5 $overload::hint_bits = 0x20000;
13 $ {$package . "::OVERLOAD"}{dummy}++; # Register with magic by touching.
14 *{$package . "::()"} = \&nil; # Make it findable via fetchmethod.
16 if ($_ eq 'fallback') {
20 if (not ref $sub and $sub !~ /::/) {
21 $ {$package . "::(" . $_} = $sub;
24 #print STDERR "Setting `$ {'package'}::\cO$_' to \\&`$sub'.\n";
25 *{$package . "::(" . $_} = \&{ $sub };
28 ${$package . "::()"} = $fb; # Make it findable too (fallback only).
32 $package = (caller())[0];
33 # *{$package . "::OVERLOAD"} = \&OVERLOAD;
35 $package->overload::OVERLOAD(@_);
39 $package = (caller())[0];
40 ${$package . "::OVERLOAD"}{dummy}++; # Upgrade the table
43 if ($_ eq 'fallback') {
44 undef $ {$package . "::()"};
46 delete $ {$package . "::"}{"(" . $_};
53 $package = ref $package if ref $package;
59 return undef unless $globref;
60 my $sub = \&{*$globref};
61 return $sub if $sub ne \&nil;
62 return shift->can($ {*$globref});
65 sub OverloadedStringify {
67 $package = ref $package if ref $package;
69 ov_method mycan($package, '(""'), $package
70 or ov_method mycan($package, '(0+'), $package
71 or ov_method mycan($package, '(bool'), $package
72 or ov_method mycan($package, '(nomethod'), $package;
77 $package = ref $package if ref $package;
78 #my $meth = $package->can('(' . shift);
79 ov_method mycan($package, '(' . shift), $package;
80 #return $meth if $meth ne \&nil;
85 my $package = ref $_[0];
86 return "$_[0]" unless $package;
87 bless $_[0], overload::Fake; # Non-overloaded package
89 bless $_[0], $package; # Back
90 $package . substr $str, index $str, '=';
94 (OverloadedStringify($_[0]) or ref($_[0]) eq 'Regexp') ?
99 sub mycan { # Real can would leave stubs.
100 my ($package, $meth) = @_;
101 return \*{$package . "::$meth"} if defined &{$package . "::$meth"};
103 foreach $p (@{$package . "::ISA"}) {
104 my $out = mycan($p, $meth);
118 %ops = ( with_assign => "+ - * / % ** << >> x .",
119 assign => "+= -= *= /= %= **= <<= >>= x= .=",
120 num_comparison => "< <= > >= == !=",
121 '3way_comparison'=> "<=> cmp",
122 str_comparison => "lt le gt ge eq ne",
126 func => "atan2 cos sin exp abs log sqrt",
127 conversion => 'bool "" 0+',
129 dereferencing => '${} @{} %{} &{} *{}',
130 special => 'nomethod fallback =');
132 use warnings::register;
134 # Arguments: what, sub
137 warnings::warnif ("Odd number of arguments for overload::constant");
140 elsif (!exists $constants {$_ [0]}) {
141 warnings::warnif ("`$_[0]' is not an overloadable type");
143 elsif (!ref $_ [1] || "$_[1]" !~ /CODE\(0x[\da-f]+\)$/) {
144 # Can't use C<ref $_[1] eq "CODE"> above as code references can be
145 # blessed, and C<ref> would return the package the ref is blessed into.
146 if (warnings::enabled) {
147 $_ [1] = "undef" unless defined $_ [1];
148 warnings::warn ("`$_[1]' is not a code reference");
153 $^H |= $constants{$_[0]} | $overload::hint_bits;
159 sub remove_constant {
160 # Arguments: what, sub
163 $^H &= ~ $constants{$_[0]};
174 overload - Package for overloading perl operations
187 $a = new SomeThing 57;
190 if (overload::Overloaded $b) {...}
192 $strval = overload::StrVal $b;
196 =head2 Declaration of overloaded functions
198 The compilation directive
205 declares function Number::add() for addition, and method muas() in
206 the "class" C<Number> (or one of its base classes)
207 for the assignment form C<*=> of multiplication.
209 Arguments of this directive come in (key, value) pairs. Legal values
210 are values legal inside a C<&{ ... }> call, so the name of a
211 subroutine, a reference to a subroutine, or an anonymous subroutine
212 will all work. Note that values specified as strings are
213 interpreted as methods, not subroutines. Legal keys are listed below.
215 The subroutine C<add> will be called to execute C<$a+$b> if $a
216 is a reference to an object blessed into the package C<Number>, or if $a is
217 not an object from a package with defined mathemagic addition, but $b is a
218 reference to a C<Number>. It can also be called in other situations, like
219 C<$a+=7>, or C<$a++>. See L<MAGIC AUTOGENERATION>. (Mathemagical
220 methods refer to methods triggered by an overloaded mathematical
223 Since overloading respects inheritance via the @ISA hierarchy, the
224 above declaration would also trigger overloading of C<+> and C<*=> in
225 all the packages which inherit from C<Number>.
227 =head2 Calling Conventions for Binary Operations
229 The functions specified in the C<use overload ...> directive are called
230 with three (in one particular case with four, see L<Last Resort>)
231 arguments. If the corresponding operation is binary, then the first
232 two arguments are the two arguments of the operation. However, due to
233 general object calling conventions, the first argument should always be
234 an object in the package, so in the situation of C<7+$a>, the
235 order of the arguments is interchanged. It probably does not matter
236 when implementing the addition method, but whether the arguments
237 are reversed is vital to the subtraction method. The method can
238 query this information by examining the third argument, which can take
239 three different values:
245 the order of arguments is as in the current operation.
249 the arguments are reversed.
253 the current operation is an assignment variant (as in
254 C<$a+=7>), but the usual function is called instead. This additional
255 information can be used to generate some optimizations. Compare
256 L<Calling Conventions for Mutators>.
260 =head2 Calling Conventions for Unary Operations
262 Unary operation are considered binary operations with the second
263 argument being C<undef>. Thus the functions that overloads C<{"++"}>
264 is called with arguments C<($a,undef,'')> when $a++ is executed.
266 =head2 Calling Conventions for Mutators
268 Two types of mutators have different calling conventions:
272 =item C<++> and C<-->
274 The routines which implement these operators are expected to actually
275 I<mutate> their arguments. So, assuming that $obj is a reference to a
278 sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n}
280 is an appropriate implementation of overloaded C<++>. Note that
282 sub incr { ++$ {$_[0]} ; shift }
284 is OK if used with preincrement and with postincrement. (In the case
285 of postincrement a copying will be performed, see L<Copy Constructor>.)
287 =item C<x=> and other assignment versions
289 There is nothing special about these methods. They may change the
290 value of their arguments, and may leave it as is. The result is going
291 to be assigned to the value in the left-hand-side if different from
294 This allows for the same method to be used as overloaded C<+=> and
295 C<+>. Note that this is I<allowed>, but not recommended, since by the
296 semantic of L<"Fallback"> Perl will call the method for C<+> anyway,
297 if C<+=> is not overloaded.
301 B<Warning.> Due to the presense of assignment versions of operations,
302 routines which may be called in assignment context may create
303 self-referential structures. Currently Perl will not free self-referential
304 structures until cycles are C<explicitly> broken. You may get problems
305 when traversing your structures too.
309 use overload '+' => sub { bless [ \$_[0], \$_[1] ] };
311 is asking for trouble, since for code C<$obj += $foo> the subroutine
312 is called as C<$obj = add($obj, $foo, undef)>, or C<$obj = [\$obj,
313 \$foo]>. If using such a subroutine is an important optimization, one
314 can overload C<+=> explicitly by a non-"optimized" version, or switch
315 to non-optimized version if C<not defined $_[2]> (see
316 L<Calling Conventions for Binary Operations>).
318 Even if no I<explicit> assignment-variants of operators are present in
319 the script, they may be generated by the optimizer. Say, C<",$obj,"> or
320 C<',' . $obj . ','> may be both optimized to
322 my $tmp = ',' . $obj; $tmp .= ',';
324 =head2 Overloadable Operations
326 The following symbols can be specified in C<use overload> directive:
330 =item * I<Arithmetic operations>
332 "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=",
333 "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=",
335 For these operations a substituted non-assignment variant can be called if
336 the assignment variant is not available. Methods for operations "C<+>",
337 "C<->", "C<+=>", and "C<-=>" can be called to automatically generate
338 increment and decrement methods. The operation "C<->" can be used to
339 autogenerate missing methods for unary minus or C<abs>.
341 See L<"MAGIC AUTOGENERATION">, L<"Calling Conventions for Mutators"> and
342 L<"Calling Conventions for Binary Operations">) for details of these
345 =item * I<Comparison operations>
347 "<", "<=", ">", ">=", "==", "!=", "<=>",
348 "lt", "le", "gt", "ge", "eq", "ne", "cmp",
350 If the corresponding "spaceship" variant is available, it can be
351 used to substitute for the missing operation. During C<sort>ing
352 arrays, C<cmp> is used to compare values subject to C<use overload>.
354 =item * I<Bit operations>
356 "&", "^", "|", "neg", "!", "~",
358 "C<neg>" stands for unary minus. If the method for C<neg> is not
359 specified, it can be autogenerated using the method for
360 subtraction. If the method for "C<!>" is not specified, it can be
361 autogenerated using the methods for "C<bool>", or "C<\"\">", or "C<0+>".
363 =item * I<Increment and decrement>
367 If undefined, addition and subtraction methods can be
368 used instead. These operations are called both in prefix and
371 =item * I<Transcendental functions>
373 "atan2", "cos", "sin", "exp", "abs", "log", "sqrt",
375 If C<abs> is unavailable, it can be autogenerated using methods
376 for "E<lt>" or "E<lt>=E<gt>" combined with either unary minus or subtraction.
378 =item * I<Boolean, string and numeric conversion>
380 "bool", "\"\"", "0+",
382 If one or two of these operations are not overloaded, the remaining ones can
383 be used instead. C<bool> is used in the flow control operators
384 (like C<while>) and for the ternary "C<?:>" operation. These functions can
385 return any arbitrary Perl value. If the corresponding operation for this value
386 is overloaded too, that operation will be called again with this value.
388 As a special case if the overload returns the object itself then it will
389 be used directly. An overloaded conversion returning the object is
390 probably a bug, because you're likely to get something that looks like
391 C<YourPackage=HASH(0x8172b34)>.
397 If not overloaded, the argument will be converted to a filehandle or
398 glob (which may require a stringification). The same overloading
399 happens both for the I<read-filehandle> syntax C<E<lt>$varE<gt>> and
400 I<globbing> syntax C<E<lt>${var}E<gt>>.
402 =item * I<Dereferencing>
404 '${}', '@{}', '%{}', '&{}', '*{}'.
406 If not overloaded, the argument will be dereferenced I<as is>, thus
407 should be of correct type. These functions should return a reference
408 of correct type, or another object with overloaded dereferencing.
410 As a special case if the overload returns the object itself then it
411 will be used directly (provided it is the correct type).
413 The dereference operators must be specified explicitly they will not be passed to
418 "nomethod", "fallback", "=",
420 see L<SPECIAL SYMBOLS FOR C<use overload>>.
424 See L<"Fallback"> for an explanation of when a missing method can be
427 A computer-readable form of the above table is available in the hash
428 %overload::ops, with values being space-separated lists of names:
430 with_assign => '+ - * / % ** << >> x .',
431 assign => '+= -= *= /= %= **= <<= >>= x= .=',
432 num_comparison => '< <= > >= == !=',
433 '3way_comparison'=> '<=> cmp',
434 str_comparison => 'lt le gt ge eq ne',
438 func => 'atan2 cos sin exp abs log sqrt',
439 conversion => 'bool "" 0+',
441 dereferencing => '${} @{} %{} &{} *{}',
442 special => 'nomethod fallback ='
444 =head2 Inheritance and overloading
446 Inheritance interacts with overloading in two ways.
450 =item Strings as values of C<use overload> directive
454 use overload key => value;
456 is a string, it is interpreted as a method name.
458 =item Overloading of an operation is inherited by derived classes
460 Any class derived from an overloaded class is also overloaded. The
461 set of overloaded methods is the union of overloaded methods of all
462 the ancestors. If some method is overloaded in several ancestor, then
463 which description will be used is decided by the usual inheritance
466 If C<A> inherits from C<B> and C<C> (in this order), C<B> overloads
467 C<+> with C<\&D::plus_sub>, and C<C> overloads C<+> by C<"plus_meth">,
468 then the subroutine C<D::plus_sub> will be called to implement
469 operation C<+> for an object in package C<A>.
473 Note that since the value of the C<fallback> key is not a subroutine,
474 its inheritance is not governed by the above rules. In the current
475 implementation, the value of C<fallback> in the first overloaded
476 ancestor is used, but this is accidental and subject to change.
478 =head1 SPECIAL SYMBOLS FOR C<use overload>
480 Three keys are recognized by Perl that are not covered by the above
485 C<"nomethod"> should be followed by a reference to a function of four
486 parameters. If defined, it is called when the overloading mechanism
487 cannot find a method for some operation. The first three arguments of
488 this function coincide with the arguments for the corresponding method if
489 it were found, the fourth argument is the symbol
490 corresponding to the missing method. If several methods are tried,
491 the last one is used. Say, C<1-$a> can be equivalent to
493 &nomethodMethod($a,1,1,"-")
495 if the pair C<"nomethod" =E<gt> "nomethodMethod"> was specified in the
496 C<use overload> directive.
498 The C<"nomethod"> mechanism is I<not> used for the dereference operators
499 ( ${} @{} %{} &{} *{} ).
502 If some operation cannot be resolved, and there is no function
503 assigned to C<"nomethod">, then an exception will be raised via die()--
504 unless C<"fallback"> was specified as a key in C<use overload> directive.
509 The key C<"fallback"> governs what to do if a method for a particular
510 operation is not found. Three different cases are possible depending on
511 the value of C<"fallback">:
518 substituted method (see L<MAGIC AUTOGENERATION>). If this fails, it
519 then tries to calls C<"nomethod"> value; if missing, an exception
524 The same as for the C<undef> value, but no exception is raised. Instead,
525 it silently reverts to what it would have done were there no C<use overload>
528 =item * defined, but FALSE
530 No autogeneration is tried. Perl tries to call
531 C<"nomethod"> value, and if this is missing, raises an exception.
535 B<Note.> C<"fallback"> inheritance via @ISA is not carved in stone
536 yet, see L<"Inheritance and overloading">.
538 =head2 Copy Constructor
540 The value for C<"="> is a reference to a function with three
541 arguments, i.e., it looks like the other values in C<use
542 overload>. However, it does not overload the Perl assignment
543 operator. This would go against Camel hair.
545 This operation is called in the situations when a mutator is applied
546 to a reference that shares its object with some other reference, such
552 To make this change $a and not change $b, a copy of C<$$a> is made,
553 and $a is assigned a reference to this new object. This operation is
554 done during execution of the C<++$a>, and not during the assignment,
555 (so before the increment C<$$a> coincides with C<$$b>). This is only
556 done if C<++> is expressed via a method for C<'++'> or C<'+='> (or
557 C<nomethod>). Note that if this operation is expressed via C<'+'>
558 a nonmutator, i.e., as in
563 then C<$a> does not reference a new copy of C<$$a>, since $$a does not
564 appear as lvalue when the above code is executed.
566 If the copy constructor is required during the execution of some mutator,
567 but a method for C<'='> was not specified, it can be autogenerated as a
568 string copy if the object is a plain scalar.
574 The actually executed code for
577 Something else which does not modify $a or $b....
583 Something else which does not modify $a or $b....
584 $a = $a->clone(undef,"");
587 if $b was mathemagical, and C<'++'> was overloaded with C<\&incr>,
588 C<'='> was overloaded with C<\&clone>.
592 Same behaviour is triggered by C<$b = $a++>, which is consider a synonym for
595 =head1 MAGIC AUTOGENERATION
597 If a method for an operation is not found, and the value for C<"fallback"> is
598 TRUE or undefined, Perl tries to autogenerate a substitute method for
599 the missing operation based on the defined operations. Autogenerated method
600 substitutions are possible for the following operations:
604 =item I<Assignment forms of arithmetic operations>
606 C<$a+=$b> can use the method for C<"+"> if the method for C<"+=">
609 =item I<Conversion operations>
611 String, numeric, and boolean conversion are calculated in terms of one
612 another if not all of them are defined.
614 =item I<Increment and decrement>
616 The C<++$a> operation can be expressed in terms of C<$a+=1> or C<$a+1>,
617 and C<$a--> in terms of C<$a-=1> and C<$a-1>.
621 can be expressed in terms of C<$aE<lt>0> and C<-$a> (or C<0-$a>).
625 can be expressed in terms of subtraction.
629 C<!> and C<not> can be expressed in terms of boolean conversion, or
630 string or numerical conversion.
632 =item I<Concatenation>
634 can be expressed in terms of string conversion.
636 =item I<Comparison operations>
638 can be expressed in terms of its "spaceship" counterpart: either
639 C<E<lt>=E<gt>> or C<cmp>:
641 <, >, <=, >=, ==, != in terms of <=>
642 lt, gt, le, ge, eq, ne in terms of cmp
646 <> in terms of builtin operations
648 =item I<Dereferencing>
650 ${} @{} %{} &{} *{} in terms of builtin operations
652 =item I<Copy operator>
654 can be expressed in terms of an assignment to the dereferenced value, if this
655 value is a scalar and not a reference.
659 =head1 Losing overloading
661 The restriction for the comparison operation is that even if, for example,
662 `C<cmp>' should return a blessed reference, the autogenerated `C<lt>'
663 function will produce only a standard logical value based on the
664 numerical value of the result of `C<cmp>'. In particular, a working
665 numeric conversion is needed in this case (possibly expressed in terms of
668 Similarly, C<.=> and C<x=> operators lose their mathemagical properties
669 if the string conversion substitution is applied.
671 When you chop() a mathemagical object it is promoted to a string and its
672 mathemagical properties are lost. The same can happen with other
675 =head1 Run-time Overloading
677 Since all C<use> directives are executed at compile-time, the only way to
678 change overloading during run-time is to
680 eval 'use overload "+" => \&addmethod';
684 eval 'no overload "+", "--", "<="';
686 though the use of these constructs during run-time is questionable.
688 =head1 Public functions
690 Package C<overload.pm> provides the following public functions:
694 =item overload::StrVal(arg)
696 Gives string value of C<arg> as in absence of stringify overloading.
698 =item overload::Overloaded(arg)
700 Returns true if C<arg> is subject to overloading of some operations.
702 =item overload::Method(obj,op)
704 Returns C<undef> or a reference to the method that implements C<op>.
708 =head1 Overloading constants
710 For some application Perl parser mangles constants too much. It is possible
711 to hook into this process via overload::constant() and overload::remove_constant()
714 These functions take a hash as an argument. The recognized keys of this hash
721 to overload integer constants,
725 to overload floating point constants,
729 to overload octal and hexadecimal constants,
733 to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted
734 strings and here-documents,
738 to overload constant pieces of regular expressions.
742 The corresponding values are references to functions which take three arguments:
743 the first one is the I<initial> string form of the constant, the second one
744 is how Perl interprets this constant, the third one is how the constant is used.
745 Note that the initial string form does not
746 contain string delimiters, and has backslashes in backslash-delimiter
747 combinations stripped (thus the value of delimiter is not relevant for
748 processing of this string). The return value of this function is how this
749 constant is going to be interpreted by Perl. The third argument is undefined
750 unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote
751 context (comes from strings, regular expressions, and single-quote HERE
752 documents), it is C<tr> for arguments of C<tr>/C<y> operators,
753 it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise.
755 Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>,
756 it is expected that overloaded constant strings are equipped with reasonable
757 overloaded catenation operator, otherwise absurd results will result.
758 Similarly, negative numbers are considered as negations of positive constants.
760 Note that it is probably meaningless to call the functions overload::constant()
761 and overload::remove_constant() from anywhere but import() and unimport() methods.
762 From these methods they may be called as
767 die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant';
768 overload::constant integer => sub {Math::BigInt->new(shift)};
771 B<BUGS> Currently overloaded-ness of constants does not propagate
774 =head1 IMPLEMENTATION
776 What follows is subject to change RSN.
778 The table of methods for all operations is cached in magic for the
779 symbol table hash for the package. The cache is invalidated during
780 processing of C<use overload>, C<no overload>, new function
781 definitions, and changes in @ISA. However, this invalidation remains
782 unprocessed until the next C<bless>ing into the package. Hence if you
783 want to change overloading structure dynamically, you'll need an
784 additional (fake) C<bless>ing to update the table.
786 (Every SVish thing has a magic queue, and magic is an entry in that
787 queue. This is how a single variable may participate in multiple
788 forms of magic simultaneously. For instance, environment variables
789 regularly have two forms at once: their %ENV magic and their taint
790 magic. However, the magic which implements overloading is applied to
791 the stashes, which are rarely used directly, thus should not slow down
794 If an object belongs to a package using overload, it carries a special
795 flag. Thus the only speed penalty during arithmetic operations without
796 overloading is the checking of this flag.
798 In fact, if C<use overload> is not present, there is almost no overhead
799 for overloadable operations, so most programs should not suffer
800 measurable performance penalties. A considerable effort was made to
801 minimize the overhead when overload is used in some package, but the
802 arguments in question do not belong to packages using overload. When
803 in doubt, test your speed with C<use overload> and without it. So far
804 there have been no reports of substantial speed degradation if Perl is
805 compiled with optimization turned on.
807 There is no size penalty for data if overload is not used. The only
808 size penalty if overload is used in some package is that I<all> the
809 packages acquire a magic during the next C<bless>ing into the
810 package. This magic is three-words-long for packages without
811 overloading, and carries the cache table if the package is overloaded.
813 Copying (C<$a=$b>) is shallow; however, a one-level-deep copying is
814 carried out before any operation that can imply an assignment to the
815 object $a (or $b) refers to, like C<$a++>. You can override this
816 behavior by defining your own copy constructor (see L<"Copy Constructor">).
818 It is expected that arguments to methods that are not explicitly supposed
819 to be changed are constant (but this is not enforced).
821 =head1 Metaphor clash
823 One may wonder why the semantic of overloaded C<=> is so counter intuitive.
824 If it I<looks> counter intuitive to you, you are subject to a metaphor
827 Here is a Perl object metaphor:
829 I< object is a reference to blessed data>
831 and an arithmetic metaphor:
833 I< object is a thing by itself>.
835 The I<main> problem of overloading C<=> is the fact that these metaphors
836 imply different actions on the assignment C<$a = $b> if $a and $b are
837 objects. Perl-think implies that $a becomes a reference to whatever
838 $b was referencing. Arithmetic-think implies that the value of "object"
839 $a is changed to become the value of the object $b, preserving the fact
840 that $a and $b are separate entities.
842 The difference is not relevant in the absence of mutators. After
843 a Perl-way assignment an operation which mutates the data referenced by $a
844 would change the data referenced by $b too. Effectively, after
845 C<$a = $b> values of $a and $b become I<indistinguishable>.
847 On the other hand, anyone who has used algebraic notation knows the
848 expressive power of the arithmetic metaphor. Overloading works hard
849 to enable this metaphor while preserving the Perlian way as far as
850 possible. Since it is not not possible to freely mix two contradicting
851 metaphors, overloading allows the arithmetic way to write things I<as
852 far as all the mutators are called via overloaded access only>. The
853 way it is done is described in L<Copy Constructor>.
855 If some mutator methods are directly applied to the overloaded values,
856 one may need to I<explicitly unlink> other values which references the
861 $b = $a; # $b is "linked" to $a
863 $a = $a->clone; # Unlink $b from $a
866 Note that overloaded access makes this transparent:
869 $b = $a; # $b is "linked" to $a
870 $a += 4; # would unlink $b automagically
872 However, it would not make
875 $a = 4; # Now $a is a plain 4, not 'Data'
877 preserve "objectness" of $a. But Perl I<has> a way to make assignments
878 to an object do whatever you want. It is just not the overload, but
879 tie()ing interface (see L<perlfunc/tie>). Adding a FETCH() method
880 which returns the object itself, and STORE() method which changes the
881 value of the object, one can reproduce the arithmetic metaphor in its
882 completeness, at least for variables which were tie()d from the start.
884 (Note that a workaround for a bug may be needed, see L<"BUGS">.)
888 Please add examples to what follows!
890 =head2 Two-face scalars
892 Put this in F<two_face.pm> in your Perl library directory:
894 package two_face; # Scalars with separate string and
896 sub new { my $p = shift; bless [@_], $p }
897 use overload '""' => \&str, '0+' => \&num, fallback => 1;
904 my $seven = new two_face ("vii", 7);
905 printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1;
906 print "seven contains `i'\n" if $seven =~ /i/;
908 (The second line creates a scalar which has both a string value, and a
909 numeric value.) This prints:
911 seven=vii, seven=7, eight=8
914 =head2 Two-face references
916 Suppose you want to create an object which is accessible as both an
917 array reference and a hash reference, similar to the
918 L<pseudo-hash|perlref/"Pseudo-hashes: Using an array as a hash">
919 builtin Perl type. Let's make it better than a pseudo-hash by
920 allowing index 0 to be treated as a normal element.
923 use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} };
931 tie %h, ref $self, $self;
935 sub TIEHASH { my $p = shift; bless \ shift, $p }
938 $fields{$_} = $i++ foreach qw{zero one two three};
940 my $self = ${shift()};
941 my $key = $fields{shift()};
942 defined $key or die "Out of band access";
943 $$self->[$key] = shift;
946 my $self = ${shift()};
947 my $key = $fields{shift()};
948 defined $key or die "Out of band access";
952 Now one can access an object using both the array and hash syntax:
954 my $bar = new two_refs 3,4,5,6;
956 $bar->{two} == 11 or die 'bad hash fetch';
958 Note several important features of this example. First of all, the
959 I<actual> type of $bar is a scalar reference, and we do not overload
960 the scalar dereference. Thus we can get the I<actual> non-overloaded
961 contents of $bar by just using C<$$bar> (what we do in functions which
962 overload dereference). Similarly, the object returned by the
963 TIEHASH() method is a scalar reference.
965 Second, we create a new tied hash each time the hash syntax is used.
966 This allows us not to worry about a possibility of a reference loop,
967 would would lead to a memory leak.
969 Both these problems can be cured. Say, if we want to overload hash
970 dereference on a reference to an object which is I<implemented> as a
971 hash itself, the only problem one has to circumvent is how to access
972 this I<actual> hash (as opposed to the I<virtual> exhibited by
973 overloaded dereference operator). Here is one possible fetching routine:
976 my ($self, $key) = (shift, shift);
977 my $class = ref $self;
978 bless $self, 'overload::dummy'; # Disable overloading of %{}
979 my $out = $self->{$key};
980 bless $self, $class; # Restore overloading
984 To move creation of the tied hash on each access, one may an extra
985 level of indirection which allows a non-circular structure of references:
988 use overload '%{}' => sub { ${shift()}->[1] },
989 '@{}' => sub { ${shift()}->[0] };
995 bless \ [$a, \%h], $p;
1000 tie %h, ref $self, $self;
1004 sub TIEHASH { my $p = shift; bless \ shift, $p }
1007 $fields{$_} = $i++ foreach qw{zero one two three};
1010 my $key = $fields{shift()};
1011 defined $key or die "Out of band access";
1016 my $key = $fields{shift()};
1017 defined $key or die "Out of band access";
1021 Now if $baz is overloaded like this, then C<$bar> is a reference to a
1022 reference to the intermediate array, which keeps a reference to an
1023 actual array, and the access hash. The tie()ing object for the access
1024 hash is also a reference to a reference to the actual array, so
1030 There are no loops of references.
1034 Both "objects" which are blessed into the class C<two_refs1> are
1035 references to a reference to an array, thus references to a I<scalar>.
1036 Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no
1037 overloaded operations.
1041 =head2 Symbolic calculator
1043 Put this in F<symbolic.pm> in your Perl library directory:
1045 package symbolic; # Primitive symbolic calculator
1046 use overload nomethod => \&wrap;
1048 sub new { shift; bless ['n', @_] }
1050 my ($obj, $other, $inv, $meth) = @_;
1051 ($obj, $other) = ($other, $obj) if $inv;
1052 bless [$meth, $obj, $other];
1055 This module is very unusual as overloaded modules go: it does not
1056 provide any usual overloaded operators, instead it provides the L<Last
1057 Resort> operator C<nomethod>. In this example the corresponding
1058 subroutine returns an object which encapsulates operations done over
1059 the objects: C<new symbolic 3> contains C<['n', 3]>, C<2 + new
1060 symbolic 3> contains C<['+', 2, ['n', 3]]>.
1062 Here is an example of the script which "calculates" the side of
1063 circumscribed octagon using the above package:
1066 my $iter = 1; # 2**($iter+2) = 8
1067 my $side = new symbolic 1;
1071 $side = (sqrt(1 + $side**2) - 1)/$side;
1075 The value of $side is
1077 ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]],
1078 undef], 1], ['n', 1]]
1080 Note that while we obtained this value using a nice little script,
1081 there is no simple way to I<use> this value. In fact this value may
1082 be inspected in debugger (see L<perldebug>), but ony if
1083 C<bareStringify> B<O>ption is set, and not via C<p> command.
1085 If one attempts to print this value, then the overloaded operator
1086 C<""> will be called, which will call C<nomethod> operator. The
1087 result of this operator will be stringified again, but this result is
1088 again of type C<symbolic>, which will lead to an infinite loop.
1090 Add a pretty-printer method to the module F<symbolic.pm>:
1093 my ($meth, $a, $b) = @{+shift};
1094 $a = 'u' unless defined $a;
1095 $b = 'u' unless defined $b;
1096 $a = $a->pretty if ref $a;
1097 $b = $b->pretty if ref $b;
1101 Now one can finish the script by
1103 print "side = ", $side->pretty, "\n";
1105 The method C<pretty> is doing object-to-string conversion, so it
1106 is natural to overload the operator C<""> using this method. However,
1107 inside such a method it is not necessary to pretty-print the
1108 I<components> $a and $b of an object. In the above subroutine
1109 C<"[$meth $a $b]"> is a catenation of some strings and components $a
1110 and $b. If these components use overloading, the catenation operator
1111 will look for an overloaded operator C<.>, if not present, it will
1112 look for an overloaded operator C<"">. Thus it is enough to use
1114 use overload nomethod => \&wrap, '""' => \&str;
1116 my ($meth, $a, $b) = @{+shift};
1117 $a = 'u' unless defined $a;
1118 $b = 'u' unless defined $b;
1122 Now one can change the last line of the script to
1124 print "side = $side\n";
1128 side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]]
1130 and one can inspect the value in debugger using all the possible
1133 Something is is still amiss: consider the loop variable $cnt of the
1134 script. It was a number, not an object. We cannot make this value of
1135 type C<symbolic>, since then the loop will not terminate.
1137 Indeed, to terminate the cycle, the $cnt should become false.
1138 However, the operator C<bool> for checking falsity is overloaded (this
1139 time via overloaded C<"">), and returns a long string, thus any object
1140 of type C<symbolic> is true. To overcome this, we need a way to
1141 compare an object to 0. In fact, it is easier to write a numeric
1144 Here is the text of F<symbolic.pm> with such a routine added (and
1145 slightly modified str()):
1147 package symbolic; # Primitive symbolic calculator
1149 nomethod => \&wrap, '""' => \&str, '0+' => \#
1151 sub new { shift; bless ['n', @_] }
1153 my ($obj, $other, $inv, $meth) = @_;
1154 ($obj, $other) = ($other, $obj) if $inv;
1155 bless [$meth, $obj, $other];
1158 my ($meth, $a, $b) = @{+shift};
1159 $a = 'u' unless defined $a;
1166 my %subr = ( n => sub {$_[0]},
1167 sqrt => sub {sqrt $_[0]},
1168 '-' => sub {shift() - shift()},
1169 '+' => sub {shift() + shift()},
1170 '/' => sub {shift() / shift()},
1171 '*' => sub {shift() * shift()},
1172 '**' => sub {shift() ** shift()},
1175 my ($meth, $a, $b) = @{+shift};
1176 my $subr = $subr{$meth}
1177 or die "Do not know how to ($meth) in symbolic";
1178 $a = $a->num if ref $a eq __PACKAGE__;
1179 $b = $b->num if ref $b eq __PACKAGE__;
1183 All the work of numeric conversion is done in %subr and num(). Of
1184 course, %subr is not complete, it contains only operators used in the
1185 example below. Here is the extra-credit question: why do we need an
1186 explicit recursion in num()? (Answer is at the end of this section.)
1188 Use this module like this:
1191 my $iter = new symbolic 2; # 16-gon
1192 my $side = new symbolic 1;
1196 $cnt = $cnt - 1; # Mutator `--' not implemented
1197 $side = (sqrt(1 + $side**2) - 1)/$side;
1199 printf "%s=%f\n", $side, $side;
1200 printf "pi=%f\n", $side*(2**($iter+2));
1202 It prints (without so many line breaks)
1204 [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1]
1206 [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912
1209 The above module is very primitive. It does not implement
1210 mutator methods (C<++>, C<-=> and so on), does not do deep copying
1211 (not required without mutators!), and implements only those arithmetic
1212 operations which are used in the example.
1214 To implement most arithmetic operations is easy, one should just use
1215 the tables of operations, and change the code which fills %subr to
1217 my %subr = ( 'n' => sub {$_[0]} );
1218 foreach my $op (split " ", $overload::ops{with_assign}) {
1219 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1221 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1222 foreach my $op (split " ", "@overload::ops{ @bins }") {
1223 $subr{$op} = eval "sub {shift() $op shift()}";
1225 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1226 print "defining `$op'\n";
1227 $subr{$op} = eval "sub {$op shift()}";
1230 Due to L<Calling Conventions for Mutators>, we do not need anything
1231 special to make C<+=> and friends work, except filling C<+=> entry of
1232 %subr, and defining a copy constructor (needed since Perl has no
1233 way to know that the implementation of C<'+='> does not mutate
1234 the argument, compare L<Copy Constructor>).
1236 To implement a copy constructor, add C<'=' => \&cpy> to C<use overload>
1237 line, and code (this code assumes that mutators change things one level
1238 deep only, so recursive copying is not needed):
1242 bless [@$self], ref $self;
1245 To make C<++> and C<--> work, we need to implement actual mutators,
1246 either directly, or in C<nomethod>. We continue to do things inside
1247 C<nomethod>, thus add
1249 if ($meth eq '++' or $meth eq '--') {
1250 @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference
1254 after the first line of wrap(). This is not a most effective
1255 implementation, one may consider
1257 sub inc { $_[0] = bless ['++', shift, 1]; }
1261 As a final remark, note that one can fill %subr by
1263 my %subr = ( 'n' => sub {$_[0]} );
1264 foreach my $op (split " ", $overload::ops{with_assign}) {
1265 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1267 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1268 foreach my $op (split " ", "@overload::ops{ @bins }") {
1269 $subr{$op} = eval "sub {shift() $op shift()}";
1271 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1272 $subr{$op} = eval "sub {$op shift()}";
1274 $subr{'++'} = $subr{'+'};
1275 $subr{'--'} = $subr{'-'};
1277 This finishes implementation of a primitive symbolic calculator in
1278 50 lines of Perl code. Since the numeric values of subexpressions
1279 are not cached, the calculator is very slow.
1281 Here is the answer for the exercise: In the case of str(), we need no
1282 explicit recursion since the overloaded C<.>-operator will fall back
1283 to an existing overloaded operator C<"">. Overloaded arithmetic
1284 operators I<do not> fall back to numeric conversion if C<fallback> is
1285 not explicitly requested. Thus without an explicit recursion num()
1286 would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild
1287 the argument of num().
1289 If you wonder why defaults for conversion are different for str() and
1290 num(), note how easy it was to write the symbolic calculator. This
1291 simplicity is due to an appropriate choice of defaults. One extra
1292 note: due to the explicit recursion num() is more fragile than sym():
1293 we need to explicitly check for the type of $a and $b. If components
1294 $a and $b happen to be of some related type, this may lead to problems.
1296 =head2 I<Really> symbolic calculator
1298 One may wonder why we call the above calculator symbolic. The reason
1299 is that the actual calculation of the value of expression is postponed
1300 until the value is I<used>.
1302 To see it in action, add a method
1307 @$obj->[0,1] = ('=', shift);
1310 to the package C<symbolic>. After this change one can do
1312 my $a = new symbolic 3;
1313 my $b = new symbolic 4;
1314 my $c = sqrt($a**2 + $b**2);
1316 and the numeric value of $c becomes 5. However, after calling
1318 $a->STORE(12); $b->STORE(5);
1320 the numeric value of $c becomes 13. There is no doubt now that the module
1321 symbolic provides a I<symbolic> calculator indeed.
1323 To hide the rough edges under the hood, provide a tie()d interface to the
1324 package C<symbolic> (compare with L<Metaphor clash>). Add methods
1326 sub TIESCALAR { my $pack = shift; $pack->new(@_) }
1328 sub nop { } # Around a bug
1330 (the bug is described in L<"BUGS">). One can use this new interface as
1332 tie $a, 'symbolic', 3;
1333 tie $b, 'symbolic', 4;
1334 $a->nop; $b->nop; # Around a bug
1336 my $c = sqrt($a**2 + $b**2);
1338 Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value
1339 of $c becomes 13. To insulate the user of the module add a method
1341 sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; }
1346 symbolic->vars($a, $b);
1347 my $c = sqrt($a**2 + $b**2);
1350 printf "c5 %s=%f\n", $c, $c;
1353 printf "c13 %s=%f\n", $c, $c;
1355 shows that the numeric value of $c follows changes to the values of $a
1360 Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>.
1364 When Perl is run with the B<-Do> switch or its equivalent, overloading
1365 induces diagnostic messages.
1367 Using the C<m> command of Perl debugger (see L<perldebug>) one can
1368 deduce which operations are overloaded (and which ancestor triggers
1369 this overloading). Say, if C<eq> is overloaded, then the method C<(eq>
1370 is shown by debugger. The method C<()> corresponds to the C<fallback>
1371 key (in fact a presence of this method shows that this package has
1372 overloading enabled, and it is what is used by the C<Overloaded>
1373 function of module C<overload>).
1375 The module might issue the following warnings:
1379 =item Odd number of arguments for overload::constant
1381 (W) The call to overload::constant contained an odd number of arguments.
1382 The arguments should come in pairs.
1384 =item `%s' is not an overloadable type
1386 (W) You tried to overload a constant type the overload package is unaware of.
1388 =item `%s' is not a code reference
1390 (W) The second (fourth, sixth, ...) argument of overload::constant needs
1391 to be a code reference. Either an anonymous subroutine, or a reference
1398 Because it is used for overloading, the per-package hash %OVERLOAD now
1399 has a special meaning in Perl. The symbol table is filled with names
1400 looking like line-noise.
1402 For the purpose of inheritance every overloaded package behaves as if
1403 C<fallback> is present (possibly undefined). This may create
1404 interesting effects if some package is not overloaded, but inherits
1405 from two overloaded packages.
1407 Relation between overloading and tie()ing is broken. Overloading is
1408 triggered or not basing on the I<previous> class of tie()d value.
1410 This happens because the presence of overloading is checked too early,
1411 before any tie()d access is attempted. If the FETCH()ed class of the
1412 tie()d value does not change, a simple workaround is to access the value
1413 immediately after tie()ing, so that after this call the I<previous> class
1414 coincides with the current one.
1416 B<Needed:> a way to fix this without a speed penalty.
1418 Barewords are not covered by overloaded string constants.
1420 This document is confusing. There are grammos and misleading language
1421 used in places. It would seem a total rewrite is needed.