90 likes | 225 Vues
This exploration highlights the limitations of traditional relational databases, such as the absence of collection types, object identity, and support for complex data types, which can lead to impedance mismatch. It discusses the promise of object-oriented databases in offering better data types and a unified language approach for persistent storage. However, it also addresses the drawbacks, such as memory management issues and the complexity of optimizing SQL-like operations. The rise of object-relational databases aims to combine the strengths of both paradigms while dealing with advanced data types, yet challenges remain in practical implementation.
E N D
Object and object-relational databases Fall 2011
What’s wrong with relational databases? • No collection types • No user-defined types • No object-IDs (no object identity and so no object semantics) • therefore: only value semantics • Impedance mismatch • No subtypes • No large or unstructured attributes
Why object-oriented? • Multiple reasons…. • Better data types • Avoid impedance problems • Object semantics • Single language – a “persistent” language
So, a new approach • A persistent programming language • maybe with selective persistence • A uniform application and database data structuring approach • Advanced and augmentable data types • Able to server more application domains • engineering? • medicine and science and geographic applications • more complex transactions run less often on smaller “tables” • business logic - moves the protection of the database upward • long term decision making?
What is an object? • Page 524 – an (OID, value) • Primitive values • Reference values • Tuple values • Set values • Plus subtyping • Plus method encapsulation • Plus augment object-oriented language with SQL-like capabilities • need path expressions
Problems with o-o databases • Cost of pulling entire objects into memory • Temptation to create large and complex objects • Difficulty in isolating SQL-like operations to optimize them • No upward compatibility • ** Didn’t actually increase applicability of database technology • no sophisticated complaints • no analytical operations • no advanced media like images, video, audio, graphics, and xml documents • no text objects or text searching • no notion of time or versions • no way to separate large volume, simple types from small types that have complex structure
Object-relational databases • Relational venders didn’t want to get blown out like H/N database venders • Objects “from the bottom up” • set valued and array valued domains • user defined types • Subtypes • Still uses two languages • Bimodel data model: • tupes • objects • see page 530
Interesting facts • Extensions are not widely used • User defined types are perhaps the most popular adaption • Newer extensions for xml, image, video, audio, and graphics data are very popular