Hi Lucene dev, I’m following up on LUCENE-15769 <https://issues.apache.org/jira/browse/LUCENE-15769> and PR #16402 <https://github.com/apache/lucene/pull/16402>. Lucene 10’s change to PathHierarchyTokenizer position increments breaks its documented ancestor-search use case: query parsers now construct a phrase rather than a synonym/disjunction query.
This surfaced in SOLR-18134 <https://issues.apache.org/jira/browse/SOLR-18134>. A Solr configuration workaround would require users to change deployed configsets and does not solve the underlying regression, so a Solr maintainer asked that we investigate correcting the tokenizer behavior itself. PR #16402 restores the prior synonym positions and includes ancestor/descendant regression coverage, but it is currently stale and needs a rebase. Could maintainers advise whether that is the desired direction, or identify the preferred alternative?
