Seth Woolley's Man Viewer

overload(3) - overload - Package for overloading perl operations - man 3 overload

([section] manual, -k keyword, -K [section] search, -f whatis)
man plain no title

overload(3)            Perl Programmers Reference Guide            overload(3)



NAME
       overload - Package for overloading perl operations

SYNOPSIS
           package SomeThing;

           use overload
               '+' => \&myadd,
               '-' => \&mysub;
               # etc
           ...

           package main;
           $a = new SomeThing 57;
           $b=5+$a;
           ...
           if(3,n) (overload::Overloaded $b) {...}
           ...
           $strval = overload::StrVal $b;

DESCRIPTION
       Declaration of overloaded functions

       The compilation directive

           package Number;
           use overload
               "+" => \&add,
               "*=" => "muas";

       declares function Number::add() for addition, and method muas() in(1,8) the
       "class" "Number" (or one of its base classes) for the assignment form
       "*=" of multiplication.

       Arguments of this directive come in(1,8) (key, value) pairs.  Legal values
       are values legal inside a "&{ ... }" call, so the name of a subroutine,
       a reference to a subroutine, or an anonymous subroutine will all work.
       Note that values specified as strings are interpreted as methods, not
       subroutines.  Legal keys are listed below.

       The subroutine "add" will be called to execute "$a+$b" if(3,n) $a is a ref-
       erence to an object blessed into the package "Number", or if(3,n) $a is not
       an object from a package with defined mathemagic addition, but $b is a
       reference to a "Number".  It can also be called in(1,8) other situations,
       like "$a+=7", or "$a++".  See "MAGIC AUTOGENERATION".  (Mathemagical
       methods refer to methods triggered by an overloaded mathematical opera-
       tor.)

       Since overloading respects inheritance via the @ISA hierarchy, the
       above declaration would also trigger overloading of "+" and "*=" in(1,8) all
       the packages which inherit from "Number".

       Calling Conventions for Binary Operations

       The functions specified in(1,8) the "use overload ..." directive are called
       with three (in(1,8) one particular case with four, see "Last Resort") argu-
       ments.  If the corresponding operation is binary, then the first two
       arguments are the two arguments of the operation.  However, due to gen-
       eral object calling conventions, the first argument should always be an
       object in(1,8) the package, so in(1,8) the situation of "7+$a", the order of the
       arguments is interchanged.  It probably does not matter when implement-
       ing the addition method, but whether the arguments are reversed is
       vital to the subtraction method.  The method can query this information
       by examining the third argument, which can take three different values:

       FALSE  the order of arguments is as in(1,8) the current operation.

       TRUE   the arguments are reversed.

       "undef"
              the current operation is an assignment variant (as in(1,8) "$a+=7"),
              but the usual function is called instead.  This additional
              information can be used to generate some optimizations.  Compare
              "Calling Conventions for Mutators".

       Calling Conventions for Unary Operations

       Unary operation are considered binary operations with the second argu-
       ment being "undef".  Thus the functions that overloads "{"++"}" is
       called with arguments "($a,undef,'')" when $a++ is executed.

       Calling Conventions for Mutators

       Two types of mutators have different calling conventions:

       "++" and "--"
           The routines which implement these operators are expected to actu-
           ally mutate their arguments.  So, assuming that $obj is a reference
           to a number,

             sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n}

           is an appropriate implementation of overloaded "++".  Note that

             sub incr { ++$ {$_[0]} ; shift }

           is OK if(3,n) used with preincrement and with postincrement. (In the
           case of postincrement a copying will be performed, see "Copy Con-
           structor".)

       "x=" and other assignment versions
           There is nothing special about these methods.  They may change the
           value of their arguments, and may leave it as is.  The result is
           going to be assigned to the value in(1,8) the left-hand-side if(3,n) differ-
           ent from this value.

           This allows for the same method to be used as overloaded "+=" and
           "+".  Note that this is allowed, but not recommended, since by the
           semantic of "Fallback" Perl will call the method for "+" anyway, if(3,n)
           "+=" is not overloaded.

       Warning.  Due to the presence of assignment versions of operations,
       routines which may be called in(1,8) assignment context may create self-ref-
       erential structures.  Currently Perl will not free self-referential
       structures until cycles are "explicitly" broken.  You may get problems
       when traversing your structures too.

       Say,

         use overload '+' => sub { bless [ \$_[0], \$_[1] ] };

       is asking for trouble, since for code "$obj += $foo" the subroutine is
       called as "$obj = add($obj, $foo, undef)", or "$obj = [\$obj, \$foo]".
       If using such a subroutine is an important optimization, one can over-
       load(7,n) "+=" explicitly by a non-"optimized" version(1,3,5), or switch(1,n) to non-
       optimized version(1,3,5) if(3,n) "not defined $_[2]" (see "Calling Conventions for
       Binary Operations").

       Even if(3,n) no explicit assignment-variants of operators are present in(1,8) the
       script, they may be generated by the optimizer.  Say, ",$obj," or ',' .
       $obj . ',' may be both optimized to

         my $tmp = ',' . $obj;    $tmp .= ',';

       Overloadable Operations

       The following symbols can be specified in(1,8) "use overload" directive:

       * Arithmetic operations
                "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=",
                "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=",

            For these operations a substituted non-assignment variant can be
            called if(3,n) the assignment variant is not available.  Methods for
            operations "+", "-", "+=", and "-=" can be called to automatically
            generate increment and decrement methods.  The operation "-" can
            be used to autogenerate missing methods for unary minus or "abs".

            See "MAGIC AUTOGENERATION", "Calling Conventions for Mutators" and
            "Calling Conventions for Binary Operations") for details of these
            substitutions.

       * Comparison operations
                "<",  "<=", ">",  ">=", "==", "!=", "<=>",
                "lt", "le", "gt", "ge", "eq", "ne", "cmp",

            If the corresponding "spaceship" variant is available, it can be
            used to substitute for the missing operation.  During "sort(1,3)"ing
            arrays, "cmp" is used to compare values subject to "use overload".

       * Bit operations
                "&", "^", "|", "neg", "!", "~",

            "neg" stands for unary minus.  If the method for "neg" is not
            specified, it can be autogenerated using the method for subtrac-
            tion. If the method for "!" is not specified, it can be autogener-
            ated using the methods for "bool", or "", or "0+".

       * Increment and decrement
                "++", "--",

            If undefined, addition and subtraction methods can be used
            instead.  These operations are called both in(1,8) prefix and postfix
            form.

       * Transcendental functions
                "atan2", "cos", "sin", "exp", "abs", "log", "sqrt", "int"

            If "abs" is unavailable, it can be autogenerated using methods for
            "<" or "<=>" combined with either unary minus or subtraction.

            Note that traditionally the Perl function int rounds to 0, thus
            for floating-point-like types one should follow the same semantic.
            If "int" is unavailable, it can be autogenerated using the over-
            loading of "0+".

       * Boolean, string(3,n) and numeric conversion
                'bool', '""', '0+',

            If one or two of these operations are not overloaded, the remain-
            ing ones can be used instead.  "bool" is used in(1,8) the flow control
            operators (like "while") and for the ternary "?:" operation.
            These functions can return any arbitrary Perl value.  If the cor-
            responding operation for this value is overloaded too, that opera-
            tion will be called again with this value.

            As a special case if(3,n) the overload returns the object itself then
            it will be used directly. An overloaded conversion returning the
            object is probably a bug, because you're likely to get something
            that looks like "YourPackage=HASH(0x8172b34)".

       * Iteration
                "<>"

            If not overloaded, the argument will be converted to a filehandle
            or glob(1,3,7,n) (which may require a stringification).  The same overload-
            ing happens both for the read-filehandle syntax "<$var>" and glob-
            bing syntax "<${var}>".

            BUGS Even in(1,8) list context, the iterator is currently called only
            once and with scalar context.

       * Dereferencing
                '${}', '@{}', '%{}', '&{}', '*{}'.

            If not overloaded, the argument will be dereferenced as is, thus
            should be of correct type.  These functions should return a refer-
            ence of correct type, or another object with overloaded derefer-
            encing.

            As a special case if(3,n) the overload returns the object itself then
            it will be used directly (provided it is the correct type).

            The dereference operators must be specified explicitly they will
            not be passed to "nomethod".

       * Special
                "nomethod", "fallback", "=",

            see "SPECIAL SYMBOLS FOR "use overload"".

       See "Fallback" for an explanation of when a missing method can be auto-
       generated.

       A computer-readable form of the above table is available in(1,8) the hash
       %overload::ops, with values being space-separated lists of names:

        with_assign      => '+ - * / % ** << >> x .',
        assign           => '+= -= *= /= %= **= <<= >>= x= .=',
        num_comparison   => '< <= > >= == !=',
        '3way_comparison'=> '<=> cmp',
        str_comparison   => 'lt le gt ge eq ne',
        binary           => '& | ^',
        unary            => 'neg ! ~',
        mutators         => '++ --',
        func             => 'atan2 cos sin exp abs log sqrt',
        conversion       => 'bool "" 0+',
        iterators        => '<>',
        dereferencing    => '${} @{} %{} &{} *{}',
        special          => 'nomethod fallback ='

       Inheritance and overloading

       Inheritance interacts with overloading in(1,8) two ways.

       Strings as values of "use overload" directive
           If "value" in(1,8)

             use overload key => value;

           is a string(3,n), it is interpreted as a method name.

       Overloading of an operation is inherited by derived classes
           Any class derived from an overloaded class is also overloaded.  The
           set(7,n,1 builtins) of overloaded methods is the union of overloaded methods of all
           the ancestors. If some method is overloaded in(1,8) several ancestor,
           then which description will be used is decided by the usual inheri-
           tance rules:

           If "A" inherits from "B" and "C" (in(1,8) this order), "B" overloads "+"
           with "\&D::plus_sub", and "C" overloads "+" by "plus_meth", then
           the subroutine "D::plus_sub" will be called to implement operation
           "+" for an object in(1,8) package "A".

       Note that since the value of the "fallback" key is not a subroutine,
       its inheritance is not governed by the above rules.  In the current
       implementation, the value of "fallback" in(1,8) the first overloaded ances-
       tor is used, but this is accidental and subject to change.

SPECIAL SYMBOLS FOR "use overload"
       Three keys are recognized by Perl that are not covered by the above
       description.

       Last Resort

       "nomethod" should be followed by a reference to a function of four
       parameters.  If defined, it is called when the overloading mechanism
       cannot find a method for some operation.  The first three arguments of
       this function coincide with the arguments for the corresponding method
       if(3,n) it were found, the fourth argument is the symbol corresponding to
       the missing method.  If several methods are tried, the last one is
       used.  Say, "1-$a" can be equivalent to

               &nomethodMethod($a,1,1,"-")

       if(3,n) the pair "nomethod" => "nomethodMethod" was specified in(1,8) the "use
       overload" directive.

       The "nomethod" mechanism is not used for the dereference operators (
       ${} @{} %{} &{} *{} ).

       If some operation cannot be resolved, and there is no function assigned
       to "nomethod", then an exception will be raised via die()-- unless
       "fallback" was specified as a key in(1,8) "use overload" directive.

       Fallback

       The key "fallback" governs what to do if(3,n) a method for a particular
       operation is not found.  Three different cases are possible depending
       on the value of "fallback":

       * "undef"       Perl tries to use a substituted method (see "MAGIC
                       AUTOGENERATION").  If this fails, it then tries to
                       calls "nomethod" value; if(3,n) missing, an exception will
                       be raised.

       * TRUE          The same as for the "undef" value, but no exception is
                       raised.  Instead, it silently reverts to what it would
                       have done were there no "use overload" present.

       * defined, but FALSE
                       No autogeneration is tried.  Perl tries to call
                       "nomethod" value, and if(3,n) this is missing, raises an
                       exception.

       Note. "fallback" inheritance via @ISA is not carved in(1,8) stone yet, see
       "Inheritance and overloading".

       Copy Constructor

       The value for "=" is a reference to a function with three arguments,
       i.e., it looks like the other values in(1,8) "use overload". However, it
       does not overload the Perl assignment operator. This would go against
       Camel hair.

       This operation is called in(1,8) the situations when a mutator is applied to
       a reference that shares its object with some other reference, such as

               $a=$b;
               ++$a;

       To make this change $a and not change $b, a copy of $$a is made, and $a
       is assigned a reference to this new object.  This operation is done
       during execution of the "++$a", and not during the assignment, (so
       before the increment $$a coincides with $$b).  This is only done if(3,n)
       "++" is expressed via a method for '++' or '+=' (or "nomethod").  Note
       that if(3,n) this operation is expressed via '+' a nonmutator, i.e., as in(1,8)

               $a=$b;
               $a=$a+1;

       then $a does not reference a new copy of $$a, since $$a does not appear
       as lvalue when the above code is executed.

       If the copy constructor is required during the execution of some muta-
       tor, but a method for '=' was not specified, it can be autogenerated as
       a string(3,n) copy if(3,n) the object is a plain scalar.

       Example
            The actually executed code for

                    $a=$b;
                    Something else which does not modify $a or $b....
                    ++$a;

            may be

                    $a=$b;
                    Something else which does not modify $a or $b....
                    $a = $a->clone(undef,"");
                    $a->incr(undef,"");

            if(3,n) $b was mathemagical, and '++' was overloaded with "\&incr", '='
            was overloaded with "\&clone".

       Same behaviour is triggered by "$b = $a++", which is consider a synonym
       for "$b = $a; ++$a".

MAGIC AUTOGENERATION
       If a method for an operation is not found, and the value for  "fall-
       back" is TRUE or undefined, Perl tries to autogenerate a substitute
       method for the missing operation based on the defined operations.
       Autogenerated method substitutions are possible for the following oper-
       ations:

       Assignment forms of arithmetic operations
                       "$a+=$b" can use the method for "+" if(3,n) the method for
                       "+=" is not defined.

       Conversion operations
                       String, numeric, and boolean conversion are calculated
                       in(1,8) terms of one another if(3,n) not all of them are defined.

       Increment and decrement
                       The "++$a" operation can be expressed in(1,8) terms of
                       "$a+=1" or "$a+1", and "$a--" in(1,8) terms of "$a-=1" and
                       "$a-1".

       "abs($a)"       can be expressed in(1,8) terms of "$a<0" and "-$a" (or
                       "0-$a").

       Unary minus     can be expressed in(1,8) terms of subtraction.

       Negation        "!" and "not" can be expressed in(1,8) terms of boolean con-
                       version(1,3,5), or string(3,n) or numerical conversion.

       Concatenation   can be expressed in(1,8) terms of string(3,n) conversion.

       Comparison operations
                       can be expressed in(1,8) terms of its "spaceship" counter-
                       part: either "<=>" or "cmp":

                           <, >, <=, >=, ==, !=        in(1,8) terms of <=>
                           lt, gt, le, ge, eq, ne      in(1,8) terms of cmp

       Iterator
                           <>                          in(1,8) terms of builtin operations

       Dereferencing
                           ${} @{} %{} &{} *{}         in(1,8) terms of builtin operations

       Copy operator   can be expressed in(1,8) terms of an assignment to the
                       dereferenced value, if(3,n) this value is a scalar and not a
                       reference.

Losing overloading
       The restriction for the comparison operation is that even if(3,n), for exam-
       ple, `"cmp"' should return a blessed reference, the autogenerated
       `"lt"' function will produce only a standard logical value based on the
       numerical value of the result of `"cmp"'.  In particular, a working
       numeric conversion is needed in(1,8) this case (possibly expressed in(1,8) terms
       of other conversions).

       Similarly, ".="  and "x=" operators lose their mathemagical properties
       if(3,n) the string(3,n) conversion substitution is applied.

       When you chop() a mathemagical object it is promoted to a string(3,n) and
       its mathemagical properties are lost.  The same can happen with other
       operations as well.

Run-time Overloading
       Since all "use" directives are executed at compile-time, the only way
       to change overloading during run-time is to

           eval 'use overload "+" => \&addmethod';

       You can also use

           eval 'no overload "+", "--", "<="';

       though the use of these constructs during run-time is questionable.

Public functions
       Package "overload.pm" provides the following public functions:

       overload::StrVal(arg)
            Gives string(3,n) value of "arg" as in(1,8) absence of stringify overload-
            ing.

       overload::Overloaded(arg)
            Returns true if(3,n) "arg" is subject to overloading of some opera-
            tions.

       overload::Method(obj,op)
            Returns "undef" or a reference to the method that implements "op".

Overloading constants
       For some application Perl parser mangles constants too much.  It is
       possible to hook into this process via overload::constant() and over-
       load::remove_constant() functions.

       These functions take a hash as an argument.  The recognized keys of
       this hash are

       integer to overload integer constants,

       float   to overload floating point constants,

       binary  to overload octal and hexadecimal constants,

       q       to overload "q"-quoted strings, constant pieces of "qq"- and
               "qx"-quoted strings and here-documents,

       qr      to overload constant pieces of regular expressions.

       The corresponding values are references to functions which take three
       arguments: the first one is the initial string(3,n) form of the constant,
       the second one is how Perl interprets this constant, the third one is
       how the constant is used.  Note that the initial string(3,n) form does not
       contain string(3,n) delimiters, and has backslashes in(1,8) backslash-delimiter
       combinations stripped (thus the value of delimiter is not relevant for
       processing of this string(3,n)).  The return value of this function is how
       this constant is going to be interpreted by Perl.  The third argument
       is undefined unless for overloaded "q"- and "qr"- constants, it is "q"
       in(1,8) single-quote context (comes from strings, regular expressions, and
       single-quote HERE documents), it is "tr" for arguments of "tr"/"y"
       operators, it is "s" for right-hand side of "s"-operator, and it is
       "qq" otherwise.

       Since an expression "ab$cd,," is just a shortcut for 'ab' . $cd . ',,',
       it is expected that overloaded constant strings are equipped with rea-
       sonable overloaded catenation operator, otherwise absurd results will
       result.  Similarly, negative numbers are considered as negations of
       positive constants.

       Note that it is probably meaningless to call the functions over-
       load::constant() and overload::remove_constant() from anywhere but
       import() and unimport() methods.  From these methods they may be called
       as

               sub import {
                 shift;
                 return unless @_;
                 die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant';
                 overload::constant integer => sub {Math::BigInt->new(shift)};
               }

       BUGS Currently overloaded-ness of constants does not propagate into
       "eval '...'".

IMPLEMENTATION
       What follows is subject to change RSN.

       The table of methods for all operations is cached in(1,8) magic(4,5) for the sym-
       bol table hash for the package.  The cache is invalidated during pro-
       cessing of "use overload", "no overload", new function definitions, and
       changes in(1,8) @ISA. However, this invalidation remains unprocessed until
       the next "bless"ing into the package. Hence if(3,n) you want to change over-
       loading structure dynamically, you'll need an additional (fake)
       "bless"ing to update(7,n) the table.

       (Every SVish thing has a magic(4,5) queue(1,3), and magic(4,5) is an entry in(1,8) that
       queue.  This is how a single variable may participate in(1,8) multiple forms
       of magic(4,5) simultaneously.  For instance, environment variables regularly
       have two forms at once: their %ENV magic(4,5) and their taint magic. How-
       ever, the magic(4,5) which implements overloading is applied to the stashes,
       which are rarely used directly, thus should not slow down Perl.)

       If an object belongs to a package using overload, it carries a special
       flag.  Thus the only speed penalty during arithmetic operations without
       overloading is the checking of this flag.

       In fact, if(3,n) "use overload" is not present, there is almost no overhead
       for overloadable operations, so most programs should not suffer measur-
       able performance penalties.  A considerable effort was made to minimize
       the overhead when overload is used in(1,8) some package, but the arguments
       in(1,8) question do not belong to packages using overload.  When in(1,8) doubt,
       test your speed with "use overload" and without it.  So far there have
       been no reports of substantial speed degradation if(3,n) Perl is compiled
       with optimization turned on.

       There is no size penalty for data if(3,n) overload is not used. The only
       size penalty if(3,n) overload is used in(1,8) some package is that all the pack-
       ages acquire a magic(4,5) during the next "bless"ing into the package. This
       magic(4,5) is three-words-long for packages without overloading, and carries
       the cache table if(3,n) the package is overloaded.

       Copying ("$a=$b") is shallow; however, a one-level-deep copying is car-
       ried out before any operation that can imply an assignment to the
       object $a (or $b) refers to, like "$a++".  You can override this behav-
       ior by defining your own copy constructor (see "Copy Constructor").

       It is expected that arguments to methods that are not explicitly sup-
       posed to be changed are constant (but this is not enforced).

Metaphor clash
       One may wonder why the semantic of overloaded "=" is so counter intu-
       itive.  If it looks counter intuitive to you, you are subject to a
       metaphor clash.

       Here is a Perl object metaphor:

         object is a reference to blessed data

       and an arithmetic metaphor:

         object is a thing by itself.

       The main problem of overloading "=" is the fact that these metaphors
       imply different actions on the assignment "$a = $b" if(3,n) $a and $b are
       objects.  Perl-think implies that $a becomes a reference to whatever $b
       was referencing.  Arithmetic-think implies that the value of "object"
       $a is changed to become the value of the object $b, preserving the fact
       that $a and $b are separate entities.

       The difference is not relevant in(1,8) the absence of mutators.  After a
       Perl-way assignment an operation which mutates the data referenced by
       $a would change the data referenced by $b too.  Effectively, after "$a
       = $b" values of $a and $b become indistinguishable.

       On the other hand, anyone who has used algebraic notation knows the
       expressive power of the arithmetic metaphor.  Overloading works hard to
       enable this metaphor while preserving the Perlian way as far as possi-
       ble.  Since it is not possible to freely mix two contradicting
       metaphors, overloading allows the arithmetic way to write(1,2) things as far
       as all the mutators are called via overloaded access(2,5) only.  The way it
       is done is described in(1,8) "Copy Constructor".

       If some mutator methods are directly applied to the overloaded values,
       one may need to explicitly unlink(1,2) other values which references the
       same value:

           $a = new Data 23;
           ...
           $b = $a;            # $b is "linked" to $a
           ...
           $a = $a->clone;     # Unlink $b from $a
           $a->increment_by(4);

       Note that overloaded access(2,5) makes this transparent:

           $a = new Data 23;
           $b = $a;            # $b is "linked" to $a
           $a += 4;            # would unlink(1,2) $b automagically

       However, it would not make

           $a = new Data 23;
           $a = 4;             # Now $a is a plain 4, not 'Data'

       preserve "objectness" of $a.  But Perl has a way to make assignments to
       an object do whatever you want.  It is just not the overload, but
       tie()ing interface (see "tie" in(1,8) perlfunc).  Adding a FETCH() method
       which returns the object itself, and STORE() method which changes the
       value of the object, one can reproduce the arithmetic metaphor in(1,8) its
       completeness, at least for variables which were tie()d from the start.

       (Note that a workaround for a bug may be needed, see "BUGS".)

Cookbook
       Please add examples to what follows!

       Two-face scalars

       Put this in(1,8) two_face.pm in(1,8) your Perl library directory:

         package two_face;             # Scalars with separate string(3,n) and
                                       # numeric values.
         sub new { my $p = shift; bless [@_], $p }
         use overload '""' => \&str, '0+' => \&num, fallback => 1;
         sub num {shift->[1]}
         sub str {shift->[0]}

       Use it as follows:

         require two_face;
         my $seven = new two_face ("vii", 7);
         printf(1,3,1 builtins) "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1;
         print "seven contains `i'\n" if(3,n) $seven =~ /i/;

       (The second line creates a scalar which has both a string(3,n) value, and a
       numeric value.)  This prints:

         seven=vii, seven=7, eight=8
         seven contains `i'

       Two-face references

       Suppose you want to create an object which is accessible as both an
       array reference and a hash reference, similar to the pseudo-hash
       builtin Perl type.  Let's make it better than a pseudo-hash by allowing
       index 0 to be treated as a normal element.

         package two_refs;
         use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} };
         sub new {
           my $p = shift;
           bless \ [@_], $p;
         }
         sub gethash {
           my %h;
           my $self = shift;
           tie %h, ref $self, $self;
           \%h;
         }

         sub TIEHASH { my $p = shift; bless \ shift, $p }
         my %fields;
         my $i = 0;
         $fields{$_} = $i++ foreach qw{zero one two three};
         sub STORE {
           my $self = ${shift()};
           my $key = $fields{shift()};
           defined $key or die "Out of band access(2,5)";
           $$self->[$key] = shift;
         }
         sub FETCH {
           my $self = ${shift()};
           my $key = $fields{shift()};
           defined $key or die "Out of band access(2,5)";
           $$self->[$key];
         }

       Now one can access(2,5) an object using both the array and hash syntax:

         my $bar = new two_refs 3,4,5,6;
         $bar->[2] = 11;
         $bar->{two} == 11 or die 'bad hash fetch';

       Note several important features of this example.  First of all, the
       actual type of $bar is a scalar reference, and we do not overload the
       scalar dereference.  Thus we can get the actual non-overloaded contents
       of $bar by just using $$bar (what we do in(1,8) functions which overload
       dereference).  Similarly, the object returned by the TIEHASH() method
       is a scalar reference.

       Second, we create a new tied hash each time(1,2,n) the hash syntax is used.
       This allows us not to worry about a possibility of a reference loop,
       which would lead to a memory leak.

       Both these problems can be cured.  Say, if(3,n) we want to overload hash
       dereference on a reference to an object which is implemented as a hash
       itself, the only problem one has to circumvent is how to access(2,5) this
       actual hash (as opposed to the virtual(5,8) hash exhibited by the overloaded
       dereference operator).  Here is one possible fetching routine:

         sub access_hash {
           my ($self, $key) = (shift, shift);
           my $class = ref $self;
           bless $self, 'overload::dummy'; # Disable overloading of %{}
           my $out = $self->{$key};
           bless $self, $class;        # Restore overloading
           $out;
         }

       To remove creation of the tied hash on each access(2,5), one may an extra
       level of indirection which allows a non-circular structure of refer-
       ences:

         package two_refs1;
         use overload '%{}' => sub { ${shift()}->[1] },
                      '@{}' => sub { ${shift()}->[0] };
         sub new {
           my $p = shift;
           my $a = [@_];
           my %h;
           tie %h, $p, $a;
           bless \ [$a, \%h], $p;
         }
         sub gethash {
           my %h;
           my $self = shift;
           tie %h, ref $self, $self;
           \%h;
         }

         sub TIEHASH { my $p = shift; bless \ shift, $p }
         my %fields;
         my $i = 0;
         $fields{$_} = $i++ foreach qw{zero one two three};
         sub STORE {
           my $a = ${shift()};
           my $key = $fields{shift()};
           defined $key or die "Out of band access(2,5)";
           $a->[$key] = shift;
         }
         sub FETCH {
           my $a = ${shift()};
           my $key = $fields{shift()};
           defined $key or die "Out of band access(2,5)";
           $a->[$key];
         }

       Now if(3,n) $baz is overloaded like this, then $baz is a reference to a ref-
       erence to the intermediate array, which keeps a reference to an actual
       array, and the access(2,5) hash.  The tie()ing object for the access(2,5) hash is
       a reference to a reference to the actual array, so

          There are no loops of references.

          Both "objects" which are blessed into the class "two_refs1" are
           references to a reference to an array, thus references to a scalar.
           Thus the accessor expression "$$foo->[$ind]" involves no overloaded
           operations.

       Symbolic calculator

       Put this in(1,8) symbolic.pm in(1,8) your Perl library directory:

         package symbolic;             # Primitive symbolic calculator
         use overload nomethod => \&wrap;

         sub new { shift; bless ['n', @_] }
         sub wrap {
           my ($obj, $other, $inv, $meth) = @_;
           ($obj, $other) = ($other, $obj) if(3,n) $inv;
           bless [$meth, $obj, $other];
         }

       This module is very unusual as overloaded modules go: it does not pro-
       vide any usual overloaded operators, instead it provides the "Last
       Resort" operator "nomethod".  In this example the corresponding subrou-
       tine returns an object which encapsulates operations done over the
       objects: "new symbolic 3" contains "['n', 3]", "2 + new symbolic 3"
       contains "['+', 2, ['n', 3]]".

       Here is an example of the script which "calculates" the side of circum-
       scribed octagon using the above package:

         require symbolic;
         my $iter = 1;                 # 2**($iter+2) = 8
         my $side = new symbolic 1;
         my $cnt = $iter;

         while ($cnt--) {
           $side = (sqrt(1 + $side**2) - 1)/$side;
         }
         print "OK\n";

       The value of $side is

         ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]],
                              undef], 1], ['n', 1]]

       Note that while we obtained this value using a nice(1,2) little script,
       there is no simple way to use this value.  In fact this value may be
       inspected in(1,8) debugger (see perldebug), but ony if(3,n) "bareStringify"
       Option is set(7,n,1 builtins), and not via "p" command.

       If one attempts to print this value, then the overloaded operator ""
       will be called, which will call "nomethod" operator.  The result of
       this operator will be stringified again, but this result is again of
       type "symbolic", which will lead to an infinite loop.

       Add a pretty-printer method to the module symbolic.pm:

         sub pretty {
           my ($meth, $a, $b) = @{+shift};
           $a = 'u' unless defined $a;
           $b = 'u' unless defined $b;
           $a = $a->pretty if(3,n) ref $a;
           $b = $b->pretty if(3,n) ref $b;
           "[$meth $a $b]";
         }

       Now one can finish the script by

         print "side = ", $side->pretty, "\n";

       The method "pretty" is doing object-to-string conversion, so it is nat-
       ural to overload the operator "" using this method.  However, inside
       such a method it is not necessary to pretty-print the components $a and
       $b of an object.  In the above subroutine "[$meth $a $b]" is a catena-
       tion of some strings and components $a and $b.  If these components use
       overloading, the catenation operator will look(1,8,3 Search::Dict) for an overloaded opera-
       tor "."; if(3,n) not present, it will look(1,8,3 Search::Dict) for an overloaded operator "".
       Thus it is enough to use

         use overload nomethod => \&wrap, '""' => \&str;
         sub str {
           my ($meth, $a, $b) = @{+shift};
           $a = 'u' unless defined $a;
           $b = 'u' unless defined $b;
           "[$meth $a $b]";
         }

       Now one can change the last line of the script to

         print "side = $side\n";

       which outputs

         side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]]

       and one can inspect the value in(1,8) debugger using all the possible meth-
       ods.

       Something is still amiss: consider the loop variable $cnt of the
       script.  It was a number, not an object.  We cannot make this value of
       type "symbolic", since then the loop will not terminate.

       Indeed, to terminate the cycle, the $cnt should become false.  However,
       the operator "bool" for checking falsity is overloaded (this time(1,2,n) via
       overloaded ""), and returns a long string(3,n), thus any object of type
       "symbolic" is true.  To overcome this, we need a way to compare an
       object to 0.  In fact, it is easier to write(1,2) a numeric conversion rou-
       tine.

       Here is the text of symbolic.pm with such a routine added (and slightly
       modified str()):

         package symbolic;             # Primitive symbolic calculator
         use overload
           nomethod => \&wrap, '""' => \&str, '0+' => \&num;

         sub new { shift; bless ['n', @_] }
         sub wrap {
           my ($obj, $other, $inv, $meth) = @_;
           ($obj, $other) = ($other, $obj) if(3,n) $inv;
           bless [$meth, $obj, $other];
         }
         sub str {
           my ($meth, $a, $b) = @{+shift};
           $a = 'u' unless defined $a;
           if(3,n) (defined $b) {
             "[$meth $a $b]";
           } else {
             "[$meth $a]";
           }
         }
         my %subr = ( n => sub {$_[0]},
                      sqrt => sub {sqrt $_[0]},
                      '-' => sub {shift() - shift()},
                      '+' => sub {shift() + shift()},
                      '/' => sub {shift() / shift()},
                      '*' => sub {shift() * shift()},
                      '**' => sub {shift() ** shift()},
                    );
         sub num {
           my ($meth, $a, $b) = @{+shift};
           my $subr = $subr{$meth}
             or die "Do not know how to ($meth) in(1,8) symbolic";
           $a = $a->num if(3,n) ref $a eq __PACKAGE__;
           $b = $b->num if(3,n) ref $b eq __PACKAGE__;
           $subr->($a,$b);
         }

       All the work of numeric conversion is done in(1,8) %subr and num().  Of
       course, %subr is not complete, it contains only operators used in(1,8) the
       example below.  Here is the extra-credit question: why do we need an
       explicit recursion in(1,8) num()?  (Answer is at the end of this section.)

       Use this module like this:

         require symbolic;
         my $iter = new symbolic 2;    # 16-gon
         my $side = new symbolic 1;
         my $cnt = $iter;

         while ($cnt) {
           $cnt = $cnt - 1;            # Mutator `--' not implemented
           $side = (sqrt(1 + $side**2) - 1)/$side;
         }
         printf(1,3,1 builtins) "%s=%f\n", $side, $side;
         printf(1,3,1 builtins) "pi=%f\n", $side*(2**($iter+2));

       It prints (without so many line breaks)

         [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1]
                                 [n 1]] 2]]] 1]
            [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912
         pi=3.182598

       The above module is very primitive.  It does not implement mutator
       methods ("++", "-=" and so on), does not do deep copying (not required
       without mutators!), and implements only those arithmetic operations
       which are used in(1,8) the example.

       To implement most arithmetic operations is easy; one should just use
       the tables of operations, and change the code which fills %subr to

         my %subr = ( 'n' => sub {$_[0]} );
         foreach my $op (split(1,n) " ", $overload::ops{with_assign}) {
           $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
         }
         my @bins = qw(binary 3way_comparison num_comparison str_comparison);
         foreach my $op (split(1,n) " ", "@overload::ops{ @bins }") {
           $subr{$op} = eval "sub {shift() $op shift()}";
         }
         foreach my $op (split(1,n) " ", "@overload::ops{qw(unary func)}") {
           print "defining `$op'\n";
           $subr{$op} = eval "sub {$op shift()}";
         }

       Due to "Calling Conventions for Mutators", we do not need anything spe-
       cial to make "+=" and friends work, except filling "+=" entry of %subr,
       and defining a copy constructor (needed since Perl has no way to know
       that the implementation of '+=' does not mutate the argument, compare
       "Copy Constructor").

       To implement a copy constructor, add "'=' => \&cpy" to "use overload"
       line, and code (this code assumes that mutators change things one level
       deep only, so recursive copying is not needed):

         sub cpy {
           my $self = shift;
           bless [@$self], ref $self;
         }

       To make "++" and "--" work, we need to implement actual mutators,
       either directly, or in(1,8) "nomethod".  We continue to do things inside
       "nomethod", thus add

           if(3,n) ($meth eq '++' or $meth eq '--') {
             @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference
             return $obj;
           }

       after the first line of wrap().  This is not a most effective implemen-
       tation, one may consider

         sub inc { $_[0] = bless ['++', shift, 1]; }

       instead.

       As a final remark, note that one can fill %subr by

         my %subr = ( 'n' => sub {$_[0]} );
         foreach my $op (split(1,n) " ", $overload::ops{with_assign}) {
           $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}";
         }
         my @bins = qw(binary 3way_comparison num_comparison str_comparison);
         foreach my $op (split(1,n) " ", "@overload::ops{ @bins }") {
           $subr{$op} = eval "sub {shift() $op shift()}";
         }
         foreach my $op (split(1,n) " ", "@overload::ops{qw(unary func)}") {
           $subr{$op} = eval "sub {$op shift()}";
         }
         $subr{'++'} = $subr{'+'};
         $subr{'--'} = $subr{'-'};

       This finishes implementation of a primitive symbolic calculator in(1,8) 50
       lines of Perl code.  Since the numeric values of subexpressions are not
       cached, the calculator is very slow.

       Here is the answer for the exercise: In the case of str(), we need no
       explicit recursion since the overloaded "."-operator will fall back to
       an existing overloaded operator "".  Overloaded arithmetic operators do
       not fall back to numeric conversion if(3,n) "fallback" is not explicitly
       requested.  Thus without an explicit recursion num() would convert
       "['+', $a, $b]" to "$a + $b", which would just rebuild the argument of
       num().

       If you wonder why defaults for conversion are different for str() and
       num(), note how easy it was to write(1,2) the symbolic calculator.  This
       simplicity is due to an appropriate choice of defaults.  One extra
       note: due to the explicit recursion num() is more fragile than sym():
       we need to explicitly check for the type of $a and $b.  If components
       $a and $b happen to be of some related type, this may lead to problems.

       Really symbolic calculator

       One may wonder why we call the above calculator symbolic.  The reason
       is that the actual calculation of the value of expression is postponed
       until the value is used.

       To see it in(1,8) action, add a method

         sub STORE {
           my $obj = shift;
           $#$obj = 1;
           @$obj->[0,1] = ('=', shift);
         }

       to the package "symbolic".  After this change one can do

         my $a = new symbolic 3;
         my $b = new symbolic 4;
         my $c = sqrt($a**2 + $b**2);

       and the numeric value of $c becomes 5.  However, after calling

         $a->STORE(12);  $b->STORE(5);

       the numeric value of $c becomes 13.  There is no doubt now that the
       module symbolic provides a symbolic calculator indeed.

       To hide the rough edges under the hood, provide a tie()d interface to
       the package "symbolic" (compare with "Metaphor clash").  Add methods

         sub TIESCALAR { my $pack(3,n,n pack-old) = shift; $pack->new(@_) }
         sub FETCH { shift }
         sub nop {  }          # Around a bug

       (the bug is described in(1,8) "BUGS").  One can use this new interface as

         tie $a, 'symbolic', 3;
         tie $b, 'symbolic', 4;
         $a->nop;  $b->nop;    # Around a bug

         my $c = sqrt($a**2 + $b**2);

       Now numeric value of $c is 5.  After "$a = 12; $b = 5" the numeric
       value of $c becomes 13.  To insulate the user of the module add a
       method

         sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; }

       Now

         my ($a, $b);
         symbolic->vars($a, $b);
         my $c = sqrt($a**2 + $b**2);

         $a = 3; $b = 4;
         printf(1,3,1 builtins) "c5  %s=%f\n", $c, $c;

         $a = 12; $b = 5;
         printf(1,3,1 builtins) "c13  %s=%f\n", $c, $c;

       shows that the numeric value of $c follows changes to the values of $a
       and $b.

AUTHOR
       Ilya Zakharevich <ilya@math.mps.ohio-state.edu>.

DIAGNOSTICS
       When Perl is run with the -Do switch(1,n) or its equivalent, overloading
       induces diagnostic messages.

       Using the "m" command of Perl debugger (see perldebug) one can deduce
       which operations are overloaded (and which ancestor triggers this over-
       loading). Say, if(3,n) "eq" is overloaded, then the method "(eq" is shown by
       debugger. The method "()" corresponds to the "fallback" key (in(1,8) fact a
       presence of this method shows that this package has overloading
       enabled, and it is what is used by the "Overloaded" function of module
       "overload").

       The module might issue the following warnings:

       Odd number of arguments for overload::constant
           (W) The call to overload::constant contained an odd number of argu-
           ments.  The arguments should come in(1,8) pairs.

       `%s' is not an overloadable type
           (W) You tried to overload a constant type the overload package is
           unaware of.

       `%s' is not a code reference
           (W) The second (fourth, sixth, ...) argument of overload::constant
           needs to be a code reference. Either an anonymous subroutine, or a
           reference to a subroutine.

BUGS
       Because it is used for overloading, the per-package hash %OVERLOAD now
       has a special meaning in(1,8) Perl. The symbol table is filled with names
       looking like line-noise.

       For the purpose of inheritance every overloaded package behaves as if(3,n)
       "fallback" is present (possibly undefined). This may create interesting
       effects if(3,n) some package is not overloaded, but inherits from two over-
       loaded packages.

       Relation between overloading and tie()ing is broken.  Overloading is
       triggered or not basing on the previous class of tie()d value.

       This happens because the presence of overloading is checked too early,
       before any tie()d access(2,5) is attempted.  If the FETCH()ed class of the
       tie()d value does not change, a simple workaround is to access(2,5) the
       value immediately after tie()ing, so that after this call the previous
       class coincides with the current one.

       Needed: a way to fix this without a speed penalty.

       Barewords are not covered by overloaded string(3,n) constants.

       This document is confusing.  There are grammos and misleading language
       used in(1,8) places.  It would seem a total rewrite is needed.



perl v5.8.5                       2001-09-21                       overload(3)

References for this manual (incoming links)