Scrolling

Scrolling is a more fine-grained approach to iterate through larger results set chunks. Scrolling consists of a stable sort, a scroll type (Offset- or Keyset-based scrolling) and result limiting. You can define simple sorting expressions by using property names and define static result limiting using the Top or First keyword through query derivation. You can concatenate expressions to collect multiple criteria into one expression.

Scroll queries return a Window<T> that allows obtaining the scroll position to resume to obtain the next Window<T> until your application has consumed the entire query result. Similar to consuming a Java Iterator<List<…>> by obtaining the next batch of results, query result scrolling lets you access the a ScrollPosition through Window.positionAt(…).

Window<User> users = repository.findFirst10ByLastnameOrderByFirstname("Doe", ScrollPosition.offset());
do {

  for (User u : users) {
    // consume the user
  }

  // obtain the next Scroll
  users = repository.findFirst10ByLastnameOrderByFirstname("Doe", users.positionAt(users.size() - 1));
} while (!users.isEmpty() && users.hasNext());

WindowIterator provides a utility to simplify scrolling across Windows by removing the need to check for the presence of a next Window and applying the ScrollPosition.

WindowIterator<User> users = WindowIterator.of(position -> repository.findFirst10ByLastnameOrderByFirstname("Doe", position))
  .startingAt(OffsetScrollPosition.initial());

while (users.hasNext()) {
  User u = users.next();
  // consume the user
}

Scrolling using Offset

Offset scrolling uses similar to pagination, an Offset counter to skip a number of results and let the data source only return results beginning at the given Offset. This simple mechanism avoids large results being sent to the client application. However, most databases require materializing the full query result before your server can return the results.

Example 1. Using OffsetScrollPosition with Repository Query Methods

interface UserRepository extends Repository<User, Long> {

  Window<User> findFirst10ByLastnameOrderByFirstname(String lastname, OffsetScrollPosition position);
}

WindowIterator<User> users = WindowIterator.of(position -> repository.findFirst10ByLastnameOrderByFirstname("Doe", position))
  .startingAt(OffsetScrollPosition.initial()); (1)

1	Start from the initial offset at position `0`.

Scrolling using Keyset-Filtering

Offset-based requires most databases require materializing the entire result before your server can return the results. So while the client only sees the portion of the requested results, your server needs to build the full result, which causes additional load.

Keyset-Filtering approaches result subset retrieval by leveraging built-in capabilities of your database aiming to reduce the computation and I/O requirements for individual queries. This approach maintains a set of keys to resume scrolling by passing keys into the query, effectively amending your filter criteria.

The core idea of Keyset-Filtering is to start retrieving results using a stable sorting order. Once you want to scroll to the next chunk, you obtain a ScrollPosition that is used to reconstruct the position within the sorted result. The ScrollPosition captures the keyset of the last entity within the current Window. To run the query, reconstruction rewrites the criteria clause to include all sort fields and the primary key so that the database can leverage potential indexes to run the query. The database needs only constructing a much smaller result from the given keyset position without the need to fully materialize a large result and then skipping results until reaching a particular offset.

Keyset-Filtering requires the keyset properties (those used for sorting) to be non-nullable. This limitation applies due to the store specific null value handling of comparison operators as well as the need to run queries against an indexed source. Keyset-Filtering on nullable properties will lead to unexpected results.

Using KeysetScrollPosition with Repository Query Methods

interface UserRepository extends Repository<User, Long> {

  Window<User> findFirst10ByLastnameOrderByFirstname(String lastname, KeysetScrollPosition position);
}

WindowIterator<User> users = WindowIterator.of(position -> repository.findFirst10ByLastnameOrderByFirstname("Doe", position))
  .startingAt(ScrollPosition.keyset()); (1)

1	Start at the very beginning and do not apply additional filtering.

Keyset-Filtering works best when your database contains an index that matches the sort fields, hence a static sort works well. Scroll queries applying Keyset-Filtering require to the properties used in the sort order to be returned by the query, and these must be mapped in the returned entity.

You can use interface and DTO projections, however make sure to include all properties that you’ve sorted by to avoid keyset extraction failures.

When specifying your Sort order, it is sufficient to include sort properties relevant to your query; You do not need to ensure unique query results if you do not want to. The keyset query mechanism amends your sort order by including the primary key (or any remainder of composite primary keys) to ensure each query result is unique.