Surveying the object persistence landscape in Java

Making Data Matrix 2d barcode in Java Surveying the object persistence landscape

Within the realm of object persistence frameworks, you ve got several viable options. Your choice of an individual framework will depend on your application s requirements, your skills, and your preference. You can think of the object

1 Most object persistence frameworks support lazy loading semantics, so while you conceptually load entire objects, the framework might load only a certain set of fields initially and lazy-load larger or less frequently used fields.

persistence landscape along two separate axes. First, the potential data stores define how data is physically stored. More interestingly, the interfaces define how you interact with the data store. Let s look at the major categories along each axis. The data store This axis answers the question, How do you persist your data Object-oriented systems store objects and links between those objects. Relational database management systems store data in related tables.

Object-oriented database management systems (OODBMS) represent data in the database as lists of objects, rather than as tables containing rows and columns. In a famous paper called The Design of a Robust Persistence Layer for Relational Databases, Scott Ambler concluded that object-oriented databases are nice but no match for relational databases. Still, some applications, like CAD systems, are inherently object-oriented. It pays to consider OODBMS for these types of problems. Vendors, such as Poet, Progress, and Versant supply the key OODBMS products in the Java space. Relational DBMS (RDBMS) represent databases as rows in a table. While this approach is not inherently object-oriented, it is ubiquitous. RDBMSs are highly efficient, due to a clean mathematical model and decades of practical experience and academic study. We understand how to maintain, tune, code, and administer them. It s meaningless to discuss alternatives that don t work well with relational databases.

Interfaces to the data store To us, the critical question is this: How does your application access data in the data store Object-oriented frameworks, for the most part, use techniques that strive for as much transparency as possible. The major distinction is proprietary versus open. Proprietary solutions can provide more flexibility; standards are more interoperable. Let s look at a breakdown of the differences between objectoriented and persistence frameworks:

Object-relational mapping frameworks (OR mappers) can convert between an object and the corresponding row-set in a set of tables in a relational database. These frameworks specialize in providing an object-oriented domain model and building the glue code that connects domain models to a database. The better frameworks are efficient. They also have highly flexible mapping technology that can adapt to a wide variety of relational schemas. OR mappers became much muddier when Oracle bought the

industry s leading OR mapper, TopLink. At this point, it s not clear if Oracle will continue to support rival databases or application servers as well as it has in the past, but several good products are more than capable. CocoBase is the main alternative to TopLink in this space, and several open source OR mappers exist. Castor is the most popular open alternative, but Hibernate is also gaining in favor.

Standardized object persistence frameworks Some alternatives attempt to provide a standard mapping to both object-oriented databases and relational databases. Within this group, the alternative with the most momentum is JDO . Most implementations have surprisingly nimble mapping technology, good performance, and perhaps one of the cleanest programming interfaces. Some JDO implementations ship with a database back end, and some focus on relational mapping. The standardization of the JDO specification by the Java Community Process board members provides us with a convenient common ground to use for discussion of object persistence tools. Strong support for JDO exists from OODBMS vendors and from thirdparty OR mapping products, but relational database vendors themselves

Many of the Java object-oriented persistence frameworks are remarkably similar. They all provide fine-grained persistence, with clean, object-oriented interfaces. For convenience, we ll use JDO in this book, but most arguments that we ll make also hold true for other persistence frameworks and mapping technologies.