The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


Tie::Hash::MultiKey - multiple keys per value


  use Tie::Hash::MultiKey;

  $thm = tie %hash, qw(Tie::Hash::MultiKey) ,@optionalext;
  $thm = tied %hash;

  untie %hash;

  ($href,$thm) = new Tie::Hash::MultiKey;

  $hash{'foo'}        = 'baz';
  $hash{'foo', 'bar'} = 'baz';
  $array_ref = ['foo', 'bar'];
  $hash{ $array_ref } = 'baz';

  print $hash{foo};     # prints 'baz'
  print $hash{bar};     # prints 'baz'

  $array_ref = ['fuz','zup'];
  $val = tied(%hash)->addkey('fuz' => 'bar');
  $val = tied(%hash)->addkey('fuz','zup' => 'bar');
  $val = tied(%hash)->addkey( $array_ref => 'bar');

  print $hash{fuz}      # prints 'baz'

  $array_ref = ['foo', 'bar'];
  $val = tied(%hash)->remove('foo');
  $val = tied(%hash)->remove('foo', 'bar');
  $val = tied(%hash)->remove( $array_ref );

  $val = tied(%hash)->delkey(); alias for above

  @ordered_keys = tied(%hash)->keylist('foo')
  @allkeys_by_order = tied(%hash)->keylist();
  @slotlist = tied(%hash)->slotlist($i);
  @ordered_vals = tied(%hash)->vals();

  $num_vals = tied(%hash)->size;
  $num_vals = tied(%hash)->consolidate;

  ($newRef,$newThm) = tied(%hash)->clone();
  $newThm = tied(%hash)->copy(tied(%new),@optionalext);

  All of the above methods can be accessed as:

  i.e.  $thm->consolidate;


Tie::Hash::MultiKey creates hashes that can have multiple ordered keys for a single value. As shown in the SYNOPSIS, multiple keys share a common value.

Additional keys can be added that share the same value and keys can be removed without deleting other keys that share that value. a value for one or more keys that already exist will overwrite the existing value and add any missing keys to the key group for that value.

WARNING: multiple key values supplied as an ARRAY to STORE and DELETE operations are passed by Perl as a single string separated by Perl's $; multidimensional array seperator. i.e.

        $hash{'a','b','c'} = $something;
        @keys = ('a','b','c');
        $hash{@keys} = $something'

This really means $hash{join($;, 'a','b','c')};

Tie::Hash::MultiKey will do the right thing as long as your keys DO NOT contain binary data the may include the $; separator character.

It is recommended that you use the ARRAY_REF construct to supply multiple keys for binary data. i.e.

        $hash{['a','b','c']} = $something;
        $keys = ['a','b','c'];
        $hash{$keys} = $something;

The ARRAY_REF construct is ALWAYS safe.

  • $thm = tie %hash,'Tie::Hash::MultiKey' ,%optional_ex

    Ties a %hash to this package for enhanced capability and returns a method pointer.

      my %hash;
      my $thm = tie %hash,'Tie::Hash::MultiKey';

    Extension of this module is discussed in detail below.

  • $thm = tied %hash;

    Returns a method pointer for this package.

  • untie %hash;

    Breaks the binding between a variable and this package. There is no affect if the variable is not tied.

    REMEMBER that if you have created a reference to the tied hash, untie will not work until that binding is broken. This means that the object will not be destroyed or garbage collected and the memory will not be reclaimed.

    i.e WRONG

      $thm = tie %h, 'Tie::Hash::MultiKey';
      ... code ...
      untie %h;
      $thm = tie %h, 'Tie::Hash::MultiKey';
      ... code ...
      undef $thm;
      untie %h;
  • ($href,$thm) = new 'Tie::Hash::MultiKey' ,%optional_ex

    This method returns an UNBLESSED reference to an anonymous tied %hash.

      input:        none
      returns:      unblessed tied %hash reference,
                    object handle

    To get the object handle from \%hash use this.

            $thm = tied %{$href};

    In SCALAR context it returns the unblessed %hash pointer. In ARRAY context it returns the unblessed %hash pointer and the package object/method pointer.

  • $val = $thm->addkey('new_key' => 'existing_key');

    Add one or more keys to the shared key group for a particular value.

      input:        array or array_ref,
      returns:      hash value
                or  dies with stack trace

    Dies with stack trace if existing_key does not exist OR if new key belongs to another key set.

    Arguments may be a single SCALAR, ARRAY, or ARRAY_REF

  • $val = ->remove('key');

  • $val = ->delkey('key'); alias for above

    Remove one or more keys from the shared key group for a particular value If this operation removes the LAST key, then it performs a DELETE which is the same as:

            delete $hash{key};

    remove returns a reverse list of the removed value's by key

      i.e.  @val = remove(something);
       or   $val = remove(something);

    Arguments may be a single SCALAR, ARRAY or ARRAY_REF

  • @ordered_keys = $thm->keylist('foo');

  • @allkeys_by_order = $thm->keylist();

    Returns all the keys in the group that includes the KEY 'foo' in the order that they were added to the %hash;

    If no argument is specified, returns all the keys in the %hash in the order that they were added to the %hash

      input:        key or EMPTY
      returns:      @ordered_keys
      returns:      () if $key is not in the %hash
  • @keys = $thm->slotlist($i);

    Returns one key from each key group in position $i.

            $thm = tie %hash, 'Tie::Hash::MultiKey';
            $hash{['a','b','c']} = 'one';
            $hash{['d','e','f']} = 'two';
            $hash{'g'}           = 'three';
            $hash{['h','i','j']} = 'four';
            @slotkeys = $thm->slotlist(1);
      will produce ('b','e', undef, 'i')

    All the keys at index '1' for the groups to which they were added, in the order which the FIRST KEY in the group was added to the %hash. If there is no key in the specified slot, an undef is returned for that position.

  • $thm->size;

    Returns the number of ITEMS in the hash (not the number of keys). Should be faster than ... scalar @values

  • $thm->consolidate;


    Consolidate all keys with the same values into common groups.

      returns: number of consolidated key groups
  • @ordered_vals = $thm->vals();

    Return a list of values in the order they were added.

  • ($href,$thm) = $thm->clone();

    This method returns an UNBLESSED reference to an anonymous tied %hash that is a deep copy of the parent object.

      input:        none
      returns:      unblessed tied %hash reference,
                    object handle

    To get the object handle from \%hash use this.

            $thm = tied %{$href};

    In SCALAR context it returns the unblessed %hash pointer. In ARRAY context it returns the unblessed %hash pointer and the package object/method pointer.

            $newRef = $thm->clone();
            $newRref->{'a','b'} = 'content'
            $newThm = tied %{$newRef};
  • $new_thm = $thm->copy(tie %new,'Tie::Hash::MultiKey');

    This method deep copies a MultiKey %hash to another new %hash. It may be invoked on an existing tied object handle or a reference to a tied %hash.

      input:        object handle OR reference to tied %hash
      returns:      object handle / method pointer
            $thm = tie %hash,'Tie::Hash::MultiKey';
            $newThm = $thm->copy(tie %new,'Tie::Hash::MultiKey');
            tie %new,'Tie::Hash::MultiKey');
            $newThm = $thm->copy(\%new);

    NOTE: this method duplicates the data stored in the parent %hash, overwriting and destroying anything that may have been stored in the copy target.


A tied multikey %hash behave like a regular %hash for most operations;

$value = $hash{$key} returns the key group value

$hash{$key} = $value sets the value for the key group

  i.e. all keys in the group will return that value

$hash{$key1,$key2} = $value assigns $value to the key key group consisting of $key1, $key2 if they do not. If at least one of the keys already exists, the remaining keys are assigned to the key group and the value is set for the entire group.

Better syntax $hash{[$key,$key]} = $value;

delete $hash{$key} deletes the ENTIRE key group to which $key belongs.

delete $hash($key1,$key2) deletes ALL groups to which $key1 and $key2 belong.

Better syntax delete $hash{[$key1,$key2]};

keys %hash returns all keys.

values %hash returns all values

NOTE: that this will not be the same number of items as returned by keys unless there are no key groups containing more than one key.

($k,$v) = each %hash behaves as expected.

References to tied %hash behave in the same manner as regular %hash's except as noted for multiple key values above.


SLICE operations will produce unusual results if you try to use regular ARRAYS to specify key groups in the slice. Tie::Hash::MultiKey %hash's only accept SCALAR or ARRAY_REF arguments for SLICE and direct assigment.

        %WRONG = (
                one     => 1,
                two     => 2,
                (3,4,5) => 12 # expands to 3 => 4, 5 => 12

        %hash = ( # OK
                one     => 1,
                two     => 2,
                [3,4,5] => 12

will produce a psuedo hash of the form:

        %hash = (
                one     => 1,
                two     => 2,
                3       => 12, --|
                4       => 12, --|
                5       => 12  --|

where the operation $hash{4} = 99 will change the hash to:

        %hash = (
                one     => 1,
                two     => 2,
                3       => 99, --|
                4       => 99, --|
                5       => 99  --|

Example: $hp = \%hash;

  @{$hp}{'one','two','[3,4,5]} = (1,2,12);

produces the same result as above. If the hash already contains a KEY of the same name, the value will be changed for all other shared keys.


If you are using ARRAY_REF's as keys (not as pointers to keys as above) they must be blessed into some other package so that

        ref $key ne 'ARRAY'

i.e. bless $key, 'KEY'; # or anything other than 'ARRAY'


Example SLICE assignments

TO tied hash

        @tiedhash{@keys} = @values;

        $hp = \%tiedhash;
        @{$hp}{@keys} =  @values;

FROM tied hash

        @values = @tiedhash{@keys};

        $hp = \%tiedhash;
        @values = @{$hp}{@keys};

NOTE: when assigning TO the hash, keys may be ARRAY_REF's as described above.

Extension of this module

This module has extension capabilities that allow adding features to the characteristics of the elements within the tied hash. For example, knowing the order that items in the hash are accessed as in a cache where older items are timed out and removed from the cache.

The extensions can be customized to a particular instance of a tied object. This means that extensions can be embodied as a new module or as customization within a Perl program for a particular object instance.


An extension 6 Required and 7 Optional callback subrefs to support the following operations:

  TIE       O   create the tied object extension
  FETCH     R   recall value operations
  STORE     R   save and update operations
  DELETE    R   delete key set + value operations
  EXISTS    O   checking to see if key exists
  NEXT      O   iterative operations (Perl each)
  COPY      R   hash copy and clone operations
  CLEAR     R   hash clear operations
  ADDKEY    O   add a key to existing key set
  DELKEY    O   delete a key from an existing key set
  REORDERK  O   operation to re-order the key indices
                that tracks the order that keys are
                added to the tied hash
  REORDERV  R   operation to re-order the value indices
                for values belonging to unique key sets or more data elements with any key name
     as required by the extension
  CONSOLD   O   operation to consolidate keys that
                have a common value

  DATAn         any scalar, array_ref, hash_ref


  require Tie::Hash::MultiKey;

  tie %x, 'Tie::Hash::MultiKey',
        TIE      =>     $subref_tie,
        FETCH    =>     $subref_fetch,
        STORE    =>     $subref_store,
        DELETE   =>     $subref_delete,
        EXISTS   =>     $subref_exists,
        NEXT     =>     $subref_next,
        CLEAR    =>     $subref_clear
        COPY     =>     $subref_copy,
        ADDKEY   =>     $subref_addkey,
        DELKEY   =>     $subref_delkey
        REORDERK =>     $subref_Korder,
        REORDERV =>     $subref_Vorder,
        CONSOLD  =>     $subref_consolidate;

  The extension may also be provisioned as a hash_ref.

NOTE: about internal re-ordering.

If the tied object has new keys or key sets added more than 2^48 times, the internal accounting mechanism will re-order the indices to prevent the pointers from converting from unique integer value to floats. Extensions that are tied either to the order of key addition or values for a key set must correct their associated pointers to match internal re-ordering.

  See:  t/Extension.t for usage and testing examples
  See:  Tie::Hash::MultiKeyCache for implementation

The callbacks return the following arguments:


  A pointer to pre-extension blessed tied hash object

  IMPORTANT: add extension storage to

        $self->[16] and beyond

  next is called ONLY if the key exists and
  is immediately followed by a call to the internal
  FETCH method. Normally no action should be done.

  A pointer to the the tied hash object
  The original key used for the call to fetch
  The internal value index hash key

NOTE: the primary key hash $self->[0] must not be touched by the $sub_next extension or it will mess up the Perl iterator.


  A pointer to the tied hash object
  A pointer to an array of the keys for the store
  The internal value index hash pointer

  A pointer to the tied hash object
  A pointer to an ordered array of the deleted keys
  A pointer to an ordered array of the deleted values

  exists is called ONLY if the key exists;

  A pointer to the the tied hash object
  The original key used for the operation


  A pointer to the tied hash object
  The reference key used to identify the key set
  The internal value index for key set
  A list of new keys added


  A pointer to the tied hash object
  The value of the key being deleted
  The internal value index for the key set
  else false

Calls extension_sub_delete if the key is the last key of a key set.


  A pointer to the tied hash object
  A pointer to the tied hash copy object
  A pointer to an array internal value index keys

  A pointer to the tied hash object
  A pointer to a hash of the reorder
  key order transfomation

        key => new_order_value

  A pointer to the tied hash object
  A pointer to a hash of the reorder to
        value hash transformation

        old_order => new_ord


  A pointer to the tied hash object
  A pointer to a hash as consolidated of
        value => [keys]
  A pointer to hash as consolidated of 
        keys => order
  A pointer to hash of
        new vi => [old vi order]
  %n2o is a map of new value indices after
  consolidation to an array of old value
  indices. i.e. if there were tow values
  belonging to different key sets then there
  would be two vi's in the old order array
  represented by the single vi key.

The internal structure of the tied hash object is as follows:


 0  =>  {       # $kh
        key     => vi     # value index for 1 & 2 below
 1  =>  {       # $vh
        vi      => value, # contains value for the key set
 2  =>  {       # $sh   pointer to hash list keys in a key set
        vi      = {key1 => order1, key2 => order2, ...},
 3  =>  vi,     # numeric value of next value index
 4  =>  or,     # numeric value of next key order
 5  =>  crumbs  # STORE key value
 6  =>  reserved
 7  =>  {       # extensions
   FETCH    => subref,  # required
   STORE    => subref,  # required
   DELETE   => subref,  # required
   COPY     => subref,  # required
   CLEAR    => subref,  # required
   REORDERV => subref,  # required
   TIE      => subref,  # optional
   EXISTS   => subref,  # optional
   NEXT     => subref,  # optional
   ADDKEY   => subref,  # optional
   DELKEY   => subref,  # optional
   REORDERK => subref,  # optional
   CONSOLD  => subref, # optional
 ... one or more data keys
   DATAn     => scalar, array_ref, hash_ref

Extension writers should store new information in the indices 16 and up.

Developers of extensions are encouraged to read the code.


Michael Robinton, <>


Copyright 2014, Michael Robinton

This program is free software; you may redistribute it and/or modify it under the same terms as Perl itself.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

1 POD Error

The following errors were encountered while parsing the POD:

Around line 666:

Expected '=item *'