Database-independent queries to the extreme in Java... or in general

Let's say I have an app that should ideally be able to use a relational database, object database, XML files, or whatever to persist its data. In the spirit of coding to interfaces instead of implementations, I have a generic DataStore interface that specifies a contract for all I/O involving the data store. This interface can be implemented by concrete classes such as RDBMSDataStore, OODBMSDataStore, XMLFileDataStore, and so on.

This works well as long as I keep the contents of the DataStore interface simple - i.e. getThis(), getThose(), saveThat(), updateThis(), etc. But as soon as I require more complicated queries, it breaks down. The XMLFileDataStore class obviously doesn't understand SQL, and the RDBMSDataStore class obviously doesn't understand XPath/XQuery. And OODBMSDataStore understands something entirely different depending on the OODBMS in use.

I could adopt a language-independent object query language, write all my queries in that and then have the concrete classes translate them into their native language, but that's a huge task, if I want to be complete.

Are there standards开发者_JS百科 or best practices for handling this kind of situation in Java? Unfortunately it seems like 99% of the world interprets "database independence" to mean "relational database independence" and ignores the object databases, XML databases, document databases, etc. entirely.

From the way I read the question, this sounds a lot like the semantic that Hibernate brings to the table for Java. It even has mode for dealing with XML as the content backing store (using Dom4J). The Hibernate API has a number of extension points that could allow the addition of an OODBMS model. Even if Hibernate turns out not to be the best solution for you (implementation-wise), I think it provides a good example of the types of patterns that can be used to solve the problems you proposed.