HISTORICAL SEARCHING

Abstract

When searching for an item x in a dictionary, let t be the number of distinct items referenced since the previous reference to x. The move-to-front-list is a widely known dictionary data structure that supports searches on x in O(t) time. We present a new selforganizing data structure, called the Historical Search Tree, which supports the search in O( log t) time, and thus exploits the temporal locality more efficiently. The Historical Search Tree also handles unsuccessful searches, insertions, and deletions efficiently. We further show how the data structure can be modified to take advantage of spatial locality.