Since my index cannot be re-indexed easily, I had to go with Erick's first
suggestion. I thought others might be interested in an example of the code
(I did this using Lucene 2.4.1):
// This code worked best to deletes documents with a null field value...
BooleanQuery nullFieldsOnlyQuery = new BooleanQuery();
MatchAllDocsQuery matchAllDocsQuery = new MatchAllDocsQuery();
// ConstantScoreRangeQuery does not throw a BooleanQuery.TooManyClauses
exception.
// A regular RangeQuery does throws a BooleanQuery.TooManyClauses exception.
// Obviously the range may be different depending on the nature of the field
involved.
ConstantScoreRangeQuery nonNullFieldsRangeQuery = new
ConstantScoreRangeQuery( theFieldName,
"0", //zero
"z",
true,
true);
// Load up the BooleanQuery
nullFieldsOnlyQuery.add( new MatchAllDocsQuery(), BooleanClause.Occur.MUST
);
nullFieldsOnlyQuery.add( nonNullFieldsRangeQuery,
BooleanClause.Occur.MUST_NOT );
// Delete the documents from the archive index
indexWriter.deleteDocuments( nullFieldsOnlyQuery );
// End of code that worked best to delete documents with a null field value.
// The following code did NOT always work to delete documents with a null
field value.
// I did get the code below to delete documents with a null field value in
my Windows XP
// environment when I simply added a document with "" for the field value.
However, it did
// not work in my CentOS environment. I imagine this failure had more to do
with the way
// those documents with a null field value ended up erroneously being
indexed in CentOS
// rather than with the difference in the OS itself.
Term[] fieldTerms = new Term[ 1 ];
fieldTerms[0] = new Term( theFieldName, "" );
indexWriter.deleteDocuments( fieldTerms );
// End of code that did NOT always work for me.
Frank Geary
--
View this message in context:
http://lucene.472066.n3.nabble.com/Searching-for-null-empty-fields-how-to-use-field-TO-tp552476p1603932.html
Sent from the Lucene - Java Users mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
