[jira] [Commented] (JENA-1652) jena-text analyzer regression

["Osma Suominen (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/JENA-1652?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16721628#comment-16721628 ] 

Osma Suominen commented on JENA-1652:
-------------------------------------

It appears to be a difference in Lucene behavior. A plain query such as {{Guppy}} will be normalized to {{guppy}} by the QueryParser, but a wildcard query such as {{Gup*}} will not be normalized and instead is kept as it is. In earlier Lucene versions, wildcard queries were lowercased by default even though the full analysis/normalization was not done unless you used AnalyzingQueryParser.

A possible workaround is to lowercase the query string in jena-text before it is given to QueryParser.parse().

I wonder if the change of behavior in Lucene was intended?

> jena-text analyzer regression
> -----------------------------
>
>                 Key: JENA-1652
>                 URL: https://issues.apache.org/jira/browse/JENA-1652
>             Project: Apache Jena
>          Issue Type: Bug
>          Components: Text
>    Affects Versions: Jena 3.10.0
>         Environment: Ubuntu 16.04
> java version "1.8.0_191"
> Java(TM) SE Runtime Environment (build 1.8.0_191-b12)
> Java HotSpot(TM) 64-Bit Server VM (build 25.191-b12, mixed mode)
>            Reporter: Osma Suominen
>            Assignee: Osma Suominen
>            Priority: Major
>             Fix For: Jena 3.10.0
>
>
> I noticed that Skosmos unit tests are failing when run with Fuseki 3.10 snapshots:
> https://github.com/NatLibFi/Skosmos/issues/828
> Digging a bit deeper, it seems that jena-text is no longer applying the analyzer on query strings as it used to in 3.9.0. The most likely reason for this change seems to be the Lucene upgrade (JENA-1621) which may have affected how analyzers are applied.
> Here is the text analyzer configuration I'm using:
> {noformat}
> <#indexLucene> a text:TextIndexLucene ;
>     ##text:directory <file:/tmp/lucene> ;
>     text:directory "mem" ;
>     text:entityMap <#entMap> ;
>     text:storeValues true ;
>     .
> <#entMap> a text:EntityMap ;
>     text:entityField      "uri" ;
>     text:graphField       "graph" ; ## enable graph-specific indexing
>     text:defaultField     "pref" ; ## Must be defined in the text:map
>     text:uidField         "uid" ;
>     text:langField        "lang" ;
>     text:map (
>          # skos:prefLabel
>          [ text:field "pref" ;
>            text:predicate skos:prefLabel ;
>            text:analyzer [ a text:LowerCaseKeywordAnalyzer ] ]
>          # skos:altLabel
>          [ text:field "alt" ;
>            text:predicate skos:altLabel ;
>            text:analyzer [ a text:LowerCaseKeywordAnalyzer ] ]
>          # skos:hiddenLabel
>          [ text:field "hidden" ;
>            text:predicate skos:hiddenLabel ;
>            text:analyzer [ a text:LowerCaseKeywordAnalyzer ] ]
>          ) .
> {noformat}
> Here is a minimal test file that I load into the default graph:
> {noformat}
> <http://example.org/guppy> <http://www.w3.org/2004/02/skos/core#prefLabel> "Guppy"@en-gb .
> {noformat}
> This is the query I'm using:
> {noformat}
> PREFIX text: <http://jena.apache.org/text#>
> SELECT * {
>   ?s text:query 'G*' .
> }
> {noformat}
> It returns one row (?s=<http://example.org/guppy>) on Fuseki 3.9.0 but nothing with today's 3.10 snapshot.
> If I change the 'G*' to lowercase 'g*' then I get the expected match also with the 3.10 snapshot. So the analyzer (which should lowercase everything and thus the case of the query string should be irrelevant) seems not to be applied for the query string.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)