Github user dsmiley commented on a diff in the pull request: https://github.com/apache/lucene-solr/pull/416#discussion_r208799313 --- Diff: solr/core/src/test/org/apache/solr/response/transform/TestChildDocTransformer.java --- @@ -242,10 +242,10 @@ private void testChildDocNonStoredDVFields() throws Exception { "fl", "*,[child parentFilter=\"subject:parentDocument\"]"), test1); assertJQ(req("q", "*:*", "fq", "subject:\"parentDocument\" ", - "fl", "subject,[child parentFilter=\"subject:parentDocument\" childFilter=\"title:foo\"]"), test2); + "fl", "id,_childDocuments_,subject,intDvoDefault,[child parentFilter=\"subject:parentDocument\" childFilter=\"title:foo\"]"), test2); --- End diff -- ah; interesting. It's logical. Is this only needed for anonymous child docs (thus \_childDocuments\_ or any/all possible relationship names that aren't necessarily just at the root level but anywhere in the hierarchy? Perhaps this is where that "anonChildDocs" ought to come into play again for backwards-compatibility sake? Well perhaps not... someone who is using anonymous child docs today will not have the nested field metadata and thus the old logic will kick in and ensure child documents are added as it was; right?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org