Subclass design problems/best practices

Question

Hello gurus -
I have a question problem regarding the domain objects I'm sticking in my cache. I have a Product object - and would like to create a few subclasses - say BookProduct and MovieProduct (along with the standard Product objects). These really need to be contained in the same cache. The issue/concern here is that both subclasses have attributes that I'd like to index AND query on.
When I try to create an index on the subclasses attributes when there are just "standard" products - I get the following error (which only exists on one of the subclasses):
2010-10-20 11:08:43.280/227.055 Oracle Coherence GE 3.5.2/463 <Error> (thread=DistributedCache:MyCache, member=2): Exception occured during index rebuild: java.lang.RuntimeException: Missing or inaccessible method: com.test.domain.product.Product.getAuthors()
So I'm not sure the indexing is working or stopping once it hits this exception.
Furthermore, I get a similar error when attempting to Filter based on that attribute. So if I want to add the following filter:
Filter filter = new ContainsAnyFilter( "getAuthors", authors );
I will receive the following exception:
Caused by: Portable(java.lang.RuntimeException): Missing or inaccessible method: com.test.domain.product.Product.getAuthors()
What is considered the best practices for this assuming these really should be part of the same names cache? Should I attempt to subclass the extractors to "inspect" the Object for its class type during indexing or applying filters? Or should I just add all the attribute in the BookProduct and MovieProduct into the parent object and just forget about subclassing? That seems to have a pretty high "yuck" factor to me. I'm assuming people have run into this issue before and am looking for some best practices or perhaps something that deals with this that I'm missing. We're currently using Coherence 3.5.2. Not sure if it matters, but we are using the POF format for serialization.
Thanks!
Chris

Answer 1

Hi Chris,
I had a similar problem. The way I solved it was to use a subclass of the ChainedExtractor that catches all RuntimeException occurring during the extraction like the following:
* {@link ChainedExtractor} that catches any exceptions during extraction and returns null instead.
* Use this for cases where you're not certain that an object contains that necessary methods to be extracted.
* F.e. an object in the cache does not contain the getSomeProperty() method. However other objects do.
* When these are put together in the same cache we might want to use a {@link ChainedExtractor} like the following:
* new ChainedExtractor("getSomeProperty.getSomeNestedProperty"). However this will result in a RuntimeException for those entries that
* don't contain the object with the someProperty. Using the this class instead won't result in the exception.
public class SafeChainedExtractor extends ChainedExtractor
     public SafeChainedExtractor()
          super();
     public SafeChainedExtractor( String sMethod )
          super( sMethod );
     @Override
     public Object extract( Object entry )
          try
               return super.extract( entry );
          catch(RuntimeException e)
               return null;
     @Override
     public Object extractFromEntry( Entry entry )
          try
               return super.extractFromEntry( entry );
          catch(RuntimeException e)
               return null;
}For all indexes and filters we then use extractors that subclassed the SafeChainedExtractor like the following:
public class NestedPropertyExtractor extends SafeChainedExtractor
     private static final long serialVersionUID = 1L;
     public NestedPropertyExtractor()
          super("getSomeProperty.getSomeNestedProperty");
//adding an index:
myCache.addIndex( new NestedPropertyExtractor(), false, null );
//using a filter:
myCache.keySet(new EqualsFilter(new NestedPropertyExtractor(), "myNestedProperty"));This way, the extractor will just return null when a property doesn't exist on the target class.
Regards
Jan