A group-select operation has been defined for relational algebra. This operation is found to be useful for efficiently reducing expressions of nonprocedural relational languages that permit natural quantifiers. Conceptually, the operation first partitions a relation into blocks of tuples that have the same value for an attribute or attribute concatenation. It then extracts each block for which a specified number of tuples meet a specified condition. The quantity of tuples for the operation is specified by means of a natural quantifier. Performance of the group-select operation will be poor with conventional file processing, making the operation more suitable for use with a database machine with an associative memory.>

A data model is presented, based on the extension of the concept of a DBTG owner-coupled set to permit static and dynamic sets and a new kind of set referred to as a virtual set. The notion of connection fields is introduced, and it is shown how connection fields may be used to construct derived information bearing set names, and hence permit the specification of (dynamic) sets which are not predeclared in a schema. Virtual sets are shown to reflect the functional dependencies which can exist within a file. A technique which permits the data model to be fully described diagrammatically by extended Bachman diagrams is described. A predicate calculus for manipulation of this data model is presented. Expressions written in this calculus are compared with corresponding expressions in a relational predicate calculus, DSL ALPHA. An argument for the relational completeness of the language is given.