How control chunk processed elements before write them to database?

I have this situation in my jt:

• Reader: Read a big csv file
• Processor: Process lines, check data using a database Select, if data exists, marks it to update. If not exists marks it to insert.
• Writer: Classifier to redirect the item to insert/update Writers.

My commit-interval is 1000. If my file contains repeated items in the same "chunk" interval. ¿How can detect that these item should be marked as updated instead insert?

File example:

Item1Code|2023-05-01|02 --> detect not exists, mark as insert
Item2Code|2023-05-02|03 --> detect not exists, mark as insert
Item3Code|2023-05-03|03 --> detect not exists, mark as insert
Item1Code|2023-05-04|03 --> detect not exists, mark as insert (should be detected as exists because It will inserted in line1)
Item4Code|2023-05-05|03 --> detect not exists, mark as insert
...
commit here <<

I only think this can be resolved using commit-interval 1. There are some tools in Spring batch to resolve this type of problem?

thanks in advance

There is no out-of-the-box tool for this specific problem. However it can be easy implemented using an in-memory cache and the ChunkListener interface.

Code example of the component:

@Component
public class InsertedCache implements ChunkListener {
    private Set<String> cache = new HashSet<>();

    @Override
    public void beforeChunk(ChunkContext context) {
        cache.clear();
    }

    public boolean isInserted(String id) {
        return !cache.add(id);
    }
}

Usage example in the ItemProcessor:

@Slf4j
public class CustomerItemProcessor implements ItemProcessor {

    @Autowired
    private InsertedCache insertedCache;

    @Override
    public Object process(Object item) {
        if (item instanceof Customer) {
            Customer customer = (Customer) item;
            if (insertedCache.isInserted(customer.getName())) {
                log.info("already iserted customer: {}", customer);
            }
        }
        return item;
    }
}