NAME

LinkedHash - A Hash map with insertion-order iteration and navigation.

SYNOPSIS

use v6;
use LinkedHash:auth<github:smls>;

my %steps is LinkedHash =
    add  => "Add all cookie ingredients.",
    mix  => "Mix until your hands hurt.",
    roll => "Roll out the dough into small balls."
;

%steps<mix> = "Mix until smooth.";    # updated but not moved
%steps<bake> = "Bake until golden.";  # inserted at the end

# Iterating:

say %steps.keys;       # (add mix roll bake)
say %steps.values;     # ..same order..
say %steps.kv;         # ..same order..
for %steps -> $pair {  # ..same order..
    ...
}

DESCRIPTION

This module provides a hash map type for Perl 6 which behaves the same as the built-in Hash type, except that it:

• Makes sure that iterating over the entries returns them in the order in which they were inserted, rather than in a randomized order.
• Has some limitations (see #LIMITATIONS).
• Will hopefully expose some kind of "linked list" navigation API in the future (see #TODO).

It's implemented by internally keeping a linked list between hash entries, where each entry knows the ones inserted directly before and after it.

Performance characteristics

The LinkedHash type makes the following computational complexity guarantees:

operation	time complexity class
looking up an entry by key	same as Hash
inserting/removing an entry	same as Hash
iterating over all entries	O(n)
starting iteration at a specific key	like lookup, but API not exposed yet
navigating to a neighboring entry	O(1), but API not exposed yet

It has the following memory characteristics:

• Requires more memory than a Hash.
• When an entry is deleted with the :delete adverb, none of it is retained (to the extent that Hash doesn't either), so you don't have to worry about memory leaks or slow-downs when adding and removing entries frequently.

API

Constructing

The assignment operator works the same as for the built-in Hash type, accepting Pairs and/or alternating keys and values:

my %fruit is LinkedHash =
    'kiwi' => 'green',
    'lemon' => 'yellow',
    'cherry' => 'red',
;

my %fruit is LinkedHash =
    'kiwi', 'green', 'lemon', 'yellow', 'cherry', 'red';

Same for the .new constructor:

my %fruit := LinkedHash.new(
    'kiwi', 'green', 'lemon' => 'yellow', 'cherry', 'red');

Subscripting

Same as for Hash - refer to the Perl 6 Subscripts documentation page.

All the adverbs (:k, :v, :kv, :p, :exists, :delete) are supported.

Filling

Values can be added to a LinkedHash in the same ways as to a Hash:

• By assigning to a %-sigiled variable bound to the LinkedHash (see #Constructing).
• By assigning to a subscript (see #Subscripting).
• Using one of the four entry-adding methods – refer to their description in the Perl 6 Hash documentation:
  • .push
  • .append
  • .classify-list
  • .categorize-list

Iterating

Same as with Hash, except that entries are returned in insertion order.

Insertion order is defined as follows:

• When a new hash entry is created (i.e. for a key that does not yet exist in the hash, as reported by the :exists adverb), then it is placed at the end.
• When the value of an existing hash entry is assigned to or otherwise modified, its place in the iteration order does not change.
• When a hash entry is deleted (with the :delete adverb), it is as if it was never there, and re-adding the same key adds it to the end again.

Refer to the description of the respective methods in the Perl 6 Hash documentation:

• .elems
• .keys
• .values
• .kv
• .pairs
• .antipairs
• .invert
• .iterator

Nagivating

Not yet implemented, see #TODO.

Serializing

Same as with Hash; Refer to the description of the respective methods in the Perl 6 documentation:

• .Str
• .gist
• .perl
• .fmt

Coercing

The following coercion method are available on LinkedHash instances, with the same behavior as on a Hash (except that insertion order is preserved when coercing to list-like types):

.Array .Bag .BagHash .Bool .Capture .Complex .FatRat .Hash .hash .Int .List
.list .Mix .MixHash .Num .Numeric .Rat .Real .Seq .Set .SetHash .Str .UInt

LIMITATIONS

• Slower than a normal Hash for things which they both support.

• Does not support direct slipping into argument lists like a normal Hash does (e.g. some-function |%foo) – Rakudo issue #1966 will need to be resolved first. As a work-around, use some-function |%foo.Hash for now.

• Does not support default values like a normal Hash does (e.g. my %foo is default(42)).

• Does not support parameterization like a normal Hash does (e.g. Hash[Int, Any]).

• There are no coercion methods from other built-in types, like are available for a normal Hash (e.g. @pairs.Hash).

Some of these may change in future releases.

THREAD-SAFETY

Same as Hash, which is to say:

• It is not safe to simultaneously modify the same LinkedHash object on different threads, or to read from it on one thread while another modifies it.

• It is safe to simultaneously read from the same LinkedHash object on different threads, or to read/modify different LinkedHash objects on different threads.

TODO

For this first alpha release, the "Linked" part is merely an implementation detail. In future releases, I'd like to expose some of the O(1) navigation features that the underlying doubly linked list makes possible, for example:

# Looking up neighboring entries (conjectural):

say %steps<mix>:pred;    # Add all cookie ingredients.
say %steps<mix>:pred:k;  # add
say %steps<mix>:succ:k;  # roll

# Starting iteration at specific entries (conjectural):

say %steps<mix>:to-end:k;    # (mix roll bake)
say %steps<mix>:to-start:k;  # (mix add)

...but I'm not yet sure how practical it really is to add adverbs to the .{ } postcircumfix operator, or if there's a better way to design this API.

Furthermore, I'd like to eventually figure out if it would be both sensible and feasible to:

• Let LinkedHash inherit from Hash (or maybe Map).

• Let .hash return self instead of .Hash.

• Let nested Hashes created by auto-vivification be LinkedHashes themselves.

• Let nested Hashes created by .push, .append, .classify-list, and .categorize-list be LinkedHashes themselves.

• Use a custom implementation for the Seqs (or associated Iterators) which are returned from the iteration methods, so as to improve the performance of things like:

  %h.values.tail   # Could immediately return the last element!
  %h.keys.reverse  # Could immediately go backwards from the last element!

• Implement the missing features listed under #LIMITATIONS.

CONTRIBUTING

If you find bugs (including any undocumented discrepancies in behavior between this type and the built-in Hash type), or want to help with implementing more features, please file an issue or send a pull request.

SEE ALSO

This module was inspired by Java's LinkedHashMap.

Similar Perl 6 modules:

• OrderedHash by FCO
• ArrayHash by zostay

AUTHOR

Sam S. (smls), [email protected]

VERSION

0.1

(Note: this is alpha-stage software; the API may still change.)

LICENSE

This is free software; you can redistribute it and/or modify it under the terms of the Artistic License 2.0 (see the accompanying LICENSE file).