Basic operators for semistructured data in a relational programming language

JRelix is a relational database implementation that supports not only traditional relational algebra and domain algebra but also complex data type and recursive nesting with powerful database programming language. This thesis documents some new features and operators of JRelix. Among them, type polymorphism (union type domain) and the relational pattern search (grep) operator are especially useful when dealing with semistructured data in a relational database.We use union type domains to increase the flexibility of rigid type definition in relational databases. In addition, we implement the grep operator in the relational algebra to facilitate queries on semistructured data. Grep returns a relation which could contain the type and name of the attribute where it finds the match and the position in that attribute and value of the match. Moreover, we also implement top-level scalar and substring function which are also very useful in a relational database language