Combining Numeric Range Query with Term Query in Lucene

12

votes

I would like to combine a numeric range query with a term query in Lucene. For example, I want to search for documents that I have indexed that contain between 10 and 20 pages and have the title "Hello World".

It does not seem possibly to use the QueryParser to generate this query for me; the range query that the QueryParser generates appears to be a text one.

I definitely would appreciate an example of how to combine a numeric range query with a term query. I would also be open taking an alternative to searching my index.

Thanks

javalucene

11

votes

Well it looks like I figured this one out on my own. You can use Query.combine() to OR queries together. I have included an example below.

String termQueryString = "title:\"hello world\"";
Query termQuery = parser.parse(termQueryString);

Query pageQueryRange = NumericRangeQuery.newIntRange("page_count", 10, 20, true, true);

Query query = termQuery.combine(new Query[]{termQuery, pageQueryRange});

5

votes

You can also create a custom QueryParser overriding protected Query getRangeQuery(...) method, which should return NumericRangeQuery instance when "page_count" field is encountered.

Like so...

public class CustomQueryParser extends QueryParser {

    public CustomQueryParser(Version matchVersion, String f, Analyzer a) {
        super(matchVersion, f, a);
    }

    @Override
    protected Query getRangeQuery(final String field, final String part1, final String part2, final boolean inclusive) throws ParseException {

        if ("page_count".equals(field)) {
            return NumericRangeQuery.newIntRange(field, Integer.parseInt(part1), Integer.parseInt(part2), inclusive, inclusive);
        }

        // return default
        return super.getRangeQuery(field, part1, part2, inclusive);    
    }
}

Then use CustomQueryParser when parsing textual queries..

Like so...

...
final QueryParser parser = new CustomQueryParser(Version.LUCENE_35, "some_default_field", new StandardAnalyzer(Version.LUCENE_35));
final Query q = parser.parse("title:\"hello world\" AND page_count:[10 TO 20]");
...

This all, of course, assumes that NumericField(...).setIntValue(...) was used when page_count values were added to documents

2

votes

You may use BooleanQuery:

var combinedQuery = new BooleanQuery();
combinedQuery.Add(new TermQuery(new Term("title","hello world")),Occur.MUST);
combinedQuery.Add(NumericRangeQuery.newIntRange("page_count", 10, 20, true, true),Occur.MUST);

0

votes

RangeQuery amountQuery = new RangeQuery(lowerTerm, upperTerm, true);

Lucene treats numbers as words, so the numbers are ordered alphabetically.

1
12
123
1234
etc.

That being said, you can still use the range query, you just need to be more clever about it.

In order to query numeric values correctly, you need to pad your integers so the same lengths (whatever your maximum supported value is)

Obviously, this doesn't work for negative numbers (since -2 < -1), and hopefully you won't have to deal with them. Here's a useful article for negatives if you do encounter them: http://wiki.apache.org/lucene-java/SearchNumericalFields

Combining Numeric Range Query with Term Query in Lucene

4 Answers