Abstract
Updating a database by means of a view requires the changes made on the viewto be properly propagated to the underlying database. This task of “translating”a view update into a suitable database update is non-trivial and it poses severalchallenges indeed. The most subtle kind of update anomaly is given by changesnot directly wanted nor explicitly made by the user, originating in the view as a“side-effect”. An additional difficulty is represented by the fact that there can bemore database updates corresponding to a given update on the view. Lastly, yetanother complication concerns updates that modify the database even thoughthis is not strictly necessary in order to reflect the changes.A general and precise understanding of the view update problem is due to theseminal work [1] by Bancilhon and Spyratos, who devise an abstract frameworkin which they formalise the problem and provide an elegant solution to it. Theyintroduce the fundamental notion of view complement, representing what is miss-ing from a view in order to have the same informative content of the underlyingdatabase. Moreover, they introduce the constant complement principle, statingthat the changes done on a view must not influence the content of its complementduring the translation process. Bancilhon and Spyratos provide no constructivecharacterisation of their approach, stating that “computational algorithms (ifthey exist) must be sought in specific problems: for example, schemata definedby functional dependencies and views derived by projections”. Most of the sub-sequent work on view updates is centred on the framework [1] and, in particular,its application to the relational model.Cosmadakis and Papadimitriou [4] consider a restricted setting that consistsof a single database relation, in which views are just projections over the attrib-utes of such a “universal” relation. They give necessary and sufficient conditionsfor the translatability of insertions, deletions and replacements under constantcomplement and they also study the complexity of finding a suitable complementthat makes a given update translatable. Even though an effective method basedon the Chase is provided for checking translatability, it is only applicable in thevery limited setting consisting of a single database relation with projective viewsand functional and join dependencies at the database level.In the context of SQL databases, Lechtenbörger [6] gives a characterisationof the constant complement principle in terms of undo operations, showing thatview updates are translatable under constant complement precisely if users havethe chance to undo all effects of their updates using further view updates. It isthen argued that testing whether this holds could be an alternative to checkingwhether users can observe all effects of their updates in the view schema, but noeffective method for doing so is proposed
Original language | English |
---|---|
Title of host publication | Nineteenth Italian Symposium on Advanced Database Systems (SEBD 2011) |
Pages | 147-154 |
Number of pages | 8 |
Publication status | Published - 2011 |