CODEWITHSUNDEEP
javalucene
Chandra
Chandra

Reputation: 777

Lucene not null query?

How can we construct a query to search for particular field to be not null?

field_name:* is not working. I tried field_name:[a* to z*] this works fine for English, but does not cover all languages.

Any alternative suggestions?

Upvotes: 26

Views: 48519

Answers (8)

kade99
kade99

Reputation: 37

In the current Lucene version FieldValueQuery is outdated. Currently you can use:

new DocValuesFieldExistsQuery(name)

This works only for SortedDocValuesFields, so you have to add them while creating the document.

doc.add(new SortedDocValuesField(name, new BytesRef(value));
doc.add(new StringField(name, value, Field.Store.Yes)); //optional

The second line enables you to retrieve the value (DocValuesFields' values cannot be read as Strings).
Note that lucene allows for multiple fields with the same name as in the aforementioned example.

Upvotes: 3

Carolyn Conway
Carolyn Conway

Reputation: 1506

For anyone else arriving late to the question, the documentation includes this little snippet:

  • where the field title has any non-null value:
    _exists_:title

Upvotes: 17

Oleksandr Martynov
Oleksandr Martynov

Reputation: 330

Have a look at org.apache.lucene.search.FieldValueQuery:

A Query that matches documents that have a value for a given field as reported by LeafReader#getDocsWithField(String).

please note, that it only works with DocValues, so you would need to change a way you create your document:

document.add(new StringField("field-name", "field-value", Field.Store.YES));
document.add(new SortedDocValuesField("field-name", new BytesRef("field-value")));

here I added two fields - you still need regular StringField to get value. You may choose to use BinaryDocValues#get() for older versions of Lucene, but as I see, it is removed in v7. Not sure, what is a proper way to retrieve a value now - please check this

Upvotes: 0

Phil
Phil

Reputation: 84

I have just started to play around with lucene (via logstash elastic search) and find that this seems to work from the kibana UI. I am not sure yet if this is some intelligence in elastic search or kibana, i just know that elastic search borrows from the lucene syntax.

application:unit-test && !exception

will return all results from my unit tests which have not had an exception

application:unit-test && exception

returns those which have a non null exception indexed. so you might try just

field

or 

!field

Upvotes: -1

Karolis
Karolis

Reputation: 386

I found this to work in some cases field:([0 TO 9] [a TO z])

Upvotes: 19

Edd
Edd

Reputation: 8570

I was having the same problem but there's a property you can set on the query parser which lets you have wildcard characters at the start of a search term.

queryParser.setAllowLeadingWildcard(true);

This solved the problem for me

Please see Wildcard at the Beginning of a searchterm -Lucene

Upvotes: 3

nikhil500
nikhil500

Reputation: 3450

This is currently not supported by Lucene. See this for a discussion.

An alternative option may be to store some pre-defined string (like nullnullnullnull) as the field value if it is null. Then you can use a negative filter to remove these records. (I don't like this much, but can't think of a better option)

Upvotes: 6

mindas
mindas

Reputation: 26703

Try field:[* TO *] or field:["" TO *]. But it's probably better to use a filter for this though.

Upvotes: 11

Related Questions