Hi,
I thought about appending a * at the end of each word like <word>*. I did
not like the idea and it becomes cumbersome when using the complex
QueryParser syntax.
I also investigated the option of overriding the getFieldQuery via
inheritance, but I doubt it works at the method does more than just return a
TermQuery. For instance it also returns PhraseQuery, MultiPhraseQuery and
BooleanQuery.
My solution to the problem is post-processing of the returned query from the
MultiFieldQueryParser.
I iterate over the query graph and replaces each TermQuery with a
PrefixQuery.
This is how I do it. Beware! It's C# code, but I'm pretty sure that you will
be able to interpret it anyway :-)
public static class QueryModifier
{
public static Query AppendWildcard(Query q)
{
BooleanQuery boolQuery = q as BooleanQuery;
if (boolQuery != null)
{
Iterate(ref boolQuery);
return boolQuery;
}
TermQuery termQuery = q as TermQuery;
if (termQuery != null)
return new PrefixQuery(termQuery.GetTerm());
return q;
}
private static void Iterate(ref BooleanQuery boolQuery)
{
BooleanClause[] boolClauses = boolQuery.GetClauses();
BooleanClause clause;
for (int i = 0; i < boolClauses.Length; i++)
{
clause = boolClauses[i];
Query clauseQuery = clause.GetQuery();
BooleanQuery boolClauseQuery = clauseQuery as BooleanQuery;
if (boolClauseQuery != null)
{
Iterate(ref boolClauseQuery);
boolClauses[i] = new BooleanClause(boolClauseQuery,
boolClauses[i].GetOccur()); ;
}
TermQuery termClauseQuery = clauseQuery as TermQuery;
if (termClauseQuery != null)
{
// Change TermQuery -> PrefixQuery to wildcard search
like <word>*
PrefixQuery prefixQuery = new
PrefixQuery(termClauseQuery.GetTerm());
boolClauses[i] = new BooleanClause(prefixQuery,
boolClauses[i].GetOccur());
}
}
boolQuery = new BooleanQuery(boolQuery.IsCoordDisabled());
foreach (BooleanClause claus in boolClauses)
{
boolQuery.Add(claus);
}
}
}
Thanks for the suggestions.
:-)
Anders Lybecker
-----Original Message-----
From: Shai Erera [mailto:[EMAIL PROTECTED]
Sent: 22. november 2007 17:04
To: [email protected]
Subject: Re: Force MultiFieldQueryParser always to use PrefixQuery
Hi
I wrote the following class:
public class AlwaysPrefixMultiFieldQP extends MultiFieldQueryParser {
public MyQP(String[] fields, Analyzer analyzer) {
super(fields, analyzer);
}
protected Query getFieldQuery(String field, String queryText) throws
ParseException {
if (field != null) {
return new PrefixQuery(new Term(field, queryText));
}
return super.getFieldQuery(field, queryText);
}
}
What it does is override getFieldQuery. If the field is not null, it creates
a new PrefixQuery. Following is a sample code:
AlwaysPrefixMultiFieldQP m = new AlwaysPrefixMultiFieldQP(new
String[] { "field" }, analyzer);
Query q = m.parse("sof was");
System.out.println(q);
This code prints: (field:sof*) (field:was*), which is I believe what you
need.
As for splitting the query (pre processing) - this is not recommended. It
may work great for space separated languages, however may produce poor
results for Asian languages (for example). The way I propose above makes use
of the Analyzer you use, thus guarantees you append a '*' to all the words
in the query.
Hope this helps.
On Nov 22, 2007 3:35 PM, Erick Erickson <[EMAIL PROTECTED]> wrote:
> The simplest way would be to pre-process the query. That
> is, just split on words and add the '*' as appropriate.
>
> Erick
>
> On Nov 21, 2007 2:16 PM, Anders Lybecker <[EMAIL PROTECTED]> wrote:
>
> > How do I force the MultiFieldQueryParser to interpret a string like
> > "dock boat" as "dock* boat*" and therefore use PrefixQuery instead of
> > TemQuery?
> >
> > The customer wants always to search with <word>* as default when
> entering
> > <word>
> >
> > :-)
> > Anders Lybecker
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [EMAIL PROTECTED]
> > For additional commands, e-mail: [EMAIL PROTECTED]
> >
> >
>
--
Regards,
Shai Erera
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.503 / Virus Database: 269.16.4/1146 - Release Date: 22-11-2007
18:55
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.503 / Virus Database: 269.16.4/1146 - Release Date: 22-11-2007
18:55
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
