Re: [Libreoffice] clucene port ...

On Fri, 2010-12-03 at 16:00 +0100, Jan Holesovsky wrote:
> Hi Michael,
>
> On 2010-12-03 at 13:57 +0000, Michael Meeks wrote:
>
> > > Start with stuff that /any/ user would see.
> > > My list would start with:
> >
> >     This is a great list :-) we should get some of it into the easy hacks
> > page I suspect (under the not so easy hacks).
> >
> > > Fulltext search in Help (i.e. <F1>) - that uses lucene and requires java
> >
> >     Right; but there is no real need for us to use Java lucene, when we can
> > use cluncene:
> >
> >     http://clucene.sourceforge.net/
>
> Even though Christian outlined some potential problems with the native
> help(s), I'd still prefer to give it a bit of research before agreeing
> that we really need to preserve Lucene / the LO internal help engine in
> general ;-)
>
> So, Jonathan/others, please, no clucene porting yet, only if it proves
> inevitable; I'll keep you informed.

BTW, I did a bit of research back in 2007 when I was looking for a good
text indexing engine (at my previous job).  I didn't find any
replacement for Lucene even though we were a C++ shop and Lucene was
written in Java.  CLucene was there back then, but the general consensus
was that CLucene was not keeping up with the development of Lucene.

I ended up building Lucene using gcj into a shared library
(liblucene.so), which, with only a bit of change in the Java code I was
able to pull off nicely.  The Lucene developers apparently make their
code as gcj friendly as they can, and some websites that use Lucene do
in fact build it with gcj into native code for better performance.

Of course the situation may have changed in the past few years, but I
thought I'd mention what I found back in 2007.

Kohei

-- 
Kohei Yoshida, LibreOffice hacker, Calc
<kyoshida@novell.com>