question on the implementation of a SetFilter

[christophe blin <cb...@tennaxia.com>]

Hi,

I'd like to implement a SetFilter like describe in
http://www.nabble.com/Re%3A-Too-many-clauses-p1145373.html

At the moment, I have a working implementation but there are some gotchas I
do not understand (i.e I take the code from RangeFilter and adapt it as
suggested by the post)

Could someone have a look at my implementation and do some suggestion about
the TODO flags (or about the code if it is not so good)

public class SetFilter extends org.apache.lucene.search.Filter {
    private String fieldName;
    private Set<String> fieldAuthorizedValues;

    public SetFilter(String fieldName, Set<String> fieldAuthorizedValues) {
        this.fieldName = fieldName;
        this.fieldAuthorizedValues = fieldAuthorizedValues;
    }

    @Override
    public BitSet bits(IndexReader reader) throws IOException {
        BitSet bits = new BitSet(reader.maxDoc());
        //builds an enum only on the inspected field
        TermEnum enumerator = reader.terms(new Term(fieldName,""));

        try {
            //TODO: why should this happen ?
            if (enumerator.term() == null) {
                return bits;
            }

            TermDocs termDocs = reader.termDocs();
            try {
                do {
                    Term term = enumerator.term();
                    //TODO: why the term can be null ? 
                    //TODO: why the term can have a field different from the
inspected one ?
                    if (term != null && term.field().equals(fieldName)) {
                        if
(this.fieldAuthorizedValues.contains(term.text())) {
                            /* we have a good term, find the docs */
                            termDocs.seek(enumerator.term());
                            while (termDocs.next()) {
                                bits.set(termDocs.doc());
                            }
                        }
                    } else {
                        break;
                    }
                }
                while (enumerator.next());

            } finally {
                termDocs.close();
            }
        } finally {
            enumerator.close();
        }

        return bits;
    }
}

-- 
View this message in context: http://www.nabble.com/question-on-the-implementation-of-a-SetFilter-tp14525027p14525027.html
Sent from the Lucene - Java Users mailing list archive at Nabble.com.


---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org

Re: question on the implementation of a SetFilter

[Paul Elschot <pa...@xs4all.nl>]

On Friday 28 December 2007 13:00:03 christophe blin wrote:
> 
> Hi,
> 
> I'd like to implement a SetFilter like describe in
> http://www.nabble.com/Re%3A-Too-many-clauses-p1145373.html
> 
> At the moment, I have a working implementation but there are some gotchas I
> do not understand (i.e I take the code from RangeFilter and adapt it as
> suggested by the post)
> 
> Could someone have a look at my implementation and do some suggestion about
> the TODO flags (or about the code if it is not so good)

The code looks good to me. Call it classic Lucene style if you will.
A TermEnum can return null the first time term() is called, note that the next()
method is called after the first call to term().
That's the way the TermEnum is implemented iirc since the first days of Lucene.

Maybe it's time to replace it with a TermIterator, or add that on top of TermEnum.

The TermEnum starts at a given field/term and iterates through all indexed
terms after that, including terms with field names ordered later than
the given field. That's why the field name must be checked in the Term.

Perhaps that could be another bit functionality for a future TermIterator?

Regards,
Paul Elschot


> 
> public class SetFilter extends org.apache.lucene.search.Filter {
>     private String fieldName;
>     private Set<String> fieldAuthorizedValues;
> 
>     public SetFilter(String fieldName, Set<String> fieldAuthorizedValues) {
>         this.fieldName = fieldName;
>         this.fieldAuthorizedValues = fieldAuthorizedValues;
>     }
> 
>     @Override
>     public BitSet bits(IndexReader reader) throws IOException {
>         BitSet bits = new BitSet(reader.maxDoc());
>         //builds an enum only on the inspected field
>         TermEnum enumerator = reader.terms(new Term(fieldName,""));
> 
>         try {
>             //TODO: why should this happen ?
>             if (enumerator.term() == null) {
>                 return bits;
>             }
> 
>             TermDocs termDocs = reader.termDocs();
>             try {
>                 do {
>                     Term term = enumerator.term();
>                     //TODO: why the term can be null ? 
>                     //TODO: why the term can have a field different from the
> inspected one ?
>                     if (term != null && term.field().equals(fieldName)) {
>                         if
> (this.fieldAuthorizedValues.contains(term.text())) {
>                             /* we have a good term, find the docs */
>                             termDocs.seek(enumerator.term());
>                             while (termDocs.next()) {
>                                 bits.set(termDocs.doc());
>                             }
>                         }
>                     } else {
>                         break;
>                     }
>                 }
>                 while (enumerator.next());
> 
>             } finally {
>                 termDocs.close();
>             }
>         } finally {
>             enumerator.close();
>         }
> 
>         return bits;
>     }
> }
> 



---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org