lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christian Reuschling <christian.reuschl...@gmail.com>
Subject Re: Empty numeric field
Date Wed, 15 Feb 2012 13:07:38 GMT
ah ok, I know what you mean. We have to read out the stored field
values later. A field can have multiple (stored) values (several
document.add(fieldable) invocations for one field). Further, we have
the problem that some field values are logically related to each
other. Since Lucene has no possibility to define relationships between
documents, and offer e.g. key-based, table-join-similar stuff, we took
this approach as a kind of approximation:

Example document:  (*fieldName: value1, value2, ..)
     * personId: personId1, personId2, personId3, personIdN
     * personName: personName1, personName2, personName3, personNameN
     * personAge: personAge1, personAge2, personAge3, personAgeN
     * tagId: tagId1, tagId2, tagId3, tagIdN
     * tagLabel: tagLabel1, tagLabel2, tagLabel3, tagLabelN
     *

In the case there is e.g. no person age known for person 2, we have to
insert an empty entry to ensure that all data for person 3 is in the
third stored field value in this document:

     * personName: personName1, personName2, personName3, personNameN
     * personAge: personAge1, EMPTYENTRY, personAge3, personAgeN

since you can't create a numeric Field instance with a null entry
(that would be wonderfull), we insert a string fieldable, so in this
example the values for 'personage1, personage3, and personageN are
NumericFields, and EMPTYENTRY is a standard string field with a
zero-length string value "". This results into exceptions when we sort
against the personAge field.

We also thinked about NaN or negative infinity as placeholder - but
these are only available at Float and Double. For Longs and Integers,
only min_value and max_value are offered. Further, we have to convert
this value back into null or something empty, whereby the empty string
value field integrates seamless as it should (with respect to the
exceptions ;) ) But maybe there is no other possibility to take the
max/min values?



2012/2/15 Uwe Schindler <uwe@thetaphi.de>:
> Hi,
>
> This looks like an XY problem (http://www.perlmonks.org/index.pl?node_id=542341). Maybe
you should first explain to us, why you need that. In Lucene fields have no "equal length"
or something like that, especially numeric fields are tokenized and contain of several tokens
separately indexed. So what do you mean with equal length? Why must this "length" be identical?
>
> The only suggestion is to index a "fake" placeholder value (like -1, infinity, NaN).
If you only need it in the "stored" fields, just store it but don't index it.
>
> Uwe
>
> -----
> Uwe Schindler
> H.-H.-Meier-Allee 63, D-28213 Bremen
> http://www.thetaphi.de
> eMail: uwe@thetaphi.de
>
>
>> -----Original Message-----
>> From: Christian Reuschling [mailto:christian.reuschling@gmail.com]
>> Sent: Wednesday, February 15, 2012 12:58 PM
>> To: java-user
>> Subject: Empty numeric field
>>
>> Hi all,
>>
>> for some reason, we need empty numeric field values (to ensure that the
>> length of the field value list is constant). We tried to add an empty String-
>> Fieldable instead in the case a value is not present, which seemed to work for
>> searching.
>> Nevertheless, when we want to sort against this field, sadly we fall into
>> exceptions.
>>
>> Is there any possibility to store empty numeric fields into a document/the
>> index?
>>
>> Thanks
>>
>> Chris
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: java-user-help@lucene.apache.org
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>

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


Mime
View raw message