[ https://issues.jboss.org/browse/ISPN-3795?page=com.atlassian.jira.plugin.... ] Sanne Grinovero commented on ISPN-3795: --------------------------------------- {quote}+1 IGNORE_RETURN_VALUE should not be honoured if prevents the indexing to work.{quote} I agree on the high level but there are some alternatives we could consider. As far as I remember we need the previous value just because we need to know which object type is being replaced. Since we allow each "type" to be mapped to its own index, this is necessary to know which index needs to be purged of the previous value. We have loads of (other) problems all caused by this commonality of allowing different types to be stored in the same cache; if we could implement the restriction of strongly type-safe Caches we would resolve the other problems and also no longer need this one. Incidentally polymorphic hierarchies in Hibernate Search get indexed in the same index, so the problem is really only triggered by the user doing something like storing _unrelated_ types: we would be fine with parent/child relations. While Hibernate Search indexed parent/child objects in the index _by default_ we would also benefit from enforcing this in Infinispan. Having strong type-safety would allow us to remove a lot more overhead as many checks currently need to be performed on a per-operation base, and the current state also requires much metadata to be computed dynamically. -- This message was sent by Atlassian JIRA (v7.2.3#72005)