3 $overload::hint_bits = 0x20000;
11 $ {$package . "::OVERLOAD"}{dummy}++; # Register with magic by touching.
12 *{$package . "::()"} = \&nil; # Make it findable via fetchmethod.
14 if ($_ eq 'fallback') {
18 if (not ref $sub and $sub !~ /::/) {
19 $ {$package . "::(" . $_} = $sub;
22 #print STDERR "Setting `$ {'package'}::\cO$_' to \\&`$sub'.\n";
23 *{$package . "::(" . $_} = \&{ $sub };
26 ${$package . "::()"} = $fb; # Make it findable too (fallback only).
30 $package = (caller())[0];
31 # *{$package . "::OVERLOAD"} = \&OVERLOAD;
33 $package->overload::OVERLOAD(@_);
37 $package = (caller())[0];
38 ${$package . "::OVERLOAD"}{dummy}++; # Upgrade the table
41 if ($_ eq 'fallback') {
42 undef $ {$package . "::()"};
44 delete $ {$package . "::"}{"(" . $_};
51 $package = ref $package if ref $package;
57 return undef unless $globref;
58 my $sub = \&{*$globref};
59 return $sub if $sub ne \&nil;
60 return shift->can($ {*$globref});
63 sub OverloadedStringify {
65 $package = ref $package if ref $package;
67 ov_method mycan($package, '(""'), $package
68 or ov_method mycan($package, '(0+'), $package
69 or ov_method mycan($package, '(bool'), $package
70 or ov_method mycan($package, '(nomethod'), $package;
75 $package = ref $package if ref $package;
76 #my $meth = $package->can('(' . shift);
77 ov_method mycan($package, '(' . shift), $package;
78 #return $meth if $meth ne \&nil;
83 my $package = ref $_[0];
84 return "$_[0]" unless $package;
85 bless $_[0], overload::Fake; # Non-overloaded package
87 bless $_[0], $package; # Back
88 $package . substr $str, index $str, '=';
92 (OverloadedStringify($_[0]) or ref($_[0]) eq 'Regexp') ?
97 sub mycan { # Real can would leave stubs.
98 my ($package, $meth) = @_;
99 return \*{$package . "::$meth"} if defined &{$package . "::$meth"};
101 foreach $p (@{$package . "::ISA"}) {
102 my $out = mycan($p, $meth);
116 %ops = ( with_assign => "+ - * / % ** << >> x .",
117 assign => "+= -= *= /= %= **= <<= >>= x= .=",
118 num_comparison => "< <= > >= == !=",
119 '3way_comparison'=> "<=> cmp",
120 str_comparison => "lt le gt ge eq ne",
124 func => "atan2 cos sin exp abs log sqrt",
125 conversion => 'bool "" 0+',
127 dereferencing => '${} @{} %{} &{} *{}',
128 special => 'nomethod fallback =');
130 use warnings::register;
132 # Arguments: what, sub
135 warnings::warnif ("Odd number of arguments for overload::constant");
138 elsif (!exists $constants {$_ [0]}) {
139 warnings::warnif ("`$_[0]' is not an overloadable type");
141 elsif (!ref $_ [1] || "$_[1]" !~ /CODE\(0x[\da-f]+\)$/) {
142 # Can't use C<ref $_[1] eq "CODE"> above as code references can be
143 # blessed, and C<ref> would return the package the ref is blessed into.
144 if (warnings::enabled) {
145 $_ [1] = "undef" unless defined $_ [1];
146 warnings::warn ("`$_[1]' is not a code reference");
151 $^H |= $constants{$_[0]} | $overload::hint_bits;
157 sub remove_constant {
158 # Arguments: what, sub
161 $^H &= ~ $constants{$_[0]};
172 overload - Package for overloading perl operations
185 $a = new SomeThing 57;
188 if (overload::Overloaded $b) {...}
190 $strval = overload::StrVal $b;
194 =head2 Declaration of overloaded functions
196 The compilation directive
203 declares function Number::add() for addition, and method muas() in
204 the "class" C<Number> (or one of its base classes)
205 for the assignment form C<*=> of multiplication.
207 Arguments of this directive come in (key, value) pairs. Legal values
208 are values legal inside a C<&{ ... }> call, so the name of a
209 subroutine, a reference to a subroutine, or an anonymous subroutine
210 will all work. Note that values specified as strings are
211 interpreted as methods, not subroutines. Legal keys are listed below.
213 The subroutine C<add> will be called to execute C<$a+$b> if $a
214 is a reference to an object blessed into the package C<Number>, or if $a is
215 not an object from a package with defined mathemagic addition, but $b is a
216 reference to a C<Number>. It can also be called in other situations, like
217 C<$a+=7>, or C<$a++>. See L<MAGIC AUTOGENERATION>. (Mathemagical
218 methods refer to methods triggered by an overloaded mathematical
221 Since overloading respects inheritance via the @ISA hierarchy, the
222 above declaration would also trigger overloading of C<+> and C<*=> in
223 all the packages which inherit from C<Number>.
225 =head2 Calling Conventions for Binary Operations
227 The functions specified in the C<use overload ...> directive are called
228 with three (in one particular case with four, see L<Last Resort>)
229 arguments. If the corresponding operation is binary, then the first
230 two arguments are the two arguments of the operation. However, due to
231 general object calling conventions, the first argument should always be
232 an object in the package, so in the situation of C<7+$a>, the
233 order of the arguments is interchanged. It probably does not matter
234 when implementing the addition method, but whether the arguments
235 are reversed is vital to the subtraction method. The method can
236 query this information by examining the third argument, which can take
237 three different values:
243 the order of arguments is as in the current operation.
247 the arguments are reversed.
251 the current operation is an assignment variant (as in
252 C<$a+=7>), but the usual function is called instead. This additional
253 information can be used to generate some optimizations. Compare
254 L<Calling Conventions for Mutators>.
258 =head2 Calling Conventions for Unary Operations
260 Unary operation are considered binary operations with the second
261 argument being C<undef>. Thus the functions that overloads C<{"++"}>
262 is called with arguments C<($a,undef,'')> when $a++ is executed.
264 =head2 Calling Conventions for Mutators
266 Two types of mutators have different calling conventions:
270 =item C<++> and C<-->
272 The routines which implement these operators are expected to actually
273 I<mutate> their arguments. So, assuming that $obj is a reference to a
276 sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n}
278 is an appropriate implementation of overloaded C<++>. Note that
280 sub incr { ++$ {$_[0]} ; shift }
282 is OK if used with preincrement and with postincrement. (In the case
283 of postincrement a copying will be performed, see L<Copy Constructor>.)
285 =item C<x=> and other assignment versions
287 There is nothing special about these methods. They may change the
288 value of their arguments, and may leave it as is. The result is going
289 to be assigned to the value in the left-hand-side if different from
292 This allows for the same method to be used as overloaded C<+=> and
293 C<+>. Note that this is I<allowed>, but not recommended, since by the
294 semantic of L<"Fallback"> Perl will call the method for C<+> anyway,
295 if C<+=> is not overloaded.
299 B<Warning.> Due to the presense of assignment versions of operations,
300 routines which may be called in assignment context may create
301 self-referential structures. Currently Perl will not free self-referential
302 structures until cycles are C<explicitly> broken. You may get problems
303 when traversing your structures too.
307 use overload '+' => sub { bless [ \$_[0], \$_[1] ] };
309 is asking for trouble, since for code C<$obj += $foo> the subroutine
310 is called as C<$obj = add($obj, $foo, undef)>, or C<$obj = [\$obj,
311 \$foo]>. If using such a subroutine is an important optimization, one
312 can overload C<+=> explicitly by a non-"optimized" version, or switch
313 to non-optimized version if C<not defined $_[2]> (see
314 L<Calling Conventions for Binary Operations>).
316 Even if no I<explicit> assignment-variants of operators are present in
317 the script, they may be generated by the optimizer. Say, C<",$obj,"> or
318 C<',' . $obj . ','> may be both optimized to
320 my $tmp = ',' . $obj; $tmp .= ',';
322 =head2 Overloadable Operations
324 The following symbols can be specified in C<use overload> directive:
328 =item * I<Arithmetic operations>
330 "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=",
331 "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=",
333 For these operations a substituted non-assignment variant can be called if
334 the assignment variant is not available. Methods for operations "C<+>",
335 "C<->", "C<+=>", and "C<-=>" can be called to automatically generate
336 increment and decrement methods. The operation "C<->" can be used to
337 autogenerate missing methods for unary minus or C<abs>.
339 See L<"MAGIC AUTOGENERATION">, L<"Calling Conventions for Mutators"> and
340 L<"Calling Conventions for Binary Operations">) for details of these
343 =item * I<Comparison operations>
345 "<", "<=", ">", ">=", "==", "!=", "<=>",
346 "lt", "le", "gt", "ge", "eq", "ne", "cmp",
348 If the corresponding "spaceship" variant is available, it can be
349 used to substitute for the missing operation. During C<sort>ing
350 arrays, C<cmp> is used to compare values subject to C<use overload>.
352 =item * I<Bit operations>
354 "&", "^", "|", "neg", "!", "~",
356 "C<neg>" stands for unary minus. If the method for C<neg> is not
357 specified, it can be autogenerated using the method for
358 subtraction. If the method for "C<!>" is not specified, it can be
359 autogenerated using the methods for "C<bool>", or "C<\"\">", or "C<0+>".
361 =item * I<Increment and decrement>
365 If undefined, addition and subtraction methods can be
366 used instead. These operations are called both in prefix and
369 =item * I<Transcendental functions>
371 "atan2", "cos", "sin", "exp", "abs", "log", "sqrt",
373 If C<abs> is unavailable, it can be autogenerated using methods
374 for "E<lt>" or "E<lt>=E<gt>" combined with either unary minus or subtraction.
376 =item * I<Boolean, string and numeric conversion>
378 "bool", "\"\"", "0+",
380 If one or two of these operations are not overloaded, the remaining ones can
381 be used instead. C<bool> is used in the flow control operators
382 (like C<while>) and for the ternary "C<?:>" operation. These functions can
383 return any arbitrary Perl value. If the corresponding operation for this value
384 is overloaded too, that operation will be called again with this value.
390 If not overloaded, the argument will be converted to a filehandle or
391 glob (which may require a stringification). The same overloading
392 happens both for the I<read-filehandle> syntax C<E<lt>$varE<gt>> and
393 I<globbing> syntax C<E<lt>${var}E<gt>>.
395 =item * I<Dereferencing>
397 '${}', '@{}', '%{}', '&{}', '*{}'.
399 If not overloaded, the argument will be dereferenced I<as is>, thus
400 should be of correct type. These functions should return a reference
401 of correct type, or another object with overloaded dereferencing.
403 As a special case if the overload returns the object itself then it
404 will be used directly (provided it is the correct type).
406 The dereference operators must be specified explicitly they will not be passed to
411 "nomethod", "fallback", "=",
413 see L<SPECIAL SYMBOLS FOR C<use overload>>.
417 See L<"Fallback"> for an explanation of when a missing method can be
420 A computer-readable form of the above table is available in the hash
421 %overload::ops, with values being space-separated lists of names:
423 with_assign => '+ - * / % ** << >> x .',
424 assign => '+= -= *= /= %= **= <<= >>= x= .=',
425 num_comparison => '< <= > >= == !=',
426 '3way_comparison'=> '<=> cmp',
427 str_comparison => 'lt le gt ge eq ne',
431 func => 'atan2 cos sin exp abs log sqrt',
432 conversion => 'bool "" 0+',
434 dereferencing => '${} @{} %{} &{} *{}',
435 special => 'nomethod fallback ='
437 =head2 Inheritance and overloading
439 Inheritance interacts with overloading in two ways.
443 =item Strings as values of C<use overload> directive
447 use overload key => value;
449 is a string, it is interpreted as a method name.
451 =item Overloading of an operation is inherited by derived classes
453 Any class derived from an overloaded class is also overloaded. The
454 set of overloaded methods is the union of overloaded methods of all
455 the ancestors. If some method is overloaded in several ancestor, then
456 which description will be used is decided by the usual inheritance
459 If C<A> inherits from C<B> and C<C> (in this order), C<B> overloads
460 C<+> with C<\&D::plus_sub>, and C<C> overloads C<+> by C<"plus_meth">,
461 then the subroutine C<D::plus_sub> will be called to implement
462 operation C<+> for an object in package C<A>.
466 Note that since the value of the C<fallback> key is not a subroutine,
467 its inheritance is not governed by the above rules. In the current
468 implementation, the value of C<fallback> in the first overloaded
469 ancestor is used, but this is accidental and subject to change.
471 =head1 SPECIAL SYMBOLS FOR C<use overload>
473 Three keys are recognized by Perl that are not covered by the above
478 C<"nomethod"> should be followed by a reference to a function of four
479 parameters. If defined, it is called when the overloading mechanism
480 cannot find a method for some operation. The first three arguments of
481 this function coincide with the arguments for the corresponding method if
482 it were found, the fourth argument is the symbol
483 corresponding to the missing method. If several methods are tried,
484 the last one is used. Say, C<1-$a> can be equivalent to
486 &nomethodMethod($a,1,1,"-")
488 if the pair C<"nomethod" =E<gt> "nomethodMethod"> was specified in the
489 C<use overload> directive.
491 The C<"nomethod"> mechanism is I<not> used for the dereference operators
492 ( ${} @{} %{} &{} *{} ).
495 If some operation cannot be resolved, and there is no function
496 assigned to C<"nomethod">, then an exception will be raised via die()--
497 unless C<"fallback"> was specified as a key in C<use overload> directive.
502 The key C<"fallback"> governs what to do if a method for a particular
503 operation is not found. Three different cases are possible depending on
504 the value of C<"fallback">:
511 substituted method (see L<MAGIC AUTOGENERATION>). If this fails, it
512 then tries to calls C<"nomethod"> value; if missing, an exception
517 The same as for the C<undef> value, but no exception is raised. Instead,
518 it silently reverts to what it would have done were there no C<use overload>
521 =item * defined, but FALSE
523 No autogeneration is tried. Perl tries to call
524 C<"nomethod"> value, and if this is missing, raises an exception.
528 B<Note.> C<"fallback"> inheritance via @ISA is not carved in stone
529 yet, see L<"Inheritance and overloading">.
531 =head2 Copy Constructor
533 The value for C<"="> is a reference to a function with three
534 arguments, i.e., it looks like the other values in C<use
535 overload>. However, it does not overload the Perl assignment
536 operator. This would go against Camel hair.
538 This operation is called in the situations when a mutator is applied
539 to a reference that shares its object with some other reference, such
545 To make this change $a and not change $b, a copy of C<$$a> is made,
546 and $a is assigned a reference to this new object. This operation is
547 done during execution of the C<++$a>, and not during the assignment,
548 (so before the increment C<$$a> coincides with C<$$b>). This is only
549 done if C<++> is expressed via a method for C<'++'> or C<'+='> (or
550 C<nomethod>). Note that if this operation is expressed via C<'+'>
551 a nonmutator, i.e., as in
556 then C<$a> does not reference a new copy of C<$$a>, since $$a does not
557 appear as lvalue when the above code is executed.
559 If the copy constructor is required during the execution of some mutator,
560 but a method for C<'='> was not specified, it can be autogenerated as a
561 string copy if the object is a plain scalar.
567 The actually executed code for
570 Something else which does not modify $a or $b....
576 Something else which does not modify $a or $b....
577 $a = $a->clone(undef,"");
580 if $b was mathemagical, and C<'++'> was overloaded with C<\&incr>,
581 C<'='> was overloaded with C<\&clone>.
585 Same behaviour is triggered by C<$b = $a++>, which is consider a synonym for
588 =head1 MAGIC AUTOGENERATION
590 If a method for an operation is not found, and the value for C<"fallback"> is
591 TRUE or undefined, Perl tries to autogenerate a substitute method for
592 the missing operation based on the defined operations. Autogenerated method
593 substitutions are possible for the following operations:
597 =item I<Assignment forms of arithmetic operations>
599 C<$a+=$b> can use the method for C<"+"> if the method for C<"+=">
602 =item I<Conversion operations>
604 String, numeric, and boolean conversion are calculated in terms of one
605 another if not all of them are defined.
607 =item I<Increment and decrement>
609 The C<++$a> operation can be expressed in terms of C<$a+=1> or C<$a+1>,
610 and C<$a--> in terms of C<$a-=1> and C<$a-1>.
614 can be expressed in terms of C<$aE<lt>0> and C<-$a> (or C<0-$a>).
618 can be expressed in terms of subtraction.
622 C<!> and C<not> can be expressed in terms of boolean conversion, or
623 string or numerical conversion.
625 =item I<Concatenation>
627 can be expressed in terms of string conversion.
629 =item I<Comparison operations>
631 can be expressed in terms of its "spaceship" counterpart: either
632 C<E<lt>=E<gt>> or C<cmp>:
634 <, >, <=, >=, ==, != in terms of <=>
635 lt, gt, le, ge, eq, ne in terms of cmp
639 <> in terms of builtin operations
641 =item I<Dereferencing>
643 ${} @{} %{} &{} *{} in terms of builtin operations
645 =item I<Copy operator>
647 can be expressed in terms of an assignment to the dereferenced value, if this
648 value is a scalar and not a reference.
652 =head1 Losing overloading
654 The restriction for the comparison operation is that even if, for example,
655 `C<cmp>' should return a blessed reference, the autogenerated `C<lt>'
656 function will produce only a standard logical value based on the
657 numerical value of the result of `C<cmp>'. In particular, a working
658 numeric conversion is needed in this case (possibly expressed in terms of
661 Similarly, C<.=> and C<x=> operators lose their mathemagical properties
662 if the string conversion substitution is applied.
664 When you chop() a mathemagical object it is promoted to a string and its
665 mathemagical properties are lost. The same can happen with other
668 =head1 Run-time Overloading
670 Since all C<use> directives are executed at compile-time, the only way to
671 change overloading during run-time is to
673 eval 'use overload "+" => \&addmethod';
677 eval 'no overload "+", "--", "<="';
679 though the use of these constructs during run-time is questionable.
681 =head1 Public functions
683 Package C<overload.pm> provides the following public functions:
687 =item overload::StrVal(arg)
689 Gives string value of C<arg> as in absence of stringify overloading.
691 =item overload::Overloaded(arg)
693 Returns true if C<arg> is subject to overloading of some operations.
695 =item overload::Method(obj,op)
697 Returns C<undef> or a reference to the method that implements C<op>.
701 =head1 Overloading constants
703 For some application Perl parser mangles constants too much. It is possible
704 to hook into this process via overload::constant() and overload::remove_constant()
707 These functions take a hash as an argument. The recognized keys of this hash
714 to overload integer constants,
718 to overload floating point constants,
722 to overload octal and hexadecimal constants,
726 to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted
727 strings and here-documents,
731 to overload constant pieces of regular expressions.
735 The corresponding values are references to functions which take three arguments:
736 the first one is the I<initial> string form of the constant, the second one
737 is how Perl interprets this constant, the third one is how the constant is used.
738 Note that the initial string form does not
739 contain string delimiters, and has backslashes in backslash-delimiter
740 combinations stripped (thus the value of delimiter is not relevant for
741 processing of this string). The return value of this function is how this
742 constant is going to be interpreted by Perl. The third argument is undefined
743 unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote
744 context (comes from strings, regular expressions, and single-quote HERE
745 documents), it is C<tr> for arguments of C<tr>/C<y> operators,
746 it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise.
748 Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>,
749 it is expected that overloaded constant strings are equipped with reasonable
750 overloaded catenation operator, otherwise absurd results will result.
751 Similarly, negative numbers are considered as negations of positive constants.
753 Note that it is probably meaningless to call the functions overload::constant()
754 and overload::remove_constant() from anywhere but import() and unimport() methods.
755 From these methods they may be called as
760 die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant';
761 overload::constant integer => sub {Math::BigInt->new(shift)};
764 B<BUGS> Currently overloaded-ness of constants does not propagate
767 =head1 IMPLEMENTATION
769 What follows is subject to change RSN.
771 The table of methods for all operations is cached in magic for the
772 symbol table hash for the package. The cache is invalidated during
773 processing of C<use overload>, C<no overload>, new function
774 definitions, and changes in @ISA. However, this invalidation remains
775 unprocessed until the next C<bless>ing into the package. Hence if you
776 want to change overloading structure dynamically, you'll need an
777 additional (fake) C<bless>ing to update the table.
779 (Every SVish thing has a magic queue, and magic is an entry in that
780 queue. This is how a single variable may participate in multiple
781 forms of magic simultaneously. For instance, environment variables
782 regularly have two forms at once: their %ENV magic and their taint
783 magic. However, the magic which implements overloading is applied to
784 the stashes, which are rarely used directly, thus should not slow down
787 If an object belongs to a package using overload, it carries a special
788 flag. Thus the only speed penalty during arithmetic operations without
789 overloading is the checking of this flag.
791 In fact, if C<use overload> is not present, there is almost no overhead
792 for overloadable operations, so most programs should not suffer
793 measurable performance penalties. A considerable effort was made to
794 minimize the overhead when overload is used in some package, but the
795 arguments in question do not belong to packages using overload. When
796 in doubt, test your speed with C<use overload> and without it. So far
797 there have been no reports of substantial speed degradation if Perl is
798 compiled with optimization turned on.
800 There is no size penalty for data if overload is not used. The only
801 size penalty if overload is used in some package is that I<all> the
802 packages acquire a magic during the next C<bless>ing into the
803 package. This magic is three-words-long for packages without
804 overloading, and carries the cache table if the package is overloaded.
806 Copying (C<$a=$b>) is shallow; however, a one-level-deep copying is
807 carried out before any operation that can imply an assignment to the
808 object $a (or $b) refers to, like C<$a++>. You can override this
809 behavior by defining your own copy constructor (see L<"Copy Constructor">).
811 It is expected that arguments to methods that are not explicitly supposed
812 to be changed are constant (but this is not enforced).
814 =head1 Metaphor clash
816 One may wonder why the semantic of overloaded C<=> is so counter intuitive.
817 If it I<looks> counter intuitive to you, you are subject to a metaphor
820 Here is a Perl object metaphor:
822 I< object is a reference to blessed data>
824 and an arithmetic metaphor:
826 I< object is a thing by itself>.
828 The I<main> problem of overloading C<=> is the fact that these metaphors
829 imply different actions on the assignment C<$a = $b> if $a and $b are
830 objects. Perl-think implies that $a becomes a reference to whatever
831 $b was referencing. Arithmetic-think implies that the value of "object"
832 $a is changed to become the value of the object $b, preserving the fact
833 that $a and $b are separate entities.
835 The difference is not relevant in the absence of mutators. After
836 a Perl-way assignment an operation which mutates the data referenced by $a
837 would change the data referenced by $b too. Effectively, after
838 C<$a = $b> values of $a and $b become I<indistinguishable>.
840 On the other hand, anyone who has used algebraic notation knows the
841 expressive power of the arithmetic metaphor. Overloading works hard
842 to enable this metaphor while preserving the Perlian way as far as
843 possible. Since it is not not possible to freely mix two contradicting
844 metaphors, overloading allows the arithmetic way to write things I<as
845 far as all the mutators are called via overloaded access only>. The
846 way it is done is described in L<Copy Constructor>.
848 If some mutator methods are directly applied to the overloaded values,
849 one may need to I<explicitly unlink> other values which references the
854 $b = $a; # $b is "linked" to $a
856 $a = $a->clone; # Unlink $b from $a
859 Note that overloaded access makes this transparent:
862 $b = $a; # $b is "linked" to $a
863 $a += 4; # would unlink $b automagically
865 However, it would not make
868 $a = 4; # Now $a is a plain 4, not 'Data'
870 preserve "objectness" of $a. But Perl I<has> a way to make assignments
871 to an object do whatever you want. It is just not the overload, but
872 tie()ing interface (see L<perlfunc/tie>). Adding a FETCH() method
873 which returns the object itself, and STORE() method which changes the
874 value of the object, one can reproduce the arithmetic metaphor in its
875 completeness, at least for variables which were tie()d from the start.
877 (Note that a workaround for a bug may be needed, see L<"BUGS">.)
881 Please add examples to what follows!
883 =head2 Two-face scalars
885 Put this in F<two_face.pm> in your Perl library directory:
887 package two_face; # Scalars with separate string and
889 sub new { my $p = shift; bless [@_], $p }
890 use overload '""' => \&str, '0+' => \&num, fallback => 1;
897 my $seven = new two_face ("vii", 7);
898 printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1;
899 print "seven contains `i'\n" if $seven =~ /i/;
901 (The second line creates a scalar which has both a string value, and a
902 numeric value.) This prints:
904 seven=vii, seven=7, eight=8
907 =head2 Two-face references
909 Suppose you want to create an object which is accessible as both an
910 array reference and a hash reference, similar to the
911 L<pseudo-hash|perlref/"Pseudo-hashes: Using an array as a hash">
912 builtin Perl type. Let's make it better than a pseudo-hash by
913 allowing index 0 to be treated as a normal element.
916 use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} };
924 tie %h, ref $self, $self;
928 sub TIEHASH { my $p = shift; bless \ shift, $p }
931 $fields{$_} = $i++ foreach qw{zero one two three};
933 my $self = ${shift()};
934 my $key = $fields{shift()};
935 defined $key or die "Out of band access";
936 $$self->[$key] = shift;
939 my $self = ${shift()};
940 my $key = $fields{shift()};
941 defined $key or die "Out of band access";
945 Now one can access an object using both the array and hash syntax:
947 my $bar = new two_refs 3,4,5,6;
949 $bar->{two} == 11 or die 'bad hash fetch';
951 Note several important features of this example. First of all, the
952 I<actual> type of $bar is a scalar reference, and we do not overload
953 the scalar dereference. Thus we can get the I<actual> non-overloaded
954 contents of $bar by just using C<$$bar> (what we do in functions which
955 overload dereference). Similarly, the object returned by the
956 TIEHASH() method is a scalar reference.
958 Second, we create a new tied hash each time the hash syntax is used.
959 This allows us not to worry about a possibility of a reference loop,
960 would would lead to a memory leak.
962 Both these problems can be cured. Say, if we want to overload hash
963 dereference on a reference to an object which is I<implemented> as a
964 hash itself, the only problem one has to circumvent is how to access
965 this I<actual> hash (as opposed to the I<virtual> exhibited by
966 overloaded dereference operator). Here is one possible fetching routine:
969 my ($self, $key) = (shift, shift);
970 my $class = ref $self;
971 bless $self, 'overload::dummy'; # Disable overloading of %{}
972 my $out = $self->{$key};
973 bless $self, $class; # Restore overloading
977 To move creation of the tied hash on each access, one may an extra
978 level of indirection which allows a non-circular structure of references:
981 use overload '%{}' => sub { ${shift()}->[1] },
982 '@{}' => sub { ${shift()}->[0] };
988 bless \ [$a, \%h], $p;
993 tie %h, ref $self, $self;
997 sub TIEHASH { my $p = shift; bless \ shift, $p }
1000 $fields{$_} = $i++ foreach qw{zero one two three};
1003 my $key = $fields{shift()};
1004 defined $key or die "Out of band access";
1009 my $key = $fields{shift()};
1010 defined $key or die "Out of band access";
1014 Now if $baz is overloaded like this, then C<$bar> is a reference to a
1015 reference to the intermediate array, which keeps a reference to an
1016 actual array, and the access hash. The tie()ing object for the access
1017 hash is also a reference to a reference to the actual array, so
1023 There are no loops of references.
1027 Both "objects" which are blessed into the class C<two_refs1> are
1028 references to a reference to an array, thus references to a I<scalar>.
1029 Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no
1030 overloaded operations.
1034 =head2 Symbolic calculator
1036 Put this in F<symbolic.pm> in your Perl library directory:
1038 package symbolic; # Primitive symbolic calculator
1039 use overload nomethod => \&wrap;
1041 sub new { shift; bless ['n', @_] }
1043 my ($obj, $other, $inv, $meth) = @_;
1044 ($obj, $other) = ($other, $obj) if $inv;
1045 bless [$meth, $obj, $other];
1048 This module is very unusual as overloaded modules go: it does not
1049 provide any usual overloaded operators, instead it provides the L<Last
1050 Resort> operator C<nomethod>. In this example the corresponding
1051 subroutine returns an object which encapsulates operations done over
1052 the objects: C<new symbolic 3> contains C<['n', 3]>, C<2 + new
1053 symbolic 3> contains C<['+', 2, ['n', 3]]>.
1055 Here is an example of the script which "calculates" the side of
1056 circumscribed octagon using the above package:
1059 my $iter = 1; # 2**($iter+2) = 8
1060 my $side = new symbolic 1;
1064 $side = (sqrt(1 + $side**2) - 1)/$side;
1068 The value of $side is
1070 ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]],
1071 undef], 1], ['n', 1]]
1073 Note that while we obtained this value using a nice little script,
1074 there is no simple way to I<use> this value. In fact this value may
1075 be inspected in debugger (see L<perldebug>), but ony if
1076 C<bareStringify> B<O>ption is set, and not via C<p> command.
1078 If one attempts to print this value, then the overloaded operator
1079 C<""> will be called, which will call C<nomethod> operator. The
1080 result of this operator will be stringified again, but this result is
1081 again of type C<symbolic>, which will lead to an infinite loop.
1083 Add a pretty-printer method to the module F<symbolic.pm>:
1086 my ($meth, $a, $b) = @{+shift};
1087 $a = 'u' unless defined $a;
1088 $b = 'u' unless defined $b;
1089 $a = $a->pretty if ref $a;
1090 $b = $b->pretty if ref $b;
1094 Now one can finish the script by
1096 print "side = ", $side->pretty, "\n";
1098 The method C<pretty> is doing object-to-string conversion, so it
1099 is natural to overload the operator C<""> using this method. However,
1100 inside such a method it is not necessary to pretty-print the
1101 I<components> $a and $b of an object. In the above subroutine
1102 C<"[$meth $a $b]"> is a catenation of some strings and components $a
1103 and $b. If these components use overloading, the catenation operator
1104 will look for an overloaded operator C<.>, if not present, it will
1105 look for an overloaded operator C<"">. Thus it is enough to use
1107 use overload nomethod => \&wrap, '""' => \&str;
1109 my ($meth, $a, $b) = @{+shift};
1110 $a = 'u' unless defined $a;
1111 $b = 'u' unless defined $b;
1115 Now one can change the last line of the script to
1117 print "side = $side\n";
1121 side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]]
1123 and one can inspect the value in debugger using all the possible
1126 Something is is still amiss: consider the loop variable $cnt of the
1127 script. It was a number, not an object. We cannot make this value of
1128 type C<symbolic>, since then the loop will not terminate.
1130 Indeed, to terminate the cycle, the $cnt should become false.
1131 However, the operator C<bool> for checking falsity is overloaded (this
1132 time via overloaded C<"">), and returns a long string, thus any object
1133 of type C<symbolic> is true. To overcome this, we need a way to
1134 compare an object to 0. In fact, it is easier to write a numeric
1137 Here is the text of F<symbolic.pm> with such a routine added (and
1138 slightly modified str()):
1140 package symbolic; # Primitive symbolic calculator
1142 nomethod => \&wrap, '""' => \&str, '0+' => \#
1144 sub new { shift; bless ['n', @_] }
1146 my ($obj, $other, $inv, $meth) = @_;
1147 ($obj, $other) = ($other, $obj) if $inv;
1148 bless [$meth, $obj, $other];
1151 my ($meth, $a, $b) = @{+shift};
1152 $a = 'u' unless defined $a;
1159 my %subr = ( n => sub {$_[0]},
1160 sqrt => sub {sqrt $_[0]},
1161 '-' => sub {shift() - shift()},
1162 '+' => sub {shift() + shift()},
1163 '/' => sub {shift() / shift()},
1164 '*' => sub {shift() * shift()},
1165 '**' => sub {shift() ** shift()},
1168 my ($meth, $a, $b) = @{+shift};
1169 my $subr = $subr{$meth}
1170 or die "Do not know how to ($meth) in symbolic";
1171 $a = $a->num if ref $a eq __PACKAGE__;
1172 $b = $b->num if ref $b eq __PACKAGE__;
1176 All the work of numeric conversion is done in %subr and num(). Of
1177 course, %subr is not complete, it contains only operators used in the
1178 example below. Here is the extra-credit question: why do we need an
1179 explicit recursion in num()? (Answer is at the end of this section.)
1181 Use this module like this:
1184 my $iter = new symbolic 2; # 16-gon
1185 my $side = new symbolic 1;
1189 $cnt = $cnt - 1; # Mutator `--' not implemented
1190 $side = (sqrt(1 + $side**2) - 1)/$side;
1192 printf "%s=%f\n", $side, $side;
1193 printf "pi=%f\n", $side*(2**($iter+2));
1195 It prints (without so many line breaks)
1197 [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1]
1199 [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912
1202 The above module is very primitive. It does not implement
1203 mutator methods (C<++>, C<-=> and so on), does not do deep copying
1204 (not required without mutators!), and implements only those arithmetic
1205 operations which are used in the example.
1207 To implement most arithmetic operations is easy, one should just use
1208 the tables of operations, and change the code which fills %subr to
1210 my %subr = ( 'n' => sub {$_[0]} );
1211 foreach my $op (split " ", $overload::ops{with_assign}) {
1212 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1214 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1215 foreach my $op (split " ", "@overload::ops{ @bins }") {
1216 $subr{$op} = eval "sub {shift() $op shift()}";
1218 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1219 print "defining `$op'\n";
1220 $subr{$op} = eval "sub {$op shift()}";
1223 Due to L<Calling Conventions for Mutators>, we do not need anything
1224 special to make C<+=> and friends work, except filling C<+=> entry of
1225 %subr, and defining a copy constructor (needed since Perl has no
1226 way to know that the implementation of C<'+='> does not mutate
1227 the argument, compare L<Copy Constructor>).
1229 To implement a copy constructor, add C<'=' => \&cpy> to C<use overload>
1230 line, and code (this code assumes that mutators change things one level
1231 deep only, so recursive copying is not needed):
1235 bless [@$self], ref $self;
1238 To make C<++> and C<--> work, we need to implement actual mutators,
1239 either directly, or in C<nomethod>. We continue to do things inside
1240 C<nomethod>, thus add
1242 if ($meth eq '++' or $meth eq '--') {
1243 @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference
1247 after the first line of wrap(). This is not a most effective
1248 implementation, one may consider
1250 sub inc { $_[0] = bless ['++', shift, 1]; }
1254 As a final remark, note that one can fill %subr by
1256 my %subr = ( 'n' => sub {$_[0]} );
1257 foreach my $op (split " ", $overload::ops{with_assign}) {
1258 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1260 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1261 foreach my $op (split " ", "@overload::ops{ @bins }") {
1262 $subr{$op} = eval "sub {shift() $op shift()}";
1264 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1265 $subr{$op} = eval "sub {$op shift()}";
1267 $subr{'++'} = $subr{'+'};
1268 $subr{'--'} = $subr{'-'};
1270 This finishes implementation of a primitive symbolic calculator in
1271 50 lines of Perl code. Since the numeric values of subexpressions
1272 are not cached, the calculator is very slow.
1274 Here is the answer for the exercise: In the case of str(), we need no
1275 explicit recursion since the overloaded C<.>-operator will fall back
1276 to an existing overloaded operator C<"">. Overloaded arithmetic
1277 operators I<do not> fall back to numeric conversion if C<fallback> is
1278 not explicitly requested. Thus without an explicit recursion num()
1279 would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild
1280 the argument of num().
1282 If you wonder why defaults for conversion are different for str() and
1283 num(), note how easy it was to write the symbolic calculator. This
1284 simplicity is due to an appropriate choice of defaults. One extra
1285 note: due to the explicit recursion num() is more fragile than sym():
1286 we need to explicitly check for the type of $a and $b. If components
1287 $a and $b happen to be of some related type, this may lead to problems.
1289 =head2 I<Really> symbolic calculator
1291 One may wonder why we call the above calculator symbolic. The reason
1292 is that the actual calculation of the value of expression is postponed
1293 until the value is I<used>.
1295 To see it in action, add a method
1300 @$obj->[0,1] = ('=', shift);
1303 to the package C<symbolic>. After this change one can do
1305 my $a = new symbolic 3;
1306 my $b = new symbolic 4;
1307 my $c = sqrt($a**2 + $b**2);
1309 and the numeric value of $c becomes 5. However, after calling
1311 $a->STORE(12); $b->STORE(5);
1313 the numeric value of $c becomes 13. There is no doubt now that the module
1314 symbolic provides a I<symbolic> calculator indeed.
1316 To hide the rough edges under the hood, provide a tie()d interface to the
1317 package C<symbolic> (compare with L<Metaphor clash>). Add methods
1319 sub TIESCALAR { my $pack = shift; $pack->new(@_) }
1321 sub nop { } # Around a bug
1323 (the bug is described in L<"BUGS">). One can use this new interface as
1325 tie $a, 'symbolic', 3;
1326 tie $b, 'symbolic', 4;
1327 $a->nop; $b->nop; # Around a bug
1329 my $c = sqrt($a**2 + $b**2);
1331 Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value
1332 of $c becomes 13. To insulate the user of the module add a method
1334 sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; }
1339 symbolic->vars($a, $b);
1340 my $c = sqrt($a**2 + $b**2);
1343 printf "c5 %s=%f\n", $c, $c;
1346 printf "c13 %s=%f\n", $c, $c;
1348 shows that the numeric value of $c follows changes to the values of $a
1353 Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>.
1357 When Perl is run with the B<-Do> switch or its equivalent, overloading
1358 induces diagnostic messages.
1360 Using the C<m> command of Perl debugger (see L<perldebug>) one can
1361 deduce which operations are overloaded (and which ancestor triggers
1362 this overloading). Say, if C<eq> is overloaded, then the method C<(eq>
1363 is shown by debugger. The method C<()> corresponds to the C<fallback>
1364 key (in fact a presence of this method shows that this package has
1365 overloading enabled, and it is what is used by the C<Overloaded>
1366 function of module C<overload>).
1368 The module might issue the following warnings:
1372 =item Odd number of arguments for overload::constant
1374 (W) The call to overload::constant contained an odd number of arguments.
1375 The arguments should come in pairs.
1377 =item `%s' is not an overloadable type
1379 (W) You tried to overload a constant type the overload package is unaware of.
1381 =item `%s' is not a code reference
1383 (W) The second (fourth, sixth, ...) argument of overload::constant needs
1384 to be a code reference. Either an anonymous subroutine, or a reference
1391 Because it is used for overloading, the per-package hash %OVERLOAD now
1392 has a special meaning in Perl. The symbol table is filled with names
1393 looking like line-noise.
1395 For the purpose of inheritance every overloaded package behaves as if
1396 C<fallback> is present (possibly undefined). This may create
1397 interesting effects if some package is not overloaded, but inherits
1398 from two overloaded packages.
1400 Relation between overloading and tie()ing is broken. Overloading is
1401 triggered or not basing on the I<previous> class of tie()d value.
1403 This happens because the presence of overloading is checked too early,
1404 before any tie()d access is attempted. If the FETCH()ed class of the
1405 tie()d value does not change, a simple workaround is to access the value
1406 immediately after tie()ing, so that after this call the I<previous> class
1407 coincides with the current one.
1409 B<Needed:> a way to fix this without a speed penalty.
1411 Barewords are not covered by overloaded string constants.
1413 This document is confusing. There are grammos and misleading language
1414 used in places. It would seem a total rewrite is needed.