String operations for a data sublanguage

The starting point of this paper is the relational model of data as explained in [i]. We use the terminology table for relation, column for attribute, and row for tuple ([I] page 53). One of the fundamental properties of the relational model is that all the entries in the tables are nondecomposable (atomic) elements ([i] page 76). Such a database is said to be normalized. The usual examples of nonatomic entries in a table are arrays and structures. But if a table has such nonatomic entries it can be transformed into one with atomic entries only ([I] page 77). Normalization simplifies the data structures and operations of the data sublanguage and allows the use of the mathematical theory of relations to give a solid foundation to database theory. It should be noted that character string operations are not included in the usual data sublanguages for databases ([i] Chapters 5-8). Indeed if an entry, like a character string, is nondecomposable then we certainly cannot decompose it by taking a substring of the entry. Yet this means that some simple queries cannot be expressed in the language. Assume first that one column in a table is NAME. Then we may wish to have a query such as: Get all rows where NAME begins with the letter tG1 .