Lucene itself doesn't provide this, but Solr does. And with Solr you
don't need your own custom Hits cache - it already provides caches in
several areas.
Erik
On May 13, 2006, at 9:12 AM, karl wettin wrote:
This might exist?
How about a list of index listeners that is called
On May 13, 2006, at 12:10 PM, Malcolm Clark wrote:
Hi everyone,
I am about to index the INEX collection (22 files with 3 files
in each-ish) using Java Lucene. I am undecided with the approach to
indexing and have left my LIA book at uni :-/
Would you recommend:
1.. indexing all
On Sun, 2006-05-14 at 07:05 -0400, Erik Hatcher wrote:
Lucene itself doesn't provide this, but Solr does.
I'm not using Solr :) But what you suggest is that I should try to fit
in in my own layer around Lucene rather than in Lucene?
May I ask why?
Not sure if I would do it that way. To me it
On 5/14/06, karl wettin [EMAIL PROTECTED] wrote:
To me it feels as the index is the
only thing that knows for sure if it has been updated.
I guess that would be whenever an IndexReader that had new deletions
is closed, or an IndexWriter that changed the segments file?
An app using Lucene
I have an out-of-memroy error when returning many hits.
I am still on Lucene 1.4.3
I have a simple term query. It returned 899810 documents.
I try to retrieve the name of each document and nothing else
and I ran out of memory.
Instead of getting the names all at once, I tried to query again
Could you share at least some pseudo-code of what you're doing in the
loop of retrieving the name of each document? Are you storing all
of those names as you iterate?
Have you profiled your application to see exactly where the memory is
going? It is surely being eaten by your own code
I'm planning to write a listener which will recieve messages from our indexing
application via SOAP - the message will indicate which indexes have been
updated as we will be using quite a few separate lucene indexes. No point
dropping all the searchers (and any caches, filters etc) if only one
Thank you for all your thoughts on this - I think we will go the route of
having two copies for each index, a working copy and the one the searchers use.
If we start off indexing like this it shouldn't be too hard to keep them in
sync.
Thanks again,
Marc
- Original Message
From:
On Sun, 2006-05-14 at 10:50 -0400, Yonik Seeley wrote:
On 5/14/06, karl wettin [EMAIL PROTECTED] wrote:
To me it feels as the index is the
only thing that knows for sure if it has been updated.
I guess that would be whenever an IndexReader that had new deletions
is closed, or an
On Sun, 2006-05-14 at 22:32 -0400, Erik Hatcher wrote:
On May 14, 2006, at 1:48 PM, karl wettin wrote:
An app using Lucene still needs to coordinate all the activity
surrounding IndexReaders and IndexWriters, including explicit
closure,
so the app will know anyway when the index has
On Sun, 2006-05-14 at 22:27 -0400, Erik Hatcher wrote:
What are the boundaries of what you call an index? Is it the
current Lucene API, or could it be a service-like layer such as Solr
on top of it?
The persistence mechanism.
So that would be a part of the the current Lucene API.
On 5/14/06, karl wettin [EMAIL PROTECTED] wrote:
On Sun, 2006-05-14 at 22:27 -0400, Erik Hatcher wrote:
What are the boundaries of what you call an index? Is it the
current Lucene API, or could it be a service-like layer such as Solr
on top of it?
The persistence mechanism.
So that would be
: As deep as possible.
:
: All the user would need to know is Directory.getListerners()
I don't really get what the point of such a low level callback mechanism
would be ... do you have uses cases where you're really going to want to
know every time someone uses a Directory object to do
On Sun, 2006-05-14 at 21:09 -0700, Chris Hostetter wrote:
Personally: I think the best way to be notified that the index is
changed at a really fine level of granularity is to just poll on
IndexReader.getCurrentVersion (and compare with
IndexReader.getVersion) ... that way you don't have to
14 matches
Mail list logo