Hi, While looking at HSEARCH-1917 which is an issue with faceting and null values, I came up with the following observations: 1/ Faceting doesn't take into account indexNullAs 2/ Faceting doesn't take into account FieldBridges 1/ In HSEARCH-1917, Hardy talked about supporting indexNullAs. Does it make sense for everyone? I don't have a strong opinion about it and if we answer no to 2/ for the time being, maybe it's better to not support it for now (but I would fix the faceting on null values anyway). 2/ FieldBridges are also not supported by faceting. In the case of a TwoWayFieldBridge, we could consider applying the FieldBridge to the values before storing them in the facet field. One thing that made me think about it is that, currently, with the Elasticsearch backend, facet values are created from the indexed field so they take into account the FieldBridge. Well, apart if it's a facet which has a name different from the field. Then the facet values are extracted from the facet values created in the Lucene document. In this case, the behavior is similar to the Lucene backend behavior. My current opinion is that we should for now fix HSEARCH-1917 by adding a if (value == null) return; at the top of addFacetDocValues and revisit the 2 subjects later for 6. Anyone against me creating an issue to centralize all the things we need to fix in FieldBridges/Document building for 6? I would rather create a single issue for now so that we keep an overview of all we need to fix before defining exactly what we want to do. -- Guillaume _______________________________________________ hibernate-dev mailing list hibernate-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hibernate-dev

On Tue, Feb 23, 2016 at 7:21 PM, Sanne Grinovero <sanne(a)hibernate.org&gt; wrote: > > I don't have a strong opinion on the usefulness on faceting on null; > would someone have a sensible example of faceting on a "null value" ? > > We can fix it but I have the impression it's just about fixing a > developer's itch because of it being inconsistent with other > components; if it helps with Elasticsearch integration then that makes > more sense to me. It's not faceting on a null value, it's the ability to support null values in faceted fields when indexing the entity. Currently, it throws a NPE on indexing. For instance, you might have entities which doesn't have a value for a given faceted field. Does it make more sense?

> But why would you fix it by "if (value == null) return;" ? Do you > think you can make it work with that? I'd rather actually index the > marker token defined by 'indexNullAs'. Do you think we would want facet results for the indexNullAs value?

> > +1 to collect notes on FieldBridge and Document change wishes. OK, will do.