Mailing-List: contact java-dev-help@lucene.apache.org; run by ezmlm
Precedence: bulk
Reply-To: java-dev@lucene.apache.org
Received-SPF: pass (hermes.apache.org: local policy)
Message-ID: <426DB487.1080605@allthingslocal.com>
Date: Mon, 25 Apr 2005 20:24:55 -0700
From: Chuck Williams <chuck@allthingslocal.com>
Organization: All Things Local
User-Agent: Debian Thunderbird 1.0 (X11/20050116)
MIME-Version: 1.0
To: java-dev@lucene.apache.org
Subject: Re: SortTest failing
References: <20050426025549.47791.qmail@web31101.mail.mud.yahoo.com>
In-Reply-To: <20050426025549.47791.qmail@web31101.mail.mud.yahoo.com>
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

Otis Gospodnetic wrote:

>Hm, Erik is not alone with unit tests failing.  My HighlighterTest
>passes (I didn't do svn update today yet), but I see SortTest failing:
>
>    [junit] Testcase:
>testNormalizedScores(org.apache.lucene.search.TestSort):  FAILED
>    [junit] expected:<0.375> but was:<0.392445>
>    [junit] junit.framework.AssertionFailedError: expected:<0.375> but
>was:<0.392445>
>    [junit]     at
>org.apache.lucene.search.TestSort.assertSameValues(TestSort.java:560)
>    [junit]     at
>org.apache.lucene.search.TestSort.testNormalizedScores(TestSort.java:365)
>
>
>Line 365 is this:
>
>assertSameValues (scoresA, getScores(multi.search(queryA,sort)));
>
>
>And this makes me think that this broke during my last commit of Wolf's
>patch for MultiSearcher and docFreq stuff.  However I did run 'ant
>test' before commit and did see BUILD SUCCESSFUL, so I'm not 100% sure.
> 
>
>Anyone else seeing this error?
>  
>
Otis, I think that is a bug in the test case that Wolf's patch has 
exposed. It creates an index consisting of two copies of "full". This 
should be equivalent to an index where every document is duplicated 
(occurs twice), not to an index where each document occurs only once. 
With the new correct idf normalizaiton, this increases all the docfreq's 
for the terms, which changes the score. Scores on the duplicated index 
should not be the same as scores on the unduplicated index.

Chuck


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