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 => '& &= | |= ^ ^=',
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 !ref $sub or $sub != \&nil;
85 return shift->can($ {*$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 one exception - see L</nomethod>).
256 The first of these is the operand providing the overloaded
257 operator implementation -
258 in this case, the object whose C<minus()> method is being called.
260 The second argument is the other operand, or C<undef> in the
261 case of a unary operator.
263 The third argument is set to TRUE if (and only if) the two
264 operands have been swapped. Perl may do this to ensure that the
265 first argument (C<$self>) is an object implementing the overloaded
266 operation, in line with general object calling conventions.
267 For example, if C<$x> and C<$y> are C<Number>s:
269 operation | generates a call to
270 ============|======================
271 $x - $y | minus($x, $y, '')
272 $x - 7 | minus($x, 7, '')
273 7 - $x | minus($x, 7, 1)
275 Perl may also use C<minus()> to implement other operators which
276 have not been specified in the C<use overload> directive,
277 according to the rules for L<Magic Autogeneration> described later.
278 For example, the C<use overload> above declared no subroutine
279 for any of the operators C<-->, C<neg> (the overload key for
280 unary minus), or C<-=>. Thus
282 operation | generates a call to
283 ============|======================
284 -$x | minus($x, 0, 1)
285 $x-- | minus($x, 1, undef)
286 $x -= 3 | minus($x, 3, undef)
289 where autogeneration results in the method for a standard
290 operator which does not change either of its operands, such
291 as C<->, being used to implement an operator which changes
292 the operand ("mutators": here, C<--> and C<-=>),
293 Perl passes undef as the third argument.
294 This still evaluates as FALSE, consistent with the fact that
295 the operands have not been swapped, but gives the subroutine
296 a chance to alter its behaviour in these cases.
298 In all the above examples, C<minus()> is required
299 only to return the result of the subtraction:
300 Perl takes care of the assignment to $x.
301 In fact, such methods should I<not> modify their operands,
302 even if C<undef> is passed as the third argument
303 (see L<Overloadable Operations>).
305 The same is not true of implementations of C<++> and C<-->:
306 these are expected to modify their operand.
307 An appropriate implementation of C<--> might look like
309 use overload '--' => "decr",
311 sub decr { --${$_[0]}; }
313 =head3 Mathemagic, Mutators, and Copy Constructors
315 The term 'mathemagic' describes the overloaded implementation
316 of mathematical operators.
317 Mathemagical operations raise an issue.
323 If C<$a> and C<$b> are scalars then after these statements
327 An object, however, is a reference to blessed data, so if
328 C<$a> and C<$b> are objects then the assignment C<$a = $b>
329 copies only the reference, leaving C<$a> and C<$b> referring
330 to the same object data.
331 One might therefore expect the operation C<--$a> to decrement
332 C<$b> as well as C<$a>.
333 However, this would not be consistent with how we expect the
334 mathematical operators to work.
336 Perl resolves this dilemma by transparently calling a copy
337 constructor before calling a method defined to implement
338 a mutator (C<-->, C<+=>, and so on.).
339 In the above example, when Perl reaches the decrement
340 statement, it makes a copy of the object data in C<$a> and
341 assigns to C<$a> a reference to the copied data.
342 Only then does it call C<decr()>, which alters the copied
343 data, leaving C<$b> unchanged.
344 Thus the object metaphor is preserved as far as possible,
345 while mathemagical operations still work according to the
348 Note: the preceding paragraph describes what happens when
349 Perl autogenerates the copy constructor for an object based
351 For other cases, see L<Copy Constructor>.
353 =head2 Overloadable Operations
355 The complete list of keys that can be specified in the C<use overload>
356 directive are given, separated by spaces, in the values of the
357 hash C<%overload::ops>:
359 with_assign => '+ - * / % ** << >> x .',
360 assign => '+= -= *= /= %= **= <<= >>= x= .=',
361 num_comparison => '< <= > >= == !=',
362 '3way_comparison'=> '<=> cmp',
363 str_comparison => 'lt le gt ge eq ne',
364 binary => '& &= | |= ^ ^=',
367 func => 'atan2 cos sin exp abs log sqrt int',
368 conversion => 'bool "" 0+ qr',
371 dereferencing => '${} @{} %{} &{} *{}',
373 special => 'nomethod fallback ='
375 Most of the overloadable operators map one-to-one to these keys.
376 Exceptions, including additional overloadable operations not
377 apparent from this hash, are included in the notes which follow.
379 A warning is issued if an attempt is made to register an operator not found
386 The operator C<not> is not a valid key for C<use overload>.
387 However, if the operator C<!> is overloaded then the same
388 implementation will be used for C<not>
389 (since the two operators differ only in precedence).
393 The key C<neg> is used for unary minus to disambiguate it from
398 Assuming they are to behave analogously to Perl's C<++> and C<-->,
399 overloaded implementations of these operators are required to
400 mutate their operands.
402 No distinction is made between prefix and postfix forms of the
403 increment and decrement operators: these differ only in the
404 point at which Perl calls the associated subroutine when
405 evaluating an expression.
407 =item * I<Assignments>
409 += -= *= /= %= **= <<= >>= x= .=
412 Simple assignment is not overloadable (the C<'='> key is used
413 for the L<Copy Constructor>).
414 Perl does have a way to make assignments to an object do whatever
415 you want, but this involves using tie(), not overload -
416 see L<perlfunc/tie> and the L</COOKBOOK> examples below.
418 The subroutine for the assignment variant of an operator is
419 required only to return the result of the operation.
420 It is permitted to change the value of its operand
421 (this is safe because Perl calls the copy constructor first),
422 but this is optional since Perl assigns the returned value to
423 the left-hand operand anyway.
425 An object that overloads an assignment operator does so only in
426 respect of assignments to that object.
427 In other words, Perl never calls the corresponding methods with
428 the third argument (the "swap" argument) set to TRUE.
429 For example, the operation
433 cannot lead to C<$b>'s implementation of C<*=> being called,
434 even if C<$a> is a scalar.
435 (It can, however, generate a call to C<$b>'s method for C<*>).
437 =item * I<Non-mutators with a mutator variant>
439 + - * / % ** << >> x .
442 As described L<above|"Calling Conventions and Magic Autogeneration">,
443 Perl may call methods for operators like C<+> and C<&> in the course
444 of implementing missing operations like C<++>, C<+=>, and C<&=>.
445 While these methods may detect this usage by testing the definedness
446 of the third argument, they should in all cases avoid changing their
448 This is because Perl does not call the copy constructor before
449 invoking these methods.
453 Traditionally, the Perl function C<int> rounds to 0
454 (see L<perlfunc/int>), and so for floating-point-like types one
455 should follow the same semantic.
457 =item * I<String, numeric, boolean, and regexp conversions>
461 These conversions are invoked according to context as necessary.
462 For example, the subroutine for C<'""'> (stringify) may be used
463 where the overloaded object is passed as an argument to C<print>,
464 and that for C<'bool'> where it is tested in the condition of a flow
465 control statement (like C<while>) or the ternary C<?:> operation.
467 Of course, in contexts like, for example, C<$obj + 1>, Perl will
468 invoke C<$obj>'s implementation of C<+> rather than (in this
469 example) converting C<$obj> to a number using the numify method
470 C<'0+'> (an exception to this is when no method has been provided
471 for C<'+'> and L</fallback> is set to TRUE).
473 The subroutines for C<'""'>, C<'0+'>, and C<'bool'> can return
474 any arbitrary Perl value.
475 If the corresponding operation for this value is overloaded too,
476 the operation will be called again with this value.
478 As a special case if the overload returns the object itself then it will
479 be used directly. An overloaded conversion returning the object is
480 probably a bug, because you're likely to get something that looks like
481 C<YourPackage=HASH(0x8172b34)>.
485 The subroutine for C<'qr'> is used wherever the object is
486 interpolated into or used as a regexp, including when it
487 appears on the RHS of a C<=~> or C<!~> operator.
489 C<qr> must return a compiled regexp, or a ref to a compiled regexp
490 (such as C<qr//> returns), and any further overloading on the return
491 value will be ignored.
495 If C<E<lt>E<gt>> is overloaded then the same implementation is used
496 for both the I<read-filehandle> syntax C<E<lt>$varE<gt>> and
497 I<globbing> syntax C<E<lt>${var}E<gt>>.
499 =item * I<File tests>
501 The key C<'-X'> is used to specify a subroutine to handle all the
502 filetest operators (C<-f>, C<-x>, and so on: see L<perlfunc/-X> for
504 it is not possible to overload any filetest operator individually.
505 To distinguish them, the letter following the '-' is passed as the
506 second argument (that is, in the slot that for binary operators
507 is used to pass the second operand).
509 Calling an overloaded filetest operator does not affect the stat value
510 associated with the special filehandle C<_>. It still refers to the
511 result of the last C<stat>, C<lstat> or unoverloaded filetest.
513 This overload was introduced in Perl 5.12.
517 The key C<"~~"> allows you to override the smart matching logic used by
518 the C<~~> operator and the switch construct (C<given>/C<when>). See
519 L<perlsyn/Switch Statements> and L<feature>.
521 Unusually, the overloaded implementation of the smart match operator
522 does not get full control of the smart match behaviour.
523 In particular, in the following code:
526 use overload '~~' => 'match';
528 my $obj = Foo->new();
531 the smart match does I<not> invoke the method call like this:
533 $obj->match([1,2,3],0);
535 rather, the smart match distributive rule takes precedence, so $obj is
536 smart matched against each array element in turn until a match is found,
537 so you may see between one and three of these calls instead:
543 Consult the match table in L<perlop/"Smartmatch Operator"> for
544 details of when overloading is invoked.
546 =item * I<Dereferencing>
550 If these operators are not explicitly overloaded then they
551 work in the normal way, yielding the underlying scalar,
552 array, or whatever stores the object data (or the appropriate
553 error message if the dereference operator doesn't match it).
554 Defining a catch-all C<'nomethod'> (see L<below|/nomethod>)
555 makes no difference to this as the catch-all function will
556 not be called to implement a missing dereference operator.
558 If a dereference operator is overloaded then it must return a
559 I<reference> of the appropriate type (for example, the
560 subroutine for key C<'${}'> should return a reference to a
561 scalar, not a scalar), or another object which overloads the
562 operator: that is, the subroutine only determines what is
563 dereferenced and the actual dereferencing is left to Perl.
564 As a special case, if the subroutine returns the object itself
565 then it will not be called again - avoiding infinite recursion.
571 See L<Special Keys for C<use overload>>.
575 =head2 Magic Autogeneration
577 If a method for an operation is not found then Perl tries to
578 autogenerate a substitute implementation from the operations
579 that have been defined.
581 Note: the behaviour described in this section can be disabled
582 by setting C<fallback> to FALSE (see L</fallback>).
584 In the following tables, numbers indicate priority.
585 For example, the table below states that,
586 if no implementation for C<'!'> has been defined then Perl will
587 implement it using C<'bool'> (that is, by inverting the value
588 returned by the method for C<'bool'>);
589 if boolean conversion is also unimplemented then Perl will
590 use C<'0+'> or, failing that, C<'""'>.
592 operator | can be autogenerated from
595 =========|==========================
609 Note: The iterator (C<'E<lt>E<gt>'>) and file test (C<'-X'>)
610 operators work as normal: if the operand is not a blessed glob or
611 IO reference then it is converted to a string (using the method
612 for C<'""'>, C<'0+'>, or C<'bool'>) to be interpreted as a glob
615 operator | can be autogenerated from
618 =========|==========================
622 abs | a1 a2 b1 b2 [*]
630 * one from [a1, a2] and one from [b1, b2]
632 Just as numeric comparisons can be autogenerated from the method
633 for C<< '<=>' >>, string comparisons can be autogenerated from
636 operators | can be autogenerated from
637 ====================|===========================
638 lt gt le ge eq ne | cmp
640 Similarly, autogeneration for keys C<'+='> and C<'++'> is analogous
641 to C<'-='> and C<'--'> above:
643 operator | can be autogenerated from
646 =========|==========================
650 And other assignment variations are analogous to
651 C<'+='> and C<'-='> (and similar to C<'.='> and C<'x='> above):
653 operator || *= /= %= **= <<= >>= &= ^= |=
654 -------------------||--------------------------------
655 autogenerated from || * / % ** << >> & ^ |
657 Note also that the copy constructor (key C<'='>) may be
658 autogenerated, but only for objects based on scalars.
659 See L<Copy Constructor>.
661 =head3 Minimal Set of Overloaded Operations
663 Since some operations can be automatically generated from others, there is
664 a minimal set of operations that need to be overloaded in order to have
665 the complete set of overloaded operations at one's disposal.
666 Of course, the autogenerated operations may not do exactly what the user
667 expects. The minimal set is:
672 atan2 cos sin exp log sqrt int
676 Of the conversions, only one of string, boolean or numeric is
677 needed because each can be generated from either of the other two.
679 =head2 Special Keys for C<use overload>
683 The C<'nomethod'> key is used to specify a catch-all function to
684 be called for any operator that is not individually overloaded.
685 The specified function will be passed four parameters.
686 The first three arguments coincide with those that would have been
687 passed to the corresponding method if it had been defined.
688 The fourth argument is the C<use overload> key for that missing
691 For example, if C<$a> is an object blessed into a package declaring
693 use overload 'nomethod' => 'catch_all', # ...
699 could (unless a method is specifically declared for the key
700 C<'+'>) result in a call
702 catch_all($a, 3, 1, '+')
704 See L<How Perl Chooses an Operator Implementation>.
708 The value assigned to the key C<'fallback'> tells Perl how hard
709 it should try to find an alternative way to implement a missing
714 =item * defined, but FALSE
716 use overload "fallback" => 0, # ... ;
718 This disables L<Magic Autogeneration>.
722 In the default case where no value is explicitly assigned to
723 C<fallback>, magic autogeneration is enabled.
727 The same as for C<undef>, but if a missing operator cannot be
728 autogenerated then, instead of issuing an error message, Perl
729 is allowed to revert to what it would have done for that
730 operator if there had been no C<use overload> directive.
732 Note: in most cases, particularly the L<Copy Constructor>,
733 this is unlikely to be appropriate behaviour.
737 See L<How Perl Chooses an Operator Implementation>.
739 =head3 Copy Constructor
741 As mentioned L<above|"Mathemagic, Mutators, and Copy Constructors">,
742 this operation is called when a mutator is applied to a reference
743 that shares its object with some other reference.
744 For example, if C<$b> is mathemagical, and C<'++'> is overloaded
745 with C<'incr'>, and C<'='> is overloaded with C<'clone'>, then the
749 # ... (other code which does not modify $a or $b) ...
752 would be executed in a manner equivalent to
756 $b = $b->clone(undef, "");
765 The subroutine for C<'='> does not overload the Perl assignment
766 operator: it is used only to allow mutators to work as described
767 here. (See L</Assignments> above.)
771 As for other operations, the subroutine implementing '=' is passed
772 three arguments, though the last two are always C<undef> and C<''>.
776 The copy constructor is called only before a call to a function
777 declared to implement a mutator, for example, if C<++$b;> in the
778 code above is effected via a method declared for key C<'++'>
779 (or 'nomethod', passed C<'++'> as the fourth argument) or, by
780 autogeneration, C<'+='>.
781 It is not called if the increment operation is effected by a call
782 to the method for C<'+'> since, in the equivalent code,
787 the data referred to by C<$a> is unchanged by the assignment to
788 C<$b> of a reference to new object data.
792 The copy constructor is not called if Perl determines that it is
793 unnecessary because there is no other reference to the data being
798 If C<'fallback'> is undefined or TRUE then a copy constructor
799 can be autogenerated, but only for objects based on scalars.
800 In other cases it needs to be defined explicitly.
801 Where an object's data is stored as, for example, an array of
802 scalars, the following might be appropriate:
804 use overload '=' => sub { bless [ @{$_[0]} ] }, # ...
808 If C<'fallback'> is TRUE and no copy constructor is defined then,
809 for objects not based on scalars, Perl may silently fall back on
810 simple assignment - that is, assignment of the object reference.
811 In effect, this disables the copy constructor mechanism since
812 no new copy of the object data is created.
813 This is almost certainly not what you want.
814 (It is, however, consistent: for example, Perl's fallback for the
815 C<++> operator is to increment the reference itself.)
819 =head2 How Perl Chooses an Operator Implementation
821 Which is checked first, C<nomethod> or C<fallback>?
822 If the two operands of an operator are of different types and
823 both overload the operator, which implementation is used?
824 The following are the precedence rules:
830 If the first operand has declared a subroutine to overload the
831 operator then use that implementation.
835 Otherwise, if fallback is TRUE or undefined for the
836 first operand then see if the
837 L<rules for autogeneration|"Magic Autogeneration">
838 allows another of its operators to be used instead.
842 Unless the operator is an assignment (C<+=>, C<-=>, etc.),
843 repeat step (1) in respect of the second operand.
847 Repeat Step (2) in respect of the second operand.
851 If the first operand has a "nomethod" method then use that.
855 If the second operand has a "nomethod" method then use that.
859 If C<fallback> is TRUE for both operands
860 then perform the usual operation for the operator,
861 treating the operands as numbers, strings, or booleans
862 as appropriate for the operator (see note).
866 Nothing worked - die.
870 Where there is only one operand (or only one operand with
871 overloading) the checks in respect of the other operand above are
874 There are exceptions to the above rules for dereference operations
875 (which, if Step 1 fails, always fall back to the normal, built-in
876 implementations - see Dereferencing), and for C<~~> (which has its
877 own set of rules - see C<Matching> under L</Overloadable Operations>
880 Note on Step 7: some operators have a different semantic depending
881 on the type of their operands.
882 As there is no way to instruct Perl to treat the operands as, e.g.,
883 numbers instead of strings, the result here may not be what you
885 See L<BUGS AND PITFALLS>.
887 =head2 Losing Overloading
889 The restriction for the comparison operation is that even if, for example,
890 C<cmp> should return a blessed reference, the autogenerated C<lt>
891 function will produce only a standard logical value based on the
892 numerical value of the result of C<cmp>. In particular, a working
893 numeric conversion is needed in this case (possibly expressed in terms of
896 Similarly, C<.=> and C<x=> operators lose their mathemagical properties
897 if the string conversion substitution is applied.
899 When you chop() a mathemagical object it is promoted to a string and its
900 mathemagical properties are lost. The same can happen with other
903 =head2 Inheritance and Overloading
905 Overloading respects inheritance via the @ISA hierarchy.
906 Inheritance interacts with overloading in two ways.
910 =item Method names in the C<use overload> directive
914 use overload key => value;
916 is a string, it is interpreted as a method name - which may
917 (in the usual way) be inherited from another class.
919 =item Overloading of an operation is inherited by derived classes
921 Any class derived from an overloaded class is also overloaded
922 and inherits its operator implementations.
923 If the same operator is overloaded in more than one ancestor
924 then the implementation is determined by the usual inheritance
927 For example, if C<A> inherits from C<B> and C<C> (in that order),
928 C<B> overloads C<+> with C<\&D::plus_sub>, and C<C> overloads
929 C<+> by C<"plus_meth">, then the subroutine C<D::plus_sub> will
930 be called to implement operation C<+> for an object in package C<A>.
934 Note that in Perl version prior to 5.18 inheritance of the C<fallback> key
935 was not governed by the above rules. The value of C<fallback> in the first
936 overloaded ancestor was used. This was fixed in 5.18 to follow the usual
937 rules of inheritance.
939 =head2 Run-time Overloading
941 Since all C<use> directives are executed at compile-time, the only way to
942 change overloading during run-time is to
944 eval 'use overload "+" => \&addmethod';
948 eval 'no overload "+", "--", "<="';
950 though the use of these constructs during run-time is questionable.
952 =head2 Public Functions
954 Package C<overload.pm> provides the following public functions:
958 =item overload::StrVal(arg)
960 Gives the string value of C<arg> as in the
961 absence of stringify overloading. If you
962 are using this to get the address of a reference (useful for checking if two
963 references point to the same thing) then you may be better off using
964 C<Scalar::Util::refaddr()>, which is faster.
966 =item overload::Overloaded(arg)
968 Returns true if C<arg> is subject to overloading of some operations.
970 =item overload::Method(obj,op)
972 Returns C<undef> or a reference to the method that implements C<op>.
976 =head2 Overloading Constants
978 For some applications, the Perl parser mangles constants too much.
979 It is possible to hook into this process via C<overload::constant()>
980 and C<overload::remove_constant()> functions.
982 These functions take a hash as an argument. The recognized keys of this hash
989 to overload integer constants,
993 to overload floating point constants,
997 to overload octal and hexadecimal constants,
1001 to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted
1002 strings and here-documents,
1006 to overload constant pieces of regular expressions.
1010 The corresponding values are references to functions which take three arguments:
1011 the first one is the I<initial> string form of the constant, the second one
1012 is how Perl interprets this constant, the third one is how the constant is used.
1013 Note that the initial string form does not
1014 contain string delimiters, and has backslashes in backslash-delimiter
1015 combinations stripped (thus the value of delimiter is not relevant for
1016 processing of this string). The return value of this function is how this
1017 constant is going to be interpreted by Perl. The third argument is undefined
1018 unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote
1019 context (comes from strings, regular expressions, and single-quote HERE
1020 documents), it is C<tr> for arguments of C<tr>/C<y> operators,
1021 it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise.
1023 Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>,
1024 it is expected that overloaded constant strings are equipped with reasonable
1025 overloaded catenation operator, otherwise absurd results will result.
1026 Similarly, negative numbers are considered as negations of positive constants.
1028 Note that it is probably meaningless to call the functions overload::constant()
1029 and overload::remove_constant() from anywhere but import() and unimport() methods.
1030 From these methods they may be called as
1035 die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant';
1036 overload::constant integer => sub {Math::BigInt->new(shift)};
1039 =head1 IMPLEMENTATION
1041 What follows is subject to change RSN.
1043 The table of methods for all operations is cached in magic for the
1044 symbol table hash for the package. The cache is invalidated during
1045 processing of C<use overload>, C<no overload>, new function
1046 definitions, and changes in @ISA.
1048 (Every SVish thing has a magic queue, and magic is an entry in that
1049 queue. This is how a single variable may participate in multiple
1050 forms of magic simultaneously. For instance, environment variables
1051 regularly have two forms at once: their %ENV magic and their taint
1052 magic. However, the magic which implements overloading is applied to
1053 the stashes, which are rarely used directly, thus should not slow down
1056 If a package uses overload, it carries a special flag. This flag is also
1057 set when new function are defined or @ISA is modified. There will be a
1058 slight speed penalty on the very first operation thereafter that supports
1059 overloading, while the overload tables are updated. If there is no
1060 overloading present, the flag is turned off. Thus the only speed penalty
1061 thereafter is the checking of this flag.
1063 It is expected that arguments to methods that are not explicitly supposed
1064 to be changed are constant (but this is not enforced).
1068 Please add examples to what follows!
1070 =head2 Two-face Scalars
1072 Put this in F<two_face.pm> in your Perl library directory:
1074 package two_face; # Scalars with separate string and
1076 sub new { my $p = shift; bless [@_], $p }
1077 use overload '""' => \&str, '0+' => \&num, fallback => 1;
1078 sub num {shift->[1]}
1079 sub str {shift->[0]}
1084 my $seven = two_face->new("vii", 7);
1085 printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1;
1086 print "seven contains 'i'\n" if $seven =~ /i/;
1088 (The second line creates a scalar which has both a string value, and a
1089 numeric value.) This prints:
1091 seven=vii, seven=7, eight=8
1094 =head2 Two-face References
1096 Suppose you want to create an object which is accessible as both an
1097 array reference and a hash reference.
1100 use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} };
1108 tie %h, ref $self, $self;
1112 sub TIEHASH { my $p = shift; bless \ shift, $p }
1115 $fields{$_} = $i++ foreach qw{zero one two three};
1117 my $self = ${shift()};
1118 my $key = $fields{shift()};
1119 defined $key or die "Out of band access";
1120 $$self->[$key] = shift;
1123 my $self = ${shift()};
1124 my $key = $fields{shift()};
1125 defined $key or die "Out of band access";
1129 Now one can access an object using both the array and hash syntax:
1131 my $bar = two_refs->new(3,4,5,6);
1133 $bar->{two} == 11 or die 'bad hash fetch';
1135 Note several important features of this example. First of all, the
1136 I<actual> type of $bar is a scalar reference, and we do not overload
1137 the scalar dereference. Thus we can get the I<actual> non-overloaded
1138 contents of $bar by just using C<$$bar> (what we do in functions which
1139 overload dereference). Similarly, the object returned by the
1140 TIEHASH() method is a scalar reference.
1142 Second, we create a new tied hash each time the hash syntax is used.
1143 This allows us not to worry about a possibility of a reference loop,
1144 which would lead to a memory leak.
1146 Both these problems can be cured. Say, if we want to overload hash
1147 dereference on a reference to an object which is I<implemented> as a
1148 hash itself, the only problem one has to circumvent is how to access
1149 this I<actual> hash (as opposed to the I<virtual> hash exhibited by the
1150 overloaded dereference operator). Here is one possible fetching routine:
1153 my ($self, $key) = (shift, shift);
1154 my $class = ref $self;
1155 bless $self, 'overload::dummy'; # Disable overloading of %{}
1156 my $out = $self->{$key};
1157 bless $self, $class; # Restore overloading
1161 To remove creation of the tied hash on each access, one may an extra
1162 level of indirection which allows a non-circular structure of references:
1165 use overload '%{}' => sub { ${shift()}->[1] },
1166 '@{}' => sub { ${shift()}->[0] };
1172 bless \ [$a, \%h], $p;
1177 tie %h, ref $self, $self;
1181 sub TIEHASH { my $p = shift; bless \ shift, $p }
1184 $fields{$_} = $i++ foreach qw{zero one two three};
1187 my $key = $fields{shift()};
1188 defined $key or die "Out of band access";
1193 my $key = $fields{shift()};
1194 defined $key or die "Out of band access";
1198 Now if $baz is overloaded like this, then C<$baz> is a reference to a
1199 reference to the intermediate array, which keeps a reference to an
1200 actual array, and the access hash. The tie()ing object for the access
1201 hash is a reference to a reference to the actual array, so
1207 There are no loops of references.
1211 Both "objects" which are blessed into the class C<two_refs1> are
1212 references to a reference to an array, thus references to a I<scalar>.
1213 Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no
1214 overloaded operations.
1218 =head2 Symbolic Calculator
1220 Put this in F<symbolic.pm> in your Perl library directory:
1222 package symbolic; # Primitive symbolic calculator
1223 use overload nomethod => \&wrap;
1225 sub new { shift; bless ['n', @_] }
1227 my ($obj, $other, $inv, $meth) = @_;
1228 ($obj, $other) = ($other, $obj) if $inv;
1229 bless [$meth, $obj, $other];
1232 This module is very unusual as overloaded modules go: it does not
1233 provide any usual overloaded operators, instead it provides an
1234 implementation for L<C<nomethod>>. In this example the C<nomethod>
1235 subroutine returns an object which encapsulates operations done over
1236 the objects: C<< symbolic->new(3) >> contains C<['n', 3]>, C<< 2 +
1237 symbolic->new(3) >> contains C<['+', 2, ['n', 3]]>.
1239 Here is an example of the script which "calculates" the side of
1240 circumscribed octagon using the above package:
1243 my $iter = 1; # 2**($iter+2) = 8
1244 my $side = symbolic->new(1);
1248 $side = (sqrt(1 + $side**2) - 1)/$side;
1252 The value of $side is
1254 ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]],
1255 undef], 1], ['n', 1]]
1257 Note that while we obtained this value using a nice little script,
1258 there is no simple way to I<use> this value. In fact this value may
1259 be inspected in debugger (see L<perldebug>), but only if
1260 C<bareStringify> B<O>ption is set, and not via C<p> command.
1262 If one attempts to print this value, then the overloaded operator
1263 C<""> will be called, which will call C<nomethod> operator. The
1264 result of this operator will be stringified again, but this result is
1265 again of type C<symbolic>, which will lead to an infinite loop.
1267 Add a pretty-printer method to the module F<symbolic.pm>:
1270 my ($meth, $a, $b) = @{+shift};
1271 $a = 'u' unless defined $a;
1272 $b = 'u' unless defined $b;
1273 $a = $a->pretty if ref $a;
1274 $b = $b->pretty if ref $b;
1278 Now one can finish the script by
1280 print "side = ", $side->pretty, "\n";
1282 The method C<pretty> is doing object-to-string conversion, so it
1283 is natural to overload the operator C<""> using this method. However,
1284 inside such a method it is not necessary to pretty-print the
1285 I<components> $a and $b of an object. In the above subroutine
1286 C<"[$meth $a $b]"> is a catenation of some strings and components $a
1287 and $b. If these components use overloading, the catenation operator
1288 will look for an overloaded operator C<.>; if not present, it will
1289 look for an overloaded operator C<"">. Thus it is enough to use
1291 use overload nomethod => \&wrap, '""' => \&str;
1293 my ($meth, $a, $b) = @{+shift};
1294 $a = 'u' unless defined $a;
1295 $b = 'u' unless defined $b;
1299 Now one can change the last line of the script to
1301 print "side = $side\n";
1305 side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]]
1307 and one can inspect the value in debugger using all the possible
1310 Something is still amiss: consider the loop variable $cnt of the
1311 script. It was a number, not an object. We cannot make this value of
1312 type C<symbolic>, since then the loop will not terminate.
1314 Indeed, to terminate the cycle, the $cnt should become false.
1315 However, the operator C<bool> for checking falsity is overloaded (this
1316 time via overloaded C<"">), and returns a long string, thus any object
1317 of type C<symbolic> is true. To overcome this, we need a way to
1318 compare an object to 0. In fact, it is easier to write a numeric
1321 Here is the text of F<symbolic.pm> with such a routine added (and
1322 slightly modified str()):
1324 package symbolic; # Primitive symbolic calculator
1326 nomethod => \&wrap, '""' => \&str, '0+' => \#
1328 sub new { shift; bless ['n', @_] }
1330 my ($obj, $other, $inv, $meth) = @_;
1331 ($obj, $other) = ($other, $obj) if $inv;
1332 bless [$meth, $obj, $other];
1335 my ($meth, $a, $b) = @{+shift};
1336 $a = 'u' unless defined $a;
1343 my %subr = ( n => sub {$_[0]},
1344 sqrt => sub {sqrt $_[0]},
1345 '-' => sub {shift() - shift()},
1346 '+' => sub {shift() + shift()},
1347 '/' => sub {shift() / shift()},
1348 '*' => sub {shift() * shift()},
1349 '**' => sub {shift() ** shift()},
1352 my ($meth, $a, $b) = @{+shift};
1353 my $subr = $subr{$meth}
1354 or die "Do not know how to ($meth) in symbolic";
1355 $a = $a->num if ref $a eq __PACKAGE__;
1356 $b = $b->num if ref $b eq __PACKAGE__;
1360 All the work of numeric conversion is done in %subr and num(). Of
1361 course, %subr is not complete, it contains only operators used in the
1362 example below. Here is the extra-credit question: why do we need an
1363 explicit recursion in num()? (Answer is at the end of this section.)
1365 Use this module like this:
1368 my $iter = symbolic->new(2); # 16-gon
1369 my $side = symbolic->new(1);
1373 $cnt = $cnt - 1; # Mutator '--' not implemented
1374 $side = (sqrt(1 + $side**2) - 1)/$side;
1376 printf "%s=%f\n", $side, $side;
1377 printf "pi=%f\n", $side*(2**($iter+2));
1379 It prints (without so many line breaks)
1381 [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1]
1383 [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912
1386 The above module is very primitive. It does not implement
1387 mutator methods (C<++>, C<-=> and so on), does not do deep copying
1388 (not required without mutators!), and implements only those arithmetic
1389 operations which are used in the example.
1391 To implement most arithmetic operations is easy; one should just use
1392 the tables of operations, and change the code which fills %subr to
1394 my %subr = ( 'n' => sub {$_[0]} );
1395 foreach my $op (split " ", $overload::ops{with_assign}) {
1396 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1398 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1399 foreach my $op (split " ", "@overload::ops{ @bins }") {
1400 $subr{$op} = eval "sub {shift() $op shift()}";
1402 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1403 print "defining '$op'\n";
1404 $subr{$op} = eval "sub {$op shift()}";
1407 Since subroutines implementing assignment operators are not required
1408 to modify their operands (see L<Overloadable Operations> above),
1409 we do not need anything special to make C<+=> and friends work,
1410 besides adding these operators to %subr and defining a copy
1411 constructor (needed since Perl has no way to know that the
1412 implementation of C<'+='> does not mutate the argument -
1413 see L<Copy Constructor>).
1415 To implement a copy constructor, add C<< '=' => \&cpy >> to C<use overload>
1416 line, and code (this code assumes that mutators change things one level
1417 deep only, so recursive copying is not needed):
1421 bless [@$self], ref $self;
1424 To make C<++> and C<--> work, we need to implement actual mutators,
1425 either directly, or in C<nomethod>. We continue to do things inside
1426 C<nomethod>, thus add
1428 if ($meth eq '++' or $meth eq '--') {
1429 @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference
1433 after the first line of wrap(). This is not a most effective
1434 implementation, one may consider
1436 sub inc { $_[0] = bless ['++', shift, 1]; }
1440 As a final remark, note that one can fill %subr by
1442 my %subr = ( 'n' => sub {$_[0]} );
1443 foreach my $op (split " ", $overload::ops{with_assign}) {
1444 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1446 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1447 foreach my $op (split " ", "@overload::ops{ @bins }") {
1448 $subr{$op} = eval "sub {shift() $op shift()}";
1450 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1451 $subr{$op} = eval "sub {$op shift()}";
1453 $subr{'++'} = $subr{'+'};
1454 $subr{'--'} = $subr{'-'};
1456 This finishes implementation of a primitive symbolic calculator in
1457 50 lines of Perl code. Since the numeric values of subexpressions
1458 are not cached, the calculator is very slow.
1460 Here is the answer for the exercise: In the case of str(), we need no
1461 explicit recursion since the overloaded C<.>-operator will fall back
1462 to an existing overloaded operator C<"">. Overloaded arithmetic
1463 operators I<do not> fall back to numeric conversion if C<fallback> is
1464 not explicitly requested. Thus without an explicit recursion num()
1465 would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild
1466 the argument of num().
1468 If you wonder why defaults for conversion are different for str() and
1469 num(), note how easy it was to write the symbolic calculator. This
1470 simplicity is due to an appropriate choice of defaults. One extra
1471 note: due to the explicit recursion num() is more fragile than sym():
1472 we need to explicitly check for the type of $a and $b. If components
1473 $a and $b happen to be of some related type, this may lead to problems.
1475 =head2 I<Really> Symbolic Calculator
1477 One may wonder why we call the above calculator symbolic. The reason
1478 is that the actual calculation of the value of expression is postponed
1479 until the value is I<used>.
1481 To see it in action, add a method
1486 @$obj->[0,1] = ('=', shift);
1489 to the package C<symbolic>. After this change one can do
1491 my $a = symbolic->new(3);
1492 my $b = symbolic->new(4);
1493 my $c = sqrt($a**2 + $b**2);
1495 and the numeric value of $c becomes 5. However, after calling
1497 $a->STORE(12); $b->STORE(5);
1499 the numeric value of $c becomes 13. There is no doubt now that the module
1500 symbolic provides a I<symbolic> calculator indeed.
1502 To hide the rough edges under the hood, provide a tie()d interface to the
1503 package C<symbolic>. Add methods
1505 sub TIESCALAR { my $pack = shift; $pack->new(@_) }
1507 sub nop { } # Around a bug
1509 (the bug, fixed in Perl 5.14, is described in L<"BUGS">). One can use this
1512 tie $a, 'symbolic', 3;
1513 tie $b, 'symbolic', 4;
1514 $a->nop; $b->nop; # Around a bug
1516 my $c = sqrt($a**2 + $b**2);
1518 Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value
1519 of $c becomes 13. To insulate the user of the module add a method
1521 sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; }
1526 symbolic->vars($a, $b);
1527 my $c = sqrt($a**2 + $b**2);
1530 printf "c5 %s=%f\n", $c, $c;
1533 printf "c13 %s=%f\n", $c, $c;
1535 shows that the numeric value of $c follows changes to the values of $a
1540 Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>.
1544 The C<overloading> pragma can be used to enable or disable overloaded
1545 operations within a lexical scope - see L<overloading>.
1549 When Perl is run with the B<-Do> switch or its equivalent, overloading
1550 induces diagnostic messages.
1552 Using the C<m> command of Perl debugger (see L<perldebug>) one can
1553 deduce which operations are overloaded (and which ancestor triggers
1554 this overloading). Say, if C<eq> is overloaded, then the method C<(eq>
1555 is shown by debugger. The method C<()> corresponds to the C<fallback>
1556 key (in fact a presence of this method shows that this package has
1557 overloading enabled, and it is what is used by the C<Overloaded>
1558 function of module C<overload>).
1560 The module might issue the following warnings:
1564 =item Odd number of arguments for overload::constant
1566 (W) The call to overload::constant contained an odd number of arguments.
1567 The arguments should come in pairs.
1569 =item '%s' is not an overloadable type
1571 (W) You tried to overload a constant type the overload package is unaware of.
1573 =item '%s' is not a code reference
1575 (W) The second (fourth, sixth, ...) argument of overload::constant needs
1576 to be a code reference. Either an anonymous subroutine, or a reference
1579 =item overload arg '%s' is invalid
1581 (W) C<use overload> was passed an argument it did not
1582 recognize. Did you mistype an operator?
1586 =head1 BUGS AND PITFALLS
1592 A pitfall when fallback is TRUE and Perl resorts to a built-in
1593 implementation of an operator is that some operators have more
1594 than one semantic, for example C<|>:
1596 use overload '0+' => sub { $_[0]->{n}; },
1598 my $x = bless { n => 4 }, "main";
1599 my $y = bless { n => 8 }, "main";
1600 print $x | $y, "\n";
1602 You might expect this to output "12".
1603 In fact, it prints "<": the ASCII result of treating "|"
1604 as a bitwise string operator - that is, the result of treating
1605 the operands as the strings "4" and "8" rather than numbers.
1606 The fact that numify (C<0+>) is implemented but stringify
1607 (C<"">) isn't makes no difference since the latter is simply
1608 autogenerated from the former.
1610 The only way to change this is to provide your own subroutine
1615 Magic autogeneration increases the potential for inadvertently
1616 creating self-referential structures.
1617 Currently Perl will not free self-referential
1618 structures until cycles are explicitly broken.
1621 use overload '+' => 'add';
1622 sub add { bless [ \$_[0], \$_[1] ] };
1624 is asking for trouble, since
1628 will effectively become
1630 $obj = add($obj, $y, undef);
1632 with the same result as
1634 $obj = [\$obj, \$foo];
1636 Even if no I<explicit> assignment-variants of operators are present in
1637 the script, they may be generated by the optimizer.
1644 my $tmp = 'obj = ' . $obj; $tmp .= "\n";
1648 The symbol table is filled with names looking like line-noise.
1652 This bug was fixed in Perl 5.18, but may still trip you up if you are using
1655 For the purpose of inheritance every overloaded package behaves as if
1656 C<fallback> is present (possibly undefined). This may create
1657 interesting effects if some package is not overloaded, but inherits
1658 from two overloaded packages.
1662 Before Perl 5.14, the relation between overloading and tie()ing was broken.
1663 Overloading was triggered or not based on the I<previous> class of the
1666 This happened because the presence of overloading was checked
1667 too early, before any tie()d access was attempted. If the
1668 class of the value FETCH()ed from the tied variable does not
1669 change, a simple workaround for code that is to run on older Perl
1670 versions is to access the value (via C<() = $foo> or some such)
1671 immediately after tie()ing, so that after this call the I<previous> class
1672 coincides with the current one.
1676 Barewords are not covered by overloaded string constants.