Introduction

The Xapian::ValueRangeProcessor was introduced in Xapian 1.0.0. It provides a powerful and flexible way to parse range queries in the users' query string.

This document describes the Xapian::ValueRangeProcessor class and its standard subclasses, how to create your own subclasses, and how these classes are used with Xapian::QueryParser.

Xapian::ValueRangeProcessor is a virtual base class, so you need to use a subclass of it. Xapian::QueryParser maintains a list of Xapian::ValueRangeProcessor objects which it tries in order for each range search in the query until one accepts it, or all have been tried (in which case an error is reported).

Each Xapian::ValueRangeProcessor is passed the start and end of the range. If it doesn't understand the range, it should return Xapian::BAD_VALUENO. If it does understand the range, it should return the value number to use with Xapian::Query::OP_VALUE_RANGE and if it wants to, it can modify the start and end values (to convert them to the correct format so that for the string comparison which OP_VALUE_RANGE uses).

StringValueRangeProcessor

This is the simplest of the standard subclasses. It understands any range passed (so it should always be the last ValueRangeProcessor) and it doesn't alter the range start or end.

For example, suppose you have stored author names in value number 4, and want the user to be able to filter queries by specifying ranges of values such as:

mars asimov..bradbury

To do this, you can use a StringValueRangeProcessor like so:

Xapian::QueryParser qp;
Xapian::StringValueRangeProcessor author_proc(4);
qp.add_valuerangeprocessor(&author_proc);

The parsed query will use OP_VALUE_RANGE, so query.get_description() would report:

Xapian::Query(mars:(pos=1) FILTER (VALUE_RANGE 4 asimov bradbury)

The VALUE_RANGE subquery will only match documents where value 4 is >= asimov and <= bradbury (using a string comparison).

DateValueRangeProcessor

This class allows you to implement date range searches. As well as the value number to search, you can tell it whether to prefer US-style month/day/year or European-style day/month/year, and specify the epoch year to use for interpreting 2 digit years (the default is day/month/year with an epoch of 1970). The best choice of settings depends on the expectations of your users. As these settings are only applied at search time, you can also easily offer different versions of your search front-end with different settings if that is useful.

For example, if your users are American and the dates present in your database can extend a decade or so into the future, you might use something like this which specifies to prefer US-style dates and that the epoch year is 1930 (so 02/01/29 is February 1st 2029 while 02/01/30 is February 1st 1930):

Xapian::QueryParser qp;
Xapian::DateValueRangeProcessor date_proc(0, true, 1930);
qp.add_valuerangeprocessor(&date_proc);

The dates are converted to the format YYYYMMDD, so the values you index also need to also be in this format - for example, if doc_time is a time_t:

char buf[9];
if (strftime(buf, sizeof(buf), "%Y%m%d", gmtime(&doc_time))) {
    doc.add_value(0, buf);
}

NumberValueRangeProcessor

This class allows you to implement numeric range searches. The numbers used may be any number which is representable as a double, but requires that the stored values which the range is being applied have been converted to strings at index time using the Xapian::sortable_serialise() method:

Xapian::Document doc;
doc.add_value(0, Xapian::sortable_serialise(price));

This method produces strings which will sort in numeric order, so you can use it if you want to be able to sort based on the value in numeric order, too.

The class allows a prefix or suffix to be specified which must be present on the values, allowing multiple NumberValueRangeProcessors to be active in the same queryparser. For example, this specifies that a prefix of "$" must be present on the first value (and may optionally be present on the second value):

Xapian::QueryParser qp;
Xapian::NumberValueRangeProcessor numrange_proc(0, "$", true);
qp.add_valuerangeprocessor(&numrange_proc);

Custom subclasses

You can easily create your own subclasses of Xapian::ValueRangeProcessor. Your subclass needs to implement a method Xapian::valueno operator()(std::string &begin, std::string &end) so for example you could implement a better version of the author range described above which only matches ranges with a prefix (e.g. author:asimov..bradbury) and lower-cases the names:

struct AuthorValueRangeProcessor : public Xapian::ValueRangeProcessor {
    AuthorValueRangeProcessor() {}

    Xapian::valueno operator()(std::string &begin, std::string &end) {
        if (begin.substr(0, 7) != "author:")
            return Xapian::BAD_VALUENO;
        begin.erase(0, 7);
        begin = Xapian::Unicode::tolower(begin);
        end = Xapian::Unicode::tolower(end);
        return 4;
    }
};

Using Several ValueRangeProcessors

If you want to allow the user to specify different types of ranges, you can specify multiple ValueRangeProcessor objects to use. Just add them in the order you want them to be checked:

Xapian::QueryParser qp;
AuthorValueRangeProcessor author_proc();
qp.add_valuerangeprocessor(&author_proc);
Xapian::DateValueRangeProcessor date_proc(0, false, 1930);
qp.add_valuerangeprocessor(&date_proc);

And then you can parse queries such as mars author:Asimov..Bradbury 01/01/1960..31/12/1969 successfully.