You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucenenet.apache.org by Andrei Alecu <an...@tachyon-labs.com> on 2009/07/29 20:53:31 UTC

Is this query possible?

Hello all,

I have a MultiSearcher searching two separate indexes.

One of the indexes (index B) has some extra terms in its documents 
compared to the other (index A).

Consider this scenario:

Index A documents:
field: fulltext

Index B documents:
field: fulltext
field: visible

I want for a query such as :

fulltext:"car" AND visible:1

To return all the matching "car" results in Index A, thus ignoring the 
'visibile' constraint if it's not defined for the document. How can I do 
that? Right now the results completely skip Index A because of no 
matching 'visible' field.

Thanks.

Re: Is this query possible? SOLVED

Posted by Andrei Alecu <an...@tachyon-labs.com>.
In case anyone is interested and has a similar requirement, here's how I 
solved this problem (source below):

Usage:

    var parser = new QueryParser("fulltext", new StandardAnalyzer());
    parser.SetDefaultOperator(QueryParser.Operator.AND);
    var searchQuery = parser.Parse("car");

    var softBooleanQuery = new SoftBooleanQuery();
    var visibleQuery = new SoftTermQuery(new Term("visible", "1"));

    softBooleanQuery.Add(searchQuery, BooleanClause.Occur.MUST);
    softBooleanQuery.Add(visibleQuery, BooleanClause.Occur.MUST);

    ... search with softBooleanQuery ...
       
Helper classes:

    public class SoftBooleanQuery : BooleanQuery
    {
        public override Query Rewrite(IndexReader reader)
        {
            restart:
            foreach (BooleanClause clause in Clauses())
            {
                var softQuery = clause.GetQuery() as SoftTermQuery;
                if (softQuery != null)
                {
                    if (softQuery.Rewrite(reader) == null)
                    {
                        // rewritten query is null, remove it from the 
clauses (see SoftTermQuery.Rewrite)
                        Clauses().Remove(clause);
                        goto restart;
                    }
                }
            }           

            return base.Rewrite(reader);
        }
    }

    public class SoftTermQuery : TermQuery
    {
        public SoftTermQuery(Term t) : base(t)
        {
        }

        public static bool IsTermIndexed(IndexReader reader, Term t)
        {
            return reader.Terms(t).Term().Field() == t.Field();
        }

        public override Query Rewrite(IndexReader reader)
        {
            Term t = GetTerm();
            // if the term does not exist in this IndexReader, return null
            return IsTermIndexed(reader, t) ? base.Rewrite(reader) : null;
        }

        public override string ToString(string field)
        {
            return base.ToString(field);
        }
    }



Andrew C. Smith wrote:
> Maybe you can perform the OR clause with the MultiSearcher and use a custom
> made QueryFilter to filter the results that you don't need out of the first
> index.
>
> ~Andrew
>
> On Wed, Jul 29, 2009 at 3:22 PM, Andrei Alecu <an...@tachyon-labs.com>wrote:
>
>   
>> Right, I was doing two separate searches before, but I was hoping there was
>> a better way.
>>
>> Thanks.
>>
>>
>> Digy wrote:
>>
>>     
>>> Either you have to reindex "Index A" and add a "visible" field with values
>>> both 0 and 1, or make two seperate searches.
>>>
>>> DIGY
>>>
>>> -----Original Message-----
>>> From: Andrei Alecu [mailto:andrei@tachyon-labs.com] Sent: Wednesday, July
>>> 29, 2009 9:54 PM
>>> To: lucene-net-dev@incubator.apache.org
>>> Subject: Is this query possible?
>>>
>>> Hello all,
>>>
>>> I have a MultiSearcher searching two separate indexes.
>>>
>>> One of the indexes (index B) has some extra terms in its documents
>>> compared to the other (index A).
>>>
>>> Consider this scenario:
>>>
>>> Index A documents:
>>> field: fulltext
>>>
>>> Index B documents:
>>> field: fulltext
>>> field: visible
>>>
>>> I want for a query such as :
>>>
>>> fulltext:"car" AND visible:1
>>>
>>> To return all the matching "car" results in Index A, thus ignoring the
>>> 'visibile' constraint if it's not defined for the document. How can I do
>>> that? Right now the results completely skip Index A because of no matching
>>> 'visible' field.
>>>
>>> Thanks.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>       
>>     
>
>

Re: Is this query possible?

Posted by "Andrew C. Smith" <po...@gmail.com>.
Maybe you can perform the OR clause with the MultiSearcher and use a custom
made QueryFilter to filter the results that you don't need out of the first
index.

~Andrew

On Wed, Jul 29, 2009 at 3:22 PM, Andrei Alecu <an...@tachyon-labs.com>wrote:

> Right, I was doing two separate searches before, but I was hoping there was
> a better way.
>
> Thanks.
>
>
> Digy wrote:
>
>> Either you have to reindex "Index A" and add a "visible" field with values
>> both 0 and 1, or make two seperate searches.
>>
>> DIGY
>>
>> -----Original Message-----
>> From: Andrei Alecu [mailto:andrei@tachyon-labs.com] Sent: Wednesday, July
>> 29, 2009 9:54 PM
>> To: lucene-net-dev@incubator.apache.org
>> Subject: Is this query possible?
>>
>> Hello all,
>>
>> I have a MultiSearcher searching two separate indexes.
>>
>> One of the indexes (index B) has some extra terms in its documents
>> compared to the other (index A).
>>
>> Consider this scenario:
>>
>> Index A documents:
>> field: fulltext
>>
>> Index B documents:
>> field: fulltext
>> field: visible
>>
>> I want for a query such as :
>>
>> fulltext:"car" AND visible:1
>>
>> To return all the matching "car" results in Index A, thus ignoring the
>> 'visibile' constraint if it's not defined for the document. How can I do
>> that? Right now the results completely skip Index A because of no matching
>> 'visible' field.
>>
>> Thanks.
>>
>>
>>
>>
>>
>>
>>
>>
>
>

Re: Is this query possible?

Posted by Andrei Alecu <an...@tachyon-labs.com>.
Right, I was doing two separate searches before, but I was hoping there 
was a better way.

Thanks.

Digy wrote:
> Either you have to reindex "Index A" and add a "visible" field with values
> both 0 and 1, or make two seperate searches.
>
> DIGY
>
> -----Original Message-----
> From: Andrei Alecu [mailto:andrei@tachyon-labs.com] 
> Sent: Wednesday, July 29, 2009 9:54 PM
> To: lucene-net-dev@incubator.apache.org
> Subject: Is this query possible?
>
> Hello all,
>
> I have a MultiSearcher searching two separate indexes.
>
> One of the indexes (index B) has some extra terms in its documents 
> compared to the other (index A).
>
> Consider this scenario:
>
> Index A documents:
> field: fulltext
>
> Index B documents:
> field: fulltext
> field: visible
>
> I want for a query such as :
>
> fulltext:"car" AND visible:1
>
> To return all the matching "car" results in Index A, thus ignoring the 
> 'visibile' constraint if it's not defined for the document. How can I do 
> that? Right now the results completely skip Index A because of no 
> matching 'visible' field.
>
> Thanks.
>
>
>
>
>
>
>

RE: Is this query possible?

Posted by Digy <di...@gmail.com>.
Either you have to reindex "Index A" and add a "visible" field with values
both 0 and 1, or make two seperate searches.

DIGY

-----Original Message-----
From: Andrei Alecu [mailto:andrei@tachyon-labs.com] 
Sent: Wednesday, July 29, 2009 9:54 PM
To: lucene-net-dev@incubator.apache.org
Subject: Is this query possible?

Hello all,

I have a MultiSearcher searching two separate indexes.

One of the indexes (index B) has some extra terms in its documents 
compared to the other (index A).

Consider this scenario:

Index A documents:
field: fulltext

Index B documents:
field: fulltext
field: visible

I want for a query such as :

fulltext:"car" AND visible:1

To return all the matching "car" results in Index A, thus ignoring the 
'visibile' constraint if it's not defined for the document. How can I do 
that? Right now the results completely skip Index A because of no 
matching 'visible' field.

Thanks.