Abstract

This talk addresses event correlation in the form of n-way joins over asynchronous event streams. Diverse domains exploit different variants of event joins, e.g., stream-relational algebra, event processing, reactive and concurrent programming. We aim to uniformly model semantic variants of joins as composable, modular and open-ended libraries, subsuming different domains and enabling novel feature compositions. We specify a syntactic theory of joins in terms of reductions over their intensional specification embedded in a lambda calculus. The key mechanism to define semantic variants are algebraic effects and handlers. Thus, the meaning of a join is a matter of context, which is a first-class notion. The end result is a framework for declaring the combination behavior of nodes in reactive computations. Our approach covers instances of joins that range from nondeterminism/speculation in the complex event processing domain to non-speculative variants, such as zip in reactive programming, and beyond. The talk will give an account of our formal model and experiences with using algebraic effects.