Temporal alignment

Abstract

In order to process interval timestamped data, the sequenced semantics has been proposed. This paper presents a relational algebra solution that provides native support for the three properties of the sequenced semantics: snapshot reducibility, extended snapshot reducibility, and change preservation. We introduce two temporal primitives, temporal splitter and temporal aligner, and define rules that use these primitives to reduce the operators of a temporal algebra to their nontemporal counterparts. Our solution supports the three properties of the sequenced semantics through interval adjustment and timestamp propagation. We have implemented the temporal primitives and reduction rules in the kernel of PostgreSQL to get native database support for processing interval timestamped data. The support is comprehensive and includes outer joins, antijoins, and aggregations with predicates and functions over the time intervals of argument relations. The implementation and empirical evaluation confirms effectiveness and scalability of our solution that leverages existing database query optimization techniques.

Querying and Learning in Probabilistic Databases

“…To resolve this issue, we enforce the time intervals of tuples with identical, non-temporal arguments to be disjoint. A relation instance that adheres to this condition is going to be termed duplicate-free [21].…”

A temporal-probabilistic database model for information extraction

“…Jensen's seminal work [18] provides the basis for most temporal database approaches today. Recently, Dignös et al [12] described how to enable the usage of sequenced semantics in temporal databases using lineage information. Intuitively, the sequenced semantics reduces a temporal operation to corresponding non-temporal operations over the individual snapshots of the database at each point in time.…”

“…1 state the marriage of DeNiro and Abbott, but with different lineages and overlapping time intervals. Following prior work on temporal DBs with data lineage [12], we next define duplicate-free temporal relations. We refer to the arguments of a fact f in a relation instance te) , where R T is the relation name, andā and t b , te are constants denoting the non-temporal and temporal arguments of f , respectively.…”

“…T is called a duplicate-free relation [12], if for all pairs of facts In other words, facts with equivalent non-temporal arguments must have non-overlapping time intervals. Deduplicating Facts.…”

“…We however refrain from full logical equivalence checks between lineage formulas (which are N P-hard). This definition is similar to change preservation in [12], which also iteratively compares sets of lineages along a query plan. Example 3.…”