A scan-driven sort facility for a relational database system

  • Authors:
  • Theo Härder

  • Affiliations:
  • -

  • Venue:
  • VLDB '77 Proceedings of the third international conference on Very large data bases - Volume 3
  • Year:
  • 1977

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper the design and implementation aspects of an integrated sort package are described which was implemented by the author as a part of an experimental relational database system called System R. The sort facility developed for variable length sort keys and tuples is discussed with regard to its anticipated use in relational database systems. It is considered to be a powerful tool to implement complex relational operations like join and projection and to support the dynamic creation of access paths and the loading and reorganizing of data clustered by field values. The input loop of the sort can be driven by various scans which allow the filtering of qualified tuples according to disjunctive normal forms of simple search arguments. Various aspects of the integration with storage and transaction management, locking and logging components are described. A number of design decisions for the sort facility concerning sorting, peerage, and merging techniques are discussed. Finally, some implementation aspects for sorting variable length keys consisting of a number of distributed fields are pointed out.