Honor� David wrote:

> It would be more efficient to store lucene index into slide BUT
> implement org.apache.lucene.store.Directory via slide API (on the server
> side) directly and not over and HTTP connection.
> (Anyway the indexer is server side.)

Yes, implementing it directly onto the Slide API is more efficient, but
not efficient enough alone I�m afraid. I got a Lucene index for a customer
that is over 2GB of size and even if this is split into several Lucene
files, many of these files are just to large to handle without some paging
I think. Further more a clean WebDAV implementation would be much more
flexible and reusable for the Lucene project to, not only Slide.

> note: Lucene juste need (Create => Write_only => Close) and (Open =>
> Random_Read_only => Close). and delete Lucene never "modify" the "file",
> but you need not to just implement org.apache.lucene.store.Directory
> Interface but also a org.apache.lucene.store.InputStream,
> org.apache.lucene.store.OutputStream, ... etc
>
> I can find some time to work on this if someone think this is something
> usefull.

Lucenes InputStream and OutputStream requires seek() even if the actual
files doesn�t. I have abused the RAMDirectory for a while now and I think
that taking the code in that implementation would be a nice start for
implementing a DAVDirectory. It already does paging into arrays, so the
work would be to save and load these arrays into slide and adjust the
maximum allowed size of each �file�. Lucene sees collections of these
paged arrays as files, so we would need a filename translation that says
that file �lucene.segment� actually is chunked out into a set of small
WebDAV resources.

> But all of this doesn't help for re-index... : (

Eh... true, but wouldnt it be cool? :-)

Mvh Karl �ie

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to