Note: Will link a Jira ticket once I get access to issues.openmrs.org.

This is part of a broader effort to modernise the DAO layer ahead of a potential Hibernate 6 migration. Happy to split further or adjust scope based on feedback.

One suggestion: consider adding a brief Javadoc on buildPredicates() explaining the filter logic for future contributors unfamiliar with the original Criteria API behaviour.

Hey @shubhangiisinghh! Great refactor! Migrating away from the legacy Criteria API is definitely the right move. I read through the changes and noticed a subtle behavioral difference regarding how the queue join translates to SQL:

1. Inner Join vs Left Join: In the original code, c.createAlias("queue", "q") was used. In the legacy Hibernate Criteria API, createAlias defaults to an INNER JOIN. In the new code (line 115), you've used root.join("queue", JoinType.LEFT).
   If a QueueEntry can ever have a null queue, this implicitly changes the result set (the old code would filter them out, the new code will include them). Even if queue is mandatory and cannot be null, an INNER JOIN is strictly more performant and faithfully matches the legacy generated SQL.

2. Avoiding Double Joins: Currently, the code uses root.get("queue") on line 113 for the .in() clause, and then explicitly creates root.join("queue", ...) on line 115. Depending on the Hibernate version, mixing implicit path navigation (get) with an explicit join on the same entity can cause JPA to generate two separate joins in the underlying SQL (one cross/inner join, one left join).

Since the old code unconditionally created the alias anyway, a safer and more exact 1:1 translation would be to unconditionally create an INNER JOIN up front, and reuse it for all the queue checks to prevent duplicate joins:

Join<QueueEntry, Queue> queueJoin = root.join("queue", JoinType.INNER);

// Use the explicit join instead of root.get("queue")
limitCollection(predicates, queueJoin, searchCriteria.getQueues());

if (searchCriteria.getLocations() != null || searchCriteria.getServices() != null) {
    limitCollection(predicates, queueJoin.get("location"), searchCriteria.getLocations());
    limitCollection(predicates, queueJoin.get("service"), searchCriteria.getServices());
}

Other than that, the limitCollection mapping for the empty list vs null checks is spot on. Nice work!