Datastream-A Language for Large Files

Abstract

Datastream is a simple language designed for writing programs that perform computations on large data files in the UNIX™ operating system. Datastream handles larger files and a wider variety of computations than most UNIX database management systems, but it provides no explicit support for updates. Its features and performance characteristics are particularly good for working with infrequently updated, mostly statistical data. This paper describes the Datastream language, outlines its evolution, and summarizes users' experience with a prototype implementation.