Partial match retrieval for non-uniform query distributions*

A good file designarid investIgatIon of reason-ably fast and efficient algorithms to perform a required task on the information stored in the file are important in data base studies. In a reasonably large class of problems one can devise and evaluate algorithms with respect to measures independent of the storage and structuring of data. But when the question is the investigation of a suitable data structure for retrieval of a specific kind, algorithms and their measures depend heavily on the proper file design. This paper addresses both file design and algorithm design when searches are to be made in a data base for queries of speciaJ type. We are conce~ed _ with information retrieval based on secondary keys. In general the secondary keys form a subset of the attributes of a record and, thus, they can not uniquely identify a record. We view a secondary key as a k-tuple