Friday, 13 January 2012

Exception Wrapping for persistence

Data can be stored in various ways, for example:
  • a relational database
  • text files
  • on the web (for example, fetching the weather forecast from a web site)
If the storage method changes, then the low level Exception objects thrown by the data access layer can also change. For example, when the data store is moved from text files to a relational database, IOException is replaced with SQLException. Otherwise there will be compilation errors throughout the code.

Removing the compilation point of view, lets look at the design point. Then consider a typical J2EE application that has a simple web layer, a business layer and a database layer. If some SQL query in the database layer goes completely wrong I will get an SQLException of some sort. In your point of view the web application (being the topmost layer) will catch this exception in the end.

This defeats the entire purpose of having a multitier (3+) architecture. I'm now dependent on the database layer in the web layer, because all exceptions are thrown all the way upwards.

The point of the PersistenceException is that it keeps the tiers independent from each other, except for the tier beneath it. The web layer doesn't care if it's an SQLException. It only needs to know something went wrong with the business logic, and more specifically with the persistence of business data objects.

In any case, the 'business code' knows more of what to do than the layers above it. If it can't do anything meaningful with it, you can either log it or perform some other action. The question is: what would a layer above it (eg. the web layer) know more of what to do than the business layer?

No comments:

Post a Comment