Early = Earliest? - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Rapport (Rapport De Recherche) Année : 2013

Early = Earliest?

Résumé

Early query answering is the core issue of memory efficient query evaluation on data streams. The idea is to select and reject answer candidates as early as possible on the stream, so that they do not have to be stored in main memory. Since earliest query answering is unfeasible for XPath, as first no- ticed by Benedikt, Jeffrey and Ley-Wild in 2008, most exist- ing streaming algorithms for XPath approximate it in some early manner, while focussing on high time efficiency. Such approximations, however, spoil all theoretical guarantees on memory efficiency. In this paper, we prove that earliest query answering is indeed feasible for positive Forward XPath queries, which have neither unsatisfiable nor valid subqueries. The core in- sight is that a variant of Colmerauer's independence property can be proven for the corresponding fragment of the FXP tree logic. Based on this independence property, we can show that the early query answering algorithm from [13], which is based on a compiler from FXP to early nested word automata, is indeed earliest for all positive FXP0 queries with neither unsatisfiable nor valid subformulas. Further- more, this algorithm outperforms most previous algorithms for XPath evaluation on XML streams in time efficiency and coverage, as shown elsewhere. Available here.

Fichier non déposé

Dates et versions

hal-00873742 , version 1 (16-10-2013)

Identifiants

  • HAL Id : hal-00873742 , version 1

Citer

Anthony Lick, Joachim Niehren. Early = Earliest?. [Research Report] inria lille. 2013. ⟨hal-00873742⟩
150 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More