Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Searching for nested Documents in Solr(J)

Tags:

java

solr

solrj

I implemented a simple adding tool for PDF documents. I first create a master document (SolrInputDocument) of all documents. It gets fields like author, filehash, keywords, 'content_type=document' and so on.

After that I generate a SolrInputDocument for every page, this object gets an id like 'parentID_p01', the page as a field value, 'content_type=page' usw.

Finally, I add all page documents to my master document with addChildDocument()..

Now my question is, how do I perform a search for a given word inside all pages of all documents for example like this:

Document1.pdf 'this is my doc1 title' [2 matches]
    [Page 14] 'Example phrase for special a <em>word</em> given by.... 
    [Page 55] 'another <em>word</em> for this test

Document2.pdf 'doc2 title' [X matches]
    [Page 1] 'given <em>word</em> in this text 
    [Page 2] '<em>words</em> hit more than fists
    [Page 99] 'some <em>words</em> of wisdom

My first idea was to simple search with 'text:word~' and then group by the parent document but I didn't find a good way :-(

It seems that nested documents are a little new to solr and I didn't find an easy solution with SolrJ.

thanks in advance

like image 954
Smoki Avatar asked Jan 16 '15 12:01

Smoki


1 Answers

I have created data on solr in below format with parent child relation. Where one insurance_accounts have multiple person's vehicle insurance. One person can have multiple vehicles like car, bike etc. I have taken person as parent and vehicle as child document.

<insurance_accounts>
            <person>
                <firstname>John</firstname>
                <lastname>Jones</lastname>
                ...other details...
                <cars>
                    <car>
                        <make>honda</make>
                        <model>accord</model>
                    </car>
                    <car>
                        <make>Nissan</make>
                        <model>Maxima</model>
                    </car>
                </cars>
                <bikes>
                    <bike>
                        <make>yamaha</make>
                        <model>passion</model>
                    </bike>
                    <bike>
                        <make>Peugeot</make>
                        <model>Vivacity</model>
                    </bike>
                </bikes>
            </person>
            <person>
                ...
            </person>
          </insurance_accounts>

In below java code, I have used solrj 4.9 to create documents and run search query on SOLR. I have processed QueryResponse to show the required result and also given solr query URL.

You can take help from the given code snippet and let me know if it works or not.

import java.util.ArrayList;
import java.util.Collection;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.LinkedHashMap;
import org.apache.solr.client.solrj.SolrQuery;
import org.apache.solr.client.solrj.impl.HttpSolrServer;
import org.apache.solr.client.solrj.response.QueryResponse;
import org.apache.solr.common.SolrDocument;
import org.apache.solr.common.SolrDocumentList;
import org.apache.solr.common.SolrInputDocument;

public class SolrNestedSearch {

    static final String SOLR_URL = "http://localhost:8983/solr/collection1";

    public static void main ( String[] args ) throws Exception {

        HttpSolrServer solr = new HttpSolrServer( SOLR_URL );

        // Docs to submit
        Collection<SolrInputDocument> batch = new ArrayList<SolrInputDocument>();

        // Parent Doc 1, a person mamed John Jones
        SolrInputDocument person1 = new SolrInputDocument();
        person1.addField( "id",            "john_jones" );
        person1.addField( "content_type",  "person"     );
        // "_t" suffix tells Solr that it's text
        person1.addField( "first_name_t",  "John"       );
        person1.addField( "last_name_t",   "Jones"      );
        // states and history used in edismax examples
        person1.addField( "states_t",      "California Nevada Idaho Maine" );
        person1.addField( "history_t",     "safe accident accident accident accident accident" );

        // child docs, the vehicles he owns
        SolrInputDocument p1_car1 = new SolrInputDocument();
        p1_car1.addField( "id",            "jj_car1"    );
        p1_car1.addField( "content_type",  "car"        );
        // For cars "make" is an alias for "manufacturer"
        p1_car1.addField( "make_t",        "Honda"      );
        p1_car1.addField( "model_t",       "Accord"     );

        SolrInputDocument p1_car2 = new SolrInputDocument();
        p1_car2.addField( "id",            "jj_car2"    );
        p1_car2.addField( "content_type",  "car"        );
        p1_car2.addField( "make_t",        "Nissan"     );
        p1_car2.addField( "model_t",       "Maxima"     );

        SolrInputDocument p1_bike1 = new SolrInputDocument();
        p1_bike1.addField( "id",           "jj_bike1"   );
        p1_bike1.addField( "content_type", "bike"       );
        p1_bike1.addField( "make_t",       "Yamaha"     );
        p1_bike1.addField( "model_t",      "Passion"    );

        SolrInputDocument p1_bike2 = new SolrInputDocument();
        p1_bike2.addField( "id",           "jj_bike2"   );
        p1_bike2.addField( "content_type", "bike"       );
        p1_bike2.addField( "make_t",       "Peugeot"    );
        p1_bike2.addField( "model_t",      "Vivacity"   );

        // Add children to parent
        person1.addChildDocument( p1_car1  );
        person1.addChildDocument( p1_car2  );
        person1.addChildDocument( p1_bike1 );
        person1.addChildDocument( p1_bike2 );

        // Add parent to batch
        batch.add( person1 );


        // Parent Doc 2, person mamed Satish Smith
        SolrInputDocument person2 = new SolrInputDocument();
        person2.addField( "id",           "satish_smith" );
        person2.addField( "content_type", "person"       );
        person2.addField( "first_name_t", "Satish"       );
        person2.addField( "last_name_t",  "Smith"        );
        person2.addField( "states_t",     "California Texas California Maine Vermont Connecticut" );
        person2.addField( "history_t",    "safe safe safe safe safe safe safe safe accident" );

        // Vehicles (child docs)
        SolrInputDocument p2_car1 = new SolrInputDocument();
        p2_car1.addField( "id",            "ss_car1"     );
        p2_car1.addField( "content_type",  "car"         );
        p2_car1.addField( "make_t",        "Peugeot"     );
        p2_car1.addField( "model_t",       "iOn"         );
        SolrInputDocument p2_bike1 = new SolrInputDocument();
        p2_bike1.addField( "id",           "ss_bike1"    );
        p2_bike1.addField( "content_type", "bike"        );
        p2_bike1.addField( "make_t",       "Honda"       );
        p2_bike1.addField( "model_t",      "Spree"       );
        // link objects and add to batch
        person2.addChildDocument( p2_car1  );
        person2.addChildDocument( p2_bike1 );
        batch.add( person2 );

        System.out.println( "Adding batch of " + batch.size() + " parent docs" );

        // Submit as a group
        solr.add( batch );
        solr.commit();



        Map<String,String> params = new LinkedHashMap<>();
        params.put( "parent_filter", "content_type:person" );
        params.put( "defType", "edismax" );
        params.put( "qf", "history_t states_t^100" );
        doQuery(solr,
                "eDismax3: Drivers matching accidents and/or California, and all of their vehicles, boost on State",
                "California accident report",
                "{!parent which=$parent_filter}",
                "*,[child parentFilter=$parent_filter]",
                params );

    }


    static void doQuery( HttpSolrServer solr, String description, String queryStr, String optFilter,
            String optFields, Map<String,String>extraParams ) throws Exception
    {
        SolrQuery q = new SolrQuery( queryStr );
        if ( null!=optFilter ) {
            q.addFilterQuery( optFilter );
        }
        if ( null!=optFields ) { 
            q.setParam( "fl", optFields );   
        }
        else {
            q.addField( "*" );  
        }
        if ( null!=extraParams ) {
            for ( Entry<String,String> param : extraParams.entrySet() ) {
                q.set( param.getKey(), param.getValue() );
            }
        }

        // Run and show results
        QueryResponse rsp = solr.query( q );
        SolrDocumentList docs = rsp.getResults();
        long numFound = docs.getNumFound();
        System.out.println( "Matched: " + numFound );
        int docCounter = 0;
        for (SolrDocument doc : docs) {
            docCounter++;
            System.out.println( "Doc # " + docCounter );
            for ( Entry<String, Object> field : doc.entrySet() ) {
                String name = field.getKey();
                Object value = field.getValue();
                System.out.println( "\t" + name + "=" + value );
            }
            List<SolrDocument> childDocs = doc.getChildDocuments();
            if ( null!=childDocs ) {
                for ( SolrDocument child : childDocs ) {
                    System.out.println( "\tChild doc:" );
                    for ( Entry<String, Object> field : child.entrySet() ) {
                        String name = field.getKey();
                        Object value = field.getValue();
                        System.out.println( "\t\t" + name + "=" + value );
                    }
                }
            }
        }

        System.out.println( "Query URL:" );
        System.out.println( SOLR_URL + "/select?" + q );
    }
}
like image 53
ManojP Avatar answered Sep 28 '22 11:09

ManojP