The Temporal Multi-Dimensional Join
Author
Jørgensen, Peter Sune
Term
4. term
Education
Publication year
2003
Abstract
Mange databaseopgaver handler om at sammenligne og kombinere data over tid. Temporale operatorer er operationer, der arbejder med tidslige data i forespørgsler. Vi præsenterer temporal multi-dimensional join (TMDJ), en enkel, parametriserbar join-operator (et konfigurerbart byggeklods), der gør det muligt systematisk og effektivt at implementere en bred vifte af avancerede tidslige operatorer. Vi formaliserer tre centrale, men ofte forvekslede semantiske egenskaber: punkt-baseret (ser på enkeltøjeblikke), interval-baseret (ser på tidsintervaller) og duplikatbevidst (behandler gentagne poster som særskilte). Vi viser, at disse semantiske egenskaber kan styres gennem en parametrisering af TMDJ. Til sidst beskriver vi en letvægtsimplementering og rapporterer eksperimenter, der viser, at avancerede tidslige operationer bygget med TMDJ er flere størrelsesordener hurtigere end tilsvarende løsninger i standard SQL.
Many database tasks involve comparing and combining data over time. Temporal operators are operations that handle time-dependent data in queries. We present the temporal multi-dimensional join (TMDJ), a simple, parameterizable join operator (a configurable building block) that enables a systematic and efficient implementation of a wide range of advanced temporal operators. We formalize three key, often confused semantic properties: point-based (looking at individual moments), interval-based (looking at time spans), and duplicate-aware (treating repeated records as distinct). We show that these semantics can be controlled by parameterizing TMDJ. Finally, we describe a lightweight implementation and report experiments showing that advanced temporal operations built with TMDJ run orders of magnitude faster than equivalent solutions in standard SQL.
[This abstract was generated with the help of AI]
Documents