Looking At Linked Lists

A linked list is a data structure that includes a 'linked' sequence of nodes. Each node contains data.

In a singly linked list each node points to the next node in the linked list:

In a doubly-linked list each node points to the next node as well as the previous node:

In order to access a single node you must traverse the entire linked list until you land on that node.

For example, to find node 4 we must start at node 1, move to node 2, then node 3, and finally to node 4:

This is very different from an array where each value can be found by using it's index, or a dictionary (hash/object) where a value can be found using it's key.

One benefit of a linked list is that you can add a node to the beginning of a linked list using constant time (O(1)). This means that subsequent nodes do not need to be re-indexed as they do in an array. The newest node simply becomes the "head" of the linked list.

Have you ever used linked lists in your applications?

If so why did you choose them over other data structures?

Originally published at http://deniseonaquest.com/looking_at_linked_lists

Comments

[Denise Pen]

Which data structures are you referring to when you say "it underpins a number of standard data structures used today". I'm assuming stacks and queues. Is there anything else?

Also, thanks for responding. I'm learning so much from reading responses to all the blog posts.

[Denys Séguret]

There are three things to know regarding linked list based structures (including double linked lists and more complex structures):

• they're very rarely the most efficient structure, even if their theoretical big O cost is pretty. They consume a lot of memory and this memory is fragmented.

• they almost always imply tricky corner cases which means a big test list and some unexpected bugs after months or years of production (true story, and that was billions of computations during 15 years until a case that I finally understood)

• sometimes they're useful... probably not for your problem, and you should have checked you didn't in fact need a ring buffer or a vector, but yes there are still some use cases...