In this release, the Spring dependencies are upgraded to the following versions:

Embracing JDK 21 LTS is one of the main themes for Spring Batch 5.1, especially the support of virtual threads from Project Loom. In this release, virtual threads can be used in all areas of the framework, like running a concurrent step with virtual threads or launching multiple steps in parallel using virtual threads.

Thanks to the well designed separation of concerns in Spring Batch, threads are not managed directly. Thread management is rather delegated to TaskExecutor implementations from Spring Framework. This programming-to-interface approach allows you to switch between TaskExecutor implementations in a transparent and a flexible way.

In Spring Framework 6.1, a new TaskExecutor implementation based on virtual threads has been introduced, which is the VirtualThreadTaskExecutor. This TaskExecutor can be used in Spring Batch wherever a TaskExecutor is required.

When using the JpaItemWriter, the JPA persistence context can quickly grow when the chunk size is large enough. This might lead to OutOfMemoryError errors if not cleared appropriately in a timely manner.

In this release, a new option named clearPersistenceContext has been introduced in the JpaItemWriter to clear the persistence context after writing each chunk of items. This option improves the memory management of chunk-oriented steps dealing with large amounts of data and big chunk sizes.

Up to version 5.0, Spring Batch provided two decorators SynchronizedItemStreamReader and SynchronizedItemStreamWriter to synchronize thread access to ItemStreamReader#read and ItemStreamWriter#write. Those decorators are useful when using non thread-safe item streams in multi-threaded steps.

While those decorators work with ItemStream implementations, they are not usable with non-item streams. For example, those decorators cannot be used to synchronize access to ListItemReader#read or KafkaItemWriter#write.

For users convenience, this release introduces new decorators for non-item streams as well. With this new feature, all item readers and writers in Spring Batch can now be synchronized without having to write custom decorators.

Up to version 5.0, the MongoItemReader provided by Spring Batch used pagination, which is based on MongoDB’s skip operation. While this works well for small/medium data sets, it starts to perform poorly with large data sets.

This release introduces the MongoCursorItemReader, a new cursor-based item reader for MongoDB. This implementation uses cursors instead paging to read data from MongoDB, which improves the performance of reads on large collections. For consistency with other cursor/paging readers, the current MongoItemReader has been renamed to MongoPagingItemReader.

Up to version 5.0, the MongoItemWriter supported two operations: upsert and delete. While the upsert operation works well for both inserts and updates, it does not perform well for items that are known to be new in the target collection.

Similar to the persist and merge operations in the JpaItemWriter, this release adds a new operation named insert in the MongoItemWriter, which is designed for bulk inserts. This new option performs better than upsert for new items as it does not require an additional lookup to check if items already exist in the target collection.