SQL/XNF-Processing composite objects as abstractions over relational data

Abstract

Complex applications, such as design appl_ications, multi-media applications, and even advanced buszness applications can benefit significantly from a database language that supports composite (or complex) Motivation and IntroductionExisting "second generation" data base systems are focussed on business applications (e.g. booking, storekeeping, cost accounting, project management and decision support). It is widely agreed now that second generation DBMSs are inadequate for a broader class of applications that deal with composite objec~ (s~ct~ed data), as opposed t~ flat rel~ tions. Such apphcauons mclude office automauon, medicine, computer aided software engineering (C_AS~). artificial intelligence (AI), hypertext, and geographical mformation systems and computer aided design (CAD).Many of these applications extensively navigate through the data. Some applications, such as CAD, process the data by complex and time consuming algorithms, and demand very high performance. For this, the data needs to be represented by data structures which ~an. be a~ccss~ very f~st. Caching of data close to the apphcauons 1s parucularly Im- portant when applications/tools. run on autonomous. w~rk stations, with remote access to mtegrated data repositones. For example, design applications deal with large af!lount of data, which is mostly organized using such !llodeling concepts as version, alternative, and co~figurauon. Those applications often work on a well-specified set of data, called working set, s~h as a particul~ version of a documen_t or a wing of an arrcraft for a particular model and vers10n. Such data is typically an aggregation of a set of othe~ documents/designs with specified versions. Usu~ly wor~mg sets are extracted from the database and loaded mto mam memory close to the applications for high performa~ce. After an application completes its work on the wo~k_ing. set, the DBMS propagates back the changes to the ongmatmg databases. Working sets are typically much smaller than the whole database. For example, in design applications, the sizes of the databases are in the gigabytes to terabytes range, whereas working sets are typically. in the range of 1 ~o 100 megabytes. Thus, loading a workmg set translates mto a data extraction where on average one tuple out of 10000 to 100000 is selected. This again calls for set-oriented query facilities for efficient data extraction, requiring powerful optimization, and high performance execution. !fie ~oncept of views is extensively used in DBMSs, allo~mg_ different applications to view the data differently. Apphcauons dealing with composite objects require this concept to be extended to allow the formation of different views over the same data. Further,these views must be updatable.We propose an extension of SQL, called SQL Exten~ed Normal Form (or XNF) to satisfy the needs of the applications discussed above. XNF is based on the concept of Composite Objects (<;0) as a coll~tion of tabl~s and rela~on ships· COs prov1de an abstractiOn mechamsm that umfies both the struct...

Incremental maintenance of aggregate and outerjoin expressions

“…Further, outerjoins have recently gained importance because data from multiple distributed databases can be integrated by means of outerjoin views [15,17,18]. Outerjoins are also extensively used in object-relational systems [19][20][21]. if the change table was either generated at an outerjoin operator or is a result of propagation of an outerjoin-change table, using the propagation equations we will derive for propagating outerjoin-change tables.…”

Navigation-Driven Evaluation of Virtual Mediated Views

“…Our navigation-driven architecture extends the virtual view mediator approach as used, e.g., in TSIMMIS, YAT, Garlic, and Hermes [15,4], [5,10]. The idea of on-demand evaluation is related to pipelined plan execution in relational [8] and object-relational [12] databases where child operators proceed just as much as to be able to satisfy their parent operator's requests. However, in the case of (XML) trees the client may proceed from multiple nodes whose descendants or siblings have not been visited yet.…”

Composite-object views in relational DBMS: An implementation perspective

“…Although the XNF language allows to derive non-updatable views, a careful view definition might circumvent some updatablility problems. More information on that can be found in [37] .…”

“…Given the prevalence of SQL as a database interface language both for application and database interoperability, we have proposed an extension to SQL to handle COs. In our approach, called SQL Extended Normal Form (for short XNF) and introduced in [37], we derive COs from relational data. That paper covers the language part giving details on syntax and semantics.…”

“…Therefore the model is closed under its language operations. More information on the XNF language, the multi-lingual API, and (update) semantics can be found in [37].…”