Unable to set _id in bulk index with raw source documents

[diegobenincasa]

I've tried to bulk index a bunch of JSON raw records into ES, and I needed to set custom _id values for them. Individual indexing works by calling "IndexQueryBuilder().withId(some_id_value)" and then calling the individual index method, but calling the "bulkIndex" method doesn't consider what was defined as the _id desired value.

Here's the code that ignores the ".withId" call:

package <ommited for safety>;

import java.util.List;
import java.util.Map;
import java.util.stream.Collectors;

import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.data.elasticsearch.core.ElasticsearchOperations;
import org.springframework.data.elasticsearch.core.IndexOperations;
import org.springframework.data.elasticsearch.core.mapping.IndexCoordinates;
import org.springframework.data.elasticsearch.core.query.IndexQuery;
import org.springframework.data.elasticsearch.core.query.IndexQueryBuilder;
import org.springframework.stereotype.Service;

@Service
public class ESService {
    
    @Autowired
    private ElasticsearchOperations esOperations;

    public void index(String baseName, Map<Integer, String> jsonDocuments, String indexName, Long exp_time) {

        IndexCoordinates indexCoordinates = IndexCoordinates.of(indexName);
        
        IndexOperations indexOps = esOperations.indexOps(indexCoordinates);
        if(!indexOps.exists()) {
            indexOps.create();
            try {
                Thread.sleep(1000);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }

        List<IndexQuery> indexQueries = jsonDocuments.keySet().stream()
            .map(id -> new IndexQueryBuilder()
                .withSource(jsonDocuments.get(id))
                .withId(id.toString()) // HERE IS THE IGNORED CALL
                .withIndex(indexName)
                .build())
            .collect(Collectors.toList());

        try {
            esOperations.bulkIndex(indexQueries, indexCoordinates);
        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}

It should be interesting (if not mandatory) that the user could set the _id for each individual record sent in the bulk request.

I was able to loop over individual IndexQuery objects and send them one by one to ES, and that correctly sets the _id value, but that increases processing time a lot - in my scenario of ~2m JSON records, elapsed time increases from 15-20 minutes (in batches of 2000 records) to ~3 hours.

[sothawo]

Your method does not compile.
jsonDocuments is of type Map<Integer, String> so the jsonDocuments.keySet().stream() provides a stream of Integer. IndexQueryBuilder.withId(String id) takes a String and no Integer and never had a different type for the id.

[diegobenincasa]

I'm sorry, I actually sent the code with that mistake. Please see my edited post.

The problem is not in the compilation (it was my mistake in copying and pasting the code here), but in the "withId" call. You can set anything there and the setting is ignored in the bulk request.

[naidusana]

I'm sorry, I actually sent the code with that mistake. Please see my edited post.

The problem is not in the compilation (it was my mistake in copying and pasting the code here), but in the "withId" call. You can set anything there and the setting is ignored in the bulk request.

Is this fix working?