An Alternative Definition of Database Relations
Abstract
Alternative definitions of tables in relational databases are given.
The definitions require one to explicitly designate the intended entities
and relationships in the key of the database tables.
They also require one to explicitly distinguish between the attribute
columns of the table and the entity or relationship columns.
The definitions allow the resolution of the 1)insertion/deletion anomalies
and 2)spurious tuples.
Neither the anomalies nor the tuples arise if one explicitly specifies
the entities, attributes and relationships that the database is meant
to depict.
The anomalies and the tuples are a result of the definition of
the database and tables and the key as uniquely identifying the
entities depicted by a line of the table (Codd).
The original definition of the tables as depicting relations makes it
somewhat cumbersome or forced to adequately depict unrelated entities
(they can be depicted as relations of degree one).
Also, the lack of a means to distinguish entities and relationships
from their attributes makes the interpretation of not only singleton
sets but also relationship sets inadequate.
One will not obtain the correct degree for the relationship if the table
depicts not only the related elements, but an attribute of the relationship
itself. Codd's definition applies only, as it is said, to purely relationship
tables. The database, however, includes tables other than pure
relationship tables.