains C<['n', 3]>, C<< 2 +
1246 symbolic->new(3) >> contains C<['+', 2, ['n', 3]]>.
1248 Here is an example of the script which "calculates" the side of
1249 circumscribed octagon using the above package:
1252 my $iter = 1; # 2**($iter+2) = 8
1253 my $side = symbolic->new(1);
1257 $side = (sqrt(1 + $side**2) - 1)/$side;
1261 The value of $side is
1263 ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]],
1264 undef], 1], ['n', 1]]
1266 Note that while we obtained this value using a nice little script,
1267 there is no simple way to I<use> this value. In fact this value may
1268 be inspected in debugger (see L<perldebug>), but only if
1269 C<bareStringify> B<O>ption is set, and not via C<p> command.
1271 If one attempts to print this value, then the overloaded operator
1272 C<""> will be called, which will call C<nomethod> operator. The
1273 result of this operator will be stringified again, but this result is
1274 again of type C<symbolic>, which will lead to an infinite loop.
1276 Add a pretty-printer method to the module F<symbolic.pm>:
1279 my ($meth, $a, $b) = @{+shift};
1280 $a = 'u' unless defined $a;
1281 $b = 'u' unless defined $b;
1282 $a = $a->pretty if ref $a;
1283 $b = $b->pretty if ref $b;
1287 Now one can finish the script by
1289 print "side = ", $side->pretty, "\n";
1291 The method C<pretty> is doing object-to-string conversion, so it
1292 is natural to overload the operator C<""> using this method. However,
1293 inside such a method it is not necessary to pretty-print the
1294 I<components> $a and $b of an object. In the above subroutine
1295 C<"[$meth $a $b]"> is a catenation of some strings and components $a
1296 and $b. If these components use overloading, the catenation operator
1297 will look for an overloaded operator C<.>; if not present, it will
1298 look for an overloaded operator C<"">. Thus it is enough to use
1300 use overload nomethod => \&wrap, '""' => \&str;
1302 my ($meth, $a, $b) = @{+shift};
1303 $a = 'u' unless defined $a;
1304 $b = 'u' unless defined $b;
1308 Now one can change the last line of the script to
1310 print "side = $side\n";
1314 side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]]
1316 and one can inspect the value in debugger using all the possible
1319 Something is still amiss: consider the loop variable $cnt of the
1320 script. It was a number, not an object. We cannot make this value of
1321 type C<symbolic>, since then the loop will not terminate.
1323 Indeed, to terminate the cycle, the $cnt should become false.
1324 However, the operator C<bool> for checking falsity is overloaded (this
1325 time via overloaded C<"">), and returns a long string, thus any object
1326 of type C<symbolic> is true. To overcome this, we need a way to
1327 compare an object to 0. In fact, it is easier to write a numeric
1330 Here is the text of F<symbolic.pm> with such a routine added (and
1331 slightly modified str()):
1333 package symbolic; # Primitive symbolic calculator
1335 nomethod => \&wrap, '""' => \&str, '0+' => \#
1337 sub new { shift; bless ['n', @_] }
1339 my ($obj, $other, $inv, $meth) = @_;
1340 ($obj, $other) = ($other, $obj) if $inv;
1341 bless [$meth, $obj, $other];
1344 my ($meth, $a, $b) = @{+shift};
1345 $a = 'u' unless defined $a;
1352 my %subr = ( n => sub {$_[0]},
1353 sqrt => sub {sqrt $_[0]},
1354 '-' => sub {shift() - shift()},
1355 '+' => sub {shift() + shift()},
1356 '/' => sub {shift() / shift()},
1357 '*' => sub {shift() * shift()},
1358 '**' => sub {shift() ** shift()},
1361 my ($meth, $a, $b) = @{+shift};
1362 my $subr = $subr{$meth}
1363 or die "Do not know how to ($meth) in symbolic";
1364 $a = $a->num if ref $a eq __PACKAGE__;
1365 $b = $b->num if ref $b eq __PACKAGE__;
1369 All the work of numeric conversion is done in %subr and num(). Of
1370 course, %subr is not complete, it contains only operators used in the
1371 example below. Here is the extra-credit question: why do we need an
1372 explicit recursion in num()? (Answer is at the end of this section.)
1374 Use this module like this:
1377 my $iter = symbolic->new(2); # 16-gon
1378 my $side = symbolic->new(1);
1382 $cnt = $cnt - 1; # Mutator '--' not implemented
1383 $side = (sqrt(1 + $side**2) - 1)/$side;
1385 printf "%s=%f\n", $side, $side;
1386 printf "pi=%f\n", $side*(2**($iter+2));
1388 It prints (without so many line breaks)
1390 [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1]
1392 [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912
1395 The above module is very primitive. It does not implement
1396 mutator methods (C<++>, C<-=> and so on), does not do deep copying
1397 (not required without mutators!), and implements only those arithmetic
1398 operations which are used in the example.
1400 To implement most arithmetic operations is easy; one should just use
1401 the tables of operations, and change the code which fills %subr to
1403 my %subr = ( 'n' => sub {$_[0]} );
1404 foreach my $op (split " ", $overload::ops{with_assign}) {
1405 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1407 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1408 foreach my $op (split " ", "@overload::ops{ @bins }") {
1409 $subr{$op} = eval "sub {shift() $op shift()}";
1411 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1412 print "defining '$op'\n";
1413 $subr{$op} = eval "sub {$op shift()}";
1416 Since subroutines implementing assignment operators are not required
1417 to modify their operands (see L<Overloadable Operations> above),
1418 we do not need anything special to make C<+=> and friends work,
1419 besides adding these operators to %subr and defining a copy
1420 constructor (needed since Perl has no way to know that the
1421 implementation of C<'+='> does not mutate the argument -
1422 see L<Copy Constructor>).
1424 To implement a copy constructor, add C<< '=' => \&cpy >> to C<use overload>
1425 line, and code (this code assumes that mutators change things one level
1426 deep only, so recursive copying is not needed):
1430 bless [@$self], ref $self;
1433 To make C<++> and C<--> work, we need to implement actual mutators,
1434 either directly, or in C<nomethod>. We continue to do things inside
1435 C<nomethod>, thus add
1437 if ($meth eq '++' or $meth eq '--') {
1438 @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference
1442 after the first line of wrap(). This is not a most effective
1443 implementation, one may consider
1445 sub inc { $_[0] = bless ['++', shift, 1]; }
1449 As a final remark, note that one can fill %subr by
1451 my %subr = ( 'n' => sub {$_[0]} );
1452 foreach my $op (split " ", $overload::ops{with_assign}) {
1453 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1455 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1456 foreach my $op (split " ", "@overload::ops{ @bins }") {
1457 $subr{$op} = eval "sub {shift() $op shift()}";
1459 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1460 $subr{$op} = eval "sub {$op shift()}";
1462 $subr{'++'} = $subr{'+'};
1463 $subr{'--'} = $subr{'-'};
1465 This finishes implementation of a primitive symbolic calculator in
1466 50 lines of Perl code. Since the numeric values of subexpressions
1467 are not cached, the calculator is very slow.
1469 Here is the answer for the exercise: In the case of str(), we need no
1470 explicit recursion since the overloaded C<.>-operator will fall back
1471 to an existing overloaded operator C<"">. Overloaded arithmetic
1472 operators I<do not> fall back to numeric conversion if C<fallback> is
1473 not explicitly requested. Thus without an explicit recursion num()
1474 would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild
1475 the argument of num().
1477 If you wonder why defaults for conversion are different for str() and
1478 num(), note how easy it was to write the symbolic calculator. This
1479 simplicity is due to an appropriate choice of defaults. One extra
1480 note: due to the explicit recursion num() is more fragile than sym():
1481 we need to explicitly check for the type of $a and $b. If components
1482 $a and $b happen to be of some related type, this may lead to problems.
1484 =head2 I<Really> Symbolic Calculator
1486 One may wonder why we call the above calculator symbolic. The reason
1487 is that the actual calculation of the value of expression is postponed
1488 until the value is I<used>.
1490 To see it in action, add a method
1495 @$obj->[0,1] = ('=', shift);
1498 to the package C<symbolic>. After this change one can do
1500 my $a = symbolic->new(3);
1501 my $b = symbolic->new(4);
1502 my $c = sqrt($a**2 + $b**2);
1504 and the numeric value of $c becomes 5. However, after calling
1506 $a->STORE(12); $b->STORE(5);
1508 the numeric value of $c becomes 13. There is no doubt now that the module
1509 symbolic provides a I<symbolic> calculator indeed.
1511 To hide the rough edges under the hood, provide a tie()d interface to the
1512 package C<symbolic>. Add methods
1514 sub TIESCALAR { my $pack = shift; $pack->new(@_) }
1516 sub nop { } # Around a bug
1518 (the bug, fixed in Perl 5.14, is described in L<"BUGS">). One can use this
1521 tie $a, 'symbolic', 3;
1522 tie $b, 'symbolic', 4;
1523 $a->nop; $b->nop; # Around a bug
1525 my $c = sqrt($a**2 + $b**2);
1527 Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value
1528 of $c becomes 13. To insulate the user of the module add a method
1530 sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; }
1535 symbolic->vars($a, $b);
1536 my $c = sqrt($a**2 + $b**2);
1539 printf "c5 %s=%f\n", $c, $c;
1542 printf "c13 %s=%f\n", $c, $c;
1544 shows that the numeric value of $c follows changes to the values of $a
1549 Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>.
1553 The C<overloading> pragma can be used to enable or disable overloaded
1554 operations within a lexical scope - see L<overloading>.
1558 When Perl is run with the B<-Do> switch or its equivalent, overloading
1559 induces diagnostic messages.
1561 Using the C<m> command of Perl debugger (see L<perldebug>) one can
1562 deduce which operations are overloaded (and which ancestor triggers
1563 this overloading). Say, if C<eq> is overloaded, then the method C<(eq>
1564 is shown by debugger. The method C<()> corresponds to the C<fallback>
1565 key (in fact a presence of this method shows that this package has
1566 overloading enabled, and it is what is used by the C<Overloaded>
1567 function of module C<overload>).
1569 The module might issue the following warnings:
1573 =item Odd number of arguments for overload::constant
1575 (W) The call to overload::constant contained an odd number of arguments.
1576 The arguments should come in pairs.
1578 =item '%s' is not an overloadable type
1580 (W) You tried to overload a constant type the overload package is unaware of.
1582 =item '%s' is not a code reference
1584 (W) The second (fourth, sixth, ...) argument of overload::constant needs
1585 to be a code reference. Either an anonymous subroutine, or a reference
1588 =item overload arg '%s' is invalid
1590 (W) C<use overload> was passed an argument it did not
1591 recognize. Did you mistype an operator?
1595 =head1 BUGS AND PITFALLS
1601 A pitfall when fallback is TRUE and Perl resorts to a built-in
1602 implementation of an operator is that some operators have more
1603 than one semantic, for example C<|>:
1605 use overload '0+' => sub { $_[0]->{n}; },
1607 my $x = bless { n => 4 }, "main";
1608 my $y = bless { n => 8 }, "main";
1609 print $x | $y, "\n";
1611 You might expect this to output "12".
1612 In fact, it prints "<": the ASCII result of treating "|"
1613 as a bitwise string operator - that is, the result of treating
1614 the operands as the strings "4" and "8" rather than numbers.
1615 The fact that numify (C<0+>) is implemented but stringify
1616 (C<"">) isn't makes no difference since the latter is simply
1617 autogenerated from the former.
1619 The only way to change this is to provide your own subroutine
1624 Magic autogeneration increases the potential for inadvertently
1625 creating self-referential structures.
1626 Currently Perl will not free self-referential
1627 structures until cycles are explicitly broken.
1630 use overload '+' => 'add';
1631 sub add { bless [ \$_[0], \$_[1] ] };
1633 is asking for trouble, since
1637 will effectively become
1639 $obj = add($obj, $y, undef);
1641 with the same result as
1643 $obj = [\$obj, \$foo];
1645 Even if no I<explicit> assignment-variants of operators are present in
1646 the script, they may be generated by the optimizer.
1653 my $tmp = 'obj = ' . $obj; $tmp .= "\n";
1657 The symbol table is filled with names looking like line-noise.
1661 This bug was fixed in Perl 5.18, but may still trip you up if you are using
1664 For the purpose of inheritance every overloaded package behaves as if
1665 C<fallback> is present (possibly undefined). This may create
1666 interesting effects if some package is not overloaded, but inherits
1667 from two overloaded packages.
1671 Before Perl 5.14, the relation between overloading and tie()ing was broken.
1672 Overloading was triggered or not based on the I<previous> class of the
1675 This happened because the presence of overloading was checked
1676 too early, before any tie()d access was attempted. If the
1677 class of the value FETCH()ed from the tied variable does not
1678 change, a simple workaround for code that is to run on older Perl
1679 versions is to access the value (via C<() = $foo> or some such)
1680 immediately after tie()ing, so that after this call the I<previous> class
1681 coincides with the current one.
1685 Barewords are not covered by overloaded string constants.
1689 The range operator C<..> cannot be overloaded.