Worked on the apache/lucene repository to implement deferred term collection for the MultiTermQuery constant-score wrapper, targeting scenarios where the number of terms is unknown. This feature, developed in Java and leveraging Lucene’s back end architecture, introduced a mechanism to delay term dictionary scans for automaton-based queries such as wildcard, regexp, prefix, and range. By enabling a cheap planning phase in ScorerSupplier and supporting early short-circuiting in conjunctions, the approach reduced unnecessary computation and improved query planning efficiency. The solution applied lazy evaluation for large term dictionaries, optimizing performance when leading wildcard queries matched a small number of documents.