Loop Elimination for Database Updates

Abstract

The additional expressive power of procedural extensions of query and update languages come at the expense of trading the efficient set-at-a-time processing of database engines for the much less efficient tuple-at-a-time processing of a procedural language. In this work, we consider the problem of rewriting for-loops with update statements into sequences of updates which do not use loops or cursors and which simultaneously carry out the action of several loop iterations in a set-at-a-time manner. We identify idempotence as the crucial condition for allowing such a rewriting. We formulate concrete rewrite rules for single updates in a loop and extend them to sequences of updates in a loop.