lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marvin Humphrey <mar...@rectangular.com>
Subject Re: Ferret's changes
Date Wed, 11 Oct 2006 03:52:34 GMT
On Oct 10, 2006, at 8:03 AM, Yonik Seeley wrote:
> Looking forward to progress on Lucy.  What is done there could
> potentially be the a future Lucene index format.

There has actually been a good bit of progress on Lucy, just nobody  
can see it.  :\  My first priority right now is finishing KinoSearch  
0.20, and I basically can't do anything else until that's done.   
Fortunately, a lot of the work going into that effort is likely to  
help out Lucy.

KS 0.20 will break backwards compatibility with earlier versions, and  
to minimize the disruption, I'm trying to cram in all the planned API  
and file-format changes into one release.  Those changes will include  
a version of the flexible indexing format we've been discussing, and  
a boolean scorer which takes position into account.  It already uses  
the KinoSearch merge model, of course, but I'm also going to  
incorporate Dave's fixed field numbers.  It may take a completely  
different approach to sorting and range queries, though I haven't  
worked out all the details on that yet.

While I've been preparing this version, I've been studying Ferret and  
making KS more Ferret-like wherever possible.  The biggest push has  
been to purge all the C functions taken from Perl's C API; KS used to  
have them everywhere, but they're almost gone now.   Object  
destruction has been abstracted out in a way that hopefully can be  
made to work with both Perl's reference-counting scheme and Ruby's  
mark-and-sweep garbage collector; callbacks from C to [your-language- 
here] have been similarly generalized.  A slew of classes have been  
refactored -- e.g. KS's PriorityQueue, which used to use Perl scalars  
as its elements, but now looks a lot more like Ferret's (and Lucene's).

Dave and I talked about collaborating by spelunking each other's  
libraries, editing, and writing tests.  That process is underway,  
just the code isn't filtering back to the Lucy repository yet.  We'll  
see how soon it does, and eventually, how much code in Lucy  
negotiates its way into Lucene.

Marvin Humphrey
Rectangular Research
http://www.rectangular.com/



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


Mime
View raw message