Brad Baxter


Data::Omap - Perl module to implement ordered mappings


 use Data::Omap;
 # Simple OO style
 my $omap = Data::Omap->new( [{a=>1},{b=>2},{c=>3}] );
 $omap->set( a => 0 );
 $omap->add( b2 => 2.5, 2 );  # insert at position 2 (between b and c)
 my $value  = $omap->get_values( 'c' );    # 3
 my @keys   = $omap->get_keys();           # (a, b, b2, c)
 my @values = $omap->get_values();         # (0, 2, 2.5, 3)
 my @subset = $omap->get_values(qw(c b));  # (2, 3) (values are data-ordered)
 # Tied style
 my %omap;
 # recommend saving an object reference, too.
 my $omap = tie %omap, 'Data::Omap', [{a=>1},{b=>2},{c=>3}];
 $omap{ a } = 0;
 $omap->add( b2 => 2.5, 2 );  # there's no tied hash equivalent
 my $value  = $omap{ c };
 my @keys   = keys %omap;      # $omap->get_keys() is faster 
 my @values = values %omap;    # $omap->get_values() is faster
 my @slice  = @omap{qw(c b)};  # (3, 2) (slice values are parameter-ordered)

 # Non-OO style

 use Data::Omap ':ALL';
 my $omap = [{a=>1},{b=>2},{c=>3}];  # new-ish, but not blessed

 omap_set( $omap, a => 0 );        # (pass omap as first parameter)
 omap_add( $omap, b2 => 2.5, 2 );  # insert at position 2 (between b and c)
 my $value  = omap_get_values( $omap, 'c' );      # 3
 my @keys   = omap_get_keys( $omap );             # (a, b, b2, c)
 my @values = omap_get_values( $omap );           # (0, 2, 2.5, 3)
 my @subset = omap_get_values( $omap, qw(c b) );  # (2, 3) (values are data-ordered)
 # There are more methods/options, see below.


This module implements the Data::Omap class. Objects in this class are ordered mappings, i.e., they are hashes in which the key/value pairs are in order. This is defined in shorthand as !!omap in the YAML tag repository:

The keys in Data::Omap objects are unique, like regular hashes.

A closely related class, Data::Pairs, implements the YAML !!pairs data type, Data::Pairs objects are also ordered sequences of key:value pairs but they allow duplicate keys.

While ordered mappings are in order, they are not necessarily in a particular order, i.e., they are not necessarily sorted in any way. They simply have a predictable set order (unlike regular hashes whose key/value pairs are in no set order).

By default, Data::Omap will add new key/value pairs at the end of the mapping, but you may request that they be merged in a particular order with the order() class method.

However, even though Data::Omap will honor the requested order, it will not attempt to keep the mapping in that order. By passing position values to the set() and add() methods, you may insert new pairs anywhere in the mapping and Data::Omap will not complain.


Normally, the underlying structure of an OO object is encapsulated and not directly accessible (when you play nice). One key implementation detail of Data::Omap is the desire that the underlying ordered mapping data structure (an array of single-key hashes) be publically maintained as such and directly accessible if desired.

To that end, no attributes but the data itself are stored in the objects. In the current version, that is why order() is a class method rather than an object method. In the future, inside-out techniques may be used to enable object-level ordering.

This data structure is inefficient in several ways as compared to regular hashes: rather than one hash, it contains a separate hash per key/value pair; because it's an array, key lookups (in the current version) have to loop through it.

The advantage if using this structure is simply that it "natively" matches the structure defined in YAML. So if the (unblessed) structure is dumped using YAML (or perhaps JSON), it may be read as is by another program, perhaps in another language. It is true that this could be accomplished by passing the object through a formatting routine, but I wanted to see first how this implementation might work.


Data::Omap version 0.06



Constructs a new Data::Omap object.

Accepts array ref containing single-key hash refs, e.g.,

 my $omap = Data::Omap->new( [ { a => 1 }, { b => 2 }, { c => 3 } ] );

When provided, this data will be loaded into the object.

Returns a reference to the Data::Omap object.

Data::Omap->order( [$predefined_ordering | coderef] );

When ordering is ON, new key/value pairs will be added in the specified order. When ordering is OFF (the default), new pairs will be added to the end of the mapping.

When called with no parameters, order() returns the current code reference (if ordering is ON) or a false value (if ordering is OFF); it does not change the ordering.

 Data::Omap->order();         # leaves ordering as is

When called with the null string, '', ordering is turned OFF.

 Data::Omap->order( '' );     # turn ordering OFF (the default)

Otherwise, accepts the predefined orderings: 'na', 'nd', 'sa', 'sd', 'sna', and 'snd', or a custom code reference, e.g.

 Data::Omap->order( 'na' );   # numeric ascending
 Data::Omap->order( 'nd' );   # numeric descending
 Data::Omap->order( 'sa' );   # string  ascending
 Data::Omap->order( 'sd' );   # string  descending
 Data::Omap->order( 'sna' );  # string/numeric ascending
 Data::Omap->order( 'snd' );  # string/numeric descending
 Data::Omap->order( sub{ int($_[0]/100) < int($_[1]/100) } );  # code

The predefined orderings, 'na' and 'nd', compare keys as numbers. The orderings, 'sa' and 'sd', compare keys as strings. The orderings, 'sna' and 'snd', compare keys as numbers when they are both numbers, as strings otherwise.

When defining a custom ordering, the convention is to use the operators < or lt between (functions of) $_[0] and $_[1] for ascending and between $_[1] and $_[0] for descending.

Returns the code reference if ordering is ON, a false value if OFF.

Note, when object-level ordering is implemented, it is expected that the class-level option will still be available. In that case, any new objects will inherit the class-level ordering unless overridden at the object level.


$omap->set( $key => $value[, $pos] );

Sets the value if $key exists; adds a new key/value pair if not.

Accepts $key, $value, and optionally, $pos.

If $pos is given, and there is a key/value pair at that position, it will be set to $key and $value, even if the key is different. For example:

 my $omap = Data::Omap->new( [{a=>1},{b=>2}] );
 $omap->set( c => 3, 0 );  # omap is now [{c=>3},{b=>2}]

(As implied by the example, positions start at 0.)

If $pos is given, and there isn't a pair there, a new pair is added there (perhaps overriding a defined ordering).

If $pos is not given, the key will be located and if found, the value set. If the key is not found, a new pair is added to the end or merged according to the defined order().

Note that set() will croak if a duplicate key would result. This would only happen if $pos is given and the $key is found--but not at that position.

Returns $value (as a nod toward $hash{$key}=$value, which "returns" $value).

$omap->get_values( [$key[, @keys]] );

Get a value or values.

Regardless of parameters, if the object is empty, undef is returned in scalar context, an empty list in list context.

If no parameters, gets all the values. In scalar context, gives number of values in the object.

 my $omap = Data::Omap->new( [{a=>1},{b=>2},{c=>3}] );
 my @values  = $omap->get_values();  # (1, 2, 3)
 my $howmany = $omap->get_values();  # 3

If one key is given, that value is returned--regardless of context--or if not found, undef.

 @values   = $omap->get_values( 'b' );  # (2)
 my $value = $omap->get_values( 'b' );  # 2

If multiple keys given, their values are returned in the order found in the object, not the order of the given keys (unlike hash slices which return values in the order requested).

In scalar context, gives the number of values found, e.g.,

 @values  = $omap->get_values( 'c', 'b', 'A' );  # (2, 3)
 $howmany = $omap->get_values( 'c', 'b', 'A' );  # 2

The hash slice behavior is available if you use tie, see below.

$omap->add( $key => $value[, $pos] );

Adds a key/value pair to the object.

Accepts $key, $value, and optionally, $pos.

If $pos is given, the key/value pair will be added (inserted) there (possibly overriding a defined order), e.g.,

 my $omap = Data::Omap->new( [{a=>1},{b=>2}] );
 $omap->add( c => 3, 1 );  # omap is now [{a=>1},{c=>3},{b=>2}]

(Positions start at 0.)

If $pos is not given, a new pair is added to the end or merged according to the defined order().

Note that add() will croak if a duplicate key would result, i.e., if the key being added is already in the object.

Returns $value.

omap_add_ordered( $omap, $key => $value );

Private routine used by set() and add().

Accepts $key and $value.

Adds a new key/value pair to the end or merged according to the defined order().

This routine should not be called directly, because it does not check for duplicates.

Has no defined return value.

$omap->get_pos( $key );

Gets position where a key is found.

Accepts one key (any extras are silently ignored).

Returns the position or undef (if key not found), regardless of context, e.g.,

 my $omap = Data::Omap->new( [{a=>1},{b=>2},{c=>3}] );
 my @pos  = $omap->get_pos( 'b' );  # (1)
 my $pos  = $omap->get_pos( 'b' );  # 1

Returns undef/() if no key given or object is empty.

$omap->get_pos_hash( @keys );

Gets positions where keys are found.

Accepts zero or more keys.

In list context, returns a hash of keys/positions found. In scalar context, returns a hash ref to this hash. If no keys given, all the positions are mapped in the hash.

 my $omap     = Data::Omap->new( [{a=>1},{b=>2},{c=>3}] );
 my %pos      = $omap->get_pos_hash( 'c', 'b' ); # %pos      is (b=>1,c=>2)
 my $pos_href = $omap->get_pos_hash( 'c', 'b' ); # $pos_href is {b=>1,c=>2}

If a given key is not found, it will not appear in the returned hash.

Returns undef/() if object is empty.

$omap->get_keys( @keys );

Gets keys.

Accepts zero or more keys. If no keys are given, returns all the keys in the object (list context) or the number of keys (scalar context), e.g.,

 my $omap    = Data::Omap->new( [{a=>1},{b=>2},{c=>3}] );
 my @keys    = $omap->get_keys();  # @keys is (a, b, c)
 my $howmany = $omap->get_keys();  # $howmany is 3

If one or more keys are given, returns all the keys that are found (list) or the number found (scalar). Keys returned are listed in the order found in the object, e.g.,

 @keys    = $omap->get_keys( 'c', 'b', 'A' );  # @keys is (b, c)
 $howmany = $omap->get_keys( 'c', 'b', 'A' );  # $howmany is 2

$omap->get_array( @keys );

Gets an array of key/value pairs.

Accepts zero or more keys. If no keys are given, returns a list of all the key/value pairs in the object (list context) or an array reference to that list (scalar context), e.g.,

 my $omap    = Data::Omap->new( [{a=>1},{b=>2},{c=>3}] );
 my @array   = $omap->get_array();  # @array is ({a=>1}, {b=>2}, {c=>3})
 my $aref    = $omap->get_array();  # $aref  is [{a=>1}, {b=>2}, {c=>3}]

If one or more keys are given, returns a list of key/value pairs for all the keys that are found (list) or an aref to that list (scalar). Pairs returned are in the order found in the object, e.g.,

 @array = $omap->get_array( 'c', 'b', 'A' );  # @array is ({b->2}, {c=>3})
 $aref  = $omap->get_array( 'c', 'b', 'A' );  # @aref  is [{b->2}, {c=>3}]

Note, conceivably this method might be used to make a copy (unblessed) of the object, but it would not be a deep copy (if values are references, the references would be copied, not the referents).


Expects no parameters. Returns the first key in the object (or undef if object is empty).

This routine supports the tied hash FIRSTKEY method.

$omap->nextkey( $lastkey );

Accepts one parameter, the last key gotten from FIRSTKEY or NEXTKEY.

Returns the next key in the object.

This routine supports the tied hash NEXTKEY method.

$omap->exists( $key );

Accepts one key.

Returns true if key is found in object, false if not.

This routine supports the tied hash EXISTS method, but may reasonably be called directly, too.

$omap->delete( $key );

Accepts one key. If key is found, removes the key/value pair from the object.

Returns the value from the deleted pair.

This routine supports the tied hash DELETE method, but may be called directly, too.


Expects no parameters. Removes all key/value pairs from the object.

Returns an empty list.

This routine supports the tied hash CLEAR method, but may be called directly, too.


An ordered mapping (as defined here) is an array of single-key hashes. It is possible to manipulate an ordered mapping directly without first blessing it with new(). Most methods have a corresponding exportable subroutine named with the prefix, omap_, e.g., omap_set(), omap_get_keys(), etc.

To call these subroutines, pass the array reference as the first parameter, e.g., instead of doing $omap->set( a => 1), do omap_set( $omap, a => 1).


Nothing is exported by default. All subroutines may be exported using :ALL, e.g.,

 use Data::Omap ':ALL';

They are shown below.

A subset may be exported using :STD, e.g.,

 use Data::Omap ':STD';

This subset includes omap_set() omap_get_values() omap_get_keys() omap_exists() omap_delete() omap_clear()

new without new()

To create an ordered mapping from scratch, simply assign an empty array ref, e.g.,

 my $omap = [];

omap_order( $omap[, $predefined_ordering | coderef] );

(See Data::Omap->order() above.)

 omap_order( $omap, 'na' );   # numeric ascending
 omap_order( $omap, 'nd' );   # numeric descending
 omap_order( $omap, 'sa' );   # string  ascending
 omap_order( $omap, 'sd' );   # string  descending
 omap_order( $omap, 'sna' );  # string/numeric ascending
 omap_order( $omap, 'snd' );  # string/numeric descending
 omap_order( $omap, sub{ int($_[0]/100) < int($_[1]/100) } );  # code

omap_set( $omap, $key => $value[, $pos] );

(See $omap->set() above.)

 my $omap = [{a=>1},{b=>2}];
 omap_set( $omap, c => 3, 0 );  # omap is now [{c=>3},{b=>2}]

omap_get_values( $omap[, $key[, @keys]] );

(See $omap->get_values() above.)

 my $omap = [{a=>1},{b=>2},{c=>3}];
 my @values  = omap_get_values( $omap );  # (1, 2, 3)
 my $howmany = omap_get_values( $omap );  # 3
 @values   = omap_get_values( $omap, 'b' );  # (2)
 my $value = omap_get_values( $omap, 'b' );  # 2
 @values  = omap_get_values( $omap, 'c', 'b', 'A' );  # (2, 3)
 $howmany = omap_get_values( $omap, 'c', 'b', 'A' );  # 2

omap_add( $omap, $key => $value[, $pos] );

(See $omap->add() above.)

 my $omap = [{a=>1},{b=>2}];
 omap_add( $omap, c => 3, 1 );  # omap is now [{a=>1},{c=>3},{b=>2}]

omap_get_pos( $omap, $key );

(See $omap->get_pos() above.)

 my $omap = [{a=>1},{b=>2},{c=>3}];
 my @pos  = omap_get_pos( $omap, 'b' );  # (1)
 my $pos  = omap_get_pos( $omap, 'b' );  # 1

omap_get_pos_hash( $omap[, @keys] );

(See $omap->get_pos_hash() above.)

 my $omap     = [{a=>1},{b=>2},{c=>3}];
 my %pos      = omap_get_pos_hash( $omap, 'c', 'b' ); # %pos      is (b=>1,c=>2)
 my $pos_href = omap_get_pos_hash( $omap, 'c', 'b' ); # $pos_href is {b=>1,c=>2}

omap_get_keys( $omap[, @keys] );

(See $omap->get_keys() above.)

 my $omap    = [{a=>1},{b=>2},{c=>3}];
 my @keys    = omap_get_keys( $omap );  # @keys is (a, b, c)
 my $howmany = omap_get_keys( $omap );  # $howmany is 3

 @keys    = omap_get_keys( $omap, 'c', 'b', 'A' );  # @keys is (b, c)
 $howmany = omap_get_keys( $omap, 'c', 'b', 'A' );  # $howmany is 2

omap_get_array( $omap[, @keys] );

(See $omap->get_array() above.)

 my $omap    = [{a=>1},{b=>2},{c=>3}];
 my @array   = omap_get_array( $omap );  # @array is ({a=>1}, {b=>2}, {c=>3})
 my $aref    = omap_get_array( $omap );  # $aref  is [{a=>1}, {b=>2}, {c=>3}]

 @array = omap_get_array( $omap, 'c', 'b', 'A' );  # @array is ({b->2}, {c=>3})
 $aref  = omap_get_array( $omap, 'c', 'b', 'A' );  # @aref  is [{b->2}, {c=>3}]

omap_exists( $omap, $key );

(See $omap->exists() above.)

 my $bool = omap_exists( $omap, 'a' );

omap_delete( $omap, $key );

(See $omap->delete() above.)

 omap_delete( $omap, 'a' );

omap_clear( $omap );

(See $omap->clear() above.)

 omap_clear( $omap );

Or simply:

 @$omap = ();



    Use Tie::IxHash if what you need is an ordered hash in general. The Data::Omap module does repeat many of Tie::IxHash's features. What differs is that it operates directly on a specific type of data structure. Whether this pans out in the long run remains to be seen.


    The code in Data::Omap is the basis for that in the Data::Pairs module. Data::Pairs also operates on an ordered hash, but allows duplicate keys.


Brad Baxter, <>


Copyright (C) 2008 by Brad Baxter

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.8 or, at your option, any later version of Perl 5 you may have available.