Analysing Temporal Relations – Beyond Windows, Frames and Predicates

This article proposes an approach to rely on the standard operators of relational algebra (including grouping and aggregation) for processing complex event without requiring window specifications. In this way the approach can process complex event queries of the kind encountered in applications such as emergency management in metro networks. This article presents Temporal Stream Algebra (TSA) which combines the operators of relational algebra with an analysis of temporal relations at compile time. This analysis determines which relational algebra queries can be evaluated against data streams, i. e. the analysis is able to distinguish valid from invalid stream queries. Furthermore the analysis derives functions similar to the pass, propagation and keep invariants in Tucker’s et al. “Exploiting Punctuation Semantics in Continuous Data Streams”. These functions enable the incremental evaluation of TSA queries, the propagation of punctuations, and garbage collection. The evaluation of TSA queries combines bulk-wise and out-of-order processing which makes it tolerant to workload bursts as they typically occur in emergency management. The approach has been conceived for efficiently processing complex event queries on top of a relational database system. It has been deployed and tested on MonetDB.