lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Becker <pbec...@dstc.edu.au>
Subject Re: Luke v 0.2 - Lucene Index Browser
Date Mon, 11 Aug 2003 20:29:40 GMT
Andrzej Bialecki wrote:

[...Luke feature requests...]

>>
>>> open the original Dokuments with the platform dependant mimetype viewer
>>
>
> Someone else already explained the problems with this... What is a 
> document in Lucene? It's a set of fields and their String values (or 
> their terms), so it's not possible to open the original document from 
> which the values were extracted, because there is no original...

We store the paths to the documents in the index, partly for exactly 
this reason, partly to get an identifier for the documents. The thing I 
could see Luke doing would be guessing that a field contains a path or 
URL and then offering to give it to BrowserLauncher, which you use anyway.

Personally I'd like to see this feature, too. It would be interesting to 
analyse an index by looking at unusual terms and then being able to see 
the full document somewhere. But I think it is hard to generalize the 
approach and it might actually break for our indexes sooner or later 
since we are thinking about storing the location relative to the indexed 
directories -- which allows us to use one index with different file 
system mappings, e.g. to index a directory locally on a Linux box which 
is then used on a Windows client via SMB. If we go down this road the 
index wouldn't contain the complete information for opening files anymore.

Just my 2c,
   Peter


Mime
View raw message