[jira] [Commented] (JENA-686) Add support for cross field conjunctive queries in jena-text

["Brian McBride (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/JENA-686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13977277#comment-13977277 ] 

Brian McBride commented on JENA-686:
------------------------------------

To meet our local needs I have attempted to address this in three steps:

  1. Created an external text indexer that produces Lucene documents that contain all the indexed properties for a resource.
  2.Create an extension point in jenatext to allow an assembler to specify the class that is used to create the Lucene document that is indexed when a dataset is updated.
  3. Create an implementation of TextDocProducer that also creates a Lucene document containing all the known indexed properties for a resource and uses update rather than add.

I have done a local solution, because I'm not clear on how best to implement a solution that supports both conjunctive searches and the ability to restrict searches to a single graph.


> Add support for cross field conjunctive queries in jena-text
> ------------------------------------------------------------
>
>                 Key: JENA-686
>                 URL: https://issues.apache.org/jira/browse/JENA-686
>             Project: Apache Jena
>          Issue Type: Improvement
>          Components: Text
>    Affects Versions: Fuseki 1.0.1, Jena 2.11.1
>            Reporter: Brian McBride
>         Attachments: TestDatasetWithBatchProducer.java
>
>
> We have a project where we are doing text search on addresses and wish to do jena text queries like "city:liverpool AND street:green".  These queries return no results, whilst queries like "street:green AND street:lane" work fine.
> The reason is that jena text indexes each property in a separate Lucene document, so there is no Lucene document matching city:liverpool AND street:green, there are two documents, one for each property.
> Given the scale of our data, we really want to do the conjunctive query in Lucene and not two separate queries and then a filter in SPARQL.
> I will attach a test case from an attempt to solve this for us to illustrate.



--
This message was sent by Atlassian JIRA
(v6.2#6252)