Hi,

Recently I've done huge migration from hibernate-search-3.0.0.GA to hibernate-search-orm-5.6.1.Final. In our project we have a lot of @Indexed entities and quite sophisticated hierarchy. There's abstract @Indexed @MappedSuperclass which has @Indexed subclasses. Simplifying thing it could be written like following:

    @MappedSuperclass
    @Indexed
    public abstract static class A {
        @Id
        @GeneratedValue
        public long id;

        @Field(store = Store.YES)
        @NumericField
        public int parentRandom = new Random().nextInt();
    }

    @Entity
    @Indexed
    public static class B extends A {
        @Field(store = Store.YES)
        @NumericField
        public int random = new Random().nextInt();
    }

I found it very interesting, that after switch all the entities started to be populated in the index for class (we're using FSDirectoryProvider as a provider) A, even though I'm trying to index B instances. Quick analysis shows, that during creation of IndexManagerGroupHolder there's wrong name passed as first argument for org.hibernate.search.indexes.impl.IndexManagerHolder#getOrCreateGroupHolder. Following that further one could see, that in org.hibernate.search.indexes.impl.IndexManagerHolder#getIndexName there appears to be index name evaluation:

//get the most specialized (ie subclass > superclass) non default index name
		//if none extract the name from the most generic (superclass > subclass) @Indexed class in the hierarchy
		//FIXME I'm inclined to get rid of the default value
		Class<?> aClass = cfg.getClassMapping( clazz.getName() );
		XClass rootIndex = null;
		do {
			XClass currentClazz = reflectionManager.toXClass( aClass );
			Indexed indexAnn = currentClazz.getAnnotation( Indexed.class );
			if ( indexAnn != null ) {
				if ( indexAnn.index().length() != 0 ) {
					return indexAnn.index();
				}
				else {
					rootIndex = currentClazz;
				}
			}
			aClass = aClass.getSuperclass();
		}
		while ( aClass != null );

which then followed by return statement:

//there is nobody out there with a non default @Indexed.index
		if ( rootIndex != null ) {
			return rootIndex.getName();
		}

If I do get comment before the loop right, this method should return the most specific index name either from @Index.name or fully-qualified class name of the @Indexed entity.

Question is is my understanding correct? If so that in my case, when indexing subclass B I should get as an index name fully qualified name of class B. Problem is, that I get fqcn A instead.

Problem appears to be serious, when one's using exclusive_index_use=true, which is default now.