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 =');
131 # Arguments: what, sub
134 $^H |= $constants{$_[0]} | $overload::hint_bits;
139 sub remove_constant {
140 # Arguments: what, sub
143 $^H &= ~ $constants{$_[0]};
154 overload - Package for overloading perl operations
167 $a = new SomeThing 57;
170 if (overload::Overloaded $b) {...}
172 $strval = overload::StrVal $b;
176 =head2 Declaration of overloaded functions
178 The compilation directive
185 declares function Number::add() for addition, and method muas() in
186 the "class" C<Number> (or one of its base classes)
187 for the assignment form C<*=> of multiplication.
189 Arguments of this directive come in (key, value) pairs. Legal values
190 are values legal inside a C<&{ ... }> call, so the name of a
191 subroutine, a reference to a subroutine, or an anonymous subroutine
192 will all work. Note that values specified as strings are
193 interpreted as methods, not subroutines. Legal keys are listed below.
195 The subroutine C<add> will be called to execute C<$a+$b> if $a
196 is a reference to an object blessed into the package C<Number>, or if $a is
197 not an object from a package with defined mathemagic addition, but $b is a
198 reference to a C<Number>. It can also be called in other situations, like
199 C<$a+=7>, or C<$a++>. See L<MAGIC AUTOGENERATION>. (Mathemagical
200 methods refer to methods triggered by an overloaded mathematical
203 Since overloading respects inheritance via the @ISA hierarchy, the
204 above declaration would also trigger overloading of C<+> and C<*=> in
205 all the packages which inherit from C<Number>.
207 =head2 Calling Conventions for Binary Operations
209 The functions specified in the C<use overload ...> directive are called
210 with three (in one particular case with four, see L<Last Resort>)
211 arguments. If the corresponding operation is binary, then the first
212 two arguments are the two arguments of the operation. However, due to
213 general object calling conventions, the first argument should always be
214 an object in the package, so in the situation of C<7+$a>, the
215 order of the arguments is interchanged. It probably does not matter
216 when implementing the addition method, but whether the arguments
217 are reversed is vital to the subtraction method. The method can
218 query this information by examining the third argument, which can take
219 three different values:
225 the order of arguments is as in the current operation.
229 the arguments are reversed.
233 the current operation is an assignment variant (as in
234 C<$a+=7>), but the usual function is called instead. This additional
235 information can be used to generate some optimizations. Compare
236 L<Calling Conventions for Mutators>.
240 =head2 Calling Conventions for Unary Operations
242 Unary operation are considered binary operations with the second
243 argument being C<undef>. Thus the functions that overloads C<{"++"}>
244 is called with arguments C<($a,undef,'')> when $a++ is executed.
246 =head2 Calling Conventions for Mutators
248 Two types of mutators have different calling conventions:
252 =item C<++> and C<-->
254 The routines which implement these operators are expected to actually
255 I<mutate> their arguments. So, assuming that $obj is a reference to a
258 sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n}
260 is an appropriate implementation of overloaded C<++>. Note that
262 sub incr { ++$ {$_[0]} ; shift }
264 is OK if used with preincrement and with postincrement. (In the case
265 of postincrement a copying will be performed, see L<Copy Constructor>.)
267 =item C<x=> and other assignment versions
269 There is nothing special about these methods. They may change the
270 value of their arguments, and may leave it as is. The result is going
271 to be assigned to the value in the left-hand-side if different from
274 This allows for the same method to be used as overloaded C<+=> and
275 C<+>. Note that this is I<allowed>, but not recommended, since by the
276 semantic of L<"Fallback"> Perl will call the method for C<+> anyway,
277 if C<+=> is not overloaded.
281 B<Warning.> Due to the presense of assignment versions of operations,
282 routines which may be called in assignment context may create
283 self-referential structures. Currently Perl will not free self-referential
284 structures until cycles are C<explicitly> broken. You may get problems
285 when traversing your structures too.
289 use overload '+' => sub { bless [ \$_[0], \$_[1] ] };
291 is asking for trouble, since for code C<$obj += $foo> the subroutine
292 is called as C<$obj = add($obj, $foo, undef)>, or C<$obj = [\$obj,
293 \$foo]>. If using such a subroutine is an important optimization, one
294 can overload C<+=> explicitly by a non-"optimized" version, or switch
295 to non-optimized version if C<not defined $_[2]> (see
296 L<Calling Conventions for Binary Operations>).
298 Even if no I<explicit> assignment-variants of operators are present in
299 the script, they may be generated by the optimizer. Say, C<",$obj,"> or
300 C<',' . $obj . ','> may be both optimized to
302 my $tmp = ',' . $obj; $tmp .= ',';
304 =head2 Overloadable Operations
306 The following symbols can be specified in C<use overload> directive:
310 =item * I<Arithmetic operations>
312 "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=",
313 "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=",
315 For these operations a substituted non-assignment variant can be called if
316 the assignment variant is not available. Methods for operations "C<+>",
317 "C<->", "C<+=>", and "C<-=>" can be called to automatically generate
318 increment and decrement methods. The operation "C<->" can be used to
319 autogenerate missing methods for unary minus or C<abs>.
321 See L<"MAGIC AUTOGENERATION">, L<"Calling Conventions for Mutators"> and
322 L<"Calling Conventions for Binary Operations">) for details of these
325 =item * I<Comparison operations>
327 "<", "<=", ">", ">=", "==", "!=", "<=>",
328 "lt", "le", "gt", "ge", "eq", "ne", "cmp",
330 If the corresponding "spaceship" variant is available, it can be
331 used to substitute for the missing operation. During C<sort>ing
332 arrays, C<cmp> is used to compare values subject to C<use overload>.
334 =item * I<Bit operations>
336 "&", "^", "|", "neg", "!", "~",
338 "C<neg>" stands for unary minus. If the method for C<neg> is not
339 specified, it can be autogenerated using the method for
340 subtraction. If the method for "C<!>" is not specified, it can be
341 autogenerated using the methods for "C<bool>", or "C<\"\">", or "C<0+>".
343 =item * I<Increment and decrement>
347 If undefined, addition and subtraction methods can be
348 used instead. These operations are called both in prefix and
351 =item * I<Transcendental functions>
353 "atan2", "cos", "sin", "exp", "abs", "log", "sqrt",
355 If C<abs> is unavailable, it can be autogenerated using methods
356 for "E<lt>" or "E<lt>=E<gt>" combined with either unary minus or subtraction.
358 =item * I<Boolean, string and numeric conversion>
360 "bool", "\"\"", "0+",
362 If one or two of these operations are not overloaded, the remaining ones can
363 be used instead. C<bool> is used in the flow control operators
364 (like C<while>) and for the ternary "C<?:>" operation. These functions can
365 return any arbitrary Perl value. If the corresponding operation for this value
366 is overloaded too, that operation will be called again with this value.
372 If not overloaded, the argument will be converted to a filehandle or
373 glob (which may require a stringification). The same overloading
374 happens both for the I<read-filehandle> syntax C<E<lt>$varE<gt>> and
375 I<globbing> syntax C<E<lt>${var}E<gt>>.
377 =item * I<Dereferencing>
379 '${}', '@{}', '%{}', '&{}', '*{}'.
381 If not overloaded, the argument will be dereferenced I<as is>, thus
382 should be of correct type. These functions should return a reference
383 of correct type, or another object with overloaded dereferencing.
387 "nomethod", "fallback", "=",
389 see L<SPECIAL SYMBOLS FOR C<use overload>>.
393 See L<"Fallback"> for an explanation of when a missing method can be
396 A computer-readable form of the above table is available in the hash
397 %overload::ops, with values being space-separated lists of names:
399 with_assign => '+ - * / % ** << >> x .',
400 assign => '+= -= *= /= %= **= <<= >>= x= .=',
401 num_comparison => '< <= > >= == !=',
402 '3way_comparison'=> '<=> cmp',
403 str_comparison => 'lt le gt ge eq ne',
407 func => 'atan2 cos sin exp abs log sqrt',
408 conversion => 'bool "" 0+',
410 dereferencing => '${} @{} %{} &{} *{}',
411 special => 'nomethod fallback ='
413 =head2 Inheritance and overloading
415 Inheritance interacts with overloading in two ways.
419 =item Strings as values of C<use overload> directive
423 use overload key => value;
425 is a string, it is interpreted as a method name.
427 =item Overloading of an operation is inherited by derived classes
429 Any class derived from an overloaded class is also overloaded. The
430 set of overloaded methods is the union of overloaded methods of all
431 the ancestors. If some method is overloaded in several ancestor, then
432 which description will be used is decided by the usual inheritance
435 If C<A> inherits from C<B> and C<C> (in this order), C<B> overloads
436 C<+> with C<\&D::plus_sub>, and C<C> overloads C<+> by C<"plus_meth">,
437 then the subroutine C<D::plus_sub> will be called to implement
438 operation C<+> for an object in package C<A>.
442 Note that since the value of the C<fallback> key is not a subroutine,
443 its inheritance is not governed by the above rules. In the current
444 implementation, the value of C<fallback> in the first overloaded
445 ancestor is used, but this is accidental and subject to change.
447 =head1 SPECIAL SYMBOLS FOR C<use overload>
449 Three keys are recognized by Perl that are not covered by the above
454 C<"nomethod"> should be followed by a reference to a function of four
455 parameters. If defined, it is called when the overloading mechanism
456 cannot find a method for some operation. The first three arguments of
457 this function coincide with the arguments for the corresponding method if
458 it were found, the fourth argument is the symbol
459 corresponding to the missing method. If several methods are tried,
460 the last one is used. Say, C<1-$a> can be equivalent to
462 &nomethodMethod($a,1,1,"-")
464 if the pair C<"nomethod" =E<gt> "nomethodMethod"> was specified in the
465 C<use overload> directive.
467 If some operation cannot be resolved, and there is no function
468 assigned to C<"nomethod">, then an exception will be raised via die()--
469 unless C<"fallback"> was specified as a key in C<use overload> directive.
473 The key C<"fallback"> governs what to do if a method for a particular
474 operation is not found. Three different cases are possible depending on
475 the value of C<"fallback">:
482 substituted method (see L<MAGIC AUTOGENERATION>). If this fails, it
483 then tries to calls C<"nomethod"> value; if missing, an exception
488 The same as for the C<undef> value, but no exception is raised. Instead,
489 it silently reverts to what it would have done were there no C<use overload>
492 =item * defined, but FALSE
494 No autogeneration is tried. Perl tries to call
495 C<"nomethod"> value, and if this is missing, raises an exception.
499 B<Note.> C<"fallback"> inheritance via @ISA is not carved in stone
500 yet, see L<"Inheritance and overloading">.
502 =head2 Copy Constructor
504 The value for C<"="> is a reference to a function with three
505 arguments, i.e., it looks like the other values in C<use
506 overload>. However, it does not overload the Perl assignment
507 operator. This would go against Camel hair.
509 This operation is called in the situations when a mutator is applied
510 to a reference that shares its object with some other reference, such
516 To make this change $a and not change $b, a copy of C<$$a> is made,
517 and $a is assigned a reference to this new object. This operation is
518 done during execution of the C<++$a>, and not during the assignment,
519 (so before the increment C<$$a> coincides with C<$$b>). This is only
520 done if C<++> is expressed via a method for C<'++'> or C<'+='> (or
521 C<nomethod>). Note that if this operation is expressed via C<'+'>
522 a nonmutator, i.e., as in
527 then C<$a> does not reference a new copy of C<$$a>, since $$a does not
528 appear as lvalue when the above code is executed.
530 If the copy constructor is required during the execution of some mutator,
531 but a method for C<'='> was not specified, it can be autogenerated as a
532 string copy if the object is a plain scalar.
538 The actually executed code for
541 Something else which does not modify $a or $b....
547 Something else which does not modify $a or $b....
548 $a = $a->clone(undef,"");
551 if $b was mathemagical, and C<'++'> was overloaded with C<\&incr>,
552 C<'='> was overloaded with C<\&clone>.
556 Same behaviour is triggered by C<$b = $a++>, which is consider a synonym for
559 =head1 MAGIC AUTOGENERATION
561 If a method for an operation is not found, and the value for C<"fallback"> is
562 TRUE or undefined, Perl tries to autogenerate a substitute method for
563 the missing operation based on the defined operations. Autogenerated method
564 substitutions are possible for the following operations:
568 =item I<Assignment forms of arithmetic operations>
570 C<$a+=$b> can use the method for C<"+"> if the method for C<"+=">
573 =item I<Conversion operations>
575 String, numeric, and boolean conversion are calculated in terms of one
576 another if not all of them are defined.
578 =item I<Increment and decrement>
580 The C<++$a> operation can be expressed in terms of C<$a+=1> or C<$a+1>,
581 and C<$a--> in terms of C<$a-=1> and C<$a-1>.
585 can be expressed in terms of C<$aE<lt>0> and C<-$a> (or C<0-$a>).
589 can be expressed in terms of subtraction.
593 C<!> and C<not> can be expressed in terms of boolean conversion, or
594 string or numerical conversion.
596 =item I<Concatenation>
598 can be expressed in terms of string conversion.
600 =item I<Comparison operations>
602 can be expressed in terms of its "spaceship" counterpart: either
603 C<E<lt>=E<gt>> or C<cmp>:
605 <, >, <=, >=, ==, != in terms of <=>
606 lt, gt, le, ge, eq, ne in terms of cmp
610 <> in terms of builtin operations
612 =item I<Dereferencing>
614 ${} @{} %{} &{} *{} in terms of builtin operations
616 =item I<Copy operator>
618 can be expressed in terms of an assignment to the dereferenced value, if this
619 value is a scalar and not a reference.
623 =head1 Losing overloading
625 The restriction for the comparison operation is that even if, for example,
626 `C<cmp>' should return a blessed reference, the autogenerated `C<lt>'
627 function will produce only a standard logical value based on the
628 numerical value of the result of `C<cmp>'. In particular, a working
629 numeric conversion is needed in this case (possibly expressed in terms of
632 Similarly, C<.=> and C<x=> operators lose their mathemagical properties
633 if the string conversion substitution is applied.
635 When you chop() a mathemagical object it is promoted to a string and its
636 mathemagical properties are lost. The same can happen with other
639 =head1 Run-time Overloading
641 Since all C<use> directives are executed at compile-time, the only way to
642 change overloading during run-time is to
644 eval 'use overload "+" => \&addmethod';
648 eval 'no overload "+", "--", "<="';
650 though the use of these constructs during run-time is questionable.
652 =head1 Public functions
654 Package C<overload.pm> provides the following public functions:
658 =item overload::StrVal(arg)
660 Gives string value of C<arg> as in absence of stringify overloading.
662 =item overload::Overloaded(arg)
664 Returns true if C<arg> is subject to overloading of some operations.
666 =item overload::Method(obj,op)
668 Returns C<undef> or a reference to the method that implements C<op>.
672 =head1 Overloading constants
674 For some application Perl parser mangles constants too much. It is possible
675 to hook into this process via overload::constant() and overload::remove_constant()
678 These functions take a hash as an argument. The recognized keys of this hash
685 to overload integer constants,
689 to overload floating point constants,
693 to overload octal and hexadecimal constants,
697 to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted
698 strings and here-documents,
702 to overload constant pieces of regular expressions.
706 The corresponding values are references to functions which take three arguments:
707 the first one is the I<initial> string form of the constant, the second one
708 is how Perl interprets this constant, the third one is how the constant is used.
709 Note that the initial string form does not
710 contain string delimiters, and has backslashes in backslash-delimiter
711 combinations stripped (thus the value of delimiter is not relevant for
712 processing of this string). The return value of this function is how this
713 constant is going to be interpreted by Perl. The third argument is undefined
714 unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote
715 context (comes from strings, regular expressions, and single-quote HERE
716 documents), it is C<tr> for arguments of C<tr>/C<y> operators,
717 it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise.
719 Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>,
720 it is expected that overloaded constant strings are equipped with reasonable
721 overloaded catenation operator, otherwise absurd results will result.
722 Similarly, negative numbers are considered as negations of positive constants.
724 Note that it is probably meaningless to call the functions overload::constant()
725 and overload::remove_constant() from anywhere but import() and unimport() methods.
726 From these methods they may be called as
731 die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant';
732 overload::constant integer => sub {Math::BigInt->new(shift)};
735 B<BUGS> Currently overloaded-ness of constants does not propagate
738 =head1 IMPLEMENTATION
740 What follows is subject to change RSN.
742 The table of methods for all operations is cached in magic for the
743 symbol table hash for the package. The cache is invalidated during
744 processing of C<use overload>, C<no overload>, new function
745 definitions, and changes in @ISA. However, this invalidation remains
746 unprocessed until the next C<bless>ing into the package. Hence if you
747 want to change overloading structure dynamically, you'll need an
748 additional (fake) C<bless>ing to update the table.
750 (Every SVish thing has a magic queue, and magic is an entry in that
751 queue. This is how a single variable may participate in multiple
752 forms of magic simultaneously. For instance, environment variables
753 regularly have two forms at once: their %ENV magic and their taint
754 magic. However, the magic which implements overloading is applied to
755 the stashes, which are rarely used directly, thus should not slow down
758 If an object belongs to a package using overload, it carries a special
759 flag. Thus the only speed penalty during arithmetic operations without
760 overloading is the checking of this flag.
762 In fact, if C<use overload> is not present, there is almost no overhead
763 for overloadable operations, so most programs should not suffer
764 measurable performance penalties. A considerable effort was made to
765 minimize the overhead when overload is used in some package, but the
766 arguments in question do not belong to packages using overload. When
767 in doubt, test your speed with C<use overload> and without it. So far
768 there have been no reports of substantial speed degradation if Perl is
769 compiled with optimization turned on.
771 There is no size penalty for data if overload is not used. The only
772 size penalty if overload is used in some package is that I<all> the
773 packages acquire a magic during the next C<bless>ing into the
774 package. This magic is three-words-long for packages without
775 overloading, and carries the cache table if the package is overloaded.
777 Copying (C<$a=$b>) is shallow; however, a one-level-deep copying is
778 carried out before any operation that can imply an assignment to the
779 object $a (or $b) refers to, like C<$a++>. You can override this
780 behavior by defining your own copy constructor (see L<"Copy Constructor">).
782 It is expected that arguments to methods that are not explicitly supposed
783 to be changed are constant (but this is not enforced).
785 =head1 Metaphor clash
787 One may wonder why the semantic of overloaded C<=> is so counter intuitive.
788 If it I<looks> counter intuitive to you, you are subject to a metaphor
791 Here is a Perl object metaphor:
793 I< object is a reference to blessed data>
795 and an arithmetic metaphor:
797 I< object is a thing by itself>.
799 The I<main> problem of overloading C<=> is the fact that these metaphors
800 imply different actions on the assignment C<$a = $b> if $a and $b are
801 objects. Perl-think implies that $a becomes a reference to whatever
802 $b was referencing. Arithmetic-think implies that the value of "object"
803 $a is changed to become the value of the object $b, preserving the fact
804 that $a and $b are separate entities.
806 The difference is not relevant in the absence of mutators. After
807 a Perl-way assignment an operation which mutates the data referenced by $a
808 would change the data referenced by $b too. Effectively, after
809 C<$a = $b> values of $a and $b become I<indistinguishable>.
811 On the other hand, anyone who has used algebraic notation knows the
812 expressive power of the arithmetic metaphor. Overloading works hard
813 to enable this metaphor while preserving the Perlian way as far as
814 possible. Since it is not not possible to freely mix two contradicting
815 metaphors, overloading allows the arithmetic way to write things I<as
816 far as all the mutators are called via overloaded access only>. The
817 way it is done is described in L<Copy Constructor>.
819 If some mutator methods are directly applied to the overloaded values,
820 one may need to I<explicitly unlink> other values which references the
825 $b = $a; # $b is "linked" to $a
827 $a = $a->clone; # Unlink $b from $a
830 Note that overloaded access makes this transparent:
833 $b = $a; # $b is "linked" to $a
834 $a += 4; # would unlink $b automagically
836 However, it would not make
839 $a = 4; # Now $a is a plain 4, not 'Data'
841 preserve "objectness" of $a. But Perl I<has> a way to make assignments
842 to an object do whatever you want. It is just not the overload, but
843 tie()ing interface (see L<perlfunc/tie>). Adding a FETCH() method
844 which returns the object itself, and STORE() method which changes the
845 value of the object, one can reproduce the arithmetic metaphor in its
846 completeness, at least for variables which were tie()d from the start.
848 (Note that a workaround for a bug may be needed, see L<"BUGS">.)
852 Please add examples to what follows!
854 =head2 Two-face scalars
856 Put this in F<two_face.pm> in your Perl library directory:
858 package two_face; # Scalars with separate string and
860 sub new { my $p = shift; bless [@_], $p }
861 use overload '""' => \&str, '0+' => \&num, fallback => 1;
868 my $seven = new two_face ("vii", 7);
869 printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1;
870 print "seven contains `i'\n" if $seven =~ /i/;
872 (The second line creates a scalar which has both a string value, and a
873 numeric value.) This prints:
875 seven=vii, seven=7, eight=8
878 =head2 Two-face references
880 Suppose you want to create an object which is accessible as both an
881 array reference, and a hash reference, similar to the builtin
882 L<array-accessible-as-a-hash|perlref/"Pseudo-hashes: Using an array as
883 a hash"> builtin Perl type. Let us make it better than the builtin
884 type, there will be no restriction that you cannot use the index 0 of
888 use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} };
896 tie %h, ref $self, $self;
900 sub TIEHASH { my $p = shift; bless \ shift, $p }
903 $fields{$_} = $i++ foreach qw{zero one two three};
905 my $self = ${shift()};
906 my $key = $fields{shift()};
907 defined $key or die "Out of band access";
908 $$self->[$key] = shift;
911 my $self = ${shift()};
912 my $key = $fields{shift()};
913 defined $key or die "Out of band access";
917 Now one can access an object using both the array and hash syntax:
919 my $bar = new two_refs 3,4,5,6;
921 $bar->{two} == 11 or die 'bad hash fetch';
923 Note several important features of this example. First of all, the
924 I<actual> type of $bar is a scalar reference, and we do not overload
925 the scalar dereference. Thus we can get the I<actual> non-overloaded
926 contents of $bar by just using C<$$bar> (what we do in functions which
927 overload dereference). Similarly, the object returned by the
928 TIEHASH() method is a scalar reference.
930 Second, we create a new tied hash each time the hash syntax is used.
931 This allows us not to worry about a possibility of a reference loop,
932 would would lead to a memory leak.
934 Both these problems can be cured. Say, if we want to overload hash
935 dereference on a reference to an object which is I<implemented> as a
936 hash itself, the only problem one has to circumvent is how to access
937 this I<actual> hash (as opposed to the I<virtual> exhibited by
938 overloaded dereference operator). Here is one possible fetching routine:
941 my ($self, $key) = (shift, shift);
942 my $class = ref $self;
943 bless $self, 'overload::dummy'; # Disable overloading of %{}
944 my $out = $self->{$key};
945 bless $self, $class; # Restore overloading
949 To move creation of the tied hash on each access, one may an extra
950 level of indirection which allows a non-circular structure of references:
953 use overload '%{}' => sub { ${shift()}->[1] },
954 '@{}' => sub { ${shift()}->[0] };
960 bless \ [$a, \%h], $p;
965 tie %h, ref $self, $self;
969 sub TIEHASH { my $p = shift; bless \ shift, $p }
972 $fields{$_} = $i++ foreach qw{zero one two three};
975 my $key = $fields{shift()};
976 defined $key or die "Out of band access";
981 my $key = $fields{shift()};
982 defined $key or die "Out of band access";
986 Now if $baz is overloaded like this, then C<$bar> is a reference to a
987 reference to the intermediate array, which keeps a reference to an
988 actual array, and the access hash. The tie()ing object for the access
989 hash is also a reference to a reference to the actual array, so
995 There are no loops of references.
999 Both "objects" which are blessed into the class C<two_refs1> are
1000 references to a reference to an array, thus references to a I<scalar>.
1001 Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no
1002 overloaded operations.
1006 =head2 Symbolic calculator
1008 Put this in F<symbolic.pm> in your Perl library directory:
1010 package symbolic; # Primitive symbolic calculator
1011 use overload nomethod => \&wrap;
1013 sub new { shift; bless ['n', @_] }
1015 my ($obj, $other, $inv, $meth) = @_;
1016 ($obj, $other) = ($other, $obj) if $inv;
1017 bless [$meth, $obj, $other];
1020 This module is very unusual as overloaded modules go: it does not
1021 provide any usual overloaded operators, instead it provides the L<Last
1022 Resort> operator C<nomethod>. In this example the corresponding
1023 subroutine returns an object which encapsulates operations done over
1024 the objects: C<new symbolic 3> contains C<['n', 3]>, C<2 + new
1025 symbolic 3> contains C<['+', 2, ['n', 3]]>.
1027 Here is an example of the script which "calculates" the side of
1028 circumscribed octagon using the above package:
1031 my $iter = 1; # 2**($iter+2) = 8
1032 my $side = new symbolic 1;
1036 $side = (sqrt(1 + $side**2) - 1)/$side;
1040 The value of $side is
1042 ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]],
1043 undef], 1], ['n', 1]]
1045 Note that while we obtained this value using a nice little script,
1046 there is no simple way to I<use> this value. In fact this value may
1047 be inspected in debugger (see L<perldebug>), but ony if
1048 C<bareStringify> B<O>ption is set, and not via C<p> command.
1050 If one attempts to print this value, then the overloaded operator
1051 C<""> will be called, which will call C<nomethod> operator. The
1052 result of this operator will be stringified again, but this result is
1053 again of type C<symbolic>, which will lead to an infinite loop.
1055 Add a pretty-printer method to the module F<symbolic.pm>:
1058 my ($meth, $a, $b) = @{+shift};
1059 $a = 'u' unless defined $a;
1060 $b = 'u' unless defined $b;
1061 $a = $a->pretty if ref $a;
1062 $b = $b->pretty if ref $b;
1066 Now one can finish the script by
1068 print "side = ", $side->pretty, "\n";
1070 The method C<pretty> is doing object-to-string conversion, so it
1071 is natural to overload the operator C<""> using this method. However,
1072 inside such a method it is not necessary to pretty-print the
1073 I<components> $a and $b of an object. In the above subroutine
1074 C<"[$meth $a $b]"> is a catenation of some strings and components $a
1075 and $b. If these components use overloading, the catenation operator
1076 will look for an overloaded operator C<.>, if not present, it will
1077 look for an overloaded operator C<"">. Thus it is enough to use
1079 use overload nomethod => \&wrap, '""' => \&str;
1081 my ($meth, $a, $b) = @{+shift};
1082 $a = 'u' unless defined $a;
1083 $b = 'u' unless defined $b;
1087 Now one can change the last line of the script to
1089 print "side = $side\n";
1093 side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]]
1095 and one can inspect the value in debugger using all the possible
1098 Something is is still amiss: consider the loop variable $cnt of the
1099 script. It was a number, not an object. We cannot make this value of
1100 type C<symbolic>, since then the loop will not terminate.
1102 Indeed, to terminate the cycle, the $cnt should become false.
1103 However, the operator C<bool> for checking falsity is overloaded (this
1104 time via overloaded C<"">), and returns a long string, thus any object
1105 of type C<symbolic> is true. To overcome this, we need a way to
1106 compare an object to 0. In fact, it is easier to write a numeric
1109 Here is the text of F<symbolic.pm> with such a routine added (and
1110 slightly modified str()):
1112 package symbolic; # Primitive symbolic calculator
1114 nomethod => \&wrap, '""' => \&str, '0+' => \#
1116 sub new { shift; bless ['n', @_] }
1118 my ($obj, $other, $inv, $meth) = @_;
1119 ($obj, $other) = ($other, $obj) if $inv;
1120 bless [$meth, $obj, $other];
1123 my ($meth, $a, $b) = @{+shift};
1124 $a = 'u' unless defined $a;
1131 my %subr = ( n => sub {$_[0]},
1132 sqrt => sub {sqrt $_[0]},
1133 '-' => sub {shift() - shift()},
1134 '+' => sub {shift() + shift()},
1135 '/' => sub {shift() / shift()},
1136 '*' => sub {shift() * shift()},
1137 '**' => sub {shift() ** shift()},
1140 my ($meth, $a, $b) = @{+shift};
1141 my $subr = $subr{$meth}
1142 or die "Do not know how to ($meth) in symbolic";
1143 $a = $a->num if ref $a eq __PACKAGE__;
1144 $b = $b->num if ref $b eq __PACKAGE__;
1148 All the work of numeric conversion is done in %subr and num(). Of
1149 course, %subr is not complete, it contains only operators used in the
1150 example below. Here is the extra-credit question: why do we need an
1151 explicit recursion in num()? (Answer is at the end of this section.)
1153 Use this module like this:
1156 my $iter = new symbolic 2; # 16-gon
1157 my $side = new symbolic 1;
1161 $cnt = $cnt - 1; # Mutator `--' not implemented
1162 $side = (sqrt(1 + $side**2) - 1)/$side;
1164 printf "%s=%f\n", $side, $side;
1165 printf "pi=%f\n", $side*(2**($iter+2));
1167 It prints (without so many line breaks)
1169 [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1]
1171 [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912
1174 The above module is very primitive. It does not implement
1175 mutator methods (C<++>, C<-=> and so on), does not do deep copying
1176 (not required without mutators!), and implements only those arithmetic
1177 operations which are used in the example.
1179 To implement most arithmetic operations is easy, one should just use
1180 the tables of operations, and change the code which fills %subr to
1182 my %subr = ( 'n' => sub {$_[0]} );
1183 foreach my $op (split " ", $overload::ops{with_assign}) {
1184 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1186 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1187 foreach my $op (split " ", "@overload::ops{ @bins }") {
1188 $subr{$op} = eval "sub {shift() $op shift()}";
1190 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1191 print "defining `$op'\n";
1192 $subr{$op} = eval "sub {$op shift()}";
1195 Due to L<Calling Conventions for Mutators>, we do not need anything
1196 special to make C<+=> and friends work, except filling C<+=> entry of
1197 %subr, and defining a copy constructor (needed since Perl has no
1198 way to know that the implementation of C<'+='> does not mutate
1199 the argument, compare L<Copy Constructor>).
1201 To implement a copy constructor, add C<'=' => \&cpy> to C<use overload>
1202 line, and code (this code assumes that mutators change things one level
1203 deep only, so recursive copying is not needed):
1207 bless [@$self], ref $self;
1210 To make C<++> and C<--> work, we need to implement actual mutators,
1211 either directly, or in C<nomethod>. We continue to do things inside
1212 C<nomethod>, thus add
1214 if ($meth eq '++' or $meth eq '--') {
1215 @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference
1219 after the first line of wrap(). This is not a most effective
1220 implementation, one may consider
1222 sub inc { $_[0] = bless ['++', shift, 1]; }
1226 As a final remark, note that one can fill %subr by
1228 my %subr = ( 'n' => sub {$_[0]} );
1229 foreach my $op (split " ", $overload::ops{with_assign}) {
1230 $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
1232 my @bins = qw(binary 3way_comparison num_comparison str_comparison);
1233 foreach my $op (split " ", "@overload::ops{ @bins }") {
1234 $subr{$op} = eval "sub {shift() $op shift()}";
1236 foreach my $op (split " ", "@overload::ops{qw(unary func)}") {
1237 $subr{$op} = eval "sub {$op shift()}";
1239 $subr{'++'} = $subr{'+'};
1240 $subr{'--'} = $subr{'-'};
1242 This finishes implementation of a primitive symbolic calculator in
1243 50 lines of Perl code. Since the numeric values of subexpressions
1244 are not cached, the calculator is very slow.
1246 Here is the answer for the exercise: In the case of str(), we need no
1247 explicit recursion since the overloaded C<.>-operator will fall back
1248 to an existing overloaded operator C<"">. Overloaded arithmetic
1249 operators I<do not> fall back to numeric conversion if C<fallback> is
1250 not explicitly requested. Thus without an explicit recursion num()
1251 would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild
1252 the argument of num().
1254 If you wonder why defaults for conversion are different for str() and
1255 num(), note how easy it was to write the symbolic calculator. This
1256 simplicity is due to an appropriate choice of defaults. One extra
1257 note: due to the explicit recursion num() is more fragile than sym():
1258 we need to explicitly check for the type of $a and $b. If components
1259 $a and $b happen to be of some related type, this may lead to problems.
1261 =head2 I<Really> symbolic calculator
1263 One may wonder why we call the above calculator symbolic. The reason
1264 is that the actual calculation of the value of expression is postponed
1265 until the value is I<used>.
1267 To see it in action, add a method
1272 @$obj->[0,1] = ('=', shift);
1275 to the package C<symbolic>. After this change one can do
1277 my $a = new symbolic 3;
1278 my $b = new symbolic 4;
1279 my $c = sqrt($a**2 + $b**2);
1281 and the numeric value of $c becomes 5. However, after calling
1283 $a->STORE(12); $b->STORE(5);
1285 the numeric value of $c becomes 13. There is no doubt now that the module
1286 symbolic provides a I<symbolic> calculator indeed.
1288 To hide the rough edges under the hood, provide a tie()d interface to the
1289 package C<symbolic> (compare with L<Metaphor clash>). Add methods
1291 sub TIESCALAR { my $pack = shift; $pack->new(@_) }
1293 sub nop { } # Around a bug
1295 (the bug is described in L<"BUGS">). One can use this new interface as
1297 tie $a, 'symbolic', 3;
1298 tie $b, 'symbolic', 4;
1299 $a->nop; $b->nop; # Around a bug
1301 my $c = sqrt($a**2 + $b**2);
1303 Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value
1304 of $c becomes 13. To insulate the user of the module add a method
1306 sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; }
1311 symbolic->vars($a, $b);
1312 my $c = sqrt($a**2 + $b**2);
1315 printf "c5 %s=%f\n", $c, $c;
1318 printf "c13 %s=%f\n", $c, $c;
1320 shows that the numeric value of $c follows changes to the values of $a
1325 Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>.
1329 When Perl is run with the B<-Do> switch or its equivalent, overloading
1330 induces diagnostic messages.
1332 Using the C<m> command of Perl debugger (see L<perldebug>) one can
1333 deduce which operations are overloaded (and which ancestor triggers
1334 this overloading). Say, if C<eq> is overloaded, then the method C<(eq>
1335 is shown by debugger. The method C<()> corresponds to the C<fallback>
1336 key (in fact a presence of this method shows that this package has
1337 overloading enabled, and it is what is used by the C<Overloaded>
1338 function of module C<overload>).
1342 Because it is used for overloading, the per-package hash %OVERLOAD now
1343 has a special meaning in Perl. The symbol table is filled with names
1344 looking like line-noise.
1346 For the purpose of inheritance every overloaded package behaves as if
1347 C<fallback> is present (possibly undefined). This may create
1348 interesting effects if some package is not overloaded, but inherits
1349 from two overloaded packages.
1351 Relation between overloading and tie()ing is broken. Overloading is
1352 triggered or not basing on the I<previous> class of tie()d value.
1354 This happens because the presence of overloading is checked too early,
1355 before any tie()d access is attempted. If the FETCH()ed class of the
1356 tie()d value does not change, a simple workaround is to access the value
1357 immediately after tie()ing, so that after this call the I<previous> class
1358 coincides with the current one.
1360 B<Needed:> a way to fix this without a speed penalty.
1362 Barewords are not covered by overloaded string constants.
1364 This document is confusing. There are grammos and misleading language
1365 used in places. It would seem a total rewrite is needed.