Spring Data Couchbase: projections with reactive repository - spring-data

Question

Reactive Couchbase repository with a method using DTO projections fails with
CouchbaseQueryExecutionException: Query returning primitive got more values than expected: 5
Is DTO projection supported by reactive repositories?
I'm trying to implement a very basic DTO projection repository method with a reactive repository: I have two Couchbase #Documents, and an interface representing a projection. A query repository method projects a few attributes from the two documents using interface projection.
After some debugging I see that the failure occurs because of this check in org.springframework.data.repository.query.QueryMethod:
public boolean isQueryForEntity() {
return getDomainClass().isAssignableFrom(getReturnedObjectType());
}
called from org.springframework.data.couchbase.repository.query.ReactiveAbstractN1qlBasedQueryexecuteDependingOnType().
Apparently the document entity is not assignable from the projection interface which is expected.
Here're the prototype classes in question:
First document:
#Document
#Data
public class DocMetadata {
#Field
private String type;
#Id
#Field
private String id;
#Field
private String title;
}
Second document:
#Document
#Data
public class Doc {
#Field
private String type;
#Id
#Field
private String id;
#Field
private String metadataId;
#Field
private Status status;
#Field
private String releaseDate;
}
Projection interface:
public interface DashboardDocProjection {
String getId(); // from Doc.id
String getReleaseDate(); // from Doc.releaseDate
String getStatus(); // from Doc.status
String getTitle(); // from DocMetadata.title
}
And the repository:
#Repository
public interface CouchbaseDocRepository extends
ReactiveCouchbaseSortingRepository<Doc, String> {
#Query("<query to select Doc join DocMetadata projecting some attributes from both>")
Flux<DashboardDocProjection> getProjection();
}
As a workaround, I can make projections work with a separate #Document representing projected attributes and a separate repository for that document type, but it requires more boilerplate code and is not so nice in general. An example mapping looks like this:
#Query("SELECT META(`doc1`).id AS _ID, META(`doc1`).cas AS _CAS, "
+ “doc2.title as title, "
+ "`doc1`.id as versionId, "
+ "`doc1`.attribute2 as attribute2 "
+ "FROM #{#n1ql.bucket} `doc1` "
+ "JOIN #{#n1ql.bucket} doc2 ON KEYS […] "
+ "WHERE `doc1`._class = "\"<doc1.class>\" "
+ "AND `doc1`.attribute1 = $1 "
+ "AND `doc1`.attribute2 = $2 "
+ "ORDER BY `doc1`.id ASC "
+ "LIMIT $3 OFFSET $4;")
Flux<DashboardEntry> findAll(
String attribute1Value, String attribute2Value, Integer limit, Integer offset);