Hi,

I am trying to find out how to search by properties of a specific nested object. My model looks something like this: X is a top level entity that contains a collection of Y, and Y has two properties: id and name. I want to find all X instances that contain a Y with specific id and name. Searching by X.Y.id and X.Y.name is not enough because it won't guarantee that id and name matched properties on the same Y instance.
Is it possible to do this?

Thanks,

You are correct, you won't have that guarantee.


Not really. You have to bear in mind that Lucene is not relational at all: we can de-normalize some kind of relations (mostly ManyToOne and OneToOne), but your case is the typical example of what is not supported by the engine without some tricks from your part.

One option would be to encode id+name in the same field. This is easy if "name" is just a keyword for which you don't want to apply an analyzer (you can't let an Analyzer blindly process your tricked value or it will split it and it won't work). If you need to analyze the name, you'll have to pre-tokenize it and then store in the field as id+token1, id+token2, ..., id+tokenN. Remember: you can index the same field multiple times.

To query such an encoded field, would need some similar token manipulation: not trivial but doable.

The good news is that Lucene is coming with a new (limited) JOIN functionality; what you need to do might become simpler in the future but the upcoming feature is going to have a significant performance hit: if you can apply the above trick that should be best.

Thanks for the answer. I came up with a solution that suites my needs: I wrote a FieldBridge for the collection, which iterates over the elements and uses each element ID (which is a Hibernate-generated UUID, so it's safe to use it) as part of the Lucene field name. Then I can use this field name in my queries with ignoreFieldBridge().

right, that looks like a good solution