Laura C. Rivero

Elicitation and conversion of hidden objects and restrictions in a database schema

Mapping a database schema from one model into another, with a higher semantic capacity, is a current research subject with application in several development fields, such as schema integration and translation, migration from legacy systems and reengineering of poor quality or no-longer accurate data models. Inclusion dependencies are one of the most important concepts in relational databases and they are the key to perform some reengineering of database schemas. Referential integrity restrictions (rir), a particular case of an inclusion constraint, requires that the set of distinct values occurring in some specified column, simple or composite (foreign key), must be a subset of the set of distinct primary key values drawn from the same domain. Pure inclusion dependencies (id), however, may apply between other pairs of attributes also (alternate keys or non-keys). Database schemas containing ids frequently reveal the presence of hidden objects and misrepresented relationships and, as a consequence, increase the effort to develop program applications and maintain the integrity. This work presents a heuristics for the conversion of schemas with ids into equivalent schemas with only rirs. In case some irreducible ids remain, a semantic interpretation of their necessity and maintenance is given.

Static detection of sources of dynamic anomalies in a network of referential integrity restrictions

Under certain circumstances, basic operations over tables in a relational database, where integrity reslxictions such as referential and null restrictions have been specified, may produce unpredictable results, not detectable by means of a static analysis of the schema. When the design includes redundancies or when the set of restrictions is contradictory it is easy to detect and prevent future exrors, but there are situations that require a dynamic analysis. In this paper, the properties of networks of referencial integrity reslfictions that contain irregularities are analyzed, and the mtomalies that may appear when data act~mlization in such environment is done are studied in order to define criteria and develop an algorithm to generate rules for proper handling of inconsistencies.