Doubly-linked list/Element definition

From Rosetta Code
Jump to: navigation, search
Doubly-linked list/Element definition
You are encouraged to solve this task according to the task description, using any language you may know.

Define the data structure for a doubly-linked list element. The element should include a data member to hold its value and pointers to both the next element in the list and the previous element in the list. The pointers should be mutable.

See also


[edit] Ada

type Link;
type Link_Access is access Link;
type Link is record
Next : Link_Access := null;
Prev : Link_Access := null;
Data : Integer;
end record;

Using generics, the specification might look like this:

type Element_Type is private;
package Linked_List is
type List_Type is limited private;
type List_Element;
type List_Element_Ptr is access list_element;
type List_Element is
Prev : List_Element_Ptr;
Data : Element_Type;
Next : List_Element_Ptr;
end record;
type List_Type is
Head  : List_Element_Ptr; -- Pointer to first element.
Tail  : List_Element_Ptr; -- Pointer to last element.
Cursor  : List_Element_Ptr; -- Pointer to cursor element.
Count  : Natural := 0; -- Number of items in list.
Traversing  : Boolean := False; -- True when in a traversal.
end record;
end Linked_List;

In Ada 2005 this example can be written without declaration of an access type:

type Link is limited record
Next : not null access Link := Link'Unchecked_Access;
Prev : not null access Link := Link'Unchecked_Access;
Data : Integer;
end record;

Here the list element is created already pointing to itself, so that no further initialization is required. The type of the element is marked as limited indicating that such elements have referential semantics and cannot be copied.

Ada's standard container library includes a generic doubly linked list. The structure of the link element is private.

[edit] ALGOL 68

Works with: ALGOL 68 version Revision 1.
Works with: ALGOL 68G version Any - tested with release algol68g-2.7.
Works with: ELLA ALGOL 68 version Any (with appropriate job cards) - tested with release 1.8-8d
File: prelude/link.a68
# -*- coding: utf-8 -*- #
MODE OBJVALUE = ~ # Mode/type of actual obj to be queued #
OBJVALUE value # ... etc. required #
PROC obj link free = (REF OBJLINK free)VOID:
prev OF free := next OF free := obj queue empty # give the garbage collector a big hint #
See also: Queue/Usage

[edit] AutoHotkey

see Doubly-linked list/AutoHotkey

[edit] Axe


[edit] BBC BASIC

      DIM node{pPrev%, pNext%, iData%}

[edit] C

struct link {
struct link *next;
struct link *prev;
int data;

[edit] C++

C++ has doubly linked list class template in standard library. However actual list noded are treated as implementation detail and encapsulated inside list. If we were to reimplement list, then node could look like that:

template <typename T>
struct Node
Node* next;
Node* prev;
T data;

[edit] C#

class Link
public int item;
public Link next;
public Link prev;

[edit] Clojure

This sort of mutable structure is not idiomatic in Clojure. Doubly-linked list/Definition#Clojure or a finger tree implementation would be better.

(defrecord Node [prev next data])
(defn new-node [prev next data]
(Node. (ref prev) (ref next) data))

[edit] Common Lisp

(defstruct dlist head tail)
(defstruct dlink content prev next)

See the functions on the Doubly-Linked List page for the usage of these structures.

[edit] D

A default constructor is implicit:

struct Node(T) {
T data;
typeof(this)* prev, next;
void main() {
alias N = Node!int;
N* n = new N(10);

[edit] Delphi

struct Node(T) {
pList = ^List ;
list = record
data : pointer ;
prev : pList ;
next : pList ;

[edit] E

This does no type-checking, under the assumption that it is being used by a containing doubly-linked list object which enforces that invariant along with others such as that element.getNext().getPrev() == element. See Doubly-Linked List#E for an actual implementation (which uses slightly more elaborate nodes than this).

def makeElement(var value, var next, var prev) {
def element {
to setValue(v) { value := v }
to getValue() { return value }
to setNext(n) { next := n }
to getNext() { return next }
to setPrev(p) { prev := p }
to getPrev() { return prev }
return element

[edit] Erlang

Using the code in Doubly-linked_list/Definition the element is defined by:

new( Data ) -> erlang:spawn( fun() -> loop( Data, noprevious, nonext ) end ).

[edit] Fortran

In ISO Fortran 95 or later:

type node
real :: data
type(node), pointer :: next => null(), previous => null()
end type node
! . . . .
type( node ), target :: head

[edit] Go

type dlNode struct {
next, prev *dlNode

Or, using the container/list package:

import "container/list"
var node list.Element
// and using: node.Next(), node.Prev(), node.Value

[edit] Haskell

Haskell in general doesn't have mutability so the following 'mutator' functions use lazy evaluation instead.

Note that unlike naive pointer manipulation which could corrupt the doubly-linked list, updateLeft and updateRight will always yield a well-formed data structure.

data DList a = Leaf | Node (DList a) a (DList a)
updateLeft _ Leaf = Leaf
updateLeft Leaf (Node _ v r) = Node Leaf v r
updateLeft new@(Node nl _ _) (Node _ v r) = current
where current = Node prev v r
prev = updateLeft nl new
updateRight _ Leaf = Leaf
updateRight Leaf (Node l v _) = Node l v Leaf
updateRight new@(Node _ _ nr) (Node l v _) = current
where current = Node l v next
next = updateRight nr new

[edit] Icon and Unicon

Uses Unicon classes.

class DoubleLink (value, prev_link, next_link)
initially (value, prev_link, next_link)
self.value := value
self.prev_link := prev_link # links are 'null' if not given
self.next_link := next_link

[edit] J

As discussed in Doubly-linked_list/Definition#J, doubly linked lists are antithetical to J's design. Defining individual elements as independent structures is even worse. Now each element of the list must contain three arrays (everything in J is an array), all so that we can implement a list.

Yo Dawg, we heard you like lists, so we put lists in your lists so you can list while you list.

Nevertheless, this is doable, though it necessarily departs from the definition specified at Doubly-linked_list/Definition#J.

create=:3 :0
'predecessor successor data'=:y
successor__predecessor=: predecessor__successor=: this

Here, when we create a new list element, we need to specify its successor node and its predecessor node and the data to be stored in the node. To start a new list we will need a node that can be the head and the tail of the list -- this will be the successor node for the last element of the list and the predecessor node for the first element of the list:

create=:3 :0
predecessor=:successor=:this=: coname''

[edit] Java

Works with: Java version 1.5+
public class Node<T> {
private T element;
private Node<T> next, prev;
public Node<T>(){
next = prev = element = null;
public Node<T>(Node<T> n, Node<T> p, T elem){
next = n;
prev = p;
element = elem;
public void setNext(Node<T> n){
next = n;
public Node<T> getNext(){
return next;
public void setElem(T elem){
element = elem;
public T getElem(){
return element;
public void setNext(Node<T> n){
next = n;
public Node<T> setPrev(Node<T> p){
prev = p;
public getPrev(){
return prev;

For use with Java 1.4 and below, delete all "<T>"s and replace T's with "Object".

[edit] JavaScript

Inherits from LinkedList (see Singly-Linked_List_(element)#JavaScript)

function DoublyLinkedList(value, next, prev) {
this._value = value;
this._next = next;
this._prev = prev;
// from LinkedList, inherit: value(), next(), traverse(), print()
DoublyLinkedList.prototype = new LinkedList();
DoublyLinkedList.prototype.prev = function() {
if (arguments.length == 1)
this._prev = arguments[0];
return this._prev;
function createDoublyLinkedListFromArray(ary) {
var node, prev, head = new DoublyLinkedList(ary[0], null, null);
prev = head;
for (var i = 1; i < ary.length; i++) {
node = new DoublyLinkedList(ary[i], null, prev);;
prev = node;
return head;
var head = createDoublyLinkedListFromArray([10,20,30,40]);

[edit] Modula-2

Link = POINTER TO LinkRcd;
LinkRcd = RECORD
Prev, Next: Link;

[edit] Nim

Node[T] = ref TNode[T]
TNode[T] = object
next, prev: Node[T]
data: T

[edit] Objeck

class ListNode {
@value : Base;
@next : ListNode;
@previous: ListNode;
New(value : Base) {
@value := value;
method : public : Set(value : Base) ~ Nil {
@value := value;
method : public : Get() ~ Base {
return @value;
method : public : SetNext(next : Collection.ListNode) ~ Nil {
@next := next;
method : public : GetNext() ~ ListNode {
return @next;
method : public : SetPrevious(previous : Collection.ListNode) ~ Nil {
@previous := previous;
method : public : GetPrevious() ~ ListNode {
return @previous;

[edit] OCaml

[edit] Imperative

type 'a dlink = {
mutable data: 'a;
mutable next: 'a dlink option;
mutable prev: 'a dlink option;
let dlink_of_list li =
let f prev_dlink x =
let dlink = {
data = x;
prev = None;
next = prev_dlink }
begin match prev_dlink with
| None -> ()
| Some prev_dlink ->
prev_dlink.prev <- Some dlink
Some dlink
List.fold_left f None (List.rev li)
let list_of_dlink =
let rec aux acc = function
| None -> List.rev acc
| Some{ data = d;
prev = _;
next = next } -> aux (d::acc) next
aux []
let iter_forward_dlink f =
let rec aux = function
| None -> ()
| Some{ data = d;
prev = _;
next = next } -> f d; aux next
# let dl = dlink_of_list [1;2;3;4;5] in
iter_forward_dlink (Printf.printf "%d\n") dl ;;
- : unit = ()

[edit] Functional

The previous implementation is the strict equivalent of the other examples of this page and its task, but in regular OCaml these kind of imperative structures can be advantageously replaced by a functional equivalent, that can be use in the same area, which is to have a list of elements and be able to point to one of these. We can use this type:

type 'a nav_list = 'a list * 'a * 'a list

The middle element is the pointed item, and the two lists are the previous and the following items. Here are the associated functions:

let nav_list_of_list = function
| hd::tl -> [], hd, tl
| [] -> invalid_arg "empty list"
let current = function
| _, item, _ -> item
let next = function
| prev, item, next::next_tl ->
item::prev, next, next_tl
| _ ->
failwith "end of nav_list reached"
let prev = function
| prev::prev_tl, item, next ->
prev_tl, prev, item::next
| _ ->
failwith "begin of nav_list reached"
# let nl = nav_list_of_list [1;2;3;4;5] ;;
val nl : 'a list * int * int list = ([], 1, [2; 3; 4; 5])
# let nl = next nl ;;
val nl : int list * int * int list = ([1], 2, [3; 4; 5])
# let nl = next nl ;;
val nl : int list * int * int list = ([2; 1], 3, [4; 5])
# current nl ;;
- : int = 3

[edit] Oforth

Complete definition is here : Doubly-linked list/Definition#Oforth

Object Class new: DNode(value, mutable prev, mutable next)

[edit] Oz

We show how to create a new node as a record value.

fun {CreateNewNode Value}
node(prev:{NewCell _}
next:{NewCell _}

Note: this is for illustrative purposes only. In a real Oz program, you would use one of the existing data types.

[edit] Pascal

type link_ptr = ^link;
data_ptr = ^data; (* presumes that type 'data' is defined above *)
link = record
prev: link_ptr;
next: link_ptr;
data: data_ptr;

[edit] Perl

my %node = (
data => 'say what',
next => \%foo_node,
prev => \%bar_node,
$node{next} = \%quux_node; # mutable

[edit] Perl 6

role DLElem[::T] {
has DLElem[T] $.prev is rw;
has DLElem[T] $.next is rw;
has T $.payload = T;
method pre-insert(T $payload) {
die "Can't insert before beginning" unless $!prev;
my $elem = ::?$payload);
$! = $elem;
$elem.prev = $!prev;
$ = self;
$!prev = $elem;
method post-insert(T $payload) {
die "Can't insert after end" unless $!next;
my $elem = ::?$payload);
$!next.prev = $elem;
$ = $!next;
$elem.prev = self;
$!next = $elem;
method delete {
die "Can't delete a sentinel" unless $!prev and $!next;
$!next.prev = $!prev;
$! = $!next; # conveniently returns next element

[edit] PL/I

define structure
1 Node,
2 value fixed decimal,
2 back_pointer handle(Node),
2 fwd_pointer handle(Node);
P = NEW(: Node :); /* Creates a node, and lets P point at it. */
get (P => value); /* Reads in a value to the node we just created. */
/* Assuming that back_pointer and fwd_pointer point at other nodes, */
/* we can say ... */
P = P => fwd_pointer; /* P now points at the next node. */
P = P => back_pointer; /* P now points at the previous node. */

[edit] PicoLisp

We use (in addition to the header structure described in Doubly-linked list/Definition#PicoLisp) two cells per doubly-linked list element:

        +-----+-----+     +-----+-----+
        | Val |  ---+---> |  |  |  ---+---> next
        +-----+-----+     +--+--+-----+
                    prev <---+

With that, 'cddr' can be used to access the next, and 'cadr' to access the previous element.

# 'cons' an element to a doubly-linked list
(de 2cons (X DLst)
(let L (car DLst) # Get current data list
(set DLst (cons X NIL L)) # Prepend two new cons pairs
(if L # Unless DLst was empty
(set (cdr L) (car DLst)) # set new 'prev' link
(con DLst (car DLst)) ) ) ) # otherwise set 'end' link
# We prepend 'not' to the list in the previous example
(2cons 'not *DLst)

For output of the example data, see Doubly-linked list/Traversal#PicoLisp.

[edit] Pop11

uses objectclass;
define :class Link;
slot next = [];
slot prev = [];
slot data = [];

[edit] PureBasic

Structure node

[edit] Python

class Node(object):
def __init__(self, data = None, prev = None, next = None):
self.prev = prev = next = data
def __str__(self):
return str(
def __repr__(self):
return repr(
def iter_forward(self):
c = self
while c != None:
yield c
c =
def iter_backward(self):
c = self
while c != None:
yield c
c = c.prev

[edit] Racket

(define-struct dlist (head tail) #:mutable)
(define-struct dlink (content prev next) #:mutable)

See the functions on the Doubly-Linked List page for the usage of these structures.

[edit] REXX

REXX doesn't have linked lists, as there are no pointers (or handles).
However, linked lists can be simulated with lists in REXX.

/*REXX program that implements various   List Manager   functions.      */
┌─┘ Functions of the List Manager └─┐
│ │
│ @init ─── initializes the List. │
│ │
│ @size ─── returns the size of the List [could be 0 (zero)]. │
│ │
│ @show ─── shows (displays) the complete List. │
│ @show k,1 ─── shows (displays) the Kth item. │
│ @show k,m ─── shows (displays) M items, starting with Kth item. │
│ @show ,,─1 ─── shows (displays) the complete List backwards. │
│ │
│ @get k ─── returns the Kth item. │
│ @get k,m ─── returns the M items starting with the Kth item. │
│ │
│ @put x ─── adds the X items to the end (tail) of the List. │
│ @put x,0 ─── adds the X items to the start (head) of the List. │
│ @put x,k ─── adds the X items to before of the Kth item. │
│ │
│ @del k ─── deletes the item K. │
│ @del k,m ─── deletes the M items starting with item K. │
└─┐ ┌─┘

call sy 'initializing the list.'  ; call @init
call sy 'building list: Was it a cat I saw'; call @put 'Was it a cat I saw'
call sy 'displaying list size.'  ; say 'list size='@size()
call sy 'forward list'  ; call @show
call sy 'backward list'  ; call @show ,,-1
call sy 'showing 4th item'  ; call @show 4,1
call sy 'showing 5th & 6th items'  ; call @show 5,2
call sy 'adding item before item 4: black' ; call @put 'black',4
call sy 'showing list'  ; call @show
call sy 'adding to tail: there, in the ...'; call @put 'there, in the shadows, stalking its prey (and next meal).'
call sy 'showing list'  ; call @show
call sy 'adding to head: Oy!'  ; call @put 'Oy!',0
call sy 'showing list'  ; call @show
exit /*stick a fork in it, we're done.*/
p: return word(arg(1),1)
sy: say; say left('',30) "───" arg(1) '───'; return
@hasopt: arg o; return pos(o,opt)\==0
@size: return $.#
@init: $.@=''; $.#=0; return 0
@adjust: $.@=space($.@); $.#=words($.@); return 0
@parms: arg opt
if @hasopt('k') then k=min($.#+1,max(1,p(k 1)))
if @hasopt('m') then m=p(m 1)
if @hasopt('d') then dir=p(dir 1)
@show: procedure expose $.; parse arg k,m,dir
if dir==-1 & k=='' then k=$.#
m=p(m $.#);
call @parms 'kmd'
say @get(k,m,dir);
return 0
@get: procedure expose $.; arg k,m,dir,_
call @parms 'kmd'
do j=k for m by dir while j>0 & j<=$.#
_=_ subword($.@,j,1)
end /*j*/
return strip(_)
@put: procedure expose $.; parse arg x,k
k=p(k $.#+1)
call @parms 'k'
$.@=subword($.@,1,max(0,k-1)) x subword($.@,k)
call @adjust
return 0
@del: procedure expose $.; arg k,m
call @parms 'km'
_=subword($.@,k,k-1) subword($.@,k+m)
call @adjust


                               ─── initializing the list. ───

                               ─── building list: Was it a cat I saw ───

                               ─── displaying list size. ───
list size=6

                               ─── forward list ───
Was it a cat I saw

                               ─── backward list ───
saw I cat a it Was

                               ─── showing 4th item ───

                               ─── showing 6th & 6th items ───
I saw

                               ─── adding item before item 4: black ───

                               ─── showing list ───
Was it a black cat I saw

                               ─── adding to tail: there, in the ... ───

                               ─── showing list ───
Was it a black cat I saw there, in the shadows, stalking its prey (and next meal).

                               ─── adding to head: Oy! ───

                               ─── showing list ───
Oy! Was it a black cat I saw there, in the shadows, stalking its prey (and next meal). 

[edit] Ruby

Extending Singly-Linked List (element)#Ruby

class DListNode < ListNode
attr_accessor :prev
# accessors :succ and :value are inherited
def initialize(value, prev=nil, succ=nil)
@value = value
@prev = prev
@prev.succ = self if prev
@succ = succ
@succ.prev = self if succ
def self.from_values(*ary)
ary << (f = ary.pop)! {|i| new i }
ary.inject(f) {|p, c| p.succ = c; c.prev = p; c }
list = DListNode.from_values 1,2,3,4

[edit] Rust

[edit] Simply using the standard library

use std::collections::LinkedList;
fn main() {
// Doubly linked list containing 32-bit integers
let list = LinkedList::<i32>::new();

[edit] The behind-the-scenes implementation

Doubly linked lists present a problem in Rust due to its ownership model. There cannot be two mutable references to the same object, so what are we to do? Below are the relevant lines (with added comments) from the std implementation (Documentation Source).

In order to circumvent the multiple mutable references, raw C-like pointers are used. Note that these cannot be dereferenced with guaranteed safety and thus dereferencing is relegated to unsafe {} blocks.

pub struct LinkedList<T> { // User-facing implementation
length: usize,
list_head: Link<T>,
list_tail: Rawlink<Node<T>>,
type Link<T> = Option<Box<Node<T>>>; // Type definition
struct Rawlink<T> { // Pointer is wrapped in struct so that Option-like methods can be added to it later (wrappers around NULL checks)
p: *mut T, // Raw mutable pointer
struct Node<T> {
next: Link<T>,
prev: Rawlink<Node<T>>,
value: T,

[edit] Sidef

var node =
data => 'say what',
next => foo_node,
prev => bar_node,
node{:next} = quux_node; # mutable

[edit] Tcl

Generally, this task should be accomplished in Tcl using list. Here we take an approach that's more comparable with the other examples on this page.
Works with: Tcl version 8.6
Library: TclOO
oo::class create List {
variable content next prev
constructor {value {list ""}} {
set content $value
set next $list
set prev ""
if {$next ne ""} {
$next previous [self]
method value args {
set content {*}$args
method next args {
set next {*}$args
method previous args {
set prev {*}$args

[edit] Visual Basic .NET

Public Class Node(Of T)
Public Value As T
Public [Next] As Node(Of T)
Public Previous As Node(Of T)
End Class
Personal tools