I learnt something interesting today: The skip list.

Skip list is a data structure that happens to do similar work as a tree. However, to make a tree optimized for searching, it has to be balanced and balancing is a cost. Skip list, however, does not enforce balancing but make the “tree” probabilistically balanced.

Actually, it is not a tree, but a list. The nodes in the list has random number of connections to decendent nodes. For example, node 1 may has three connections and it is pointing to node 2, node 10 and node 15, node 2 has only one connection and it is pointing to node 3. In general, node \(n\) must point to node \(n+1\) but there may be some other connections to node \(n+k\).

The more levels of connections you have, the faster you can search. Details are mentioned in the following paper:

The sample code is also available from the author: ftp://ftp.cs.umd.edu/pub/skipLists/

Bibliographic data

   title = "Skip lists: A probabilistic alternative to balanced trees",
   author = "William Pugh",
   journal = "Communications of the ACM",
   volume = "33",
   number = "6",
   pages = "668--676",
   year = "2000",
   url = "ftp://ftp.cs.umd.edu/pub/skipLists/skiplists.pdf",