Isn't Lucene's ParallelReader meant to address such use cases?  Don't ask me 
for details, the actual use of PR always seemed a bit fuzzy to me because of 
its requirement to keep docIDs in sync.

 Otis
--
Sematext -- http://sematext.com/ -- Lucene - Solr - Nutch




________________________________
From: Noble Paul നോബിള്‍  नोब्ळ् <[email protected]>
To: [email protected]
Sent: Thursday, February 19, 2009 8:12:28 PM
Subject: An ExternalIndexField implementation with multicore

hi,

Just the way we have an ExternalFileField is it possible to refer to a
field (ExternalIndexField) in another index ( which lives in another
core)?

I would not want to search on that field but I may wish to use it to
filter or sort or as a ValueSource in a Function

The usecase is as follows.
--------------------------------------
I have a large index with huge docs which changes less frequently
(think of a mailbox). The user may arbitrarily apply/remove tags on
that. but I may not wish to reindex the mails where the tags are
applied. I want to just add a small doc mail-unique_id and the tag
into another index in another core. When I query, I wish to  apply a
filter of the label or when i retrieve the mail details I want to get
the tags (stored field) applied to that.

Another one.
I have a huge index of products which the users can vote up or down
(say popularity). I may want to add the add the popularity of the item
into another index and when I query I wish to sort by the popularity.

the commits on the other external index will be more frequent than the
main index.

What are the challenges in implementing something like this? I wish to
raise a Jira issue if it looks feasible


-- 
--Noble Paul

Reply via email to